uses

science fora changing world

Proceedings
UJNR WORKSHOP ON
SOIL-STRUCTURE INTERACTION
Menlo Park, California
September 22-23, 1998

JI/^AVvAA^

Editors & Conveners:

Mehmet Celebi1
and
Izuru Okawa2
Open-File Report 99-142
1999
This report is preliminary and has not been reviewed for conformity with U.S. Geological Survey
editorial standards (or with the North American Stratigraphic Code). Any use of trade, product or firm
names is for descriptive purposes only and does not imply endorsement by the U. S. Government.

U.S. Department of the Interior

ILS. Geological Survey

1 U. S. Geological Survey (MS977), 345 Middlefield Rd, Menlo Park, Ca. 94025
2 IISEE, Building Research Institute, Tsukuba, Japan
 ACKNOWLEDGEMENTS

The workshop was sponsored by both United States (US) and Japan. On the US side, the
National Science Foundation (CMS-980762) and the United States Geological Survey (USGS)
were the main sponsors. On the Japan side, several institutions including the Building Research
Institute (BRI) and the Public Works Research Institute (PWRI) of the Ministry of Construction
of Japan provided funds in support of this workshop. The conveners gratefully acknowledge the
financial support that enabled successful planning and execution of the workshop

Mehmet elebi (USGS, Menlo Park, Ca.)

Izuru Okawa (BRI, Japan)

DISCLAIMER

The technical papers presented in this document are the responsibility of the authors. The United
States Geological Survey does bear any responsibility for the accuracy of the calculations,
figures and presentations contained in individual papers.
 TABLE OF CONTENTS

Acknowledgements .........................................................................................2
Disclaimer ...................................................................................................2
Table of Contents .............................................................................................3
Forward .......................................................................................................6
Resolution of the Workshop .................................................................................7
Summary of Discussions and Recommendations ........................................................8
Workshop Program .......................................................................................... 14
Workshop Participants List ................................................................................ 17
Workshop Photo .............................................................................................22

TECHNICAL PAPERS OF THE WORKSHOP

Author and Title Page Numbers

Iguchi, Michio and Yasui, Yuzuru 1-1 to 1-17

State of the Art on Soil - Structure Interaction Researches Relating to Recent
Strong Earthquakes in Japan

Ostadan, Farhang and White, William H. 2-1 to 2-36

Lateral Seismic Soil Pressure: An Updated Approach

Karkee, Madan B., Sugimura, Yoshihiro, and Fujiwara, Kaoru 3-1 to 3-13
Design of Piles Considering the Deformation Response Under the Action
of Earthquake Shaking

Veletsos, A.S. and Younan, A.H. 4-1 to 4-10

Response to Horizontal Group Shaking of Cantilever Retaining Walls

Mizuno, Hatsukazu, Sugimoto, Michio, liba, Masanori, Mori,

Toshihiro, and Hirade, Tsutomu 5-1 to 5-9
Dynamic Behavior of Pile Foundations in Liquefaction Process - Shaking
Table and Oscillator Tests Utilizing Big Shear Box

Grouse, C.B. 6-1 to 6-17

Energy Dissiptation in Soil-Structure Interacture: A Consultant's
Perspective

Jonathan P. Stewart 7-1 to 7-21

Empirical Assessment of Soil-Structure Interaction Effects from Strong
Motion Recordings

Tamori, Shin'ichiro, liba, Masanori, and Kitagawa, Yoshikazu 8-1 to 8-8

Dynamic Response of Soil-Pile-Building Interaction System in Large
Strain Levels of Soils
Borja, Ronaldo, Chao, Heng-Yih, and Lin, Chao-Hua 9-1 to 9-12
Nonlinear SSI Analysis

Bazan-Zurita, E., Bazan-Arias, N.C., and Bielak, J. 10-1 to 10-17

Three Dimensional Seismic Response of Building-Foundation Systems

Ghanem, Roger 11-1 to 11-10

A Statistical Measure of Fit for Analytical Models Conditioned on
Experimental Results

Ghiocel, D.M., Ph.D. 12-1 to 12-23

Uncertainties of Soil-Structure Interaction Analysis: Significance
Modeling and Examples

Tseng, Wen S., and Penzien, Joseph 13-1 to 13-11

Seismic Soil-Foundation-Structure Interaction in Bridges

Fukuwa, Nobuo, Ghannad, M.. Ali, Tobita, Jun, and Nishizaka 14-1 to 14-15
Analytical and Experimental Studies on the Effect of Soil Structure Interaction on
Damping, Natural Frequency and Effective Input of Buildings

Aydmoglu, M. Nuray, Polat, S. Seref, and Beyen, Kemal 15-1 to 15-15

Soil-Structure Interaction Analysis via Fixed-Base System Subjected to a Modified
Ground Motion

Tamura, Keiichi, Nakao, Yoshiro, and Honda, Riki 16-1 to 16-12

Sudy on Strong Ground Motions for the Application to Seismic Design of Structures -
Semi-empirical Method for Ground Motion Estimation and Non-linear Response
Spectra

Trifunac, M.D., and Todorovska, M.L, 17-1 to 17-20

Relative Flexibility of a Building Foundation

Kitada, Yoshio, Kinoshita, Masamitsu, Iguchi, Michio, and

Fukuwa, Nobuo 18-1 to 18-14
Soil-Structure Interaction Effect on an NPP Reactor Building - Activities ofNUPEC;
Achievements and the Current Status

Man Mok, Chin, Chang, C.-Y., Settgast, Randolph, Wang, Z.-L.,

Gonnermann, Helge M., and Chin, C.-C. 19-1 to 19-30
Dynamic Soil-Foundation Structure Interaction Analyses of Large Caissons
Kitagawa, Yoshikazu, Matushima, Yutaka, Yamazaki, Yutaka,
Kawamura, Soichi, and Inoue, Yoshio 20-1 to 20-7
Seismic Design Procedure of Building Structures Including Soil-Structure Interaction
Effect

Ashford, Scott A. 21 -1 to 21 -5
Experimental Facilities in the United States for Soil-Structure Interaction Research

Okawa, Izuru, Kashima, Toshihide, and Koyama, Shin 22-1 to 22-9

Dense Instrumentation in BRI Building and Surrounding Ground

gelebi, Mehmet 23-1 to 23-12

Planning and Implementation of a Soil Structure Interaction Experiment
Luco, J. Enrique 24-1 to 24-32
Equivalent 1-DOF Model for Structural Control Including Soil-Structure Interaction
Effects

Luco, J. Enrique and de Barros, Francisco C.P. 25-1 to 25- 51

Seismic Response of a Pipeline Embedded in a Layered Medium
 FORWARD

During the May 1997 meeting of the Panel on Wind and Seismic Effects of the United States -
Japan Natural Resources Development Program (UJNR) held in Tsukuba, Japan, it was decided
to hold a workshop on "Soil-Structure Interaction " (SSI) in 1998. The panel assigned Dr/ M.
Celebi of USGS, Menlo Park, Ca., and Dr. I. Okawa from BRI, Tsukuba, Japan to serve as
organizers of the workshop. Subsequently, the organizers decided to hold this workshop in
Menlo Park, California on
September 22-23, 1998.

The workshop participants who are experts in the field from Japan and the United States
met in Menlo Park, California, presented technical papers and discussed topics including but not
limited to (a) current methods of SSI used in design/analyses processes in both Japan and the
United States, (b) recent research that is being carried out, (c) experimental SSI research
arrays and/or facilities developed and that are in the process of being developed and (d)
searching ways to cooperate on future SSI research. The aim of the workshop was to cover
the following topics:

1. Current Methods of Practice of SSI in the US and Japan

a. Geotechnical Point of View
b. Structural Point of View
2. Code Provisions and Limitations.
3. Observed Data.
4. Observational Arrays and Testing Facilities - Current Status and Future Needs.
5. Recent Research Results and How To Implement Them Into Practice.
6. Additional Research Needed.
7. Additional Observational Arrays and Testing Facilities Needed.

While it was impossible to cover all of these subjects, the workshop provided a venue to discuss
a variety of issues related to soil-structure interaction. One of the important issues revealed
during these discussions, summarized in Section XX (?) of the proceedings, is that funding for
SSI research has not increased in either the US or Japan. Consequently, the number of published
papers related to SSI research has been steady and has not increased during the last few years. A
recommendation made by the participants to organize an International Association for Soil-
Structure Interaction (LASSI) is aimed to improve the funding for SSI research and improve the
communication between the researchers and the practicing engineers. Furthermore, it was
pointed out that should such an organization be formalized, it would be a means to organize
special purpose conferences and workshops on the specific subject.

As the host convener of the workshop, I thank all of the participants for their attendance and
enthusiastic presentations and discussions. I look forward to future UJNR-SSI Workshops.

Mehmet felebi
Menlo Park, Ca.
September 1998
 RESOLUTION OF PARTICIPANTS

U.S. - Japan Workshop on Soil-Structure Interaction

Menlo Park, California, September 22-23, 1998

WHEREAS, soil-structure interaction can have major influence on the seismic

performance of important structures, such as buildings, dams, bridges, and nuclear
power plants, and thus affect public safety; and

WHEREAS, the methodologies for including soil-structure interaction effects in

assessing seismic performance of such structures are inadequate; and

WHEREAS, present-day design codes provide little guidance for treating soil-
structure interaction effects;

NOW THEREFORE BE IT RESOLVED THAT

(1) research to advance soil-structure interaction methodologies be given high priority and
that design provisions related thereto be introduced into the codes, thus enhancing the
seismic safety of structures designed accordingly, and

(2) cooperation between the U.S. and Japan, with focus on advancing both state-of-the-art
and state-of-the-practice of treating soil-structure interaction be strengthened, and

(3) future additional UJNR-SSI meetings be organized, and

(4) an international organization be established to promote research, education, and improved

design practice as related to soil-structure interaction.

Participants
US-Japan Workshop on SSI,
Menlo Park, Ca.
September 22-23, 1998
 SUMMARY OF DISCUSSIONS AND RECOMMENDATIONS

Soil-structure-interaction (SSI) effects may be either beneficial or detrimental to the performance

of structures. When beneficial, by incorporating SSI effects in the seismic code calculations,
more cost-effective designs are possible. For some situations, such as the design or retrofitting of
bridges, dams or buried structures, etc., an appropriate inclusion of SSI effects in seismic
calculations may bring large design cost savings to our society. There is an urgent need for
performing comparative cost-benefit reviews with and without considering rigorously the SSI
effects for different types of constructions. On the other hand, when it is determined by
calculations that SSI effects can be detrimental to the performance of structures, by mere
recognition and taking effective measures, safety and better performance can be achieved.

1. Present Status of SSI:

To promote practical application of SSI evaluation procedures, practicing engineers must

first be convinced of the need for SSI evaluations. To render such evaluations a
necessity, SSI evaluation procedures must become an integral part of the total seismic
analysis and design process. Current building codes, which are based on SSI response
behavior of a single-degree-of-freedom (SDOF) SSI system and have incorporated only
the SSI effects of period elongation and damping increase for the fundamental mode of a
structure system, do not address the total effect of SSI (such as the additional effects of
"scattered" seismic input motions, and global as well as local soil non-linearity); as such,
they do not promote the use of proper SSI evaluation procedures in the design process.
In the past, SSI research has concentrated on solutions for gross dynamic response
behaviors of simple linear SSI systems. Recent research also tends to focus on studying
the SSI problems that can be solved with simple linear theories. To further the SSI
research, it is time that the research be advanced beyond the studies of simple linear SSI
systems and should start to develop realistic SSI evaluation procedures needed for
practical design purposes, e.g. evaluations of nonlinear soil-resistance behaviors and soil-
foundation interface pressures.

To date, evaluation of seismic SSI effects has placed emphasis on seismic system
demand, i.e., seismic SSI system response behavior. It is time to extend evaluation to SSI
system performance, which requires the evaluation of not only the system demands but
also the corresponding (strength and ductility) capacities. In the context of SSI, the
system capacities of interest are the capacities of the soil-foundation interaction system.
In fact, any realistic evaluation of the SSI system demand must incorporate realistic
constitutive behavior of the soil-foundation system up to its allowable capacity limit.

Experimental research should not be limited to the confirmation of SSI system response
behavior. It must be designed and conducted in a manner in order to improve the SSI
system modeling and to facilitate assessment of the SSI system performance up to its
performance limit.
To facilitate practical applications, SSI researchers must also develop and make available
to practicing engineers a set of reliable and easy-to-use computer software for them to
conduct realistic SSI evaluations.
The number of papers on SSI both in Japan and US has been steady during the last few
years. This implies that support for SSI research has not increased in recent years.

SSI is interdisciplinary (geotechnical and structural) and hence tends to be poorly

understood by both sides. There is a big gap between SOA (state-of-the-art) and the
knowledge of practicing engineers.

SSI is too complex to define exactly. We can define conditions where SSI is not
important, however. Let's define what we know, where contributions can be made, and
improve our knowledge transfer. We should not emphasize code-oriented research too
much, we need to communicate to practitioners the essential aspects of the problem.

Need to distinguish between heavy nuclear power plants and ordinary buildings. Nuclear
plants are already being designed with consideration of SSI. For buildings, there are
cases where SSI is not important. These cases should be identified.

Our knowledge of ordinary building structures is limited, so there is a need to emphasize

SSI research for ordinary types of buildings.

There are virtually no full-scale experiments on buildings.

Single-degree-of-freedom (SDOF) structure with rigid foundation is the most common

type of research topic. Much research has been done on this subject.

Flexible foundations with multi-degree-of-freedom (MDOF) structures are difficult to

analyze and there is very little research done on this topic.

Individual footings beneath each column are also difficult to analyze, and there is very
little work done on this subject.

There is virtually no field performance data on SSI. Existing data is inadequate.

Interpretation of field data from earthquakes is important to verify methodologies.

There is a lot of research on pile foundations. However, there is great need to translate
that accumulated knowledge into practice.

We need more detailed experiments.

In general, Linear Elastic Analysis is good for:

i. buildings on surface foundations
ii. building-soil-building interaction
iii. single building with embedded foundation
 Linear analyses is not so bad. Past experience shows linear models are here to stay.
They've been around for a long time, despite some nonlinear alternatives, and they will
remain. However, linear elastic analysis has shortcomings for building-soil-building
interaction when the foundations are embedded.

The standpoint of practitioners: Is SSI a necessary aspect of the design process? We

think it is, but how do we demonstrate that? Need more than period lengthening and
foundation damping; these are not useful to practicing engineers. We need to translate
our research results into better demand predictions for structures. SSI enters the design
process through pressure on foundations. Need further research on this.

SSI is significant in the context of performance-based design.

Community studying SSI is shrinking due to limited funding priority place on SSI by
NSF. If we speak as one voice, we can make an impact on the NSF (like the structural
control and tsunami people). Let's create a web site to advance the issue (post research
findings, etc.).

Design of Nuclear Power Plants was a major stimulus to SSI. Since practically no new
nuclear plants are being design, such stimulus has vanished.

Recent earthquakes show that there is a high level of nonlinearity in soil over broad area.
This nonlinearity may have lead to SSI effects which saved these buildings. We need to
investigate this.

Need dialogue between experimentalists and analysts.

2. Additional SSI Research:

Seismic earth pressures against retaining walls, considering non-linear aspects such as
gapping.
Comparative studies of non-linear vs. linear SSI to evaluate where non-linear analysis is
important.
More work is needed on pile foundations. For example, observed damage of piles due to soil
displacement suggests that we need to consider soil displacement, not just structural inertia,
when designing piles. How the two actions can be superimposed is of vital importance.
More work is needed for underground structures such as tunnels and pipelines.
How do we estimate the degree of nonlinearity in soil and its effects.
Need more work on flexible foundations.
More work needs to be done on seismic soil pressures against walls
SSI is more involved than just the first mode period lengthening ratio (T / T) due to
interaction and ground motion variability. The load paths for inertial load, especially near
the ground line need to be considered.

10
 Nonlinear SSI may be very important for severe earthquakes. Need simple models for
nonlinear SSI
We are only recently getting accustomed to > Ig ground motions. Pushover is becoming
more common, need to properly account for SSI in such analyses, especially near the failure
state.
If the movement is towards Performance Based Design, then we need to understand the
uncertainties.
There is a need for SSI research for near field ground motions. Such effects may be very
localized.
From the geotechnical point of view, an important issue is the damage in piles at soil
boundaries significantly below the ground surface due to lateral flow. Before liquefaction,
soil is fairly elastic and strains are important in determining soil properties. After
liquefaction, soil behaves as liquid. Need to distinguish between liquefied and non-liquefied
soil in our SSI formulations.
Energy absorption by liquefied soil is significant, adds extra damping.
Level of energy dissipation depends on when liquefaction occurs in the time history.
Liquefaction doesn't help much if it occurs late. There is evidence of this from Wildlife
Liquefaction Site (see Holzer, T. L., Youd, T. L., and Hanks, T. C., 1989, Dynamics of
Liquefaction during the 1987 Superstition hills, California, earthquake, Science, v. 244, pp.
56-69).
We need to be concerned about 5-to-10 story buildings subjected to near-field pulses. The
long period energy content of these motions means that period lengthening would increase
the base shear.
There is a need to address in future engineering activities the large uncertainty associated
with SSL We know that the earthquake motions are random, the soil properties are random,
local motion spatial variation is random, etc. So, there is an objective need in the future to
approach these aspects more consistently using probabilistic models. In addition, for
improving a seismic design or for a costly retrofit of a highway concrete bridge, it is
essential to do some probabilistic SSI analyses, and try to calibrate the deterministic design
based on risk assessment comparisons. Therefore, it is important that NSF envisions this
need for future.

3. Better Field Observations:

In general, there have been some successes in experimental work and use of observed
data. These can be summarized in three ways: (a) Lotung-type of experiment with
very good instrumentation for a specific type of structure, lots of comparisons
between theory and experiment, (b) in-depth studies of typical building structures and
(c) studies of many buildings, look at trends that can be easily understood by many
engineers (e.g. Stewart, Ph.D. thesis - Stewart, J.P., and Stewart, A. F., 1997,
Analysis of soil-structure interaction effects on building response from earthquake
strong motion recordings at 58 sites, Report No: UCB/EERC 97/01, Richmond, Ca.).
However, there is still great need for developing and/or improving the current field
observation systems such that these systems will better enable
experimental verification of analytical procedures (e.g., in Europe, the
research is on verifying SSI provisions in Eurocode.).

11
 interpretation of available field data
additional instrumentation to obtain proper SSI response data (e.g. most
instrumented buildings have inadequate vertical sensors to calculate rocking
effects, and in some cases, if physically possible, additional free-field
instruments and downhole accelerographs should be deployed).
evaluation of the influence of free-field displacements on piles.
understanding the soil pressures against foundation elements such as
basement walls. More work needs to be done on seismic soil pressures against
walls.

We need reliable experimental data for verification of simple analysis schemes

4. Transfer of Knowledge:

There is a big gap between state-of-the-art and the knowledge of practicing engineers.
Therefore,
it is necessary to simply be able to demonstrate to the practicing engineers when and
if SSI is important.
simple and practical tools and procedures are needed for transfer of knowledge to
practicing engineers.
efforts should be made to include SSI in building codes.
efforts should be made to incorporate SSI methodologies in favorite computer
software such as SAP.
efforts should be made to demonstrate to the profession the usefulness in
incorporating SSI in their designs. The design engineers should be appraised of the
fact that incorporation of SSI procedures can be, in some cases, financially beneficial.

There is considerable research on pile foundations. However, there is great need to

translate that accumulated knowledge into practice.

There is a significant need for knowledge transfer on the issue of damping.

Graduate students need to be taught SSI - this will help bridge that gap between
researchers and practitioners. All grad student qualifying exams should have SSI
questions.

Need practical tools and agreed-upon computer codes.

The data should not be used to calibrate a design code. Rather, we need to understand
simple problems well, then develop good code formulations for design based on the
insight gained from these simple models. Instrumentation needs to be detailed enough to
guide us through the process.

12
5. Data Exchange Between US and Japan:

This is of vital importance for researchers on both sides. As an example, Professor Iguchi
wrote a book (along with 21 authors) on SSL Two-thousand (2000) copies of this book
were distributed in Japan through lectures to engineers. The US side may desire to have
the book translated.
There are impressive experiments in Japan for SSI, we need to become more familiar
with them.
We should recommend that there be better information exchange between US and Japan
Japanese experimental data is extremely valuable - must relate it to available theoretical
models. Perhaps we in the US could contribute our expertise to such an effort.

6. Other Issues:

SSI practices should be pushed into the codes. In that case, the industry will use it. Thus,
it will be necessary to teach it. Under these conditions, funding agencies will have to fund
such activities. The code committees are receptive now to SSI. Therefore, this should be
followed to fruition.
We need an inside advocate in the NSF and USGS to get SSI funding. Let's speak as one
voice to develop this inside advocacy. We should also look into new funding sources
such as the insurance industry, and gas and oil companies.
Funding for SSI research will increase when the industry has a demand for SSI, e.g.
nuclear industry in 1970-1980s. Our models should not just be elastic, but should
consider soil strength.
There is an acute need for an international organization to promote SSI. The workshop
participants agreed that it is important to promote research, education and design
applications of soil-structure interaction, and to initiate specific steps towards its
mandatory implementation in the design codes. Maria Todorovska proposed that an
international association be set up to serve as an organizational framework to carry out
this important task. The workshop participants endorsed this idea and agreed to promote
the establishment of such an organization. It was further agreed that Maria Todorovska
will contact the leading researchers in the US and abroad to discuss this idea further and
to take specific steps towards its completion. The Japanese workshop participants agreed
to contact other researchers in Japan.
It was agreed that the 12WCEE (to be held in Auckland, New Zeland, in year 2000) is a
convenient place for a meeting which can finalize the creation of an International
Association. A special session on soil-structure interaction at this conference would be
appropriate. Maria Todorovska proposed to organize such a session.

13
 PROGRAM
U.TNR-SSI WORKSHOP
Vallombrosa Center, 250 Oak Grove Avenue, Menlo Park, Ca

MONDAY - SEPTEMBER 21,1998

Japanese participants arrive in the morning. US participants arrive in the afternoon/evening and
the following morning.

TUESDAY - SEPTEMBER 22, 1998

8:00 - 9:00 AM Breakfast & Registration

[Session Chairs: Okawa, Celebi]
9:00 - 9:20 AM Welcome Comments by Mehmet Celebi & Izuru Okawa
9:20 - 10:00 AM (1) "States of the Arts on the Research on SSI in Japan" (Iguchi)
10:00 -10:20 AM Coffee Break
[Session Chairs: liba, Veletsos]
10:20-11:00 AM (2" Lateral seismic soil pressure" (Ostadan )
11:00- 11:30 AM (3) " On the design of pile foundation using response deformation
method ( Sugimuro & Karkee)
11:30- 12:00 AM (4) "Dynamic Soil-Foundation Structure Interaction Analyses of Large
Caissons" (Mok)

12:00- 13:00 Lunch at Vallombrosa Center

[Session Chairs: Kitagawa, Ostadan]
13:00-13:30 PM (5) "Dynamic Response of Cantilever Retaining Walls" (Veletsos)
13:30-14:00 PM (6) "Dynamic behavior of pile foundation in liquefied Process-Shaking
Table Tests Utilized Big Shear Box" (liba)
14:00-14:30 PM (7) "Energy Dissipation in Soil-Structure Interaction" (Crouse)
14:30-15:00 PM (8) "Empirical Assessment of SSI Effects from Strong Motion
Recordings" (Stewart)
15:00 -15:30 PM Coffee Break
[Session Chairs: Iguchi, Crouse]
15:30-16:00 PM (9) "Dynamic Response of Soil-pile-building Interaction System in
Large Strain Levels" (Tamori)
16:00-16:30 PM (10) "Non-linear SSI Analyses" (Borja)
16:30-17:00 PM (11) "SSI Effects on damping and natural frequency and effective input
motion" (Fukuwa)
17:00- 17:30 PM (12) "The use of nolinear response spectra to seismic design" (Nakao)
17:30- 18:OOPM Discussions & Recommendations

18:00 -19:00 PM Barbacue Dinner at Vallombrosa Center

19:30- 20:30 PM Review Discussion of Draft Resolution & Reseach Issues

14
WEDNESDAY 23, SEPTEMBER, 1998
8:00 - 9:00 AM Breakfast
[Session Chairs: Fukuwa, Stewart]
9:00 - 9:30 AM (13) "Reliability & Probabilistic Methods in SSI based on Observational
Data" (Ghanem)
9:30 - 10:00 AM (14) "Methods of analysis of SSI Effects in Bldgs & Underground
Structures" (Luco)

10:00- 10;30AM Coffee Break

[Session Chairs: Tamori, Bielak]
10:30 - 11:00 AM (15) "Soil-structure Interaction Effect on an NPP Reactor Building
(Testing by NUPEC, Achievements and the Current Status)" (Kinoshita)
11:00 - 11:30 AM (16) "Uncertainties in SSI Analysis: Modeling and Examples" (Ghiocel)
11:30 - 12:00 AM (17) "Soil-structure interaction analysis via fixed-based system subjected
to a modified ground motion" (Aydinoglu)

12:00 -13:00____Lunch at Vallombrosa Center

[Session Chairs: Sugimori, Todorovska]
13:00 - 13:30 PM (18) "Seismic Design Procedure of Building Structures including SSI
Effect" (Kitagawa)
13:30 - 14:00 PM (19) "Experimental Facilities in the US that are being (and can be) used
for SSI Research" (Ashford)
14:00 - 14:30 PM (20) "Some full-scale experimental results on soil-structure interaction"
(Todorovska & Trifunac)
14:30- 15:00 PM (21) "Soil-Foundation structure Interaction of bridges" (Tseng &
Penzien)

15:00 -15:30 PM Coffee Break

[Session Chairs:Karkee, Ashford]
15:30 - 16:00 PM (22) "Three dimensional response of building-foundation systems"
(Bielak)
16:00 - 16:30 PM (23) "Dense Instrumentation in the BRI building and surrounding soil"
(Okawa)
16:30 - 17:00 PM (24) " Development of an SSI Experiment" (Celebi)
17:00- 18:OOPM Discussions & Recommendations

18:00 -19:00 PM Dinner at Vallombrosa Center

15
THURSDAY 24. SEPTEMBER, 1998

8:00- 9:00 AM Breakfast

9:00 -10:00 AM Current Research Issues on Soil Structure Interaction
10:00 -10:15 AM Coffee Break
10:15 -12:00 AM Future Needs in Observation and Research (Analysis and
Experiments)/Adoption of Resolution/Conclusion
12:00 -14:00____Lunch
14:00 - 17:30 PM Visit to USGS/Menlo Park

FRIDAY 25, SEPTEMBER. 1998

8:00 AM Breakfast
9:00 -12:30 AM John Blume Earthquake Eng. Center, Stanford university & Lunch
PM Visit to Nearby Experiment Sites is being arranged for the
participants

Adjourn

16
 LIST OF PARTICIPANTS

U.S. PARTICIPANTS

Professor Scott Ashford Dr. Dan M. Ghiocel

Structural Engineering. Stress Technology Inc
9500 Oilman Dr., Mail 0085 1800 Brighton-Henrietta Town Line Road
University of California at San Diego, Rochester, New York 14623
La Jolla, Ca. 92093-0085 (T) 716-424-2010, (F) 716-272-7201
(T) 619-822-0431, (F) 619-534-6373 [email protected]
[email protected]
Professor Enrique Luco
Professor Jacobo Bielak Dept. of Civil Engineering
Dept. of Civil & Environ. Eng. University of California
Carnegie-Mellon University La Jolla, Ca. 92093-0085
Pittsburg,PA15213 (T) 619-534-4338, (F) 619-822-2260
(T) 412-268-2958, (F) 412-268-7813 [email protected]
[email protected]
C. -M. Mok
Professor Ronaldo I. Borja Geomatrix Consultants Inc.
Dept. of Civil and Environmental Engineering 100 Pine St., Suite 1000
Stanford University San Francisco, Ca. 94111
Stanford, CA 94305-4020 (T)415-434-9400, (F) 415-434-1365
(T) 650-723-3664, (F) 650-723-7514 [email protected]
[email protected]
Dr. Farhang Ostadan
Dr. Mehmet elebi Chief Engineer, Bechtel Corp.
USGS (MS977) 50 Beale Street, P.O. Box 193965
345 Middlefield Rd. San Francisco, Ca. 94119-3965
Menlo Park, Ca. 94025 (T) 415-768-3734, (F) 415-768-3734
(T) 650-329-5623, (F) 650-329-5163 fostadan @bechtel .com
[email protected]
Professor Joseph Penzien
Dr. C. B. Crouse Internationl Civil Eng. Consultants,
Principal, Dames & Moore 1995 University Ave., Suite 119,
2025 First Ave., Suite 500 Berkeley, Ca. 94704
Seattle, WA 98121 (T) 510-841-7328, (F) 510-8417438
(T) 206-728-0744, (F) 206-727-3350 penzien @ icec.com
seacbc @ dames.com
Professor Shahram Pezeshk
Professor Roger G. Ghanem Civil Engineering Dept.
201LatrobeHall, University of Memphis
The John Hopkins University Campus Box 526570
Baltimore, MD 21218 Memphis, TN 38152
(T) 410 516 7647 , (F) 410 516 7473 (T)901-678-4727, (F)901 -678-3026
ghanem @ jhu.edu [email protected]

17
Mr. Maurice Power, Professor Misha Trifunac
Principal Engineer Civil Engineering Department,
Geomatrix Consultants Inc. KAP216A&D,MC2531
100 Pine St., Suite 1000 Univ. of Southern California,
San Francisco, Ca. 94111 Los Angeles, CA 90089-2531
Tel: 415-434-9400, (T) 213-740-0570,(F) 213-744-1426
Fax:415-434-1365
Dr. Wen Tseng
Professor Jonathan Stewart Intemationl Civil Eng. Consultants,
Civil & Environ. Eng. Dept. 1995 University Ave., Suite 119,
5731-HBoelter Hall Berkeley, Ca. 94704
University of California (T) 510-841-7328, (F) 510-8417438
Los Angeles, Ca. 90095-1593 [email protected]
(T) 310-206-2990, (F) 310-206-2222
j stewart @ seas.ucla.edu Professor Anestis S. Veletsos
Brown & Root Professor of Civil Eng.,
Professor Luis E. Suarez, Civil Eng. Dept. P.O. Box 1892,
Civil Engineering Dept. Rice University
University of Puerto Rico Houston, Texas 77251
Mayaguez, Puerto Rico 00681-5000 (T) 713-285-5291, (F) 713-285-5268
(T)787-832-4040,(F)787-832-8260
[email protected] Mr. Bela Palfalvi
General Services Administration, Technical
Professor Maria I. Todorovska Services Branch (9PCT), Office of Property
Civil Engineering Department, Development, 450 Golden Gate Avenue,
KAP216A&D,MC2531 3rd Floor West,
Univ. of Southern California, San Francisco, Ca. 94102-3400
Los Angeles, CA 90089-2531 (T) 415-522-3183, (F)415- 522-3114
(T) 213-740-0616, (F) 213-744-1426 bela.palfal vi @ gsa.gov
[email protected]
Dr. Roger Borcherdt
Dr. Seluk Toprak USGS
USGS, MS977 345 Middlefield Road
Menlo Park, Ca. 94025 Menlo Park, Ca. 94025
(1)650-329-4864, (F)650-329-5163 (T)650-329-5619, (F)650-329-5163
stoprak @ us gs. gov [email protected]

Dr. William Joyner Dr. Cliff Astill (in Absentia)

USGS National Science Foundation
345 Middlefield Road 4201 Wilson Blvd.
Menlo Park, Ca. 94025 Arlington, VA. 22230
(T)650-329-5640, (F)650-329-5163
[email protected]

18
 JAPANESE PARTICIPANTS

Prof. Nobuo Fukuwa Prof. Yoshikazu Kitagawa

Center for Cooperative Research in Department of Structural Engineering , Faculty
Advanced Science & Technology of Engineering, Hiroshima University, 1 -4-
Nagoya University, Furo-cho, Chikusa-ku, 1 ,Kagamiyama,Higashi Hiroshima 739-
Nagoya 464-8603, Japan 8527,Japan
(T) 81-52- 789-3757, (F) 81-52- 789-3768 (T) 81-824- 24-7794, (F) 81-824- 24-7794
[email protected] [email protected]

Dr. Masanori liba Prof. Shin-ichiro Tamori,

Head, Geotechnical Engineering Division Department of Architecture and Civil
Structural Engineering Department Engineering, Faculty of Enginnering
Building Research Institute, Ministry of Shinshu University, 500 Wakasato, Nagano,
Construction, Tatehara-1, Tsukuba, Nagano 380, Japan
Ibaraki 305-0802, Japan (T) 81-26- 224 2750,(F)81-26- 224-2750
(T)81-298- 64-6637,(F)81-298- 64-6773 [email protected]
[email protected]
Dr. Yoshihiro Nakao
Prof. Michio Iguchi Ground Vibration Division,
Dept. of Architectural Engineering, Faculty Earthquake Disaster Prevention Research
of Science and Engineering, Center, Public Works Research Institute
Science University of Tokyo, (PWRI), Ministry of Construction, 1, Asahi,
Yamazaki 2641, Noda City 278-8510, Japan Tsukuba-shi, Ibaraki-ken, 305 Japan.
(T) 81-471-24-1501,(F) 81-471-25-7533 (T)81-298-64-2211, (F)81-289 64-0598
[email protected] [email protected]

Dr. Madan Karkee Mr. Hidemi Nakashima

GEOTOP Corporation, Research Div. Structural Design Engineer, Shimizu
l-16-3Shinkawa,4F Corporation, [c/o EERC, University of
Chuo-ku, Tokyo, 104-0033, Japan California at Berkeley, 1301 South 46th
(T)81-3- 5543-4601,(F)81-3- 5543-4610 Street, Richmond, California 94804-4698]
[email protected] (T)l-510-231-9597, (F) 1-510-231-9471
[email protected] /
Dr. Masamitsu Kinoshita Hidemi @ppp.bekkoame.ne.jp
Nuclear Power Engineering Corporation
(NUPEC), Shuwa-Kamiyacho Bldg. Dr: Michihiro OHORI
Toranomon 4-3-13, Researcher, Obayashi Corporation
Minato-ku, Tokyo 105, Japan [c/o John Blume EE Center, Stanford University
(T)81 -3-3434-5695,(F)81 -3-3434-9487 Home address:
kitada @ nupec. or. j p 600 Willow Rd.#25,
Menlo Park, CA 94025, USA]
(T) 1-650-328-9450
E-mail: [email protected]

19
Dr. Izuru Okawa,
Head, Building Engineering Division
IISEE, Building Research Institute, Ministry of
Construction, Tatehara-1, Tsukuba, Ibaraki 305-
0802,Japan
(T)81-298-64-6758, (F)81-298- 64-6777
[email protected]

Prof. Yoshihiro Sugimura

Dept. of Architecture and Building Science,
Graduate School of Engineering, Tohoku
University, Aoba Aramaki-06, Aobaku, Sendai,
980-8579, Japan
(T)81-22- 217-7867, (F)81-22- 217-7869
[email protected]

20
 VISITING PARTICIPANTS

Professor Nuray Aydinoglu Mr. Wen Ruizhi

Bogazici University, USGS, Menlo Park
Kandilli Rasathanesi, [visiting from: IEM, Harbin, China
81220 Cengelkoy, Istanbul, Turkey
(T)90-216-332-9701, (F) 90-216-308- Dr. Li Hongjing
0183 USGS, Menlo Park
[email protected] [visiting from: IEM, Harbin, China

Professor Mustafa Erdik Seunghyun Kim

Princeton Univ. & Bogazici University, Student -UCLA
Kandilli Rasathanesi,
81220 Cengelkoy, Istanbul, Turkey Chao-Hua Lin
(T)90.216.332.1711,(F) 90.216.308.0163 Student-Stanford
[email protected]
Kossi Sama
Student-Stanford

21
 \mitm
\ v ** ^V-t * - i

Back Row: M. Trifunac, E. Luco, S. Pezeschk, R. Ghanem, I. Okawa, S. Toprak, L. Suarez, Y. Nakao, M. Kinoshita,
J.Stewart (in the back), A. Veletsos, S. Ashford (in the back), H. Li (in the back), S. Tamori, M. Ohori (in the back), N.
Fukuwa, MErdik , H. Nakashima, C.B. Grouse

Front Row: M. Qelebi, W. Tseng, J. Bielak, M. Todorovska, Y. Kitigawa, Y. Sugimura, M. Iguchi, M. Karkee, M. liba
 STATE OF THE ART
ON
SOIL-STRUCTURE INTERACTION RESEARCHES RELATING TO
RECENT STRONG EARTHQUAKES IN JAPAN

By Michio Iguchi (1) and Yuzuru Yasui (2)

ABSTRACT: This report reviews the soil-structure interaction (SSI) researches relating to three major
earthquakes recently occurred in Japan, which include the 1993 Kushiro-oki earthquake, aftershocks of
the 1993 Hokkaido-Nansei-oki earthquake and the 1995 Hyogo-ken Nanbu earthquake. The review work
is focussed on the simulation analyses conducted on the basis of simultaneous earthquake records
observed both in buildings and in the surrounding soil during these major earthquakes, as well as other
small events observed at the site. One of the most interesting results reviewed in this report is the ratio of
response spectra for the observed motions in a building to those for motions recorded on the soil surface.
Marked difference of the ratios for small ground motions and for the strong motions in the Kushiro-oki
earthquake is detected. It is also discussed that the nonlinear response analyses conducted with use of the
two-dimensional FEM for soil-structure system have resulted in considerable deviation from the
observations during the strong motions in the Kushiro-oki earthquake. Regarding the Hyogo-ken Nanbu
earthquake, on the other hand, satisfactory results of the simulation analyses are overviewed. Furthermore,
special emphasis is placed on a study of effective input motions into structures investigated on the basis of
the simultaneous observations recorded both in structures and the surrounding soil during the Hyogo-ken
Nanbu earthquake. The compared results of the peak values recorded on the foundation with those on the
free-field soil surface indicate distinctly the reduction of effective input motions into structure. The
effective input rates, which indicate the ratios of peak values of the effective input motion to the
corresponding free-field motion, are approximately 0.7 for accelerations and 0.9 for velocities. In addition
to the topics, a general trend of SSI researches for the last ten years in Japan is overviewed.

INTRODUCTION analysis methods can be chosen at our disposal among

various methods (Iguchi and Akino 1993). In the analysis
Where is the goal of soil-structure interaction (SSI) of an actual soil-structure system, however, we are obliged
researches? Although the objective may be set at various to introduce many simplifications, idealizations and
points, it is of no doubt that one of the goals is to improve assumptions in making mathematical soil models, as well
the accuracy of assessment of structural safety against as the interface between the soil and foundation. Thus,
earthquakes by taking into account the SSI effects in there are still a great gap between an actual system and the
response analyses of structures. This will lead to mitigation mathematical model.
of earthquake damage to structures. In order to fill the gaps and to rationalize the
The seismic response analyses have become possible for assumptions, comparative studies between the observed
complicated soil-structure models with taking into account and the analyzed results, so called simulation analyses
various factors owing to great progress of SSI researches in must be accumulated. Furthermore, the establishment of
the last three decades. As for methodology, for example, prediction method for structural responses to strong

(1) Professor, Department of Architectural Engineering, Faculty of Science and Technology, Science University of Tokyo, Noda City,
Japan.
(2) General Manager, "vibration Engineering Department, Technical Research Institute, Obayashi Corporation, Kiyose City, Japan.

1-1
earthquake ground motions with taking into account the described in the following chapters.
SSI effects is also needed. In the simulation or prediction
work, in either case, the accumulation of the densely OVERVIEW OF CURRENT TREND OF SSI RESEARCHES IN
observed records of soil-structure system becomes JAPAN
essential. In particular, the accumulation of many reliable
strong motion records and the establishment of a A technical meeting of Architectural Institute of
methodology to extract the SSI effects from the records Japan (AIJ) is held every year and a number of papers are
will be the key not only to capture substantially the SSI presented at the meeting. Many scientists, engineers and
effects but also to improve the accuracy of assessment of graduate students of various fields such as building science,
structural safety against intense earthquakes. structural mechanics, geotechnical engineering, earthquake
In Japan, the simultaneous earthquake observations both engineering, etc. gather in one place to exchange
in buildings and on the surrounding soil have been information or to have a discussion. The papers presented
performed since the early 1970s, and many observed at the meeting reflect the academic interest of the
records have been accumulated so far. These data, however, participants, and therefore it gives a clue to know a
have been scarcely made the best use of in tlie SSI tendency of researches in Japan. In what follows, the
researches. This may be due to lack of methodologies to current research trend on SSI in Japan is overviewed
isolate the SSI effects from the limited records or to through the papers presented at the meeting. Fig. 1 shows a
insufficient observation records to extract the effects. change of the number of papers on SSI presented at the
Furthermore, the accumulated data are limited to those for meeting during the last ten years. As seen from the figure,
small to medium ground motions, and there have been few the number of presented papers varies from around 60 to
strong motion records. It was not until the 1995 Hyogo-ken 80. It is noticed that a decline of the number had appeared
Nanbu earthquake that a certain number of simultaneous for several years from 1992 but it has recovered soon after
records have been observed for a strong earthquake and at 1995 when Hyogo-ken Nanbu earthquake occurred.
various sites. Based on the observations, a number of Though the interest in SSI has been unchanged among
studies have been presented on SSI effects for the researchers, it is needed to extend the importance of SSI
earthquake. The objective of this report is to summarize the more widely to younger generation, as well as engineers
SSI researches relating to recent strong earthquakes concerned in seismic design of structures.
including the Hyogo-ken Nanbu earthquake. Special Fig. 2 shows the change of methodology used in SSI
emphasis is placed on the simulation analyses and effective analyses. It is evidently seen from the figure that SSI
input motions into structures investigated on the basis of researches relating to analytical procedures have been in
the simultaneous observations recorded both in structures decline and the thin layered element method (TLM) is
and the surrounding soil during the earthquakes. In gradually increasing to the contrary. The analytical
addition to the topics, a general trend of SSI researches for solutions may have less meaning considering from
the last ten years in Japan is overviewed. These will be practical point of view; but can play a significant roll to be

SSO 199! 19*2 1993 I9S4 1995 I99C 1991

Fig. 1. The change of the number of papers presented Fig. 2. The change of methodology used in SSI analyses.
at annual meeting of AIJ.

1-2
 the benchmark to check results obtained by other
numerical procedures. The exploitation of simplified
methods is important to have means to obtain results easily
and quickly and the methods can be used to make up for
the analytical methods. Nevertheless, the number of papers
tends to be decreasing regrettably for these several years.
The exploitation of simplified methods should be more
evaluated.
Finally, Fig. 3 shows the change of subjects of SSI
researches. It is clearly seen from the figure that the 198B 1989 1990 1991 1S92 1993 1994 1995 1996 1997

gradual increase of papers dealing with pile groups is in

contrast to the decrease of papers on embedded
foundations. It is also noticed from the figure that after Fig. 3. The change of subjects of SSI researches.
1995, when the Hyogo-ken Nanbu earthquake occurred,
the number of papers dealing with pile groups has were 815 gals and 919 gals in north-south and east-west
increased suddenly. It is also noted that the majority of directions, respectively, which showed larger values than
papers are related to the pile groups for these two years. those observed on the free field. In spite of high intensity
This is due to heavy damage to pile groups during the of earthquake motions, damage to structures was very
earthquake. At the same time, the spread of recognition of slight not only at the site but also in the surrounding area.
importance of soil-pile-structure interaction is another The possibilities of the intense ground motions and the
reason. slight damage had been studied from various viewpoints
such as local site effects of surface geology, SSI effects,
SSI RESEARCHES ON RECENT STRONG EARTHQUAKES the effect of frozen soil, the strength capacities of the
structures against seismic loads and so on. Some peak
The 1993 Kushiro-oki Earthquake characteristics appeared in the response spectra of the
In January 1993, a strong earthquake struck Kushiro recorded motions could be explained by using a detailed
city in Hokkaido named the 1993 Kushiro-oki earthquake soil-structure model, but not enough to explain the
(M = 7.8). The simultaneous earthquake records were magnitude of the spectra (Yasui and Takano 1994; Dan
observed both in a building of Kushiro branch of the 1994). Also, the reasons of large accelerations in the
Japan Meteorological Agency (JMA) and on the soil building compared with the surrounding free-field
surface at the site during the earthquake. The peak motions are left unsettled. More detailed results
horizontal ground accelerations at the site were 711 gals investigated in the Extensive Research Committee on the
(N63E) and 637 gals (N153E) recorded by a strong 1993 Kushiro-oki earthquake established in the AIJ are
motion accelerometer (SMAC-MD) of the Building provided elsewhere (AIJ 1994).
Research Institute (BRI). The peak accelerations observed The earthquake motions have been observed thereafter
in the building with an accelerograph (JMA-87) of JMA and some other records were observed at the site. Making

5.0 5.0

1.0 10.0 o.i 10.0

PERIOD CSECD PERIOD CSEC)

(a) (b)

Fig. 4. Response spectral ratios (JMA/BRI) (a) for small event and
(b) for Kushiro-oki earthquake (after Dan 1995).

1-3
 use of these observations, comparative studies were the observed motions on the ground floor. In evaluation of
performed on response spectra for other small events and the free-field motions, the foundation was assumed to be
the Kushiro-oki earthquake (Dan 1995). One of the most rigid in plane. The peak value of the estimated
noticeable results presented in the paper may be found in acceleration motion on the free surface was 345 gals in
the difference between response spectral ratios for small EW direction, which indicated large amplification of the
events and those for the strong motions of the Kushiro-oki input motion into the structure (Dan et al. 1997). It may
earthquake. Fig. 4(a) and (b) show the ratios of the be interpreted that the difference between the free-field
acceleration response spectra for the motions observed in motion and the records at the ground floor has resulted
the building (JMA) to those for the free-field motions from SSI effects for the building. In order to confirm the
(BRI). Fig. 4(a) is for a small event with peak acceleration large SSI effects presented in the paper, however, careful
(PA) of 19.3 gals and Fig. 4(b) is for the Kushiro-oki examination will be needed on effects of the location of
earthquake with PA of 919 gals on the soil surface. The accelerometers and the assumption of rigid foundation as
observed results are compared in the figures with those well.
computed by two-dimensional (2-D) FEM with taking
non-linearity of the soil into account. An inspection of The 1995 Hyogo-ken Nanbu Earthquake
these results indicates that for small earthquake the ratio Some earthquake records were observed on the ground,
becomes smaller than 1.0 in shorter periods less than 0.3 inside of structures and simultaneously both in buildings
sec, while for the Kushiro-oki event the ratio becomes and on the surrounding soil during the 1995 Hyogo-ken
greater than 1.0 in wider range of 0.1 to 1.0 sec. It is also Numbu earthquake (M = 7.2). Based on the records, some
noticed that the computed results can not fully explain the simulation analyses have been conducted for heavily or
measured results in magnitude for the strong motions. slightly damaged structures with consideration of SSI
Recalling the fact that the 2-D models tend to effects and many suggestive results have been presented
overestimate larger values of damping coefficients than so far. Kurimoto (1996) performed a simulation analysis
the 3-D models (Luco and Hadjian 1974), the discrepancy for a 41-story reinforced concrete building taking SSI
might be mainly due to employment of 2-D model instead effects into account when subjected to the observed
of 3-D model. It should be also noticed that the tendency motion at 70m below the ground surface. For this
recognized in the observed results in Fig. 4(b) is building, accelerometers are densely instrumented not
consistent with the parametric studies of nonlinearity only in the superstructure but also hi the soil (Yasui et al.
effects of the surrounding soil on the structural response 1995). The peak horizontal accelerations observed on the
presented by Miyamoto et al. (1995). ground floor during the main shock were 86.8 gals in EW
and 60.7 gals in NS directions. The methodology used in
The Aftershock of 1993 Hokkaido-Nansei-oki Earthquake the analysis was 2-D FEM incorporated with non-linearity
In July 1993, different large accelerations were of soil. Fairly good agreement between the observed and
observed on the ground floor of two-story school building the computed was presented both in the soil and hi the
during an aftershock (M = 6.5) of the 1993 Hokkaido- superstructure. This paper also indicates that the lateral
nansei-oki earthquake (M = 7.8). The peak horizontal and soil of the embedded foundation becomes less effective
vertical accelerations observed were 393 gals (NS), 1,569 due to non-linearity of the soil when the foundation is
gals (EW) and 575 gals in UD (up-down) direction. The supported on a firm soil. Prior to this study, Fujimori et al.
building is L-shaped of about 55m long in one side and (1995) had studied the foundation input motions for the
supported on the soil having shear wave velocities of building subjected to ground motions observed during the
about 70 to 170 m/sec to the depth of 8.6m. In spite of the Hyogo-ken Nanbu earthquake and the aftershocks as well.
surprisingly large accelerations, there was no damage to The comparison work indicated insignificant difference
structural elements of the building. Nonlinear simulation between the main and aftershocks. Perhaps, this is due to
analyses of the structure were conducted with taking into small rigidity of the lateral soil of the embedded
consideration of SSI effects in order to explain the no foundation not only for the main shock but also for the
structural damage to the building (Dan et al. 1997). As a aftershocks.
consequence of detailed studies, the reason of no damage Tamura et al. (1996) conducted SSI analyses for
was attributed to high strength capacities of the building. damaged steel framed reinforced concrete buildings of
The surface ground motions were also estimated based on 13-story and 11-story with embedded foundations.

1-4
Through response analyses, the horizontal motion at the (300 to 800 gals) and velocities (50 to 120 cm/sec) were
base of the foundation was found to be almost same as the observed on the ground (BPJ 1996). Some of these values
free-field motion at the level. This indicates a possibility exceeded the earthquake intensity level which have been
of a method that the response analyses of superstructure generally supposed in the present seismic design (the
could be evaluated with the base fixed at the bottom of second stage of design) in Japan, i.e. about 400gals in
the foundation and being subjected to the free-field acceleration and 40 to 50 cm/sec in velocity. Despite
motions at the depth of soil. Similar results have been higher intensities of the earthquake ground motions,
presented by Hayashi et al. (1997). It was also pointed out damage to buildings designed based on the present
that the damping factor of the SSI system associated with seismic code was not so severe. One of the possibilities of
non-linearity of soil is dependent on the rigidity of the the reduced damage has been considered to be attributable
bearing soil (Tamura et al. 1996). to the effects of SSI. To confirm the SSI effects, some
Earthquake motions were observed for buildings other studies have been presented in addition to the
supported by piles and simulation analyses have been previously described simulation analyses. Hayashi et al.
performed for the structures. Yokoyama (1996) and (1995) and Hayashi (1996) have discussed a possibility
Kowada et al. (1997) have conducted simulation analyses that the minor damage in the severely shaken area may be
of the soil-pile-stmcture systems subjected to the attributable to uplift of foundations.
observed motions. In the analyses, Penzien's models In order to develop the discussions about the SSI
(Penzien et al. 1964) for the soil-pile system was effects from a different viewpoint, Hayashi et al. (1998)
successfully used to explain the observed motions of have performed systematically nonlinear response
superstructures. analyses by supposing mid-to-low-rise buildings be
located in the severely damaged area in Kobe city.
EFFECTIVE INPUT MOTIONS IN THE HYOGO-KEN NANBU Through the numerical investigation, a possibility of
EARTHQUAKE reduction of effective input motions into structure was
suggested. The reduction rates were estimated to be about
Review of the Researches 30% for accelerations and 10% for velocities. It was also
In addition to the findings obtained through simulation shown that the reduction of effective input motions due to
analyses, there are some noticeable studies focused on SSI effects is pronounced for the low-rise buildings.
why the damage to structures had been comparatively The effective input motions into structure have also
minor in the Hyogo-ken Nanbu earthquake considering been studied for a different eight-story building of steel
the intense earthquake ground motions. In the severely structure by assuming incident wave motions at the
damaged Kobe City area, high intensity of accelerations bedrock, 31m below the ground surface (Kaneta et al.

Table 1. Outline of structures.

Takami Floral Nanko power plant Nanko power plant Osaka Institute of Murano drinking water
Bldg. Smokestack (Kowada Main bldg.(Doi et Technology (Ooba & plant (Kobori et al.
(Yokoyama 1996) et al. 1997) al. 1994) Mimura 1995) 1981)
structure type RC frame RC Steel SRC frame RC frame
Cast-in place Upper:SC piles UppenSC piles Cast-in place Cast-in place
foundation type
Concrete piles Lower:PHC piles Lower: PHC piles Concrete piles Concrete piles
construction area 989.5 m2 271.6 13817.0 788.0 6531.4
height of structure 98.2m 200 31.0 65.65 37.7
depth of foundation 8.3m 6.3 4.0 12.3 5.2 14.8
length of piles 22.75 m 64.5 62~70 20.3 12 16
diameter of piles 2.3m 0.8 0.8 1.8, 2.0 2.0
number of piles 56 273 2449 24 422
Vs : subsurface layers 140~240m/s 160~350 160~350 130~1BO -
Vs : bearing stratum 380 m/s 340 340 320 -
natural period NS.EW
1.41, 1.45 sec. 1.87(NS) 0.93(NS) 0.83, 0.92 0.437,0.439
seismometer
GL-1.5m GLrO.lm GL-O.lm GL-1.5 m GL-l.Om
on the ground surface
seismometer GLOm B2F
1FL 1FL 1FL
on the foundation (GL- 10.8m)

1-5
 1996). The analyses were performed by use of 2-D FEM
incorporated with an equivalent linearization method. The
numerically obtained results have indicated that the peak
acceleration responses of foundation at the ground floor \ Osaka Insuule ol Technology

were reduced by 16% comparing to the surface ground

motions.
To confirm the SSI effects, it is desirable to study the
effects based on recorded earthquake motions. A study of
the effective input motions conducted by Yasui (1996)
and Yasui et al. (1998) is substantial one investigated
based on the records simultaneously observed both in Fig.5. Location of structures.
buildings and on the ground One of the features of the
research is that, while the sites are limited, the effective 700 O NS
input motions into structures have been investigated for D EW
A UD
strong earthquake motions. The effective input motion is 600 estimated (NS)
------ least-squares approx.
defined as the response of foundation during earthquakes,
500
which includes the effects of inertia! and kinematic
interactions (Kausel et al. 1978). 400

300
Comparison of Simultaneously Observed Motions in
Structures and Soil 200

The simultaneous records obtained both in structures

100
and on the surrounding soil have been available at several
sites in the vicinity of Kobe area. Table 1 shows an
0 100 200 300 400 500 600 700
outline of the structures at four different sites, which
maximum accleration at the ground surface (gal)
include four buildings and one smokestack. The location
of the sites is shown in Fig. 5. In this figure, location of Fig. 6. Peak accelerations at the ground surface
Buildings A (Shin-Nagata Urban Complex) and Building and on foundations.
B (NTT Kobe Ekimae Bldg) is included, which will be
discussed afterward
120 O NS
Fig. 6 shows the compared results of the peak a EW
A UD
accelerations recorded on the foundations with those on 100 estimated (NS)
least-squares approx.
the soil surface. Similarly, Fig. 7 shows the compared
results of the peak velocities numerically obtained by 80

integration of the acceleration records. The location of the

60
observation points in the ground was 0.1 to 1.5m below
the soil surface and the records may be considered to be
40
the surface motions. The seismic sensors were installed
on the ground floor for three buildings and a smokestack, .E 20
and the rest was instrumented on the second basement
(10.8m below ground line). The results obtained on the
0 20 40 60 80 100 120
second basement may be considered to be those on the
maximum velocity at the ground surface (kine)
ground floor level because of high rigidity of the
basement. In Figs 6 and 7, symbols o, o and A correspond Fig. 7. Peak velocities at the ground surface
to the results of NS (north south), EW (east-west) and UD and on foundations (kine = cm/sec).
(up-down) components, respectively. The slope of the line
obtained by the least-squares method is 0.7 for the those of the soil surface. In other words, the peak
observed accelerations. The value represents the weighted -accelerations on foundations have reduced by 30%
average ratio of peak accelerations on the foundations to comparing to the surface ground motions. As for velocity,

1-6
 on the other hand, the slope is 0.9, which exhibit less
reduction rate comparing to the case of accelerations. It
& 103
will be noticed that while the vertical components are E

included in above described results, the reduction rates

are not affected even when they are excluded. Fig. 8
shows the comparison of response spectra (5% damping)
for accelerations recorded on the foundation and on the
soil surface (1.5m below soil surface) (Yokoyama 1996).
Exhibited are for the records having showed the largest 10'
0.1 1.0
acceleration among the sites shown in Table 1. It is period (sec.)
.noticed from the figure that the reduction of the effective
input motions can be recognized in wide range of periods Fig. 8. Response spectra of motions on the ground surface
less than 2 sec. and on the foundation (5% damping).
The results in the severely shocked area, so called
heavily damaged belt zone, are not included in Figs 6 and due to the structural vibration is negligible small
7, as the simultaneous earthquake records have not been comparing to the site nonlinearity or the primary
obtained in the area. To make good the loss, the surface nonlinearity (Roesset andTassoulas 1982).
ground motions are estimated numerically based on Letting the time histories of horizontal foundation input
records observed on the foundations. A method of
motion and free-field motion on the surface be uc (t) and
backward analyses will be described in what follows
(Yasui et al. 1996). In Figs 6 and 7, thus obtained results UF (t), and corresponding Fourier transforms be /c (eo)
are plotted by A and B together with other observed
results.
and UF (eo), respectively, the foundation input motion

Estimation Method ofSurface Ground Motions

Uc (to) can be expressed in terms of the transfer function
The primary step is to construct a vibration model of
superstructure and the restoring force characteristics of
of the foundation input motion H(eo) as follows.
each story. Next step is to compute the inelastic response
for the fixed base model when subjected to the observed
Uc (o>) = H(o>)UF (o>) (1)
motions on the foundation or alternative motions recorded
on basement of the structure. Thus calculated results are Let the total shear force response of the superstructure
compared to the observed records at different floor levels
be q(t) and its Fourier transform be 0(0)), which have
in order to confirm whether the constructed vibration
model can appropriately reproduce the actual response of been evaluated hi the primary step, then the horizontal
the structure during the earthquake. equation of motion of the foundation can be expressed as
The second step is to evaluate the impedance functions follows.
and the foundation input motions for the layered soil
- moa> 2 (Uc (6>) + U0 (eo))+Kff (a>)U0 (o>) = Q(co) (2)
models at the sites. In the analyses, the thin layered
element method is employed that lias capability of taking
In which /?; is mass of foundation, KH (co) is the
into account the embedded foundations (Tajimi 1984;
Takano et al. 1992). If strain dependency of soil is known horizontal impedance function of the foundation and
a priori and the soil system could be replaced by an U0 (CD) is the sway motion of the foundation due to the
equivalent linear medium, the surface ground motions can
be estimated by an iterative procedure described below. In horizontal shear force. The Fourier transform of the
the backward analyses, the followings are assumed: (1)
observed horizontal motion on foundation UB (co) may
Rocking motion is negligible; (2) incident waves impinge
to foundation vertically; (3) a foundation is bonded
be expressed by the sum of Uc (co) and U0 (cai). Thus,
completely to the soil; and (4) the effect of local
nonlinearity on the soil surface resulting from the stresses
>) (3)

1-7
Substituting from equation (3) into (2) yields, shape of the foundation
Initial stiffness of soil layers
estimated from N-value

Up (co) = \UJca)-
H(fo)\ * Ka (co)

(4)
The time history of ground motion on the free surface
uF (t) can be calculated by an inverse Fourier transform of
end of loop

Up (co). Based on thus obtained u F (t), the free-field

motions of soil system can be evaluated by an equivalent

observed wave al the basement
linearization method with taking into account the strain and
base shear lime history obtained from
non-inea/ simulation
dependency of a layered soil medium. Then follows the
Fig. 9. Flow diagram of backward analysis.
analyses of the impedance function KH (co) and a

transfer function of the foundation input motion H(co)

for the modified soil constants. Using the newly obtained E-W Section

KH (co) and H(co), the free surface motions may be

Location of
recalculated by equation (4). This procedure must be i Seismographs
repeated until it converges within an admissible limit. Fig.
9 shows a schematic figure of above mentioned procedure.
One of the features of this procedure is to make it possible
to estimate the free-field motions on soil surface with
consideration of both nonlinearities of soil and
superstructure. In the above formulation, only a sway
motion of the foundation is taken into account. This Fig. 10. Plan and section views of Building A.
assumption is valid as far as following two examples are
concerned, that was confirmed by
Table 2. Observed maximum accelerations (gal).
observing the transfer functions of
NS EW UD-
soil-structure-foundation system to
observed Interpolated Observed interpolated Observed interpolated
an incident seismic wave with
24F 635* 956 302* 354 327 -
consideration of the effect of - - -
5F 379* 407 183
rocking motion. - -
B1F 315* 354 121 119

Estimation of Free-Field Motion for Table 3. Soil parameters.

Building A -The Shin-Nagata initial converged
Vs' h'
Urban Complex of the Housing and depth (m) soil classification Vs density
(m/sec) (ton/in3) (m/sec)
Urban Development Corporation
GLO-2. 8 Fill 125.0 1.8 114-122 0. 03-0. 05
This building is built in front of 2. 8-5. 0 Sand 210.2 1.8 200 0.04
Shin-Nagata station and a high rise 5. 0-6. 8 Clay 209.7 1.8 192 0.04
building of 25-story with three 6. 8-10. 9 sand and gravel 324.3 1.9 239 0. 10
basement floors and of steel framed 10. 9-14. 0 Clay 229.2 1.8 198 0.06
reinforced concrete structure (Sawai 14. 0-17. 4 sand and gravel 337.0 1.9 219 0.13

et al. 1996). This is located in 17.4-21.8 medim stiff clay 271.9 1.8 233-236 0.06
21.8-26.3 silty fine sand 313.9 1.8 276 0.05
severely shaken zone of the
26. 3-29. 8 medium stiff clay 278.3 1.8 233 0.07
intensity W of JMA (the Japan
29. 8-37. 0 stiff "clay 296.0 1.8 250-252 0.06
Meteorological Agency) scale in the 37. 0-40. 0 gravelly fine sand 450.0 1.9 306 0.12

1-8
 24th floor : N-S direction
_ 1000

1.00

-1000
20.0

~ 1000 stillness reduction ratio (day)

stillness reduction ratio (sand)
damping factor (day)
damping (actor (sand)

0.00
lO" 1C"3
-1000 strain
15.0 20.0
time (sec.)

Fig. 11. Comparison of computed motions Fig. 12. Characteristics of strain dependency of soil.
and the observed.

Hyogo-ken Nanbu earthquake. Fig. 10 shows a plan view type models, and were determined by the sum of those of
of an upper story and a cross section in the short direction structural frames and the nonstructural planes. The
of the structure. damping factor of superstructure was assumed to be 3%
As for the damage of this structure, the damage in the for the first mode and to be proportional to the initial
long side direction (NS direction) was larger than that in stiffness of each story. The fundamental period of the
the short direction. The damage in the long direction was superstructure when fixed at the first basement floor is
concentrated especially on beams and non-structural walls 1.07 sec, and the system period including the effect of soil
of the inner frames. There were detected large shear is 1.08 sec.
cracks in the non-structural walls and shear failure in The inelastic response analyses were performed for the
beams and around openings of the partitions as well. fixed model of superstructure subjected to the horizontal
Accelerographs (SMA-1) have been installed in the seismic motion recorded at the basement floor. It should
building since it was built, and successive seismic be noticed that while the obtained results are for the fixed
observation has been made thereafter (Kusakabe 1997). In base condition, the results could be interpreted to have
the Hyogo-ken Nanbu earthquake, the seismic motions included the effects of SSI. This is valid as far as the
were successfully recorded at three points, i.e. 24th and 5* rocking motion is negligible. The computed results of
stories, and on the first basement. Some parts of the acceleration responses of NS component at the 24* and 5th
records, however, were truncated around peaks as the floors are compared to the observed motions as shown in
peak values of the motions have largely exceeded the Fig. 11. It is noticed that the agreement between two is
values expected at first. The truncated parts of the excellent up to 10 sec.
motions were interpolated by use of a spline function, and Next step is to compose a soil model. Since the soil
peak accelerations of the records and reproduced motions data of wave velocities of each layer were not available
are shown in Table 2. The values with asterisks in Table 2 around the site, the soil constants were assumed based on
indicate the peak values of the truncated records. The the standard penetration test values (SPT N-values) given
backward analysis of the free-field surface motions at this in the boring logs. Table 3 shows the initial soil constants
site was made in the NS direction, which showed the
assumed in the analyses. The strain-shear modulus (G-y )
maximum value among the records.
The lumped mass model of the superstructure used in
and strain-damping factor (h-y) relationships, which are
inelastic response analyses was determined by a threee
dimensional (3-D) static analysis of the structure when required in the equivalent linearization analysis, were
fixed at the first basement floor (Sawai et al. 1996). The determined based on the existing soil data (Ishihara 1976)
3-D model consists of six planes of structure that include as shown in Fig. 12. The free-field motions on soil surface
two non-structural planes with beams and nonstructural were estimated by the aforementioned method using the
walls above fifth floors. The restoring force soil model. Figs 13 and 14 show the horizontal impedance
characteristics of each story were assumed to be tri-linear function and foundation input motion. The figures include

1-9
 \ ' i
the results for the initial soil constants (step-0) and those T~ ~] I

'

for the successive iteration (step-1) are shown. In the 1" 2.0 107
"c ?'
.__'

,i

iterative analyses, the equivalent soil constants have o f .-ii

Sc 0.010
converged by a single iteration. Thus obtained results may
be considered as the soil constants to have shown during I -2.0 107
__
step-0 real
step-0 imag.
step-1 real
S
s
the earthquake. Table 3 includes the estimated shear wave step-1 imag.
-4.0 107
1 10
frequency (Hz)
velocities Vs and damping factor ti of each soil layer
Fig. 13. Impedance functions of the foundation.
during the strong earthquake ground motions.
The estimated free-field motion on the soil surface is
compared to NS component of the observed motions at
the first basement as shown in Fig. 15. As for peak values,
the observed peak acceleration 354 gals at the first
basement level after the correction of the records using
the spline function is smaller than the estimated free-field
peak value 515gals, which corresponds to about 30%
reduction of input motion to superstructure. Fig. 16 shows
distribution of the computed peak accelerations in the soil frequency (Hz)

and superstructure, and the observed peak accelerations

Fig. 14. Foundation input motions of the foundation.
are also included. It may be noticed that the amplification
of the free-field motions above the base of foundation is
significant and also the observed motions at the basement
- observed al the basement
is almost same as the free-field motion at the base of 600 - estimated tree-Held motion

foundation (Tamura et al. 1996; Hayashi et al. 1997).

In order to confirm the reduction of the effective input I.
motions into superstructure, comparison is made for the 354gal
response motions at the ground floor when subjected to -600
2.0 4.0 6.0 8.0 10.0 12.0
the observed motion at the first basement and to the
time (sec.)
estimated motion on the free surface. Fig. 17 shows the
compared results of time histories of the calculated
acceleration motion on the ground floor and the estimated Fig. 15. Estimated Acceleration motion on the ground
surface and observed motion on the first basement
surface motions. As for peak values, the calculated motion
on the ground floor of 398gals is about 23% smaller than
the estimated value of 515gals on the soil surface. The
comparison of peak velocities resulted in 65 cm/sec on
the soil surface and 58 cm/sec at the ground floor, and it
corresponds to about 10% reduction. These results are
plotted by A in Figs 6 and 7, which show a similar
tendency as those of the observed ones. The compared f 20

results of the response spectra (5% damping) for

calculated acceleration motion at the ground floor and for
Iree field
estimated motion on the soil surface are shown in Fig. 18. structure
observed (miaipolaled)
It may be observed from the figure that a slight reduction o obwrved (original)
of input motions at around 1.0 sec and larger reduction
200 400 600 BOO 1000
may be recognized in periods below 0.4 sec. This may be maximum response acceleration (gal)

interpreted as the effects of the foundation input motion

shown in Fig. 14. Fig. 16. Distribution of the peak accelerations
of the structure and in the soil.

1-10
 -simulated response at the 1st floor
- estimated Iree-lield motion

lo

. 515gal
-600
2.0

time (sec.)
10.0
Fig. 17. Comparison of estimated surface motion
and calculated motion on the ground floor. Fig. 18. Comparison of response spectra.
O maximum response of the primary frame
(Input: simulaled wave at the 1st floor)
a maximum response of the secondary frame
(input: simulated wave at the 1st floor) sei!smo! iraphs seismographs
maximum response of the primary frame
(input: estimated free-field motion) ' ' ,920 ^
i* maximum response of the secondary frame |
(input: estimated free-field motion) '5.050 Is

'5.050 ^
'5.050 F

>,050 I*

5.050 ^
3V3 =*C =*C 3C
s
^ 5.050 r s
VfiL
III
6.500 T1 seismographs
*"""""
|24F|
1.0 2.0 3.0 4.0 5.0 6.0
' 6.100
|
r*
interstory displacement 6 (cm)

41.600
Fig. 19. Maximum shear force responses 41.600
Section Site Plan & B3F Plan
of each story.
Fig. 20. Plan and section views of Building B.

In order to investigate the effects of SSI on the The building is located in front of Kobe station, and is
response of the superstructure, the response analyses were a steel framed reinforced concrete structure of eight-story
conducted for the structure when fixed at the base of the above the ground and with three basement floors
first story and excited by the estimated motion on the soil (Ninomiya 1996). The base of foundation is directly
surface. The results are compared to the response to the supported on a soil layer of gravel with sand beneath 14.5
calculated motion at the ground floor, which may be m below the ground level. Fig. 20 shows the cross section
considered to be equal to the response to the observed view in east-west direction and a plan of the third
motion at the basement level. It should be noticed that the basement.
former result does not include the effects of SSI and, on The building suffered damage to shear walls with
the other hand, the latter includes the effects of SSI. openings at the second to fifth stories in the north-south
Therefore, the differences between these two may be direction, and slight cracks of about 1mm width were
interpreted as the effects of SSI. The compared results of detected in other shear walls. Accelerometers (SMAC-B2)
the maximum relative displacement of each story are had been installed on the third basement and on eighth
shown in Fig. 19. It may be seen from the figure that the story of the building since it was built, and seismic
difference between two is small at sixth and ninth stories. Observation had been carried out successively. In the
Whereas the difference between two become pronounced Hyogo-ken Nanbu earthquake, the seismic motions were
above 12th floors, and especially the responses at 16* and successfully recorded, and the records of aftershocks as
24th stories calculated with consideration of the SSI well. The peak accelerations observed during the Hyogo-
effects are smaller than those calculated without the ken Nanbu earthquake are shown in Table 4.
effects of SSL The backward estimation of the free-field motions was
performed in the north-south direction of the structure
Estimation of Free-Field Motion for Building B -NTT (long side direction), which showed the largest value
Building in Front ofKobe Station- among the records.

1-11
 3.U W i i i rrT-\ " i i i | T
II

8th floor : N-S direction

f 1.5 107 '.
* *

o ... --
T^.

g 0.010 >" -~.

,
*S
T3
10.0 15.0 step-0 real
I" -1.5 107 step-0 imag.
time (sec.) ~ step-3 real
step-3 imag.
Fig. 21. Comparison of computed response -a n m7
1
frequency (Hz)
at the 8th story and observed motion.
Fig. 23. Impedance functions of the foundation.

frequency (Hz)
Fig. 22. Nonlinear characteristics of soil.
Fig. 24. Foundation input motions of the foundation.

A lumped mass system of the fixed base used in viscous type. The damping factor was assumed to be 3%
inelastic response analysis is determined on the basis of for the first mode and to be proportional to the initial
the model used in the structural design. The restoring stiffness of each story. It should be emphasized that the
force characteristics of each story was assumed to be of model is made for the purpose of a backward estimation
Takeda's model (Takeda et al. 1970) and determined by of the free-field motions and not for evaluation of the
modifying the yield strength and the second stiffness after structural design of this building. The calculated period of
cracks which had been used in design of the structure. the structure when fixed at the bottom of the basement
.Special consideration was made in the modification so was 0.75 sec and the period including the effects of soil
that the analyzed results correspond to the observed was 0.76 sec.
acceleration motions at the eighth story, as well as the The inelastic response analyses were conducted for the
relative displacement between the eighth and the third motion recorded at the basement. The computed response
basement floor. The damping was assumed to be of motion at the eighth floor is compared to the observed

Table 4. Peak accelerations of observed motions (gal).

NS EW UD
8F 881 504 408
B3F 331 154 169

Table 5. Soil parameters.

Initial converged
Vs Vs' h'
depth (m) soil classification Density
(m/sec) (tonV) (m/sec)
GLO-2. 0 Fill 90.0 1.6 66-85 0. 04-0. 15
2. 0-5. 0 Sand 130.0 1.8 46-73 0. 22-0. 28
5. 0-10. 0 sand and gravel 190.0 1.9 75-105 0. 22-0. 27
10. 0-20. 0 Sand 250.0 1.9 100-134 0. 23-0. 27
20.0-38.0 Sand 410.0 2.0 249-294 0. 16-0. 20
38. 0-46. 0 Clay 410.0 2.0 345-351 0. 09-0. 10
46. 0-52. 0 sand and gravel 410.0 2.0 221-228 0. 22-0. 23
52. 0-54. 0 sand and gravel 360.0 1.9 147 0.27
54. 0-65. 0 Clay 360. 0 1.9 248-259 0. 16-0. 17

1-12
 simulated response at the 1st floor
......... estimated free-field motion

3 0

-800
5.0 10.0 15.0
15.0
time (sec.)
time (sec.)

Fig. 25. Comparison of estimated surface motion Fig. 27. Comparison of estimated surface motion
and observed motion on the 1* basement. and computed motion on the ground floor.

10.0

Fig. 28. Comparison of response spectra.

-tree field
-structure
observed

200 400 GOO BOO 1000

maximum response acceleration (gal)
(input: simulated wave at me in now!
maximum response rt me primary frame
(Inpul: estimated Iree-fleM motion)

Fig. 26. Distribution of the peak accelerations 02 4 E 8 10 12

of the structure and in the soil. interstory displacement <5(cm)

Fig. 29. Maximum shear force response

of each story.

results as shown in Fig. 21. Fairly good agreement low strain level of the soil, and those for the final soil
between the analyzed and the observed may be seen constants after convergence are shown. In the calculation,
except the coda part of the motion after 13 sec. three times of iteration was needed to converge. Table 5
The soil model, on the other hand, was determined shows the final shear wave velocities and damping factors
based on the data obtained by PS logging tests that were of soil layers together with the corresponding initial
carried out after the earthquake. Table 5 shows the initial values. Remarkable reduction of shear wave velocities
soil constants of the soil model. As for strain dependency from the initial values is seen.
of soil (G-y and h- y relationships), a modified Ramberg- Thus estimated free-field motion on soil surface is
compared to the observed motion on the third basement
Osgood model was employed in the analyses. Fig. 22 floor as shown in Fig. 25. Regarding the peak values, the
shows the strain dependency models of silt and sand used estimated peak acceleration on the soil surface was
in the backward analyses. 633gals and has increased by 90% comparing to the
The Free-field motion on the soil surface was estimated observed peak on the third basement floor.
by means of the iterative procedure described earlier. Figs Comparing the free-field motion shown in Fig. 25 to
23 and 24 show impedance functions and foundation the previous results of Building A, followings may be
input motions for the foundation. In these figures, the observed: (1) for the case of Building A, the motions on
results for the initial soil constants, which correspond to a the soil surface and the associated motion in the building

1-13
are similar as shown in Fig. 15; (2) for the case of estimated free field motion and for the calculated motion
Building B, on the contrary, a pronounce difference on the ground floor are shown. The results indicate the
between the motions on the soil surface and in the reduction of the effective input motion observed through
structure may be recognized in magnitude and in phase as the response spectra and the reduction may be recognized
well. This is mainly due to the difference of the in wider range of periods below 1.5 sec comparing to the
nonlinearity rates of soil between the two cases. In other case of Building A shown in Fig. 18.
words, the nonlinearity rates have affected greatly on the Further comparison is made for the inelastic responses
results of the foundation input motions as shown in Figs of superstructure with base fixed at the ground floor when
14 and 24 for Building A and Building B, respectively. subjected to both the estimated free-field motion on the
It may be noticed from these figures that the foundation soil surface and the calculated motion on the ground. The
input motion for Building A is scarcely affected by computed results of relative story displacements are
nonlinearity of soil up to around 10 Hz. For the case of shown in Fig. 29. It may be noticed from the figure that
Building B, on the other hand, a pronounced difference the maximum responses to the estimated motion on the
may be detected in the results of the foundation input free surface become large at all stories except the eighth
motion for the initial soil model and the final one. Fig. 24 story, especially at intermediate stories of fifth and sixth.
indicates also that the foundation input motion is It should be noticed that the increase of input motions as
amplified by strong non-linearity of soil in frequencies large as 30% have caused a larger difference in inelastic
more than 2 Hz. responses, that indicates the importance of evaluating
Fig. 26 shows the distribution of peak accelerations of properly the effective input motions to structures.
structure and in the soil as well. In this figure the peak
values of the observed motions are also plotted. The result Summary ofDiscussions
indicates a large amplification of the ground motions in The followings can be summarized based on the study
the surface layers above -5 m. It is also noted that the described in this chapter, which is focussed on the
observed horizontal peak value in the basement is almost reduction effects of effective input motions into structures
same in magnitude as'the free-field motion at the bottom during the Hyogo-ken Nanbu earthquake.
of the foundation. This suggests that the earthquake 1) The compared results of the peak values recorded on
motions transmitted to structure predominantly from the the foundation with those on the free-field soil surface
bottom of the foundation and less from the side of the have indicated distinctly the reduction of effective
foundation as far as this building is concerned. input motions into structure. The effective input rates,
In order to investigate the reduction effects of the which indicate the ratios of peak values of the
effective input motions at the ground surface level, the effective input motion to the corresponding free-field
estimated free-field motion on the soil surface is motion, were approximately 0.7 for accelerations and
compared to the response motion on the ground floor of 0.9 for velocities. It is interesting to compare the
the structure subjected to the horizontal motion observed results with those for the Olive View Medical Center
at the third basement. In the response analysis, the base is building during the Northridge earthquake (Ms = 6.8)
fixed at the base of the third basement of the structure. of 1994 and the Whittier earthquake (Ms = 5.9) of
Fig. 27 shows the compared results of acceleration time 1987 (Celebi 1997). In the Northridge earthquake, the
histories estimated on the soil surface and the computed reduction rates of accelerations at the level of ground
motion on the ground floor. Observing the peak values, it floor were 0.9 for NS component and 0.69 for EW
may be seen that the peak value of free-field motion on component. On the other hand, the rates for the
the soil surface 633 gals has reduced to 423 gals at the Whittier Earthquake, in which the peak accelerations
ground floor, and the reduction rate became 33%. As for at this site were about one seventh to fifteenth of those
velocities, the peak value of 105 cm/sec on the soil of the Northridge earthquake, were 1.0 and 1.20 for
surface reduced to 101 cm/sec on the ground floor, and NS and EW components, respectively. Perhaps it is
the reduction rate was 4%. These values are plotted in worth notice that the reduction rates may be affected
Figs 6 and 7 by the mark B. The reduction rates are not only the frequency characteristics of ground
consistent with those obtained by the observed motions motions but also the intensity of the ground motions.
for accelerations, and are somewhat smaller for velocities. Further substantial studies are needed to reach the firm
In Fig. 28, the response spectra (5% damping) for the conclusion about this subject.

1-14
2) In order to make up for the lack of the simultaneously against earthquakes. From such a viewpoint, this report
observed data in the severely shaken area during the lias reviewed the SSI researches relating to three major
Hyogo-ken Nanbu earthquake, the backward analyses earthquakes recently occurred in Japan. In the review
of the free-field motion on the soil surface on the basis work emphasis was placed on the simulation analyses
of the motion observed in the structure were conducted based on simultaneously observed earthquake
performed at sites of two buildings. The estimated motions, as well as researches on input motions into
free-field surface motions to corresponding motions in structure for the major earthquakes and small events
the buildings exhibited the same tendencies as those observed at the same sites. In most simulation analyses, 2-
obtained based on the observed motions. D FEM incorporated with non-linearity of soil has been
3) The reduction effects were recognized in periods less used. Though the method seems to give satisfactory
than 1.0 sec by comparing the response spectra for results in the simulation analyses for small ground
free-field motions and for the motions at the ground motions, not to be acceptable for strong earthquakes.
floor. Among the reviewed researches, the most noticeable
4) The horizontal peak accelerations at the base of the results was the difference of response spectral ratios, that
foundation are almost same as the peak value of the indicate the ratios of response spectra for the observed
free-field motion at the corresponding level. At the motion in a building to those for motions recorded on the
same time, it will be inferred that above mentioned soil surface, between small and strong earthquakes
reduction effects are mainly due to the amplification of observed at the same site. A distinct result lias been
surface soil layers above the base of foundation. recognized between the response spectral ratio for small
5) To evaluate properly the effective input motions into events and that for the Kushiro-oki earthquake at a
structure is important in the inelastic response analyses specific site, in which the ratio showed greater than 1.0 in
of structures. wide period range less than 1.0 sec.
Another emphasis was also placed on a study of
This chapter is compiled based on the paper (Yasui et al. effective input motions into structures investigated on the
1998) and the summarized is outcomes of an activity done basis of the simultaneous observations recorded both in
in Sub-Working Group 5-2 (chaired by Yasui) organized structures and the surrounding soil during the Hyogo-ken
under the Special Research Committee of the Hyogo-ken Nanbu earthquake. The compared results of the peak
Nanbu earthquake established in the Architectural values recorded on the foundation with those on the free-
Institute of Japan (AIT). Many fruitful discussions field surface have indicated distinctly the reduction of
exchanged among the members are gratefully effective input motions into structure. The effective input
acknowledged. rates, which indicate the ratios of peak values of the
effective input motion to the corresponding free-field
CONCLUDING REMARKS motion, were approximately 0.7 for accelerations and 0.9
for velocities. It is worth while to notice that the result of
It was said that the present state of SSI analyses has reduced input motions into structures seen during the
reached a mature stage of development (Gazetas 1983). Hyogo-ken Nanbu earthquake is different in tendency
However, it should be also noticed that due to existing from one observed at a specific site during the Kushiro-
limitations in each analysis method we are obliged to oki event. The distinct tendencies detected between two
introduce many simplifications, idealizations and earthquakes, the Kushiro-oki and the Hyogo-ken Nanbu
assumptions in making mathematical soil-structure earthquakes, at the different sites may be attributed to
models and this gives rise to a great gap between an different degree of nonlinearity of the surrounding soil.
actual system and the mathematical model. The What we have extracted from the observed motions
examination of the assumptions and simplifications for during three major earthquakes is perhaps a portion of
validity ought to be made through simulation analyses on nonlinear SSI phenomena occurred during the
the basis of the earthquake motions, if possible, observed earthquakes. It is desirable to increase our knowledge of
simultaneously in a building and on the surrounding soil. nonlinear SSI effects based on the observations.
In particular, the accumulation of simulation analyses for One of the remained subjects on SSI to be tackled is
strong ground motions are indispensable in order to .development of simple methods to evaluate stresses in
improve the reliability for assessment of structural safety piles with taking nonlinearity of soil into account and

1-15
input motions into superstructures supported on pile reinforced concrete structure during strong motions,"
groups when subjected to strong earthquake motions. Summaries of Technical Reports of Annual Meeting,
Accumulation of the observations of earth pressures on AIJ, Vol. B-2, pp397-398, (in Japanese).
piles during intense earthquakes and strains of piles is Gazetas, G (1983). "Analysis of machine foundation
also needed to validate the methods. vibrations: state of the art", Int. J. of Soil Dyn. and
Earthq. Engrg, Vol. 2, No. 1, pp2-42.
ACKNOWLEDGEMENTS Hayashi, Y, M. Mori, K. Watanabe, K. Tamura, M. Kaneko,
T. Saito, and H. Yokota, (1995). "Simulation analyses
The authors would like to express their gratitude to M. of a damaged building during the 1995 Hyogo-ken
Nakamura of Technical Research Institute of Obayashi Nanbu Earthquake," Technical Research Report of
Corporation and T. Mutoh of Science University of Tokyo Shimizu Corporation, Vol.62, pp. 37-49, (in Japanese).
for their helpful cooperation in preparation of this report. Hayashi, Y. (1996). "Damage reduction effect due to
basemat uplift of buildings," J. of Struct and Constr.
REFFERENCES Engrg, AIJ, No.485, pp.53-62, (in Japanese).
Hayashi, Y, J. Miyakoshi, K. Tamura, and H. Kawase
Architectural Institute of Japan (1994). "A general (1997). "Peak ground velocity evaluated from damage
symposium on the 1993 Kushiro-oki earthquake," ratio of low-rise buildings during the Hyogo-ken
The Extensive Research Committee on the 1993 Nanbu Earthquake," J. of Struct and Constr. Engrg,
Kushiro-oki Earthquake, (in Japanese). AIJ, No.494, pp.59-66, (in Japanese).
Building Research Institute (1996). "Final reconnaissance Hayaslii, Y, T. Fujimori, Y Yasui and M. Iguchi (1998).
report on the 1995 Great Hanshin-Awaji Earthquake "Effects of soil-structure interaction in heavily
Disaster," Building Research Institute, (in Japanese). damaged zone in the 1995 Hyogo-ken Nanbu
Celebi, M. (1997). "Response of Olive View Hospital to Earthquake," J. of Struct, and Constr. Engrg, AIJ.
Northridge and Whittier Earthquakes," J. of Struct. (submitted).
Engrg, Proc. ASCE, Vol. 123, No. 4, pp389-396. Iguchi, M. and K. Akino (1993). "Experimental and
Dan, K. (1994). "Nonlinear interaction of soil and analytical studies for soil-structure interaction: State-
structure observed in the earthquake motions at the of-the-art Report," Principal Division Lecture at the
Kushiro district meteorological observatory 12th Int. Conf. on Struct. Mech. in React. Tech.,
building," A General Symposium on The 1993 (Stuttgart).
Kushiro-oki Earthquake, AIJ, pp85-92, (in Japanese). Ishihara, K. (1976). "Introduction to geological dynamics,"
Dan, K. (1995). "Simulation of soil and structure Kajima Shuppankai, pp. 199-201, (in Japanese).
interaction observed in the earthquake motions at the Kaneta, R., M. Hisano and M. Takagi (1996). "Dynamic
Kushiro district meteorological observatory," J. of analysis of a damaged buildings by the 1995
Struct, and Constr. Engrg, AIJ, Vol. 470, pp75-84, (in Hyogoken-nanbu earthquake (Part 1 Effect of the
Japanese). basement floor on the input motion to the building),"
Dan, K., M Kikuchi, A. Fukukita and K. Nishimura Summaries of Technical Reports of Annual Meeting,
(1997). "Study on large acceleration motions AIJ, Vol. B-2, pp517-518, (in Japanese).
recorded during the largest aftershock (M6.5) of the Kausel, E., R. V. Whitman, J. P. Morray and F. Elsabee
1993 Hokkaido-Nansei-oki earthquake," J. of Struct, (1978). "The spring method for embedded
and Costr. Engrg, AIJ, Vol. 495, pp45-52, (in foundations," Nuclear Engrg and Design, Vol. 48,
Japanese). pp377-392.
Doi, T, A. Kowada, T. Matsumura, T. Kida, and N. Maeda Kobori, T, T. Kamata, and K. Kanda (1981). "Earthquake
(1994). "A study on dynamic characteristics of thermal observation at a multistory water treatment plant,"
power plant structure; (Part 1) Dynamic characteristics Proc. of Architectural Research Meetings, Kinki
of turbine building," Summaries of Technical Papers District, AIJ, pp.77-80, (in Japanese).
of Annual Meeting, AIJ, Vol. B, pp!155-1156, (in Kowada, A., N. Maeda, S. Mori, T. Dceda and Y. Maekawa
Japanese). (1997). "Verification of a high smokestack for the
Fujimori, T., Y. .Yasui, K. Wakamatsu and A. Nobata 1995 Great Hanshin earthquake motion records; (Part
(1995). "Foundation input motions of a multistory 1: Characteristics of the strong motion records, the

1-16
 design model and safety judgement)," Summaries of substructure method; (Part 6) Formulation of grouped
Technical Reports of Annual Meeting, AIJ, Vol. B-2, pile embedded foundation case," Summaries of
pp357-358, (in Japanese). Technical Papers of Annual Meeting, All, Vol. B,
Kurimoto, O. (1996). "Nonlinear seismic response of 41- pp.595-596, (in Japanese).
story reinforced concrete structure," Summaries of Takeda, T., M. A. Sozen and N. N. Nielsen (1970).
Technical Reports of Annual Meeting, AIJ, Vol. B-2, "Reinforced concrete response to simulated
pp403-404, (in Japanese). earthquakes," J. of Struct. Engrg, Proc. ASCE. Vol.
Kusakabe, K. (1997). "Earthquake response of embedded 96, ST12, pp2557-2573.
high rise building due to 1995 Hyogo-ken Nanbu Tamura, K., Y. Hayashi, M. Mori and I. Takahashi (1996).
earthquake," Report by the Special Research "Evaluation of effective input motion of damaged
Committee of the Hyogo-Ken Nanbu Earthquake in buildings in 1995 Hyogoken-Nanbu earthquake: (Part
the Kinki District, AIJ, pp!79-192, (in Japanese). 2) Effective Input motion and damping," Summaries
Luco, J. E. and A. H. Hadjian (1974). "Two-dimensional of Technical Reports of Annual Meeting, AIJ, Vol. B-
approximations to the three-dimensional soil-structure 2, pp515-516, (in Japanese).
interaction problem," Nuclear Engrg and Design, Vol. Yasui, Y. and S. Takano (1994). "Analysis of soil and
31,ppl95-203. structure interaction on Kushiro district
Miyamoto, Y, Sako, Y, Koyamada, K. and Miura, K. meteorological observatory building and the
(1995). "Seismic response analyses of a building foundation of strong motion seismograph," A General
subjected to the 1995 Hyogo-ken Nanbu Earthquake Symposium on The 1993 Kushiro-oki Earthquake,
with consideration of site condition (part-3)," Proc. of AIJ, pp75-84, (in Japanese).
the 66th Architectural Research Meetings, 1995, Kanto Yasui, Y, K. Wakamatsu, T. Fujimori and S. Yagi (1995).
District, AIJ, pp.73-76, (in Japanese). "Aseismic motions of 41-story reinforced concrete
Ninomiya, T. (1996). "NTT Kobe-Ekimae Building; - tube structure; Part 11: Seismic observation (2) -The
Seismic records and dynamic response analysis-," Hyogo-ken Nanbu earthquake -," Summaries of
Symp. on the Hyogo-Ken Nanbu Earthquake in the Technical Reports of Annual Meeting, AIJ, Vol. B-2,
Research Committee of Earthquake Resistant Structure pp947-948, (in Japanese).
in the Kinki District, AIJ, pp26-34, Nov., (in Japanese). Yasui, Y. (1996). "Effects of soil-structure interaction,"
Ohba, S. and K. Mimura (1995). "Influence of intensity of First symposium on the Hyogo-Ken Nanbu
ground motion on natural period of a building," Proc. Earthquake, The Special Research Committee of the
of Architectural Research Meetings, Kinki District, AIJ, Hyogo-Ken Nanbu Earthquake, AIJ, pp. 19-22, (in
pp205-208, (in Japanese). Japanese).
Penzien, J., C. F. Scheffy and R. A. Parmelee (1964). Yasui, Y, S. Takano, M. Nakamura, and K. Takata (1996).
"Seismic analysis of bridges on long piles," J. Engrg "A study on observed earthquake records in a 25-story
Mech. Div., ASCE, EM3r pp223-254. SRC residential building during the Hyogo-Ken
Roesset, J. M. and J. L. Tassoulas (1982). "Nonlinear Nanbu Earthquake," Symp. on the Hyogo-Ken Nanbu
structure interaction: An overview," Earthquake Earthquake, The Research Committee of Earthquake
Ground Motion and Its Effects on Structures, ed. by S. Resistant Structure in the Kinki District, AIJ, pp.35-44.
K. Datta, AMD-Vol. 53, ASME, pp59-76. Nov., (in Japanese).
Sawai, N., H. Eto, K. Tsuda, and K. Takata (1996). Yasui, Y, M. Iguchi, H. Akagi, Y. Hayashi and M.
"Response analysis of a 25-story SRC building during Nakamura (1998). "Examination on effective input
the 1995 Hyogoken-Nanbu (Kobe) earthquake," motion to structures in heavily damaged zone in the
Concrete Journal, Vol. 34, No. 11, pp.37-41, Nov., (in 1995 Hyogo-ken Nanbu Earthquake," J. of Struct,
Japanese). and Constr. Engrg, ALT, No. 512, (to appear).
Tajimi, H. (1984). "Predicted and measured vibration Yokoyama, H. (1996), "Earthquake observation and
characteristics of a large-scale shaking table dynamic analysis of a high-rise RC apartment housing
foundation," Proc: 8* World Conf. on Earthq. Engrg, subjected to the Hyogo-ken Nanbu Earthquake," Symp.
M)L HI, pp873-880. on the Hyogo-Ken Nanbu Earthquake, The Research
Takano, S., H. Maeno and Y. Yasui (1992). "Seismic Committee of Earthquake Resistant Structure in the
response analysis of grouped pile foundation using Kinki District, ALT, ppl-17, Nov, (in Japanese).

1-17
LATERAL SEISMIC SOIL PRESSURE

AN UPDATED APPROACH

By

Farhang Ostadan

William H. White

Bechtel National
San Francisco, California

Presented as part of

US-Japan SSI Workshop

September 22-23,1998

United States Geological Survey

Menlo Park, California

2-1
INTRODUCTION

The effect of ground motion on retaining walls was recognized by Okabe (1924) and Mononobe and
Matsuo (1929) following the great Kanto Earthquake of 1923 in Japan. The method proposed by
Mononobe and Okabe, currently known as the M-O method, was based on the Coulomb's theory of static
soil pressure developed more than 200 years ago. In the last 30 years, a great deal of research work both in
the analytical and in experimental areas has been performed to evaluate the adequacy of the M-O method or
to extend the method for specific applications. Discussion of the all the research work on the seismic soil
pressure is extensive and is beyond the scope of this study. Rather, only the milestones that have influenced
the design practice are described below.

Seed and Whitman (1970)

In 1970, the M-O method and the associated analytical relationships were simplified by Seed and Whitman
(1970) for design of earth retaining structures for dynamic loads. Using the charts, the designer only needs
to know the basic properties of the backfill (the angle of internal friction) and the peak ground acceleration
to obtain the seismic soil pressure. As suggested by Seed and Whitman, the basic assumptions used in the
development of the M-O method should always be considered in design applications. These assumptions
are:

The backfill materials are dry cohesionless materials.

The retaining wall yields equally and sufficiently to produce minimum active soil pressure.
The active soil pressure is associated with a soil wedge behind the wall which is at the point of
incipient failure and the maximum shear strength is mobilized along the potential sliding surface.
The soil behind the wall behaves as a rigid body and the acceleration is uniform in the soil wedge
behind the wall.

Whitman et al. (1979,1990.1991)

The effect of some of the limiting assumptions used in the M-O method above has been investigated by,
among others, Whitman et al. (1979, 1990,1991) and Nadim and Whitman (1984). The non-yielding wall
conditions and the amplifications of the motion in the soil mass were found to be significant in some cases.
However, no practical tools were proposed for design applications to circumvent the limiting assumptions
used in the M-O method. Judging from the results of model tests by several researchers, Whitman (1990)
found that use of the M-O method for design of relatively simple gravity walls up to 30 ft high is
acceptable. However, for higher walls and non-yielding walls, he recommended a more careful analysis be
performed.

2-2
 Richards and Elms (1979)
One of the more important recent developments in characterizing the seismic soil pressure for retaining
walls was the work performed by Richards and Elms (1979). Using the M-O method and the Newmark's
sliding-block analogy, the authors proposed a displacement-controlled method which incorporates basic
ground motion parameters (maximum acceleration and maximum velocity) and reduces the seismic soil
pressure based on the acceptable amount of the wall movement. In practice, the method is currently used
for designing walls for which limiting horizontal displacements are of no concern.

Wood (1973)
While the M-O method was developed for yielding walls, Wood (1973) developed an equivalent static
elastic solution for seismic soil pressure for non-yielding walls. The solution is based on finite element
analysis of a soil-wall system for a wall resting on a rigid base and a uniform soil layer behind the wall. In
general, Wood's solution amounts to a lateral force that acts about 0.63 times the height of the wall above the
base of the wall which corresponds approximately to a parabolic distribution of soil pressure unlike M-O's
inverted triangular distribution. Wood's solution predicts seismic soil pressure larger (by a factor of 2 to 3)
than the pressure predicted by the M-O method. The elastic solution proposed by Wood has been adopted
by ASCE Standards for Nuclear Structures (1986) and has been used in many applications. Wood's solution
requires knowledge of the maximum ground acceleration along with the density and Poisson's ratio of the
soil to obtain the seismic soil pressure behind the wall.

Matsuzawa et al. (1984), Ishibashi et al. (1985)

To address saturated backfill conditions and to include the hydrodynamic forces, the M-O method was
extended by Matsuzawa et al. (1984) and Ishibashi et al. (1985). A comprehensive summary of the all the
M-O based methods and their applications to various retaining wall conditions are documented in a recent
US Army publication (Ebeling and Morrison, 1992).

Veletsos et al. (1994a. 1994b)

More recently, Veletsos and Younan (1994a, 1994b) developed an analytical model to compute seismic soil
pressure for rigid vertical walls resting on a rigid base. The proposed model is based on the series of
elastically supported semiinfinte horizontal bars with distributed mass to model the soil medium behind the
wall. The model was developed for vertically propagating shear waves with the assumption that horizontal
variation of vertical displacements in the soil medium is negligible. In this model, contrary to Wood's
equivalent static solution, amplification of motion in the soil medium behind the wall is considered. The
model highlights the effects of several parameters including the frequency of vibration on the seismic soil
pressure magnitude and distribution. The model was subsequently expanded for application to cylindrical
vaults and storage buildings (Veletsos and Younan, 1994c; 1995).

2-3
Significance of Seismic Soil Pressure in Design

Seed and Whitman (1970) summarized damage to wall structures during earthquakes. Damage to retaining
walls with saturated backfills is typically more dramatic and is frequently reported in the literature.
However, damage reports of walls above the water table are not uncommon. A number of soil retaining
structures were damaged in the San Fernando earthquake of 1971. Wood (1973) reports that the walls of a
large reinforced concrete underground reservoir at the Balboa Water Treatment Plant failed as a result of
increased soil pressure during the earthquake. The walls were approximately 20 ft high and were restrained
by the top and bottom slabs.

Damage has been reported for a number of underground reinforced concrete box-type flood control
channels. Richards and Elms (1979) report damage to abutment of bridges after the 1968 earthquake in
Inangahua, New Zealand. Out of the 39 bridges inspected, 24 showed measurable movement and 15
suffered damage on bridge abutments. In the Madang earthquake of 1970 in New Guinea, the damage
patterns were similar. Out of 29 bridges repaired, some experienced abutment lateral movements as much
as 20 inches. Reports on failed or damaged bridge abutments indicate mainly settlement of the backfill and
pounding of the bridge superstructure against the abutment in longitudinal and transverse directions.

Nazarian and Hadjian (1979) also summarized damage to soil-retaining structures during past earthquakes.
Damage to bridges has also been reported from various earthquakes including 1960 Chilean, 1964 Alaskan,
1964 Nigata, 1971 San Fernando, and 1974 Lima. Most of the reported damage can be attributed to the
increased lateral pressure during earthquakes.

Numerous damage reports are available from recent earthquakes which report damage to the embedded
walls of buildings. However, contribution of the seismic soil pressure to the damage can not be quantified
since the embedded walls often carry the inertia load of the superstructure with cracks extending in all
directions in the walls of the buildings. On the other hand, simple structures, such as underground box-type
structures, retaining walls, and bridge abutments have suffered damage due to the increased soil pressure.
All of these reports and others not mentioned highlight the significance of using appropriate seismic soil
pressure in design.

RECENT EXPERIMENTS AND OBSERVATIONS

Lotting Experiment
Soil-structure interaction (SSI) effects play a significant role in the dynamic response of critical structures
and internal components. Recognizing these effects, the Electric Power Research Institute (EPRI) with the
cooperation from Taiwan Power Company (TPC) and the United States Nuclear Regulatory Commission
(NRC) sponsored a large-scale experiment in the earthquake active area of Lotung, Taiwan. The objective
2-4
of the experiment was to evaluate the SSI analyses methodologies and to reduce uncertainties in the design.
In this experiment, a 1/4-scale containment model was constructed. Instrumentation was installed both in
the containment model and at the site. Since completion of the model and its instrumentation in October
1985, a number of recordings from earthquakes ranging in Richter magnitude 4.5 to 7.0 have been made at
the site. The information on site condition, soil properties, and structural drawings were distributed to
selected teams from the industry and academia (a total of 13 groups one which was Bechtel) to predict the
responses on a round-robin basis. The results of this extensive experiment and follow up studies are
published in several EPRI reports (EPRI, 1989; EPRI, 1991). The senior author also participated in the
studies performed by Bechtel.

The Lotung site is a relatively flat with a relatively soft surface layer with thickness of 200 ft to 260 ft (60 m
to 80 m) overlying deep alluvium stratum. The soil properties in terms of low-strain shear and compression
wave velocities were measured at the site. The shear wave velocity is about 100 m/sec increasing to 250
m/sec at the depth. In addition cyclic laboratory testing was performed on soil samples and the strain-
dependent soil properties were obtained.

The instrumentation for the experiment is extensive and consists of accelerometers and pressure gages in
the model and in the free-field. Pressure gages were installed beneath the basemat for monitoring uplifting
and bonding/de-bonding of the basemat from the supporting soil layer. In addition, pressure gages were
also installed on the perimeter of the containment shell to measure seismic lateral soil pressure.

A number of earthquakes up to magnitude 7 were recorded at the site. For the purpose of this study, only
the records from earthquake event LSST07 are used. The LSST07 event occurred on May 20, 1986 at
about 40 miles(66.2 km) from the Lotung experiment. This event had a Richter magnitude of 6.5. The
peak ground acceleration in the free-field at the ground surface were 0.16g, 0.2 Ig, and 0.04g in the east-
west, north-south, and vertical directions, respectively.

A typical recorded rime histories of seismic soil pressure is shown in Figures 1. Most time histories show a
drift in the response and substantial residual pressure at the end of the shaking. Some of the pressure time
histories have also been examined by Chang et al. (1990). As suggested by Chang et al., the drift in the
time history and the residual pressure are attributed to the compaction of the backfill material during
shaking and particle re-arrangement of the materials in the soil near the instrument For this reason the
recorded pressure time histories were corrected to eliminate the drift and the residual pressure in order to
obtain the peak transient stresses. The corrected pressure time history is also shown in Figure 1 with
positive sign indicating pressure and negative sign indicating extension.

2-5
The seismic soil pressure shown in Figure 1 is the normal stress component with the direction normal to the
body of the containment shell in the North-South direction. The magnitude of the stress is a function of the
relative motion of the containment and the surrounding soil and the soil properties. In the Lotung
experiment, the relative motion was caused primarily by the rigid body rocking motion of the containment
shell. To evaluate the effect of rocking motion on the lateral seismic soil pressure, frequency contents of
the rocking motion are compared with the frequency contents of the pressure time history at one location, as
shown in Figure 2. Comparison of the pair of spectra shows that, while the nature of the spectral amplitudes
are different and are expected to have different amplitudes, the frequency content of the two motions are
very similar, particularly at the rocking frequency of the containment shell (2.2 Hz).

The overall comparison of the results (see Ostadan and White, 1997) indicates that the seismic soil pressure
is caused by the relative motion of the structure with respect to the surrounding soil and as such it is a SSI
response. This implies that the seismic soil pressure will not only be affected by the soil properties and the
characteristics of the ground motion, but also the structural properties as well as the size of the structure
and its foundation embedment.

Finally, the result of the SSI analysis using the computer program SASSI (Lysmer et al., 1981) in terms of
seismic soil pressure was obtained and compared with the recorded pressure in terms of spectral amplitudes
in Figures 3

Other Observations From Recent Field and Experimental Data

In recent years, several field and laboratory experiments have been conducted to resolve the complexities
associated with the seismic soil pressure and to develop a more realistic design parameter for the design of
embedded structures. A summary of the selected recent investigations is presented below.

Case 1 - Deeply Embedded Reactor Building

Hirota et al. (1992) have collected and studied the soil pressure data from instrumented buildings since
1989. Specifically, the data from a deeply embedded reactor building (embedment depth of 120 ft) in a
suburb of Tokyo have been presented and evaluated. The data from a total of eight earthquake records are
presented. The principal conclusions of the study are as follows:

The seismic soil pressure is significantly affected by the low-frequency content of the earthquake
motion.
Comparison of the pressure time history with the derived relative displacement time history between
the structure and the far-field shows similar characteristics in phase and amplitude.

2-6
Case 2 - Deeply Embedded Structure
Matsumoto et al. (1991) and Watakabe et al. (1992) present the results of a study using the recorded data
for a deeply embedded building in a suburb of Tokyo. The site consists of a soft alluvial layer with a
thickness of 120 ft underlain by a much stiffer formation. The shear wave velocity of the upper layer ranges
from 300 ft/sec to 1000 ft/sec. The building foundation rests on the stiff formation. The records from a
total of 21 earthquakes have been collected and examined. The main points of the investigation are as
follows:
Frequency content of the soil pressure was examined by comparing the normalized response spectra of
the soil pressure with the normalized velocity spectra of the motion in the soil layers at the respective
elevations. The shapes of the normalized spectra closely matched.
The finite element method employed was able to predict the soil-interaction effects. This conclusion
confirms the use of finite element and soil-structure interaction techniques to predict seismic soil
pressure.

Case 3 - Underground LNG Storage Tanks

Koyama et al. (1988,1992) collected and examined the earthquake and seismic soil pressure records from
two large scale Liquid Natural Gas (LNG) underground storage tanks. The instrumented tanks are large
diameter concrete tanks (200 ft diameter, 120 ft high). The site soil is a medium dense sand with a shear
wave velocity of 1300 ft/sec. Over the 8-year period, records from 70 earthquakes have been collected and
examined. The authors concluded that the seismic soil pressure is strongly correlated to the to the
acceleration and the relative displacement of the tank and the ground.

In addition to the field experiments, a number of laboratory tests have been recently performed Kazama and
Inatoi (1988) and Itoh and Nogami (1990). Evaluation of the test results showed that:

The dynamic soil pressure is amplified near the resonant frequency of the backfill sand.
The effect of soil nonlinearity on the peak dynamic pressure can be observed by increasing the
amplitude of the vibration.
The dynamic soil pressure distribution is consistent with the relative displacement between the ground
and the caisson.
Finite element analysis methods are able to reproduce the measured data.
At the soil column resonant frequency, the seismic soil pressure acts in the direction of the basement
movement to drive the structure, whereas at the structural resonant frequency, the dynamic pressure
acts in the opposite direction of the basement movement to restrain the movement of the structure.

2-7
Recognition of the Problem and Objective of the Study

In spite of the much better understanding of the soil-wall interaction behavior that have evolved over the
years, the M-O method continues to be widely used despite many criticisms and its limitations. As stated
above, the method was developed for gravity retaining walls with cohesionless backfill materials. In design
applications, however, the M-O method or any of its derivatives is commonly used for below ground
building walls. In this regard, the M-O method is one of the most abused methods in the geotechnical
practice.

In view of the overwhelming information and evidence on the dynamic behavior of buildings, some of
which was outlined above, the United States Nuclear Regulatory Commission (US NRC, 1991) recently
issued a position paper on the subject of the seismic soil pressure. Pertinent excerpts are quoted as follows:

"The use of the M-O method of analysis to compute pressure on embedded walls of structures like
the nuclear island (NI) structure of..... is not considered appropriate since the development of the
limit conditions in the soil requires wall movements which are most likely inappropriate for SSI
conditions anticipated. The M-O approach will generally lead to a lower bound estimate for soil
loads (using active state conditions in the soil) since the soil in the active wedge is assumed to
transfer part of the load to the soil below through its own shear strength..."

It is the objective of this study to develop a simple and practical method to predict lateral seismic soil
pressure for building walls.

The walls of the buildings are often of the non-yielding type. The movement of the walls is limited due
to the presence of the floor diaphragms, and displacements to allow development of the limit-state
conditions are unlikely to develop during the design earthquake.
The frequency content of the design motion is fully considered. Use of a single parameter as a measure
of design motion such as peak ground acceleration may misrepresent the energy content of the motion,
at frequencies important for soil amplifications.
Appropriate soil properties are included in the analysis. For soil dynamic problems, the most important
soil property is the shear wave velocity followed by the material damping, Poisson's ratio, and density
of the soil.
The method is flexible to allow for consideration of soil nonlinear effect where soil nonlinearity is
expected to be important.
The interaction between the soil and the building is represented. This includes consideration for the
building rocking motion, amplification and variation of the motion in the soil, geometry, and
embedment depth of the building.

2-8
SIMPLIFIED METHOD TO PREDICT LATERAL SEISMIC SOIL PRESSURE FOR BUILDING
WALLS ON ROCK OR FIRM FOUNDATIONS

In this section, the dynamic characteristics of lateral seismic soil pressure for buildings with basemat resting
on rock or firm soil layers are examined and a simplified method for predicting seismic soil pressure is
presented. It is assumed that the building walls are effectively rigid. The condition that the basemat rests
on a firm soil layer also simplifies the problem in that the rocking vibration of the buildings becomes
insignificant. With this assumption, the embedment ratio of the building (embedment depth to basemat
width) will not play a role in the results. The extension of the method for buildings embedded in deep soil
layers is presented in the next section.

To investigate the characteristics of the lateral seismic soil pressure, a series of seismic soil-structure
interaction analyses was performed using the Computer Program S ASSI. A typical SASSI model of a
building basement is shown in Figure 4. The embedment depth is designated by H and the soil layer is
identified by the shear wave velocity, Vs, the Poisson's ratio, v, total mass density, p, and the soil material
damping, P. The basemat is resting on rock or a firm soil layer. A column of soil elements next to the wall
is included in the model in order to retrieve the pressure responses from the results.

For this analysis, the acceleration time history of the input motion was specified at the top of the rock layer
corresponding to the basemat elevation in the free-field. In order to characterize the dynamic behavior of
the soil pressure, the most commonly used wave field consisting of vertically propagating shear waves was
specified as input motion. The frequency characteristics of the pressure response were examined using
harmonic shear waves for a wide range of frequencies. For each harmonic wave, the amplitude of the
normal soil pressure acting on the building wall at three locations (Elements 2, 10, and 15 in Figure 4) was
obtained. The pressure responses are presented in terms of pressure transfer function amplitudes which are
the ratio of the amplitude of the seismic soil pressure in the respective element to the amplitude of the input
motion (Ig harmonic acceleration) in the free-field for each harmonic frequency. The analyses were
performed for a building with embedment of 50 ft and soil shear wave velocities of 500, 1000, 1500, and
2000 ft/sec, all with the Poisson's ratio of 1/3. The material damping in the soil was specified to be 5%.
The transfer function results for Element 2 (see Figure 4) are shown in Figure 5. As shown in this figure,
the amplification of the pressure amplitude takes place at distinct frequencies. These frequencies increase
as the soil shear wave velocity increases. The amplitude of soil pressure at low frequency was used to
normalize the amplitude of the pressure transfer functions for each element. The frequency axis was also
normalized using the soil column frequency which was obtained from the following relationship:

f=Vs/(4xH) (1)

2-9
In the above equation, Vs is the soil shear wave velocity and H is the embedment depth of the building.
The normalized transfer functions are shown in Figure 6. As shown in this figure, the amplification of the
pressure is about the same for all the shear wave velocities considered. In all cases the maximum
amplification takes place at the frequency corresponding to the soil column frequency. Similarly, the results
for points in the mid-height and bottom of the wall were examined (Ostadan and White, 1997). These
results also showed the same characteristics described above.

Examining the dynamic characteristics of the normalized pressure amplitudes (such as those shown in
Figure 6), it is readily evident that such characteristics are those of a single degree-of-freedom (SDOF)
system. Each response begins at a value of one and increases to a peak value at a distinct frequency and
subsequently reduces to a small value at high frequency. Dynamic behavior of a SDOF system is
completely defined by the mass, stiffness and associated damping constant. It is generally recognized that
response of a SDOF system is controlled by the stiffness at low frequency, by damping at resonant
frequency, and by the inertia at high frequencies.

Following the analogy for a SDOF system and in order to characterize the stiffness component, the pressure
amplitude at low frequencies for all elements (Elements 1 through 15 in Figure 4) was obtained and plotted
as shown in Figure 7 in terms of the normalized height (Y/H, H=50 ft; Y is the distance from the base of the
wall as shown in Figure 4). The pressure amplitudes at low frequency are almost identical for the wide
range of the soil shear wave velocity profiles considered. The sudden increase shown at the top of the
profile is due to the zero stress boundary condition near the ground surface and can be improved if finer
elements are used. However, it is also generally recognized that soils particularly at shallow depths with
low confining pressure have low shear strength and are subject to softening during vibration. For this
reason, the normalized pressure profile was adjusted to have a vertical tip as shown in Figure 7. The shape
of the normalized pressure will be used as a basis to determine seismic soil pressure along the height of the
building wall. This will be discussed after the seismic soil pressure is examined for cases in which input
motion is specified at the ground surface level.

A similar series of parametric studies were also performed by specifying the input motion at the ground
surface level (Ostadan and White, 1997). The results of these studies also showed that the seismic soil
pressure in normalized form can be represented by a single degree-of-freedom (SDOF) system. For both
cases considered, the low frequency pressure profiles depict the same distribution of the pressure along the
height of the wall as shown in Figure 7. This observation is consistent with the results of the analytical
model developed by Veletsos and Younan (1994a). Since all the soil-structure interaction analyses were
performed for the Poisson's ratio of 1/3, the pressure distribution was adjusted for the soil's Poisson's ratio
using the factor recommended by Veletsos and Younan (1994a). The \ff v factor is defined by:

2-10
For the Poisson's ratio of 1/3 used in the analysis, \ff v is 1.897. Use of lf/ v in the formulation allows

correction of the soil pressure amplitude for various Poisson's ratios. The adjusted soil pressure
distribution is also shown in Figure 7. Using the adjusted pressure distribution, a polynomial relationship
was developed to fit the normalized pressure curve. The relationship in terms of normalized height, y =
Y/H (Y is measured from the bottom of the wall and varies from 0 to H), is as follows:

p(y) = -.0015 + 5.05y - 15.84y2 + 28.25y3 - 24.59y4 + 8.14y5 (3)

The area under the curve can be obtained from integration of the pressure distribution over the height of the
wall. The total area is 0.744 in terms of normalized wall height or 0.744H for the wall with the height H.

Having obtained the normalized shape of the pressure distribution, the amplitudes of the seismic pressure
can be also obtained from the concept of a SDOF. The response of a SDOF system subjected to earthquake
loading is readily obtained from the acceleration response spectrum of the input motion at the damping
value and frequency corresponding to the SDOF. The total load is subsequently obtained from the product
of the total mass times the acceleration spectral value at the respective frequency of the system.

To investigate the effective damping associated with the seismic soil pressure amplification and the total
mass associated with the SDOF system, the system in Figure 4 with wall height of 50 ft and soil shear wave
velocity of 1500 ft/sec was subjected to six different input motions in successive analyses. The motions
were specified at the ground surface level in the free-field. The acceleration response spectra of the input
motions at 5% are shown in Figure 8. The motions are typical design motions used for analyses of critical
structures. From the set of six shown in Figure 8, two motions labeled EUS local and distant are the design
motions for sites in Eastern US with locations close and far away from a major fault. The ATC SI motion
is the ATC recommended motion for SI soil conditions. The WUS motion is the design motion for a site
close to a major fault in Western US. The RG1.60 motion is the standard site-independent motion used for
nuclear plant structures. Finally, the Loma Prieta motion is the recorded motion from the Loma Prieta
earthquake scaled to 0.3g maximum acceleration. This motion is used in the analysis as described in later
sections. All motions are scaled to O.SOg and limited to frequency cut-off of 20 Hz for use in the analysis.
The cut-off frequency of 20 Hz reduces the peak ground acceleration of the EUS local motion to less than
0.30g due to high frequency content of this motion as shown in Figure 8.

2-11
The maximum seismic soil pressure values at each depth obtained from the analyses for the various input
motions are shown in Figure 9. The amplitudes of the pressure vary from one motion to the other with
larger values associated with use of RG1.60 motion. Using the pressure profiles in Figure 9, the lateral
force acting on the wall for each input motion was computed. The lateral force represents the total inertia
force of a SDOF for which the system frequency is known. The system frequency for the case under
consideration is the soil column frequency which is 7.5 Hz based on Eqn (1). The total force divided by the
spectral acceleration of the system at 7.5 Hz at the appropriate damping ratio amounts to the mass of the
SDOF. To identify the applicable damping ratio, the acceleration response spectrum of the free-field
response motions at the depth of 50 ft were computed for all six motions shown in Figure 8 for damping
ratios of 5, 10, 20, 30,40, 50, and 60 percents. Knowing the total force of the SDOF, the frequency of the
system, and the input motion to the SDOF system, the relationship in the form proposed by Veletsos and
Younan (1994a) was used to compute the total mass and the damping of the SDOF system. For the total
mass, the relationship is

m = 0.50 x p x H2 x \f/ v (4)

where p is the mass density of the soil, H is the height of the wall, and Iff v is the factor to account for the

Poisson's ratio as defined in Eqn (2). In the analytical model developed by Veletsos and Younan, a
constant coefficient of 0.543 was used in the formulation of the total mass. Study of the soil pressure
transfer functions and free-field response motions at the depth of 50 ft showed that spectral values at the soil
column frequency and at 30% damping have the best correlation with the forces computed directly from the
SSI analysis. In the Veletsos and Younan's model, a damping of 27.5 + P percent has been proposed where
P is the material damping of the soil (%). For the case of 5% soil material damping, the proposed spectral
damping amounts to 32.5%. However, as shown by Ostadan and White (1997), the spectral values of the
various motions considered are insensitive to the spectral damping ratios at the soil column frequency of
7.5. The various motions, however, have significantly different spectral values at the soil column
frequency. This observation leads to the conclusion that while the frequency of the input motion
particularly at the soil column frequency is an important component for magnitude of the seismic soil
pressure, the spectral damping ratio selected is much less important in terms of pressure amplitudes. The
role of soil material damping is discussed by Ostadan and White (1997).

Simplified Method: Computational Steps

To predict the lateral seismic soil pressure for below ground building walls resting on firm foundation and
assuming rigid walls, the following steps should be taken: '

2-12
1. Perform free-field soil column analysis and obtain the ground response motion at the depth
corresponding to the base of the wall in the free-field. The response motion in terms of acceleration
response spectrum at 30% damping should be obtained. The free-field soil column analysis may be
performed using the Computer Program SHAKE (Schnabel et al., 1972) with input motion specified
either at the ground surface or at the depth of the foundation basemat. The choice for location of
control motion is an important decision that needs to be made consistent with the development of the
design motion. The location of input motion may significantly affect the dynamic responses of the
building and the seismic soil pressure amplitudes.

2. Use Eqn (4) to compute the total mass for a representative SDOF system using the Poisson's ratio and
mass density of the soil.

3. Obtain the lateral seismic force from the product of the total mass obtained in Step 2 and the
acceleration spectral value of the free-field response at the soil column frequency obtained at the depth
of the bottom of the wall (Step 1).

4. Obtain the maximum lateral seismic soil pressure at the ground surface level by dividing the lateral
force obtained in Step 3 by the area under the normalized seismic soil pressure, 0.744 H.

5. Obtain the pressure profile by multiplying the peak pressure with the pressure distribution relationship
shown in Eqn (3).

One of the attractive aspects of the simplified method is its ability to consider soil nonlinear effect. The
soil nonlinearity is commonly considered by use of the equivalent method and the strain-dependent soil
properties. Depending on the intensity of the design motion and soil properties, the effect of soil
nonlinearity can be important in changing the soil column frequency and therefore, amplitude of the spectral
response at the soil column frequency.

Accuracy of the Simplified Method

The simplified method outlined above was tested for building walls with heights of 15, 30 and 50 ft using
up to six different time histories as input motion. The results computed directly with SASSI are compared
with the results obtained from the simplified solution. A typical comparison is shown in Figure 10. More
extensive validation of the method is presented by Ostadan and White (1997).

Comparison to Other Commonly Uses Solutions

The seismic soil pressure results obtained for a building wall 30 ft high embedded in a soil layer with shear
wave velocity of 1000 ft/sec using the M-O, Wood and the proposed simplified methods are compared in
2-13
Figure 11. For the simplified method, the input motions defined in Figure 8 were used. The M-O method
results in the smallest pressure values. This is understood since this method relies on the wall movement to
relieve the pressure behind the wall. Wood's solution generally results in the maximum soil pressure and is
independent of the input motion as long as the peak acceleration is 0.3 g. The proposed method results in a
wide range of pressure profiles depending on the frequency content's of the input motion, particularly at the
soil column frequency. For those motions for which the ground response motions at the soil column
frequency are about the same as the peak ground acceleration of the input motion, e.g., RG1.60 motion, the
results of the proposed method are close to Wood's solution. Similar trend in the results is observed if sum
of the lateral forces and the overturning moments from the above three methods are compared (Ostadan and
White, 1997).

The simplified method was extended for application to soil layered system and soil deposits with parabolic
distribution of the shear modulus. The extended method and its verification are discussed by Ostadan and
White (1997).

SIMPLIFIED METHOD TO PREDICT LATERAL SEISMIC SOIL PRESSURE FOR BUILDINGS

IN DEEP SOIL SITES

One of the distinct dynamic characteristics of a building in a deep soil site is its rocking vibration which has
a significant role on distribution of the pressure depending on the embedment ratio (embedment depth
versus plan dimensions), dynamic properties of the soil, and frequency contents of the ground motion under
consideration.

Mita and Luco (1989) have reported the harmonic response of an embedded square foundation subjected to
vertically propagating shear waves. The results adopted from the authors but modified to reflect the same
nomenclature used in this report are shown in Figure 12. The results are for a square foundation with plan
dimensions of 2B x 2B and embedment depth H. The halfspace is characterized by the shear wave velocity
of Vs. The free-field motion has a unit amplitude at the ground surface at each harmonic frequency. The
horizontal translational motion of the foundation (D) at the middle point corresponding to the basemat
motion and the normalized rocking motion represented in terms of HxT are shown in terms of
dimensionless frequency ratio &i- coxH/Vs where T is the angle of rocking rotation and o is the circular
frequency at each harmonic frequency under consideration. The dimensionless frequency is a measure of
the harmonic shear wave length as compared to the embedment depth H. The free-field motion
corresponding to the basemat depth (depth of H) in the free-field shows decreasing amplitude with
increasing frequency. At the soil column frequency of f = Vs/(4xH), the dimensionless frequency al is
1.57 at which the amplitude of the free-field motion is zero. The foundation motion is a function of the
frequency of vibration and the embedment ratio (H/B).

2-14
In order to examine the effect of rocking motion on seismic soil pressure, a series of SSI analyses were
performed using the soil shear wave velocities of 500, 1000, 1500, and 2000 ft/see. In all cases, the wall
height considered was H=50 ft but the foundation width (2B) was changed successively from 50 ft, to 100
ft, 200 ft, and to 400 ft, resulting in embedment ratios of B/H=0.5, 1, 2, and 4. The input motion was
specified at the basemat level in the free-field. A typical result in terms of amplitude of pressure transfer
function is shown in Figures 13. For each soil case, the results from all three elements are clustered
together with the same peak frequency which leads to the conclusion that (1) the soil column frequency
continues to be the most significant frequency for the response in terms of maximum value of the seismic
soil pressure, and (2) the frequency of the peak response is not affected by the embedment ratio. However,
the distribution of the maximum soil pressure in terms of amplitude of the pressure in Elements 2,5, and 10
is significantly affected by the rocking motion of the building and thus the embedment ratio. The effects of
rocking motion on distribution of maximum seismic soil pressure for four different aspect ratios are shown
in Figure 14. As shown, for buildings with narrow width, the rocking motion tends to reduce the amplitude
of the soil pressure at top of the wall.

The results of the parametric studies performed for deep soil sites were also examined in detail. Limitation
of space prohibits detail discussion of the studies performed. The computational steps for deep soil sites
are, however, similar to the rigid case and consist of the following:

. 1. Perform free-field soil column analysis and obtain the response motion in terms of acceleration
response spectrum at 30% damping at the depth corresponding to the basemat elevation in the free-
field.
2. Obtain the soil column frequency using Eqn (1) and obtain the spectral value at the soil column
frequency using the results of Step 1.
3. Use the following relationship to obtain the lateral force acting on the wall:

F = ccxpxH2 xSax4/v (5)

where p is the mass density of the soil, H is height of the wall, Sa is the spectral value of the free-field
response obtained in Step 2, and Vv is the function that considers the effect of soil Poisson's ratio and
can be obtained using Eqn (2). In order to represent the effect of the embedment ratio and reduction
of soil pressure due to rocking motion as well as its increase beyond the rigid base cases for wide
buildings, the parameter ex is defined in the equation above. This parameter was determined from
back-calculation of the lateral force obtained from soil pressure and the shear stress under the basemat
to hold the equilibrium offerees in the horizontal direction. Using the results of the all the parametric
studies, the following values were obtained for a:

2-15
 Embedment Ratio, B/H Parameter a
0.50 0.27
1.0 0.43
2.0 0.62
4.0 0.92

4. Obtain the maximum soil pressure by dividing the lateral force obtained from Step 3 to the area under
the soil pressure curve provided in Eqns(6) through (9) below depending on the embedment ratio. For
an embedment ratio that falls in between the ratios considered, use interpolation.

Embedment ratio of B/H =0.50

p(y) = -2.58y3 + 0.32 y2 +2.46 y - 0.03 (6)
Maximum pressure at the depth y = 0.625
Area under the curve = 0.632H
Point of application for resultant force, Y = 0.55H

Embedment ratio of B/H=1.0

p(y) = 0.60y3 -3.09y2 + 3.34y - 0.025 (7)
Maximum pressure at the depth y = 0.625
Area under the curve = 0.77H
Point of application for resultant force, Y = 0.5 8H

Embedment ratio of B/H=2.0

p(y) = -1.33y4 + 4.38y3 - 5.66y2 + 3.44y + 0.17 (8)
Maximum pressure at top of the wall y = 1
Area under the curve = 0.832H
Point of application for resultant force, Y = 0.57H

Embedment ratio of B/H=4.0

p(y) = -O.OSSy2 + 0.47y + 0.61 (9)
Maximum pressure at top of the wall y = 1
Area under the curve = 0.82H
Point of application for resultant force, Y = 0.54H

5. Multiply the maximum lateral soil pressure from Step 4 by the relationships provided in Eqns (6)
through (9) to get the pressure distribution depending on the embedment ratio of the foundation under

2-16
 consideration. Judgment should be exercised to obtain the distribution for embedment ratios in
between the four embedment ratios considered above.

The simplified method for deep soil sites was also tested extensively for a wide range of soil properties and
foundation embedment ratios (Ostadan and White, 1997).

A comparison of the simplified method with the M-O and Wood's methods for a building with four
different embedment ratios is shown in Figure 15. The results clearly demonstrates the effect of the rocking
motion on distribution of the seismic soil pressure.

SUMMARY AND CONCLUSION

The Mononobe-Okabe (M-O) method was developed in the 1920's. Since then, a great deal of research
work has been performed to evaluate its adequacy and to improve it. The method is, strictly speaking,
applicable to gravity retaining walls which, upon experiencing seismic loading, undergo relatively large
movement to initiate the sliding wedge behind the wall and to relieve the pressure to its active state.
Unfortunately, the method has been and continues to be used extensively for embedded walls of the
buildings as well. Recent field observations and experimental data, along with enhancements in analytical
techniques have shown that hardly any of the assumptions used in the development of the M-O method are
applicable to building walls. The data and the follow up detail analysis have clearly shown that the seismic
soil-pressure is an outcome of the interaction between the soil and the building during the seismic excitation
and as such is function of all parameters that affect soil-structure interaction (SSI) response. Some of the
more recent observations and experimental data were presented in the paper. The new understanding of the
attributes of seismic soil pressure prompted the United Sates Nuclear Regulatory Committee (NRC) to
reject the M-O and the M-O based methods for application to critical structures. At this time, while
elaborate finite element techniques are available to obtain the soil pressure for design, no simple method has
been proposed for quick prediction of the maximum soil pressure, thus hindering the designer's ability to
use an appropriate method in practice. To remedy this problem, the current research was conducted to
develop a simple method which incorporates the main parameters affecting the seismic soil pressure for
buildings.

Using the concept of the single degree-of-freedom, a simplified method was developed to predict maximum
seismic soil pressures for buildings resting on firm foundation materials. The method incorporates the
dynamic soil properties and the frequency content characteristics of the design motion in its formulation. It
was found that the controlling frequency that determines the maximum soil pressure is that corresponding to
the soil column adjacent to the embedded wall of the building. The proposed method requires the use of
conventionally-used simple one-dimensional soil column analysis to obtain the relevant soil response at the
base of the wall. More importantly, this approach allows soil nonlinear effects to be considered in the
2-17
process. The effect of soil nonlinearity can be important for some applications depending on the intensity
of the design motion and the soil properties. Following one-dimensional soil column analysis, the proposed
method involves a number of simple hand calculations in order to arrive at the distribution of the seismic
soil pressure for design. The accuracy of the method relative to the more elaborate finite element analysis
was verified for a wide range of soil properties, earthquake motions, and wall heights.

The method was extended to include buildings on deep soil sites . The complexity of the seismic soil
pressure for such cases is compounded by the rocking motion of the structure. The rocking motion is in
turn, a function of soil properties, frequency content of the design motion, and embedment ratio of the
structure. A wide range of parametric studies were performed that cover many practical cases. The steps
for the analysis are similar to the steps outlined for buildings on rock except that an appropriate pressure
distribution curve should be selected to observe the effect of the embedment ratio. Similarly, the accuracy
of the proposed method was verified against a more detailed SSI analysis.

REFERENCES
ASCE 4.86 (1986). "Seismic Analysis of Safety-Related Nuclear Structures and Commentary on Standard
for Seismic Analysis of Safety Related Nuclear Structures," Published by the American Society of Civil
Engineers.

Chang, C. Y., Power, M. S., Tang, Y. K., and Tang, H. T. (1990). "Analysis of Dynamic Lateral Earth
Pressures Recorded on Lotung Reactor Containment Model Structure," Proceedings of 4th US National
Conference on Earthquake Engineering, May 20-24, Palm Springs, CA.

Ebeling, R. M., and Morrison, Jr., E. E. (1992). "The Seismic Design of Waterfront Retaining Structures,"
US Army Technical Report ITL-92-11, Prepared by Information Technology Laboratory, Department of the
Army, Waterways Experiment Station, Corps of Engineers, Vicksburg, Mississippi.

EPRI (1989). "Proceedings: EPRI/NRC/TPC Workshop on Seismic Soil-Structure Interaction Anslysis

Techniques Using Data From Lotung, Taiwan," Electric Power Research Institute Publication No. NP-
6154, 2 Volumes, March.

EPRI (1991). "Post-Earthquake Analysis and Data Correlation for the 1/4-Scale Containment Model of the
Lotung Experiment," Electric Power Research Institute Publication No. NP-7305SL, October.

Hirota, M., Sugimoto, M., and Onimaru, S. (1992). "Study on Dynamic Earth Pressure Through
Observation," Proceedings of 10th World Conference on Earthquake Engineering, July, Madrid, Spain.

2-18
Ishibashi, I., Matsu,zawa, H., and Kawamura, M. (1985). "Generalized Apparent Seismic Coefficient for
Dynamic Lateral Earth Pressure Determination," Proceedings of 2nd International Conference on Soil
Dynamics and Earthquake Engineering, the Queen Elizabeth II, New York to Southampton, June/July.

Itoh, T., and Nogami, T. (1990). "Effects of Surrounding Soils on Seismic Response of Building
Basements," Proceedings of 4th US National Conference on Earthquake Engineering, May 20-24, Palm
Springs, CA.

Kazama, M., and Inatomi, T. (1988). "A Study on Seismic Stability of Large Embedded Rigid Structures,"
Proceedings of Ninth World Conference on Earthquake Engineering, August 2-9, Tokyo-Kyoto, Japan.

Koyama, K., Watanabe, O., and Kusano, N. (1988). "Seismic Behavior of In-Ground LNG Storage Tanks
During Semi-Long Period Ground Motion," Proceedings of 9th World Conference on Earthquake
Engineering, August 2-9, Tokyo-Kyoto, Japan.

Koyama, K., Kusano, N., Ueno, H., and Kondoh, T. (1992). "Dynamic Earth Pressure Acting on LNG in-
Ground Storage Tank During Earthquakes," Proceedings of 10th World Conference on Earthquake
Engineering, July, Madrid, Spain.

Matsuzawa, H., Ishibashi, I., and Kawamura, M. (1984). "Dynamic Soil and Water Pressures of Submerged
Soils," ASCE, Journal of GeotechnicalEngineering, Vol. Ill,No. 10, September.

Matsumoto, H., Arizumi, K., Yamanoucho, K., Kuniyoshi, H., Chiba, O., and Watakabe, M. (1991).
"Earthquake Observation of Deeply Embedded Building Structure," Proceedings of 6th Canadian
Conference on Earthquake Engineering, June, Toronto, Canada.

Mita, A. and Luco, J. E. (1989). "Impedance Functions and Input Motions For Embedded Square
Foundations," ASCE, Journal of Geotechnical Engineering, Vol. 115, No. 4, April.

Mononobe, N. and Matuo, H. (1929). "On the Determination of Earth Pressures During Earthquakes,"
Proceeding of World Engineering Congress, Tokyo, Vol. 9, Paper 388.

Nadim, F. and Whitman, R.V. (1984). "Coupled Sliding and Tilting of Gravity Retaining Walls During
Earthquakes," Proceedings of 8th World Conference on Earthquake Engineering, July, San Francisco, CA.

Nazarian, H., and Hadjian, A. H. (1979). "Earthquake Induced Lateral Soil Pressure on Structures," ASCE,
Journal of Geotechnical Engineering, Vol. 105, No. GT9, September.
2-19
Nuclear Regulatory Commission (NRC, 1991), Proposed Staff Position on Dynamic Lateral Earth Pressures
on Earth Retaining Walls and Embedded Walls of Nuclear Power Plant Structures, Civil Engineering and
Geoscience Branch, Division of Engineering, Office of Nuclear Reactor Regulation.

Okabe, S. (1924). "General Theory of Earth Pressures and Seismic Stability of Retaining Wall and Dam,"
Journal of Japan Society of Civil Engineers, Vol. 12, No. 1.

Lysmer et al. (1981). "SASSI - A System for Analysis of Soil-Structure Interaction," Report No.
UCB/GT/81-02, University of California, Department of Civil Engineering, Berkeley, California.

Ostadan, F., White, W. H. (1997). "Lateral Seismic Soil Pressure-An Updated Approach," Bechtel
Technical Grant Report, Bechtel Corporation, San Francisco, California.

Richards Jr., R. and Elms, D. (1979). "Seismic Behavior of Gravity Retaining Walls," ASCE, Journal of
Geotechnical Engineering, Vol. 105, No. GT4, April.

Schnabel, P. B., Lysmer, J., Seed, H. B., "SHAKE - A Computer Program for Earthquake Response
Analysis of Horizontally Layered Sites," Earthquake Engineering Research Center, University of California,
Berkeley, Report No. EERC 72-12, December 1972.

Seed, H. B., and Whitman, R. V. (1970). "Design of Earth Retaining Structures for Seismic Loads," ASCE
Specialty Conference on Lateral Stresses in the Ground and Design of Earth Retaining Structures, June.

Veletsos, A, and Younan, A. H. (1994a). "Dynamic Soil Pressure on Rigid Vertical Walls", Earthquake
Engineering and Soil Dynamics, Vol. 23, 275-301.

Veletsos, A, and Younan, A. H. (1994b). "Dynamic Modeling and Response of Soil-Wall Systems," ASCE,
Journal of Geotechnical Engineering, Vol. 120, No. 12, December.

Veletsos, A, and Younan, A. H. (1994c). "Dynamic Soil Pressures on Rigid Cylindrical Vaults,"
Earthquake Engineering and Structural Dynamics, Vol. 23, 645-669

Veletsos, A, and Younan, A. H. (1995). "Dynamic Modeling and Response of Rigid Embedded Cylinders,"
ASCE, Engineering Mechanics, October.

2-20
Veletsos, A, Parikh, V. H., Younan, A. H., and K. Bandyopadhyay (1995). ''Dynamic Response of a Pair of
Walls Retaining a Viscoelastic Solid," Brookhaven National Laboratory, AUI, Department of Advanced
Technology, Associated Universities, Inc., UPTON, New York.

Watakabe, M., Matsumoto, H., Fukahori, Y., and Shikama, Y., Yamanouchi, K., and Kuniyoshi, H. (1992).
"Earthquake Observation of Deeply Embedded Building Structure," Proceedings of 10th World Conference
on Earthquake Engineering, July, Madrid, Spain.
Whitman, R. V. (1979). "Dynamic Behavior of Soils and Its Application to Civil Engineering Projects,"
Proceedings of 6th Pan American Conference, Lima, Peru.

Whitman, R. V., (1990). "Seismic Design and Behavior of Gravity Retaining Walls," Proceedings of a
Specialty Conference on Design and Performance of Earth Retaining Structures, ASCE, Cornell University,
June 18-21.

Whitman, R. V., and Christian, J. T. (1990). "Seismic Response of Retaining Structures," Pola Workshop
Meeting, San Pedro, CA.

Whitman, R. V. (1991). "Seismic Design of Earth Retaining Structures," Proceedings of 2nd International
Conference on Recent Advances in Geotechnical Earthquake Engineering and Soil Dynamics, March 11-15,
St. Louis, Missouri.

Wood, J. H. (1973). "Earthquake Induced Soil Pressures on Structures," Doctoral Dissertation, EERL 73-
05, California Institute of Technology, Pasadena, CA.

2-21
 Figure 1
Original Soil Pressure Record
100

50 .__. Mov ing Window Correction

P4N11 .14 -Original Record

^^1 bH--\rj '

zCi^ All "

-50 - -

-100
10 15 20 25 30 35 40
Time (second)

Corrected Soil Pressure Record

100

50 ... _. _.___ P4N1.14 - Corrected Record

-50 ------

-100
10 15 20 25 30 35 40
Time (second)

Lotungpl.xls, N114plot
7/28/98
 Figure 2
Comparison of Response Spectra (5% Damping)
Lotung Records, Event LSST07. N-S Component
Rocking Motion and Soil Pressure
160 160

140 F4N Horizontal Motion @ El. -1.14 m

140

o P4N.1.14 Soil Pressure Record

N)
I
N)

100
Frequency (Hz)

P4 Mrxls. 4N Rck EP
7/27/98
 Figure 3
Soil Pressure Response Spectra (5% Damping)
Comparison of Recorded and Computed Soil Pressure Records
EPRI Lotung 1/4 Model, Event LSST07, Station P4N1.14, N-S Direction

P4N1.14-Computed

P4N 1.14 -Recorded

0.1 10 100
Frequency (Hz)

P4_ars.xls, ARS_P4N114
7/27/98
 Figure 4
SASSI MODEL SOIL ELEMENTS

Element 2

Soil:
vs, p, p, v

NJ Element 10
I
NJ
Ul

X
o

Y
m
Element 15

RIGID BASE
ZB

Fig3l.xls, R_Base
7/27/98
 Figure 5
Amplitude of Soil Pressure Transfer Functions
Rigid Halfspace. Rigid Wall. H=50 ft, v=1/3, (3=0.05
Element 2 -- D/H =0.075. Input Motion at the Base of the Wall
40000
Vs = 500 ft/s

--- Vs = 1000 ft/s

Vs = 1500 ft/s

- - Vs = 2000 ft/s

a\

8 10 12 14 16 18 20
Frequency (Hz)

Tfbase.xls,ELEM2_T
7/27/98
 Figure 6
Amplitude of Soil Pressure Transfer Functions
Rigid Halfspace, Rigid Wall, H = 50 ft, v = 1/3, p = 0.05
Element 2 -- D/H = 0.075. Input Motion at the Base of the Wall

Vs = 500 ft/s
- Vs = 1000 ft/s
Vs = 1500 ft/s
Vs = 2000 ft/s

0.25 0.5 0.75 1 1.25 1.5

Normalized Frequency

Tfbase.xls,ELEM2_T (2)
7/27/98 7/27/98
 Figure 7
Amplitude of Soil Pressure Transfer Function
Rigid Halfspace, Rigid Wall. H=50 ft, v=1/3, (3=0.05
Average Amplitude in Freq. Range 0.10 - 0.25 Hz
1g Harmonic Input Motion

2000 4000 6000 8000 10000 12000

ADJUSTED
PRESSURE
NORMALIZED for
POISSON'S FACTOR

2-28

Tfsurf.xls,LOWFRQ_PRS_NORM (2)
7/27/98
 ARS_6EQ_5T

Figure 8
Acceleration Response Spectra (5%)
Comparison of Different Input Earthquake Motions: Free-Field
At Ground Surface, Cutoff Freq. = 20 Hz.

EUS- LOCAL
EUS DISTANT
ATCS1
WUS
RG1.60
LOMA PRIETA

0.1 10 100
Frequency (Hz)

H30_aeqs.xls 7/27/98
 Figure 9
Maximum Seismic Soil Pressure - psf
WALL 50 ft, Vs=1500 ft/sec, v=l/3, (3=5%

500 1000 1500 2000 2500 3000

2-30

Eqs.xls,EPS_sum
7/27/98
 Figure 10
Maximum Seismic Soil Pressure (psf)
WALL 30 ft, Vs=1000 ft/sec, v=l/3, (3=5%
LOMA PRIETA Motion
300 600 900 1200 1500

g
UJ

2-31

Ht_aeqs.xls,H30_LMP
7/27/98
 Figure 11
Maximum Seismic Soil Pressure (psf)
WALL 15 ft, Vs=500 ft/sec, v=1/3, (3=5%
All Motions
200 600 800 1000
15.00

12.50

BUS LOCAL

EUS DISTANT

WUS

ATC

LOMA PRIETA

RG1.60

-B M-Ofor0.3g

Wood Soln for 0.3g

0.00

2-32

Ht_aeqs.xIs.H15_ALL (2)
7/27/98
 Figure 12
Normalized Amplitude Response Ratio of Horizontal D and Rocking H*T Motions
For Square Embedded Foundations

-4 Free-Field
--DforH/B=0.50
-D-H*TforH/B=0.50
-*-DforH/B=1.0
-*-H*TforH/B=1.0
- -DforH/B=1.5
H*TforH/B=1.5

2.0 2.5 3.0

Frequency Ratio a1 = to* H / Vs

Luco.xls.luco pit
7/27/98
 Figure 13
Amplitude of Soil Pressure Transfer Functions
B/H=0.50, H=50 ft, v=1/3, (3=0.05
Input Motion at the Base of the Wall
40000

ELEM 2, Vs=500 ft/sec

ELEM 10, Vs=500 ft/sec
ELEM 15, Vs=500 ft/sec
ELEM 2, Vs=1000 ft/sec
ELEM 10, Vs=1000 ft/sec
ELEM 15, Vs=1000 ft/sec
ELEM 2, Vs=100 ft/sec
ELEM 10, Vs=1500 ft/sec
ELEM 15, Vs=1500 ft/sec
- ELEM 2, Vs=2000 ft/sec
ELEM 10, Vs=2000 ft/sec
- - - - ELEM 15, Vs=2000 ft/sec

10
Frequency(Hz)

Tfst.xls,B5HAEL_PL
7/27/98
 Figure 14
Maximum Seismic Soil Pressure
H=50 ft, Vs=500 ft/sec, v=1/3, p=5%
RG1.60 Motion at the Base
2000 4000 6000 8000 10000 12000

2-35

Stsum_rg.xls,v1_pl
7/27/98
 Figure 15
Maximum Seismic Soil Pressure
H=50 FT, Vs=1000 ft/sec, v=1/3, p=5%
RG1.60 Motion

500 1000 1500 2000 2500 3000 3500 4000

2-36

Rx.xIs.VS1000_PLT
7/27/98 7/27/98
 DESIGN OF PILES CONSIDERING THE DEFORMATION RESPONSE
UNDER THE ACTION OF EARTHQUAKE SHAKING

By Madan B. Karkee1, Yoshihiro Sugimura2 and Kaoru Fujiwara3

ABSTRACT: There have been several instances of damage to piles at deeper part, generally near the
soil layer interfaces, during the past earthquakes. Such damages are inherently difficult to detect
and repair, mandating adequate provision in the design to make them as unlikely as possible.
Nonlinear response analysis of soil-pile-structure system considering a two dimensional FEM
model shows distinctly large ground response forces near soil layer interfaces, demonstrating the
nature of stresses that may develop in piles due to distinct stiffness contrast between soil layers.
While such detailed analysis is rather impractical for the general design application, current
practice of designing the pile for a single concentrated load representing the inertia effect of the
superstructure involves implicit disregard of the actions on piles attributable to the ground
deformation response. A simple approach to account for the ground deformation response actions
on piles is proposed and the potential for its use in practical design application is discussed.

Results of the nonlinear response analysis on the soil-

INTRODUCTION
pile-building interaction system of a 35 storied reinforced
concrete building based on the two-dimensional finite
Investigations on the damage to piles during the past
earthquakes provide some basic information concerning element model (Sugimura et al. 1997) is discussed.
the nature of failures in piles at locations with deep soil Considering three simple variations in soil condition,
other structural details remaining the same, it is shown
deposit under strong ground shaking. Examples include
the January 17, 1995 Hyogoken-Nambu earthquake that the influence of soil layering on the stresses
(Karkee & Kishida 1997, Karkee et al. 1997, Matsui & developed in piles during earthquake shaking can be very
dominant. Of particular interest is the clear dependence of
Oda 1996 etc.) and the June 12, 1978 Miyagiken-oki
the building superstructure inertia itself on the nature of
earthquake (Sugimura 1981, 1987). Remarkably
the soil layering system interacting with the pile. That is,
significant instance of the damage is reported to have
for a given incident motion specified for a general region,
occurred at deeper parts along the pile, particularly in
the superstructure inertia itself depends on the different
relatively longer piles. Evidently, the location of pile
damage at the intermediate part in longer piles also tends levels of excitation based on the extent of the nonlinear
to coincide with changes in soil layering, giving rise to response (Karkee et al. 1992) depending on the local site
condition. In addition, the ground deformation response
stiffness contrast interfaces.
for the same input earthquake motion is very dependent
The stresses developed in piles due to the soil-pile- on the local site condition. Results of the response
structure interaction under earthquake shaking consist of analysis show the need for adequate consideration of the
the superstructure inertia effects as well as the kinematic inertia as well as the ground deformation response effects.
effects of ground response. The latter effects are termed
While the detailed finite element analysis is known to
simply as 'ground response effects' in this paper. Relative
adequately depict the response of the soil-pile-structure
magnitude of the inertial and the ground response actions
system under earthquake excitation, computational effort
depends on the ground condition as well as the level of
can be formidable for its application to everyday design
excitation. Generally, long piles penetrating a deep
practice. Considering the current practice of utilizing a
layered deposit, particularly where there is a sudden
single concentrated load at the top of the pile to represent
change in soil stiffness, are likely to be exerted by large
ground response forces. However, only the inertia effects the inertial actions, there is a need to develop a simple
tend to be explicitly accounted for in the seismic design design method that can account for the ground response
effects realistically. Sugimura (1992) proposed the use of
practice for piles. The horizontal force to be resisted by
distributed load to represent the effect of ground response
the pile consists of the inertia of the building, and the
on piles. Presumably, the nature and the magnitude of the
basement if applicable (BCJ 1984), with no recognition
distributed load should reflect the local soil condition and
of the ground response effects explicitly.

1 Manager, Research Division, GEOTOP Corporation, 1-16-3 Shinkawa, 4F, Chuo-ku, Tokyo, 104-0033, Japan.
2 Professor, Department of Architecture & Building Science, Grad. School of Engineering, Tohoku University, Sendai, Japan
3 Sr. Managing Director, Suzuki Architectural Design Office, Yamagata, Japan

Karkee, Sugimura & Fujiwara/ 1

3-1
its dynamic characteristics. A simple approach for the highway bridge (HBC, 1995). Again the pile has cracked
consideration of these aspects in the estimation of the at the middle part where the soil stiffness decreases
distributed load is proposed for evaluation of the ground suddenly as indicated by the N-value distribution. The
deformation response effects to be considered in general cracking of the piles at deeper part noted in Fig. 1 were
design practice. Preliminary simulations clearly illustrate detected by core boring and borehole camera. Fig. 2
the potential of the approach to realistically represent the shows the damage to pile during the Miyagiken-oki
ground response action on piles that may be expected earthquake (Sugimura & Oh-oka, 1981) in a L-shaped
during earthquakes. building where the outer comer joining the two wings had
undergone a large settlement. This is a case of a pile on
the inner side of one the wings. The pile foundation
DAMAGE TO PILES IN PAST EARTHQUAKES
damage was considered to be due to inertial forces of the
superstructure. However, the cracks seem to align with
Reports on investigation of damage to foundations
changes in soil layering.
during past earthquakes provide ample instances of
damage at the intermediate part of a pile. The location of Clearly, the damage to piles at deeper part is more
the damage in piles may provide some indication of the
problematic in the event of rehabilitation and recovery
dominance of either the inertia effects of the
after the earthquake. While there are successful cases of
superstructure or of the kinetic effects of ground
repair of the damage to pile near the top (e.g. Karkee &
response. Generally the inertial forces may be considered
Kishida 1997), damages at deeper part of the pile are by
to result in the failure of piles near the top, while the
nature much more difficult to detect and repair. It is
ground response effects may be expected to result in
imperative that the design approach for piles at
damages at the deeper part, particularly where there is a
earthquake regions should particularly strive to make the
abrupt change in the soil stiffness.
damage to piles at the deeper part less likely.
nFPTiir.i
DEfTH(i) * 40
mUE DEPTH.) B 2V ,UUE4 ,
N VALUE
DEPTHd 20 40 60
B
FOOT IIG
FOOTING

10

-(6)
X -(7)
-it) -0.3
-(9)
ID-
-0.7 "
20- 20-

CRACK IIDTH:
4>IIOO
if
(3):Gn.(4):4.(G):|ii.(7)::
(0):3.OTHERS:HAIR CRACKS H

3fl~
(a)
20 #!!!
4>I50
(b)
FIG. 2. Damage to pile in a building structure during the
FIG. 1. Damage to piles during the Hyogoken-Nambu
1978 Miyagiken-oki earthquake (Sugimura & Oh-oka 1980)
earthquake: (a) Building (AIJ 1996) (b) Highway (HBC 1995)
Extremely large actions on piles may manifest due to
A typical case of damage to a pile supporting a
failure of the ground (e.g. liquefaction) accompanied by
building during the Hyogoken-Nambu earthquake (AIJ,
lateral spreading (e.g. Tokimatsu et al. 1996) and detailed
1996) is shown in relation to the ground profile in Fig.
investigation of the local site for such possibility should
l(a). The pile is seen to develop cracks near the top as
be included while considering ground response effects in
well as at the middle part around where the soil profile
the seismic design of piles. Even when the complete
changes into a stiffer layer. It seems plausible that the
ground failure does not occur, the excessive nonlinear
change from a softer to a stiffer soil layer may have
contributed to the cracking at the middle part of the pile. response of the soft soil layer can result in large
deformation response in the piles. Thus the deformation
Similarly, Fig. l(b) shows the damage to a pile in a

Karkee, Sugimura & Fujiwara/ 2

3-2
 response of piles during strong shaking can include the is the well known El Centre NS record and the other is the
ground shaking effects as well as the ground failure Hyogoken-Nambu earthquake record of the Kobe marine
effects. observatory (Kobe JMA NS). The El Centre motion is
scaled to a maximum velocity of 50cm/s while the Kobe
record is used as it is. The corresponding peak
FFINITE ELEMENT ANALYSIS FOR NONLINEAR
accelerations are 510.8cm/s2 and 818.0cm/s2 respectively.
SEISMIC RESPONSE OF THE SYSTEM
The response spectra of the two input motions are given
in Fig. 5, where it is seen that the spectral velocity
In Japan, 20 to 45 storied reinforced concrete ordinates for the Kobe record are significantly larger than
buildings have been common for apartment building those for El Centre in the period range 0.3-3.0s.
structures. The natural period of these buildings range
from 1.2s to 2.5s, falling into a range in the design spectra
where the spectral velocity ordinates tend to be uniform. The results of the response analysis
Generally these buildings are supported on cast-in-place Some of the time histories of the response forces for
concrete piles with enlarged base. A 35 storied building is El Centra input are shown in Fig. 6. From the time history
considered a representative of these structures and is of the building inertia, the predominant period of the
analyzed in detail for its seismic response characteristics. building can be noted to be about 2.9s. The corresponding
Basement & wall details
period of the building was about 3.3s for Kobe JMA NS.

n ii
O >"
>"
Superstructure
( Condensed Model) f^
~ 20
o
Rigid Beam' D.
Slit Foundation Beam 2
Pile u ra **"
Soil
' Wall G 00 >
( Finite Element) / /Pile 30
Surface of Soil v /
Free-
BASE
Field Sandy
End of Pile -30m v gravel
Viscous
Boundary
p=20.6kN/m
Viscous Boundary

a- soil b- soil c- soil

(35 stories of the building are condensed into 7 lumped masses) FIG. 4. Details of three soil conditions used in the analysis,
the building & foundation structure remaining the same.
FIG. 3. 2-D finite elements in combination with the lumped
mass model representing the soil-pile-building system.

The schematic finite element model for nonlinear

dynamic response analysis is given in Fig. 3. The
basement slab in Fig. 3 is assumed to have no direct
contact with the soil underneath. Three simple soil
profiles shown in Fig. 4 and designated as a-soil, b-soil
and c-soil overlying a sandy gravel layer, typically found
in urban areas in Japan at depths of 30m or more, is
considered. Three sets of response analysis are carried out
for the three soil conditions, other structural details
remaining the same. Details of the building structure and 0.1 1.0 10.0
the method of analysis, together with the assumed Period(sec)
nonlinear behavior of soil and concrete are as given by
Sugimura et al. (1997). FIG. 5. Response sectra of El Centra & Kobe motions

Base input motion for the analysis Compared to the elastic period of 2.1s, the period of
Two input motions are considered to investigate the the building has elongated by a factor of 1.4 and 1.6
effect of relative difference in the level of excitation. One 'respectively, larger factor indicating stronger shaking in
case of the Kobe input. Fig. 6 also shows that the

Karkee, Sugimura & Fujiwara/ 3

3-3
basement inertia is larger in stiffer soil (a-soil) compared The maximum shear force in piles is largest in case of
to soft soils (b and c-soils). The time histories for pile top the c-soil for both the input motions and occurs at the
shear force contain long as well short period components, interface of the two soil layers. The cracking of piles at
unlike the building inertia, where the component the layer interfaces observed during past earthquakes, as
attributable to the predominant period of the building noted in Fig. 1, may have been caused by this tendency of
dominates. large shear force around the soil stiffness contrast
interfaces.
(MN)
Fig. 8 shows the maximum bending moment in piles.
Again a very strong influence of the soil layering is
apparent from the distinctly different shape of the
maximum bending moment distribution in case of the c-
15.0 soil. Larger moment at the soil layer interface is seen in
c-soil for both the El Centre and the Kobe input. Larger
bending moment at deeper part of the pile is particularly
problematic because it tends to act in combination with
the decreased axial load with depth generally observed in
piles. In addition to larger moment at the layer interface, it
15.0
is seen in Fig. 8 that the bending moment at the pile head
is also largest in case of the c-soil, for both the input
motions. That is, the pile head moment as well as the
moment in the intermediate part can be expected to give
the worst condition when soil stiffness contrast exists
JSJ-
15.0
over the pile length.
c-soil
It may be noted that the maximum shear force, as well
as the maximum bending moment, tends to large in the
FIG. 6. Time histories of some typical response forces intermediate part of the pile when large soil stiffness
contrast exists over the pile length.

Yield Moment

-30
0.0 50.0 100.0 50.0 100.0 150.0 ~5JO 10.0 J15.0 115.0
Shear Force (MN) Shear Force (MN) Bending Moment (MNm) Bending Moment (MNm)

FIG. 7. Maximum shear force response of the building and FIG. 8. Maximum bending moment distribution in piles
the piles for the El Centra & the Kobe input motions. under the action of the El Centra & the Kobe input motions

Fig. 7 shows the maximum shear force response of the Another interesting result is that the bending moment
structure corresponding to the two incident motions. In distribution in a-soil is nearly half of that in b-soil in case
the current design practice for the tall buildings in Japan, of El Centre input, while the bending moment distribution
the piles are generally designed to have the ultimate shear for a-soil and b-soil is practically same in case of the
capacity. The ultimate shear capacity is defined as the Kobe input. This may be attributed to the much higher
sum of 1.5 times the base shear capacity of the building level of shaking in case of the Kobe input, with the
(64.0MN) and the seismic coefficient for the basement increased nonlinear effect resulting in similar stiffness in
times the weight of the basement (98.4MN). Assuming a-soil and b-soil at higher level of excitation (Karkee et al.
the seismic coefficient for the basement to be 0.2, the 1993, 1992). This may be considered a possibility
ultimate shear capacity works out to be 115.8MN. This is because the soil stiffness degradation tends to saturate at
designated as the design force in Fig. 7.

Karkee, Sugimura & Fujiwara/ 4

3-4
 strain level attributed to the higher level of that of the building system. However, the contribution of
tation. both the building system and the foundation system to the
pile top shear force clearly depends on the soil condition.
nponents of the building and the foundation
-systems in the response of the total system Fig. 11 shows distribution of the maximum shear force
in a single pile. While the magnitude and the nature of
\ttempt is made to separate (Sugimura et al. 1997) distribution of maximum shear force is strongly
esponse of the total system into those of the building dependent on the ground condition, the contribution of
the foundation sub-systems. This is done to evaluate the superstructure inertia to the shear force in piles is
relative contribution of the two sub-systems to the distinctly smaller than that of the foundation system.
onse of the system as a whole. The contribution of the
ding superstructure and that of the soil and foundation Thus the shear force obtained based solely on the
le response of the total system depicted in Fig. 3 is inertia without regard to the local site condition tend to
obtained. Figs. 9 and 10 show time histories of the grossly underestimate the maximum shear forces that
onse actions contributed by the building and the may be expected during strong ground shaking. In fact,
idation sub-systems respectively. the maximum shear force responses of the total system
and that of the foundation system are seen to be
[t may be noted in Fig. 9 that the building inertia and practically coincident in all the three soil types in Fig. 11,
rile top shear force tend to be in opposite phase, while particularly at deeper part, indicating domination of the
. are dominated by the predominant period of the foundation part. The result shows that the shear forces in
ding as noted above. Figs. 9 and 10 show that the piles may even be represented by the response of the
ribution of the foundation system consists of foundation portion alone.
ively short period components when compared to
(MN) (MN)

15.0 15.0

o.o 5.0 Time(sec) 10.0 15.0

15.0
a-soil b-soil -c-soil
c-soil

9. Response forces due to building sub-system FIG. 10. Response forces due to foundation sub-system

1 Total System I Building System 1

' ^\
' fT^
_^~.+

X- -10
,,
~^~
1.
Q -20 J o"' ~ - -20
%
*L
^
V?7^ x
a
a-soil
b-soil *

*
V
}
t, k
.
o
i-soQ
b-soil
c-sotl
-30
00 1.0 2.0 3.0
3. " 0.0 1.0 2.0 3.0 1.0 2.0
Shear Force (MN) Shear Force (MN) Shear Force (MN)

FIG.11. Maximum shear force in a pile due to the total system and the building and foundation sub-systems

Karkee, Sugimura & Fujiwaraf 5

3-5
 Building System

-10
.' d

-20

-3C
5.0 10.0 15.0 0.0 5.0 10.0 15.0 0.0 5.0 10.0 15.0
Bending Moment (MNm) Bending Moment (MNm) Bending Moment (MNm)

FIG. 12. Maximum bending moment in a pile due to the total system and the building and foundation sub-systems

Similar to the case of maximum shear force, Fig. 12

shows distribution of maximum bending moment in a Principle of beams on elastic foundations
pile. Again, the contribution of the foundation system to Considering a Winkler soil model and assuming the
the maximum bending moment distribution is distinctly pile to be a massless beam on elastic foundation, the basic
large and practically same as that of the total system. This equation relating the horizontal deflection y of the pile
follows logically from Fig. 11, because integration of may be given by Equation 1. Here, D is the diameter, E is
shear force gives the bending moment. However, the Young's modulus and / is the sectional area moment
maximum bending moment does not necessarily of inertia of the pile. The constant k,, is the coefficient of
correspond to the maximum shear force, and it is horizontal subgrade reaction of the soil and* is the depth
important to note strong domination of foundation system in soil. The soil movement f(x) in Equation 1 represents
in the bending moment acting on the piles. the free field ground response displacement during the
earthquake and equating it to zero gives the equation for
The results of nonlinear response analysis of the the static loading case.
soil-pile-building system depicted in Fig. 3 clearly
illustrate the inadequacy of the current practice of
considering the inertia of superstructure and basement (1)
(BCJ, 1984) with implicit disregard of the soil condition
in the seismic design of piles. Considering that finite lfk,f>f(x) in Equation 1 is considered to be a forcep(x)
element analysis discussed above is impractical for at a depth x required to displace the soil there byf(x),
everyday design, the need to develop a suitable method to then we have:
adequately account for the ground response effects is
evident.
(2)
SIMPLE DESIGN APPROACH CONSIDERING
GROUND RESPONSE Equation 2 can be solved for a given distribution of
load p(x) along the pile length. From a practical point of
As mentioned above, seismic design practice in Japan view, solutions for three simple load distribution,
allows for inertial forces to be resisted by the pile. consisting of concentrated load, uniformly distributed
However, the reports on damage to piles during past load and triangularly distributed load, can be
earthquakes, as well as the results of finite element appropriately combined to depict a more general
response analysis, show strong domination of ground distributed load approximately. These solutions are in fact
response effects on the internal forces developed in piles. already available from Hetenyi (1946). This framework
A simple method based on the principle of beam on for the solution of a beam on elastic foundation is utilized
elastic foundation, together with an approach to evaluate in the proposed simple analysis method to indirectly
the distributed load, is proposed for adequate account for the kinematic effects of ground response.
consideration of the ground response effects in the
seismic design of piles. Preliminary results indicate that The current design method (MKS units)
the proposed approach has the potential for evaluation of
the ground deformation response effects on piles. Based on the Japanese design guide (Sugimura,
-1988), the design external force Qp for earthquake loading

Karkee, Sugimura & Fujiwara/ 6

3-6
 is given in terms of the base shear Q and the number of may be given in terms of the peak velocity of the input
piles n as: motion Vmax by Equation 7.

= (!-) ; a = 1-0.2- (3) (7)

Here, a is the participation factor of embedment given The parameter <p in Equation 7 is the 'ground period
in terms of the height above ground HB and the depth of elongation factor' representing the extent of elongation in
embedment Hp both in meters. Based on the engineering the ground period due to the nonlinear effect (Karkee et
judgment derived from parametric analysis, it is al. 1992, 1993) during strong ground shaking, such that <f>
recommended that the value of a should not be greater ^1. If all the soil layers at a site are assumed to contribute
than 0.7. The design guide suggests that the internal equally to the ground period elongation, a constant <p may
forces and displacement can be computed based on the be assumed for all the soil layers. The introduction of the
theory of beam on elastic foundation. For a given degree parameter 0 constitutes an attempt to account for the
of fixity ar at the pile head and a horizontal concentrated effect of nonlinear soil response in a simple manner. From
load Qp acting at the top, horizontal displacement at pile the dynamic analysis with the El Centre input (Sugimura
head y0 is obtained from Equation 4. When not et al. 1997) the, predominant ground period TP was seen
determined from horizontal loading test, it is suggested to be about twice the elastic fundamental ground period
that k,t in kg/cm3 be estimated from the Equation 6, where TG. This corresponds to a value of 0 equal to about 2.0. In
E0 is the elastic modulus of soil in kg/cm2 and D is in cm. comparison, the Japanese guideline (BCJ, 1992)
recommends a value of 2.2 for strong shaking at bay areas
with deep soil deposit. It seems reasonable to assume the
(4) overall ground period elongation factor $ at a deep soil
4EI/53
site to be in the range of 2.0 to 3.0.

For a given value of the ground period elongation

(5) factor <p at a site, consideration of a variable value 0, for
different soil layers may be considered appropriate
depending on the site condition. One of the way this could
)~4 (6) be done is to assume the value of fa for the i"' soil layer of
thickness H and shear wave velocity Vs to be in
As can be noted, the design method considers only the proportion to the value of H/VS such that overall ground
inertia effect of the superstructure and needs to be period elongation factor <j> remains the same. Thus, if the
recognized as such. While the method may be adequate to value of H/VS for the i'h layer is defined as r-n then the
account for the inertia effects, it does not attempt to period elongation factor $ for the layer may be given by
addresses the kinematic effects due to interaction between &%, where 6 is given by Equation 8.
the pile and the surrounding soil that may undergo
significant nonlinear response during strong ground
shaking. From detailed analysis, it is noted above that the & Where (8)
ground response effects can be very significant.

The proposed method for the evaluation of With the displacement Ug relative to the bottom of the
ground response effects ith layer obtained from Equations 7, the ground
As is noted above in Equation 2, if the magnitude and displacement f(z) relative to the bottom of the soil layer,
where z varies from 0 at the top to H at the bottom of each
the nature of the distributed \oadp(x) could be determined
taking into consideration the local site condition and soil layer, may be given by the cosine function of Equation 9.
Once the displacements relative to the bottom of each
nonlinearity, it may be possible to solve the problem of
the beam on elastic foundation to estimate the forces layer is computed, the overall displacement/C^ relative to
acting on piles due to the kinematic effects of ground the pile toe can be easily obtained, such ihatp(x) is given
by Equation 10.
response. For this the displacement response f(x) of the
free field under the action of the earthquake needs to be
estimated. Considering the first mode of the free field (9)
motion, the elastic fundamental period of a soil layer of
thickness H and shear wave velocity Vs may be
approximated by 4H/VS, such that the maximum p(x) = kh Df (*); Q<zxxL (10)
displacement Ug occurring at the top of the /"' soil layer One of the crucial aspect in the proposed method is the
proper evaluation of the coefficient of subgrade reaction

Karkee, Sugimura & Fujiwara/ 7

3-7
kh in Equation 10, because its value affects the magnitude with even better contrast in case of variable 0. The ground
of the distributed load p(x) to a large extent once f(x) is displacement near the pile top is about 100mm.
determined. In this respect, it may be noted that the
elongation of the ground period by 0, times noted in The coefficient of subgrade reaction kh for the ground
Equation 7 corresponds to a soil stiffness degradation by response effect is obtained from Equation 11 by assuming
l.'fy2. This can be adequately accounted for in estimating the Poisson's ratio v of the soil to be 0.45. Resulting
the value of kh, which may be computed from the distribution of kh for constant 0 is shown in Fig. 14(a).
Equation 11 derived by Vesic (Poulos and Davis 1980) Assumption of a variable 0 based on Equation 8 results in
considering an infinite beam on elastic foundation. It may a kh distribution of Fig. 14(b). The distributed load p(x)
be noted that the expression for kh by Vesic accounts for obtained from Equation 10 corresponding to the two cases
the modulus of the soil as well as that of the pile material. of constant and variable 0 are shown in Fig. 14(c). The
Evidently, the soil modulus utilized to estimate kh should cosine distribution of the load is approximated by
consider the extent of nonlinear response that may be uniformly varying loads at discrete intervals in Fig. 14(c).
expected during strong ground shaking.

YS (rajs) Soil Movement (mm)

(11) .0 100 200 300 0 20 40 60 80 100
u 1 1 ' r 'i -i ! T- 'i

In Equation 11, El is the bending stiffness of the pile 2 - c- soil k c- soil jl

as defined earlier, v is the Poisson's ratio of soil and A is 4 _

the 'stiffness degradation factor' indicating the extent to

which the soil stiffness reduces during strong shaking !6 -

resulting in the ground period elongation by 0, as H 8 -

mentioned above. If there is a single soil layer over the fc

0.10 -

pile embedment depth, the value of A for it may be taken 2

as equal to <f?, otherwise the stiffness degradation factor A, e 12 '^K^JiV-"'!-'"
for the /"' layer is equal to 0/, where 0, is obtained from - 14
Ifir^-'
Equation 8. -"'-' f'''.T'
I"
It may be noted that although the effect of soil 18 -{Jv^J^- '
stiffness degradation due to nonlinear response would be f; . Constant <p
20 i\-^JC': Variable <p
clearly crucial in the evaluation of the kinematic stresses
T>
in the piles due dynamic soil-pile interaction, it is also
likely to be important in evaluating the inertia effects. FIG. 13. Shear wave velocity profile for c-soil and the
However, the simple design method proposed here is ground displacement for constant & variable values of <p
primarily concerned with the aspects of soil response
effects, and no attempt is made here to investigate the Load (MN/m)
effect of soil nonlinearity on the evaluation of the inertial 0123
forces in piles.
c-soil c- soil
Computation based on the proposed method
Attempt is made to compute bending moment and
shear force distribution in piles for the case of c-soil by D. 8
assuming the value of 0 to be 2.0 corresponding to the El
10
Centra input case mentioned above. As noted in Figs. 11
and 12, the ground response forces are most significant in 12
c-soil, which consists of a clayey soil layer underlain by a
stiffer sandy soil layer as shown in Fig. 4.
16

Fig. 13 shows the shear wave velocity profile together 18

with the soil displacement estimated by Equation 9. Fig.
20
13 shows the soil displacement distribution for the
constant <p case as well as for the variable 0 based on 22
Equation 8. The ground displacement depicts the shear (a) Kh Distribution (b) Kh for (c) Distributed load
wave velocity profile logically, giving much larger for constant <f> variable 0 cases (a) and (b)
displacement of softer layer compared to the stiffer layer, FIG. 14. Distribution of subgrade reaction for constant and
variable 0and the corresponding distributed loads on piles

Karkee, Sugimura & Fujiwaraf 8

3-8
 basement. When this fact is recognized, the maximum
The distributed lateral load on the pile shown in Fig. shear force of about 2.5 MN and the maximum bending
14(c) may not act all at once throughout the duration of moment of about 5.5MNm around the middle part of the
ground snaking, specially in case of multiple soil layers. pile in Fig. 16 are qualitatively comparable to those of
This may be understood from different possible modes of 2.5MN and 8.0MN-m in Figs. 11 and 12 respectively.
ground movement. This was previously confirmed This indicates that the simple solution proposed here is
(Sugimura et al. 1997) by the elastic eigenvalue analysis capable of accounting for the ground response effects in
of the three systems corresponding to the three soil piles that may be expected during ground shaking.
conditions of Fig. 4. It was found that the second mode of
c-soil showed the worst of the three soil type cases, The bending moment near pile top in Fig. 16 is of the
indicated by large bending moment and shear force order of about 20.0MNm compared to about 13.0MNm
around the two soil layers. To indirectly account for such for c-soil in Fig. 12. In contrast, the maximum bending
effects in the simple design method proposed here, it moment of 5.5MNm at the middle part in Fig. 16 is
would be logical to consider different combinations of smaller than that of about S.OMNm in Fig. 12. This
distributed load over different sections of the pile length. difference is most likely because of the assumption of full
restraint of pile top against rotation, which is unlikely to
Fig. 15 shows three possible distributed load be the case due to nonlinear response of the concrete piles
combinations assuming a constant value of <p, and in during strong earthquake shaking. In fact, the yield
consideration to the two distinct soil layers. The shear moment of the pile is about 14.0MNm as indicated in Fig.
force and bending moment diagrams for the three 8. Beyond yield level the pile top would tend to rotate,
distributed load cases of Fig. 15 and the kh distribution of resulting in only a partial restraint against rotation. When
Fig. 14(a) are given in Fig. 16. this condition is accounted for in the analysis by assuming
a certain degree of restraint against rotation rather than
the full restraint, the maximum bending moment obtained
from the detailed nonlinear analysis can be closely
approximated by the proposed simple method. For
example, assuming the degree of restraint to be 0.7 in Fig.
15 would logically amounts to limiting the moment at pile
top to the yield level. Under this condition, the bending
moment at the middle part of the pile in Fig. 16 would be
closer to about S.OMNm obtained from the detailed
29HN/I .29HN/n
nonlinear response analysis.
Shear Force (MN) Bending Moment (MN-m)
-3-2-10 1 2 -20 -10 0 10 20

(a)Full (b)Upper (c)Lower

FIG. 15. Three possible distributed load cases in the c-soil
consisting of two distinct soli layers over the pile length

The shear force and bending moment diagrams for the

three distributed load cases representing the ground
response effects is obtained by assuming the pile top to be
restrained against rotation with the pile tip free. It is seen
in Fig. 16 that there is a large shear force near the
interface of the two soil layers similar to that seen in Fig.
22
11, and that the bending moment too is large around there. -3 -2 -1

FIG. 16. Shear force and bending moment diagrams in a

It may not be logical to make a point to point pile considering the three distributed load cases in Fig. 15
quantitative comparison of the shear force and the
bending moment diagrams in Fig. 16 with those for c-soil Investigations based on large number of simulations
in Figs. 11 and 12, which are actually the envelop of would be required for further refinement of the proposed
maximum forces rather than the actual distribution. In method for actual application. However, the values of the
addition, Figs. 11 and 12 also include the inertia of the coefficient of subgrade reaction kh and the ground

Karkee, Sugimura & Fujiwara/ 9

3-9
displacement near pile top, applicable for realistic shear wave velocity, Poisson's ratio, thickness of soil
estimation of the ground response effects based on the layers etc are utilized.
proposed simple method, tend to be significantly different
from those recommended in the current design practice. To indirectly account for the different modes of
Evident the values of kh recommended in Japanese code ground movement in the proposed simple method, it is
for the seismic design based on the inertial forces can not logical to consider different combinations of distributed
be adopted for the evaluation of ground response forces load depending on the number of distinct soil layers over
the simple method proposed here. From this standpoint, it the pile length. Worst combination of the distributed load
seems logical to evaluate the inertial forces and the over different sections of the pile may be considered in
ground response forces separately, assuming different evaluating the ground response effects.
values of subgrade reaction kh as applicable, and then
design the pile for the envelop distribution of the largest The ground period elongation factor for a given site
of the two response forces along the pile length. depends on the local site condition as well as the level of
earthquake excitation. For extreme level of excitation at
CONCLUSIONS
sites with deep soil deposit, where the ground
deformation response effect is likely to dominate, may be
assumed to be in the range 2.0 to 3.0. The soil stiffness
Reports on the investigation of the damage to
degradation may be obtained from the ground period
foundations during past earthquakes show ample
instances of the failure of piles that can be directly or elongation factor.
indirectly attributed to the deformation response of the
The coefficient of subgrade reaction obtained based on
ground. Nonlinear response analysis based on the
the proposed method seems to work out to be much
combination of a finite element and lumped mass model
smaller than that recommended in the Japanese code for
indicates large response forces at the soil layer interfaces.
seismic design of piles, which directly accounts for the
The result demonstrates the importance of ground
inertial forces alone. Further investigation is necessary for
response effects in the seismic design of piles, while also
confirmation, but considering that the effects of the
illustrating the inadequacy of the present design practice
ground deformation response is directly related to the
that directly accounts for the superstructure inertia effects
extent of nonlinear response of the ground, it may be
alone.
logical to consider smaller value of subgrade reaction in
evaluating the ground deformation response effects. This
When the response of the total soil-pile-building
means evaluation of the inertia and the ground
system is decomposed into those of the building system
deformation response effects on pile by considering
and the foundation system, the contribution of the
different values of subgrade reaction as appropriate. It
foundation part is by far dominant, and comparable to that
of the total system. In addition, the contribution of the may even be possible for the coefficient of subgrade
reaction to assign a value of zero to depict a liquefied
building system itself is strongly dependent on the soil
condition. The results clearly demonstrate the influence layer that is not expected to provide any horizontal
of the local site condition on the inertial actions as well as support to the piles.
on the ground deformation response effects on piles.
The degree of restraint of the pile head against rotation
may be a major consideration in the evaluation of the
While demonstrating the importance of considering
bending moment distribution in piles due to ground
the ground deformation response in the design of piles,
deformation response effects. If the moment at the pile
the finite element response analysis also illustrates the
head exceeds the yield level, partial restraint at the pile
different levels of earthquake excitation to which the
top may be logically assumed to limit the moment there to
building superstructure might be subjected depending on
the soil condition of the site. the yield level. Conversely, if the moment at a section
exceeds the yield moment capacity, the moment may be
The proposed simple approach, based on the beam on redistributed to limit the moment to yield level, are
elastic foundation framework, is seen to be promising in relevant aspects for research and investigation.
capturing the essence of the ground response effects on
piles. The results indicate that it may be possible to
evaluate the nonlinear response of a soil-pile-building
system by such simple method provided the necessary APPENDIX I. REFERENCES
parameters are selected to reflect the dynamic
characteristics of the ground adequately. The essential AIJ (Architectural Institute of Japan). (1996). "Report on the
parameters in the proposed method include peak velocity, survey of building foundation damage due to the
ground period elongation factor, soil stiffness degradation Hyogoken-Nambu earthquake." AIJ Kinki branch
factor and degree of restraint at the pile head. In addition, foundation group, July 2996. (in Japanese)
the dynamic characteristic of ground such as the elastic

Karkee, Sugimura & Fujiwara/ 10

3-10
BCJ (Building Center of Japan). (1992). "Guideline for Sugimura, Y. (1988). "Japan's foundation design guide."
evaluation of design earthquake ground motion for Building research and practice, the journal of CIB, the
building design." Published by BCJ (in Japanese). int. council of building research studies and
documentation, Vol.16, No.2:109-121.
BCJ (Building Center of Japan). (1984). "Guideline on
Seismic design of building foundation and exemplary Sugimura, Y. (1987). "Earthquake damage of pile
design practice." Published by BCJ (in Japanese). foundation in Japan." International society of soil
mechanics and foundation engineering; Proc. 8th Asian
HBC (Highway Bridges Committee). (1995). "Survey report regional conf., Vol.2: 245-246.
on damages to highway bridges during the Hyogoken-
Nambu earthquake." Committee for earthquake disaster Sugimura, Y. (1981). "Earthquake damage and design
prevention in highway bridges, December 1995. (in method of piles." International society of soil mechanics
Japanese) and foundation engineering,; Proc. 10th int. conf.,
Stockholm. Vol. 2: 865-868.
Hetenyi, M. (1946). "Beams on elastic foundation." The
university of Michigan press, Ann Arbor. Sugimura, Y. & Oh-oka, H. (1981). "Report on the damage
of precast prestressed concrete piles during the 1978
Karkee, M. B. & Kishida, H. (1997). "Investigations on a off-Miyagi Prefecture earthquake." Building research
new building with pile foundation damaged by the institute, ministry of construction, (in Japanese)
Hyogoken-Nambu (Kobe) earthquake." The structural
design of tall buildings, Vol. 6: 311-332, John Wiley & Tokimatsu, K., Mizuno, H. & Kakurai, M. (1996). "Building
Sons, Ltd. damage associated with geotechnical problems." Soils
and foundations; special issue on geotechnical aspects of
Karkee, M. B., Nagai, K., Ogura, H. & Kishida, K. (1997). the January 17, 1995 Hyogoken-Nambu earthquake:
"Common behavior of building foundations during the 219-234.
Hyogoken-Nambu earthquake." International exchange
committee, Kansai branch of JGS (ed.), KIG forum on
geotechnical engineering in recovery from urban
earthquake disaster; Proc. int. symposium, Kobe, Japan,
January 9-10, 1977: 209-218. APPENDIX II. DISTRIBUTED LOAD METHOD

Karkee, M. B., Sugimura, Y., Tobita, J. & Sato, K. (1993). Based on the formulations on the beams on elastic
"Potential effects of long period components in incident foundations by Hetenyi (1946), Sugimura (1992) has
motion on the nonlinear ground response." Architectural proposed the use of distributed load method for the
institute of Japan (AIJ); Journal of struct, constr. engng, analysis of the horizontal resistance of piles. The
No.449: 69-82. solutions for a pile of semi-infinite length with the top end
fixed are given in the tabular form. Tables II-1, II-2 and
Karkee, M. B., Sugimura, Y. & Tobita, J. (1992). "Scaling a II-3 correspond to the solutions for concentrated,
suite of ground motions for compatible levels of uniformly distributed and triangularly distributed load
nonlinear ground response." AIJ Journal of struct,
constr. engng, No.440: 29-42. cases. The solutions for the three loading cases can be
readily combined to cover any general .loading pattern
Matsui, T. & Oda, K. (1996). "Foundation damage of assuming the superposition principle.
structures." Soils and foundations; Special issue on
geotechnical aspects of the January 17, 1995 The three tables can be utilized for the computation of
Hyogoken-Nambu earthquake: 189-200. displacement y, rotation 6, bending moment M and shear
force Q, at any point along the length of the pile, by
Poulos, H. G. & Davis, E. H. (1980). "Pile foundation multiplying the coefficient value x by the multiplier \i.
analysis and Design." John Wiley & Sons, Singapore. The explanations for the values of the multiplier given in
Table II-l are applicable to all the three Tables.
Sugimura, Y., Fujiwara, K., Ohgi, T. & Karkee, M. B.
(1997). "Seismic behavior of piles supporting tall
buildings and the consideration of ground response It may be noted that the relations given in Tables II-l,
effects in design." Tall buildings in seismic regions; II-2 and II-3 are applicable for a pile of semi-infinite
Proc. 4"' conf., Los Angeles, California, May 9-10,1977: length with the top end fixed. Similar simple formulations
303-317. can be developed for a pile of finite length and for
different boundary conditions generally encountered in
Sugimura, Y. (1992). "A proposal on analysis of horizontal practice.
resistance of pile based on load distribution method."
Japanese society of soil mechanics and foundation
engineering; Proc. 27th national conf., Kochi, June 2-4,
1992: 1637-1640. (in Japanese)

Karkee, Sugimura &. Fujiwara/ 11

3-11
 Table II-1: Concentrated load

Coefficient
Concentrated Load Multiplier \i
Values
Explanations
/\ 1\ ^ X *<<*. *<*x.

P
B, ft " *l\k h D
\/ Jc = SEI/3 3 3fl
/\
a
\ fP w
^ ^/ b = x c -x b
n
P -B2a
"c 4EI/3 2 2fl

B lb =e- f)b cos(/3b)

-ii B. B. B 2b = e ~ Pb sin(y3^)
\r
X
kh : Coefficie at of
subgra de reaction P
"id -B,
B 4 a = e"^{cos(y3fl)-sin(/?a)}

Table II-2 : Uniformly distributed load

Uniform Load Coefficient Multiplier |i

Values
0 xc <xa Xa ^ Xe ^ Xb xc >xb
/V \ /\ X
y
xc P B,-BW B.-B*
(
3
1\ /\ y* 8EI/3*
a
\/ \/
P n P **-. B3a~ B3b B3a' B3b
r b
P "c 8EI/33
p L>
\ -> \] /
p
M " ^
B- B2a B^ Blb B*-B.
x^ r

* -% B-B. B4a ~ B4b B4a' B4b

Karkee, Sugimura & Fujiwara/ 12

3-12
 Table II-3 : Triangularly distributed load

Triangular Load Coefficient Multiplier \JL

Values
0 ^y xc <xa x.**<*xt xc >xb
A /\ /\ ^ X

\1
/\ /\
Po
Vc ~ 16EIp 5 h ,..w D
fj D
^ fy ^O f?Li D
^f/^tflfj Ti ^T/yD
A /?/-*

a
\ \/ n Po D D i flit D
b D _!_/ ? _l_ &ll /? O
B la B lb + PhB 3a
> /\
>
> h

V
1 +
>
*
\ \/ \/ - ip*k
Po D /? _i_ O /?/i / ?

1 ?x
o^ c Po
40 2 h BI. - B2t + (ShB,, ~ B 2a ~ B 2b + fihBta ,,-,,...

Karkee, Sugimura & Fujiwaraj 13

3-13
 RESPONSE TO HORIZONTAL GROUND SHAKING OF
CANTILEVER RETAINING WALLS

A. S. Veletsos 1 and A. H. Younan2

1 Department of Civil Engineering, Rice University, Houston, Texas, USA

2 EQE International, Inc., Houston, Texas, USA

ABSTRACT

A broad overview is presented of the response to horizontal ground shaking of vertical, flexible
cantilever walls retaining a uniform, linear viscoelastic stratum of constant thickness and semi-
infinite extent in the horizontal direction. The response quantities examined include the
magnitude and distribution of the dynamic wall pressures and displacements, and the maximum
values of the total dynamic wall force or base shear and of the overturning base moment. Special
attention is given to the effects of very long-period, effectively static excitations. The effects of an
earthquake ground motion are then expressed as the products of the corresponding static effects
and appropriate amplification or deamplification factors. It is shown that the flexibility of the wall
may affect significantly the results, and that the maximum dynamic effects for flexible walls may
be significantly smaller than those for non-deformable, rigid walls.

INTRODUCTION

Despite the multitude of studies that have been carried out over the years, the dynamic response of
retaining walls is far from being well understood. There is, in particular, a lack of fundamental
information on the effects of some of the major parameters that influence the dynamic response of
such systems.

Previous studies of the problem may be classified into two groups: (1) elastic analyses in which
the backfill is presumed to respond as a linearly elastic or viscoelastic material; and (2) limit-state
analyses, such as the venerable Mononobe-Okabe approach (Mononobe and Matuo 1929, Okabe
1924), in which the wall is considered to displace .sufficiently at the base to mobilize the full
shearing capacity of the backfill. Detailed summaries of these contributions have been presented
by Nazarian and Hadjian (1979), Prakash (1981), Whitman (1991), and Veletsos and Younan
(1995).

The most comprehensive previous study of the elastic response of the system is the one presented
by Wood (1973), in which the wall was considered to be rigid. In the present study, the
assumption of linear response for the retained material is preserved, but the wall is considered to
be flexible. The objective is to assess the effects of this flexibility on the magnitude and
distribution of the wall pressures and displacements, and on the maximum values of the total wall
force or base shear and of the overturning moment induced by horizontal ground shaking. The
4-1
material presented is essentially a summary of some of the more important information reported
in a series of recent contributions by the authors (1994, 1995, 1997).

SYSTEM CONSIDERED

Shown in Figure 1, the system examined consists of a fixed-based, vertical cantilever wall
retaining a semi-infinite stratum of uniform viscoelastic material. The wall is considered to be of
uniform thickness and finite rigidity, and the retained material is considered to be free at its upper
surface and bonded to a non-deformable, rigid base. The bases of the wall and stratum are
presumed to experience a space-invariant horizontal motion, the acceleration of which at any time
t is x (t) and its maximum value is Xg . Material damping for both the medium and the wall is
considered to be of the constant hysteretic type.

The properties of the soil stratum are defined by its mass density p, shear modulus of elasticity
G, Poisson's ratio v, and material damping factor 8, which is considered to be the same for both
shearing and axial deformations. The factor 8 is the same as the tan 8 factor used in some
previous studies of foundation dynamics and soil-structure interaction and twice as large as the
damping factor P expressed in terms of the critical coefficient of damping. The properties of the
wall are defined by its thickness tw , mass per unit of surface area \iw , Young's modulus of
elasticity Ew , Poisson's ratio v w , and damping factor 8W which, like 8, is twice as large as the
corresponding factor expressed in terms of the critical damping value.

Figure 1: System considered

METHOD OF ANALYSIS

The method of analysis used has been described in Veletsos and Younan (1997) and Younan,
Veletsos and Bandyopadhyay (1997), and it is summarized only briefly here.
Fundamental to the method is the assumption that, under the horizontal excitation considered, no
vertical normal stresses develop anywhere in the medium, i.e. d = 0. It is further assumed that
the horizontal variations of the vertical displacements for the medium are negligible, so that the
horizontal shearing stresses T can be expressed as T = G*(3w/3v), where u is the horizontal
displacement of an arbitrary point of the medium relative to the moving base, G* = G( 1 + i8) is

4-2
the complex-valued shear modulus, and i = . The reliability of these assumptions has been
confirmed by Veletsos and Younan (1994) by comparing the solutions for rigid walls with those
obtained by Wood's 'exact' solution (1973).
The steps involved in the analysis are as follows:

First, the horizontal displacements of the medium relative to the moving base are expressed as
linear combinations of the natural modes of vibration of the stratum when it is considered to act
as an unconstrained, vertical cantilever shear-beam. Similarly, the relative displacements of the
wall are expressed as linear combinations of its natural modes, i.e. those of a clamped-free
beam. These expressions satisfy the conditions of zero displacement at the rigid base, and of
zero stresses and forces at the top boundary.
Next, the differential equation governing the motion of the medium in the horizontal direction is
solved subject to the condition of compatible displacements at the interface of the medium and
wall. For the purpose of satisfying this boundary condition, each natural mode of vibration of
the beam is expanded in terms of the corresponding modes of the retained medium. The result
of this step are medium displacements and wall pressures that are functions of the as yet unde-
termined wall displacements.
Finally, the wall displacements are evaluated by satisfying the dynamic equilibrium of the
forces acting on the beam using Lagrange's equations of motion.

The analysis is first implemented for harmonic excitations. The responses to arbitrary transient
ground motions are then evaluated by Fourier transform techniques.

PRESENTATION OF RESULTS

Consideration is first given to the responses obtained for excitations the dominant frequencies of
which are small compared to the fundamental natural frequency of the soil-wall system. This is
tantamount to neglecting the dynamic amplification effects of the retained medium. Such
excitations and the resulting effects will be referred to as 'static', a term that should not be
confused with that normally used to represent the effects of gravity forces. A maximum dynamic
effect for an arbitrary excitation is then expressed as the product of the corresponding static effect
and an appropriate amplification or deamplification factor.

Static Effects

In Figure 2, the total force or base shear per unit of wall length induced by the maximum static
values of the lateral inertia forces, Pst , is plotted as a function of the dimensionless measure of
the wall flexibility

in which Dw represents the flexural rigidity per unit of wall length and is given by

4-3
 D, = (2)
12(1-v*)

The results are normalized with respect to pXgH2 . Also shown is the distance h from the base to
the line of action of the resultant of the wall force, normalized with respect to the wall height H.
Referred to as the effective height, h represents the distance by which Pst must be multiplied to
yield the static value of the overturning moment per unit of wall length. The wall in these
solutions, and all others that follow, is considered to be massless, and Poisson's ratio for the
retained medium is taken as v = 1/3.
It is observed that the flexibility of the wall affects significantly both the magnitude of the wall
force and the line of action of its resultant. Increasing the wall flexibility reduces the horizontal
extensional stiffness of the retained material relative to its shearing stiffness, and this reduction, in
turn, increases the proportion of the inertia forces transmitted by horizontal shearing action to the
base, and decreases the proportion transmitted to the wall. Since the effective height also
decreases with increasing wall flexibility, the reduction for the overturning base moment is
generally much greater than for the corresponding shear. The reductions are substantial even for
rather small values of dw . For example, Pst is reduced from 0.941 pXgH2 for a rigid wall to
0.561 pXgH2 for a system with dw = 10. The corresponding reduction in effective height is from
0.6/f to 0.375#, and that for base moment is from 0.565pX^#3 to Q.2\QpXgH3 .

Figure 2: Normalized values of base shear and effective heights in wall of

statically excited systems with different wall flexibilities; v = 1/3,

4-4
The effects of wall flexibility on the magnitude and distribution of the wall pressures, pst (T\) , is
shown in Figure 3 in which T] = y/H . For rigid walls (dw = 0), the pressures increase almost as
a quarter-sine from zero at the base to a maximum at the top, whereas for the flexible walls, there
is a sharp change in the intensity of the pressure near the top, with the pressure decreasing and
changing signs.

0.8-

0.6-

0.4-

0.2-"

-1

Figure 3: Distributions of wall pressures for statically excited systems; v = 1/3 ,

The maximum value of the wall displacement, which naturally occurs at the top, may be
expressed either in terms of the wall properties as

(3)

or, more conveniently, in terms of the properties of the retained medium as

pXH _ XgH
~ C ~ C 2 ~~ (4)

where v^ = *jG/p = the shear-wave velocity for the medium, and c l and c2 are dimensionless
factors that are functions of the wall flexibility factor dw and are interrelated by

= c (5)

4-5
The dependence of these two coefficients on dw is shown in Figure 4.

0.5 n

0.1-

0.05H
^

0.01 I I I
10 20 30

Figure 4: Coefficients C| and c2 in expressions of maximum displacements relative

to base for statically excited systems with different wall flexibilities;

The deflection configurations of the wall of systems with different values of the relative flexibility
factor dw are shown in Figure 5. The results are normalized to a unit value at the top. It is
observed that, within the range of dw values examined, the results are quite similar. As dw * 0 ,
the configuration naturally tends to that obtained for the pressures exerted on a non-deflecting,
rigid wall.

As a measure of values of maximum wall displacements that may be encountered in practice,

consider a concrete wall of height H = 4.6 m (15 ft) and thickness tw = 0.46m ( 1 .5 ft) retaining
a medium with shear-wave velocity vs = 122 m/sec (400 ft/sec) and unit weight y = pg = 1.6 t/
m3 (100 pcf), where g is the gravitational acceleration. With Ew = 20684 MPa (3000 ksi) and
V^^O.17, the displacement factor c2 in Equation (4) becomes 0.427; and for a peak ground
acceleration Xg = Q.3g, the static value of the maximum wall displacement becomes 0.039
percent of the wall height. Even with a dynamic amplification factor of 2.0, which, based on
information presented in the following sections, represents a reasonable maximum for intense
earthquake ground motions, the resulting maximum dynamic displacement is below the 0. 1 to 0.4
percent range widely accepted as representing the displacement values required to induce a limit
state in the backfill material (Clough and Duncan, 1990). This information is offered as a
confirmation of the applicability of the elastic solutions for the systems examined here.

Dynamic Effects

As already noted, the maximum dynamic effects for an arbitrary transient excitation may
conveniently be expressed as the product of the corresponding static effects and appropriate

4-6
 max

Figure 5: Distributions of wall displacements relative to base for statically

excited systems with different wall flexibilities; v = 1/3, \LW = 0

amplification or deamplification factors, AF. Figure 6(a) shows the values of this factor for the
total force or base shear per unit of wall length for systems subjected to the first 6.3 sec of the N-
S component of the ground motion recorded during the 1940 El Centre, California earthquake.
The peak acceleration of this motion is Xg = 0.312 g. Four values of the wall flexibility factor dw
are considered, with the results plotted as a function of 7*p the fundamental natural period of the
stratum when it is considered to respond as an unconstrained, cantilever shear-beam. This period
is given by T^ = 4H/vs . The wall in these solutions is presumed to be massless, the material
damping factors for the wall and retained medium are taken as 5^= 0.04 and 8 = 0.10 (i.e., 2
percent and 5 percent of critical damping, respectively), and Poisson's ratio for the medium is
taken as v = 1/3.
The plots in Figure 6(a) are similar to, but by no means the same as, the non-dimensionalized
response spectra for similarly excited, viscously damped single-degree-of-freedom systems.
Specifically, for low-natural-period, stiff media, the AF = 1, indicating that the peak value of the
dynamic base shear is equal to its corresponding static value. With increasing medium flexibility
or 7*j, the amplification factors increase, and after attaining nearly horizontal plateaus, they reach
values that may be substantially lower than unity. Within the practically important period range of
Tl = 0.1 to 0.5 sec, for which the amplification factors are nearly constant, these factors are
relatively small, particularly for the stiffer walls with the lower values of dw . The average
amplification factor within this range varies from 1.33 for a rigid wall to 1.68 for a wall with dw =
5 and 1.82 for a wall with dw = 15. These relatively low values are due to the capacity of the
medium to dissipate energy by radiation of waves to the far field. The greater the wall stiffness
relative to that of the retained medium, or the smaller the dw , the greater is the capacity of the

4-7
 1.5
30

15

AF 1-5

0.5

0.5
(b)
il in I I ILL
0.02 0.1 0.02 0.1
sec sec

Figure 6: (a) Amplification factors for maximum wall force or base shear per unit of wall
length for systems with different wall flexibilities subjected to El Centro
earthquake record, and (b) Normalized values of this force; p,w = 0 , 5W = 0.04 ,
v= 1/3 and 6 = 0.1

wall to reflect the waves impinging on it and to dissipate them by radiation to the far field.
Further insight into the magnitude of the maximum dynamic force or base shear per unit of wall
length, Pmax , may be gained from Figure 6(b), in which this quantity is replotted normalized with
respect to the common factor pXgH . As before, four values of d w are considered, with the
results plotted as a function of the natural period of the retained medium, j. It is observed that
increasing dw reduces the dynamic wall wal force over the full range of 7\ the reductions being
largest for the low and medium values of
Figure 7 shows the normalized values of the effective height h for the seismically excited systems
examined. It is observed that this height, which represents the distance by which the maximum
dynamic wall force or base shear must be multiplied to yield the corresponding base moment, is
insensitive to variations in Tv Accordingly, it may, for all practical purposes, be considered to
have the values displayed in Figure 2 for statically excited systems. It further follows that the
dynamic amplification factor for base moment may be taken equal to that for the wall force or
base shear. These simple relations are consequences of the fact that the response of the system is
basically dominated by its fundamental mode of vibration.

Effect of Wall Mass

For the systems examined so far, the wall was presumed to be massless. The effect of the wall
mass is twofold: (1) it modifies (generally decreases) the wall pressures induced by the retained
medium; and (2) it induces additional forces on the wall. The net effect is generally an increase in

4-8
the magnitude of the wall forces.

A precise evaluation of these effects is beyond the scope of this contribution. However, a
reasonable approximation to the additional base shear may be obtained from the expression
meXg (AF) , where me = the effective wall mass, Xg = is the maximum ground acceleration, and
AF = the appropriate amplification factor for the massless wall. The effective mass me may be
considered to decrease from the total wall mass for a non-deforming, rigid wall to 70 percent of
the total mass for a wall with dw = 10, and to 60 percent of the total mass for a wall with dw = 30.
In the computation of the corresponding base moment, the effective height h may be taken equal
to that determined for the massless wall.

0.8

0.6
h_
H
0.4

0.2

i i i i mi I I I I I MM
0.02 0.1 1
sec
Figure 7: Normalized values of effective height for systems with different wall
flexibilities subjected to El Centre earthquake record; p,w = 0 , 8W = 0.04 ,
v = l/3 and 8 = 0.1

CONCLUSIONS

The magnitudes and distributions of the wall displacements, wall pressures and associated forces
induced by horizontal ground shaking in the systems examined are quite sensitive to the flexibility
of the wall. Increasing this flexibility reduces the horizontal extensional stiffness of the retained
medium relative to its shearing stiffness, and this reduction, in turn, decreases the proportion of
the soil inertia forces that gets transferred to the wall and, hence, the forces developed in it.
For realistic wall flexibilities , the total wall force or base shear may be of the order of one-half or
less of that obtained for non-deformable, rigid walls, the reduction in base moment being even
larger. The reduced effects may well be in reasonable agreement with those obtained by the

4-9
Mononobe-Okabe method. This agreement, however, does by no means constitute a validation of
the latter method, which represents an approximate, limit-state analysis for the problem.
Even for the 1940 El Centre, California earthquake ground motion record, the maximum wall
displacement relative to the moving base for realistic wall flexibilities is found to be less than the
values of 0.1 to 0.4 percent of the wall height normally accepted as the minimum required to
develop a limit state in the retained material.

ACKNOWLEDGMENT

This study was carried out under Projects 558223 and 568821 from Brookhaven National
Laboratory, Upton, New York. This support is acknowledged gratefully.

REFERENCES

Clough, G. W. and Duncan, J. M. (1990). Earth pressures, Foundation Engineering Handbook,

Ed. by H. Y. Fang, Chapman & Hall, NY, USA, 223-235.
Mononobe, N. and Matuo, H., On the determination of earth pressures during earthquakes. Proc.
world engrg. congress, Tokyo, Japan, 1929, 9, Paper No. 388.
Nazarian, H. N., and Hadjian, A. H., Earthquake-induced lateral soil pressures on structures. J. of
the Geotech. Engrg. DivASCE, 1979,105(9), 1049-1066.
Okabe, S., General theory of earth pressure and seismic stability of retaining wall and dam. J.
Japan soc. civil engrs.,1924, 12.
Prakash, S., Analysis of rigid retaining walls during earthquakes, Proc. Int. Conf. on Recent Adv.
in Geotech. Earthquake Engrg. and Soil Dyn., University of Missouri, Rolla, MO, USA, 1981,
III, 1-28.
Veletsos, A. S. and Younan, A. H., Dynamic modeling and response of soil-wall systems. J. of
geotech. engrg. ASCE, 1994,120(12), 2155-2179.
Veletsos, A. S., and Younan, A. H., Dynamic soil pressures on vertical walls. Proc. Third Int.
Conf. on Recent Adv. in Geotech. Earthquake Engrg. and Soil Dyn., University of Missouri,
Rolla, MO, USA, 1995, HI, 1589-1604.
Veletsos, A. S. and Younan, A. H., Dynamic response of cantilever retaining walls. /. of geotech.
engrg. ASCE, 1997,123(2), 161-172.
Whitman, R. V, Seismic design of earth retaining structures. Proc. 2nd Int. Conf. on Recent Adv.
in Geotech. Earthquake Engrg. and Soil Dyn., University of Missouri, Rolla, MO, USA, 1991,
n, 1767-1777.
Wood, J. H. (1973). Earthquake-induced soil pressures on structures, Report EERL 73-05,
Earthquake Engineering Research Laboratory, California Institute of Technology, Palo Alto,
CA, USA.
Younan, A. H., Veletsos, A. S. and Bandyopadhyay, K. (1997). Dynamic response of flexible
retaining walls, Report BNL-525I9, Brookhaven National Laboratory, Upton, NY, USA.

4-10
 Dynamic Behavior of Pile Foundations in Liquefaction Process
- Shaking Table and Oscillator Tests Utilizing Big Shear Box -
by
Hatsukazu MIZUNO1, Michio SUGIMOTO2, Masanori IIBA1, Toshihiro MORP and Tsutomu HIRADE1

ABSTRACT : The paper presents shaking table tests and oscillator tests utilizing big shear box
to clarify pile foundation behavior in liquefaction process. It is, in principle, impossible for
liquefaction process to satisfy similitude ratios in reduced models of prototype water-saturated
sands in centrifugally accelerated field. Therefore, we carried out shaking table tests in near-to
fullscale models of water-saturated sands and piles to break through the above-mentioned bottle
neck. The effect of ground water table depth on liquefaction and pile behavior is examined. And
the effect of an excess pore water pressure dissipation method, that is adopted as a countermeasure
against liquefaction, is also evaluated. In a case of low ground water table, the liquefaction is not so
severe and bending moments of piles is reduced extremely.

INTRODUCTION pile foundations based on an expanded sway and rocking

model in the linear region of soil properties.
The research was conducted under a series of a activity 2)a proposal of evaluation method on internal stresses of
related to a project of the Construction Technology, pile foundations based on the Penzien model (lumped mass
Research and Development(Ministry of Construction) model) in the linear and nonlinear region of soil
entitled "Development of technology for earthquake properties(including detailed discussion of evaluation of
disaster prevention in large metropolitan areas" in several constants in the model).
collaboration with researches and engineers of 3)the comparison of the results between by the expanded
universities and private companies. Purposes of the SR model and by the Penzien method.
subcommittee in the project are to clarify seismic actions to 4)Verification through the comparison with the results of
building foundations or substructures under nonlinear more rigorous analyses and shaking table tests on soil
ground vibration and during liquefaction, and to develop grounds.
the seismic design method of the building foundations or The paper presents the results and discussions on the
substructures through incorporating dynamic soil-structure shaking table tests of the pile foundations utilizing the big
interaction based on real phenomena. On the way, in 1995, shear box of about 6m, 11.6m and 3.1m in height, length
the Hyogoken-nanbu earthquake occurred and many and width, respectively. It is, in principle, impossible for
reports about pile damages were shown. liquefaction process to satisfy similitude ratios in reduced
Contents of the subcommittee are a proposal of design models of prototype water-saturated sands in centrifugally
model of buildings incorporating dynamic soil-structure accelerated field. A main reasons for using the big shear
interaction and shaking table tests of pile foundations box is to carry out the experiment in near-to fullscale
utilizing a big shear box. The contents of the former are as models of water-saturated sands and piles.(Ref. 1) The
follows; shaking table which was used in the experiment is large
l)a proposal of evaluation method on internal stresses of shaking table(maximum weight is 500 tonf) of Disaster
Prevention Research Institute of Scientific Technology
Agency.
1)Building Research Institute, Ministry of Construction
2)Construction Technology Institute, Takenaka-Koumuten, DYNAMIC PROPERTIES OF SOIL
Inc.
3)Constmction Technology Institute, Kumagai-Gumi, Inc. The shaking table test on dynamic properties of the non-
5-1
liquefied ground whose soil surface is dry(in this case the random wave excitation with white noise, a static test, that
height of the ground is 3.48m). The soil is called is, horizontal force to a pile head, and an oscillator test, that
Kasumigaura sands whose physical properties are shown in is, dynamic excitation with a constant frequency to the pile
Table 1. A uniformity coefficient is about 3. Figure 1 head during dissipating stage of water pressure were
presents strain dependency of equivalent shear stiffness conducted. An earthquake wave form using in the
and equivalent damping factor base on the dynamic triaxial experiment is a north and south component record at 32m
test(a confining pressure: 0.5kgf/cm2). in depth observed in Port Island in Kobe city in 1995. The
In order to investigate a change of fundamental dynamic maximum acceleration of input excitation was adjusted to
properties of the soil ground, which is caused by soil be 500 gals.
settlement due to repeated excitation, the test of random Production(reproduction) methods of soil deposits are a
noise excitations with maximum acceleration of 60 gal was drop of sands into water(Case CD), and a boiling from
carried out. Figure 2 presents dynamic properties(the bottom of soil deposit by high water pressure(Case(D -
Fourier spectrum ratio of ground surface to bottom and 1st (D). Table 3 shows heights and relative densities of
mode shape) of the non-liquefied ground. In case of sands produced soil deposits.
with dry surface, a distribution of shear stiffness of soil Figure 3 represents measurement points in the
deposits with depth is similar to that assumed to be experiment. In this figure, a ground water level of each
proportional to a square root of overburden pressure of soil. experimental series and a position of drains for the
When the shear stiffness is constant with depth, the countermeasure against liquefaction by dissipation in case
vibration mode shape is the function of cosine. Also when @ are also drawn. In the ground (measured lines G-l to
the shear stiffness is proportional to depth, the soil deposit G-5), accelerometers and excess pore water transducers
has the vibration mode shape which much changes near the were installed. In the piles (P-l), accelerometers, excess
surface. pore water transducers and earth pressure transducers(on
the both sides perpendicular to vibration direction) were set
SHAKING TABLE TEST OF PILE FOUNDATION IN Also strain gauges were attached on pile surfaces (P-l and
LIQUEFIED SOIL UTILIZING BIG SHEAR BOX P-2) to measure bending moments and shear forces. In the
foundation and the shear box, accelerometers and
3.1 Experimental Series and Purpose of Experiment displacement transducers were installed.
The dimensions of soil deposit in the big shear box used in
the experiment are about 6m, 11.6m and 3.1m in height, 3.2 Dynamic Properties and Vibration Mode of Soil
length and width, respectively. A model of pile foundation Ground
is made of steel and is 40cm, 10cm and 5.82m in width, Dynamic properties of soil ground under random wave
thickness and length, respectively. excitation (max. acceleration 30 gals), are drawn in Fig. 4.
Table 2 shows the experimental series and purposes of Predominant frequencies of soil ground are about 5.3 5.6
each experiment In case CD, (D and (D, the purposes are Hz and 6.85 Hz in Case CD - (D, and Case ,
to clarify the behaviors of pile foundation during respectively.
liquefaction, influences of production methods of soil The vibration mode at the predominant frequency in
deposit on the behaviors and to confirm the reproduction of Case CD is shown in Fig. 5. In the figure, the vibration
dynamic behaviors by boiling production method. In case modes corresponding to the soil grounds with constant
(D, that is to investigate the effect of a countermeasure shear stiffness through depth and with stiffness
against liquefaction by dissipation method of excess pore proportional to overburden pressures are also drawn. The
water pressure using vertical drains. In case (D, that is to distribution of shear stiffness of produced soil ground is
confirm the effect of level of ground water on liquefaction similar to that in the proportional stiffness to overburden
process(low ground water level is an idea of the pressures. The distribution in saturated soil ground is
countermeasures against liquefaction). different from that in soil ground consisted of dry sand as
In each case, in addition to earthquake excitation and shown in Fig. 2. Under the small vibration(acceleration),
5-2
the pore water seems to contribute the stiffness of the of the shear stiffness is less in the soil deposit with non-
ground. Through the contribution of pore water to stiffness, saturated soil layer, the displacement responses of soil
the excess pore water will increase under the large deposit is less. The displacement of the shear box in Case
vibration and the effective stress will decreases. (D is about one fourth of that in Case (D. The amount of
pole bending moments is corresponding to the difference of
3.3 Liquefaction Phenomena Under Earthquake the displacement of shear box.
Excitation From a result of comparison of responses between in
(1) Comparison by Production Method of Soil Deposit Cases (D and (D, the excess pore water pressure and its
Time histories of main measuring points in Cases (D and build-up process, the acceleration of foundation and the
(D are illustrated in Fig. 6. The maximum response bending moment of piles have the similar responses. It
distribution of accelerations, excess pore water pressures shows that the effect of reducing increase of excess pore
and pile bending moments are drawn in Fig. 7. The water pressure by vertical drains is not remarkable under
production method in Case (D and Cases d) and (D are the earthquake level in the experiment. However, in the
the drop of sands into water and the boiling from bottom of dissipation process of the excess pore water pressures after
soil deposit by high water pressure, respectively. The the excitation, the dissipation rate is 4 times by vertical
difference of the process of excess pore water pressures drains, as shown in Fig. 10.
between in Case (D and in Case (D is little except that it (3) Behavior of Piles
takes more time to increase the excess pore water pressure The distribution of bending moment of piles and excess
in depth less than 2m from the soil bottom in Case (D than pore water pressures at several stages of excess pore water
in case (D. The excess pore water pressure and its process, pressure level are drawn in Fig. 11. During the increase
the acceleration of foundation and the bending moment of process of excess pore water pressures, the distribution that
piles have almost the same responses in Cases (D and (D the amount of bending moments is large in the depth near
except little difference of soil ground density. From results ground surface(shallow depth) can be seen. There are
that the distribution of maximum excess pore water enough subgrade reactions to piles in deeper soil layers.
pressure and, maximum acceleration of foundation and the After the increase of excess pore water pressures in all of
bending moment of piles through depth in Cases (D to (D, the layers, the bending moment distribution of piles is
it is clarify that the liquefied soil deposits are reproduced linearly changed and the reduction of subgrade reactions
by the boiling. due to liquefaction is verified.
(2) Effect of Countermeasure Against Liquefaction
The time histories of responses and the distribution of OSCILLATOR TEST OF PILE FOUNDATION
maximum responses in Cases (D and (D are shown in
Figs. 8 and 9, respectively. The ground water levels are the hi order to investigate relationships between the subgrade
ground surface and 1.4m in depth from the surface in Cases reaction of the soil and the excess pore water pressure, an
(D and (D, respectively, hi Case (D, all of the soil deposit oscillator(an eccentric moment is lOOkg.cm) was installed
is saturated. On the other hand, the soil deposit in depth on the footing of the pile foundation. After the earthquake
less than 1.4m is not saturated and the soil deposit in depth excitation, the oscillator test was conducted in several
more than 1.4m is saturated, hi Case (D. The amount of stages of dissipated process of the soil deposit A excited
increase of excess pore water pressures in saturated soil way of the oscillator is a sweep excitation with increase or
ground of case (3) is larger than that in soil ground with decrease of frequency between 3 and 16 Hz in time period
non-saturated soil layer of Case (D. hi Case (D, there is of 50 seconds. Following results are those under the
little increase of excess pore water pressures in non- frequency increase excitation in the dissipated process of
saturated soil layers. The maximum acceleration of soil Case (D.
deposit is larger in Case (D than that hi Case (D. On the (l)Responses of Footing and Pile
contrary, The maximum bending moment of piles is less. The time history of excess pore water pressure in the
As the decrease of the effective stress, that is, the decrease dissipated process after the earthquake excitation are

5-3
shown in Fig. 12. The broken lines means a time period subgrade reactions are obtained to be differentiated twice
when the excess pore water pressures were not measured with respect to a time t. The displacements are obtained to
just after the earthquake excitation and during final stage of be integrated twice with respect to a time. Boundary
dissipation. The excess pore water pressure is zero before condition of the displacement at the pile tip is zero and that
the earthquake excitation. Symbols Tl to T6 means several at pie head is equal to the displacement of the footing
stages in dissipating process. Though the increase of excess measured in the experiment. The coefficient of the
pore water pressure occurs during the oscillation test, the subgrade reaction is a division of the subgrade reaction by
influence on the whole dissipation process seems to be the displacement and the pile width.
little. The distributions of excess pore water pressures at The distributions of the pile displacements and the
the several stages are drawn in Fig. 13. In the figure, the subgrade reactions obtained from the pile bending
distribution of the maximum excess pore water pressures moments are shown in Fig. 17. In the figure, the
during earthquake excitation is plotted. The dissipation of distribution of displacement obtained from the measured
excess pore water pressures starts in the deeper layer and acceleration and the distribution of subgrade reaction
the dissipated layers move to shallow layers gradually with obtained from the measured earth pressures. The
time. approximate polynomial equations are compatible to the
Resonant curves of the foundation displacement at the distribution of measured data except a little difference of
stages Tl to T6 are illustrated in Fig. 14. With the values. The coefficients of the subgrade reactions is drawn
dissipation, the resonant frequencies are high and the in Fig. 18. The coefficient of subgrade reactions with
amplitudes at the resonant frequencies is low. When the nearly zero displacement is omitted in the graph. In early
dissipation goes to a certain extent, the resonance of the dissipation stage, the coefficients are large in the relatively
soil deposit(about 5 Hz) appears again by getting the shear deep layers. With reproducing shear modulus of soil, the
stiffness of the soil. The distribution of the pile bending depth at the maximum of the coefficients is gradually
moments at the resonant frequencies of each stage is shown shallow.
in Fig. IS. The maximum of pile bending moments occurs
at the pile head in every case. The second maximum points CONCLUSIONS
of bending moments are 2.5 to 4 m from ground bottom
and are more shallow with dissipation. The results of the shaking table tests of the pile
(2)Subgrade Reaction During Dissipation foundation during liquefaction and the oscillator tests of
The subgrade reactions and coefficients of subgrade that during dissipation are summarized as follows.
reaction to piles are estimated from the data of bending 1) It is clarify that the liquefied soil deposits are
moments of piles. reproduced by the boiling except little difference of soil
The distribution of earth pressures to the pile at the ground density.
resonant frequencies of each stage is presented in Fig. 16. 2) The effect of low ground water level on preventing from
The earth pressure transducers were installed at the both liquefaction is remarkable.
sides in the excitation direction. The earth pressures are 3) The effect of the vertical drains(the countermeasure
obtained from values of outer and inner method against liquefaction) is not remarkable in the
transducers(difference between two). The depth at the increasing process of excess pore water pressure during
maximum earth pressures moves at shallow depth with the liquefaction. But the vertical drains are effective on
dissipation. The change of depth at the maximum bending dissipating the excess pore water pressure after
moments and the maximum earth pressures in the ground excitation.
means the reappearance of soil stiffness with dissipation. 4) The subrgade reaction is reproduced with dissipation.
The distribution of bending moments is approximately The deeper layers are early reproduced.
expressed by a polynomial equation. The subgrade 5) The shear stiffness is reproduced with dissipation and
reactions do not act to the piles in the depth where the . the resonant frequency of the soil ground appears.
settlement of ground surface occurs during dissipation. The
5-4
AKNOWLEDGEMENT of the design for earthquake resistance (building part) in
the Development of Technology for Earthquake Disaster
The research was conducted in a subcommittee under the Prevention in Large Metropolitan Areas, annual reports
project entitled "Development of Technology for of 1993 - 1997 (in Japanese)
Earthquake Disaster Prevention in Large Metropolitan 2)Ishihara K., T. Kagawa and C. Minowa et al. : Design
Areas" in collaboration among Building Research Institute, and Setup of Large scale Liquefaction Reproducing
Housing & Urban Development Corporation and Building Facilities, Proc. of 31th presentation of the Japanese
Construction Society(The subcommittee of design for Society of Soil Mechanics and Foundation Engineering,
earthquake resistance of building foundations, chairman 1996.6, pp.1189-1190 (in Japanese)
is Dr. Yutaka Matsushima, Professor of Tsukuba 3)Iguchi M. et al.: Evaluation of Seismic Behavior of Pile
University) The experiment was carried out a working Foundation and Seismic Action on Piles by Shaking
group under the subcommittee (chairman is Dr. Michio Table Test of Large-scale Shear Box, Summary of annual
Iguchi, Professor of Science University of Tokyo). The meetings of Architectural Institute of Japan, 1997.9,
authors express our thanks members of the subcommittee pp.369-396 (in Japanese)
and the working group for their sincere advice to conduct 4)Mizuno H. et al.: Dynamic Behavior of Pile Foundations
the experiment. in Liquefaction Process - Shaking Table Tests Utilizing
Big Shear Box -, Proc. of 10th Japan Earthquake
REFERENCES Engineering Symposium, 1998.11 (in Japanese, now
adjusting the paper for final submission)
l)Building Research Institute : reports of a subcommittee

Tablel Soil Test Result

ouu
shear
Test item Result 500
' rnodU
--^
density (g/cm^) 2.718 400
gravel fraction $6)
sand fraction (%)
silt fraction (%)
day fraction (%)
0.8
948
3.5
0.9
300
200
I:
100
uniformity meflfiripnt aoi rt
D50(mm) 0.311
natural water content 5.71 0.1 1 10 100 Q1 1 10 100
shorsbanft) shea-shrift)

figl Equivalent Shear Modules & Dumping

20 1 fl=gnnata nt
G=k"z
G=SQRT(Z)
Experimental results
0 15
I

110
CO

10 16 20
(Hz)

H&2 Spectral Ratio on Random Wave(60gal)

5-5
 Table2 Test Series and Objects

Test Case Shaking Table Teat(Oinput) Static Oscil-

: & Ean- Earthquake wave , gal Load laterTest
Test Ground fio^n (Port island GL-32mJSFs) (before Object
Condition. gliaVp
30gal
600 200 re-500 table)
Case Qj Oi check the liquefaction
Drop of Sand into Cte r^*afinenljrim pni=U2 Ou at seismic behavior of
Water, water niftBipfltinn ground & foundations
fevelOL
Case C& Oj a seismic behavior of
Produced by Qts *600gal Ou at ground & foundations
boflinfii (compare Case (1)
water level G.L.
Case Ol5 reappearance of
i _j
Reproduced by 0^75 04 Cb Oio OM DeiDre seismic behavior of
bailing, eartiurualffi ground & foundations
water level G-L meai jured dissipation
wave (compare Case d)
Case seismic behavior of
Reproduced by 03 Qs 0; a ground & foundations
boiling with Qua under the drain
drain, meai lured dissipiation (compare Case
water level G.L
Case 04 seismic behavior of
Reproduced by O5&7 Oa a a Ow at ground & foundations
Killing; M LH^nflfl HflT^ at lew water level
measured disapabon
water level (compare Case d$
GJ*-L4m
OmdexiTkst Order

LfiOQ
Ace. transducer O Pore pressure meter X Strain geuge
> Dlsp.transducer A Earth pressure
I Water level G-6 cell
jWater levei^G-1 G-2P-1MP-2 Base

6,820

1^3 Outline ofTest Specimen and Measurement

5-6
 TableS Ground Condition
Test Surface Dry Void Wot Relative Water
Cuo LavaKm) density ratio density densfty Untfm)
(ft/cm1) Cl/cm") (W)
CD 5.B2 1.50 0.81 155 38.7 urface
(2) 5.65 1.59 0.71 2JOO 63 surface
(3) 554 1.62 0.68 22 71.7 surface
552 1.62 0.87 2.03 73J surface
5.39 1.69 0.64 2SJ5 83.3 surfac-1.4

fr=5.3ZHz

10 l5FreqCHz320

Fig.4 Spectral Ratio due to Random Wave(30gal)

_ j.i _
0 0.2 0.4 O.B 0.8 1 9 '
f '
- - -J

Fig.5 Vibration Mode 1

1
,i
1
-1
1
_T_
5
-4 _ i. _

O Pore pressure
Acc.tnnsducer
' Disp. transducer i*-i' I

- T ~
1
X Strain gauge
i 1

0 5 10 15 20
0 250 500 750 1000
ding Mon
Acc.(gtt) <POel>
<Pile-Bisc>

Fig.7 Max. Distribution ofPore Water Pressure,

Acceleration and BendingMoment (Case (D~

|- Croud fipQS

__-^
Gced
OBeC

&a -20
-600
10 15 feed 20 IS (sec) 20 15 (sec) 20

Rg.6 Time KGstray die to Earthquake Excitation (Case*

5-7
600r1 Ground AX17 ca<3 -Ground wp 5 r 20 Pile! PS1 r ' Cased)
'- Caie
- yV^KX^,
10

1M
/A
Cue Q 0

-600L
Cue
-10 V !
-20

i" 40 Base EE03 CfceCE

Cue

-6TO
<: -40
f\ _ /\
i?J ' v* x^

tt 5 10 15 (sec) 2

| Ground iptS

5 i " i_ i_ ^f *-
i i ____
4 ~ r ~ r ~
3 3 i i Cue
I3 >_ L _L _
i i
10 15 (sec) 20
2 - H -r-r-
-{Ground vplS
1
'
; i"
0
2-5 5 7.5 10 0 2 >0 500 750 10( 0 5 10 IS 20
ller preisuredf/n1) Ace. (gal) HoBEnl(tf-n)
<G-2> < ?ile'Base> l-Bend> 0 500 100D 1500 (sec) 2000
%9CornparisonofMax. ReqponseDistribufan Eg.lOPcre Water Pressure During Diss^otfon
\wfcGroundWaterLevd (Case. )

MUHt ratio Water prwsursOvf/cra*) Moment ratio Moment ratio

00 05 10 15 Q.O 02 0.4 ao oo os i jo ao (L2 a4 0.6 ox> as 1.0 ao 0.2 0.4 ae
8 e

70

Ffell Distribute ofBendi^

5-8
 WP02 -- - WPD4 0.5
T1(4.3Hz)[
WP06 - - - - WPOB
WP10
JS 0.4

0.3
.
v E

S 01
5

0.0
8 10 12 14 16
Time(m)

fig.12 Pore Water Pressure in Dissipation process after Hg.14 Resonant Curves of Footing
Earthquake Excitation Displacement

-1012345 -1.5 0-1234

1.5
Pore vate pressire(tF/nf) <G-1> Bending momertftnv'tonO Earth pressLre(tf/m!Aorf)

Fig.13 Distribution of Pore Fig.15 Distribution of Bending Hg.16 Distribution of Earth Pressure
Water Pressure at several stages Moment at several stages(Pile2) at several stages(Pilel)

ttnv'tonf) SLtpafe
-05 0 OS -10000 -SCO 0 5000 10000 15000
-Q001 0 0001 0.002 0X103 0004 01X6
-1

Rg.17 Pile Displacement, Subgrade Reaction, Coefficient of Subgrade Reaction obtained by Polynomial Approximation

5-9
 ENERGY DISSIPATION IN SOIL-STRUCTURE INTERACTION:
A CONSULTANT'S PERSPECTIVE

By: C.B. Crouse1

ABSTRACT: Energy dissipation as a means of reducing the seismic response of

structures has become a popular topic among researchers and structural engineers
who have developed and implemented devices, such as friction dampers, fluid
dampers, and isolators, in the retrofit of structures. However, a natural source of
energy dissipation is the interaction between a structure, its foundation, and the
supporting soil medium. This interaction can be significant and potentially
beneficial in certain situations, resulting in large reductions in seismic response.
Unfortunately, this phenomenon is often ignored by many structural engineers
because of their lack of knowledge of (1) theoretical principles governing soil-
structure interaction (SSI), (2) circumstances when SSI is potentially important, (3)
the field measurements of modal damping ratios, and (4) methods to estimate modal
damping ratios in soil-structure systems. This knowledge gap exists primarily
because the SSI subject is not generally taught at the undergraduate or graduate
levels in university civil engineering departments. The theoretical principles,
involving wave-propagation theory, boundary-value problems, and soil and
structural mechanics/dynamics, are daunting to most civil engineering students with
design-oriented career goals. Nevertheless, assuming the frequency-dependent
foundation impedance functions can be obtained for a particular SSI system from
the literature or from a consultant, relatively simple and practical systems-
identification methods can be used to estimate the composite modal damping ratios
for the significant modes of vibration. SSI experiments and theoretical calculations
using these simple models have yielded relatively large modal damping ratios in
certain situations for structures such as short-span bridges, offshore concrete gravity
platforms, nuclear power plant containments, fuel storage tanks, short to mid-rise
buildings, and nuclear waste processing plants.

1 Principal Engineer, Dames & Moore, 500 Market Place Tower, 2025 First Avenue, Seattle, Washington 98121
6-1
INTRODUCTION

This paper, which is intended primarily for the professional structural engineer engaged in seismic
design, first reviews the general state of SSI practice in the consulting engineering/structural design
professions from the standpoint of energy dissipation in SSI systems. In dynamic analysis of
structures, modal damping ratios between 0.03 and 0.07 (with a nominal value of 0.05) are typically
used hi practice. These values are generally acknowledged as structural damping ratios and are
usually valid for a rigid or nearly rigid-base model of the structure. However, these values are often
used in flexible-base models, which can result in an overestimation of the seismic loads.

In the section, Composite Modal Damping, technical justification for higher damping ratios for
flexible-base models is provided by reviewing the concept of composite modal damping for soil-
structure systems and the factors that affect it. Next, experimental data on modal damping of soil-
structure systems are reviewed in the section, Measurements of Composite Modal Damping. In the
subsequent section, Estimate of Composite Modal Damping, a simple systems identification
procedure is described for estimating modal damping ratios hi SSI systems with one or more
foundations and frequency-dependent foundation impedance functions. Details of an example
calculation of the modal damping ratios illustrating this procedure for a liquid fuel storage tank are
described at the end of this section. A discussion of the implications of energy dissipation in SSI
for seismic design is presented in the last section of this paper.

STATE OF PRACTICE

The state of the practice for modeling energy dissipation hi SSI analysis for structures other than
nuclear power plants is typically as follows. The structural engineer usually performs the analysis
and constructs a model of the structure from the element library hi a commercial software program
for structural dynamics. Most of these element libraries have rotational and translational springs of
constant stiffness that can be attached to the base of the structural model to simulate the foundation-
soil interaction. The structural engineer will usually consult a geotechnical engineer for these
foundation stiffnesses, but will not usually request estimates of the foundation damping. The
structural analysis software typically solves the equations of motion using modal superposition.

The input motion is usually defined by the geotechnical engineer or engineering seismologist and
consists of design spectra corresponding to damping ratios specified by the structural engineer, who
typically selects values in the aforementioned 0.03 - 0.07 range.

This practice of incorporating SSI effects has remained fairly constant during the last 20 years.
While more structural engineers recognize the importance of including foundation flexibility in
their models, there has been a reluctance to properly incorporate the damping into the system. Most
popular commercial structural dynamics software programs do not include viscous damping

6-2
004\MISC\ING.DOC
 elements. Even if the provision for specifying the element damping were available in these
programs, most engineers would have difficulty specifying the viscous damping constants for each
element because, unlike the relatively simple methods for generating the element stiffnesses, the
procedures for estimating the values of a system damping matrix, are more complex. Available
procedures are obscure, unknown, or difficult to understand by many structural engineers. Until
they become familiar and comfortable with these procedures, the usual practice of arbitrarily
adopting modal damping values around 0.05 will continue.

Many structural engineers are presently using software for nonlinear inelastic dynamic analysis of
structures. The hysteretic damping from the specification of the load-deflection characteristics of
the structural elements is automatically included in the structural models, but the damping for the
foundation elements is more difficult to specify because of the anelastic (hysteretic) damping of the
soil and the frequency-dependent radiation damping of the foundation-soil medium.

In the penultimate section of this paper, a simple example is provided illustrating the calculation of
modal damping ratios for the fundamental mode of vibration of a tank-foundation system. Even in
this example, the geotechnical engineer needed to recognize and properly estimate the two
components of foundation damping (hysteretic and radiation).

COMPOSITE MODAL DAMPING

The composite modal damping ratio for each mode of vibration of a soil-structure system depends
on the foundation damping, the structural damping, and the nature and degree of interaction
between the structure and supporting soil. The foundation damping consists of the material (or
hysteretic) damping of the soil and the radiation damping associated with the generation and
propagation of seismic waves into the soil medium by the motion of the foundation relative to the
free-field earthquake motion. The material damping primarily depends on strain induced in the soil
during the shaking (e.g., Seed and Idriss, 1970), whereas the radiation damping depends on the
elastic properties of the surrounding soil and the shape and embedment of the foundation. For a
given soil profile and foundation geometry, the radiation damping depends on the mode of
foundation vibration, e.g., vertical translation, horizontal translation, rotation, or a combination of
translation and rotation (Richart et al., 1970; Gazetas, 1983).

Generally, foundation damping is highest for vertical and horizontal vibration and lowest for
rocking motion. However, even for rocking modes, the damping ratios can be significantly larger
than the nominal 5% critical damping ratio typically assumed for structures such as buildings and
bridges. Thus, foundation damping is normally much higher than structural damping. In
qualitative terms, the composite modal damping for a given soil-structure system with given
amounts of foundation and structural damping will depend on the amount of deformation in the
structure relative to the foundation movement. For example, the composite modal damping for stiff

004\MISC\ING.DOC 6-3
structures on flexible soils is expected to be greater than the composite modal damping for flexible
structures on stiff soils. This conclusion is easily seen in the limiting cases of an infinitely -rigid
structure on a flexible soil or a flexible structure founded on hard bedrock (i.e., rigid-base
structure).

The next two sections discuss the measurement and prediction, respectively, of composite modal
damping.

MEASUREMENTS OF COMPOSITE MODAL DAMPING

Composite modal damping ratios have been measured in a variety of soil-structure systems
including simple footings, pile foundations, multistory buildings, bridges, and a scale-model
nuclear plant containment structure. These damping ratios were obtained from soil-structure
responses measured during forced vibration tests or earthquake excitations. The results for several
soil-foundation and soil-foundation-structural systems familiar to the author are presented below.
A comprehensive compilation of SSI parameters (including modal damping ratios) estimated from
earthquake responses recorded at 58 building sites is presented in Stewart and Stewart (1997).

Concrete Footings

Forced harmonic vibration tests were conducted on small rectangular footings, approximately
1.2m x 1.2m in plan dimensions with thicknesses ranging from 0.1 m to 0.6m. The footings, to
which strong-motion accelerographs were attached to measure earthquake ground motions, were
underlain by moderately stiff soil with shear-wave velocities on the order of 150 m/s. One test was
conducted in Jenkinsville, South Carolina (Crouse et al., 1984) and the others were conducted in
southern California (Crouse and Hushmand, 1989). The Jenkinsville footing, which is 0.6 m thick
and embedded approximately 0.45 m in the surrounding soil, is shown in Figure 1, and the southern
California footings are shown in Figure 2. The modal damping ratios for these footings are
summarized in Table 1 below.

TABLE 1. Measured Modal Damping Ratios

for 1.2m x 1.2m Concrete Footings

Location Damping Ratio

Jenkinsville, SC 0.20 - 0.21
Parkfield, CA 0.30-0.40
El Centre, CA 0.30-0.40

The vibration modes corresponding to these damping ratios were coupled translation-rocking.

004\MSC\ING.DOC
6-4
A similar vibration test was conducted in El Centre, California on a larger rectangular foundation
with dimensions of 2.44m x 2.44m x 0.10m thick (Grouse and Hushmand, 1989, 1990). This
foundation, also founded on soil with a shear-wave velocity of approximately 150 m/s, supported
an 2.44m high rigid masonry block structure (Figure 3). The modal damping ratio for this system
was approximately 0.29 (Grouse et al., 1992).

Scale-Model Concrete Containment Structure

In 1985 the Electric Power Research Institute (EPRI) and Taiwan Power Company (TPC)
constructed a 1/4 scale-model of a nuclear power plant containment structure (Figure 4) in Lotung,
Taiwan. The containment structure was a cylindrical reinforced concrete shell (outer radius =
5.25m) that was 15.2m high and attached to a 0.91m thick circular concrete mat. The average
shear-wave velocity measured in the upper 6.1m of soil was approximately 150 m/s. The Nuclear
Regulatory Commission (NRC) conducted forced harmonic vibration tests on the structure. The
instrumentation on the structure also recorded motions from several earthquakes.

The modal damping ratios measured during the forced vibration test and during the May 20, 1986
earthquake of magnitude 6.5 are listed in Table 2 (Tajimi Engineering Services, Ltd, 1989).

TABLE 2. Measured Modal Damping Ratios for

Concrete Containment Structure

Event Estimation Method Damping Ratio

Forced Vibration Test Bandwidth 0.10
Resonant Amplification 0.15-0.17
Energy 0.13-0.21

5/20/86 Earthquake System Identification 0.22

The mode of vibration yielding these damping ratios was primarily foundation rocking with a
relatively small component of sway deformation of the containment. The associated resonant
frequency was 3.8 Hz during the forced vibration test and 2.3 Hz during the earthquake. The lower
frequency and higher damping during the earthquake is attributed to greater nonlinear (hysteretic)
soil behavior.

Bridges

Vibration measurements on bridges during forced vibration tests and during a large earthquake
were used to estimate modal damping ratios for three bridge-foundation-soil systems. Analysis of
these data appears in a paper by Crouse and Werner (1995). Table 1 from that reference
(reproduced with minor edits as Table 3 below) lists the relevant information about the bridges,
Horsethief, Meloland Road Overcrossing (MRO), and Moses Lake, which are one, two, and three

004\MISC\ING.DOC
spans, respectively, and which are supported on moderately stiff soil at their abutments and pier
foundations.

TABLE 3. Characteristics of Bridges

Name and Length Type of Type of

Location (m) Construction Abutment Foundation Type Foundation Soils
Abutment Pier Abutment Pier
Horsethief 31.4 RC box girder Monolithic Footing - Stiff sand -
Corona, CA
MRO 63.4 RC box girder Monolithic Wood Piles Wood Piles Stiff clay Mod. stiff sand,
clay and sand
El Centra, CA
Moses Lake, WA 43.3 RC I girder Seat Footing Footing Dense silty Very stiff
sand sandy silt

The mode shapes, natural frequencies, and modal damping values were compiled from the test data
and the strong motion records. Table 4 contains the relevant modal data from each bridge-vibration
data set; only the data from modes with a significant transverse component are included in the
table. In these modes, the SSI was thought to be potentially significant. Listed in the third through
seventh columns of Table 4 are the natural frequencies, selected modal deflections, and modal
damping ratios. The modal deflections listed are the average transverse translation of the abutments
(}& )> me maximum transverse deflection of the deck (Rd \ and the maximum vertical deflection of
the deck (Rd ). For these data sets, Rd was primarily due to torsion of the deck about the
longitudinal axis. The transverse modal deflections at the pier foundations of the MRO and Moses
Lake bridges were not included in Table 4 because they were roughly an order of magnitude
smaller than the abutment deflections, ( Ja ).

A parameter, 9 , was defined to characterize the amount of SSI in each mode of vibration of the
bridge-foundation-soil system in terms of the modal deflection terms:
9= & m

The rationale for this expression is as follows. The experimental data for bridges reveal high
damping in transverse modes of vibration where the abutment transverse deflection is a significant
fraction of the transverse deck deflection. In these modes, the vertical deflection of the deck is
relatively small. Modes with vertical deflections that are much larger than the transverse
deflections of the abutment have relatively small damping. Longitudinal modes of vibration are
expected to have fairly large damping also, but these modes are difficult to excite and are not
considered to be as important as transverse modes from the standpoint of earthquake performance.
Thus, the parameter 9 does not incorporate longitudinal deflection terms.

6-6
004\MISC\ING.DOC
 If significant soil-structure interaction is defined as values of 6 > 0.1, then all modes satisfying this
criterion in Table 4 have modal damping ratios > 6%. The range of is between 6% and 26%.
For 0.05 < 6 < 0.1, modal damping ratios, > 6% were also observed for the second transverse
mode of the MRO. For 9 < 0.05, < 0.05 for the data in Table 4. The correlation between 6 and
is clearly seen in these data, which strongly suggest that damping ratios , > 0.05 are justified
for bridges in cases where the SSI is significant.

TABLE 4. Modal Damping Ratios for Transverse Modes of Vibration of Bridges

Bridge Test Case Transverse Natural Modal Deflections Modal

Mode No. Frequency SSI
Damping, Parameter
(Hz) ew e
z R'd
&

1. MRO-vib. test 1 3.3 0.11 0.31 0.18 6.2 0.22

(quick release - 2 12.7 0.029 0.096 0.30 8.8 0.073
21 kip load) 3 22.1 ~0 0.19 0.38 3.2 ~0

2. MRO-vib. test 1 3.2 0.094 0.33 0.14 6.5 0.2

(quick release - 2 13.2 0.02 0.10 0.32 11.6 0.05
141 kip load) 3 22.4 ~0 0.96 0.34 3.4 ~0

3. MRO-1979EQ 1 2.5 0.66 1.0 0.37 10-26 0.48

4. Horsethief 1 6.4 <0.01 0.03 0.27 3.5 ~0
(forced 2 8.2 0.8 1.25 0.39 15.0 0.49
harmonic
vibration)
5. Moses Lake 1 6.6 0.31 1.0 0.22 ' 9.1 0.25
(forced 2 7.4 0.26 1.0 0.45 6.8 0.18
harmonic
vibration- test 1)
6. Moses Lake 1 6.5 0.24 1.0 0.26 6.2 0.19
(forced 2 7.1 0.40 1.0 0.42 8.5 0.28
harmonic
vibration - test 2)

ESTIMATION OF COMPOSITE MODAL DAMPING

i
This section first describes a systems-identification method for the estimation of modal damping in
SSI systems where the foundation impedance functions are frequency and strain dependent. Next,
the method is illustrated in an example calculation of composite modal damping for a liquid natural
gas (LNG) tank.

6-7
004\MISC\ING.DOC
Systems-Identification Method

Systems-identification methods typically have been used to estimate system parameters (e.g.,
natural frequencies, damping ratios, and foundation stifrhesses) from response data recorded during
vibration tests or earthquakes. These methods also can be adapted to SSI analyses for structures in
the design stage in cases where the structures are supported on multiple foundations and/or where
the frequency dependence of the foundation impedance functions is significant. For such cases,
closed-form solutions for modal damping ratios do not presently exist, although solutions have
been developed for the case of single foundations supported on a medium in which the foundation
impedance functions (stiffness and damping coefficients) are frequency independent (e.g., Luco,
1981;ASCE, 1986).

The systems-identification method, presented herein, is applicable to linear SSI systems and is as
follows. Equations of motion are derived in the frequency domain for the SSI system in
generalized spatial coordinates and in modal coordinates. These two sets of equations are used to
develop transfer functions (one for generalized coordinates and another for modal coordinates) that
express the ratio of the motion at some point on the structure to the free-field motion as a function
of frequency. The transfer function of the model in generalized coordinates is presumed to be
known because the structural masses, moments of inertia, stifmesses and damping constants, as
well as the foundation impedance functions, are known or can be computed. On the other hand, the
natural frequencies and damping ratios in the transfer function for the SSI model based on modal
coordinates are unknown. The values of these parameters are varied until the two transfer functions
are similar.

In the generalized coordinate system, the specification of the structural damping is straightforward
for a 1 degree-of-freedom (d.o.f.) representation of the structure, which is an acceptable
approximation for a number of SSI systems. In this case, the damping coefficient, c, is computed
using the formula, c=2$<Jkm , for a 1 d.o.f. oscillator, where is the assumed modal damping ratio
for the oscillator (fixed-base structure), and k and m are the known oscillator stiffness and mass,
respectively. In cases where a multi-d.o.f. model is required for the structure, the structural
damping matrix can be computed by fairly simple matrix algebra (e.g., Tsai, 1969) by (1) assigning
modal damping ratios (Q) for each mode of vibration of the fixed-base model of the structure, and
(2) assuming these ratios are factors in the diagonal elements, 2 co/, of the diagonal damping
matrix of the modal equivalent of the fixed-base structure, which is assumed to possess classical
normal modes.

For many applications in practice, the foundation impedance functions can be obtained from the
literature (see Appendix A of WSDOT (1993)) or from commercially available computer codes,
e.g., DYNA3 (Novak et al, 1991), SUPELM (Kausel, 1992).

6-8
004\MISC\ING.DOC
 The impedance functions in the literature are usually presented in graphs or tables for assumed
values of the material damping of the soil. In the case of uniform material damping of the soil
medium, the published impedance functions can be easily adjusted to account for a different
material damping (Wong and Luco, 1981). This material damping ratio can be estimated from
published strain-dependent damping ratio curves or can be derived from laboratory tests on soil
samples. In either case, the effective shear strain in the soil must be estimated based on the design
ground motion.

Example Calculation of Composite Modal Damping for a LNG Tank

Dames & Moore recently participated in a project involving the seismic design of a large
cylindrical steel flat-bottom LNG tank in a region of moderate seismic activity. The tank was to be
supported on a mat foundation resting on the surface of improved soil. The structural designer was
planning to conduct a linear dynamic analysis using the design response spectra corresponding to
the appropriate modal damping ratios associated with the impulsive and convective modes of the
tank-fluid-foundation-soil system.

The composite modal damping of primary interest was for the fundamental impulsive mode. The
model used to compute this ratio was similar to the model of Veletsos and Tang (1990). It
consisted of a single-degree-of-freedom oscillator representing the impulsive mode of the fixed-
base tank attached to a circular foundation mat supported on a visco-elastic half space. The
relevant parameters of the model for the Operating Basis Earthquake (ORE) and Safe Shutdown
Earthquake (SSE) levels of shaking (0.2 g and 0.4 g, respectively) are summarized in Table 5.

TABLE 5. Model Parameters for LNG Tank Example

Model Element Parameter Parameter Value

QBE SSE
Tank Shell& Fluid Impulsive Liquid Mass 3.35x107 kg 3.35x107 kg
Impulsive Mode Frequency 3.9 Hz 3.9 Hz
Impulsive Mode Damping Ratio 0.03 0.07
Height of Impulsive Mass 11.9m 11.9m
Height of Liquid, H 31.8m 31.8m
Inner Tank Radius, a 40.0m 40.0m
Tank Foundation Radius 44.5m 44.5m
Mass 1.23x107 kg 1.23x107 kg
Mass Moment of Inertia 7.10x109 kg-m2 7.10x109 kg-m2
Half Space Shear-Wave Velocity 263 m/s 204 m/s
Density 1.76gm/cc 1.76gm/cc
Poisson's Ratio 0.3 0.3
Material Damping Ratio 0.10 0.15

004VMISCMNG.DOC
6-9
The modal damping ratio was estimated by deriving equations for the Transfer Function (TF)
between the tank-displacement motion and the free-field, ground-displacement motion induced by
the earthquake. One TF was expressed in terms of the generalized coordinates, mass, damping, and
stiffness quantities, while the other TF was expressed in terms of the modal quantities (i.e., natural
frequencies, modal damping ratios, mode shapes, and participation factors). The material damping
ratio of the soil was incorporated by modifying the foundation impedance functions for a circular
disc on an elastic half space without material damping. The procedure for the modification is given
in Gazetas (1983) or Wong and Luco (1981).

The modal damping ratio for the fundamental mode of the system that provided similarity in the
moduli of the two TFs was estimated. Using best estimates for the parameters of the tank-
foundation soil model (Table 5), modal damping ratios of 0.19 and 0.28 were estimated for the
ORE and SSE, respectively. Comparisons between the resulting two TFs of the OBE and SSE are
shown in Figures 5a and 5b, respectively.

The relatively high modal damping ratios estimated for the OBE and SSE are the result of the
significant interaction between the tank foundation and underlying soil. These results are consistent
with those in Veletsos and Tang (1990) in the sense that composite modal damping values much
higher than the structural damping value are expected for tanks with small liquid height to inner
tank radius ratios (H/a) on relatively flexible soils.

DISCUSSION

Experimental test results and predictions from theoretical models clearly demonstrate that relatively
large composite modal damping ratios are possible when SSI effects are significant. Furthermore,
when properly substantiated by appropriate SSI analysis procedures, the use of relatively large
composite damping values for the computation of seismic loads is accepted practice in the U.S.
nuclear power industry. All of this experimental/theoretical evidence and the nuclear industry
precedent suggest that in the case of other important structures, composite modal damping values
larger than the structural damping ratios should be considered in the calculation of the seismic
loads. However, the composite modal damping values determined by theoretical models are not
necessarily those that should be used in final design. The final modal damping values should
consider uncertainties associated with the SSI model and its parameters, relevant experimental data,
and the degree of conservatism desired for the design.

One potentially beneficial effect that was not considered in the SSI model for the tank example is
the filtering or reduction in high frequency ground motion by the passage of seismic waves across
the tank foundation. During the last 25 years, this effect has been observed in buildings with
foundation areas similar to that of the example tank. The high-frequency filtering phenomenon for
foundations of this size on soils of roughly the same stiffness has been observed for frequencies

004\MISCMNG.DOC 6-10
 greater than about 2 Hz (e.g., Yamahara, 1970; Grouse and Jennings, 1975; Newmark et al., 1977;
Fenves and Serino, 1992; Stewart and Stewart, 1997). The filtering, or kinematic interaction, is
caused by (1) incoming seismic waves at angles of incidence less than 90 (vertical propagation),
and (2) incoherence in the ground motion due to wave scattering from inhomogeneities in the local
geology and anelastic attenuation.

The size of the example tank foundation (~ 80 m diameter), the flexibility of the underlying soil, and the
estimated fundamental impulsive frequency of the tank-soil system (~ 3 }/2 Hz) were factors collectively
suggesting that filtering of ground motion at this frequency may occur, thus reducing the impulsive
seismic load on the tank. The observational data indicate that a modest reduction in the OBE and SSE
design spectra at short periods to account for this effect (which is equivalent to increasing the damping)
was appropriate. However, the inclusion of this effect in the development of site-specific design spectra
for particular structures should be coordinated with the structural engineer performing the dynamic
analysis and design. The basis and amount of any reduction in the design spectra due to kinematic
interaction should be well documented by the professional developing the design spectra.

APPENDIX. REFERENCES

ASCE, 1986. Seismic Analysis of Safety-Related Nuclear Structures and Commentary on Standard
for Seismic Analysis of Safety-Related Nuclear Structures: American Society of Civil
Engineers, ASCE Standard 4-86, September.

Grouse, C.B., and B. Hushmand, 1989. Soil-structure interaction at CDMG and USGS accelograph
stations: Bull. Seism. Soc. Am., Vol. 79, No. 1, February, p. 1-14.

Grouse, C.B. and B. Hushmand, 1990. Soil-structure interaction and nonlinear site response at the
Differential Array accelerograph station: Proceedings of 4th U.S. National Conf. on Earthq.
Eng., Palm Springs, CA, May, Vol. 3, p. 815-823.

Grouse, C.B. and S.D. Werner, 1995. Estimation of modal damping for bridges: Proceedings of
4th U.S. Conf. on Lifeline Earthq. Eng., San Francisco, CA, August, p. 408-415.

Grouse, C.B., and P.C. Jennings, 1975. Soil-structure interaction during the San Fernando
earthquake: Bull. Seism. Soc. Am., Vol. 65, No. 1, February, p. 13-36.

Grouse, C.B., E. Trahern, and T. Price, 1992. Nonlinear site response at the Differential Array
station during the 1979 Imperial Valley, California, earthquake: Report to U.S. Nuclear
Regulatory Commission and National Science Foundation, March.

Grouse, C.B., G.C. Liang, and G.R. Martin, 1984. Experimental study of soil-structure interaction
at an accelograph station: Bull. Seism. Soc. Am., Vol. 74, No. 5, October, p. 1995-2013.

004\MISC\ING.DOC 6-11
Fenves, G.L. and G. Serino, 1992. Evaluation of soil-structure interaction in buildings during
earthquakes: Calif. Strong Motion Instrumentation Program, CSMIP/92-01, June.

Gazetas, G., 1983. Analysis of machine foundation vibrations, state of art: J. Soil Dyn. and Earthq.
Eng.,Vol.2,No. l,p.2-42.

Kausel, E., 1992. SUPELM, Ver. 2.0, Foundations embedded in layered media: Dynamic
stiffnesses and response to seismic waves.

Luco, J.E., 1981. Linear soil-structure interaction: Lawrence Livermore Laboratory Report
(UCRL-53011) to U.S. Nuclear Regulatory Commission (NUREG/CR-1780).

Newmark, N.M., W.J. Hall, and J.R. Morgan, 1977. Comparison of building response and free-
field motion in earthquakes: Proceedings 6th World Conf. Earthq. Eng., New Delhi, India,
Vol. 3, p. 972-978.

Novak, M., M. Sheta, L. El-Himawy, H. El-Marsafawi, and O. Ramadan, 1991. DYNA3, A

Computer Program for Calculation of Foundation Response to Dynamic Loads, User
Manual: Univ. of Western Ontario Geotechnical Research Centre, Rept. No. GEOP90-02,
Vol. 1, Rev. 2

Richart, F.E., J.R. Hall and R.D. Woods, 1970. Vibrations of Soils and Foundations. Prentice-
Hall, Englewood Cliffs, NJ.

Seed, H.B. and I.M. Idriss, 1970. Soil moduli and damping factors for dynamic response analysis.
University of California, Berkeley, Earthquake Engineering Research Center, ERRC 70-10.

Stewart, J.P. and A.F. Stewart, 1997. Analysis of soil-structure interaction effects on building
response from earthquake strong motion recordings at 58 sites. University of California,
Berkeley, Earthquake Engineering Research Center, Report No. UBC/EERC-97/01,
February.

Tajimi Engineering Services, Ltd., 1989. Lotung SSI analysis using a 3-D harmonic Green's
function method: Proceedings EPRI/NRC/TPC Workshop on Seismic Soil-Structure
Interaction Analysis Techniques using Data from Lotung, Taiwan, Vol. 2, EPRI NP-6154,
March, p. 16-1 to 16-50.

Tsai, N-C, 1969. Influence of Local Geology on Earthquake Ground Motion: Ph.D. Thesis,
California Institute of Technology, Pasadena, CA, 201 p.

004\MISC\ING.DOC 6-12
Veletsos, A.S. and Y. Tang, 1990. Soil-structure interaction effects for laterally excited liquid
storage tanks: J. Earthq. Eng. and Struct. Dyn., Vol. 19, p. 473-496.

WSDOT (Washington State Department of Transportation), 1993. Manual of Practice,

Foundation-Soil Interaction Analysis of Bridges, Vol. 1, Report by Dames & Moore and
Inca Engineers.

Wong, H.L. and J.E. Luco, 1981. Identification of sensitive parameters for soil-structure
interaction: Lawrence Livermore Laboratory Report (UCRL-15493) to U.S. Nuclear
Regulatory Commission (NUREG/CR-3044).

Yamahara, H., 1970. Ground motions during earthquakes and the input loss of power to an
excitation of buildings: Soils and Foundations, Vol. 10, No. 2, Tokyo, Japan, p. 145-161.

6-13
004\MISC\ING.DOC
 HUT

ONE ACCELEROMETER .
LOCATION FOR hh
FIELD EXPERIMENTS

r-
VbJ
FOUNDATION SOIL

Fig 1. (a) Jenkinsville, South Carolina, accelerograph station, (b) Soil-structure interaction
model of station.

004\MISaiNG.DOC
6-14
 PLAN
CDMG USGS
Y CHOLAME IE & FAULT ZONE 3 EL CENTRO STATION 6

1.27
1.22

t.27
i: 1.14

-1
0.51
;:
1" 'I
T 0.1
0.10 "
0.10
r^
1 _!__
t"""?. avafuti
4-1
0.46-0.61

ELEV.

Fig. 2. Foundations of CDMG (Cholame IE and Fault Zone 3) and USGS (Station 6)
accelerograph stations. Numbered arrows indicate locations and directions of applied shaker
forces. Dimensions in meters.

i
0.15 _

2.44 MASONRY
BLOCK

0.71

2.44

Fig. 3. Differential Array station. Dimensions in meters.

6-15
 0.30 4.95

1.07

15.2

T
4.57

1
0.91

r
Fig. 4. Cross section of 1/4-scale containment model. Dimensions in meters.

6-16
Lft

H
Transfer Function Modulus Transfer Function Modulus
 EMPIRICAL ASSESSMENT OF SOIL-STRUCTURE INTERACTION
EFFECTS FROM STRONG MOTION RECORDINGS

Jonathan P. Stewart1

Abstract: System identification analyses are used to evaluate soil-structure interaction effects
for 77 strong motion data sets at 57 building sites which encompass a wide range of structural
and geotechnical conditions. Kinematic interaction effects on the "input" motion at the bases of
structures are found to be relatively modest in many cases, whereas inertial interaction effects on
the structural response to these motions can be significant. To quantify inertial interaction
effects, fixed- and flexible-base modal vibration parameters are used to evaluate first-mode
period lengthening ratios (T / T) and foundation damping factors (Q ) The response of some
structures is dominated by inertial interaction (e.g. T / T = 4, 0 ~ 30%), whereas others undergo
negligible SSI (e.g. T / T 1, 0 ~ 0). Simplified analytical formulations adapted from Veletsos
and Nair (1975) and Bielak (1975) are used to predict inertial interaction effects, and are found to
be reasonably accurate relative to empirical results, with some limitations for long-period
structures. A collective examination of the empirical and predicted results reveals a pronounced
influence of structure-to-soil stiffness ratio on inertial interaction, as well as secondary influences
from structure aspect ratio and foundation embedment, type, shape, and flexibility.

INTRODUCTION

Documentation of seismic case history data is a critically important step towards

understanding and reliably characterizing complex problems in geotechnical earthquake
engineering. Few empirical studies of soil-structure interaction (SSI) have been performed due
to the previously limited amount of strong motion data from sites with instrumented structures
and free-field accelerographs. In contrast, analytical formulations for SSI are numerous, ranging
from complex, three-dimensional finite element analysis procedures capable of incorporating
fully nonlinear dynamic soil modeling (e.g. Borja, 1992) to simplified substructure techniques
suitable for implementation in building codes (e.g. Veletsos and Nair, 1975). While some
sophisticated analytical models have been verified using recorded data from nuclear reactor
structures or scaled models thereof (e.g. Valera et al., 1977; Bechtel, 1991), empirical studies
incorporating a large number of building sites with strong motion recordings are lacking.
In this paper available earthquake strong motion data, much of which has only recently
become available, are analyzed to evaluate the effects of inertial interaction on structural

1 Asst. Prof., Civil and Env. Engrg. Dept., University of California, Los Angeles, CA 90095-1593

7-1
response for a range of geotechnical and structural conditions. The results are used to verify
simplified inertial interaction analysis procedures modified from Veletsos and Nair (1975) and
Bielak (1975). Kinematic interaction, which modifies foundation-level motions relative to free-
field motions, is a second order effect for many buildings, and is not the primary subject of this
paper. The paper is organized into separate sections describing the analysis procedures, database,
empirical data, and the calibration of the SSI design procedures. These results are a summary of
the findings in Stewart et al. (1998).

ANALYSIS METHODS

Overview of Design Procedure for SSI

For analysis of inertial interaction effects, the objectives are predictions of first-mode period
lengthening ratio T/T and foundation damping factor Q As shown in Fig. 1, simple
procedures for evaluating these effects employ a model consisting of a single degree-of-freedom
structure resting on a foundation-soil system represented by an impedance function. The
impedance function is calculated for a rigid disk foundation resting either at the surface of
(Veletsos and Nair, 1975) or embedded into (Bielak, 1975) a uniform visco-elastic halfspace.
As shown in Fig. 2, the motivation for characterizing T/T and 0 ls tnat tnev can be used
to estimate flexible-base modal parameters (T, ), which in turn are used in response spectrum-
based approaches for evaluating design-level seismic base shear forces and deformations in
structures. The parameters needed for analysis of T / T and 0 are:
Soil conditions: shear wave velocity Vs and hysteretic damping ratio P which are
representative of the site stratigraphy and the level of ground shaking; representative soil
Poisson's, ratio v.
Structure/Foundation Characteristics: effective height of structure above foundation level, h\
embedment, e\ and foundation radii which match the area and moment of inertia of the actual
foundation, ru and re.
Fixed Base 1st Mode Parameters: period and damping ratio, T and .

Using these data, the impedance function is evaluated at the flexible-base period of the structure,
T. The frequency dependent and complex-valued impedance terms are expressed in the form
kj =kj (a0> D)+i(DCj (a0> D) (1)
where y denotes either deformation mode u (translation) or 0 (rocking), CO is angular frequency
(radians/sec.), ao is a dimensionless frequency defined by ao = cor/Vs, r = foundation radius, Vs =

7-2
soil shear wave velocity, and i) = soil Poisson ratio. Terms kj and Cj consist of a combination of
static foundation stiffness (Kj) and dynamic modifiers otj and (3j as follows:

vs

The terms otj and (3j express the frequency dependence of the impedance, and are computed
differently for surface (Veletsos and Verbic, 1975) and embedded (Bielak, 1975) foundations.
Foundation radii are computed separately for translational and rotational deformation modes to
match the area (Af) and moment of inertia (If) of the actual foundation (i.e. ru = >/Af/7C, re =
4V4If/7C). The Bielak formulation includes a rigorous model of dynamic basement wall-soil
interaction, assuming perfect wall-soil bonding. An approximate analysis of embedment effects
can be made with the Veletsos and Nair model by increasing the static stiffness according to the
well known guidelines of Kausel (1974), and using otj and (3j terms for surface foundations
(Elsabee and Moray, 1977).
Stewart et al. (1998) outlined several considerations associated with the application of these
procedures to realistic foundation and soil conditions. These can be summarized as follows:
1 . Representation of nonlinear soil response and nonuniform soil profiles as a visco-elastic
halfspace. Strain dependent soil properties are evaluated with site response analyses which
are used to calculate an equivalent hysteretic damping ratio and a degraded shear wave
velocity profile. The effective profile velocity is taken as the profile depth divided by the
shear wave travel time through the degraded profile. Profile depth is taken as ru.
2. Representation of non-circular foundations. While noncircular foundations with aspect
ratios < 4: 1 can generally be represented as equivalent disks (Roesset, 1980), radiation
dashpot coefficients for rocking can be underestimated by such procedures (Dobry and
Gazetas, 1986). Correction factors can be adapted from the Dobry and Gazetas results.
3. Representation offlexible foundations. The impedance of flexible base mats with thin
perimeter walls or rigid concentric interior and perimeter walls can be reasonable well
represented by rigid foundation models (Liou and Huang, 1994; Riggs and Waas, 1985).
However, the rigid disk model is inadequate for buildings with rigid central cores, and should
be modified according to the results of Iguchi and Luco (1982).
The basic procedures for rigid disk foundations on or in halfspaces were modified according to
(1) to (3) above, and are subsequently referred to as the "modified Veletsos" (MV) and "modified
Bielak" (MB) formulations.

Use of System Identification to Evaluate SSI Effects

The objective of system identification analyses is to evaluate the unknown properties of a
system using a known input into, and output from, that system. For analyses of seismic structural

7-3
response, the "system" has an unknown flexibility that generates a known difference between
pairs of input and output strong motion recordings. For example, as indicated in Fig. 3,
parameters describing the fixed-base system are evaluated from input/output pairs that differ only
by the structural deformation (u). Likewise, parameters describing the flexible base system are
evaluated from strong motion pairs whose difference results from foundation flexibility in
translation (Uf) and rocking (0), as well as structural flexibility. A comparison of fixed- and
flexible-base modal parameters provides a direct quantification of SSI effects.
There are two principal system identification procedures:
1. Nonparametric procedures evaluate complex-valued transmissibility functions from the input
and output recordings without fitting an underlying model. These transmissibility functions
represent an estimate of the ratio of output to input motion in the frequency domain, and are
computed from smoothed power and cross-power spectral density functions of the input and
output motions. Modal frequencies and damping ratios are estimated from peaks in the
transmissibility function amplitude (Pandit, 1991; Ljung, 1987; Fenves and DesRoches,
1994).
2. Parametric procedures develop numerical models of transfer functions, which represent the
ratio of output to input motion in the Laplace domain. The amplitude of the transfer function
is a surface in the Laplace domain. The locations of peaks (or poles) on this surface can be
related to modal frequencies and damping ratios. Parameters describing transfer function
models are estimated by minimizing the error between the model output and recorded output
in the discrete time domain using least squared techniques. The transfer function surface can
be estimated by minimizing cumulative error for the entire time history (Safak, 1991), or by
recursively minimizing error for each time step using a window of time immediately
preceding that time step (Safak, 1988).
The evaluation of vibration frequencies and damping ratios from transmissibility functions
can be problematic (especially for damping), because the shape of the functions is dependent on
details associated with the computation of the spectral density functions such as the number of
points in the Fast Fourier Transform and the windowing procedures used (Pandit, 1991).
Parametric procedures provide a relatively rigorous modeling of system response, because the
transfer function for a given set of time histories is only dependent on two user-defined
parameters: the delay between the input and output and the number of modes used in the
analyses (i.e. the order of the model). When these parameters are selected judiciously, the modal
frequencies and damping ratios can be reliably evaluated for linear structures. Hence, parametric
identification techniques were used here for the evaluation of modal vibration parameters.
Further details on the identification procedures are provided in Stewart et al. (1998).

7-4
Evaluation of Modal Parameters for Various Base Fixity Conditions
Three cases of base fixity are of interest in analyses of SSI: (1) fixed-base, representing only
the flexibility of the structure, (2) flexible-base, representing the combined flexibility of the
complete soil-structure system, and (3) pseudo flexible-base, representing flexibility in the
structure and rocking in the foundation. Pseudo flexible-base parameters are of interest because
they can sometimes be used to approximate flexible-base parameters or to estimate either fixed"
or flexible-base parameters.
Stewart et al. (1998) evaluated the types of input and output strong motion recordings that
are necessary to evaluate fixed-, flexible- and pseudo flexible-base vibration parameters of
structures with parametric identification procedures. While roof translations are always used as
output, the input motions for various base fixity conditions vary as indicated in Fig. 3.
Recordings of free-field, foundation, and roof level translations, as well as base rocking, are
needed to evaluate directly both fixed- and flexible-base modal parameters of a structure.
Instrumented buildings often lack sensors for recording base rocking or free-field
translations. For such cases either fixed-base parameters (missing base rocking) or flexible-base
parameters (missing free-field translations) cannot be evaluated directly from system
identification analyses. Stewart et al. (1998) derived expressions to estimate either flexible- or
fixed-base parameters using "known" modal parameters for the two other cases of base fixity.
The estimation procedures operate on the premise that differences between known parameters
can be used to calibrate the foundation impedance at the structure's period; the calibrated
impedance can then be used to estimate the unknown parameters. These estimation procedures
extend significantly the number of sites for which SSI effects can be empirically evaluated.

DATABASE

Two classes of sites are used in this study: Class 'A* sites, which have a free-field
accelerograph and a structure instrumented to record base and roof translations (and in some
cases, base rocking as well), and Class 'B' sites, which have structures instrumented to record
base rocking as well as base and roof translations, but have no free-field accelerographs. This
section presents criteria employed for the selection of 'A' sites. The *B' sites are simply those
with the stated structural instrumentation.
Each 'A' site was reviewed for the following: (1) the free-field instrument is not so close to
the structure as to be significantly affected by structural vibrations, and (2) the free-field
instrument is not so far from the structure that free-field and foundation-level motions exhibit
significant incoherence.

7-5
The check for contamination of free-field motion by structural vibrations is made by examining
power spectral density and coherency functions for the free-field and foundation motions. High
coherencies between the two motions at modal frequencies, or spectral peaks in free-field
motions at modal frequencies, indicate potential contamination. Results for all sites considered
in this study can be found in Stewart and Stewart (1997). Significant contamination of free-field
data was only found at two sites, and in both cases arose from vibrations of structures other than
the subject structure that were near the free-field seismograph.
The incoherence between foundation-level and free-field motions is assumed to follow the
empirical models developed using data from the Lotung, Taiwan LSST array (Abrahamson et al.,
1991) and SMART1 array (Abrahamson, 1988). A minimum coherency of 0.8 was enforced,
yielding maximum free-field/structure separations of about 800 m for 1 Hz structures, 450 m for
2 Hz structures, and 150 m for 4 Hz structures.
Suitable free-field instruments were sought for virtually all instrumented structures in
California, and 44 sites were identified (plus one additional structure in Taiwan). An additional
13 structures in California were considered in this study as 'B' sites. The 'A' and 'B' sites are
listed in Table 1. For the 57 sites, 74 processed data sets are available as a result of multiple
earthquake recordings at 13 sites.
Fifteen California earthquakes contributed data to this study with magnitudes ranging from
4.8 to 7.3. Moderate to low level shaking (pga < 0.1 to 0.2g) is well represented in the database
(50 data sets), while a moderate amount of data (24 data sets) is available for more intense
shaking (pga > 0.2g).
The foundation conditions at the sites include 23 buildings with piles or piers, and 34 with
footings, mats, or grade beams. Most buildings are not embedded (36) or have shallow single-
level basements (14). Only seven buildings have multi-level basements. The buildings range
from single story warehouses to high-rise office buildings. Lateral force resisting systems
include shear walls, frames, and base isolation.

EMPIRICAL EVALUATION OF INTERACTION EFFECTS

Comparison of Free-Field and Foundation-Level Structural Motions

A simple investigation of kinematic and inertia! interaction effects can be made by
comparing indices of free-field and foundation motions. Shown in Fig. 4 for free-field and
foundation-level motions at 'A' sites are (a) peak horizontal accelerations and (b) 5%-damped
spectral accelerations at the flexible-base period of the structure ( T). The T values were
established from system identification analyses. Second order polynomials are fit to the data in
Fig. 4 using linear regression.

7-6
 The data in Fig. 4 indicate that peak foundation-level accelerations are de-amplified relative
to the free-field, especially in embedded structures. Earlier studies utilizing smaller databases
had similar findings (e.g. Poland et al., 1993). Conversely, spectral accelerations at T for
foundation motions are generally negligibly de-amplified for surface foundations (open circles in
Fig. 4), and only modestly de-amplified for embedded foundations (solid dots in Fig. 4). These
different de-amplification levels at different spectral periods can be attributed to frequency
dependent kinematic de-amplification effects which are maximized at low periods (i.e. T=0),
coupled with potential contributions of inertial interaction to foundation motions for periods near
T. As it is the spectral acceleration at T that best simply describes the ground motion
controlling structural response, for design purposes, there appears to be little useful ground
motion de-amplification on surface foundations relative to the free-field, and only modest de-
amplification on most embedded foundations (average reduction of 20%). However, as indicated
in Fig. 4(b), significant reductions on the order of 40% can occur in individual cases (typically
deeply embedded foundations).
Although significant further study is needed to more fully evaluate kinematic interaction
effects, the data in Fig. 4 suggest that for purposes of engineering design, free-field and
foundation-level motions are often comparable in amplitude. Hence, a more significant SSI
effect would appear to be the modification of structural response associated with the flexibility of
foundation support. These inertial interaction effects are examined in the remainder of this paper
through evaluations of period lengthening ratios and foundation damping factors.

Period Lengthening and Foundation Damping

System identification analyses were performed for the 57 sites considered in this study using
procedures outlined in Stewart et al. (1998). Modal vibration periods and damping ratios were
evaluated for the fixed-base (T,) and flexible-base (T, ) cases. These parameters are listed in
Table 1, along with the calculated period lengthening ratio T/T, foundation damping factor

0 = - u (T/Tj , and dimensionless structure-to-soil stiffness ratio 1/a = h/(Vs T) , where h

= effective structure height and Vs = effective soil shear wave velocity.
Each site was assigned a confidence level based on the quality of available geotechnical data
and the accuracy/uncertainty associated with the identification. These confidence levels are
indicated in Table 1, with "A" indicating acceptable confidence, "L" indicating low confidence,
and "U" indicating unacceptable confidence. Unacceptable confidence is associated with one of
the following situations:

7-7
 Ul: Reliable flexible-base parameters could not be developed due to significant incoherence
between foundation and free-field motions.
U2: The structure was so stiff that the roof and foundation level motions were essentially
identical, and hence the response could not be established by system identification.
U3: Fixed-base ('A' sites) or flexible-base ('B' sites) parameters could neither be estimated
nor obtained directly from system identification.
U4: Reliable parametric models of structural response could not be developed for unknown
reasons.
Presented in Fig. 5 are the variations of T / T and Q witn l/c fr sites wnere there is an
"acceptable" or "low" confidence level in the modal parameters. Also shown are second-order
polynomials fit to the acceptable confidence data by regression analysis, and analytical results by
Veletsos and Nair (1975) for h/r = 1 and 2. Both T/ T and Q are seen to increase with I/a, and
the best fit lines through the data are similar to the Veletsos and Nair curves.
There is significant scatter in the data in Fig. 5, although much of this results from
systematic variations in T / T and Q associated with factors such as structure aspect ratio,
embedment, foundation type, and foundation shape and flexibility effects. In addition, o is
influenced by the hysteretic soil damping (p), which varies with soil type.
Results from several sites help to illustrate the strong influence of I/a on inertia! interaction
effects. The most significant inertial interaction occurred at site A46 (T/ T 4 and
tJo ~ 30%), which has a stiff (T ~ 0.1 sec) cylindrical concrete structure (h=14.3 m, r= 4.9 m)
and relatively soft soils (Vs ~ 85 m/s), giving a large I/a of about 1.5. Conversely, the inertial
interaction effects are negligible at site A21(T/T1 and Q ~ 0%), which has a relatively
flexible (T 0.8-1.0 sec) base-isolated structure (h=6.7 m, ru=21.6 m) that is founded on rock
(Vs ~ 300 m/s), giving a much smaller I/a value of 0.02-0.03. These two sites represent the
extremes of inertial interaction. More typical SSI effects occur at sites B14 (T / T = 1.14 and
Co 3.4%) and Al-tr (T / T = 1.57 and 0 ~ 15.4% ). The structures at both sites are shear
wall buildings with periods of T = 0.49 and 0.15 sec, respectively, and are founded on medium-
stiff soils (Vs = 256 and 213 m/s), combining to give I/a 0.12 at B14 and I/a 0.29 at Al-tr.
The results from these four sites illustrate that both T/ T and 0 increase with increasing I/a.
To examine the influence of parameters other than I/a on SSI effects, the data in Fig. 5 were
sorted according to aspect ratio (h/re), foundation type (piles or piers vs. shallow foundations),
embedment ratio (e/r), and lateral force resisting system by Stewart et al. (1998). The trends

7-8
resulting from these regressions are relatively weak, as the influence of the respective parameters
could not be readily isolated from each other given the limited scope of the database.
Nonetheless, some dependence on aspect ratio was found, with larger period lengthening and
smaller damping for structures with h/re > 1 than for structures with h/re < 1. This is consistent
with trends from the analytical models. Well-defined trends in data sorted according to the other
parameters were not identified.

CALIBRATION OF SIMPLIFIED ANALYSIS PROCEDURES FOR INERTIAL

INTERACTION

Period lengthening ratios and foundation damping factors were evaluated by the Modified
Veletsos (MV) procedure for each data set in Table 1. These factors were also evaluated by the
Modified Bielak (MB) procedure for embedded structures.

Modified Veletsos (MV) Formulation

Deviations in MV predictions of T/T and Q relative to empirical values are shown in Fig.
6 for sites with acceptable and low confidence designations. Also plotted are best fit second-
order polynomials established from regression analyses on data from acceptable confidence sites.
For most sites, the predictions are accurate to within absolute errors of about 0.1 in TV T and
3% damping in 0 fr I/0 = 0 to 0.4. The regression curves indicate no significant systematic
bias in predictions of either T/T or 0 UP to l/a = 0-4- However, there is a downward trend in
the best fit curve for damping for I/a > 0.5 (beyond the range on Fig. 6) due to a significant
underprediction of J0 at site A46 (I/a = 1.5) which results from a pronounced embedment effect
at this site that is not fully captured by the MV formulation.
The results from several sites help illustrate the general findings of Fig. 6. The minimal
inertia! interaction effects at site A21 (I/a = 0.02 to 0.03, T / T 1 and 0 0% ) are well
predicted by the MV analyses, as is typical for sites with I/a < 0.1. The predictions are also
generally satisfactory for sites with intermediate I/a values such as B14 and Al-tr (I/a = 0.12,
I/a = 0.29). At these sites, period lengthenings of 1.14 and 1.57 are over- and under-predicted
by absolute differences of about 0.11 and 0.06, respectively, while foundation damping factors of
3.4 and 15.4% are underpredicted by absolute differences of 2.3 and 4.8%, respectively. The
large inertial interaction effects at site A46 (I/a = 1.5, T / T ~ 4.0 and Q ~ 30% ) are predicted
to within an absolute difference of about 0.4 for period lengthening, but damping is
underpredicted by an absolute difference of about 14%. With the exception of the damping

7-9
results at site A46 (where there is a significant embedment effect), these results indicate that
predictions of T / T and 0 by tne MV procedure are reasonably good considering the breadth
of conditions represented in the database.
There are several noteworthy outliers in Fig. 6. When the residuals in Fig. 6 are considered
with respect to the magnitude of the observed SSI effect, the most significant outliers for period
lengthening are seen to be site A34 and several long period structures (A4, B3). The unusual
results at site A34 may be associated with erroneously high shear wave velocity measurements
(Stewart et al., 1998). The long period structures at sites A4 and B3 are founded on soft Bay
Mud soils in the San Francisco Bay Area, and were subject to negligible period lengthening (a
common system identification result for all long-period structures). The soft soils at sites A4 and
B3 lead to overpredictions of period lengthening, suggesting an error in the model. It appears
from these data that the simple single-degree-of-freedom models on which the MV and MB
formulations are based are incapable of adequately modeling SSI effects in long period structures
with significant higher mode responses.

Effect of Embedment: Comparison of "Modified Veletsos" and "Modified Bielak"

Predictions
Plotted in Fig. 7 are deviations between analytical and empirical results for three data sets,
(1) MV predictions for buildings with surface foundations, (2) MV predictions for buildings with
embedded foundations, and (3) MB predictions for buildings with embedded foundations. As
before, the best fit curves are second-order polynomials established from regression analyses.
The regression curves in Fig. 7 indicate that T / T is slightly over-predicted for embedded
structures (by either MV or MB), and more accurately predicted for surface structures. The
differences between MV and MB predictions are generally minor (e.g. absolute differences of
about 0.02 at.A20-tr, 0.02 at A23) for typical values of I/a (i.e. < 0.4). At site A46 (I/a = 1.5),
the absolute difference between the predictions is about 1.2, which is modest compared to the
empirical value of T / T ~ 4.0.
The accuracy of o predictions by the MV methodology are comparable for surface and
embedded structures. However, there are disparities between the MB and MV 0 predictions
for embedded structures which increase with I/a (e.g. absolute differences of 0.7% at A23, I/a =
0.11; 2.7% at A20-tr, I/a = 0.17; 10% at A46, I/a = 1.5). The regression curves are primarily
controlled by the shallowly embedded foundations (e/r < 0.5), which are the most numerous in
the database. For such cases, MV predictions are typically more accurate than MB predictions,
as shown by the regression curves in Fig. 7, and as illustrated by sites A20 (e/r = 0.27) and A26

7-10
(e/r = 0.41). However, there are systematic errors in MV predictions for more deeply embedded
foundations. These errors are not surprising because only the MB formulation incorporates
dynamic basement wall/soil interaction effects into the foundation impedance function. As
shown by individual labeled sites in Fig. 7, MV predictions of 0 are generally too low for
relatively deeply embedded structures with continuous basement walls around the building
perimeter such as A46 (e/r = 0.92) as well as A9, B12, and A16-L (e/r > 0.5). Other structures in
the database with e/r > 0.5 had negligible foundation damping (i.e. Q < 1%) which was
overestimated by both MV and MB predictions (i.e. A16-tr and B13). Hence, it appears that MB
predictions of 0 are generally more accurate than MV predictions for structures with e/r > 0.5
and significant SSI effects. These differences are most pronounced at site A46, where the MB
and MV predictions of 0 =27 and 17% can be compared to the empirical value of 30%.
In summary, the accuracy of period lengthening predictions by the MV methodology are
reasonably good for surface and shallowly embedded structures, and differences between the MV
and MB predictions are generally minor for I/a values of common engineering interest (I/a <
0.4). Accuracies of MV damping predictions are generally acceptable for surface and shallowly
embedded structures (e/r < 0.5). For deeper embedment (e/r > 0.5), MB damping predictions are
generally more accurate. These results suggest that the dynamic basement-wall/soil interaction
modeled by the MB procedure can be important for deeply embedded foundations.

Other Effects
The adequacy of the MV/MB analysis procedures to capture the influence of factors such as
structural aspect ratio, foundation type, structure type, foundation shape, and foundation
flexibility were investigated by Stewart et al. (1998). The influence of aspect ratio and structure
type were adequately captured by the analyses.
Foundation type was found to have a minor influence on prediction accuracy, indicating that
the shallow disk foundation models provided reasonable evaluations of SSI for many structures
with pile or caisson foundations. However, many of the deep foundation sites for which this
trend was established have fairly stiff surficial soils and no marked increase in stiffness across
the depth of the foundation elements. For such cases, it is reasonable that dynamic foundation
performance would be strongly influenced by the interaction of surface foundation elements (e.g.
pile caps, base mats, footings) with soil. Conversely, for several sites with pile foundations and
soft soils, the shallow foundation models were found to underpredict foundation damping.
Foundation shape effects were generally found to be minor for the structures in the database.
That is, the exclusion of correction factors by Dobry and Gazetas (1986) for shape effects did not
adversely effect the accuracy of damping predictions. Foundation flexibility effects were found

7-11
to be significant for the only structure in the database with a stiff central core of shear walls and a
integral connection between the foundation for the walls the foundation for the remainder of the
structure. For this site (B2), corrections to the impedance function adapted from the results of
Iguchi and Luco (1982) substantially improved predictions of period lengthening.

CONCLUSIONS AND RECOMMENDATIONS

Summary of Findings
Available strong motion data suggests that foundation-level and free-field spectral
accelerations at the period of principal interest in structural design (i.e. the first-mode flexible-
base period, T ) are similar for structures with surface foundations, and that foundation-level
spectral accelerations are generally only modestly de-amplified (averaging about 20%) for
embedded foundations. Since the free-field and foundation level ground motions therefore
appear to be comparable, this study has focused principally on evaluating the effects of inertial
interaction on structural response.
Inertial interaction effects for buildings are expressed in terms of the lengthening of first-
mode period (T/ T) and the damping associated with soil-foundation interaction ( Q )
Simplified analytical procedures for predicting T/T and Q include Modified Veletsos (MV)
and Modified Bielak (MB) approaches that can be adapted for a wide range of conditions.
Based on the database of 57 sites compiled for this study, the factor with the greatest
influence on T / T and o Is me rati f structure-to-soil stiffness as quantified by the parameter
I/a = h/(Vs-T). When I/a is nearly zero, T/T and Q values are about unity and zero,
respectively, whereas at the maximum observed value of I/a = 1.5 at site A46, interaction effects
dominated the structural response (T/T 4 and Q ~ 30%). Additional factors which can
significantly affect inertial interaction include the structure's aspect ratio (h/re) and foundation
embedment and flexibility. For the majority of sites in the database, other factors such as the
type of structural lateral force resisting system as well as foundation type and shape, were found
to have a relatively small influence on SSL

Recommendations
Inertial SSI effects can be expressed by a period lengthening ratio ( T / T ) and foundation
damping factor (Q ) These factors are used to estimate flexible-base fundamental-mode
parameters, which in turn are used in response spectrum based approaches for evaluating base

7-12
shear forces and deformations in structures (e.g. Fig. 2). A key finding of this research is that
these inertial interaction effects can generally be reliably predicted by the MV analysis procedure.
However, several caveats apply to this basic recommendation:
1. Inertial interaction effects were generally observed to be small for I/a < 0.1 (i.e. T / T < 1.1
and Q < 4%), and for practical purposes could be neglected in such cases.
2. For structures with embedment ratios greater than 0.5, the MB methodology should be used
in lieu of MV to appropriately model the extra radiation damping contributed by dynamic
soil/basement-wall interaction.
3. Damping results for pile supported structures on soft foundation soils (Vs < 500 fps) should
be interpreted with caution, as the damping is likely to exceed the values predicted from
simplified analyses (which assume shallow foundations) due to soil-pile interaction effects.
4. Period lengthening for long-period (T > 2 sec.) structures with significant higher-mode
responses is negligible and can be neglected.
5. Corrections to rocking damping values for foundation shape effects are generally small and
can be neglected without introducing significant errors.

ACKNOWLEDGMENTS

Support for this project was provided by the U.S. Geological Survey, Department of the
Interior (USGS award number 1434-HQ-97-GR-02995) and the National Science Foundation.
The views and conclusions contained in this document are those of the authors and should not be
interpreted as necessarily representing the official policies, either expressed or implied, of the
U.S. Government.

REFERENCES

Abrahamson, N.A. (1988). "Empirical models of spatial coherency of strong ground motion,"
Proc. 2nd Workshop on Strong Motion Arrays, Taipei, Taiwan, 65-92.
Abrahamson, N.A., Schneider, J.F., and Stepp, J.C. (1991). "Empirical spatial coherency
functions for application to soil-structure interaction analyses," Eq. Spectra, EERI, 7(1), 1-27
Bechtel Power Corporation (1991). "A synthesis of predictions and correlation studies of the
Lotung soil-structure interaction experiment," Rpt. No. EPRINP-7307-M, EPRI, October.
Bielak, J. (1975). "Dynamic behavior of structures with embedded foundations," /. Eq. Engrg.
Struct. Dynamics, 3(3), 259-274.
Borja, R.I., Smith, H.A., Wu, W-H., and Amies, A.P. (1992). "A methodology for nonlinear
soil-structure interaction effects using time-domain analysis techniques," Rpt. No. 101,
Blume Earthquake Engrg. Ctr., Stanford Univ., June.

7-13
Building Seismic Safety Council, BSSC (1995). "NEHRP Recommended provisions for seismic
regulations for new buildings, Part 1, Provisions and Part 2, Commentary" Rpt. No. FEMA
222A, Federal Emergency Management Agency, Washington D.C.
Dobry, R. and Gazetas, G (1986). "Dynamic response of arbitrarily shaped foundations," /.
Geotech. Engrg., ASCE, 112(2), 109-135.
Elsabee, F. and Morray, J.P. (1977). "Dynamic behavior of embedded foundations," Rpt. No.
R77-33, Dept. of Civil Engrg., Massachusetts Inst. Technology.
Fenves, G.L. and DesRoches, R. (1994). "Response of the northwest connector in the Landers
and Big Bear Earthquakes," Rpt No. UCB/EERC-94/12, Earthquake Engrg Research Ctr.,
Univ. of California.
Iguchi, M. and Luco, J.E. (1982). "Vibration of flexible plate on viscoelastic medium," /. Engrg.
Mech., ASCE, 108(6), 1103-1120.
Kausel, E. (1974). "Forced vibrations of circular foundations on layered media," Rpt. No. R74-
11, Dept. of Civil Engrg., Massachusetts Inst. Technology.
Liou, G.-S. and Huang, P.-H. (1994). "Effect of flexibility on impedance functions for circular
foundations," /. Engrg. Mech., ASCE, 120(7), 1429-1446.
Ljung, L. (1995). System Identification Toolbox, Users Guide, The Math Works, Inc.
Pandit, S.M. (1991). Modal and Spectrum Analysis, John Wiley, New York, NY.
Poland, C.D., Soulages, J.R., Sun, J., and Mejia, L.H. (1994). "Quantifying the effect of soil-
structure interaction for use in building design," SMIP-93, Seminar on Seism. & Engrg.
Aspects of Recent Strong Motion Data, CDMG, 43-54.
Roesset, J.M. (1980). "A review of soil-structure interaction," in Soil-structure interaction: the
status of current analysis methods and research, J.J. Johnson, ed., Rpt. No. NUREG/CR-
1780 and UCRL-53011, U.S. Nuclear Regulatory Com. and Lawrence Livermore Lab.
Riggs, H.R. and Waas, G. (1985). "Influence of foundation flexibility on soil-structure
interaction," /. Eq. Engrg. Struct. Dynamics, 13(5), 597-615.
Safak, E. (1988). "Analysis of recordings in structural engineering: adaptive filtering,
prediction, and control," Open File Rpt. 88-647, U.S. Geological Survey, October.
Safak, E. (1991). "Identification of linear structures using discrete-time filters," /. Struct.
Engrg., ASCE, 117(10), 3064-3085.
Stewart, J.P., Seed, R.B., and Fenves, G.L. (1998). "Empirical evaluation of soil-structure
interaction effects," Rpt. No. PEER-98/??, Pacific Earthquake Engineering Research Center,
Richmond, CA (in press).
Stewart, J.P. and Stewart, A.F. (1997). "Analysis of soil-structure interaction effects on building
response from earthquake strong motion recordings at 58 sites," Rpt. No. UCB/EERC 97/01,
Earthquake Engineering Research Ctr., Univ. of California.
Valera, J.E., Seed, H.B., Tsai, C.F., and Lysmer, J. (1977). "Soil structure interaction effects at
the Humboldt Bay Power Plant in the Ferndale Earthquake of June 7, 1975," Rpt. No.
UCB/EERC 77/02, Earthquake Engrg. Research Ctr., Univ. of California, Berkeley.
Veletsos, A.S. and Nair V.V. (1975). "Seismic interaction of structures on hysteretic
foundations," /. Struct. Engrg., ASCE 101(1), 109-129.
Veletsos, A.S. and Verbic, B. (1973). "Vibration of viscoelastic foundations," /. Eq. Engrg.
Struct. Dynamics, 2(1), 87-102.

7-14
 Table 1: Compilation of first-mode parameters for 'A' and 'B1 sites
Transverse Longitudinal

1 Station Eqk.

$
g
.c
E
03
1
2
%) ru re
(ft.) (ft.)
T T
(sec.) (sec.)
?
(%)
5
(%)
J[
O
T/T !o fu
(ft.)
re T T
(ft.) (sec.) (sec.)
?
(%)
C
(%)
j_
O
T/T Eo o
O
'A' Sites
1 Eureka Apts. PT 0.18 31 0 701 5.1 57 42 0.24 0.15 19.6 15.9 0.29 1.57 15.4 57 77 0.25 0.22 12.8 5.5 0.20 1.09 8.6 A
2 Fortuna Market PT 0.12 22 0 772 5.6 115 1120.37 0.36 34.0 26.0 0.08 1.04 11.2 115 122 0.31 0.29 39.4 18.0 0.10 1.08 25.1 L
PTA 0.19 786 5.8 0.35 0.34 17.2 15.0 0.08 1.05 4.1 0.29 0.28 25.6 17.6 0.10 1.03 9.6 L
4 Emeryville PPP LP 0.25 218 0 448 6.8 87 94 2.50 2.45 13.0 7.4 0.20 1.02 6.1 87 94 2.58 2.65 12.6 5.9 0.18 1.00 8.9 A
5 Hayward City Hall LP 0.05 84 0 2210 0.8 66 54 1.16 1.11 3.6 3.5 0.03 1.04 0.7 66 82 0.87 0.85 3.2 4.2 0.04 1.03 0.0 A
6 Hayward 13-St. LP 0.09 141 0 63 64 U3 U3 U3 U3 U3 U3 U3 63 64 U3 U3 U3 U3 U3 U3 U3
7 Hollister1-St. LP 0.36 30 0 502 7.7 97 75 0.73 0.71 26.9 19.0 0.08 1.03 9.3 97 130 U2 U2 U2 U2 U2 U2 U2 A
8 Piedmont Jr. HS LP 0.08 25 0 1820 1.3 52 46 0.18 0.16 4.8 2.2 0.09 1.16 3.4 52 59 0.17 0.17 7.0 5.1 0.08 1.00 2.0 A
9 PVPP CGA 0.21 48 32 970 3.2 53 37 0.53 0.53 8.2 5.2 0.09 1.00 3.0 53 77 U2 U2 U2 U2 U2 U2 U2 L
10 Richmond CH LP 0.13 33 10 768 3.7 75 54 0.30 0.28 9.2 3.9 0.16 1.08 6.1 75 106 0.27 0.26 14.4 1.4 0.16 1.03 13.1 A
11 San Jose 3-St. LP 0.27 35 0 2642 0.9 86 68 0.64 0.67 20.6 24.8 0.02 1.00 0.0 86 111 0.80 0.66 20.3 23.6 0.02 1.00 0.0 A
12 El Centra Bldg. IMP 0.24 54 0 484 6.6 61 55 0.74 0.50 16.0 23.4 0.23 1.47 8.8 61 69 1.23 1.25 33.9 36.6 0.09 1.00 0.0 A
13 Indio 4-St. LD 0.09 56 15.5 695 4.1 69 58 0.71 0.67 10.3 7.5 0.12 1.05 3.8 69 85 0.66 0.64 10.9 10.9 0.13 1.03 1.0 A
14 Lancaster 3-St. WT 0.07 26 0 908 2.1 54 47 0.20 0.20 13.4 12.4 0.14 1.00 1.1 54 63 U2 U2 U2 U2 U2 U2 U2 A
15 Lancaster 5-St. NR 0.07 40 0 1001 2.5 99 83 0.73 0.69 10.0 7.5 0.06 1.06 3.6 99 121 0.72 0.71 8.2 9.1 0.06 1.02 0.0 A
16 Lancaster Airfield NR 0.08 45 6 953 1.7 11.9 11.9 0.34 0.27 9.9 24.5 0.18 1.28 0.0 11.9 11.9 0.33 0.24 8.9 13.6 0.20 1.34 3.3 L
17 Loma Linda VA NR 0.06 50 0 1415 3.4 246 248 0.29 0.25 15.0 5.8 0.14 1.17 11.3 246 250 0.32 0.29 10.1 5.6 0.12 1.09 5.7 A
18 Long Beach 7-St. WT 0.07 58 0 615 3.5 49 46 U3 U3 U3 U3 U3 U3 U3 49 53 1.12 1.14 5.5 6.5 0.08 1.00 0.0 A
20 Long Beach VA NR 0.07 98 23 1143 2.9 84 83 0.58 0.51 3.1 4.6 0.17 1.13 0.0 84 87 0.58 0.55 4.5 4.1 0.16 1.05 1.0 A
21 LA 2-St. FCCB SM 0.11 22 0 1006 1.3 71 59 0.79 0.79 9.4 13.3 0.03 1.01 0.0 71 88 0.79 0.82 21.5 17.7 0.03 1.00 3.9 L
LD 0.05 1010 1.2 0.97 0.98 14.2 14.9 0.02 1.00 0.0 0.90 0.90 18.4 18.9 0.02 1.00 0.0 L
NR 0.32 981 1.7 0.92 0.95 34.4 40.1 0.02 1.00 0.0 0.83 0.84 33.4 39.5 0.03 1.00 0.0 L
22 LA 3-St. Bldg. NR 0.28 46 22.5 980 5.9 130 128 U1 U1 U1 U1 U1 U1 U1 130 134 U1 U1 U1 U1 U1 U1 U1 L
23 LA 6-St. Bldg. NR 0.25 56 14 630 5.2 21.4 20.6 0.89 0.82 5.5 6.9 0.11 1.08 0.0 21.4 22.7 U3 U3 U3 U3 U3 U3 U3 A
24 LA 6-St. Garage NR 870 (tr) 7.3 (ir)
0.22 40 0 640 (L) 6.2 (L) 159 154 0.52 0.51 6.6 6.1 0.09 1.04 1.1 54 47 0.44 0.28 6.1 6.5 0.23 1.60 4.5 A
25 LA 7-St. Hos. LD 0.04 68 0 1148 1.2 110 110 1.14 1.18 11.6 12.1 0.05 1.00 0.0 110 110 1.09 1.11 10.1 13.1 0.05 1.00 0.0 A
NR 0.49 1065 2.5 1.19 1.27 27.1 29.3 0.05 1.00 0.0 1.21 1.19 21.1 28.0 0.05 1.01 0.0 A
26 LA 7-St. Bldg. NR 0.47 66 13.5 548 7.6 33 30 0.66 0.63 9.2 16.1 0.19 1.04 0.0 33 37 1.04 1.09 7.0 11.5 0.11 1.00 0.0 L
27 LA15-SI. Bldg. LD 0.03 174 0 1161 1.2 131 110 3.15 3.20 3.9 3.0 0.05 1.00 1.0 131 161 3.09 3.10 2.1 1.8 0.05 1.00 0.3 A
NR 0.19 1120 2.0 3.12 3.20 8.5 2.8 0.05 1.00 5.6 3.07 3.09 8.8 2.0 0.05 1.00 6.8 A
28 LA19-St. Bldg^ NR 0.28 220 38 980 5.9 92 76 3.24 3.45 U1 U1 0.07 1.00 U1 92 113 3.72 3.89 U1 U1 0.06 1.00 U1 A
29 LA Hollywood SB WT 0.21 96 9 930 2.8 59 42 1.80 1.77 9.1 5.4 0.06 1.01 3.9 59 86 U4 U4 U4 U4 U4 U4 U4 L
NR 0.39 879 4.4 2.10 2.05 18.3 15.4 0.05 1.02 3.9 0.80 0.75 8.0 8.5 0.15 1.06 0.9 L
30 LA Wadsworth NR 0.25 78 17 981 6.3 189 199 1.00 0.92 9.3 9.3 0.09 1.08 1.9 189 199 U3 U3 U3 U3 U3 U3 U3 A
31 Newport Beach LD 0.04 94 0 1009 2.3 61 52 0.84 0.70 4.7 2.9 0.13 1.19 3.0 61 74 A
NR 0.11 969 3.2 0.86 0.75 3.2 8.5 0.13 1.16 0.0 0.77 0.67 3.9 4.4 0.14 1.14 0.9 A
32 Norwalk 12400 WT 0.23 70 13.5 730 6.8 93 83 U3 U3 U3 U3 U3 U3 U3 93 106 1.48 1.54 9.6 8.5 0.06 1.00 1.1 A
 33 Norwalk 12440 WT 0.23 72 15 794 7.3 142 1051.32 1.32 2.0 1.8 0.07 1.00 0.2 142 195 1.20 1.22 3.5 3.0 0.07 1.00 0.5 A
NR 0.08 906 4.0 1.28 1.30 6.3 4.9 0.06 1.00 1.4 1.20 1.22 4.6 3.4 0.07 1.00 1.2 A
34 Palmdale 4-St. NR 0.08 24 0 1575 1.7 69 49 0.20 0.12 18.5 24.1 0.12 1.66 13.1 69 100 0.20 0.16 12.4 4.9 0.09 1.22 9.7 A
35 Pomona 2-St. WT 0.06 28 10.5 1246 1.6 59 57 0.26 0.25 8.7 5.5 0.09 1.02 3.6 59 63 0.27 0.26 5.8 8.6 0.09 1.02 0.0 A
UP 0.21 1178 3.2 0.29 0.29 9.2 4.9 0.08 1.01 4.4 0.30 0.30 11.2 12.1 0.08 1.00 0.0 A
36 Pomona 6-St. UP 0.21 53 12.5 1148 3.1 50 43 U1 U1 U1 U1 U1 U1 U1 50 59 U1 U1 U1 U1 U1 U1 U1 L
LD 0.07 1188 2.2 1.26 1.07 9.3 13.4 0.04 1.17 1.0 1.20 0.87 9.8 9.5 0.05 1.39 6.2 L
37 Rancho Cue. LJC RD 0.04 56 14 1172 1.4 120 87 0.60 0.59 4.3 3.7 0.08 1.03 0.9 120 170 0.60 0.60 5.6 4.2 0.08 1.00 1.4 A
WT 0.06 1157 1.7 0.65 0.63 4.4 5.0 0.08 1.02 0.0 0.66 0.65 8.4 6.1 0.07 .01 2.5 A
UP 0.24 1060 4.5 0.76 0.77 4.7 7.8 0.07 1.00 0.0 0.75 0.77 6.0 6.7 0.07 .00 0.0 A
LD 0.11 1039 4.4 0.87 0.85 11.2 12.5 0.06 1.01 0.0 0.89 0.87 17.2 17.4 0.06 .02 0.9 A
NR 0.07 1114 2.8 0.76 0.75 4.6 6.9 0.07 .02 0.0 0.81 0.79 7.5 8.9 0.06 .02 0.0 A
38 San Bern. 3-St. LD 0.09 29 0 883 3.8 78 77 0.56 0.52 6.2 7.2 0.06 .07 0.4 78 80 0.57 0.55 7.6 10.9 0.06 .03 0.0 L
39 San Bern. 5-St. NR 0.07 52 13 1233 2.6 95 86 0.63 0.65 9.1 5.2 0.06 .00 3.9 95 107 0.50 0.51 7.5 6.2 0.08 .00 1.3 L
40 San Bern. 9-St. LD 0.10 74 0 848 3.4 55 52 2.01 2.01 5.0 6.8 0.04 .00 0.0 55 59 2.05 2.08 7.4 6.0 0.04 .00 1.3 A
41 San Bern. CQC NR 0.04 38 0 1011 2.4 114 114 0.51 0.51 2.6 3.4 0.07 .01 0.0 114 114 0.93 0.91 4.5 4.0 0.04 1.02 0.7 A
42 Santa Susana NR 0.28 91 0 4460 1.0 23.5 20.5 0.54 0.53 19.1 15.1 0.04 1.02 4.9 23.5 27.6 0.54 0.55 11.4 8.3 0.04 1.00 3.1 L
43 Seal Beach 8-St. LD 0.05 83 16 933 2.6 103 90 1.28 1.26 11.5 13.2 0.07 1.02 0.0 103 121 1.16 1.12 13.7 16.2 0.08 1.04 0.0 A
NR 0.08 911 3.3 1.21 1.18 6.6 7.1 0.08 1.03 0.0 1.09 1.06 9.1 8.7 0.09 1.03 1.1 A
44 Sylmar Hos. WT 0.05 63 0 1506 1.9 126 132 0.30 0.27 9.1 9.9 0.16 1.10 1.7 126 132 0.29 0.25 9.5 8.6 0.17 1.17 4.1 A
NR 0.84 1119 9.1 0.38 0.37 18.9 19.4 0.15 1.04 1.7 0.34 0.26 23.4 17.5 0.21 1.29 15.2 A
45 Venture 12-St. NR 0.06 69 0 886 2.8 62 47 0.71 0.53 4.0 5.7 0.15 1.34 1.6 62 83 A
46 Lotung Reactor L07 0.11 47 15 275 9.7 16.3 18.3 0.49 0.12 30.6 3.0 1.45 4.14 30.6 16.3 16.3 0.45 0.11 31.0 3.0 1.54 4.01 31.0 A
'B1 Sites
1 Milpitas 2-St. LP 0.14 26 0 649 4.1 6.4 10.9 0.25 0.24 15.3 21.9 0.17 1.06 0.0 A
2 San Bruno 9-St. LP 0.11 66 0 916 3.6 72 59 1.10 0.97 12.5 11.4 0.07 1.13 4.7 A
3 San Fran. 47-St. LP 0.16 414 27 478 4.3 86 80 5.16 5.03 U4 U4 0.17 1.03 U4 A
4 San Fran. Trans. LP 0.12 475 51 801 5.9 98 99 U3 U3 U3 U3 U3 U3 U3 A
5 San Jose 10-St. LP 0.12 61 0 768 4.1 65 49 0.48 0.29 6.7 18.6 0.27 1.64 2.5 A
6 San Jose 13-St. LP 0.10 109 13.5 725 3.5 83 84 2.16 2.13 1.0 1.3 0.07 1.01 0.0 83 84 2.19 2.17 2.8 2.2 0.07 1.01 0.6 A
7 Walnut Crk 10-St LP 0.10 89 14 1405 1.2 32 26 0.77 0.66 6.6 13.3 0.10 1.17 0.0 A
9 El Segundo 14-St. NR 0.13 114 0 899 4.1 69 64 U3 U3 U3 U3 U3 U3 U3 A
10 LA 9-St. NR 0.16 89 13 878 3.7 50 38 1.25 1.25 10.3 7.8 0.08 1.00 2.5 65 67 1.06 1.04 11.5 13.0 0.10 1.03 0.0 A
11 LA17-SI. LD 0.04 91 0 1190 0.9 76 35.2 0.96 0.85 4.1 3.4 0.09 1.13 1.7 A
NR 0.26 1140 1.4 1.05 0.90 4.9 3.9 0.09 1.17 2.5 A
12 LA 32-St. NR 0.11 300 54 1339 1.2 93 91 1.94 1.84 7.6 7.1 0.12 1.06 1.7 93 97 U3 U3 U3 U3 U3 U3 U3 A
13 LA 54-St. NR 0.14 414 57.5 1317 3.4 96 87 5.81 5.70 7.7 15.0 0.06 1.02 0.0 A
14 Whittier 8-St. NR 0.17 48 0 842 3.2 64 28.5 0.56 0.49 12.1 12.9 0.12 1.14 3.4 A
h= effective structure height 0.7 full height U1-U4 denote unacceptably low confidence results that are unreported Lateral force resisting systems:
e= embedment depth A, L denote acceptable and low confidence in results, respectively SW = shear wall, masonry or cone.
Vs= soil shear wave velocity blank entries = insuffiencient strong motion data to evaluate modal parameters DWF = dual wall/frame system
P = soil hysteretic damping ru i"e =foundation radii matching area and moment of inertia of actual foundation CF = concrete frame
Vs and (3 evaluated from soii profiles in Stewart (1997) = h/(v,.T) SF = steel frame
Earthquakes: CGA=Coalinga Aftershock, IMP=lmperial Valley, LD=l_anders, LP=Loma Prieta, L07=l_otung Event 7, NR=Northridge, Bl = base isolated
PT=Petrolia, PTA=Petrolia Aftershock, RD=Redlands, SM=Sierra Madre, UP=Upland, WT=Whittier
 uf h0 u

A\\\ A\\\

Fig. 1: Simplified model for analysis of inertial interaction

0.8
spectral accelerations @
spectral accelerations @

.Spectra drawn for foundation , = Flexible-base period, damping ratio

input motion, which is identical (includes SSI effects)
to free-field motion if kinematic
interaction is ignored , = Fixed-base period, damping ratio
(neglects SSI effects)
CO

0.0
0 1
T (sec)
Fig. 2: Schematic showing effects of period lengthening and foundation damping
on design spectral acceleration using smoothed spectral shape. Sa can
increase or decrease due to SSI.

7-17
 u +uf + H6 + U

Input Output
Flexible-Base ug ug + uf +H9 + U
Pseudo Flexible-Base u g + Uf ug + uf + H9 + u
Fixed-Base u +u, +H9 ug + uf +H9 + u

Fig. 3: Motions used as inputs and outputs for system identification

of structures

0.8 o 0.8
4) Embedded Foundation (E) Embedded Foundation (E)
a*
O Surface Foundation (S) O Surface Foundation (S).
O
10.6
TO

(b)
0.0
0.2 - 0.4 0.6 0.8 0.0 0.2 0.4 0.6 0.8

Peak Acceleration, Free-Field (g) Spectral Ace. @ t, Free-Field (g)

Fig. 4: Comparison of free-field and foundation-level structural motions:

(a) peak acceleration data, and (b) 5%-damped spectral accelerations at T

7-18
 Acceptable Conf.
A46 y<

O Low Conf. . (1 *X \A2-L(pt) (1.5A 360%)^ ~

1.8 20

***' B5* A1-tr

A44^. (nr) *
1.6 * '/ ' ~
A1-tr 4 & Best fit, acceptable
Veletsos and Nair, ^/ / A34-tr confidence sites \
h/r = 1to2 ,?/
T
T
A12-tr*

/ /
//

/
/
W*) A2-trJpt) A1^-tr /

X-''
1.4 o ' / /
10
i A1|-L /// * * ^,''/
H
VD / / / X
A16-tr A44-L(rtr) ( .*' /' X ,.'
/ ,' * x^ -\ Veletsos and Nair,
o 4 x / \^h/r=1-2,P = 10% .
* / / /^ Best fit, acceptable
1.2 o ;* A1^V-M4-'"' confidence sites * x^ / X - -"""
** ^-L(wt)-''' * ..-^"""

* * / .-''.- '
*0 .^ X * '' .--"' B5
*** / .. .-
V^':::: V. *
1.0 0
0 0 "^21 0.1 0.2 0.3 04 0 0 X A21 0.1 0.2 0.3 0.
1/o = h/(Vs *T) 1/c = h/(Vs *T)

Fig. 5: Period lengthening ratio and foundation damping factor for sites sorted by confidence level,
and analytical results from Veletsos and Nair (1975). (tr=transverse, L=longitudinal direction)
 0.5 15
Acceptable conf.
Low conf.
10
Best fit, acceptable
confidence sites

* * AI^L A24-L B5

ih- A21
I 0.0
I Q.
to o" A46
o o *< o *
O 00 (1.5,-14%)
A4*
-5
A46
(1.5,0&-0.4) A34-tr

24-L
-10

-0.5 A34-tr . -15

0.0 0.1 * 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4
= h/(Vs *T) h/(Vs *T)

Fig. 6: Errors in "modified Veletsos" formulation for sites sorted by confidence level
(tr=transverse, L=longitudinal direction)
0.5 15
+ e = 0 (mod. Veletsos)
0> --- e>0(mod.Veletsos)
Q ....... e > o (mod. Bielak) 10
A46: Bie.~
A46: Bie.
(1.5, 0.8 & 1.2)

A46: Vel
(1.5.0&-0.4; B13
0.0
Q.
o"

-5

Best fit curves -

-10

A2-L(pt)

-0.5 -15
0, 0 0.1 * 0.2 0.3 04 0.0 0.1 0.2 0.3 0.4
h/(Vs *T) h/(Vs *T)

Rg. 7: Errors in "modified Veletsos" and "modified Bielak" formulations for surface and embedded structures
 DYNAMIC RESPONSE OF SOIL-PILE-BUILDING INTERACTION
SYSTEM IN LARGE STRAIN LEVELS OF SOILS
By Shin'ichiro Tamori1, Masanori liba2 and Yoshikazu Kitagawa3
ABSTRACT: A series of shaking-table tests of a scaled soil-pile-building model were performed
in order to study the effects of the plastic deformation of soil on dynamic characteristics of the soil-
pile-building interaction system. Results showed the natural frequency and amplification factor
decreased by 40% and 60%, respectively, when shear strain of soil was 4.2x10"2. Dynamic response
analyses, which combined the Sway-Rocking model and an equivalent linearization method, were
done. The maximum acceleration of the building was underestimated when amplitude of input
motion was 600 cm/s2, because the amplification factor of the rocking motion were overestimated
in this case. This facts was caused by underestimation of the damping effects for the rocking
motion of the foundation.

INTRODUCTION The initial shear modulus, G, (strain being 1.0 x 10"5),

shear modulus at large strain levels, Gs, and damping factors,
When designing a building, it is important to evaluate hg, were obtained by tri-axial compression tests in which
earthquake performance of a building including non-linear ambient stress were kept at 1.0 kg/cm2 and exciting frequency
soil-building interaction effects during an earthquake. Many was l.OHz. The shear modulus and damping factor of the
method(Novak and Sheta 1980; Darbe and Wolf 1988; plastic soil material, Plasticine, has strain dependency similar
Motosaka et. al.1992, etc.) have been proposed to evaluate to those of actual clayey soils.
the effects, but they are too complex for a practical design
process. In the practical designing of a building, analytical o PLASTICINE
methods should be simple so that , for example, an ---- Seed-ldriss
equivalent linearization method, like SHAKE(Schnabel et. al. 1.2 - - - Ishihara
1972), have been used frequently to evaluate ground response. - Nishigaki
1 ------Hara
But, in the case of the non-linear soil-building interaction
Tay lo r- P arto n
system, the accuracy of the method had not been tested 0.8 Anderson
enough.
In this study, a series of shaking table tests were done in
0
order to evaluate the effect of plastic deformation of soils on
0.4
dynamic characteristics of soil-pile-building interaction
system. Dynamic response analyses, which combined Sway- 0.2
Rocking model and an equivalent linearization method, of the
tests were also done to evaluate the accuracy of this analytical
method. 1.E-06 1-05 1.E-04 1.E-03 1.E-02 1.E-01
Shear Strain

PLASTIC MATERIAL FOR GROUND MODEL (a) Shear modulus ratios versus shear strain

Plastic material for the artificial ground model used in this 30

study was made of Plasticine and oil. Plasticine , being a O PLASTICINE
mixture of calcium-carbonate and oil, has been used as a - - - Seed-ldriss
25
- - - Ishihara
model material for plastic deformation processing of steel, 3?
since it has restoring force curves similar to high- r 20 - -.-- -Hara
- " " Taylot Parton
temperature steel( Cook 1953). I
JS 15
Fig. 1 shows the soil characteristics, strain-shear modulus
and strain-damping factor relationships for actual clayey
soils and Plasticine, which is the plastic soil material used
in this shaking table tests.
0
1 Associate Prof. , Department of Architecture and Civil
Engineering., Faculty of Engineering., Shinshu University, 1 .E-06 1 .E-05 1.E-04 1.E-03 1 .E-02 1 .E-01
Nagano, Japan. Shear Strain
2 Head, Geotechnical Div., Structural Department., Building
Research Institute, Ministry of Construction, Tsukuba, Japan. (b) Damping factor versus shear strain
3 Professor, Department of Structural Engineering., Cluster IV
Faculty of Engineering., Hiroshima University, Higashi Fig.1 Soil characteristics
Hiroshima, Japan.

8-1
 OUTLINE OF SHAKING TABLE TESTS

The similarity which proposed by Buckingham was used in

modeling the building and the ground soils. The scale factors
calculated from this formula are summarized in Table 1.
This similarity is applicable to non-linear soil dynamics
when the soil model material has a shear modulus-strain and a
damping factor-strain relations similar to those of the
prototype(Kagawa 1987). Under these conditions the ratio of
shear forces in the model and the prototype were kept ^ -
approximately equal to that of the damping forces for wide
strain levels of soil.
Fig. 2 shows a outline of the building and the ground model
together with the location of the measurement apparatus.
Two dwelling units of 11-story buildings were modeled in the
transverse direction. Table 2 shows the natural frequency and
damping factor of the building model. The building model was
made of steel weight and it's columns were made of steel
plates. The building foundation was made of aluminum and
BT3 B H6 BH. BT. CH. SH. FH: Acceleroneter
acryl plates. Four cylinder-shaped( 0 38mm, length is 487mm) (Horizontal)
M_ BV:Acceleroneter(Vertical)
pile models were made of steel plate attached with rubber, BHS -j i_ 11 : Bending Kouent
they were set at the corners of the foundation. P: Axial Force D:Displaceaent
BH4 _j 1.000
The ground model has a block shape and its size is
2xl.46xO.6m. Stainless plates were set at both side ends in BH3 ^
400 . 300 300

transverse direction of the ground to prevent vertical motion DIBH2 _j

of the ground. The central part(0 800mm, depth is 387mm) of
BV1.^~
I r P
BV2 SHS /
fi FH3
the ground model was made from Plasticine and oil. The jg\ T|
remaining portions of the model were composed of 5 1
H227 ci 4
polyacrylamid and bentnite, and remained elastic throughout POLYACRYLAMAID ^13 SHJ ir: LT3

nz - 3
the tests. Table 3 shows characteristics of the ground.
Damping factors were obtained by a free torsional vibration
4
BENTONITE -- 1
'- CJ2 i ^PI3
FH2
o u?
CD

cr>
LT3
test and shear wave velocity was obtained by the P-S wave \ / P23
_A_
""' ~^~ FHI
propagation tests. PLASTICINE/ Cfil UNIT(aa)
Two. earthquake records in which the time length was
corrected according to the similarity were used for the input Fig.2 Building and ground model
ground motion: 1968 Hachinohe EW and 1940 El Centro NS.
Maximum acceleration of the input motions were set as 100,
300 and 600 cm/s2 on the shaking table.
Table 2 Characteristics of building model

Characteristics
Foundation Building of fixed base building
Natural Damping
Table 1 Similitude ratios Size Weight Height Weight
Factor
Freq.
(cm) 0*9 (cm) (kg9
(Hz) (%)
Item Ratio(Model/Prototype)
30
Soil Density kgf/cm3 1/7, 1 X 6.79 78.7 28.4 18.8 0.22
Length cm l/^. 1/40 30
Acceleration cm/s2 1 1
Displacement cm 1/^L 1/40
Mass kgf.sYcm I/ 7? ^. 1/6.4X10"
Shear Modulus kgf/cm2 I/ ?7 ^. 1/40 Table 3 Characteristics of ground model
Frequency 1/s V ^ 6.325
Velocity cm/s j~/ V ^- 1/6.325 Upper layer
Lower layer
Stress kgtfs2 1/1? ^t 1/40 Item (GL~GL-45cm)
(GL-45~60cm)
Strain 1 1 Center Edge
Vs(m/s) 23.7 18.4 36.0
Damping factor(%)* 6.63* 5.57 6.05
Densiry(gtfcm3) 1.57 1.17 1.41
*Strain level is 3.6 XlO"

8-2
 Nogami (1977) were employed.
RESULTS OF THE TESTS (2) Dynamic stiffness and damping factor of piles for vertical
motion were calculated by the D.G.C.(Kobori et. al. 1970)
Fig. 3 shows first natural frequency estimated by spectral of the rigid plate, which has same cross section of the pile.
ratios of BH6/SH5 (see Fig. 2) . The shear strain shown in Vertical stiffness of a end bearing pile, which was
Fig. 3 is maximum strain that calculated from displacement at obtained by a wave propagation theory(Nogami and
BH1, CH3 and CH4. Novak 1976), was significantly larger than those obtained
Fig. 4 shows the amplitude of the spectral ratio at the by the D.G.C., so that we neglected the stiffness of the end
natural frequency. The natural frequency was decreased by bearing pile.
40% and the amplitude of spectral ratio was 60% at most (3) Dynamic stiffness and damping factor of the bottom of the
when the shear strain of soil was 4.20xlO"2. foundation was calculated by the D.G.C.
In this study, the dynamic stiffness of the soil-pile-
foundation system was calculated by the sum of the dynamic
stiffness of the piles and that of the bottom of the foundation.
The equivalent linearization method was employed in order
-rrr - A El Centre NS to consider plastic deformation of soils. Soil stiffness,
damping factor and strain of soil were determined as follows:
---i-- Hachinohe EW
O
A
to
OO
C?\
IO
O
J 1 -
Shear modulus, Gs, and damping factor, hg, of the soil were
Natfrequenrcayl(Hz) - _ _ H _ - ~

IT - + -i -t t- determined by the tri-axial compression tests according to the

i i i i
----,--
1

~ ~r i
- T - __ i. "!*iir r
following equation modified by the Hardin-Drnevich
model(Hardin and Drnevich 1972):
1 1 1 11
--- -- ~ -1t -t1 ^1 1*-
i
i
- - - -t - -I - - -f -
1.01
-

(1)
1
1
1
i iii i G, 1+ 0.96 (r, / 0.002072 ) K25S
l.OE-03 l.OE-02 l.OE-01 hg = 0.035 + 0.145(1- G5 /Gr ) (2)
Shear strain Where G, is the initial shear modulus and 7 s is shear strain
of the soil.
Strain of the soil caused by wave propagation, 7 wave was
estimated by displacement at SH3, SH4 and SH5. Soil strain
Fig. 3 Natural frequency versus shear strain of soil caused by the foundation was estimated from the maximum
relative displacement of the foundation, ubjna]t, as follows:
Displacement of soil at depth z, ub(z), was assumed to be
determined by eq. (3).
O A El Centre NS
t t
_' , -- Hachinohe EW
".()- g';/r-"..m
Bl + z I y/bc
(3)
A
BK where Bl is a constant, b and c are the width of half the
S3 - B" - r foundation in the vibration and transverse direction,
respectively.
h > 4- - t-

This formula was proposed by Kobori et. al.(1972)

0
By averaging the strain from z = 0 to H, considering energy
l.OE-03 l.OE-02 l.OE-01 caused by the displacement, Ut,(s), the equivalent maximum
Shear strain strain of the soil caused by displacement of the foundation,
7 base, becomes,

Fig. 4 Amplification factors versus shear strain of soil (4)

dz
where H = b.

THEORETICAL MODEL Equivalent shear strain of the soil, 7 eq> which has
determined soil stiffness and damping factor, is
The theoretical model employed in this study was the Sway-
Rocking(S-R) model, and an equivalent linearization method r eq=o.7ori.o( 7,^+ r wave) (5)
was used for dynamic response analyses. Dynamic stiffness
and the damping factor for sway and rocking motion were
calculated as follows:
(1) Dynamic stiffness and damping factor of piles for
horizontal and rocking motion proposed by Novak and

8-3
 plastic soil material to investigate the soil-pile-building
RESULTS OF ANALYSES interaction system in large strain levels of soils. Dynamic
analyses of the test, which incorporated Novak's and Kobori's
Figs. 5, 6 and 7 show spectral ratios , where input motion method and an equivalent linearization method were used to
was Hachinohe EW. In these figures, UR is rocking motion determine dynamic stiffness of foundation and piles.
at the top of the building and UH is relative deformation of the Results of the analyses were as follows;
building. Ratio of the equivalent strain divided by maximum (1) Ratio of equivalent strain divided by maximum strain was
strain was set to 0.7 in this case(see. eq.(5).) Fig. 8a shows the set from 0.7 to 1.0. It must be changed by the order of the
first natural frequencies detected form spectral ratio of tests or character of input motion. Difference in natural
BH6/SH5 (see Fig. 2) and 8b shows amplification factors, frequency obtained by the analyses were within 20% and
which are the amplitude of the spectral ratio at the first natural those of maximum acceleration were within 30%.
frequency. Fig. 8c shows maximum acceleration and 8d shows (2) Transfer functions for the rocking motion at the natural
maximum shear strain of the soil beneath the foundation. frequency was overestimated when maximum acceleration
As shown in Figs.8a~8d, difference in natural frequencies of input motion was 100 cm/s2 and those were
by the test and by the analyses were within about 20%. For underestimated when maximum acceleration of input
amplification factor and maximum acceleration, the difference motion was 600 cm/s2. In conclusion, the method used to
became 30%. When the maximum input acceleration was 100 evaluate the damping effects of the rocking motion should
cm/s2, the amplification factor was overestimated(see Fig. 5a be reconsidered in order to improve the accuracy of the
also). Figs. 5a ~5d show that transfer function of rocking analyses.
motion was overestimated. When the maximum input motion
became 600 cm/s2 , the amplification factor and maximum REFERENCES
acceleration at BH6 were underestimated by the analysis.
From Fig. 7b , this fact was caused by underestimation of the Cook, P. M. (1953). "Forging research - use of Plasticine
amplification factor of the rocking motion. models." Metal Treatment and Drop Forging 20: 534.
So when maximum acceleration of the input motion was Darbre, G. R. and Wolf, J.P. (1988). "Criterion of stability and
100 cm/s2, the damping effects of the rocking motion were implementation issues of hybrid frequency-time-domain
underestimated and when the maximum acceleration was 600 procedure for non-linear dynamic analysis," Earthquake
cm/sz , the effects were overestimated. Engineering and Structural Dynamics, 16, 569-581.
Figs. 9a~9c show comparisons of maximum acceleration Hardin, B. O. and Drnevich, V. P. (1972). "Shear modulus and
at the model building by the tests and analyses. When damping in soils, Design equation and curves," Journal of
maximum acceleration of input motion was 600 cm/s2, the Soil Mech. and Foundation, Div. ASCE, 98(SM7), 667-692.
maximum acceleration at the upper part of the building was Kagawa, T. (1987). "On the similitude in model vibration tests
underestimated by the analysis. on earthquakes." Proc. Of Japan Society of Civil
Figs. 10~13 show results, where input motion was El Engineering, 275: 69-77
Centro NS. In this case, ratio of the.equivalent strain divided Kobori, T. and Suzuki, T. (1970). "Foundation vibration on a
by maximum strain was set to 1.0. We have done tests by viscoelastic multi-layered medium," Proc. of the 3rd
using Hachinohe EW first, and then by using El Centro NS. Japanese Earthquake Engineering Symposium: 493-500.
So, this ratio may be affected by the order of the test. As Kobori, T., Soji, Y, Minai, R., Suzuki, T. and Yuwakasaki, Y.
shown in Figs. 3 and 4, the maximum shear strains beneath the T. (1972). "Effects of soil and geological conditions on
foundation for Hachinohe EW is as twice as those for El structural response in Osaka area," Proc. of International
Centro NS. So, In the case of El Centro NS, the soil had Conf. on Microzonation for Safer Construction, Research
already experienced strain level lager than that occurred and Application, 719-734.
during the test. Motosaka, M., Wolf, J. P. and Nagano, M. (1991) "Application
As shown in Figs. 13a and 13b, difference of natural of recursive evaluation of convolution integral in nonlinear
frequency, according to the tests and the analyses were within dynamics based on transfer function of linear system in
10% and, for amplification factors, the differences were within frequency domain,", Proc. of IV-ICCCBE, 243-250.
25%. Results of maximum acceleration of BH6 and of Nogami, T. and Novak, M. (1996). "Soil-pile interaction in
maximum shear strain by the analyses were in agreement with vertical vibration," Earthquake Engineering and Structural
those of the test. The amplification factor was overestimated Dynamics, 4, 263-281.
when maximum input motion was 100 cm/sz(see Fig. lOa) and Novak, M. and Nogami, T. (1977). "Soil-pile interaction in
was underestimated when maximum input motion was 600 horizontal vibration," Earthquake Engineering and
cm/s2 (see Fig. 12a). This was caused by the difference of Structural Dynamics, 5, 263-281.
damping effects of rocking motion by the analyses and those Novak, M. and Sheta, M. (1980). "Approximate approach to
by the tests also in this case (see Figs. lOb and 12b). contact effects of piles," Proc. of Session on Dyn. Response
Fig. 14 shows a comparison of maximum acceleration of of Pile Foundation: Analytical Aspects, ASCE National
the building by the tests and analyses. In this case, results by Conv., 53-79.
the tests and the analyses agree well. Schabel P. B., Laysmer, J. and Seed H. B.(1972). "A computer
program of earthquake response analysis of horizontally
CONCLUSION layered sites," Report No. EERC 72-12, College of
Engineering, University of California Berkeley.
This study involved performing shaking table tests on elasto-

8-4
 10 20 30 40

Hz
(b) UR/SH5 (b) UR/SH5 (b) UR/SH5

3
2.5
2
1.5
1
0.5
0
10 20 30 40

Hz
(c)BH1/SH5 (c) BH1/SH5
(c) BH1/SH5

10 20 30 40
Hz
(d) UH/SH5 (d) UH/SH5 (d) UH/SH5

Fig. 5 Spectral ratios Fig. 6 Spectral ratios Fig. 7 Spectral ratios

(Hachinohe 1968 EW, 10Ocm/s/s) (Hachinohe 1968 EW, 300cm/s/s) (Hachinohe 1968 EW, 600cm/s/s)

8-5
 ' \
10 _ _ j _ _ _ _ j. _ _ -- i~ 10 ,_j__L_j-___y__

. _ _i _ _ ~ ~ ^ ' f .J_-^_^-X-;--
S
CO

15
8

6
i
"A" - - \- - -
8
6 ~i--:*i -t--
i i ix i
< 4 7 i 4 ---I->T-T----T--
i ' '^^ ' '
2 2 . _ jr- ^._j____i__
r I i ' ' ' '
n y i i ,'11111
n
0 2 4 6 8 10 12 0 2 4 6 8 10 12
Test Test
(a) Natural frequency(Hz) (b) Amplification factor

u.uo x'
______
* /
S 4000 g) U.Ut
X
en ^ (A
> "<5 X
CO

< onnn - - -y
'

: n no
A
r
S
n n if

2000 4000 6000 0 0.02 0.04 0.06

Test Test
(c) Max. ace. at (d) Max. strain of sojl
BH6(cm/s/s)

Fig.8 Comparison results of tests and analyses (Hachinohe EW)

12

10

I 6
w
4

0
0 500 1000 1500 0 2000 4000 0 2500 5000
Max. acc.(cm/s/s) max. acc.(cm/s/s) Max. acc.(cm/s/s)

(a) 100cm/s/s (b) 300cm/s/s (c) 600cm/s/s

Fig. 9 Maximum acceleration at building (Hachinohe EW)

8-6
 12 test
10 analysis

10 20 30 40 10 20 30 40
Hz Hz
(a) BH6/SH5 (a) BH6/SH5 (a) BH6/SH5

10 20 30 40 10 20 30 40
Hz Hz
(b) UR/SH5 (b) UR/SH5 (b) UR/SH5

3
2.5
2
1.5
1
0.5
0
10 20 30 40 10 20 30 40 10 20 30 40
Hz Hz
Hz
(c) BH1/SH5 (c) BH1/SH5
(c) BH1/SH5

10 10 20 30 40
Hz
(d) UH/SH5 (d) UH/SH5 (d) UH/SH5

Fig. 10 Spectral ratios Fig. 11 Spectral ratios Fig. 12 Spectral ratios

(El Centro 1940 NS, 100cm/s/s) (EL Centro 1940 NS, 300cm/s/s) (EL Centro 1940 NS, 600cm/s/s)

8-7
 ,
14
10 --- _ _ _ _ _ ~<l ^ _ . 12
*__ 10
8
<! 8
6
4 -x / 4
'
2 2
n 0
0 2 46 8 10 12 0 2 4 6 8 10 12 14
Test
Test
(a) Natural frequency(Hz) (b) Amplification factor

3000

2000 * g 0.02
CO
'
+
15 i.**

< 1000 < 0.01

X
n
0 1000 2000 3000 0 0.01 0.02 0.03
Test Test
(c) Max. acc.(cm/s/s) (d) Max. shear strain

Fig.13 Comparison results of tests and analyses (El Centre NS)

12 12

10 10

8 8

6 f e
w W

4 4

2 2

0 0
0 500 1000 0 1000 2000 0 1000 2000 3000
Max. acc.(cm/s/s) Max. acc.(cm/s/s) Max. acc.(cm/s/s)

(a) lOOcm/s/s (b) 300cm/s/s (c) 600cms/s/

Fig. 14 Maximum acceleration at building ( El Centra NS)

8-8
 Nonlinear SSI analysis
by
Ronaldo I. Borja, Heng-Yih Chao and Chao-Hua Lin
Department of Civil & Environmental Engineering
Stanford University, Stanford, CA 94305-4020, USA

Abstract
Accurate representation of soil-structure interaction (SSI) effects is a crucial part of
earthquake engineering analysis. The SSI model must be robust enough to capture
nonlinear 3D effects, as well as accommodate the spectrum of frequencies of interest to the
analyst. This paper investigates the potential of standard nonlinear finite element (FE)
procedures for 3D analysis of SSI systems. The analysis uses the Lotung Large-Scale
Seismic Test (LSST) problem as a case study. Nonlinear ground response is a ubiquitous
feature of the soil behavior in Lotung; thus the analysis utilizes a direct method in which the
entire soil-foundation-structure system is modeled and analyzed in a single step.

1. Introduction
Ground motions are generally influenced by the presence of structures, and structural
motions are in turn influenced by the compliance of the supporting subsoils. These
coupling phenomena are due to soil-structure interaction (SSI) effects, which are absent
when the structure is founded on solid rock with extremely high stiffness. Dynamic
analysis of SSI effects requires varying levels of rigor depending on the type of analysis
(linear or nonlinear) and geometrical contraints (2D or 3D). In general, the computational
challenge lies in modeling nonlinear effects in a general three-dimensional setting. This
paper discusses the performance of a standard nonlinear finite element (FE) program, called
SPECTRA, for 3D analysis of SSI systems.
Inclusion of nonlinear effects in SSI analysis eliminates the convenience offered by the
principle of superposition, which allows a separate treatment of inertia! and kinematic
interactions before obtaining the combined response. Instead, nonlinear analyses are best
carried out by direct method, which entails modeling and analysis of the entire soil-
structure system in a single step. Section 2 describes some computational issues and
challenges relevant to a faithful modeling of SSI effects, as well as discusses some aspects
necessary for selecting a sound framework for nonlinear FE analysis by direct method.
In Section 3 we report the performance of a standard (implicit) nonlinear FE program,
SPECTRA, for estimating the ground response in Lotung incorporating SSI effects. The
type of analysis pursued is based on elastoplastic modeling with deviatoric plasticity under
the assumption of infinitesimal deformation. The solution is based on a total stress
formulation in which the soil solid and fluid are assumed to move as one body (undrained
condition). The Lotung Large-Scale Seismic Test (LSST) problem has been selected as a
prototype case study because of the relevance of its design to the goals of accurate SSI
modeling.

2. Framework for nonlinear SSI analysis

We consider an analytical platform for nonlinear SSI analysis embodied in a time-domain
FE model. The global iterative solution strategy is Newton-Raphson iteration with line

9-1
search, but the technique should also be able to accommodate quasi-Newton, modified
Newton, and even PCG-based equation solving techniques (each technique has its own
strengths and limitations). The specific platforms for each of the model components are
outlined and described below.
2.1. Constitutive model and stress-point algorithm. We seek a constitutive model that best
describes the material response. Models based on plasticity theory are on top of the
author's hierarchical list. The stress-point integration algorithm must be at least first-order
accurate and unconditionally stable. This is achieved with the use of standard implicit
return-mapping algorithm in computational plasticity. Explicit stress-point integration
algorithms for elastoplasticity are unacceptable they simply create unnecessary stability
problems and are not accurate enough for large load steps. Hypoplasticity models rank
next on the author's list of hierarchical models. Unconditionally stable stress-point
algorithms are also available for this class of model and must be utilized whenever
possible. Regardless of the type of model, the stress-point integration algorithm must be
linearized consistently as there is so much efficiency to be gained by using the consistent
tangent operator.
2.2. Finite deformation model. A finite deformation theory based on multiplicative
plasticity and implemented using product formula algorithm is on top of the author's
hierarchical list. This formulation has considerable advantage over the conventional
hypoelastic formulation. In the first place, there is no question as to what objective stress
rate must be used (Jaumann rate, Green-Naghdi rate, etc.), and so problems associated
with the use of the Jaumann stress rate, for example, do not exist. Furthermore, the model
has a hyperelastic basis which imposes no restriction on the elastic strains (unlike the
hypoelastic formulation which requires that the elastic strains be small). Finally, the
product formula algorithm can accommodate the standard return maps of infinitesimal
plasticity without loss of objectivity under rigid-body rotations. Quite recently, the author
has implemented the proposed technique in a nonlinear consolidation FE code, and the
results are very encouraging (Borja et al. 1998).
2.3. Time-integration algorithm. The classical Newmark family of algorithms has become
the backbone feature of many structural dynamics FE codes over the years, and has indeed
performed quite successfully in the geometrically linear case. Quite recently, this algorithm
has been found to fail to conserve energy and total angular momentum for the geometrically
nonlinear case. This result has a profound impact on the development of robust
mathematical models since conservation laws play a central role in classical mechanics; in
particular, conservation of angular momentum is crucial in motions with significant rigid-
body rotation, such as structures undergoing rocking motion. The impact of this discovery
on SSI research remains largely unexplored. Simo et al. (1992) have suggested time-
stepping algorithms that conserve energy and total angular momentum for general nonlinear
Hamiltonian systems, but this class of algorithms remains untested for earthquake
engineering analysis applications.
2.4. Liquefaction model. The problem of lateral flows and liquefaction-induced large
ground movement of saturated soils during and following an earthquake is a subject of
considerable importance in SSI modeling. Models based on Biot's two-phase mixture
theory and cast within the finite deformation model are possible, as described in Sec. 2.2.
The constitutive model should be capable of replicating hysteretic volume change behavior
to allow pore pressure buildup. Pore pressure buildup and the attendant liquefaction
phenomena have profound impacts on the responses of soil-structure systems.
2.5. Strain localization model. A problem not typically covered by the standard finite
element approximation is strain localization, particularly in the soil medium. Strain

9-2
localization effects are particularly important in simulating cracking of concrete, shear
banding in geomaterials, and structural collapse. Traditionally, cracking and softening
were treated as a constitutive response, although strictly speaking, they are a structural
response (see review by Read and Hegemier 1984). The SSI model should be robust
enough to accommodate strain localization as a structural response, and to allow modeling
of damage, limit states, and collapse.
An ideal FE analysis code must accommodate the above features, among others, or a
faithful simulation of the SSI phenomena is not possible. Although developing a package
with all of the above features is difficult, it is possible to include at least some of the most
important SSI aspects in the analysis. To illustrate the rigors of a 3D nonlinear FE analysis
of SSI phenomena, the next section describes a FE modeling of a soil-structure system in
Lotung, Taiwan, as it responded to the earthquake of May 20, 1986.

3. SSI analysis of Lotung LSST problem

Lotung is a seismically active region in northeastern Taiwan, and was the site of two
scaled-down nuclear containment structures (1/4-scale and 1/12-scale models) constructed
by the Electric Power Research Institute, in cooperation with Taiwan Power Company, for
SSI research (Tang et al. 1990). The local geology at the test site has been established
from shear wave velocity and field boring tests. On May 20, 1986, a strong earthquake,
denoted as the LSST7 event, with magnitude 6.5, epicentral distance of 66 km, and focal
depth of 15.8 km shook the test site. Two downhole arrays located approximately at 3 m
and 49 m from the edge of the 1/4-scale model, herein called DHA and DHB arrays,
respectively, recorded the downhole motions at depths of 0, 6, 11, 17, and 47 m (Fig. 1).
In this paper, we will analyze the downhole motions recorded by both arrays using a
nonlinear FE code SPECTRA.

DHB
1/4-SCALE
MODEL

3-COMPONENT
ACCELEROMETERS

Figure 1. Location of surface and downhole instrumentation, LSST site: (a) plan; (b) elevation.

For purposes of 3D analysis, a full-scale FE model for the 1/4-scale structure and
foundation is shown in Fig. 2. Array DHB is located at the edge of the mesh, while array
DHA is located approximately 3 m from the edge of the structure. Both arrays are located
along the northern arm, as indicated in the figure (see also Fig. 1). The mesh consists of
4,320 eight-noded trilinear brick elements, with a total of over 13,000 degrees of freedom.

9-3
Free-field motions are applied at the bottom and side boundaries. Before a full SSI
analysis can be carried out, it is necessary that the input free-field motions be first
determined. In the following we describe a numerical model for generating the input free-
field ground motion consistent with the SSI model for the Lotung problem.

DMA

DHB

DEPTH = 0 m

DEPTH = 47 m

Figure 2. FE mesh for Lotung LSST case study.

Figure 3. FE mesh for nonlinear ground response analysis: (a) soil column model; (b) stick FE mesh.

The input free-field motions are generated by the same FE code assuming the case of
vertically propagating waves. Fig. 3 shows a FE mesh consisting of column and stick
elements representing the 47-m deep soil column. The special element shown in Fig. 3(b)
has been included in the library of the code SPECTRA specifically for nonlinear ground
response analysis. The vertical discretization of the soil layer is consistent with the full 3D
mesh of Fig. 2. Each stick element contains 3 DOFs at the nodes: two horizontal and one
vertical. The constraints imposed by the condition of vertically propagating waves are that

9-4
the two normal horizontal strains as well as the shear strain on the horizontal plane must be
equal to zero, and that for the general nonlinear case the three kinematical components of
motion must be coupled. Previous studies suggest that the angle of incidence of the
seismic waves during the May 20, 1986 event is only around 6 degrees relative to the
vertical direction (Chang et al. 1990), which justifies the simplifying assumption of
vertically propagating waves.
The soil in Lotung is modeled using bounding surface plasticity theory with a vanishing
elastic region in which the hardening modulus is interpolated by an exponential hardening
function (Borja and Amies 1984). The material parameters for this soil include the elastic
bulk and shear moduli as well as the exponential hardening parameters, which have been
determined for the Lotung soil from shear and compressional wave velocity profiles as well
as from moduli ratio degradation curves available for the LSST site. Details of how the
model parameters have been determined for the LSST site are described by Borja et al.
(1998) and Borja and Lin (1998).
Figures 4, 5 and 6 show respective east-west (EW), north-south (NS) and up-down (UD)
free-field motions predicted by the program SPECTRA using the stick FE model of Fig. 3,
superimposed with the free-field motions recorded by downhole array DHB (the
assumption that array DHB is sufficiently far from the structure and that it recorded
basically free-field motion is crucial for the rest of the analysis to be meaningful).
Calculations were carried out on a 266-MHz Pentium n PC. The close agreement between
the recorded and predicted responses is noteworthy, which implies that it is possible to
generate reasonably accurate free-field motions with a soil column model that allows for a
full kinematical coupling of all three components of motion. Computer runtimes are in the
order of 2 minutes on the PC for a time-domain analysis consisting of about 1000 time
steps and 4-5 iterations per time step.

5o
O'1i
o (a) DEPTH, 0 M

10 15

(b) DEPTH, 11 M

10 15

- DHB47/EW
INPUT MOTION

(c) DEPTH, 47 M

10 15
TIME

Figure 4. EW free-field acceleration (m/sec2)-time (sec) history: Lotung LSST7 case study.

9-5
 2.5
2.0
1.5
1.0
0.5
0.0
-0.5
-1.0 (a) DEPTH, 0 M
-1.5
10 15

2.5
2.0 -DHB11/NS
1.5 - SPECTRA
1.0
0.5
0.0
-0.5
-1.0 (b) DEPTH, 11 M
-1.5
10 15

2.5
2.0 - DHB47/NS
1.5 - INPUT MOTION
1.0
0.5
0.0
-0.5
-1.0 (c) DEPTH, 47 M
-1.5
10 15
TIME

Figure 5. NS free-field acceleration (m/sec2)-time (sec) history: Lotung LSST7 case study.

0.50

0.25

0.00

-0.25
(a) DEPTH, 0 M
-0.50
10 15

0.50

0.25

0.00

-0.25
(b) DEPTH, 11 M
-0.50
10 15

0.50
DHB47/UD
0.25 INPUT MOTION

0.00

-0.25
(c) DEPTH, 47 M
-0.50
1Q 15
TIME

Figure 6. UD free-field acceleration (m/sec2)-time (sec) history: Lotung LSST7 case study.

9-6
Assuming that array DHB recorded truly free-field responses and that the soil profiles and
the input seismic motions are essentially the same for arrays DHA and DHB, the effect of
SSI can be inferred by comparing the motions recorded by the two arrays. Figures 7, 8
and 9 compare the EW, NS and UD motions recorded by arrays DHA and DHB. Note that
the recorded motions at depth of 47 m are essentially the same, but the responses are quite
different at shallower depths. Considering that DHA is only 3 m away from the structure,
we can postulate that this difference is due to SSI effects. We will follow this idea and
proceed with the nonlinear SSI analysis of the Lotung problem.
SSI effects are now investigated using the program SPECTRA along with the full 3D FE
mesh shown in Fig. 2. Calculations were carried out on CRAY C90 supercomputer at San
Diego Supercomputer Center (SDSC). The 3D mesh of Fig. 2 was the biggest that the
authors could construct for this problem at the time of the analysis, given the limited CRAY
storage and CPU time made available by SDSC. There is certainly more room for
improvement in so far as refining the mesh is concerned, since this mesh is admittedly too
coarsely discretized on the horizontal plane compared to the much finer mesh discretization
of the soil layers in the vertical direction. However, for ground motions dominated by
horizontal sidesway action, this mesh has a resolution comparable to that of the stick model
used for nonlinear ground response analysis.
The 1/4-scale nuclear plant structure is modeled as a hollow cylindrical tank made of elastic
trih'near brick elements and integrated using the standard 8-point Gauss integration rule.
The dimensions of the model and properties of the material are similar to those of the
prototype structure. The soil is modeled as elastoplastic brick elements integrated using the
B-bar method to alleviate mesh locking in the nearly incompressible regime

Z
o -DHB
-DHA

(a) DEPTH, 0 M

10 15

2
O

o (b) DEPTH, 11 M

10 15

2 -DHB
-DHA
I
s
nl
8-
(c) DEPTH, 47 M

10 15
TIME

Figure 7. EW acceleration (m/sec2)-time (sec) history: comparison between DHA and DHB.

9-7
 Z 2
O

ffi

(a) DEPTH, 0 M
8-*
10 15

DHB
DHA

s
O (b) DEPTH, 11 M

10 15

Figure 8. NS acceleration (m/sec2)-time (sec) history: comparison between DHA and DHB.

0.5
-DHB
0.25 -DHA

jjj
"J -0.25
(a) DEPTH, 0 M
-0.5
10 15

0.5
-DHB
0.25 -DHA

-0.25
(b) DEPTH, 11 M
-0.5
10 15
0.5
DHB
0.25 DHA

-0.25
(c) DEPTH, 47 M
-0.5
10 15
TIME

Figure 9. UD acceleration (m/sec2)-time (sec) history: comparison between DHA and DHB.

9-8
The solution procedure goes as follows. Free-field motions are computed from the
nonlinear ground response analysis described previously. Then, the computed motions are
applied at the base and side boundaries of the mesh in Fig.2 for SSI analysis. Iterations are
performed at each time step via a composite Newton-PCG iteration algorithm, in which
Newton's method is applied globally at each time step to solve the nonlinear problem and
the PCG iteration is applied locally to solve the linearized problem. For purposes of
executing the PCG algorithm, a global elastic tangent operator is employed as the
preconditioner. The solutions generally converged in 4-5 iterations based on a maximum
residual error norm of 0.001%.
Results of the analysis are shown in Figs. 10, 11 and 12. In these figures, the computed
motions are shown with and without SSI effects. The latter results (no SSI) were obtained
from the nonlinear ground response analysis discussed previously (cf. Figs. 4, 5 and 6).
Note that SSI effects are again seen to be more evident at shallower depths, consistent with
the results of comparisons between the motions recorded by DHA and DHB as shown in
Figs. 7, 8 and 9. That the difference between the results of Figs. 10, 11 and 12 may
indeed be attributed to SSI effects is explainable as follows: if the structure in Fig. 2 were
removed and the 3D mesh were still used in the SSI analysis, the computed motions on
DHA would be identical to the free-field motion applied on DHB.
Implicit 3D nonlinear analyses are computer-intensive as affirmed by the following
statistics: for a 3D mesh with about 4.5K elements, 5K nodes, and 13K DOFs similar to
that shown in Fig. 2, a time-domain analysis composed of 1,000 time steps and 4 to 5
iterations per time step would require about 4 CPU-days on the CRAY C90 supercomputer
utilizing one processor.

(b) DEPTH, 11 M

10 15

(c) DEPTH, 47 M

10 15
TIME

Figure 10. EW acceleration (m/sec2)-time (sec) history with and without SSI: Lotung LSST7 case study.

9-9
 SSI
NO SSI

(a) DEPTH, 0 M

10 15

o -SSI

g
ec
ui
- NO SSI

ui
o (b) DEPTH, 11 M

10 15

(c) DEPTH, 47 M

10 15
TIME

Figure 11. NS acceleration (m/sec2)-time (sec) history with and without SSI: Lotting LSST7 case study.

0.5
SSI
0.25 NO SSI

o
-0.25
(a) DEPTH, 0 M
-0.5

10 15

0.5
SSI
0.25 NO SSI

-0.25
(b) DEPTH, 11 M
-0.5

10 15

0.5
SSI
0.25 NO SSI

-0.25
(c) DEPTH, 47 M
-0.5

10 15
TIME

Figure 12. UD acceleration (m/sec2)-time (sec) history with and without SSI: Lotting LSST7 case study.

9-10
These above statistics look astounding, although it is somewhat comforting to see that at
least this task is now feasible. The new parallel vector 14-processor CRAY T90
supercomputer available in SDSC is similar to the CRAY C90 systems but each processor
is about 1.7 times faster (24 Gflops peak speed). Thus, we can expect a similar
improvement in the computer runtime with this faster machine.

4. Closure
Implicit 3D nonlinear FE analyses are computer-intensive but can provide detailed
information on the effects of SSI without undue simplifications of the geometrical
configurations and the boundary conditions. Advances in the computer hardware will
enhance the accuracy of the simulations and must be accompanied by parallel developments
of more advanced analytical platforms for nonlinear SSI analysis.

Acknowledgments
Financial support for this research was provided by the Earthquake Hazard Mitigation
Division of National Science Foundation under contract Nos. CMS-9114869 and CMS-
9613906 through the direction of Dr. C.J. Astill. The writers would like to thank Dr. H.T.
Tang and Electric Power Research Institute for making the digitized data for the LSST7
event available.

References
Borja, R.L and Amies, A.P. (1994). Multiaxial cyclic plasticity model for clays, J.
Geotech. Engrg., ASCE 120, 1051-1070 (1994).
Borja, R.L, Tamagnini, C., and Alarcon, E. (1998). Finite strain elastoplastic
consolidation, Part 2: Finite element implementation and numerical examples,
Comput. Methods Appl. Mech. Engrg., 159, 103-122.
Borja, R.L, Chao, H.Y., Montans, K, and Lin, C.H. (1998). Nonlinear ground response
at Lotung LSST site, J. Geotech. and Geoenvironmental Engrg., ASCE, in review.
Borja, R.L and Lin, C.H. (1998). Modeling nonlinear ground response of nonliquefiable
soils, Earthquake Engrg. Struc. Dyn., in review.
Chang, C.Y., Mok, C.M., Power, M.S., Tang, Y.K., Tang, H.T., and Stepp, J.C.
(1990). Equivalent linear and nonlinear ground response analyses at Lotung seismic
experiment site, Proc. 4th U.S. Nat. Conf. Earthquake Engrg, Palm Springs,
California, 1, 327-336.
Chao, H.Y. and Borja, R.I. (1998), Nonlinear dynamic soil-structure interaction analysis
and application to Lotung problem, John A. Blume Earthquake Engineering Report,
preprint.
Read, H.E. and Hegemier, G.A. (1984). Strain softening of rock, soil, and concrete-A
review article, Mech. Mater., 3, 271-294.

9-11
Simo, J.C., Tarnow, N. and Wong, K.K. (1992). Exact energy-momentum conserving
algorithms and symplectic schemes for nonlinear dynamics, Comput. Methods
Appl. Mech. Engrg. 100, 63-116.
Tang, H.T., Tang, Y.K., and Stepp, J.C. (1990). Lotung large-scale seismic experiment
and soil-structure interaction method validation, Nuclear Engrg. Des., 123, 197-
412.

9-12
 THREE DIMENSIONAL SEISMIC RESPONSE OF BUILDING
FOUNDATION SYSTEMS

By E. Bazan-Zurita1 , N. C. Bazan- Arias2 and J. Bielak3

Consultants, Inc.,
Pittsburgh, PA
Department of Civil and Environmental Engineering, University of Pittsburgh
Pittsburgh, PA
Department of Civil and Environmental Engineering, Carnegie-Mellon University
Pittsburgh, PA

ABSTRACT

This paper provides means for quantifying the importance of three-dimensional soil-structure
interaction (SSI) during earthquakes on elastic building-foundation systems localized on sites
where SSI effects can be significant. The building is modeled as an elastic one-story three-
dimensional structure resting on an embedded foundation. A parametric study is conducted to
assess the effects of the foundation impedance, building eccentricity, and excitation characteristics
on the seismic response. The results indicate that it is often not conservative to ignore SSI in the
seismic analysis of torsionally eccentric structures.

INTRODUCTION

The seismic design of buildings, according to current building codes, is based on static or
dynamic analyses that consider elastic behavior. Effects of inelastic behavior during strong
earthquakes are taken into consideration by reducing by a global factor the forces obtained in the
elastic analysis. Such a factor is derived from comparisons of the response of simple elastic and
inelastic models, and is justified in practice on the basis of the overall performance of different
types of buildings during actual earthquakes. Elastic analysis remains, therefore, the most widely
used approach for the seismic design of building-foundation systems.

The objective of this study is to evaluate the impact of three-dimensional soil-structure

interaction (SSI) on the elastic response of building-foundation systems under earthquake
excitation. In pursuing this objective, we have conducted a parametric analysis of an idealized
single-story three-dimensional model, meant to represent the first two translational and the first
torsional modal response of actual buildings, using two prototype foundation conditions: one very
stiff (fixed base) and the other simulating soft soil conditions. We focus our attention on the two

10-1
response quantities that are most important in seismic design of three-dimensional systems, namely
base shear force and dynamic torsional eccentricities.

FORMULATION

The system under investigation is shown in Figure 1. It consists of an elastic viscously damped
structure with mass ml and three degrees of freedom (two horizontal perpendicular translations
along the X, Y axes and one rotation around the vertical axis, Z). The structure's properties
represent the three (two translational and one torsional) "fundamental modes" of a multistory
building with floors stiff enough in their own plane to be considered as infinitely rigid horizontal
diaphragms. This structure rests on a rigid foundation with mass mo, embedded in a flexible soil,
and with no slippage allowed between the base and the soil. The soil flexibility allows translational
and rotational displacements related to the X and Y axes and torsional rotations around the Z axis.
Formulated thus, the system has eight significant degrees of freedom, namely, two horizontal
translations and one torsional rotation at the top mass, ml5 and two horizontal translations, two
rocking rotations and one torsional rotation at the base mass, m^. The position of the lumped
masses ml and m^ (center of masses of the diaphragms) can be eccentric in both the X and Y
directions. The seismic input motion occurs in the X direction. This SSI system was modeled by a
three-dimensional two-story column whose first story represents the foundation and its second the
structure. The program ETABS (Wilson et al, 1975) was used to perform the numerical
calculations.

PLANAR SYSTEMS

To establish a basis of comparison for three dimensional behavior, we have studied planar SSI
systems with no torsional eccentricities and masses moving only along the X horizontal axis. In
this case only three degrees of freedom remain active: horizontal translation of the top mass in the
X axis, horizontal translation of the base mass relative to free field motion, and rotation of the
system in the X-Z plane. If there is no interaction, the relevant dynamic properties of the
superstructure are the fixed-base period, T1? and the damping ratio, 1- When SSI is significant,
the most relevant foundation parameters are its mass, mo, the SSI translational (horizontal) and
rocking stiffness coefficients, kv, and k^, and the corresponding SSI damping ratios, v and

To examine design situations, we have used the response spectrum method to analyze planar
SSI systems considering two types of input spectra: 1) a constant design acceleration for all
periods, and 2) a hyperbolic spectrum with design acceleration that decreases inversely with the
period. The two spectra are depicted on Figure 2. n^ has been taken to be 0.2 ml throughout this
paper. We have assumed that the translational and rocking SSI stiffness are kv = a (5 ki) and k^ =
a (ki H2) varying a such that T'/Ti varies from 1.0 (a very large) to 1.33 (a = 1). T' is the
fundamental SSI period. SSI damping usually increases the damping ratio of the fundamental
mode of a building foundation system. This effect could be included by modifying the design
spectra. However, we opted to ignore it since our interest is focused in the changes of seismic

10-2
response attributed to differences in modes and periods of vibration when SSI is included. In any
event, beneficial SSI damping effects should be carefully evaluated in practice since soil layering,
separation of the foundation from the surrounding soil, and the presence of adjacent structures
could decrease them significantly.

Table 1 summarizes the resulting dynamic base shear normalized by the product of mi times the
spectral acceleration, S', corresponding to T. Since the values in the table are very close to unity,
it is concluded that the seismic shear forces can be accurately calculated as miS'. The errors are
smaller than 3.0 percent and are well covered by the factors of safety included in design spectra
and regulations. However, it is clearly important to calculate accurately the fundamental SSI
period to determine the correct value of S'.

TABLE 1
NORMALIZED BASE SHEAR FORCE, V/( miS1) FOR PLANAR SYSTEMS
Factor T'/Ti Flat Hyperbolic
a Spectrum Spectrum
10000 1.000 1.000 1.000
10 1.037 1.004 1.005
3 1.120 1.010 1.012
1 1.329 1.020 1.027

DYNAMIC PROPERTIES OF 3-D SYSTEMS

We have considered two types of three-dimensional systems. Firstly, it has been assumed that
the masses m^ and n\Q both have the same eccentricity along the Y axis, Le., in the direction
perpendicular to the input motion, and no eccentricity along the X axis. The floor and the
foundation mat of the model are both taken to be square, with side dimension b and torsional
radius of gyration b/V6. We have also considered that the structure has the same translational
stiffness ki in the X and Y directions, and that the SSI translational and rocking SSI stiffnesses are
respectively the same in both horizontal directions.

The torsional stiffness of the structure has been varied such that the ratio of the fixed-base
uncoupled torsional period, Te, to the fundamental uncoupled fixed-base translational period, TI,
equals 0.6, 0.8 and 1.0. These values represent, respectively, buildings where the lateral stiffness is
concentrated in the exterior portions of its plan, distributed more or less uniformly in plan, or
concentrated at the central part of the building's plan (see Figure 3). For the base case studied
herein, we have set kv = 5ki and k$ = ki H2 resulting in T'/Ti = 1.33 for symmetric systems. The
torsional SSI stiffness, ke, was taken as 1.5 k^, which represents a mat foundation embedded in a
viscoelastic halfspace. For a second basic SSI case, the values of kv and k$ remained the same but
kewas reduced to 0.15 k^. This could correspond, for instance, to structures founded in a soil

10-3
layer underlain by bedrock or with flexible foundation. To examine systematically systems where
torsion is relevant, the static eccentricity has been varied from zero to 0.3 times the building's plan
dimension, b.

For this type of torsionally eccentric systems, the first and third modes of vibration exhibit
coupled translations and rotations. The second mode is uncoupled, with displacements only in the
Y direction, and has the same period as the corresponding planar system. Figure 4 shows the first
and the third natural periods of these SSI systems. All values have been normalized by the
fundamental fixed-base period of the corresponding symmetric system. It is noted that the first
period ratio of eccentric systems is always larger than unity, in agreement with the known result
that torsional eccentricities amplify the fundamental period of the associated symmetric structure.
The increase of the fundamental period with the eccentricity is greater when Te/Ti approaches
unity, for both fixed base and SSI systems. Conversely, the normalized third periods are always
smaller than the corresponding ratio Te/Ti. In all cases incorporation of SSI results in longer
periods. The effects of SSI on the first period are always more significant, showing increasing
factors greater than 1.33, since the SSI parameters were set to amplify the translational period by
33 percent when there is no eccentricity. The SSI effects on the third period are small when ke
was taken as 1.5 k^,, but more noticeable when ke was reduced to 0.15 k^.

To examine cases with eccentricities in two directions, systems for which the two masses mi
and m0 have the same static eccentricity, est, along both the X and Y axes have been analyzed.
The properties of these systems are the same as those for the cases with eccentricity in only one
direction. The difference is the additional eccentricity in the Y direction. The first and third modes
of vibration are again coupled. The second mode has no torsional component and has the same
period as the corresponding planar system. Figure 5 shows the first and the third natural periods of
these doubly eccentric SSI systems, normalized by the fundamental fixed-base period of the
corresponding symmetric system. The trends of these periods are the same as those for systems
eccentric in only one direction. However, the increasing and decreasing amounts are always more
pronounced. Again, the SSI effects are more noticeable when ke assumes the reduced value of
0.15 k^ and the impact of torsional eccentricities is more noticeable as Te/Ti approaches unity.

RESPONSE OF 3-D SYSTEMS WITH ECCENTRICITY IN ONE DIRECTION

The three-dimensional systems with eccentricity in only one direction described in the previous
section have been analyzed with the response spectrum method, including all modes of vibration in
the calculations. The seismic input consists of two simplified design spectra: flat and decreasing
with period. Since these systems have natural periods with very close values, a complete quadratic
combination rule was used to combine modal responses. The base shear forces obtained using a
flat spectrum for Te/Ti = 0.6, 0.8 and 1.0 are presented in Figure 6. The forces have been divided
by the product of the mass mi times the spectral acceleration, S' (in this case constant). In all
cases, whether with or without SSI, an increase in the eccentricity decreases the shear force, with
reductions as large as 25 percent for fixed bases systems with Te/Ti = 1. The reductions are

10-4
considerably smaller when SSI effects are significant. This clearly indicates that ignoring SSI will
not necessarily lead to conservative results in the dynamic elastic analysis of eccentric structures.

The dynamic eccentricities resulting from a flat spectrum are presented in Figure 7. These
eccentricities have been normalized by the corresponding static eccentricity, thus illustrating the
dynamic amplification factors of static torsional moments. Such factors increase significantly when
the uncoupled torsional period of the structure approaches the uncoupled translational period (i.e.,
when Te/Ti approaches unity). Values as high as 5.5 occur in fixed-base systems with small
eccentricities, indicating that in practice torsional moments can be significant even in buildings that
are almost symmetric. SSI always reduces the dynamic amplifications of the eccentricities; but,
again, if the SSI torsional stiffness becomes small, these reductions also become appreciably
smaller or even negligible for systems with large eccentricities. As a rule, the beneficial effects of
SSI decrease as the static eccentricity increases, particularly when the uncoupled translational and
torsional periods are close to each other. Sound design recommendations aim at preventing the
tuning of translational and torsional uncoupled periods on the basis of this observation.

The normalized shear forces and dynamic eccentricities corresponding to the hyperbolic
spectrum are presented in Figures 8 and 9. The variations of shear force with dynamic properties
and SSI parameters are very similar to those shown in Figures 6 and 7 for a flat spectrum. The
maximum differences are approximately 10 percent. This indicates that by dividing the shear
forces by the spectral ordinate the normalized response becomes almost independent of the type of
spectrum. On the other hand, it highlights again the importance of calculating accurately the SSI
periods of vibration since the seismic response is almost directly proportional to spectral
accelerations.

The comparison of Figures 7 and 9 shows that the ratio of the dynamic eccentricity, edyn, to the
static eccentricity, est, is in most cases greater for the hyperbolic spectrum (Figure 9) than for the
flat spectrum (Figure 7). The differences are the result of different relative participations of the
various modes of vibration of the system in the calculation of shear forces and torsional moments.
This emphasizes the importance of calculating accurately higher modes of vibration that exhibit
significant torsional rotations.

RESPONSE OF 3-D SYSTEMS ECCENTRIC IN TWO DIRECTIONS

To examine three-dimensional cases with eccentricities in two directions, systems for which the
two masses mi and mo have the same static eccentricity, e^, along both the X and Y axes have been
analyzed. The properties of these systems have been selected to be the same as those in the
previous section with the exception of the additional eccentricity in the Y direction. The input
spectrum is flat. Table 2 presents the normalized shear forces in the horizontal X and Y directions
and the dynamic torsional eccentricity as a function of the static eccentricity, e^,. It is immediately
apparent that all these response quantities are significantly affected by the eccentricity. While the
shear in the direction of the ground motion, Vx, decreases with increasing values of e^, the
opposite occurs for the torsional moments and the shear in the perpendicular direction, Vy, which

10-5
now is not nil as a result of the coupling of displacement in the modal shapes. The effects of SSI
and of the eccentricity on Vx and on the dynamic eccentricity follow the same trends as in systems
with eccentricity in only one direction, but are always numerically higher.

TABLE 2
SEISMIC RESPONSE OF SYSTEMS
ECCENTRIC IN Two DIRECTIONS. FLAT SPECTRUM

Shear Force in X, Vx/(miS') Shear Force in Y, Vy/(miS') Dynamic Eccentricity, tdynftst

Te/Ti 0.6 0.8 1.0 0.6 0.8 1.0 0.6 0.8 1.0
FIXED BASE
e = 0.00 1.000 1.000 1.000 0.000 0.000 0.000 NA NA NA
0.05 0.994 0.958 0.832 0.058 0.164 0.348 1.62 2.73 4.40
0.10 0.963 0.843 0.716 0.210 0.412 0.513 1.58 2.40 3.04
0.15 0.897 0.752 0.672 0.381 0.537 0.568 1.59 _j 2.19 2.43
0.20 0.825 0.706 0.655 0.503 0.587 0.593 1.62 2.01 2.11
0.30 0.738 0.676 0.649 0.609 0.626 0.620 1.63 1.78 1.80
SSI, Ke = 1.5 K*
e = 0.00 1.020 1.020 1.020 0.000 0.000 0.000 NA NA NA
0.05 1.019 1.013 0.990 0.030 0.064 0.138 1.30 1.67 2.44
0.10 1.009 0.977 0.891 0.117 0.229 0.383 1.30 1.63 2.22
0.15 0.982 0.904 0.792 0.242 0.406 0.529 1.30 1.63 2.08
0.20 0.935 0.829 0.736 0.373 0.525 0.592 1.33 1.65 1.96
0.30 0.831 0.744 0.696 0.553 0.624 0.696 1.42 1.64 1.76
SSI,Ke =0.151^
e = 0.00 1.020 1.020 1.020 0.000 0.000 0.000 NA NA NA
0.05 1.014 1.000 0.948 0.062 0.114 0.223 1.63 2.18 3.26
0.10 0.980 0.919 0.813 0.224 0.345 0.468 1.59 2.03 2.67
0.15 0.911 0.821 0.735 0.401 0.508 0.566 1.60 1.96 2.31
0.20 0.837 0.758 0.701 0.523 0.585 0.605 1.63 1.89 2.07
0.30 0.751 0.709 0.681 0.627 0.640 0.638 1.65 1.75 1.80

Table 3 presents similar results for the hyperbolic spectrum. Note that the normalized shear force
in the horizontal X direction is in all cases higher than for the flat spectrum. This reflects a larger
contribution of higher modes of vibration (with smaller periods) when the spectrum increases for
decreasing periods. For the same reason, the normalized shear force in the Y direction is also
higher than for the flat spectrum. On the other hand, for most cases, the ratio of the dynamic
torsional eccentricity to the static eccentricity is smaller than the ratio corresponding to the flat
spectrum. The explanation is that in the calculation of the dynamic eccentricity the dynamic
torsional moment is divided by a relatively larger dynamic shear force. Again, all these response

10-6
quantities are appreciably affected by the eccentricity, following the same trends as in the case of
the flat spectrum.

TABLES
SEISMIC RESPONSE OF SYSTEMS
ECCENTRIC IN Two DIRECTIONS. HYPERBOLIC SPECTRUM

Shear Force in X, V^n^S1) Shear Force in Y, Vy/Cm^ 1) Dynamic Eccentricity, &AyJ&si

Te/T, 0.6 0.8 1.0 0.6 0.8 1.0 0.6 0.8 1.0
FIXED BASE
e = 0.00 1.000 1.000 1.000 0.000 0.000 0.000 NA NA NA
0.05 0.998 0.972 0.906 0.059 0.169 0.379 1.76 2.91 4.29
0.10 0.980 0.888 0.845 0.216 0.441 0.603 1.69 2.49 2.81
0.15 0.931 0.832 0.853 0.400 0.602 0.719 1.66 2.16 2.10
0.20 0.879 0.826 0.890 0.542 0.694 0.804 1.64 1.87 1.68
0.30 0.840 0.882 0.997 0.699 0.822 0.951 1.52 1.45 1.24
SSI, Ke = 1.5 K*
e = 0.00 1.027 1.027 1.027 0.000 0.000 0.000 NA NA NA
0.05 1.027 1.024 1.007 0.037 0.069 0.144 1.41 1.82 2.62
0.10 1.023 1.000 0.933 0.128 0.242 0.410 1.39 1.75 2.32
0.15 1.005 0.946 0.867 0.256 0.433 0.589 1.37 1.71 2.09
0.20 0.970 0.892 0.849 0.394 0.574 0.691 1.38 1.67 1.85
0.30 0.896 0.861 0.892 0.603 0.730 0.822 1.40 1.52 1.46
SSI, Ke = 0.15K
e = 0.00 1.027 1.027 1.027 0.000 0.000 0.000 NA NA NA
0.05 1.025 1.015 0.976 0.065 0.118 0.233 1.80 2.35 3.39
0.10 1.003 0.956 0.879 0.234 0.365 0.513 1.74 2.15 2.70
0.15 0.950 0.945 0.845 0.425 0.594 0.656 1.71 2.00 2.21
0.20 0.897 0.854 0.855 0.568 0.666 0.744 1.69 1.84 1.85
0.30 0.861 0.877 0.931 0.725 0.798 0.876 1.58 1.53 1.41

CONCLUSIONS

The results presented in this paper show that the seismic response of a building-foundation
system including soil structure interaction can be significantly different from that calculated with a
fixed-base model. Simple single-story models have been used for rapid evaluation of SSI effects
showing that changes of tens of percent occur in the shear force on the structure and on the
torsional moments. The SSI effects are mainly the result of the increase in the fundamental period,
which leads to a different spectral ordinate. However, in three-dimensional eccentric structures

10-7
SSI also produces appreciable changes in higher natural periods and modes of vibration, affecting
significantly dynamic eccentricities and shear forces in the direction perpendicular to the seismic
input.

For the three-dimensional fixed-base eccentric systems examined in this paper, an increase in
eccentricity reduces the shear force up to 35 percent when the static eccentricity is 0.3 times the
building plan dimension and the ratio Te/Ti is equal to 1.0. These reductions are appreciably offset
when SSI is significant. SSI also leads to even greater reduction in the dynamic amplifications of
static eccentricities as compared to the associated fixed-base systems. The reductions are,
however, much smaller when the torsional SSI stiffness is small compared to the rocking SSI
impedance. In general, the beneficial effects of SSI diminish as the static eccentricity increases,
particularly when the uncoupled translational and torsional periods are close to each other.

The above results point out to the convenience of incorporating SSI in the elastic dynamic
analysis to avoid unconservative calculations of the seismic response. While simple formulas such
as the one proposed by Jennings and Bielak (1973) provide rapid and accurate estimates of the SSI
uncoupled fundamental translational period, the torsional SSI parameter should also be carefully
evaluated and incorporated in the dynamic modal analysis of three-dimensional structures. Special
attention should be paid to factors such as layering, proximity to other structures, stiffness of the
foundation members, and embedment. It is pertinent to recall that torsional moments are favorable
or detrimental to the vertical frames or walls of a building depending on their location in plan, and
that a smaller moment is not favorable in every instance. This highlights the importance of
estimating torsional eccentricities as accurately as possible.

The findings of this study reinforce design recommendations oriented to increase the torsional
stiffness of a building to prevent tuning of translational and torsional uncoupled periods as well as
to avoid excessive eccentricities. This can be achieved by properly locating in plan the vertical
systems that resist the seismic forces. With the help of modern computer programs, effects of
torsion and of SSI can be expeditiously incorporated in a three-dimensional dynamic analyses.
Significant changes in the periods and modes of vibration with respect to the fixed-base uncoupled
properties may indicate the need for reviewing the basic structural and foundation layout.

REFERENCES

Jennings, P.C., and J. Bielak. "Dynamics of Building-Soil Interaction." Bull. Seism. Soc. Am.,
1973 63 (1), 9-48.
Wilson, E.L., Holings, J.P. and Dovey, H.H., "Three Dimensional Analysis of Building Systems
(extended version)", Report No. UCB/EERC 75-13, Earthquake Engineering Research Center,
University of California, Berkeley, California, 1975.

10-8
 x^ *>x
X x-^

"Sv*^^ ' frvCfr-S-

1 x

Figure 1. Three-Dimensional SSI System

10-9
 Flat

S '
Hyperbolic

Socl/T

Figure 2. Flat and Hyperbolic Design Spectra

10-10
 same in both
directions
x

Te /Tir=0.6

Te / Tj = 0.8

Figure 3. Plan View of Three-Dimensional Systems

10-11
 Normalized Period Normalized Period Normalized Period
P P r* p p -* * < *
o o o O S S 8 8 b o o -* -* i*
>n o b In o cn c
8 I ' ' ' | o Q O O O O C
o ] 1 Q 4 i 1 i i ' '
t i o
'f i j
CD r i
4*. J
o ' : :
to M 0 i
^ 3
S 1
CO b
ri j : g J
3 : ii
:

O P
Lk p
Idk -
3 O i"i o ;i :, j p
Lk
i :i j
O
o 0) i1 i 4
,1 '.
J
fa '1 o-1 ^II i :
o p
I n
H
M o
, ;' H
-T1
* '
cn
;' : P o
w S
\ ;i
B1 ;i ; il S
O EL ; ^
n ' ' n, :i '
& HH i 3it
o> 2? ii i
2. p
I 1 ' :! i
oB. O ro 0
ii o
S'l s: o 'i ' ro
,!i 8 :l ! i o il i 1
O W ;' ; i i ; i ;i ;
ST-
O

HK ;l ; 1
ii
i r i i
n
M ;i i
o 1 1 0 . i'
ro ro
i cn cn iili
ir i
'i i11
,i i
O i : i ; i
P : ;i :ii
p i 'i i1
CO
o
r
i
'i
iii o i i i
w I U-J 1 LJ 1 o .. -v
CO
3. s?
Q.
s
rt-
S. a
 Normalized Period Normalized Period
p P -* r* N p p -A _i K)
08 s 2 g 8 g 8
p
b
o
TT
p
8
on
oo
P
o

o II
II 2 p
o H CO w b)
I
LO H

H
I1
p
KJ
tn - 1
: i
p : i
co
o co
3.
 1.10

1.00 -.
o
O o.90 -

0.80 -I-
O

0.30

0.70
0.00 0.05 0.10 0.15 0.20 0.25 0.30

1.10

0.70
0.00 0.05 0.10 0.15 0.20 0.25 0.30

Normalized Static Eccentricity

- - -NoSSI, Ke= 1.

Figure 6. Normalized Base Shear Force along X Axis,

Flat Spectrum
10-14
 = 0.6T1

c
0)
o
o
LJ
TO
0)

010 0.15 0.20 0.25 0.30

Te = 0.8T,
3.00

~ 2.60

2.20 ?
LJ
TO
O
SN 1-80 - -
ra

1.40
0.05 0.10 0.15 0.20 0.25 0.30

5.50
>
-4-

_O

4-> 4.50 - -
O
o
o
LJ
3.50 - *
0)
"w
2.50-

1.50
0.05 0.10 0.15 0.20 0.25 0.30

Normalized Static Eccentricity

---NoSSI, Ke=1.

Figure 7. Normalized Dynamic Eccentricity Perpendicular to the

X Axis, Flat Spectrum
10-15
 Normalized Shear Force Normalized Shear Force Normalized Shear Force
p
bo
P
(D
*
b
o o o

p
g

00
00

ffl
CD
II II
p
o b
I H
 Normalized Eccentricity Normalized Eccentricity Normalized Eccentricity

GO
GO
z
o
V)
fi)

a>
CO II II
so' II p O
o H bo b>
I
H H H
5?

 A Statistical Measure of Fit for Analytical Models
Conditioned on Experimental Results
Roger Ghanem
The Johns Hopkins University
Baltimore, MD 21218

Introduction
The general question this paper tries to answer can be phrased as follows: Given data for
some response quantity associated with an soil-structure-interaction (SSI) experiment, with
what level of confidence can any particular mathematical model be claimed to represent
the physics of the experiment ?
Mathematical models of physical phenomena usually attempt to address a very well
defined aspect in isolation of the other aspects involved. This is necessitated by the com-
plexity of nature and the restriction of our methods of inquiry to certain forms of logical
statements. These isolated aspects of nature usually represent the behavior of a given sys-
tem in a particular and isolated mode of operation. Since it is generally not possible to
isolate and examine a physical system in this way, it should be expected that predictions
from mathematical models would generally disagree with the observed behavior of the sys-
tems they purporte to model. By generalization and analogy, however, models permit us,
starting from an observed fact, to follow a certain chain of logic in order to deduce and
predict a number of additional facts. It should be remembered, though, that only the first
fact is certain, and all the others are merely probable. As noted by Poincare [4], however,
it is far better to predict with certainty than to never have predicted at all.
In addition to uncertainties regarding the applicability of a certain model to a given
situation, which is typically associated with the model having captured the physical mech-
anisms known to be at play, uncertainties are also present in identifying the parameters to
be used in that model. These parameters are usually obtained through some fitting proce-
dure. Were the model perfect, some set of parameters would produce a perfect fit between
the predictions of the model and the observed data. Given the approximate nature of the
model, however, fitting errors will necessarily be incurred in the estimated values of these
parameters. The question identified at the beginning of the paper can then be restated as
one of deducing the confidence to be attached to the model given the observed scatter, or
uncertainty, in its parameters. Implicit here is the assumption that these parameters are
the parameters that produce a best fit for the predictions of the model against experimental
data. A closely related and equally important question is that of specifying the maximum
level of uncertainty to be tolerated in the data if a decision about accepting or rejecting a
particular model is desired with a specified confidence.

11-1
General Formulation
In trying to model physical phenomena, certain intrinsic principles have to be satisfied
by any mathematical model attempting to rationalize observed behavior. Such principles
may consist of some equilibrium or conservation laws whose applicability is not questioned
within the confines of the problem at hand. Consider the set of models, denoted by S,
consisting of all models that satisfy such consistency requirements. Further, consider a
model M.(6\ ,..., Op ) S parametrized by the set (6\ ,..., Op), and denote by At s , the model
whose parameter set is equal to a specific set 6S indexed by 5. Moreover, let qM s = q(.Ms)
symbolize the functional dependence of some predictable of the model on the parameter set
6. The symbol qs will also be used when the specific model M. is clear from the context. The
quantity qs could, for instance, represent the stress, or acceleration predicted by the model
M s at some point in the domain of interest. Finally, the symbol q will be used to denote
the observed value of the physical quantity qs attempts to predict. Although the parameter
set for a given model is uncertain (i.e. can be viewed as having a probability measure
induced on it), the relationship between this set and the predictables, qs , is deterministic,
and is completely specified through the functional At. Thus, once the confidence in 6S has
been quantified, the confidence in qs is uniquely determined. A probabilistic framework will
be used to represent the uncertainty in the problem. Probabilistic inquiries of the form,
P(Q Q) will be posed, where a capital letter will be used to denote a random variable
and a calligraphic letter will be used to denote a set to which the random variable may
belong; whenever this set consists of an interval on the real line, then lower case letter will
be used to delineate the interval. Given the above definitions, the following equality holds,

PQ(O) = P(Q e Q) = P(Q(M(0)) eq) = P(0 e T) , (i)

where T denotes a subset of the of the set of possible parameters. The probability statement
on the right hand side can be deduced following the curve fitting of the observations to the
predictions of the model. The statement to the left of the equality describes the probability
of the prediction falling in the set Q when the parameters are in the set T. Although
conceptually straightforward, evaluating the probabilistic characterization of Q requires a
significant computational effort, and is typically performed using a Monte Carlo simulation
procedure. In the next section, the spectral stochastic finite element procedure will be
implemented as a more efficient way of computing the transformation from probability
measures on the parameters to probability measures on the predictions. It should be noted,
before proceeding, that the transformation between the two probability measures is uniquely
defined by the model A1 and can therefore be perceived as a characteristic of the model, in
the same way that a transfer function is a characteristic of a dynamical system. Once the
expression on the left hand side of equation (1) has been evaluated for all relevant sets Q,
however, hypothesis testing for the model given the observations can be readily performed.
Thus, let the set Qa with confidence level a be specified in the sense that

P(Q QQ ) = 1 - a . (2)
Then an observation q within this set would corroborate the mathematical model with a
confidence of at least 1 a. Clearly, the largest such a is of most value in the present

11-2
 k, jn

Xl Xg

far field near field

Figure 1: Two-Degree-of-Freedom Model for Foundation-Soil Interaction.

context, and can be used to define the confidence in the model provided by the data.

Physical Model
An often used model for soil-structure interaction consists of an elastic structure supported
by an elastic foundation of Winkler type. Particularly in the case of pile foundations, and
in an effort to provide for different soil behavior in near and far field regions, each spring in
the Winkler reaction model is sometimes replaced by a two layer model, consisting of three
masses connected by two springs. This model features the mass of the pile, that of the near
field soil, and the far field soil. The springs connecting these masses represent the different
elastic reactions that characterize the near and far field regions. Considering the pile-soil
system to be part of a shaking table experiment, whereby the base of the pile is fixed to
the shaking table, each section of the pile, along its length, oscillates with some amplitude.
Considering a slice through the soil-pile system, and neglecting interactions between this
layer and neighboring layers, the three-mass model for that layer can be perceived as being
subjected to a base excitation due to the oscillation of the pile.^ Referring all motion to a
coordinate system fixed on the pile at that layer, a two-degree-of freedom system is obtained
as shown in Figure (1). Assume damping, of various sources, to be negligible. Although an
oversimplification, the present model will serve the purpose of introducing the concepts set
forth in the present paper.

Approximation of the Uncertainty in the Predictions

The stochastic finite element method presents a general heading under which a host of
procedures have been developed for propagating the uncertainty from the model parameters
to the model predictions. In particular, the spectral formulation will be used herein [2]. A
typical equation of motion for a linear system, neglecting various damping agents, can be
formulated in the frequency domain as,

[K - u'MJ X = U (3)
where K, M, X, and U denote respectively, the stiffness matrix, the mass matrices, the
Fourier amplitude vector of the displacement and that of the excitation. In the case of a
base excitation, the input motion consists of a vector of M.I, multiplied by the Fourier

11-3
amplitude of the base acceleration. In that case, the motion X refers to the relative mo-
tion of the various masses with respect to the base motion. Clearly, equation (3) is a
representation of a mathematical model M. that attempts to represent the pile-structure
interaction. This model depends on a set of variables 0 (mi, m2, &i, 2). Assume that the
masses and stiffnesses are modeled as random variables with a known mean and variance.
In order to completely characterize the probabilistic structure of these parameters, higher
order statistics or a closed form expression for their probability distribution function would
normally be required. A Gaussian distribution will be assumed here in order simplify the
presentation. The mass and stiffness matrices can thus be rewritten as,
2 2
M = M+ Y. &M'- = E &M>- (4)
i=i i=o
and
2 2
K = K + >K,- = 5>K,-, (5)
j=l {=0

where an overbar denotes an average quantity, and f,- and 77; denote normalized uncorrelated
random variables,

<ifct> = 0. (6)

Next, the solution process X is expanded as a polynomial in the random variables describing
the material properties resulting in,
P P

i=i i=o
where the set {V'i} denotes orthogonal polynomials in the variables & and 77;, the first few
of which are given as,

,?7-l,--.} (8)

The orthogonality of these polynomial is interpreted to mean that they are uncorrelated,

<&> = 0, i>l, <Mj> = <^i>fij ViJ. (9)

Substituting the above expansions into the transformed equation of motion results in the
equation,

Equality in this last equation is construed to be in the weak sense. Thus projecting the
equation on each of the V'i used to approximate the solution and averaging, results in,

11-4
 /_ _ x / P 2 PI
(K - w2M,-) <$>Xk + <*Wk>Ki - ^ <&
\j=0 i=l j=0 i=l
2
* = 0,...,P. (11)

These are P sets of equations, each with as many degrees of freedom as the physical set-up
has. These equations can be solved for the deterministic coefficients X;. Once these have
been computed, statistical realizations of the solution process can be readily computed by
synthesizing the polynomial expansion of the solution process. From these, probability of
various events can be easily computed. Alternatively, if less probabilistic information is
enough for the problem at hand, the mean vector and the covariance matrix of X can be
obtained as,

X = X0 , (12)
and
P
Cov(X, X) = <(X - X0 )(X - X0 )T > = <^>X,-Xf , (13)

respectively.

Probabilistic measures of Closeness

Consider the quantities Xi and Xi in the above problem to constitute the set of relevant
predictables. Thus the random variable Q is a 2-dimensional vector representing the motion
of the near-field and field masses. Given the probabilistic characterization of Q, confidence
intervals and tests of hypothesis regarding the closeness of the model predictions to the
observations can be readily conducted. Thus assume that a number of experiments are
conducted and a set of n observations of Q has been made. Let the mean of the observed
set and its standard deviation be denoted, respectively, as q and sq . The standard devi-
ation of the model prediction is assumed known and can be computed according to the
procedure described in the previous section. Given the mean and standard deviation of the
predictions, and assuming they are Gaussian, a confidence interval can be constructed at
some a significance level for the mean and variance of the observations. If the measured
statistics of the observations fall within their respective intervals, then the observations
are consistent, at the specified level of significance, with the model predictions. Otherwise,
they are not.
Clearly, these confidence intervals depend on the level of uncertainty in the parameters
of the model. Thus, if a perfect fit has been achieved for the parameters, resulting in a
variance of zero, the corresponding variance of the model prediction is also zero, and the
only observation that will corroborate the model is the observation that coincides with the

11-5
prediction. This, however, is highly unlikely as the model, not being perfect, will neces-
sarily result in parameters that have a finite non-zero scatter around their nominal values.
The larger this scatter is, the more tolerant the model is of non-matching observations.
Clearly, a good model should be characterized by a narrow scatter in the computed values
of its parameters, but it is ultimately the statistical scatter in the predicted values that
controls the suitability of the model at the specified level of significance. Confidence in-
terval estimates usually relate the number of observations, to the level of significance and
the acceptable error. In their simplest form, these estimates can be developed as closed
form expressions between the various quantities involved. This, however,' is usually true
for predicted quantities having a Gaussian probability distribution, which is rarely the case
for most mathematical models. In such more realistic situations, Monte Carlo simulation
procedures must be relied upon to delineate the boundaries of the confidence domain. As
indicated above, the spectral stochastic finite element procedure is ideal for such situations
as it greatly expedites the simulation process.

Numerical Example
In the numerical example, the following numerical values were assumed for the physical
parameters involved,

h = 2k2 = 7339500N/m2 (14)

k2 = Es = 3669750N/m2 (15)
mi = 1281kg (16)
m2 = 12810kg (17)

where Es denotes the elastic Young's modulus for the soil. These numerical values corre-
spond roughly to the properties of Ottawa standard medium sand with shear wave velocity
of about 30m/s. The mass of the near field is taken to correspond to the mass of a disk of
radius 0.5m around the pile. The scatter in all these material properties is well documented
[3]. In particular, shear wave velocities are known to vary with depth, frequency, level of
compaction, and to be polluted by reflections and scatter. Figure (2) shows the shear wave
velocity measured in a shaking table experiment, at various sections throughout the depth
of the soil [5]. Accelerometers were placed at 5 different depths in the sand box, the shear
wave velocity as obtained from processing the data from all the pairs of sensors was com-
puted, and is shown in the Figure as a function of frequency. It is clear that estimated
value for V8 depends to a great extent on which two sensors are used.
Figure (3) shows the mean frequency domain response associated with four different
sets of the coefficients of variation. Figures (4) shows the coefficient of variation of the
predictions associated with these same two cases. On each of these two figures are shown,
on the same plot, the coefficients of variation obtained from first order approximation for
the solution process, along with a second order approximation. As expected, the effect of
the higher order approximation on the statistics of the solution decreases with the coefficient
of variation in the parameters. This figure indicates the level of variability in the model

11-6
predictions that would be consistent with the level of variability observed in the model
parameters. As expected, the variation in the coefficients of variation of X\ and X? as a
function of frequency follows closely that of their mean values. It is worth noting that the
tolerated error in the neighborhood of the resonant frequencies is significantly larger than
that at other frequencies. This indicates that, given the present model, while the magnitude
of the response amplitude at resonant frequencies can be afforded to deviate significantly
from their observed values, the location of these peaks along the frequency axis have a lower
margin of tolerance. Clearly, these observations are only valid for the very approximate
model used herein. Analogous conclusions regarding other models would involve producing
similar curves associated with that model.

0.0.1 Conclusions
The paper emphasized the availability, as an intrinsic part of any mathematical model ofa
physical system, of statistical diagnostic tools for assessing the closeness between observed
data and the model. It is recommended that an effort be made to develop "neighborhood
spheres" for various mathematical models currently in use for SSL

References
[1] Fares, N., and Maloof, R. (1997), "Probabilistic framework for detecting and
identifying anomalies," Probabilistic Engineering Mechanics, Vol. 12, No. 2, pp
63-73.

[2] Ghanem, R., and Spanos, P. (1991), Stochastic Finite Elements: A Spectral
Approach, Springer-Verlag.

[3] Lamb, T. and Whitman, R. (1969), Soil Mechanics, John Wiley.

[4] Poincare, H., (1905) Science et Hypothese,

[5] Tazoh, T., and Shimizu, K., "Nonlinear seismic behavior of pile foundation
structure systems," Proceedings of the 10th World Conference on Earthquake
Engineering, Madrid, Spain, 1992, pp. 1807-1810.

11-7
 09 or oz o

OOSZ 0001 0 OSZ OSI 0$ OSZ OS I OS

(P>)SA

00 001 OOZ 001 OS

( P) )SA

009 OOZ 0
(P))SA

Figure 2: Typical Shear Wave Velocities Based on Experimental Data.

11-8
 20 40 60 80
Frequency, Hz
COV kl=0.3. COV k2=0.3. COV ml=0.3. COV m2=0 3

20 40 60 80
Frequency, Hz
COV lcl=0.3. COV k2=0.3. COV ml=0.1. COV m2=0.3

20 40 60 80
Frequency, Hz
COV lcl=0.3. COV k2=0.2, COV ml=0.. COV m2=0.

Frequency, Hz
COV kl=0.3. COV k2=0.1. COV ml=0 3. COV nV2=0.

Figure 3: Mean Value of the Frequency Domain Response.

11-9
 0 20 40 60 80
Frequency, Hz
COV kl=0.3. COV k2=0.3. COV ml=0.3, COV m2=0.3

0 20 40 60 80
Frequency, Hz
COV kl=0 3. COV k2=0.3. COV ml=0.1. COV m2=0.3

Near Field
o FBI Field

O
U ^
o

20 40 60 80
Frequency, Hz
COV kJ=0.3, COV k2=0.2. COV mj=0., COV m2=0.

20 40 60 80
Frequency, Hz
COV kl=0.3. COV k2=0. J. COV ml=0.3. COV m2=0.

Figure 4: Mean Value of the Frequency Domain Response.

11-10
Uncertainties of Seismic Soil-Structure Interaction Analysis: Significance, Modeling and Examples

D. M. Ghiocel, Ph.D.
STI Technologies, Inc., Rochester, NY 14623

ABSTRACT: The aim of this paper is to (i) discuss the significance of different uncertainty sources on seismic
soil-structure interaction (SSI), (ii) review the engineering current practice for assessing SSI uncertainty effects
using probabilistic models and finally (iii) propose a new procedure for an accurate probabilistic SSI analysis.
The intention of the paper is not to address all significant SSI aspects, but only few of these which based on
author's opinion are not consistently reflected by the current state-of-engineering practice. Several shortcomings
of the current engineering practice for assessing structural risks for critical facilities are pointed out. The
proposed procedure uses for the idealization of seismic input and soil properties stochastic field models. Its
implementation offers a significant advancement for performing probabilistic seismic SSI analyses.

INTRODUCTION embedment effects on system stiffness and

vibration energy radiation
The factors influencing SSI are a myriad. This is structure nonlinear behavior, which may be
due to the complexity of seismic SSI phenomenon. more ductile or brittler, including
A short list of major factors influencing SSI may stiffness degradations and damping increase
include: local contact interface nonlinearities between
soil and foundation
Wave propagation:
control motion, including intensity, SSI response depends drastically on both the
directionality, frequency content seismic environment and structure-foundation-soil
wave composition, including internal waves, P ensemble dynamic characteristics.
and S, with surface waves, Rayleigh The seismic SSI uncertainties are usually
and Love, and other wave types divided in two major source types of uncertainties,
spatial variation of ground motion with depth namely: (i) uncertainties due to inherent
and distance, including motion randomness in natural phenomena induced by
incoherency and wave passage effects earthquakes and in material properties, and (ii)
soil nonlinear behavior as a function of shear uncertainties due to modeling uncertainties in SSI
strain in soil, soil stability models and assumptions.
To illustrate the contributions of the two types
Soil-structure interaction: of uncertainty sources, the probabilistic seismic
wave scattering effects or kinematic interaction response of a nuclear power plant (NPP) is
dynamic characteristics of structure-foundation- considered (Ghiocel et al., 1994). Figure 1 shows
soil ensemble, including the simulated in-strucrure spectra for the Reactor
Building (RB), at the basemat and the top of the
12-1
containment, and for the Auxiliary Building (AB), There are two major avenues for improving SSI
at the roof level. The SSI effects are larger for the modeling uncertainties: (i) improve deterministic
RB than for AB. The random variability in the SSI prediction models and (ii) improve probabilistic
spectral response is higher at the top elevations than models. These two avenues are discussed in the next
for the basemat. This indicates that the SSI two sections.
uncertainties are mostly propagated through the
rocking motion than through the horizontal motion. DETERMINISTIC PREDICTION MODELS
The contributions of two types of uncertainty on
seismic response of the two NPP buildings are There are many significant SSI aspects with
quantified in Figure 2. The two spectral curves significant impact on accuracy of seismic structural
correspond to coefficient of variation curves which predictions which are not appropriately considered
were computed assuming that the uncertainties are by the present state-of-engineering practice. Herein,
only few of these aspects, subjectively selected, are
addressed.

Table 1. Structural Fragility Analysis Results for the investigated NPP

Building PR Pu PC Median HCLPF

Capacity Capacity
Reactor Building 0.43 0.32 0.54 2.31g 0.70g
(Basemat Failure)
Auxiliary Building 0.33 0.29 0.44 2.50g 0.90g
(Steel Columns)
Penetration Area 0.32 0.27 0.42 1.19g 0-45g
(Concrete Wall)
Intake Structure 0.31 0.26 0.40 l-40g 0-55g
Diesel Gen. Building 0.32 0.25 0.41 1.24g 0.49g
(Concrete Slab)

due to inherent randomness in the input motion One major aspect which is a significant barrier
frequency content and soil properties, and that the against SSI modeling accuracy is the limitation of
uncertainties are due to both the inherent currently available computational tools for
randomness and the modeling uncertainties, performing efficiently rigorous nonlinear SSI
respectively. The final results of the seismic analyses, including both wave propagation aspects
probabilistic risk assessment (SPRA) calculations and soil/structure nonlinear behavior aspects. Using
for the investigated NPP are shown in Table 1. The the most currently applied computer programs such
results indicate that the two uncertainty sources, i.e. as SASSI, CLASSI, DRAIN, ADINA, ANSYS,
randomness and modeling, contribute almost ABAQUS, etc. there are strong limitations for
equally to the total seismic response uncertainty. rigorous nonlinear SSI analyses. The limitations are
They are typical for NPPs founded on soil sites and due to the computational effort, program capability
consistent with the present state-of-the-engineering and professional qualification and effort associated
knowledge and practice. The paper focus is limited with the use of different computer programs. As a
to SSI modeling uncertainties. consequence of this situation, the SSI practical
procedures include significant conservatism to
cover the simplified assumptions made. On the
12-2
 other side, simplified investigations may generate nonsymmetric structures with large mass
an uneconomical design due to higher stresses in eccentricities.
structures. One may think more seriously to the
potential savings coming from applying a more Figure 4 shows the seismic pressure computed
refined SSI analysis while designing or retrofitting a on the lateral wall of a typical, flexible buried waste
concrete highway bridge within the US. Any storage tank (WST) filled with liquid, using a
simplified conservative assumption on modeling of Beredugo-Novak lumped parameter SSI model
SSI effects, including structural/soil nonlinear (Miller and Costantino, 1994) and a SASSI model.
behavior and local wave propagation/scattering The computed pressure distribution has different
effects, may induce additional cost of millions of shapes for the two SSI models. Further, Figure 5
dollars. shows the effect of SSI by comparing the seismic
Other significant SSI aspects which need more response of an isolated WST with that of two WST
attention and more adequate consideration in the model. The two WST are identical and separated by
future engineering practice are related to the a short horizontal distance, being coupled through
evaluation of (i) torsional effects induced in the soil medium. The surface input excitation is the
structures with mass eccentricities and large size same for both comparative cases. The wave shadow
foundations due to motion incoherency, including effect (Xu et al., 1994) is visible in high frequency
wave passage effects, and structure-soil-structure range. For the two WST coupled model there is a
interaction effects between neighboring structures, significant drop in the high frequency components
especially for massive, deeply embedded or buried from the bottom tank level to the surface due to the
structures. An important modeling SSI uncertainty wave shadow effect. However for the bending
is related to the computation of seismic pressures on moments in the tank shell the wave shadow effects
embedded walls and deep foundations. Other aspect appears to be less significant.
which in practice sometime is of a great interest is Figures 6 through 8 illustrates the results
the local soil instability effects, especially computed for a Reactor Building (RB) on a pile
liquefaction, on SSI response. To highlight the SSI foundation in a relatively soft and liquefiable sand
aspects selected case study results are briefly deposit (Ghiocel et al, 1996). Figure 6 shows the
discussed as shown Figures 3 through 10. SASSI model of the RB including the pile
Figure 3 shows the in-structure spectra foundation. Figure 7 shows the free-field
computed in an axisymmetric model of a Reactor liquefaction analysis results computed using an
Building founded on a soft soil, i.e. shear wave effective-stress computer program, LASS
velocity of 1000 fps, at the basemat at the edge due (Ghabousi and Dikmen, 1977-1984), and assuming
to torsional accelerations and at the top of internal the water table located just below ground surface.
structure due to horizontal translational The liquefaction analyses indicated that there is a
accelerations. The motion incoherency was potential liquefiable sand layer between 1m and 6m
idealized using a Luco-Wong model (Luco and depths. SSI analysis was performed using the
Wong, 1986) with a coherence parameter of 0.30, equivalent soil properties computed from the free-
which corresponds to an upperbound of field analysis assuming that limited liquefaction
incoherency. For this value, the computed peak takes place between 1m and 6m depths. It was
acceleration due to torsional motion is 30% of the assumed that liquefaction starts at the beginning of
peak acceleration due to horizontal translation. The the earthquake, and that it surrounds the pile
SSI calculations were done with the ACS foundation in all horizontal directions. The
SASSI/PC computer program (Ghiocel, 1997). This assumption is very drastic, so that the corresponding
computer program is based on the original SASSI results represent an upper bound of the pile
program, but has significant additional capabilities, foundation response. As expected, the effect of
including motion incoherency and multiple liquefaction on pile forces was severe as illustrated
excitation options. Torsional motions induced by in Figure 8. There is a major redistribution in the
incoherency can have a severe effect on seismic forces and moments in the piles due to
12-3
liquefaction, which shows a large increase in the lognormal format for SPRA reviews is its
upper part of the piles, underneath basemat, where mathematical simplicity for implementation. The
liquefaction occurred. It should be noted that the lognormal distribution format is based on a
variability hi the pile axial forces is larger than in mathematical expedience by combining subjective
the pile bending moments. inputs with a multiplication scheme for fragility
evaluation.
PROBABILISTIC PREDICTION MODELS Using the lognormal format approach (Kennedy
et al., 1980), a structural fragility curve which is a
Most of probabilistic seismic SSI analyses function of hazard parameter, A, is defined in terms
currently applied in practice, usually for critical of the median capacity, A, times the product of two
facilities, use a lognormal format and base on random factors, E R andey, representing the inherent
simple technical guidelines such as those for the randomness about the median value and the
SPRA for NPP (Reed and Kennedy, 1994). These uncertainty in the median value as follows:
guidelines are a sort of modeling "recipes", in
which the effects of different SSI uncertainties are
generically quantified. The bad part of such a = Ae R 8 u (1)
simple approach is that the probabilistic modeling
may be crude and that the quantified uncertainty The two random factors are assumed to be
effects given in guidelines are based on limited lognormal random variables with median of unity
research investigations and measurements; so that and logarithmic standard deviation pR and pus
may not reflect the particularity of a SSI problem. respectively. If there is no modeling uncertainty
Because of this, such types of modeling "recipes" (only randomness) failure of probability as a
may impact sometimes negatively on the quality of function of hazard parameter is computed using the
a SSI prediction, especially when the particularity standard normal cumulative function O(.) by
of the problem is significant. It should be
understood that such simplified probabilistic
approaches with questionable modeling (2)
simplifications, which were considered as feasible
and versatile for practitioner engineers of the '80s,
when the computational resources were low and If the modeling uncertainty is included then
probabilistic modeling was in infancy, should be
replaced in the next future. Desirably, a r ln(- A / A exp[pO-'
-^ (p)])
^
probabilistic SSI analysis has to accurately [P f > P (3)
Pu
determine, by itself, the effects of different
uncertainties for a specific SSI problem and not to which computes the probability for which the
assume them. Several criticisms of the actual state- failure probability p f exceeds p given hazard
of-engineering practice are discussed in this section. parameter value A (Kennedy et al.,1980). Using the
Before doing this, a brief review of the lognormal lognormal format, the probabilistic dynamic
format is presented (Kennedy el., 1980). structural response for a hazard parameter reference
level and probabilistic structural capacities are
Lognormal Format expressed as products of different factors (Kennedy
et al., 1980, Reed and Kennedy, 1994). The basic
Lognormal format has been used extensively in two properties of lognormal model are (i) the
the past two decades for developing seismic median of a product of lognormal distributed
structural fragilities for critical facilities including random variables is equal to the product of the
SSI effects. At this time the lognormal format is the medians and (ii) the logarithmic standard deviation
most popular probabilistic format in engineering
practice. One of the main reasons for using
12-4
of a product is the square root of sum of squares of the probabilistic SSI analysis, but introduces a
the individual logarithmic standard deviations. significant modeling uncertainty due to the highly
nonlinear relationship between in-structure spectral
In the early '80s, it has been considered by peaks and the soil stiffness. This nonlinear
engineering experts that the accuracy of the relationship is shown in Figure 10. The use of a
probabilistic distribution in the region of fragility single deterministic SSI analysis for computing the
curve tails is not essential for a SPRA. median response for the reference level earthquake
Unfortunately, this is true only if the major risk (other questionable concept for simplifying the
contributors correspond to hazard parameter values probabilistic analysis) may introduce artificially
far away from probability distribution tails, high spectral peaks.
especially from the lower tail (Kennedy et al., The last criticism discussed herein is related to
1980). More recently, comparative studies have the computation of the structural capacity using the
indicated that the lognormal assumption for lognormal format and the SPRA guidelines when
probability distribution applied in conjunction with significant SSI effects are present. The use of
multiplicative models for structural response and inelastic energy absorption factors computed using
structural capacity may produce crude results for the methods suggested in the SPRA guidelines is
risk assessments (Hwang et al., 1987). In drastically unconservative. Those methods are
conjunction with the lognormal format, the use of a calibrated for fixed-base structure without SSI
single reference level of the hazard parameter effects. If the fixed-base inelastic factors are
(assumed to be representative for the median combined with SSI results the overall safety margin
structural capacity) for performing the SSI analysis, is incorrectly estimated, as the inelastic absorption
which has been applied in practice for expediency, and SSI effects are not independent and their
may introduce significant inaccuracies. This is combination must take into account their
especially true when significant risk contributors dependencies. Otherwise, a double-counted margin
come from random events defined by hazard is introduced. This should be avoided when the
parameters sensibly lower than reference level. A Newmark modified spectra method or the spectral
significant drawback of lognormal model is that the averaging method derived for fixed-base structures
median capacity is insensitive to modeling are combined with SSI effects. The changes in
uncertainty (Ellingwood, 1994). natural frequencies and structural damping due to
One criticism of the lognormal format and its structural nonlinear effects have considerably less
application based on the SPRA guidelines is related impact on a structural system having a large
to probabilistic definition of seismic motion effective damping (including SSI energy radiation
frequency content. As suggested by SPRA phenomenon), than on a fixed-base structure with
guidelines, the coefficient of variation of the low damping which is very sensitive to changes in
spectral shape of a Uniform Hazard Spectrum its frequency and damping. Kennedy et al., 1985,
(UHS) varies in the range of 0.25-0.30, which is based on a limited number of cases, gives two
typical for the Newmark-Hall spectrum type for the different expressions for the median inelastic factor,
WUS, but less appropriate for the UHS type for the
BUS. Figure 9 illustrate a typical UHS of a BUS
NPP. It should be noted from this figure that the FM - 1) + 1 for fixed-base
coefficient of variation of spectral shape varies in
the range of 0.80-1.00, which is far larger than that structures (4)
recommended by the SPRA guidelines.
Another criticism is related to typical including SSI (valid
applications of the lognormal format in conjunction for a stiff soil) (5)
with SPRA guidelines for computing in-structure
response spectra using the so-called "median output where j^s is the story drift ductility factor. For
to median input" rule. This rule largely expedites example, for a story ductility factor j^s =5.0 the
12-5
 inelastic absorbtion factor is 1.8 for fixed-base cases response variability. However, the LLNL study did
and only 1.3 for SSI cases. For a NPP structure not involve any probabilistic methodology. The
founded on a softer soil the difference is obviously BNL study focused on nuclear containment
larger. Further research on this important aspect is structures using linear random vibration theory to
needed. calculate limit state probabilities under random
The above criticisms are only a few of the many seismic loads. The BNL departed from the current
possible criticisms and, probably not the severest format suggested by the SPRA guidelines for NPP
ones against the current state-of-engineering (Reed and Kennedy, 1994). However, the BNL
practice for probabilistic SSI analysis, more specific methodology is restricted to superficial rigid
for NPP practice. These methodological circular foundations on a visco-elastic half-space.
deficiencies can be improved if the lognormal For realistic situations including arbitrary shaped
format is applied in conjunction with extensive and/or flexible foundations, partially embedded or
simulations using random sampling techniques. buried structures, oblique seismic waves, non-
This implies significant higher costs of probabilistic uniform soil layering the BNL methodology is not
SSI analysis due to larger computational and human directly applicable. The proposed probabilistic
effort. It also requires highly qualified engineers on approach rigorously addresses these aspects.
both SSI modeling and probabilistic modeling. A A significant advantage of the proposed
cost-effective alternative is to use specialized probabilistic approach is that the loads and system
computer programs with user friendly interface for parameters can be more accurately described by
performing probabilistic SSI. In this idea, a random fields (time-space variability) rather than
probabilistic approach is proposed in this section random variables (point variability) as assumed in
(Ghiocel and Ghanem, 1999). This approach was the current SPRA reviews. Earthquake motion and
recently implemented and applied in conjunction soil properties are properly represented by random
with SASSI models for SSI computations. fields (Ghiocel, 1996, Ghiocel et. al 1996).
The above discussions suggest the need of a The key idea of the proposed approach is to
case-by-case type of probabilistic SSI methodology provide a global complete description of the
and compuational tool, capable of including the stochastic system response surface. The proposed
particularity of a SSI problem. Specifically, SSI approach has two implementation steps. The first
effects coming from deep embedment, structure- step involves an expeditious condensation of the
soil-structure interaction, motion incoherency for basic random processes via the KL expansion. The
large size foundations, local soil instability should second step evaluates the coefficients of a stochastic
be more carefully considered in engineering orthogonal polynomial expansion of system
practice. As shown in this paper each of these response. After the coefficients of polynomial
effects may significantly affect seismic SSI expansion are obtained, simulation of points on the
response. system response surface can be obtained. Finally,
probabilistic structural risks can be directly
Proposed Approach evaluated once the expansion of stochastic response
surface is calculated.
The most extensive studies performed in the Using Karhunen-Loeve (KL) expansion (Loeve,
past on probabilistic (seismic) SSI, which were 1977, Ghanem and Spanos, 1991) a continuous
supported by Nuclear Regulatory Commission random property field, p(x,6), is expanded
(NRC), are those performed by Lawrence according to equation
Livermore National Laboratory (LLNL, Shieh et al.,
1985) and by Brookhaven National Laboratory
(BNL, Pires et al, 1985). The LLNL study was (6)
based on a large number of case studies with the
where 6 denotes the random dimension, p.
aim of identifying the most significant variables for
seismic SSI effects and their influence on structural represents a certain scale of fluctuation of the field p
12-6
 while ,; represents its random magnitude and hence variable can be represented by a single term. In
the random contribution of that particular scale to physical systems, it can be expected that material
the overall property field. Both the property field properties vary smoothly at the scales of interest in
and its various scales are global quantities and most applications, and therefore only a few terms in
depend on the spatial position x, they can also be the KL expansion can capture most of the
multi-variate quantities. In the case where the uncertainty in the process. It should be noted that hi
material property in question is a random variable, comparison with other series representations, the
the above sum is reduced to a single term. KL expansion has the minimum number of terms,
The KL expansion of a stochastic process or in other words the minimum number of random
e(x,0), is based on the spectral expansion of its variables for random field decomposition.
covariance function R^foy). Here, x and y are For seismic SSI problems, of a particular
used to denote spatial coordinates, while the interest are positive random fields, such as the
argument 0 indicates the random nature of the amplitude of as a function of frequency or soil
corresponding quantity. The covariance function stiffness and hysteretic damping profiles as
being symmetrical and positive definite, by functions of depth, which are positive quantities.
definition, has all its eigenfunctions mutually Thus, a new development consisting of a
transformed space KL expansion was used for
orthogonal, and they form a complete set spanning
the function space to which e(x,6) belongs. It can representing the positive non-normal random fields.
The basis of this development is to find a mapping
be shown that if this deterministic set is used to
between the positive non-normal random field and
represent the process e(x,6), then the random
an associated normal random field (Grigoriu, 1997).
coefficients used in the expansion are also
In particular, the treatment of lognormal processes
orthogonal. The expansion then takes the following is particularly expeditious given a number of
form analytic expressions that are available regarding it.
For SSI response, the covariance function is not
known apriori, and hence the KL expansion cannot
be used to represent it. Since the SSI solution
process is a function of the material properties and
where e(x) denotes the mean of the stochastic seismic input, the entries of the nodal response
vector c can be formally expressed as a nonlinear
process, and {^(6)} form a set of orthogonal
functional of a set (^ (6)} used to represent the
random variables. Furthermore, {<j>j(x)} are the
material and seismic input stochasticity. It has been
eigenfunctions and {A,;} are the eigenvalues, of the
shown that this functional dependence can be
covariance kernel, and can be evaluated as the
expanded in terms of polynomials in gaussian
solution to the following integral equation
random variables, referred to as Homogeneous (or
Polynomial) Chaoses (Cameron, 1947).
(8) The expansion of SSI response takes on the
following form (Ghanem and Spanos, 1991):

where A denotes the spatial domain over which the

process e(x,0) is defined. The most important u(x,t,6) = a0 (x,t)ro + 2,a, (x,t)r,(^, (6))
aspect of this spectral representation is that the
spatial random fluctuations have been decomposed (9)
into a set of deterministic functions in the spatial
variables multiplying random coefficients that are
independent of these variables. The closer a process In equation 9, the symbol Fn (j ,...,, j )
is to white noise, the more terms are required in its denotes the Homogeneous (or Polynomial) Chaos
expansion, while at the other limit, a random
12-7
(Kallianpur, 1980, Wiener, 1938) of order n in the Polynomial Chaoses are orthogonal with respect
variables (.,...,;). Introducing a one-to-one to the Gaussian probability measure, which makes
mapping to a set with ordered indices denoted by them identical with the corresponding
(ViCG)} and truncating the Homogeneous (or multidimensional Hermite polynomials (Grad,
1949). From the above equation it is obvious that
Polynomial) Chaos expansion after the pfll term, the integration domains spans a large
equation 10 can be rewritten as multidimensional space, the dimensionality being
given by the number of elementary standard normal
random variables defining the set{^}. The
(10) multidimensional integral given in equation (11)
j=o
can be computed using various integration
procedures including Gauss-Hermite quadrature or
These polynomials are orthogonal in the sense
efficient simulation techniques. For actual
that their inner product < xj/jij/ k > , which is
integration an innovative stratified sampling
defined as the statistical average of their product, is technique was employed. An alternate approach
equal to zero for j * k. A complete probabilistic using advanced stochastic finite element concepts is
characterization of the solution process u(x,t,6) is described elsewhere (Ghiocel and Ghanem, 1999).
obtained once the deterministic coefficients For getting a faster convergence in the case of
uj (x,t)have been calculated. A given truncated non-normal processes, a transformed space
series can be refined along the random dimension representation of non-normal processes was used.
either by adding more random variables to the set Therefore, a logarithmic transformation was applied
{j} or by increasing the maximum order of at the level of the extreme responses before
expanding it in a Homogeneous Chaos. Then the
polynomials included in the Homogeneous (or
expansion was performed in a transformed space for
Polynomial) Chaos expansion. The first refinement
which the corresponding process is closer to a
takes into account higher frequency random
normal process. Finally, the non-normal process
fluctuations of the underlying stochastic process,
was determined using an inverse transformation,
while the second refinement captures strong non-
specifically an exponential transformation. This
linear dependence of the solution process on this
transformation is expressed mathematically by
underlying process (Ghanem and Spanos, 1991).
Using the orthogonality property of
polynomials, the coefficients of the Homogeneous (13)
Chaos of the solution process can be computed by <vj/

This significantly has speeded up the

fork=l,...K convergence and has improved the accuracy of the
computed series expansions for extreme-value
One of the key factors for obtaining an efficient responses.
numerical implementation of the stochastic
approach based on Homogeneous Chaos expansion Earthquake Motion Description
is related to the computation of the inner products or
averages <\j/ k u> in equation 11. This can be Earthquake ground acceleration was represented
by a segment of a (non)stationary random process
rewritten in an explicit integral form
(nonstationarity was introduced by using a
deterministic intensity shape function) with zero
mean, known frequency content and spatial
(12)
correlation structure. This stochastic representation
is conditional to the given zero-period peak ground
12-8
 acceleration (ZPGA) level. For evaluation of the (GD/GD )4
overall seismic structural risk all the ZPGA levels, (15)
i.e. the seismic hazard curve at the site, should be
considered. For each ZPGA level, the frequency
content of earthquake motion is described locally, in (iii) Ruiz-Penzien spectrum multiplied by a low-
a point at ground surface, by either a acceleration pass first-order filter and
probabilistic response or a power spectral density
function. The three earthquake motion components
S(G3) =
were assumed to be statistically independent. The
spatial correlation structure of ground motion field, (16)
which is a function of frequency, was defined by a
coherency spectrum matrix.
Local (Point) Description: Typically in
engineering practice probabilistic site-specific (iv) Brookhaven National Lab (BNL) spectrum
ground response spectra were defined for hazardous
facilities (LLNL, 1993, EPRI, 1991). The
probabilistic ground spectra are usually described
by three digitized spectral response curves
computed for 15%, 50% and 85% non-exceedance
probability assuming a lognormal distribution of where GO f , GO p and f ,4 p are the frequency and
amplitudes. Herein, the probabilistic ground bandwidth of the filters.
spectrum was assumed as an one-dimensional These analytical forms are widely accepted by
lognormal random field in frequency domain with the earthquake engineering community, being the
certain bandwidth characteristics given by the soil most popular ones. The Kanai-Tajimi spectrum, (i),
deposit behavior as a second-order linear filter for was the first of the above expression to be proposed.
incoming seismic waves. The spectral amplitude The Ruiz-Penzien spectrum, (ii), was intended to
field was modeled by a lognormal random field adjust the low frequency content of Kanai-Tajimi
using a transformed KL expansion. As an alternate spectrum at frequency equal to zero. The improved
of local description of earthquake ground motion, Ruiz-Penzien spectrum, (iii), reduces the high
the power spectral density may be input instead of a frequency content of the Kanai-Tajimi spectrum.
probabilistic spectra. Four analytical expressions The BNL spectrum, (iv), has a lower high frequency
were considered for the power spectral density content than the Kanai-Tajimi spectrum removes
(Piresetal., 1985): the singularity of the displacement spectral power
density at zero frequency (Pires et al., 1985).
(i) Kanai-Tajimi spectrum (spectral shape similar Spatial. Variation (Incoherency): For an
to the acceleration transfer function of single incoherent wave field the unlagged coherence for
degree of freedom subjected to a base two point motions i and k can be defined as
excitation) (Abrahamson et al, 1990):

CohUik ((D) = Cohitk (o>) A(io>,X, -Xk )

S(CO) = (14) exp[i(D(X Di -X Dk )/VXDi_XDk ] (18)

(ii) Ruiz-Penzien spectrum where A(ico,Xj-X k ) is a decaying function of

frequency starting from unit value which gives the
relative power of the wave field described by a
S((D) = plane wave at all frequencies. The term
exp[i(D(X Di -XDk )/VXDi_XDk ] in equation 18
12-9
represents in the frequency domain the phase angle (co) = C oh(|Xj - Xk |,co) =
between the two point motions due to the wave (20)
passage effect along the direction D. Parameter
VXDi-XDk is the apparent seismic wave velocity
defined by the projected distance between the two in which y is the coherence parameter and Vs is the
points along the direction D. If the wave field is shear wave velocity in the soil. The above analytical
perfectly described by a single plane wave, the expression compared with others given in the
function A^cOjX; - X k ) is equal to unity. technical literature based on experiment fitting
For two one-dimensional random time series (Hoshiya and Ishii, 1983, Harichandran and
representing an unidirectional seismic motion Vanmarcke, 1986, etc.) has the advantage of a
components in two arbitrary points of the soil theoretical support based on the analytical
deposit, j and k, the (narrow band) coherence is formulation of shear wave propagation in random
defined by a complex function of frequency media (Uscinski, 1977). Luco and Wong, 1986,
suggested that the coherence parameter has generic
values in the range of 0.10 to 0.30.
S u (co)
Coh. k () = 11/2 (19)
(ii) Abrahamson model (Abrahamson, 1991,
1993), defined by

where S jk (co) is the cross-spectral density CohClXj- Xk |,fi>) =

ik
function for two points j and k, and 8^(00) is the Tanh{(al + a 2 |X; - Xk |)[exp[- (21)
auto-spectral density for point j (similar for point
k). The coherence describes the similarity of the two
point motions. Generally, in engineering
applications, the so-called "lagged" coherency where al, a2, bl, b2 and c are model parameters.
spectrum or "lagged " coherence are used These parameters can be introduced by the user,
(Abrahamson et al., 1990). The lagged coherency otherwise by default the values (Abrahamson, 1990)
includes only the amplitude randomness and are used, i.e. al=2.55, a2=-0.012, bl=0.115,
removes the wave-passage randomness. From b2=0.00084, c=0.878 and k=0.35. These parameters
physical point of view, the lagged coherence may be defined as random variables.
represents the fraction of the total power of seismic Assuming that the seismic wave field can be
motion which can be idealized by a single modeled by a plane wave, an element of the cross-
deterministic plane wave motion called the coherent spectral density matrix of multidimensional motion
motion. Usually in the current earthquake random field can be derived analytically
engineering language, the lagged coherence is
called simply coherence. More generally than the ,k (co) = [S ; . (co)S k k (co)] I/2 Coh(|xi - Xk |,co)
"lagged" coherence, the "unlagged" coherence (22)
includes the wave-passage random effects. exp[ico(XDi -XDk )/Vxi_Xk ]
Based on the experimental evidence of different
records of past earthquakes, the following analytical for each pair i,k of point motions.
forms for the coherence function were considered: To implement the random field model of
incoherent soil motion, the coherence matrix is
(i) Luco-Wong model (Luco and Wong, 1986), decomposed via KL expansion. The motion
defined by incoherency effects are are larger for higher
frequency components than for lower frequency
components. Usually, the effect of incoherency is to

12-10
reduce translational motion and rocking motion and appropriately a set of independent random variables,
increase torsional motion. than by a continuous random field with a well-
established correlation structure expandable in a KL
Soil Property Description series.

Soil properties were assumed to be Example Application

homogeneous in a horizontal plane and therefore
they were idealized as one-dimensional random The proposed approach was applied to a typical
fields, i.e. random varying profiles with depth. Reactor Building (RB) subjected to earthquake
Specifically, the randomness in soil dynamic motion. The probabilistic SSI response was
properties was considered by variabilities in shear compared with a deterministic SSI response
modulus, hysteretic damping and Poisson ratio. computed using the current practice for NPP. The
First, the soil deposit was discretized in a geometric finite element model used for seismic soil-structure
layering with varying properties. Soil shear interaction analysis is shown in Figure 11 (Lysmer
modulus at low strains, G max , was idealized as an et al., 1988). This SSI computational model
one-dimensional lognormal random field in the represents a typical SASSI model for seismic design
vertical direction having a non-stationary mean and basis calculations of a reactor building. The
an assumed correlation length for same material superstructure is modeled by beam elements and the
type. basemat is modeled by solid elements. Rigid links
This idealization is considered to be are introduced to transmit the rocking motion from
significantly more realistic and less conservative the superstructure stick to the basemat. The ACS
than the assumption of perfect correlation currently SASSI computer program (Ghiocel, 1997) was used
applied for parametric deterministic SSI studies. For for both the free-field analysis and the SSI analysis,
soil layering including different materials, a set of performed either probabilistically or
multiple random fields may be considered. The deterministically.
shape (nondimensional variation) of the shear Deterministic analysis was done for a seismic
modulus - shear strain curve, G(y)/G max -y was input defined by the design ground spectrum
modeled by a random field along the shear strain associated to a 84% probability of nonexceedance.
axis with a non-stationary mean. The mean curve A spectrum compatible accelerogram was generated
was assumed to have an arbitrary shape which is for SSI analyses. As shown in Figure 12 the
either Introduced by the user or by default stored in computed response spectra of the generated
the program database. The same modeling accelerogram slightly envelopes the given design
assumption used for the shear modulus curve was spectrum. Soil properties were be the best-estimate
considered for the hysteretic damping - shear strain values (median). In accordance to the current
curve, D(y). seismic design requirements, two additional
For implementation, the soil property fields extreme bounds, 0.50 times best-estimate and 2.00
were decomposed via KL expansion. The statistics times best-estimate values were considered. The
of the soil property field models, including final results of the deterministic analysis are
correlation length parameters, were derived by obtained by enveloping the results for the three soil-
calibrating the mathematical models with structure interaction analysis for the three set of
experimental data available. values of soil parameters.
For probabilistic analysis, the earthquake input
Structural Properties was defined by a probabilistic response ground
spectrum as shown in Figure 13. The four spectral
Structure damping and stiffness parameters were curves corresponds to mean, median and 16% and
assumed to be random variables. This assumption is 85% nonexceedance probability estimates. The
based on the fact that the random variation of these probability distribution was assumed to be
parameters within the superstructure are
12-11
 lognormal. The lognormal spectral amplitude field between 73 and 144 are the coefficients of the
was represented using a transformed KL expansion. second-order polynomials (without coupling). It is
The correlation length along frequency axis was to be noted that only less than half of the number of
selected depending on the desired bandwidth of basic random variables have significant
simulated spectra (function of damping). The contributions. Larger contributions come from
number of frequency steps to describe the spectral linear terms than from nonlinear terms. However, it
shape was 100. The smaller the correlation length is very difficult for the complex soil-structure
is, the narrower the spectral peaks are. Figure 14 problem to preliminary establish with are the most
illustrates the ensemble statistics (for significant variables. There is a need to get more
nonexceedance probabilities of 15%, 50%, 85% and insights on this aspect in the future.
mean) of the probabilistic model of ground response Figure 22 shows a comparison between
spectrum for a set of 100 realizations. Few deterministic and probabilistic analysis results, both
simulated realizations are shown in Figure 15.For in terms of in-structure response spectra.
probabilistic soil-structure interaction analysis the Deterministic estimates corresponds to very low
effect of motion incoherency was considered using nonexceedance probability levels. Having in mind
a Luco-Wong model with a y parameter of 0.20. the additional conservatism, introduced in the
The resulted spatial variation of motion amplitude overall seismic evaluation by the seismic hazard
for different frequencies is plotted in Figure 16. definition and the evaluation of structural elements
Soil properties were defined assuming that the or equipment capacities, it appears that the current
low strain soil shear modulus and hysteretic deterministic SSI analysis procedure is overly
damping profiles (variation with depth) are conservative.
lognormal random fields. Figure 17 shows the
probabilistic shear modulus profile (statistically CONCLUDING REMARKS
estimated profiles are included). Plotted curves
correspond to mean, median and 16% and 84% The paper addresses the effects of SSI modeling
nonexceedance probability. A transformed space uncertainty on seismic response, discusses
KL expansion was used to represent these shortcomings of current state-of-engineering
lognormal positive fields. The variation of practice on probabilistic SSI for hazardous facilities,
nondimensional shear modulus and hysteretic and proposes a new accurate procedure for
damping versus shear strain were modeled as performing probabilistic SSI anatysis. SSI modeling
normal random fields decomposable directly in uncertainty effects are illustrated using the results
original space in KL expansion. Simulated from different case studies. The proposed procedure
variations are shown in Figure 19. represents a significant advancement for performing
Structural properties are described using random probabilistic seismic SSI analyses of hazardous
variables. Specifically, the Young elastic modulus facilities.
and the material damping ratio were assumed to be The proposed approach based on a stochastic
normal random variables with a coefficient of series representation of SSI response offers
variation of 0.25. The means were assumed to be accuracy, efficiency and significant modeling
0.80 of the linear elastic modulus and 8%, advantages in comparison with the currently SPRA
respectively. approaches. The proposed approach addresses
A comparison of probabilistic response efficiently large number of variables problems such
computed using the proposed approach (using 100 as dynamic SSI problems and handles random field
solutions) and a Monte Carlo simulation (using 500 models, useful for idealization of dynamic loading
solutions) is shown in Figure 20. and system parameters. In addition, the proposed
Figure 21 shows the coefficients of the approach is capable of handling large variability and
transformed Homogeneous Chaos expansion using highly nonlinear problems.
72 basic random variables. Between 1 and 72 are
the coefficients of the first-order polynomials, and
12-12
ACKNOWLEDGEMENTS Probabilistic Mechanics and Structural
Reliability, Worcester, MA, August.
The paper is partially based upon work Ghiocel, D.M. et al. (1996) Probabilistic Seismic
supported by the National Science Foundation Analysis Including Soil-Structure Interaction,
under a SBIR project with the Award Number DMI- 7th ASCE MD/STD Joint Speciality Conference
9660321. on Probabilistic Mechanics and Structural
Reliability, Worcester, MA, August.
SELECTED REFERENCES Ghiocel, D. M. (1997) ACS SASSI/PC - An
Advanced Computational Software for System
Abrahamson, N. A. et al. (1990). Spatial Variation Analysis of Soil-Structure Interaction on
of Strong Ground Motion for Use in Soil- Personal Computers, ACS Report, Cleveland,
Structure Interaction Analyses, 4th U.S. Ohio.
National Conference on Earthquake Ghiocel, D. M., Wilson, P., Stevenson, J. D.,
Engineering, Palm, Springs, Vol.1. (1995). Evaluation of Probabilistic Seismic FRS
Cameron, R. and Martin W (1947). The Including SSI Effects, the 13th SMIRT
Development of Nonlinear Fuctionals in Series Conference, Invited Paper, Vol. M, Porto
of Fourier-Herrnite Functionals, Ann. Math, 48, Alegre.
385-392. Ghiocel, D.M., et al., (1990) "Evaluation of SSI and
EPRI (1989). Probabilistic Seismic Hazard SSSI Effects on Seismic Response of Nuclear
Evaluation at Nuclear Plant Sites in the Central Heavy Buildings by Different Approaches", the
and Eastern United States: Resolution of 10th European Conference on Earthquake
Charleston Issue, Report NP-6395-D. Engineering, Moscow.
Ghaboussi, J. and Dikmen, S.U (1977). LASS II - Ghiocel, D. M., et al.,(1987) "Structural Reliability
Computer Program for Analysis of Seismic of Prestressed Concrete Nuclear Containment
Response and Liquefaction of Horizontally Under Combined Loads", the 9th SMIRT, Vol.
Layered Sands, Report UILU-ENG-79-2010. M, Laussan.e
Ghanem, R., Ghiocel, D.M.(1996). A Comparative Ghiocel, D. M.,(1986) "Probabilistic Seismic Soil-
Analysis of FORM/SORM and Polynomial Structure Interaction Analysis", 8th ECEE
Chaos Expansion for Highly Nonlinear Systems Conference, Lisbon
ASCE/EMD Speciality Conference, Ft. Hwang et al. (1987). Seismic Probabilistic Risk
Lauderdale. Assessment for Nuclear Power Plants, NCEER-
Ghanem, R., and Spanos, P. (1991). Stochastic BNL-Report 87-0011.
Finite Elements: A Spectral Approach, Kallianpur, G. (1980). Stochastic Filtering Theory,
Springer- Verlag. Springer Verlag.
Ghiocel, D. M. and Ghanem, R. (1999). A Kennedy, R.P, Cornell, C.A., Campbell, R.D.,
Stochastic Finite Element Approach for Seismic Kaplan,S., Perla, H.F. (1980). Probabilistic
Soil-Structure Interaction, Journal of Earthquake Seismic Safety Study of An Existing Nuclear
Engineering and Structural Dynamics, ASCE Power Plant. Nuclear Eng. & Des., Vol. 59.
(submitted). Kennedy, R.P., Kincaid, R.H., Short, S.A. (1985) .
Ghiocel, D.M. and Ghanem, R. (1998). Stochastic Engineering Characterization of Ground
Seismic SSI Using Homogeneous Chaos Motion. Task II, NUREG/CR-3805, Vol. 2.
Expansion, the 12th ASCE Engineering LLNL (1993) Eastern US Seismic Hazard
Mechanics Speciality Conference, San Diego, Characterization Update, Lawrence Livermore
May. National Laboratory Report UCRL-ID-115111
Ghiocel, D.M. (1996) Seismic Motion Incoherency Loeve, M.,(1948) Probability Theory, McGraw-
Effects on Dynamic Response, 7th ASCE Hill.
EMD/STD Joint Speciality Conference on Lysmer, J.,Tabatabaie-Raissi, Tajirian, F.Vahdani,
S., and Ostadan, F.(1988), SASSI - A System
12-13
 for Analysis of Soil - Structure Interaction, Including Soil-Structure Interaction, USNRC
Report UCB 1988, Geotechnical Engineering, ReportNUREG/CR-4329, BNL-NUREG-5190.6
University of California, Berkeley. Reed , J.W, Kennedy, R.P. 1994. Methodology for
Luco, J. and Wong, H. L. (1986). Response of a Developing Seismic Fragilities.
Rigid Foundation to a Spatially Random Shieh, L.C., et al. 1985. Simplified Seismic PRA:
Ground Motion, Earthquake Engineering & Procedures and Limitation, NUREG/4331,
Structural Dynamics. UCID-20468.
Miller, C.A. and Costantino, C.J. (1994). " Seismic Xu, J., Bandy opadhy ay, K., Miller, C.A.,
Induced Pressures in Buried Vaults", Natural Costantino, C.J. (1994) "Spacing Effects on
Hazard Phenomena and Mitigation, ASME, Seismic Responses of Underground Waste
PVP-Vol.217,p.3-U. Storage Tanks", Natural Hazard Phenomena and
Pires, J., Hwang, H., and Reich, M., (1985). Miitigation^SAffi, PVP-Vol 217, p. 13-18.
Reliability Evaluation of Containments Wiener, N (1938). The Homogeneous Chaos, Amer.
J. Math, 60, p. 897-936.

DMGijah
D: \ghiocel\ssiawk98. doc

12-14
 BASH MAT ELEVATION 8.5 FT AUXILIARY STRUCTURE-DIR.N-S- SIMULATIONS
CONTAINMENT SHELL ELEVATION 163.0 FT ELEVATION 117.5 FT - EAST OF SF POOL
1.00

0.80

0.40

0.20

0.00
0.1 1 10
0.1 1 10

FREQUENCY(Hz) FREQUENCY(Hz) FREQUENCY(Hz)

a) RB, basemat b) RB, top on containment c) AB, roof level

to
I
Figure I. In-structure spectra in the RB and AB

CONTAINMENT SHELL - ELEVATION 163.0 FT Br Be

0r Be

1.00
1.00
|--!-y EFFECT OF SPECTRAL
:
COMPOSITE VARIATION SyTj AMPLITUDE-SOIL STIFFNESS 0.80
0.80 ...:.../....i..: NONLINEAR RELATIONSHIP
'.::'.: co MPOS f EV AR iAfidN"""^^-i-i : ^n:
0.60
0.60 'Jf ""." ^^.^.^.. ....... .;...;. .!..:...;..;.
..
^m- Trrrrr"
-\^^ ^
0.40 ! >..
0.40 ^ ^<y>^_v^^^;^^ . ..Vv^
...... .. . . : : : ; ! : : :
i
0.20 *' RAN
0.20
DOMN E SS VA RIATION .;:::!::;:r /-
:::?! RANDOMNESS VARIATION i:::;::;:ti ssi EFFECTS ARE LARGER ;;;;;;;::;|;;:::j:':i::j::j:;:..s SIEFF ECTA RE SMALLER
0.00
o.oo
0.1 10
0.1 10
FREQUENCY(Hz)
b) AB, roof level
a) RB, top of containment
Figure 2. Coefficients of variation; for the in-structure spectra
FRS - TOP OF THE INTERNAL STRUCTURE (RB) FRS - BASEMAT LEVEL DIRECTION Y (RB)
SOFT SOIL - VsMOOO FPS SOFT SOIL - Vs-1000 FPS
1.60

INCOH-H

INCOH-L

100
100
FREQUENCY (Hz)
FREQUENCY fHz^
b) Basemat, torsional motion
a) Top of internal structure, horizontal motion
Figure 3. Effect of incoherency on in-structure spectra
Buried Radioactive Waste Storage Tank
Comparative SSI Results for Va-1000 fps
2.00

0.00
! 6 10 14 18 22 26 30 34

Depth (ft) from Vault Top Level,El.-16ft

Figure 4. Seismic pressure on WST wall

 Buried Radioactive Waste Storage Tank Buried Radioactive Waste Storage Tank
Comparative SSI Results for Vs-1000 fps\ Comparative SSI Results (or Vs-1000 (ps
1.00
Tank TOP Tank Bottom
1.00

0.80
Coupled Coupled
0.61

0.41

Isolated
0.22 "
Isolated

0.02 0 02
0.5 1 10 0 5 1

Frequency (Hz) Frequency (Hz)

to
I a) Surface motions b) Bottom tank motions
Figure 5. Response spectra for the WST model

TIME:

Stick Model
SSI Model

Basemat

STR,10.sec
Sidewalls
-120 PO,10.sec

150
0.50 3.40 6.30 9.20 12.10 15.00 DEPTH (m)
Piles

Figure 6. Structure-pile foundation SSI model Figure 7. Pore pressure and effective stresses in soil
 REACTOR BUILDING - NO LIQUEFACTION REACTOR BUILDING - LIQUEFACTION
MAXIMUM AXIAL FORCES IN PILES (SRSS) 0.40
MAXIMUM AXIAL FORCES IN PILES (SPSS)
300
,,.. ...
0.30
.... _;
r' _
- MAX ... . *
<
S 0.20
..j..]..
X
AVG ' ' r. m\ i *
'*' ' ..;.....

-! ;
^. 0.10

i i* ^
,.. -|4- _.__.
n
0.00
20248

DEPTH(ffl)
Frequency (Hz)

REACTOR BUILDING NO LIQUEFACTION REACTOR BUILDING - LIQUEFACTION

MAXIMUM BENDING MOMENT IN PILES (SRSS) MAXIMUM BENDING MOMENT IN PILES (SRSS) Figure 9. Typical UHS for a BUS NPP site
300 BOO

700
Linear Median
600
' MAX
Spectral Amplitude

AVG, /

200

100 s
0
2 0 2 4 6 B 10

DEPTH(m)

f(G-) f(Gm) f(G+) G- Gm

Figure 8. Forces and moments in piles

Figure 10. Highly nonlinear relationship between

spectral peaks and soil stiffness
CO
I

Figure 11. RB SSI model (after Lysmer et al., 1988)

DETERMINISTIC (DESIGN) GROUND SPECTRUM PROBABILISTIC GROUND SPECTRUM (Damp.-5%
ZPGA-0.20g, Damping-5% 84% Probability Statistical Estimations Logn. Dlstrlb.
COMPUTED DESIGN MEAN MEDIAN 16% 84%

1.00
1.00

\
\

0.00
0.00
0.5 1 10
0.5 1

Frequency (Hz) Frequency (Hz)

Figure 12. Deterministic ground response spectra Figure 13. Probabilistic ground response spectra
 Lognormal Dlstr. - 50 VARIABLES, COV-45% Lognormal Random Field - KL Exp. (100)

1.00 1.00
I
0.83

CD - 6/
"* * ^\r-
..../TT-T
/

0 :\

7
^
0.50
s

u
u -^
/-
.^- \
< 0.33 \
'/ / : ; :^
0.17
.\^/* S:
,^^^:\ :
0.00 o.oo
0 .1 1 10 0.1 1 10

to Frequency(Hz) Frequency (Hz)

I
to
o Figure 14. Statistical estimate of probabilistic spectra Figure 15. Simulated ground ground spectra

REACTOR BLDG. BASEMAT REACTOR BLDG. BASEMAT

FREQUENCY - 1.0 Hz REACTOR BLDG.BASEMAT
FREQUENCY 12.0 Hz
FREQUENCY- 18 0 Hz
NO MOTION INCOHERENT NO MOTION - INCOHERENT NO MOTION INCOHERENT

100

-100
-145 87 29 29 -31 31 31 31 155
DIRECTION X DIRECTION X DIRECTION X

Figure 16. Motion incoherency effects on input

 Go Probabilistic Profiles Simulated Go Profiles - KL Expansion
Statistical Estimation Logn. Dlstr. Lognormal Dlatr. - 20 VARIABLES, COV-40%

10

O
o

is
o o
O JC

5
<D
V)

100 100

Figure 17. Statistics and realizations of low strain

shear modulus profile

Statistical Shear Modulus Curves,Geff/Go Statistical Damping Ratio Curves,Deff

Normal Random Field 11 variables Normal Random Field - 11 variables

1.00

Vv
0.80
ss\
0.60
v\\

0.40
A

0.20

0.00
-4 -2

Shear Straln(exponent of 10) In percent Shear Straln(exponent of 10) In percent

Figure 18. Statistical curves for variations of soil
properties with shear strain
 Simulated Shear Modulus Curves,Geff/Go Simulated Damping Ratio Curves,Deft
Normal Random Field 11 variables Normal Random Field 11 variables
1.00

0.80

0.60

0.40

0.20

0.00
-3 -2 -1 0

Shear Straln(Exponent of 10) In percent Shear Straln(Exponen1 of 10) In percent

Figure 19. Simulated curves for variations of soil
properties with shear strain
to TOP OF CONTAINMENT-MCS(500) VS THCE(100)
I
to ACCELERATION AT 2.00 HZ FREQUENCY Containment Shell - Base Moment
to
MCS(SOO) HCE(100) Transformed Chaos Expansion Coefficients
2.00
10
9
B 1.80
Q.
8 CO

7
1.20
6
5
0.80
4
3
0.40
2
1
0.00
10 12 24 36 48 60 72 84 06 106 120 132 144

ACCELERATION (G) Number of Coefficient

Figure 20. Comparison of probabilistic results Figure 21. Coefficient of chaos expansion
 TOP OF CONTAINMENT SHELL Deterministic TOP OF INTERNAL STRUCTURE- Deterministic
Variation ol Soil Properties, Damping-2% Variation of Soil Propertlea. Damping-2%

ro 0.1 1 10
I
ro
to Frequency (Hz)

Top of Internal Structure Probabilistic

Top of Containment Shell Probabilistic " - - ...Damping-2%
Damping-2%

8 : : : ::l
3.50
7
>;!; '! : '! (!
3.00
6
-A - MEAN
a 2.SO
MEAN

a s . 1 .......
MEDIAN
s
MEDIAN
i}
-;. A""""
2.00
15%

1\
2 4 15% \
o 1.50
I 3 85%
85%

1 k ^ ht v . ....
\ ^s*5sU
->-sS: ^
1.00

0.50 -*
r v
^i -^
^ HV..
MM

M
^ n-
MM*

0
j^ 0.00
0.1
*m^^m^ff1*
' 10
0 .1 1 10

Frequency (Hz)
F reque nc) (><z)

Figure 22. Probabilistic vs. deterministic in-structure

response spectra
 SEISMIC SOIL-FOUNDATION-STRUCTURE INTERACTION IN BRIDGES

By Wen S. Tseng1 and Joseph Penzien1

ABSTRACT: Discussed herein are state-of-the-art methodologies used in assessing soil-

foundation-structure interaction (SFSI) in bridges during seismic events, including (1) generating
free-field ground-motion inputs, (2) determining foundation impedances and "scattered" input
motions, (3) conducting global demand analyses of the complete soil-foundation-structure system,
and (4) performing foundation capacity analyses for assessing foundation system performance.

INTRODUCTION It is because of this most recent experience that

the importance of coupled soil-foundation-structure
Prior to the 1971 San Fernando, California interaction (SFSI) on the dynamic response of bridge
earthquake, nearly all damages to bridges during structures during earthquakes has been fully realized.
earthquakes were caused by ground failures, such as In treating this problem, two different methods have
liquefaction, differential settlement, slides, and/or been used (1) the so-called "elasto-dynamic" method
spreading; little damage was caused by seismically developed and practiced in the nuclear power
induced vibrations. Vibratory response considerations industry for large foundations, and (2) the so-called
had been limited primarily to wind excitations of large "empirical p-y" method developed and practiced hi
bridges, the great importance of which was made the offshore oil industry for pile foundations. Each
apparent by failure of the Tacoma Narrows suspension method has its own strong and weak characteristics,
bridge hi the early 1940s, and to moving-loads and which generally are opposite to those of the other,
impact excitations of smaller bridges. thus restricting their proper use to different types of
The importance of designing bridges to withstand bridge foundation. By combining the models of these
the vibratory response produced during earthquakes two methods hi series form, a hybrid method has
was revealed by the 1971 San Fernando earthquake, been developed by Tseng and Penzien (1998) which
during which many bridge structures collapsed. makes use of the strong features of both methods,
Similar bridge failures occurred during the 1989 Loma while minimizing their weak features. While this
Prieta and 1994 Northridge, California earthquakes, hybrid method may need some further development
and the 1995 Kobe, Japan earthquake. As a result of and validation at this time, it is fundamentally sound;
these experiences, much has been done recently to thus, it is expected to eventually become a standard
improve provisions in seismic design codes, advance procedure hi treating seismic SFSI of large bridges
modelling and analysis procedures, and develop more supported on different types of foundation.
effective detail designs, all aimed at insuring that newly
designed and retrofitted bridges will perform FREE-FIELD SEISMIC GROUND MOTIONS
satisfactorily during future earthquakes.
Unfortunately, many of the existing older The first step hi conducting a seismic performance
bridges in the U.S.A. and other countries, which are evaluation of a bridge structure is to define the seismic
located in regions of moderate to high seismic input to the coupled soil-foundation-structure system.
intensity, have serious deficiencies which threaten In a design situation, this input is defined hi terms of
life safety during future earthquakes. Because of this the expected free-field motions hi the soil region
threat, aggressive actions have been taken hi surrounding each bridge foundation. It is evident that
California, and elsewhere, to retrofit such unsafe to precisely characterize such motions is practically
bridges so as to bring their expected performances unachievable within the present-state-of-knowledge of
during future earthquakes to an acceptable level. To seismic ground motions. Therefore, it is necessary to
meet this goal, retrofit measures have been applied to use a rather simplistic approach hi generating such
the bridge superstructures, piers, abutments, and motions for design purposes. The procedure most
foundations. commonly used for designing a large bridge is to (1)
generate a three-component (two horizontal and
vertical) set of accelerograms representing the free-field
'Principal and Senior Principal, International Civil ground motion at a "control point" selected for the
Engineering Consultants, Inc., 1995 University Avenue.
Suite 119, Berkeley, CA 94704, USA bridge site, and (2) characterize the spatial variations of

13-1
 the free-field motions within each soil region of interest apparent horizontal wave velocity and (b) ground
relative to the control motions. motion coherency functions; therefore, the seismic
The control point is usually selected at the surface input motions developed for the various pier-
of bedrock (or surface of a firm soil stratum in case of a foundation locations need to be compatible with the
deep soil site), referred to herein as "rock outcrop", at values specified for these two additional parameters.
the location of a selected reference pier; and the free- Having specified the design seismic ground-
field seismic wave environment within the local soil motion parameters, the steps required in establishing
region of each foundation is assumed to be composed the pier-foundation location-specific seismic input
of vertically propagating plane shear (S) waves for the motions for a particular bridge are: (1) develop a three-
horizontal motions and vertically propagating plane component (two horizontal and vertical) set of free-
compression (P) waves for the vertical motions. For a field rock-outcrop motion time-histories which are
bridge site consisting of relatively soft top soil deposits compatible with the design target ARS and associated
overlying competent soil strata or rock, the assumption design ground motion parameters applicable at a
of vertically propagating plane waves over the depth of selected single control-point location at the bridge site
the foundations is reasonably valid as confirmed by (these motions are referred to herein simply as the
actual field downhole array recordings. "response-spectrum-compatible" time-histories of
The design ground motion for a bridge is normally control motion), (2) generate response-spectrum-
specified in terms of a set of parameter values compatible time-histories of free-field rock-outcrop
developed for the selected control point which include a motions at each bridge pier-support location such that
set of target acceleration response spectra (ARS) and a their coherencies relative to the corresponding
set of associated ground motion parameters for the components of the response-spectrum-compatible
design earthquake, namely (a) magnitude, (b) source- motions at the control-point and at other pier-support
to-site distance, (c) peak ground (rock-outcrop) locations are compatible with the wave passage
acceleration (PGA), velocity (PGV), and displacement parameters and the coherency functions specified for
(PGD), and (d) duration of strong shaking. For large the site (these motions are referred to herein as
important bridges, these parameter values are usually "response-spectrum-and-coherency-compatible" rock-
established through regional seismic investigations outcrop motions), and (3) carry out free-field site
coupled with site-specific seismic hazard and ground response analyses for each pier-support location to
motion studies; whereas, for small bridges, it is obtain the time-histories of free-field soil motions at
customary to establish these values based on generic specified discrete elevations over the full depth of each
seismic study results such as contours of regional PGA foundation using the corresponding response-spectrum-
values and standard ARS curves for different general and-coherency-compatible free-field rock-outcrop
classes of site soil conditions. motions as inputs.
For a long bridge supported on multiple piers
which are in turn supported on multiple foundations Rock-Outcrop Motions At Control-Point
spaced relatively far apart, the spatial variations of Location
ground motions among the local soil regions of the
foundations need also be defined in the seismic input. Given a prescribed set of target acceleration
Based on the results of analyses using actual earthquake response spectra (ARS) and a set of associated design
ground motion recordings obtained from strong motion ground motion parameters for a bridge site as described
instrument arrays, such as the El Centre differential previously, the objective herein is to develop a three-
array in California and the SMART-1 array in Taiwan, component set of time-histories of control motion that
the spatial variations of free-field seismic motions have (1) provides a reasonable match to the corresponding
been characterized using two parameters, namely, (a) target ARS and (2) has time-history characteristics
apparent horizontal wave propagation velocity (speed reasonably compatible with the other specified
and direction) which controls the first-order spatial associated ground motion parameter values. In the
variations of ground motion due to the seismic wave past, several different methods have been used for
passage effect and (b) a set of horizontal and vertical developing rock-outcrop time histories of motion
ground-motion "coherency functions" which quantifies compatible with a prescribed set of target ARS,
the second-order ground-motion variations due to including (1) the response-spectrum compatibility
scattering and complex three-dimensional wave time-history adjustment method, (2) the source-to-
propagation (Abrahamson, 1992). Thus, in addition to site numerical model time-history simulation method,
the design ground motion parameter values specified (3) the multiple actual recorded time-history scaling
for the control motion, characterizing the design method, and (4) the connecting accelerogram
seismic inputs to long bridges needs to include the two segments method. At the present time, the first of
additional parameters mentioned above, namely, (a) these four methods is considered most suitable and

13-2
 practical for bridge engineering applications, currently used are the apparent horizontal seismic wave
particularly the method developed by Lilhanand and speed, V, and its direction angle relative to an axis
Tseng (1988) which is based on earlier work by Kaul normal to the longitudinal axis of the bridge. Studies of
(1978). In this case, small local perturbations are strong- and weak-motion array data including those in
added in a systematic manner to the initial or starting California, Taiwan, and Japan show that the apparent
acceleration time history at those times horizontal speed of S-waves in the direction of
corresponding to the occurrences of the response propagation is typically in the 2-3 km/sec range (Chang
spectral values. As a result, the phasing et al., 1986 and Abrahamson, 1992). In applications,
characteristics (wave sequence or pattern) in the the apparent wave-velocity vector showing speed and
initial, or starting, time history are largely direction must be projected along the bridge axis giving
maintained. It is important therefore that the initial the apparent wave speed in that direction. To be
time history be selected carefully. realistic, when becomes small, a minimum angle for
Each three-component set of starting 0, say 30 degrees, should be used in order to account
accelerograms for a given bridge site should preferably for waves arriving in directions different from the
be a set recorded during a past seismic event that has specified direction.
(1) a source mechanism similar to that of the The spatial coherency of the free-field components
controlling design earthquake, (2) a magnitude within of motion in a single direction at various locations on
about 0.5 of the target controlling earthquake the ground surface has been parameterized by a
magnitude, and (3) a closest source-to-site distance complex coherency function defined by the relation
within 10 km of the target source-to-site distance. The
selected recorded accelerograms should have their
PGA, PGV, and PGD values and their strong shaking i,j =1,2,..., n locations (1)
durations within a range of 25% of the target values
specified for the bridge site and they should represent
free-field surface recordings on rock, rock-like, or a in which Sy (/CD) is the smoothed complex cross-
stiff soil site; no recordings on a soft site should be power spectral density function and SH (co) and Sjj (co)
used. For a close-in controlling seismic event, e.g. are the smoothed real power-spectral density (PSD)
within about 10 km of the site, the selected functions of the components of motion at locations i
accelerograms should contain a definite velocity pulse
andy. The notation /co in the above equation is used to
or the so-called "fling". When such recordings are not
indicate that the coefficients S^(KO) are complex-valued
available, Method (2) described previously can be used
to generate a starting set of time histories having an (contain both real and imaginary parts) and are
appropriate fling or to modify the starting set of dependent upon excitation frequency co. Based on
recorded motions to include the desired directional analyses of strong motion array data, a set of generic
velocity pulse. coherency functions for the horizontal ,and vertical
Having selected a three-component set of starting ground motions has been developed (Abrahamson, et
time histories, the horizontal components should be al., 1991).
transformed into their principal components and the Given a three-component set of response-
corresponding principal directions should be evaluated spectrum-compatible time histories of rock-outcrop
(Penzien and Watabe, 1975). These principal motions developed for the selected control-point
components should then be made response-spectrum location and a specified set of wave-passage parameters
compatible using the time-domain adjustment and "target" coherency functions as described above,
procedure described above or the standard frequency- response-spectrum-compatible and coherency-
domain adjustment procedure (Hao, Oliviera, and compatible multiple-support rock-outcrop motions
Penzien, 1989; Silva and Lee, 1987; and Bolt and applicable to each pier-support location of the bridge
Gregor, 1993). Using the latter procedure, only the can be generated using the "marching method"
Fourier amplitude spectrum, not the phase spectrum, is developed by Hao, Oliveira, and Penzien (1989) and
adjusted iteratively. extended later by Tseng, Lilhanand, and Yang (1993).
For a long bridge located close-in to the
Rock-Outcrop Motions At Bridge Support controlling seismic source, attenuation of motion with
distance away from the control-pier location should be
Locations
considered. This can be achieved by scaling the
Characterization of the spatial variations of rock- generated motions at various pier locations by
outcrop motions for engineering purposes is based on a appropriate scaling factors determined from an
set of wave passage parameters and ground motion appropriate ground motion attenuation relation. The
coherency functions. The wave passage parameters acceleration time histories generated for all pier

13-3
locations should be integrated to obtain their been published by Vucetic and Dobry (1991) for clay
corresponding velocity and displacement time histories, and by Sun, Golesorkhi and Seed (1988), and the
which should be checked to ensure against having Electric Power Research Institute (EPRI) for sand. The
numerically-generated baseline drifts. Relative shear modulus is given in its nondimensional form
displacement time-histories between the control-pier G/Gmax where Gmax is the in-situ shear modulus at very
location and successive pier locations should also be low strains ( y < lO'Vo). The shear modulus G must be
checked to ensure that they are reasonable. The rock obtained from cyclic shear tests, while Gmax can be
outcrop motions finally obtained should then be used in obtained using Gmax = pVs2 in which p is mass density
appropriate site response analyses to develop the of the soil and Vs is the in-situ shear wave velocity
corresponding free-field soil motions required in obtained by field measurement.
conducting the SFSI analyses for each pier location. For generating vertical free-field motions produced
by vertically propagating compression waves, the
Soil Motions at Bridge Support Locations needed soil parameters are the low-strain constrained
elastic modulus, Ep = pVp2, where Vp is the compression
The seismic inputs to large bridges are defined in wave velocity, and the corresponding damping ratio.
terms of the expected free-field soil motions at discrete The variations of these soil parameters with
elevations over the entire depth of each foundation. compressive strain have not as yet been well
Such motions must be evaluated through location- established. At the present time, vertical site response
specific site-response analyses using the corresponding analyses have generally been carried out using the low-
previously-described rock-outcrop free-field motions as strain constrained elastic moduli, Ep, directly and the
inputs to appropriately-defined soil/bedrock models. strain-compatible damping ratios obtained from the
Usually, these models are based on the assumption that horizontal response analyses, but limited to a maximum
the horizontal and vertical free-field soil motions are value of 10%, without any further strain-compatibility
produced by upward/downward propagation of one- iterations. For soils submerged in water, the value of
dimensional shear and compression waves, Ep should not be less than the compression wave
respectively, as caused by the upward propagation of velocity of water.
incident waves in the underlying rock or firm soil Having generated acceleration free-field time
formation. Consistent with these types of motion, it is histories of motion using the SHAKE computer
assumed that the local soil medium surrounding each program, the corresponding velocity and displacement
foundation consists of uniform horizontal layers of time histories should be obtained through single and
infinite lateral extent. Wave reflections and refractions double integrations of the acceleration time histories.
will occur at all interfaces of adjacent layers, including Should unrealistic _drifts appear in the displacement
the soiltoedrock interface, and reflections of the waves time histories, appropriate corrections should be
will occur at the soil surface. Computer program applied. Should such drifts appear in a straight line
SHAKE (Schnabel et al, 1972, Idriss and Sun, 1991) is fashion, it usually indicates that the durations specified
most commonly used to carry out the above-described for Fourier transforming the recorded accelerograms
one-dimensional type of site-response analysis. For a are too short; thus, increasing these durations will
long bridge having a widely varying soil profile from usually correct the problem. If the baseline drifts
end to end, such site response analyses must be depart significantly from a simple straight line, this
repeated for different soil columns representative of the tends to indicate that the analysis results may be
changing profile. unreliable, in which case, they should be carefully
The cyclic free-field soil deformations produced at checked before being used. Time histories of free-field
a particular bridge site by a maximum expected relative displacement between pairs of pier locations
earthquake are usually of the nonlinear hysteretic form. should also be generated and then be checked to judge
Since the SHAKE computer program treats a linear the reasonableness of the results obtained.
system, the soil column being analyzed must be
modelled in an equivalent linearized manner. To obtain SOIL-FOUNDATION INTERACTION
the equivalent linearized form, the soil parameters in
the model are modified after each consecutive linear Basic to solving the seismic SFSI problem for .a
time-history response analysis is complete, which bridge is the interaction between the combined
continues until convergence to strain-compatible structure-foundation system and its supporting soil
parameters are reached. In generating horizontal free- medium, which, for analysis purposes, can be
field motions produced by vertically propagating shear considered to be a full half-space. The fundamental
waves, the needed equivalent-linear soil parameters are step in solving this problem is to characterize the
the shear modulus G and the hysteretic damping ratio constitutive relations between the dynamic forces
P. These parameters, as functions of shear strain y have acting on each foundation of the bridge at its interface
13-4
boundary with the soil and the corresponding obtained for foundations having relatively simple
foundation morions, expressed in terms of the soil/foundation interface geometries, e.g. rectangular,
displacements, velocities, and accelerations. Such cylindrical, or spherical soil/foundation interface
forces are called herein the "soil-foundation interaction geometries, supported in simple soil media. In practical
forces". For a bridge subjected to externally applied applications, the dynamic Green's functions are often
static and/or dynamic loadings, such as dead, live, obtained in numerical forms based on a finite element
wind, and wave loadings, these soil-foundation discretization of the half-space soil medium and a
interaction forces are functions of the foundation corresponding discretization of the soil/foundation
morions only; however, for a bridge subjected to interface boundaries using a computer program such as
seismic conditions, they are functions of the free-field SASSI (Lysmer et al. 1981), which has the capability of
soil morions as well. properly simulating the wave radiation boundary
Let h be the total number of degrees-of-freedom conditions at the far-field of the half-space soil
(DOF) of the bridge's foundations as defined at their medium. The use of finite element soil models to
soil/foundation interface boundaries; A (0> *(') and evaluate the dynamic Green's functions in numerical
form has the advantage that foundations having
iih(t) be the corresponding foundation displacement, arbitrary soil/foundation interface geometries can be
velocity, and acceleration vectors, respectively; and easily handled; it, however, suffers from the
uh (t), uh (t) and uh (t) be the free-field soil disadvantage that the highest frequency, i.e. cut-off
displacement, velocity, and acceleration vectors in the h frequency, of motion for which a reliable solution can
DOF, respectively; and lety/, (?) be the corresponding be obtained is limited by size of finite element used for
soil-foundation interaction force vector. Using these modelling the soil medium.
notations, characterization of the soil-foundation Having evaluated the dynamic Green's functions
interaction forces under seismic conditions is usually using the procedure described above, the desired soil
expressed in linear (or equivalent linear) form as shown impedance matrix can then be obtained by inverting,
by the matrix equation frequency-by-frequency, the "soil compliance matrix",
which is me matrix of Green's function values
(2)
evaluated for each specified frequency to. Because the
dynamic Green's functions are complex-valued and
frequency-dependent, the coefficients of the resulting
in which matrix GM(/co) is a complex, frequency- soil impedance matrix are also complex-valued and
dependent coefficient matrix called herein the "soil frequency-dependent The real parts of the soil-
impedance matrix". Depending upon the type of impedance coefficients represent the dynamic
foundation, this matrix and the associated free-field soil stiffnesses of the soil medium which also incorporate
input motion vector uh(to) to be used for seismic the soil inertia effects; the imaginary parts of the
demand analysis purposes can be established utilizing coefficients represent the energy losses resulting from
different soil models as described below. both soil material damping and radiation of stress
waves into the far-field soil medium. Thus, the soil
Elasto-Dynamic Model impedance matrix as developed reflects the overall
dynamic characteristics of the soil medium as related to
For a large bridge foundation such as a large the motion of the foundation at the soil/foundation
spread footing, caisson, or single or multiple shafts interfaces.
having very large diameters, for which the Because of the presence of the foundation
nonlinearities occurring in the local soil region excavation cavities in the soil medium , the vector of
immediately adjacent to the foundation are small, the free-field soil motions uh (i&) prescribed at the
soil impedance matrix GM(/CD) can be evaluated soil/foundation interface boundaries has to be derived
utilizing the dynamic Green's functions (dynamic from the seismic input motions of the free-field soil
displacements of the soil medium due to harmonic medium without the foundation excavation cavities.
point-load excitations) obtained from the solution of a The derivation of the motion vector iTA(/eo) requires
dynamic boundary-value problem of a linear damped- the solution of a dynamic boundary-value problem for
elastic half-space soil medium subjected to harmonic the free-field half-space soil medium having foundation
point-loads applied at each of the h DOF on the excavation cavities subjected to a specified seismic
soil/foundation interface boundaries. Such solutions wave input such that the resulting solution satisfies the
have been obtained in analytical form for a linear stress-free conditions at the surfaces of the foundation
damped-elastic continuum half-space soil medium by excavation cavities. Thus, the resulting seismic
Apsel (1979). Because of complexities in the analytical
solution, dynamic Green's functions have only been response motions, A(ioo), reflect the effects of seismic

13-5
wave scattering due to the presence of the cavities. designated herein as G^A (0) to differentiate it from the
This scattering effect is much more important for a
large bridge foundation, such as a large caisson or a soil impedance matrix G'A (/GO) defined previously.
group of very-large diameter shafts, man it is for a
foundation having small characteristic dimensions, The construction of the "p-y", "t-z," and "Q-d"
such as a slender-pile group; and, it is more significant curves depends only on the strength parameters but not
for foundations supported in soft soil sites than in stiff the stiffness parameters of the soil medium; thus, the
soil sites. effects of global soil nonlinearities on the dynamic
The characterization of the soil impedance matrix stiffnesses of the soil medium, as caused by soil-shear-
utilizing the elasto-dynamic model of the soil medium modulus decrease and soil-damping increase as
as described above requires soil material functions of free-field-motion-induced soil shear
characterization constants which include (a) mass strains, can not be incorporated into the soil impedance
density, p, (b) shear and constrained elastic moduli, G matrix developed from these curves. Furthermore,
and Ep (or shear and compression wave velocities, Vs since these curves are developed on the basis of results
and V^, (c) and constant-hysteresis damping ratio, p. from field tests in which there are no free-field ground-
As mentioned previously, the soil shear modulus motion-induced soil deformations, the effects of such
decreases, while the soil hysteresis damping ratio global soil nonlinearities on the soil strength
increases as functions of soil shear strain induced in the characterization parameters and hence the "p-y", "t-z,"
free-field soil medium due to the seismic input and "Q-d" curves can not be incorporated.
motions. The effects of these so-called "global soil
nonlinearities" can be easily incorporated into the soil Because of the small cross-sectional dimensions of
impedance matrix based on an elasto-dynamic model slender piles, the seismic wave scattering effect due to
by using the free-field-motion-induced strain- the presence of pile cavities is usually negligible; thus,
compatible soil shear moduli and damping ratios as the the scattered free-field soil input motions UA(/G>) in this
soil material constants in the evaluation of the dynamic case are often taken to be the same as the free-field soil
Green's functions. For convenience of later motions when the cavities are not present.
discussions, the soil impedance matrix, 6^(10),
characterized using an elasto-dynamic model will be Hybrid Model
denoted by the symbol G'hh (ico) . From the discussions in the above two sections, it
is clear that characterization of the soil-foundation
Empirical "p-y" Model interaction forces for demand analysis purposes can be
achieved using either an elasto-dynamic model or an
For a slender-pile group foundation for which soil empirical "p-y" model for the soil medium, each of
nonlinearities occurring in the local soil regions which has its own merits and deficiencies. The elasto-
immediately adjacent to the piles dominate the behavior dynamic model is capable of incorporating soil inertia,
of the foundation under loadings, the characterization damping (material and radiation), and stiffness
of the soil resistances to pile deflections has often relied characteristics; and, it can incorporate the effects of
on empirically derived "p-y" curves for lateral global soil nonlinearities induced by the free-field soil
resistance and "t-z" and "Q-d" curves for axial motions in an equivalent linearized manner. However,
resistance. For such a foundation, the characterization it suffers from the deficiency that it does not allow for
of the soil impedance matrix needed for demand easy incorporation of the effects of local soil
analysis purposes can be made by using the secant nonlinearities. On the contrary, the empirical "p-y"
moduli derived from the commonly used nonlinear "p- model can properly capture the effects of local soil
y", "t-z," and "Q-d" curves. Since the development of nonlinearities in an equivalent linearized form; but, it
these empirical curves has been based upon static or suffers from the deficiencies of not being able to
pseudo-static test results, it does not incorporate the soil properly simulate soil inertia and damping effects, and
inertia and material damping effects. Thus, the it can not treat the effects of global soil nonlinearities.
resulting soil impedance matrix developed from the Since the capabilities of the two models are mutually
secant moduli of the "p-y", "t-z," and "Q-d" curves complimentary, it is logical to combine the elasto-
reflects only the static soil stiffnesses but not the soil dynamic model with the empirical "p-y" model in a
inertia and soil material damping characteristics. series form such that the combined model has the
Hence, the soil impedance matrix so obtained is a real- desired capabilities of both models. This combined
valued constant coefficient matrix applicable at the zero "hybrid" model is presented in more detail in the
frequency (co = 0); it, however, is a function of the publication by Tseng and Penzien (1998).
foundation displacement amplitude. This matrix is

13-6
 DEMAND ANALYSIS PROCEDURE in which Kijt Cy, and My are the standard stiffness,
damping, and mass matrices, respectively, which would
The seismic response of a complete bridge system appear in the system's equations of motion if expressed
involves interactions between the structure and its in the time domain. For a constant-hysteresis-damped
supporting foundations and between the foundations linear system, the impedance matrices are given by
and their surrounding soil media. To develop the
equations of motion governing the response of this i,j = s,f,g (5)
system in discrete (finite element) form, let s denote the 'J 'J'O

number of degrees-of-freedom (DOF) in the structure,

excluding its / DOF at the structure/foundation in which Jf*. is a complex stiffness matrix obtained by
interface locations, and let g denote the number of
DOF in the foundations, also excluding the / DOF but assembling individual finite-element matrices Kt{m) of
including the h DOF at all soil/foundation interfaces the form
as defined previously. Corresponding with those DOF,
let vectors us (/), uf (/), and ug (/) contain the total-
displacement time histories of motion at the DOF s,f,
and g, respectively.
where A(m) denotes the standard elastic stiffness matrix
Linear Modelling for finite element m as used in the assembly process to
obtain matrix K^ and p(m) is a damping ratio specified
Since the soil medium surrounding all foundations appropriately for the material used in finite element m
is continuous and of infinite extent, a rigorous model of (Clough and Penzien, 1993).
a complete bridge system must contain stiffness and The hysteretic form of damping represented in Eq.
damping coefficients which are dependent upon the (5) is the more appropriate form to use for two reasons
excitation (or response) frequencies. Such being the (1) it is easy to accommodate different damping ratios
case, the corresponding equations of motion of the for the different materials used in the system, and (2)
complete system having n DOF (n = s +f+ g) must the resulting modal damping is independent of
rigorously be represented in the frequency domain. excitation (or response) frequency o>, consistent with
Considering the coupled structure-foundation test evidence showing that real damping is indeed
system as a free-free (no boundary constraints) system essentially independent of this frequency. As noted by
having externally applied forces -//, (t) acting in the h the form of Eq. (4), viscous damping is dependent upon
DOF, its equations of motion can be expressed in the frequency CD, contrary to test results; thus, preference
frequency-domain form should definitely be given to the use of hysteretic
damping for linear systems which can be. solved in the
frequency domain. Hysteretic damping is
unfortunately incompatible with solutions in the time
a (3) domain.
Vector -fh(io>), which makes up the last h
components in force vector fg(i) appearing in Eq.
in which us (/co), / (/co), ug (/co), and fg (/co) are the (3), represents, as defined previously, the internal soil-
Fourier transforms of vectors us (t), / (t), ug (t), and foundation interaction forces at the soil/foundation
fg(t\ respectively; and matrices D^-O'to) , i,j = s,f,g, are interfaces when the entire coupled soil-foundation-
structure system is responding to the free-field soil
the corresponding impedance (dynamic stiffness) input motions. Therefore, to solve the SFSI problem,
matrices. The g components in vectors ug (/co) and this vector must be characterized in terms of the
fg(i(a) are ordered such that their last h components foundation displacement vector HA(J<O) and the free-
make up vectors /,(/co) and -fh (/co), respectively, with
all other components being equal to zero. field soil displacement vector uh(i&) . As discussed
For a viscously-damped linear structure- previously, for demand analysis purposes, this vector
foundation system, the impedance matrices D^KD) are can be linearized to the form
of the form
-A(fa>) = Ghh (iG>){uh(ia>)-uh(iG>)} (7)

D,(/CO) = + iG>Ct. -co 2M& i,j-s.f,g (4)

13-7
in which -fh (/CD) represents the force vector acting on in its linear form with the full viscous damping matrix
the foundations from the soil medium and Ghh (ia)) is C being expressed in the Rayleigh form
the soil impedance matrix which is complex-valued and
C = (11)
frequency-dependent.

Substituting Eq. (7) into Eq. (3), the equations of in which M and K are the full mass and elastic-stiffness
motion of the complete bridge system become matrices shown in Eq. (10) and CLK and fiR are constants
assigned numerical values which will limit the modal
damping ratios to levels within acceptable bounds over
a range of modal frequencies dominating the seismic
response.
For a time-domain solution of Eq. (10) in its
0 modified nonlinear form, all parameters in the equation
must be real (no imaginary parts) and frequency
in which independent. It remains therefore to modify the soil
impedance matrix Gy,/,(/co) so that when introduced into
Eq. (7), the inverse Fourier transform of -fh (/co) to the
(to) J" 1 ;7(to) _ time domain will yield a vector - fh (/) having no
|_0 Gh .(i(>>)] ' g frequency dependent parameters. To accomplish this
objective, separate GM (o) into its real and imaginary
Vector Gh (i(d)uh (to) is the free-field soil "seismic parts in accordance with
driving force" vector, in which the free-field soil
displacements in vector UA (/CO) result from scattering (12)
of incident seismic waves propagating to the bridge site
as explained previously. in which G^,(a>) and G^(co) are real functions of co.
Then approximate these functions using the relations
Nonlinear Modelling

When large nonlinearities develop in the structure- G*(co) = (13)

foundation subsystem during a seismic event,
evaluation of its performance requires nonlinear where the real constants in matrices KM, , MM , and
modelling and analysis in the time domain. In mis
case, the standard linear equations of motion of the Chh are assigned numerical values to provide best fits
complete system as expressed by to the individual frequency-dependent functions in
matrices G^(co)and G^(co) over the frequency range
0 of major influence on seismic response. Typically,
applying these best fits to the range 0 < co < 4
M Mfg -^ fg
-,? radians/second, corresponding to the range 0 </< 2
0 -/* Hz, where /= co/27t, is adequate for most large bridges.
(10) In this fitting process, it is sufficient to treat MM as a
MO 0 diagonal matrix, thus affecting only the diagonal
*M fft 0 functions in matrix GM (co) . The reason for selecting
VO L/W
the particular frequency-dependent forms of Eqs. (13)
is that when they are substituted into Eq. (12), which in
turn is substituted into Eq. (7), the resulting expression
must be modified appropriately to characterize the
nonlinearities for use in a step-by-step numerical for fh(i(ri) can be Fourier transformed to the time
solution. Usually, it is the third term on the left-hand domain yielding
side of this equation which must be modified to
represent the nonlinear-hysteric force-deformation
behavior taking place in the individual finite elements
of the system. The second term in this equation, . Substituting -fh (f) given by this equation for the last h
representing viscous damping forces, is usually retained components in vector^ (f), with all other components

13-8
in fg (0 being equal to zero, and then substituting the the time domain using Eq. (15). In this case, the
resulting vector^ (f) into Eq. (10) gives coupled nonlinear equations of motion are solved using
standard step-by-step numerical integration procedures.
M This one-step direct approach is simple and
straight forward to implement for a structural system
supported on a single foundation, such as a building.
However, for a long multiple-span bridge supported on
many independent foundations, a very large system of
equations and an associated very large number of
a V
seismic free-field inputs in vector i/K (/co) result,
making the solution computationally impractical,
especially when large nonlinearities are present in the
equations of motion. In this case, it is desirable to
simplify the problem by finding separate solutions to a
in which set of smaller problems and then combine the solutions
in steps so as to achieve the desired end result. The
multiple-step substructuring approach described
nn subsequently is ideally suited for this purpose.
0 M
Multiple-Step Substructuring Approach
showing that no frequency-dependent parameters
remain in the equations of motion, thus allowing the For long bridges supported on multiple
standard time-domain solution procedure to be used for foundations, the support-separation distances are
sufficiently large so that each foundation subsystem can
solving them. Usually, the terms CK ug(t) and be treated as being independent of the others; therefore,
Mgg ug(t) on the right-hand side of Eq. (15) have the soil impedance matrix for each foundation will be
small effects on the solution of this equation; however, uncoupled from those of the other foundations. In this
the importance of their contributions should be case, to simplify the overall problem, each foundation
checked. Having modified the third term on the left- subsystem can be analyzed separately to obtain a
hand side of Eq. (15) to its nonlinear hysteretic form, boundary impedance matrix called the "foundation
the complete set of coupled equations can be solved for impedance matrix" and a consistent boundary force
displacements us (t\ uf (f), ug (t) using standard step-by- vector called the "foundation driving-force vector",
step numerical integration procedures. both of which are associated with the DOF at its
structure/foundation interface. Having obtained the
foundation impedance matrix and associated driving
SOLUTION PROCEDURES
force vector for each foundation subsystem, all such
matrices and vectors can be combined into the
One-Step Direct Approach
equations of motion for the total structure as a free-free
In this approach, the equations of motion are system, resulting in (s+f) DOF present in the structure-
solved directly in their coupled form. If the system is foundation subsystem rather than the (s+f+g) DOF
treated as being fully linear (or equivalent linear), the present in the complete soil-structure-foundation
solution can be carried out in the frequency domain system. This reduced set of equations having (s+f)
using Eq. (8). In doing so, the complete set of complex DOF can be solved much more efficiently than solving
algebraic equations are solved separately for discrete the equations for the complete system having (s+f+g)
DOF as required by the one-step direct approach.
values of o over the frequency range of interest
yielding the corresponding sets of displacement vectors
CAPACITY EVALUATIONS
us (/to), uf (/co), and ug (/co). Having obtained these
vectors for the discrete values of co, they are inverse The objective of the capacity evaluation is to
Fourier transformed to the time domain giving vectors determine the most probable levels of seismic
us (i), Uf(i), ug (i). The corresponding time histories of resistance of the various elements, components, and
internal forces and/or deformations in the system can subsystems of the bridge. The resistance capacities
then be obtained directly using standard finite-element provided by this evaluation, along with the
procedures. corresponding demands, provide the basis for judging
If the structure-foundation subsystem is modelled seismic performance of the complete bridge system
as a nonlinear system, the solution can be carried out in

13-9
during future earthquakes. In the domain of SFSI as 5. Penzien, J. and Watabe, M., "Simulation of 3-
discussed here, the capacity evaluation focuses on soil- Dimensional Earthquake Ground Motions", J. of
foundation systems. Earthquake Eng. & Str. Dyn., Vol. 3, No. 4,1975.
For a bridge subjected to static loadings, the soil-
foundation capacities of interest are the load resistances 6. Hao, H., Oliviera, C. S., and Penzien, J., "Multiple-
and the associated foundation deflections and Station Ground Motion Processing and Simulation
settlements. Their evaluation constitutes the bulk of the Based on SMART-1 Array Data," Nuc. Eng. and
traditional foundation design problem. When the Des., Vol. ffl, No. 6,2229-2244,1989.
bridge is subjected to oscillatory dynamic loadings,
including seismic, the static capacities mentioned above 7. Silva, W. J. and Lee, K., "WES RASCAL Code
are, alone, insufficient in the process of judging soil- for Synthesizing Earthquake Ground Motions,"
foundation performance. In this case, it is necessary to State-of-the-Art for Assessing Earthquake Hazards
assess entire load-deflection relationships, including in United States, Report 24, Army Engineers
their cyclic energy dissipation characteristics, up to Waterway Experimental Station, Miscellaneous
load and/or deformation limits approaching failure Paper 5-73-1,1987.
conditions in the soil-foundation system. Because of
the complexity of this assessment, the capacity 8. Bolt, B. A. and Gregor, N. J., "Synthesized Strong
evaluation must be simplified in order to make it Ground Motions for the Seismic Condition
practical. This is usually done by treating each soil- Assessment of Eastern Portion of the San
foundation system independently and by subjecting it to Francisco Bay Bridge", Report No. EERC 93-12,
simplified pseudo-static monotonic and/or cyclic Earthquake Engineering Research Center,
deformation-controlled step-by-step patterns of loading, University of California, Berkeley, 1993.
commonly referred to as the "push-over" analysis.
9. Chang, C. Y., Power, M. S., Idriss, I. M.,
ACKNOWLEDGEMENT Sommerville, P. G., Silva, W. and Chen, P. C.,
"Engineering Characterization of Ground Motion,
For a more comprehensive treatment of soil- Task II; Observation Data on Spatial Variations of
foundation-structure interaction, including example Earthquake ground Motion", NUREG/CR-3805,
solutions, the reader is referred to Ref. 1 by W. S. Vol. 3, U.S. Nuclear Regulatory Commission,
Tseng and J. Penzien, upon which this condensed paper Washington, D.C., 1986.
is based.
10. Tseng, W. S., Lilhanand, K., and Yang, M. S.,
REFERENCES "Generation of Multiple-Station Response-
Spectra-and-Coherency-Compatible Earthquake
1. Tseng, W. S. and Penzien, J., "Soil-Foundation- Ground Motions for Engineering Applications",
Structure Interaction," Chapt 42 of "Handbook of Proc. 12th Int. Conference on Struct. Mech. in
Bridge Engineering" W. F. Chen and L. Duan, Reactor Technology, Paper No. KOI/3, Stuttgard,
Editors, CRC Press LLC, Cat No. 74-34. Germany, August 15-20,1993.

2. Abrahamson, N. A., "Spatial Variation of 11. Schnabel, P. B., Lysmer, J., and Seed, H. B.,
Earthquake Ground Motion for Application to "SHAKE - A Computer Program for Earthquake
Soil-Structure Interaction," Report No. TR- Response Analysis of Horizontally Layered Sites,"
100463, Electric Power Research Institute, March Report No. EERC 72-12, Earthquake Engineering
1992. Research Center, University of California,
Berkeley, 1972.
3. Lilhanand, K. and Tseng, W. S., "Development
and Application of Realistic Earthquake Time 12. Idriss, I. M. and Sun, J. I., "User's Manual for
Histories Compatible with Multiple-Damping SHAKE 91," Center for Geotechnical Modeling,
Design Response Spectra," Proc. 9th World University of California, Davis, 1992.
Conference of Earthquake Engineers, Tokyo-
Kyoto, Japan, August 2-9,1988. 13. Vucetic, M. and Dobry, R., "Effects of Soil
Plasticity on Cyclic Response", J. of Geotech.
4. Kaul, M. K., "Spectrum-Consistent Time-History Eng., ASCE, Vol. 117, No. 1,89-107,1991.
Generation", J. of Eng. Mech. Div., ASCE, Vol.
104, No. EM4,781,1978.

13-10
14. Sun, J. I., Golesorkhi, R., and Seed, H. B., 16. Lysmer, J., Tabatabaie-Raissai, M., Tajirian, F.,
"Dynamic Moduli and Damping Ratios for Vahdani, S., Ostadan, F., "SASSI-A System for
Cohesive Soils", Report No. UBC/EERC-88/15, Analysis of Soil-Structure Interaction", Report No.
Earthquake Engineer Research Center, University UCB/GT/81-02, Department of Civil Engineering,
of California, Berkeley, 1988. University of California, Berkeley, 1981.

15. Apsel, R. J., "Dynamic Green's Function for 17. Clough R. W. and Penzien, J., "Dynamics of
Layered Media and Applications to Boundary Structures" Second Edition, McGraw-Hill, Inc.
Value Problems", Ph.D. Thesis, University of 1993.
California, San Diego, 1979

13-11
 ANALYTICAL AND EXPERIMENTAL STUDIES ON THE EFFECT OF SOIL-
STRUCTURE INTERACTION ON DAMPING, NATURAL FREQUENCY AND
EFFECTIVE INPUT MOTION OF BUILDINGS
Nobuo Fukuwa1 , M. AH Ghannad2 , Jun Tobita3 and Riei Nishizaka4

ABSTRACT: The effect of Soil-Structure Interaction on the dynamic response of buildings is studied through
analytical and experimental approaches. As the analytical approach, simplified models with different levels of
simplicity are employed in order to study the SSI effect on the natural period and damping of buildings.
Parametric studies are conducted for the cases of structures located on the surface of either a homogeneous
half-space or a layered half-space soil medium. The results are then compared with current seismic code
provisions. The general trends leamt from the analytical study are compared with experimental results on real
buildings through both microtremor measurements and seismic observations.

1. INTRODUCTION change in the soil and the structured physical characteristics.

However, it is impossible to develop an exact deterministic
It has been known for many years that the flexibility of soil
solution for the soil-structure system due to the complexity of
under structures affects the structural response to dynamic loads.
soil material properties, involved geometry of building foundation
The effect of Soil-Structure Interaction (SSI), thus, has been
and the complicated nature of earthquake ground motions. Thus,
considered in the design of massive and critical facilities such as
for the sake of simplification, it is necessary to make numerous
nuclear power plants. However, the SSI effect on dynamic
assumptions so that the problem at hand is amenable to solution;
behavior of ordinary building type structures has not attracted
and it is the art of engineering to make models as simple as
much attention. Also, the experimental studies on dynamic
possible. To this end, it is necessary to confirm the achieved
behavior of such buildings are rare in contrast to the laige number
analytical results through comparison with the results of
of such studies on high-rise buildings and nuclear facilities. This
experimental studies. In this paper, simplified models for the
may be due to the fact that the experimental evaluation of dynamic
soil-structure system, with different levels of simplicity, are
characteristics of short and squatty buildings can be quite difficult.
employed in order to study the SSI effect on the natural period
Additionally, the final results can be subjective and may render
and damping of buildings. Parametric studies are conducted for
different values, depending on the techniques used and the the cases of structures located on the surface of either a
personal judgement applied. homogeneous half-space or a layered half-space soil medium.
The most significant event intervention in this regard has been
The general trends learnt from the analytical study are then
the 1978 inclusion of SSI in the tentative provisions ofATC3-06
compared with experimental results on real buildings through
in the United States (ATC 1978). The seismic codes of other
both microtremor measurements and seismic observations.
countries including Japan, however, continue to have no
provisions addressing SSI (Earthquake Resistant Regulations
1988). In fact, it seems that even in the United States the
beneficial use of SSI have not been sufficiently exploited. Also, 2. ANALYTICAL STUDY
the regulations have remained unchanged since their first 2.1 ANALYSIS MODEL
appearance in 1978 (NEHRP 1994). On the other hand, studies
have indicated that the level of damages caused by recent Figure 1 shows the conventional soil-structure model where
earthquakes, particularly in case of short and stiff buildings, has the structure is simply modeled as a shear building model with
been predominantly related to the effects of SSI (Celebi 1997; lumped mass and mass moment of inertia in each story and the
Hayashi et al. 1998; Wallace and Moehle 1990; Yasui and soil is replaced by the sway and rocking springs and dashpots,
Tokimatsu 1998). Hence, there is much opportunity in further i.e., ks, kr, cs and c.. Here, as a further simplification, the same
analytical and experimental studies on this matter mass, 777, mass moment of inertia, j, height, /;, and stiffness, k,
The key factors in the seismic design of buildings are the are considered for all stories and it is believed that such a simple
natural period, damping ratio and the effective input motion and structure model is quite adequate for the purpose of this study
all of which are affected by SSI phenomenon. Therefore, it is The other substructure, the unbounded soil, however, should be
necessary to examine the variation of these factors due to the treated more carefully. The soil model should consider the
unboundedness of the soil and should satisfy the radiation
1. Prof., Center for Cooperative Research in Advanced Science & Technology, condition in the soil. It also must be simple enough to allow the
Nagoya University. Japan, Dr. Eng. required parametric studies to be carried. In this regard, the
2. Former Graduate Student at Nagoya University, Japan, Dr. Eng. concept of cone models based on one-dimensional wave
3. Assoc. Prof., School of Engineering, Nagoya University, Japan, Dr. Eng.
4. Research Associate, Graduate School of Engineering, Nagoya University,
propagation theory is used. The cone models, in spite of their
Japan, Dr. Eng. simplicity, are able to capture most of the basic and important
14-1
 nth floor distribution of mass and stiffness in the structural model allows
the determinant of its stiffness matrix in the fixed-base state can
in .j .... (ii-llih floor be presented explicitly by polynomials through direct cofactor
expansion method (Fukuwa and Ghannad 1996). Having the
determinant of the stiffness matrix for the fixed base structure,
the related expression for the case of the structure located on the
sway and rocking springs and dashpots can be expressed by
1st floor polynomials readily. The complex eigenvalues of the soil-
structure model can then be calculated by solving the respective
Foundation
polynomials. The equations, however, would be nonlinear due
to the frequency dependency of the soil representative springs
and dashpots'coefficients. Special techniques such as searching
FIG. 1. The soil-structure model in the complex plane thus should be employed in this regard.
Once having the eigenvalues of the system, the damped natural
concepts of more rigorous solutions (\Vblf 1994). frequencies, orf, and the modal damping ratios, , can be
calculated. The detailed discussion on the method as well as
2.2 TRANSFER MATRIX METHOD related formulations can be found in (Fukuwa et al. 1995).

The simplified structural model of Fig. 1 may be considered

2.4 APPROXIMATION THROUGH A SIMPLIFIED
as a periodic structure with slight modifications. Thus, the 3DOF MODEL
transfer function for the structure itself may be calculated by
transfer matrix method which is a powerful method for dealing Figure 2 shows the simplified 3-degree of freedom (3DOF)
with periodic structures. The transfer function for the soil- model where the superstructure -building- is replaced by its modal
structure system can then be constructed readily by introducing effective mass,m%ir , and modal effective stiffness, ktir , providing
the effect of soil related degrees of freedom. It can be shown the same modal frequency as the original multi degree of freedom
that the transfer function for the^'th storey of the sway-permitted (MDOF) model of Fig.l. Also, H represents the modified
model will be as follows (Fukuwa et al. 1995) effective height of the structure defined as follows
(2n-2y- + l)'-w (3a)
1f. cos-
u,= 2 r
K-')< (2n + l)y (2/i' + 3)V
2
where

where (3b)
^* Ar,+/GJc5 r r
and 77 is the effective height of the structure. Also, r is the radius
of the (equivalent) circular foundation. It is shown that//w
and /; is the number of stories and i = J^l. Also, ca and E*t/r. are introduces the effect of floor's mass moment of inertia, j, into
the circular frequency of excitation and the hysteretic damping the model implicitly (Ghannad 1998). The effect may be
ratio in the structure, respectively. considerable for the case of short and squatty buildings. The
foundation is replaced by mass m and the soil is replaced by
2.3 COFACTOR EXPANSION METHOD complex sway and rocking springs, Kt and Kf. The dynamic
properties of the system are then evaluated through parametric
It is well-known that standard eigenvalue analysis methods
eigenvalue analysis. It can be shown that the first complex
are not applicable to soil-structure systems due to
eigenfrequency of the system will be (Ghannad et al. 1998)
nonproportionality of the damping matrix. Although there are
well-established techniques for handling even nonclassical
damped systems, they are not applicable when the frequency (4)
dependency of soil stiffness is also considered. Also, the
application of such techniques is limited to cases with real where (O is the natural circular frequency of the structure in the
stiffness matrices where the use of complex damping for
modeling the material damping in the soil or structure is not
applicable (Hurty and Rubinstein 1964). That's while the results
of experimental studies on structures show more compatibility
with the concept of complex damping model (hysteretic damping) H
than the commonly used viscous form of damping. The hysteretic
damping model has also been proposed as the best possibility
for modeling the material damping in the soil (Kausel and Roesset
K,
1974). As an alternative method capable to deal with above
mentioned problems, explicit presentation of the determinant of
the system's stiffness matrix is introduced here. The uniform FIG. 2. The simplified 3DOF model

14-2
fixed-base state. Also, nonproportional damping matrices. Transfer functions for the
three mentioned cases are drawn in Fig.3 in comparison to those
K. of the fixed-base structure model. The transfer functions are
(5a)
computed based on the response of the top story in comparison
to the soil surface. The results are shown for three different
(5b) building models with different number of stories. As shown, the
The material damping in the soil and structure may also be effect of SSI on the eigenfrequencies of the system (the
included in the formulations as the hysteretic form of damping frequencies related to peak values) is overestimated for the first
by using the correspondence principle. Using compatible values model with proportional damping matrix. The same conclusion
for the effective mass, effective stiffness and effective height of is applicable to any other damping model which leads to real
building with those of MDOF model, the results would be in full eigenvalues (Ghannad 1998). Additionally, the peak values
agreement for the two models (Ghannad et al. 1998). The 3DOF related to the second and higher modes are obviously
model, however, needs much less effort to be solved. underestimated for the case of stiffness proportional damping
model. It is because the stiffness proportional damping model
2.5 TRANSFER FUNCTIONS leads to high damping ratios for higher modes of vibration which
are in fact much larger than the actual damping in the system.
In order to clarify the importance of damping model in SSI
These effects are seen more clearly for short and moderate height
studies, the transfer function of soil-structure systems with
buildings.
different damping models are compared to each other As a
From another point of view, Fig.4 shows the response of model
representative example, an ordinary reinforced concrete building
(2) at different story levels. The results are shown for the first,
with different number of stories located on a square 35x35 meters
the last and the middle story of three different building models
foundation is studied. The shear wave velocity in soil under the
with different number of stories. The results show a clear phase
foundations considered to be V=200m/s. As the damping
difference in the vibration of different parts of the structure for
mechanism in the soil-structure system, three different models
all three cases. In the other words, the mode shapes would be
are examined as follows: (1) Only material damping in the soil
complex, whose elements differing in phase as well as in
and structure, modeled as 1% complex damping (hysteretic
amplitude. This means that models which lead to real eigenvalue
damping), (2) Complex material damping in the soil and structure
analysis and real mode shapes, such as those with proportional
(1 %) and radiation damping in the soil based on the concept of
damping models, can not be a suitable choice.
cone models and (3) Stiffness proportional damping as the
material damping in the soil and structure (1% at the first mode
2.6. STRUCTURES ON THE SURFACE OF A
of vibration) and radiation damping in the soil based on the
HOMOGENEOUS HALF-SPACE
concept of cone models. The first damping model leads to a
proportional damping matrix for the system, which can be The effect of SSI on the natural frequencies and the damping
diagonalized by the same transformation that uncouples the ratios of buildings located on the surface of a homogeneous half-
undamped system, whereas the two other models result to space soil medium is studied in this section through the

Fixed-base structure
- - Sway-permitted (Only Hystet material damp, in soil & structure)
Sway-permitted (Hystec material damp, in soil & structure + radiation damp, in soil
Sway-permitted (Stiffness prop, material damp, in soil & structure + radiation damp, in soil
100.0 100.0

10.0 10.0

1.0 1.0

0.0 0.5 1.0 1.5 2.0 0.0 0.2 0.4 0.6 0.8 1.0 0.00 0.05 0.10 0.15 0.20 0.25
3.14

-3.14
0.0 0.5 1.0 1.5 2.0 0.0 0.2 0.4 0.6 0.8 1.0
0.00

-3.14 i
0.00 0.05 0.10 0.15 0.20 0.25

( a ) 5-story building ( b ) 10-story building ( c ) 40-story building

FIG. 3. Transfer functions for systems with different damping models
14-3
 The highest story The middle story The lowest story
100 100

o 10 10

I l
: .r
': \
O.I

-7T
0.5 I 1.5 0 0.5 1.5 0 0.2 0.4 0.6 0.8 1
Real(co)/con Real(to)/(0n Real(co)/(fl()
(a) 2-story building (b) 5-story building (c) 10-story building

FIG. 4. Transfer functions for different parts of the building (Sway-permitted model)

application of cofactor expansion method. The effective aspect 1980; AIJ 1996). No material damping, however, is addressed
ratio of the building, H/r, and the dimensionless frequency for the structure. The results apparently show more dramatic
interaction effect for higher values of (o,,)^ in all cases as would
r-co, be expected. Consequently, drastic change in the first natural
KL, - (6) period, and also very high damping ratios are seen for stiff
structures located on soft soils. It should be mentioned that the
are considered as the key parameters whereas the other soil and practical range of interest for (alt) is different for buildings with
structure parameters are set to some typical values. In (6),ot is different aspect ratios. For example, (a())fir would hardly exceed
the circular frequency of the fixed-base building model andV^ is
0.5 for ordinary buildings with H/r - 5. The limit value may be
the shear wave velocity in soil. Figure 5 shows the results of
considered about 1.5 and 2.0 for buildings with the effective
parametric studies where a damping ratio,ol.,=3% is assigned to
aspect ratios 2 and 1.5, respectively. The results also point to
the material damping for low strain levels in the soil (Kokusho
severe effect on the natural period of slender buildings due to
rocking effect. However, the SSI effect on the damping ratios
are more important for squatty buildings. The results of Fig. 5
are independent of the number of stories and also are in full
agreement with the results of 3DOF model. The results of Fig.5
may be presented in a new fashion as the variation the systems
damping ratio with the change in the natural period of the

ATC3-06 ----- This study]

0 0.5 1 1.5 2 2.5 0.25

( a ) The first natural period 0.20

0.15 -

0.10 -

0.05 -

0.00

TIT
fix
( b) The soil induced damping ratio (^=0.03)
FIG. 6. Damping ratio as a function of the
FIG. 5. The effect of soil on the dynamic properties of system change in the natural period of the building.
14-4
 underlain half-space increases, the reflection of waves at the
interface of two media results to lower damping for the soil-
structure system. Additionally, for the limit state of a soil layer
on rigid rock a cutoff frequency equal to the fundamental natural
frequency of the layer exists below which no radiation occurs.
Using the latter concept, the ratio of the fundamental period of
layer, Tlmvr, to that of the soil-structure system, T, is introduced
as radiation index, /, for verification of existence or absence of
radiation in soil.
FIG. 7. The layered half-space soil model
' laver 4d
building. This format is generally more desirable in practical
structural design. Figure 6 shows the results in this new format The same concept has been used byATC3-06 regarding buildings
in comparison to theATC3-06 regulations. As shown, the results located on a stratum of a soft soil underlain by a much stiffer
of this study are in a good agreement with the graphs suggested soil. According to ATC3-06, the soil induced damping ratios
by ATC3-06. It is also shown that the results of the two studies related to the half-space case shall be scaled down by the factor
are in a good agreement even for higher levels of strain in soil of square value of radiation index, i.e.,/ 2. Equation (8) may be
(Ghannad 1998). Although no material damping is addressed rewritten in the following format.
for the structure in this study, it should be kept in mind that the
internal damping of structures is also subject to change due to T
/ = 1 1aver
K)*
SSI (Novak 1975). The change in the structural damping has (9)
been approximated by some researchers as follows (\eletsos
1977;AIJ 1996) This thus allows to study the variation of radiation index as a
function of (fl(() for any specific soil-structure system. Figure 8
shows the results of such studies for different cases. The graphs
(7)
have been drawn for a range of (fl,,)ffr which is practically
important for buildings with each specific aspect ratio. As shown,
in which w; and 5fl._ are the internal damping ratio of the the radiation index for slender buildings (///r = 5) would be
structure in the fixed-base state and when located on flexible much less than 1.0 even for sites with deep layers. Therefore,
soil, respectively. The same approximation has been also used for different sites with different depths of the layer, slender system
by ATC3-06. Equation (7) is valid for systems with viscous type would be in "no radiation" zone and consequently depth of the
of material damping in the structure. The assumption of hysteretic layer doesn't play any important role. On the other hand, for the
type of damping for the structure, as adopted here, however, leads case of squat buildings, the radiation index may be below or
to an exponent 2 (instead of 3) in the right-hand side of (7) (W)lf above the radiation limit depending on depth of the layer
1985; Ghannad et al. 1998). Consequently, the depth index, d/r, would be an important factor
for the case of buildings with moderate or low aspect ratios.
2.7 STRUCTURE ON THE SURFACE OF A LAYERED Another factor which plays an important role in the level of soil
HALF-SPACE SOIL MEDIUM induced damping ratio is the difference between the stiffness of
Figure 7 shows the layered half-space soil model which the layer's material and that of the underlain half-space, i.e., the
consists of a soil layer located on surface of a homogeneous soil stiffness contrast index which has not been addressed byATC3-
half-space. The mass density and Poisson's ratio are considered 06. As a result, different systems with the same radiation index
to be the same for the materials of the two media. However the may have quite different levels of damping ratio due to difference
shear wave velocity is different for them. As the stiffness of in the stiffness contrast index. Consequently, the radiation index

1.00 LOO 0.50

^ 0.80 0.80 ~ 0.40

x
<u x
0)
H 0.60 0.60 H 0.30
o o
"ii
'5 0-4o '1
5 0.40 '% 0.20
1
& 0.20 &> 0.20 & 0.10

0.00 0.00 0.00

0.5 1.5 0.5 1.5 0.1 0.2 0.3 0.4 0.5

(a) 77//-= (b) H/r = 2 (c) H/r =

FIG. 8. Variation of radiation index,/ ( V?/VtL = 2 )

14-5
 Directly computed for the layered half-space case -- Scaling down the results of the half-space case I
0.25 0.25 0.25 0.25

0.20 0.20 0.20 0.20

0.15 0.15 0.15 0.15

0.10 0.10 0.10 0.10

0.05 0.05 0.05 0.05

000 0.00 0.00 0.00

1 1.2 1.4 1.6 1.8 2 i i.2 1.4 1.6 1.8 2 1 1.2 1.4 1.6 1.8 2 1 1.2 1.4 1.6 1.8 2

r/r TIT,
(a) (b) H/r = \.5 (c) H/r = (d) H/r =

(a) dlr=\ , v"/V5L =2

0.25 0.25 0.25

0.20 0.20

0.15 - 0.15

0.10 - 0.10 0.10

0.05 - 0.05 0.05

0.00 0.00 0.00

1 1.2 1.4 1.6 1.8 2 1 1-2 1-4 1.6 1.8 2 1 1.2 1.4 1.6 1.8 2 ] 1.2 1.4 1.6 1.8 2
TIT TIT TlTr,i TlTr

(a) H/r = (b) H/r = l.5 (c) H/r = 2 (d) /// = 5

( b ) dlr = 2 , VS"/VSL = 2
FIG. 9. Verification of ATC3-06 suggested approach for calculation of soil induced damping ratio in the layered half-space case

alone can not be used as a suitable criteria for estimation of natural frequencies above the cutoff frequency limit. On the
damping capacity in soil. On the other hand, considering the other hand, the results for very slender structures, such as those
effect of stiffness contrast index does complicate the method with H/r = 5, would be underestimated in all cases (almost no
and makes it far from a code suited approach. So, for the sake of damping). Although this conservatism involved inATC3-06 for
simplicity it may be desirable to set the stiffness contrast index tall buildings is partly compensated by the requirement that
to a practically conservative value, say V"/VSL = 4, compatible system's damping ratio, with or without considering the SSI
with the qualitative description of layered half-space sites by effect, be no less than 5%, there is no justification for
ATC3-06. However, even doing so, it is necessary that the effect overestimation of damping ratios for short and moderate height
of the two other key factors, the depth index and the effective buildings which is not in the safe side!
aspect ratio of the building, be included in the construction of
the scaling factor more explicitly. Figure 9 shows the resulted
soil induced damping ratios for different circumstances. In each 3. EXPERIMENTAL STUDIES BASED ON
figure, the obtained values for damping ratio of the system by MICROTREMOR MEASUREMENTS
scaling down the half-space case results are compared with those
Ambient vibration measurements have always been a fast and
directly achieved through the complex eigenvalue analysis. The
efficient way of determining the dynamic properties of existing
results are related to a typical stiffness contrast index v"/V^ - 2. buildings. Recently, even more reliable results have been
According to results, damping ratios obtained by application of achieved through the use of high precision measurement devices,
ATC3-06 recommendation tend to become overestimated by more accurate techniques and high speed computers. Here, also,
increase in depth index value and by decrease in the effective the method is employed to study the effect of the size of buildings
aspect ratio of the structure. This is mainly because ATC3-06 and the flexibility of the underlain soil on the dynamic
doesn't consider any reduction in the damping of systems with characteristics of soil-structure systems.
14-6
 Selected buildings j
* Borings j

,\1* f\Jttra,
1 ^ 31 ^\ ^T^s^,
\i ^N^v^-vNx.'

FIG. 10. The location of selected buildings in the campus

TABLE 1. The selected buildings' information

No. of Height Struc. Pile Embe- Construe. Average
No. Building Ref. name Stories (m) System length dment Year N-value
(m) (m)
1 Faculty of Eng. No. 1 Engl 3+1 15.3 RC - 1.2 1951-70 14.6
2 Faculty of Eng. No.2 EngZ 3+1 15 RC - 1.1 1 954-65 10.8
3 Faculty of Eng. No. 3 Eng3 4 15.8 RC 5-16 1.5 1 962-70 12.5
4 Faculty of Eng. No.4 Eng4 4 14.75 RC 6 1.25 1964-70 17.2
5 Faculty of Eng. No. 5 Eng 5 6 22.17 RC 5-33 1.7 1967-69 14.4
6 Faculty of Eng. No. 7 Eng7 4 15.2 RC 20 1.2 1971-89 16
7 Faculty of Eng. No.8 Eng8 4 16 RC 9-10 2.8 1987 28.9
8 Faculty of Eng. No.9 Eng9 6 22.2 RC 12 2.3 1982-93 22.9
9 Faculty of Science_A2 Sc.A2 5 18.1 RC 9 1.5 1979 17.2
10 Faculty of Science_E Sc.E 5+1 19.7 RC 8 1.7 1967-85 21.6
11 Faculty of Agriculture Agr. 6 21 RC - 2 1966 23.1
12 Inst. for Hydrospher. Hyd. 5 18.8 RC 9 1.6 1971 30.2
13 International Residence Res. 8 33.2 RC 15 2 1981-88 27.8
14 Faculty of Eng. No.1 (new) Eng1_N 10 39.3 SRC 45 8.05 1995 11.6

3.1 BUILDINGS IN NAGOYA UNIVERSITY depth. Although only the groundfilled parts are shown in the
figure, it is interesting to say that the difference between the
3.1.1 OUTLINE OF BUILDINGSAND SOIL level of the excavated and filled areas reaches up to 50 meters in
CONDITIONS some parts. Generally speaking, the northwestern and
northeastern parts of the campus are the earthfilled areas and
Microtremor tests were conducted on fourteen buildings in
consequently have looser soil. The other areas, especially the
the Higashiyama campus of Nagoya University It should be
central part of the campus have been mainly excavated and better
mentioned that because of heavy earthworks during the years,
soil condition is expected in these areas. The buildings were
the topography of the campus has been drastically changed from
selected among the existing RC buildings considering some
the original situation. Therefore, the different parts of the campus
parameters which are of interest, e.g. the number of stories, plan's
have different soil conditions. On the other hand, most of the
shape and size, soil condition and also the availability of boring
selected buildings have almost a typical plan of educational type
data. It should be mentioned that totally, there are 202 boring
buildings, i.e. a slender rectangular with almost the same width.
data available in the campus which are mainly related to the
So, the effect of soil condition on the eigenproperties of buildings
northern and central parts of the campus as it can be seen in
can be studied well. Figure 10 shows the Higashiyama campus
Fig. 10.
and the location of the selected buildings. The numbers on the
figure refer to the order number of buildings in the first column
3.1.2 MEASUREMENTS AND ANALYSIS METHODS
of Table 1 where the information of the selected buildings have
been summarized. The change of topography of the campus due The microtremor measurements were performed for NS and
to heavy earthworks from 1936 to 1991 is shown in the EW directions which are related to the transverse and longitudinal
background where darker shades indicate an increase in filling dimensions of buildings, respectively. The responses were
14-7
measured at the first floor and top of the buildings as well as on over the first 10 meters from the ground surface. Since the height
the ground surface. The measurements on the ground surface of the building affects the severity of SSI effect, the buildings
were done at a distance from the building which are not affected are divided into two categories according to their height: 1) Short
by the building's vibration. In the case of the Engl_N building, buildings including 3 and 4-story buildings and 2) Moderate
the test was repeated for different stages of construction, i.e. after height buildings including 5 and 6-story buildings. The results
completion of each story. of the transfer function fitting method for the second category
The moving coil type seismometers with natural period of are shown in Fig. 12. Although the number of data is not so
1.0 second were used to measure the responses simultaneously large, a general trend of lower damping ratios and shorter periods
at the above mentioned locations and the velocity was measured for buildings located on stiffer soils can be observed. Howevei;
in all cases. The signals, after amplification and low pass filtering the results for the shorter buildings (the first category) don I show
(f. =30 Hz) are digitized at the rate of 100 samples per second. any clear trend and thus are not presented here. More detailed
In all cases, the measurements were conducted three times in discussion can be found in (Ghannad et al. 1997a, 1997b).
ten-minute intervals, providing the total length of 30 minutes.
The Fast Fourier Transform was computed for every 2048 points 3.2 LOW-RISE R.C. SCHOOL BUILDINGS IN
leading to a total of 87 specimens which were used for ensemble NAGOYA CITY
averaging.
For the parameter estimation methods, two techniques are 3.2.1 OUTLINE OF BUILDINGSAND SOIL
used: 1, Transfer function fitting method and 2, Random CONDITIONS
Decrement (RD) method. In the former method, the dynamic Totally sixty seven 3-story elementary school buildings were
properties of the building are estimated by finding a known examined in Nagoya city (Yagi et al. 1998). All the buildings
system whose transfer function can be matched well with the have almost a typical plan of educational buildings, i.e., a slender
observed one (Tobita 1996) whereas the latter technique is based rectangular with the same width. The structural system is also
on the extraction of the free vibration motion from the recorded the same for all buildings which is consisted of 2-span reinforced
data by superimposing a sequence of intervals with the same concrete frames with shear walls and multi-span frames with
phase (Jeary 1986;Tamura et al. 1996). few walls in the transverse and longitudinal directions,
respectively. However, the foundation type and soil condition
3.1.3 RESULTS AND DISCUSSION are different for them. Figure 13 shows the location of the
The estimated values for the period and damping ratios in the buildings in Nagoya city. As shown, almost half of buildings are
NS direction of the examined buildings have been summarized located on alluvium deposits. Also, almost half of them have
in Table 2. Three sets of results presented for each building are pile foundations.
related to the following cases respectively: 1) Using the transfer
3.2.2 MEASUREMENTS AND ANALYSIS METHODS
function fitting method by considering the responses at the ground
as the input for calculating the observed transfer functions, 2) Almost the same devices and methodology as described in Sec.
The same as case #1 but using the responses of the first floor as 3.1.2 are also used here. However, for the parameter estimation,
input, 3) Using the RD method. It is believed that the results of only the transfer function fitting method is used. As mentioned
the second case can be considered as a quasi-fixed base model before, the evaluation of dynamic characteristics of low-rise
because the sway effect is not included. Although the results are buildings would be difficult through conduction of microtremor
not generally compatible with this quasi-fixed base idea, there tests. It is because a well-shaped transfer function can not be
are some cases where this idea works well. For example, the achieved except under very well-controlled conditions due to
effect of SSI can be studied well in the well-performed case of the low amplitude of vibrations and due to the remarkable effect
Engl_N building by comparing the 6th and 10th columns of of soil under buildings. Figure 14 and 15 show the results of
Table 2. The higher effect for the lower number of stories can two cases which have been selected as the representatives for
also be seen clearly. Also, the results show higher damping ratios well and poor estimations. Using the same classification for all
for the longitudinal direction (EW) which is compatible with cases. Fig. 16 shows the statistical results based on the soil type,
the results of the analytical study. the type of foundation and the examined direction of buildings.
The results using the RD method are drawn in Fig. 11 for both Generally speaking, better performance is seen for the cases of
of NS and EW directions. The figure represents a general spread foundations on stiffer soils. On the other hand, comparison
tendency of lower damping ratios for taller buildings which may of the Fourier amplitudes at the three mentioned levels in Fig. 14
be interpreted as the effect of SSI. Although the site's soil reveals that the effect of SSI is much lower for the well-
condition is not the same for all buildings, generally one may performed case of Building Y. Thus, the effect of SSI may be
conclude that the SSI effect is higher for shorter buildings. More considered as one of the parameters responsible for the distortion
specifically, the results of the Engl_N building, which are related of transfer functions.
to the different number of stories but the same soil condition,
clearly lead to the same conclusion. From another point of view, 3.2.3 RESULTS AND DISCUSSIONS
the results show higher dispersion in the left hand side of the
graph, i.e., for shorter buildings. This can be partly explained Using the idea of quasi-fixed base model introduced in
by SSI effect. For evaluation of the effect of soil condition on Sec.3.1.3, the results for the frequency of the superstructure are
the dynamic properties of the building, the natural periods and drawn versus the frequency of the soil-structure system in Fig. 17.
damping ratios are plotted versus the average N-value of the soil Only the results of well-performed cases are shown. The results

14-?
 TABLE 2. The estimated parameters for NS direction
Average rms Transfer Function Fitting Method RD Method
Ref . name (m.kine) Top / 1 st floor Top / Ground
@ ground top freq (Hz) I S/P Error freq (Hz) I S/P Error freq (Hz) f,
Engl 0.213 0.397 5.13 0.134 0.102 1.82% 4.82 0.146 0.081 2.57% 5.08 0.117
Eng2 0.170 0.244 5.82 0.137 0.094 0.67% 5.51 0.134 0.093 0.86% 5.47 0.128
Eng3 0.098 0.196 - - - - 5.25 0.310 0.142 2.29% 5.08 0.137
Eng4 0.380 0.486 6.93 0.243 0.113 3.42% 6.34 0.113 0.115 1.02% 6.64 0.132
Eng5 0.638 1.347 3.30 0.119 0.089 0.85% 3.06 0.057 0.048 1.50% 2.93 0.075
Eng7 0.374 0.273 6.64 0.184 0.115 0.45% 5.22 0.204 0.225 2.49% 5.08 0.161
Eng8 0.264 0.278 5.54 0.107 0.051 0.85% 5.15 0.093 0.057 1.46% 5.08 0.095
Eng9 0.277 0.406 4.72 0.156 0.076 1.43% 4.05 0.103 0.066 1.37% 3.71 0.089
Sc.AZ 0.339 0.897 5.04 0.220 0.068 0.80% 4.48 0.096 0.040 2.70% 4.49 0.088
Sc.E 0.212 0.697 4.08 0.272 0.158 0.27% 3.48 0.055 0.031 1.43% 3.52 0.096
Hyd. 0.362 0.814 5.93 0.177 0.077 0.87% 5.36 0.122 0.060 3.57% 5.47 0.093
Agr. 0.152 0.251 3.70 0.189 0.080 3.70% 3.59 0.117 0.038 5.22% 3.32 0.114
Res. 0.235 0.809 3.37 0.054 0.018 1.41% 3.33 0.022 0.021 3.08% 3.32 0.056
Engl_N(3) 0.190 0.240 7.56 0.224 0.137 0.94% 9.10 0.150 0.261 0.96% 9.38 0.120
Engl_N(4) 0.196 0.305 5.67 0.194 0.093 0.66% 5.71 0.450 0.270 0.93% 4.30 0.203
Engl_N(5) 0.200 0.414 4.78 0.126 0.074 2.30% 4.60 0.217 0.177 1.61% 5.08 0.084
Engl_N(6) 0.208 0.483 3.05 0.086 0.076 5.55% 2.90 0.107 0.097 4.64% 2.73 0.101
Engl_N(7) 0.166 0.484 2.93 0.070 0.045 0.98% 2.74 0.065 0.056 1 1 .60% 2.73 0.079
Engl_N(8) 0.167 0.528 2.67 0.071 0.039 0.56% 2.46 0.072 0.053 1.72% 2.54 0.104
Engl_N(9) 0.190 0.628 2.35 0.057 0.034 0.71% 2.23 0.059 0.043 1.00% 2.15 0.079
Engl_N(10) 0.204 0.715 2.09 0.051 0.030 0.54% 2.00 0.031 0.018 1.82% 2.06 0.042
: Participation factor

O NS D EW NS(Engl_N> EW (Engl_N | O NS D EW NS(Engl_N) EW(Engl_N)|

0.6
n-2 5 I i I I I I I

0.5 -
. 11
0.4 . - -
o <2 "I 5
CD
S o., Lo

0.2
,
to *%*t
C
1 do I "'
a
\ Q
C

O.I
:
n i___i______i___i_
5 10 15 20 25 30 35 40 45 10 15 20 25 30 35 40 45

Height (m) Height (m)

(a) Period of vibration (b) Damping ratio

FIG. 11. Distribution of period and damping ratio with the height of buildings (RD method)

Sc.A2 Sc.E Sc.A2 Sc.E

Eng5 Eng9 Eng5 i Eng9
EngI_N(6)-, -Hyd. Engl_N(6)-i |-Agr. |-Hyd.
U.J3
075 B | i I _!
U
O 0 NS c EW| 1 0 NS 0 E\V 1

0.3 ...... ...D 0.20 _

So o
ft 0.15
0.25 tr *M

0.2
s D Cn ".
0.10 - o
D ;

iD
Q DO Q
0.15 0.05
0 @ O"B ;
n i 0.00 1 1 1 1 1
10 15 20 25 30 35 10 15 20 25 30 35
N-value N-value
(a) Period of vibration (b) Damping ratio

FIG. 12. The results of transfer function fitting method for moderate height buildings
14-9
 Alluvium layer depth
; -ICOm CJ
Dilu- ["Spread iSSwa |
c
3 vium L Pile
3 -SOm 3
a
oo

0
-40m to Allu- r Spread M 1 good estimation
vium L Pile 3^PSpSJw3 ' 1
-20m
/
poor estimation
e -10m
Dilu- T Spread mm i
CJ
c/5 vium L Pile
> M~\
Allu- r Spread
rt
O Pile foundation H vium L Pile ^i48^BS$ssrawftSK:SWf! 1
Spread foundation
0 5 10 15 20 25 30 35

FIG. 13. Soil condition in Nagoya city and the location of FIG. 16. The quality of transfer function estimation
investigated buildings for different conditions
1U

9
.
-
FrNatcequueranl cy
Superstructure
(Hz * *

^
OS
CO
U>
-J. * A '
-
/
5 10 5 10
Frequency (Hz) Frequency (Hz) - m rj A * x ' E Spread. Alluvium
(a) Building Y (b) Building S A A/ A Spread. Deluvium
' Pile. Alluvium
^' A Pile. Deluvium
FIG. 14. Examples of Fourier spectra for two representative buildings
XI II II

56789 10
Natural Frequency of
Soil-structure System (Hz)
(a) Longitudinal direction

* HB * "
A A AD> '
Eg
i# *>'
*A,
A
d
A
Spread, Alluvium
Spread, Deluvium
5 Pile, Alluvium
X A Pile, Deluvium
4
6789 10
Natural Frequency of
Soil-structure System (Hz)
(a) Building Y (good estimation) (b) Building S (poor estimation)
(b) Transverse direction
FIG. 15. Good and poor estimations of transfer function and coherence FIG. 17. Natural frequencies of superstructure
compared with those of soil-structure
systems
2.0

,z = 29, /? = 0.43
y = 0.21 A- -0.47
0.0 ' ' ' i = I 0.0
56789 10 567 8 9
Natural frequency (Hz) Natural frequency (Hz)
(a) Longitudinal direction (b) Transverse direction
FIG. 18. Correlation between the aseismic performance index (Is) and the natural frequency of the superstructure
14-10
clearly show lower natural frequencies for soil-structure systems buildings. A clear trend of lower natural frequencies for higher
which is an explicit consequence of SSI effect. levels of vibration is observed. However, the results for damping
From another point of view, Fig. 18 shows a correlation ratios are almost inconclusive.
between the observed natural frequencies of buildings and their
aseismic performance index, /5. This in turn points to the
possibility of easy and rapid evaluation of the aseismic 5. DYNAMIC BEHAVIOR OF A 10-STORY BUILDING
performance of buildings through microtremor measurements. IN NAGOYA UNIVERSITY
The building #1 studied in Sec.4 is studied in more detail in
4. SEISMIC OBSERVATION IN NAGOYA UNIVERSITY this section (Tobita et al. 1997, 1998b). The building as shown
in Fig.25 is a 10-story Steel Reinforced Concrete (SRC) structure
The vibration data from six instrumented buildings in the with irregular plan. The building is located on deep soft soil
Higashiyama campus of Nagoya University were collected during deposits (building No. 14 in Fig. 14) and consequently has pile
three weak and moderate earthquakes (Tbbita et al. 1998a). The foundation with long piles up to 45 meters. The dynamic behavior
collected data were then analyzed in order to study the efect of of the building was studied through precise microtremor
SSI on the dynamic characteristics of buildings as well as on the measurements. The observation points are also shown in Fig.25.
effective input motions. Figure 19 shows the outline of The translational and torsional mode shapes of the building
investigated buildings which are numbered in descending order are drawn in Fig.26 where the effect of building's irregularity is
of their area. Also shown are the locations of seismometers. seen clearly. Table 6 shows the sway and rocking ratios computed
The information of the investigated buildings are summarized based on displacements at the top story for different points of
in Table 3. It should be mentioned that the buildings #1, #3 and the plan. Rocking ratios even up to 50% is observed for the
#4 in this table are, respectively, the same buildings #14, #4 building at the eastern side. The ratio is 26% at the western side
and #8 in Table 1 which were studied in Sec.3.1. More detailed where the foundation width is bigger. But even at the western
results on building #1, i.e., the 10-story building will be discussed side, the rocking ratio is much laiger than would be expected for
in Sec.5. buildings supported by pile foundations. Such laige values of
Figure 20 shows the acceleration response of the buildings at rocking ratio may be considered as a result of long piles and of
roof, 1st floor and ground levels. Also shown are the Fourier softness of the underlain soil.
spectral ratios for the two cases of top/1 st floor and top/ground. The spatial vibration shape of the building at the top floor are
It is believed that the former case may be considered as a also presented in Fig. 27 in a different style. Deformation of the
representative for the fixed-base structure whereas the latter case first floor slab is shown in Fig.28. An obvious out-of-plane
includes the SSI effect too. The results thus show a clear change deformation is observed which is mainly due to different rocking
in the natural frequency of the system due to SSI for the 6-story motion in the two sides of the building. This complex behavior
and the 4-story buildings. Regarding the vertical vibration of is among parameters which complicate the study of SSI effect
buildings, remarkable difference is seen among the response of on this building.
the three buildings.
The peak acceleration and velocity at the ground level and
the first floor of all buildings have been summarized inTable 4 6. CONCLUDING REMARKS
for the case of Mar. 16,1997 earthquake. Investigating the results
reveals that there is more input motion loss for buildings with The effect of SSI on the dynamic properties of structures was
larger foundations. The loss of effective input motion is also studied analytically and experimentally As a result, the dominant
studied through construction of Fourier spectral ratio graphs in effect of SSI for the cases of low and medium-rise buildings was
Fig.21. The results are drawn versus nondimensional frequencies. clearly recognized. The main concluding remarks are as follows:
- Simplified methods were adapted for estimation of frequency
The loss of input motion is seen clearly in the results, especially
for the range of higher frequencies. and damping ratio of soil-structure systems which may be used
The results of natural frequencies and damping ratios for the for the design of ordinary buildings. The results were then
superstructure are drawn versus the corresponding values of the compared with currently available provisions.
soil-structure system in Fig.22. The results for natural frequencies - The natural frequency and damping ratio of a few low and
clearly show lower values for soil-structure systems which can medium-rise buildings were evaluated through microtremor
be interpreted as SSI effect. Also, the effect is more drastic for measurements. The effect of size of building on the severity of
larger earthquakes (EQ2>EQ3>EQ1). Regarding the damping SSI effect was then studied through comparison of the results.
ratios, the results for building #3 are compatible with the general Also, microtremor tests were conducted on several buildings with
expectation of higher damping ratios for soil-structure systems. the same size but located on different soil conditions. It was
However, the results of building #4 show a diiferent trend which concluded that identification of dynamic characteristics of
may be due to the effect of neighboring building's vibration buildings can be quite difficult for cases in which SSI is dominant.
- The dynamic behavior of three buildings was studied during
through structure-soil-structure interaction. An example of the
curve fitting process for the building #1 is shown in Fig.23. three low and medium size earthquakes. The loss in the effective
input motion was clearly observed, especially for buildings with
Finally, Fig.24 shows the amplitude dependency of the results.
The estimated natural frequencies and damping ratios for different large foundations. Also studied was the amplitude dependency
cases with various levels of vibration amplitude are drawn in the of the results which leads to lower frequencies and laiger damping
same figure versus the peak acceleration values at top of ratios for more severe excitations.

14-11
 46m

G.L.

O 16.500V
Plan
FIG. 19. Plan and elevation of investigated buildings and pickup points (O rground, :structure).
TABLE 3. Outline of investigated buildings TABLE 4. Maximum acceleration and velocity values
during Mar. 16. 1997 earthquake
No Structure Floor Heiglil(m) Area(m2) Foundation type Av. Vs (m/s) Max. Acce. (jal Max. Vel. (kine )

(i) SRC 10 39.3 1,541 Pile (45m) 202 G.L. IF G.L. IF G.L. IF G.L. 1 IF G.L. IF G.L. IF
C2) RC 3 (12) 1,370 R.C. Pile (315) to 72.1 48.4 97.2 60.6 37.0 18.1 3.95 3.85 5.81 5.42 .80 .59
.61
RC 4 17.9 1,155 Pile (6m) 288 (2) 54.0 52.7 63.2 49.8 34.7 37.9 2.49 4.04 4.58 3.28 .58
RC 6 22.3 604 P.C. Pile (12m) 303 (a> 50.3 57.3 73.5 50.0 31.4 24.1 2.8V 4.02 4.51 4.05 .81 .93
PO 44.2 4? 7 55 <> 587 388 ?05 ?S3 369 4.69 4.40 91 88
RC 2 (9) 290 Spread and R.C. Pile (5m) 291 61.2 49.1 69.7 61,3 37,5 31.1 3.79 4.40 4.50 3.86 .73 .63
(6) RC 1 (15) 150 R.C. Pile (7m) 305 54.5 61.3 74.0 80.4 57.6 51.6 3.19 4.32 5.26 4.84 .96 .90

NS Oct. 5, 1996 (DEW Oct. 5. 1996

r
.a 1.0
T
Mar. 16, 1997
May 24, 1997
Mar. 16. 1997
May 24. 1997

|
0.0
1.0 2.0 3.0 4.0 " 00 1.0
3.0 NSy 6'55Hz 3'
I ,

1.0 2.0 3.0 4.0 ' 0.0 1.0 2.0 3.0 4.0
lO.SHz 3.0
NS EW
T

0.0 i rt)
10.1 Hz ?o
NS EW

.\J J\J H

Building (6 story), EW
0.0 1.0 2.0 3.0 4.0 1.0 2.0 3.0 4.0
iMvV^ - r 15.7HZ 3.0

EW

r" . .x
.
j 0
jJJurJJMfVt^^u.,
,...,,,,",.,,.
I . 2.0 NS

1.0 2.0 3.0 4.0 "00 1.0

Building (6 stoiy), UD 3.0 21.7Hz 3.0

a
NS EW
max=I30cm/s2 - 10 6 Z '
EQ.I EQ.I
EQ.2 EQ.2
J. EQ.3 EQ.3,
1.0
<JA^
2.0 3.0 4.0 ( 1.0 2.0 3.0 4.0
31.7Hz
t,\J JV *\J
Nondimensional Freq. CO^JA /2Vs
Building ( 10 story), EW
FIG. 21. Fourier spectral ratio (1 st floor / ground)
max=123cm/s2 20 plotted versus nondimensional frequency

FIG. 20. Response acceleration (cm/s2) during Mar. 16,

20 30 40 10 15 1997 earthquake, and Fourier spectral ratios
Building ( 10 story), UD (top/1st floor, top/ground)
14-12
 2.1 5.0
Building
_
O -"
2.0 (4 stoiy)

f(f 1-9 5.4.6 QEQI

C"C/3
K
t3 1.8
1|
crco

II

11 3.5
A n

il"
o-co

OEQ.3

z.| 1.7 "3 ^^ 4.2

CO QEQ.2

1.6 3.0 4.0

1.6 1.7 1.8 1.9 2.0 2.1 3.0 3.5 4.0 4.5 5.0 4.0 4.2 4.4 4.6 4.8 5.0
Natural Frequency of Natural Frequency of Natural Frequency of
Superstructure (Hz) Superstructure (Hz) Superstructure (Hz)
0.10 0.10 0.10
EQ.3

E 0.08 ^ E 0.08 0.08

"o H
EQ.l
EQ.2 x
| w 0.06 |f 0.06 EQ.2 If 0.06
oi K EQ-3.
co 5 op3
EQ.l OEQ.2
If 0.04 'I I 0.04 |1 0.04
15
[:%0^ Building o|
OT 0.02 Building
EQ.l O Q'i
^ 0.02 " 0.02 >EQ2 Building
EQ.I (10 story) (6 story) (4 story)
0.00 0.00 0.00
0.00 0.02 0.04 0.06 0.08 0.10 0.00 0.02 0.04 0.06 0.08 0.10 0.00 0.02 0.04 0.06 0.08 0.10
Damping Ratio of Superstructure Damping Ratio of Superstructure Damping Ratio of Superstructure

FIG. 22. Correlation between estimated parameters of superstructure and those of the soil-structure system

10.0

-18<
4.0 6.0 8.0 10.0
Frequency (Hz)
FIG. 23. Transfer functions fitting for building #1 (Mar. 16, 1997 earthquake)

2.00 El
ini crotremor
(1995) LN.
i i
5-:; EQ, Top/lF
\rr, Top/i F
Z.JO
microtremor
I 2.25 - (1995)
5" 2.15
HEW"-; EQ, Top/lF
MT, Top/lF
1.90
111 III | 01iQ, Top/GL
I 2.05
II Hill Mj'i 1 ' - no EQ, Top/GL
(1997)
!rV ( i j DIVTT, Top/GL (1997) VfT, Top/GL
1.80 Fin
Tff
i i It
in
>hooii"
r^
o 1
-J Si i QS

| 1.85
^t 1111 O r^
O 1
1.70 D - Fypho -in 0
997) z 1.75 (1997 )
1.60 Mil III b 1.65 in 0
0. 01 0.1 1 10 100 1000 0.01 0.1 1 10 100 1000
ri -rn 0.07
0.05 u
0 o " 0.06 [ O o
0.04 P 3- -
n ., | 0.05
0.03 * ^w ^ 0.04
-]
1^ H -
EQ, Top/lF - !>-_-. '1 0.03 EQ. Top/lF
V
(gffli
0.02 MT,Top/lF i'r
fil ! - 1 0.02 MT, Top/lF
T
!
0.01
n no
O EQ, Top/GL
D MT, Top/GL

.01 0.1 1
1 10 100
NS
ui
1000
o on
0.01
O EQ, Top/GL
0.01 D MT, Top/GL

0.1 1
1
10 100
EW
M
1000
Peak Ace. at Top (cm/s: Peak Ace. at Top (cm/s-)

FIG. 24. Amplitude-dependency of natural frequencies and damping ratios for building #1
14-13
 side. NS comp.
si side, NS comp.
Icr. NS comp
ler. EW comp.
h side. EW comp.

IF.
0 5 10 15 20 25 30 35
(umXsec)

FIG. 26. Mode shapes of translational and torsinal modes

estimated by the amplitude of displacement

TABLE 5. Sway and rocking ratios for NS and EW directions

West, NS comp. Center, NS comp. East, NS comp. Center, EW comp.
Plan (2nd - 10th floor)
Ratio Amp. Ratio Amp. Ratio Amp. Ratio Amp.
(%) (/jmXsec) (%) t/j mXsec) (%) lli mXscc) (%) (fimXsec)
Total Top Disp. 100 31.5 100 19.7 100 24.6 100 14.6
FIG. 25. Outline of the building and Rocking 26 8.2 50 12.3 22 3.2
microtremor observation points Sway 8 2.5 7 1.4 8 2.0 13 1.9
Elastic Deform. 66 20.8 48 11.8 69 10.1
Ground 2 0.6 3 0.6 3 0.6 5 0.7

l-.
-- _i.:ui^ 111 a --4

. -
Ccmer
.9.50
10.30
9.55
- 9.60
; : h e -
........
10.40
10.45 9.65
i : 1! * 10.50 9.70
- 10.55
i I lit 10.60 9.75
! ! I: S " 10.65 9.80
10.70 9.85
10.75
ii ! i . HI: Souihsid* - '- 10-S9 9.90

(a) NS translational mode (1.82Hz) (b) EW translational mode (1.92Hz) (c) torsional mode (1.82Hz)

FIG. 27. Horizontal vibration shapes of the top floor of translational and torsional modes

T=79.00 sec T=79.05 sec 7=79.10 sec

FIG. 28. Out-of-Plain deformation of the 1st floor (NS Translational Mode, 1.82Hz)
14-14
 - The dynamic behavior of a 10-story building was studied Kausel, E. & Roesset, J. (1974). "Dynamic stiffness of circular
foundations", Engng. Mech. Div.,ASCE, Vol.101, 771-785
precisely through dense microtremor measurements. Large
rocking motions were observed for the building in spite of Kokusho,T (1980). "Cyclic triaxial test of dynamic soil properties for
existence of pile foundation. Different levels of rocking motion wide strain range", Soil and Foundations, Vol.20,45-60
were observed for the two sides of the building due to change in NEHRP Recommended provisions for the development of seismic
the size of foundation. This leads to out of plane deformation of regulations for new buildings (1994). The Building Seismic Safety
floors which, in turn, complicates the experimental evaluation Council for the Federal Emergency Management Agency
of dynamic characteristics of the building. Novak, M. (1975). "Additional note on the efect of soil on structural
Finally, the need to improve the reliability of simplified response", Earthq. Engng. Struct. Dynam., Vol.3, 312-315
formulations was pointed out which may be achieved by
Tobita, J. (1996). "Evaluation of nonstationary damping characteristics
increasing the quality and quantity of experimental studies on of structures under earthquake excitations'Vor/io/ of Wind Engineering
ordinary buildings. and Industrial Aerodynamics, Vol.59, Nos.2,3, 283-298
Tobita, J., Fukuwa, N., Nishizaka, R. and Nishiyama,T. (1997).
"Evaluation of damping properties of structures based on high density
REFERENCES
earthquake observation in Nagoya University Higashiyama Campus",
D&D '97 Symposium, JSME, 41-44, (in Japanese with English abstract).
Architectural Institute of Japan (AIJ) (l996)An Introduction to Dynamic
Soil-Structure Interaction (In Japanese) Tobita, J., Nishiyama,T. and Fukuwa, N. (I998a). "Observed soil-
Applied Technology Council (1978). Tentative Provisions for the structure response characteristics of low- and middle-rise reinforced
concrete buildings", 5th Symposium on SoilStmciure Interaction, AIJ,
Development of Seismic Regulations for Buildings ATC Publications,
ATC3-06 131-136, (in Japanese).

Celebi, M. (1997). "Response of Olive View Hospital to Northridge Tobita, J., Nishiyama,T., Fukuwa, N., Nishizaka, R. and Murahashi, R.
and Whittier earthquakes",/. Stnic. Engng.,ASCE, Vol.123, No.4,389- (1998b). "Three-dimensional vibration characteristics of a 10-story SRC
396 building observed by microtremor test", 10th JEES, (in Japanese with
English abstract), (will be published).
Earthquake Resistant Regulations -A World List (1988). International
Association for Earthquake Engineering,Tsukuba, Japan Veletsos, A.S. (1977). "Dynamic of structure-foundation systems".
Structural and Gcotechnical Mechanics, A Volume Honoring N.M.
Fukuvva, N., Ghannad, M.A. &Yagi, S. (1995). "A study on the effect Newmark (Editor: W.J. Hall), Prentice-Hall, Inc. Englewood Clifs, N.J.,
of soil-structure interaction on the eigenproperty of structure' V- Struct. 333-361
Constr. Engng., Transactions of Arch. Inst. of Japan (AIJ), No.475, 35-
44, (In Japanese ) Wallace, J.W. and Moehle, J.P(1990). "Evaluation ofATC requirements
for soil-structure interaction using data for the 3 March 1985 Chile
Fukuwa, N. andGahnnad, M.A. (1996). "Soil-structure interaction effect earthquake", Earthquake Spectra, Vol.6, No.3, 593-611
on the eigenproperties of structure",/)roc. J1WCEE, Acapulco, Mexico,
Paper No. 949 Wolf, J.P. (1985). Dynamic Soil-Stnicture Interaction, Prentice-Hall,
Englewood Cliffs, N.J.
Ghannad, M.A., Tobita, J., Fukuwa, N., Nishizaka, R. and Koide, E.
(1997a). "A study on the effect of soil-structure interaction on the Wolf, J.P. (1994). Foundation Vibration Analysis Using Simplified
dynamic properties of RC structures based on the microtremor records", Physical Models, Prentice-Hall, Englewood Clifs, N.J.
/. Structural Engng., Architectural Inst. of Japan (AIJ),Vol.43B, 441- Yagi, S., Tobita, J., Fukuwa, N. and Nishizaka, R. (1998). "Estimation
450 of dynamic characteristics and effect of soil-structure interaction at R.C.
buildings based on the microtremor tests",5f/i Symposium on Soil
Ghannad, M. Ali, Fukuwa, N. andTobita, J. (1997b), "The effect of
Structure Interaction, AIJ, 125-130 (in Japanese).
soil-structure interaction on the damping of structure",/Yoc. 7th
International Conf. on Structural Safety and Reliability (ICOSSAR'97) Yasui, Y. and Tokimatsu, K. (1998). "Outline of damage to buildings
Kyoto, Japan, Paper No. WeB 1-01 and foundations during the Hyogo-ken Nanbu Earthquake of 1995",
Proceeding of the 5th Symposium on Dynamic Soil-Stiucture Interaction
Ghannad, M.A., Fukuwa, N. and Nishizaka, R. (1998). "A study on the
frequency and damping of soil-structure systems using a simplified Effect, Sponsored by Architectural Institute of Japan (AIJ),Tokyo, Japan,
3-12
model",/. Structural Engng , Architectural Inst. of Japan (AIJ),Vol.44B,
85-93
Ghannad, M.A. ( 199$).A Study on the Effect ofSoil-Stmctwc Interaction
on the Dynamic Pivperties of Structures Using Simplified Methods,
Thesis presented to Graduate School of Engineering, Nagoya Univin
partial fulfillment of the requirements for the degree of Doctor of
Engineering
Hayashi, Y., Yasui, Y. and Yoshida, N.(1998). "Effects of soil-structure
interaction in heavily damaged zone during Hyogo-ken Nanbu
Earthquake", Proceeding of the 5th Symposium on Dynamic Soil-
Structure Interaction Effect, Sponsored by Architectural Institute of
Japan (AIJ), Tokyo, Japan, 13-24
Hurty, W.C. and Rubinstein, M.E (1964) Dynamics of'Structures,
Prentice-Hall Inc.

14-15
 SOIL-STRUCTURE INTERACTION ANALYSIS
VIA FIXED-BASE SYSTEM
SUBJECTED TO A MODIFIED GROUND MOTION

M.Nuray Aydmoglu1, .eref Polat2 and Kemal Beyen3

ABSTRACT : Analysis of seismic soil-structure interaction is reformulated to define an equivalent

(modified) ground motion and an equivalent pseudo acceleration for the first mode response of a given
fixed-base structure under a given earthquake. Equivalent ground acceleration corresponds to the total
acceleration of the "rigid" structure defined at the centre of modal inertia forces developed in the first
vibration mode of the fixed-base structure. A typical example dealing with an extremely stiff and
stocky building actually constructed in the city of Dinar (Turkey) after 1995 Dinar earthquake on
relatively soft soil conditions exhibited a considerable amplification in terms of equivalent (modified)
earthquake ground acceleration. Set of numerical results also presented in this contribution deals with
the generation of modified pseudo-acceleration response spectra under the same earthquake (Dinar
1995) for typical shear wall buildings. Such buildings are selected to represent those constructed by
the so-called "tunnel formwork* system in Turkey during the last decade for mass-construction of
multi-storey residential buildings. Response spectra ordinates which are calculated and plotted by
varying the soil stiffness and the building aspect ratio, clearly demonstrate the expected effects of soil-
structure interaction particularly on stiff, short-period buildings.

INTRODUCTION

The idea behind the present paper has originated from the observation of extensive use
of the so-called "tunnel formwork" system in Turkey during the last decade for mass-
construction of multi-storey residential buildings. Such buildings are made of merely
reinforced concrete walls and slabs resulting in extremely stiff structural systems
(Fig. 1). When they are supported through raft foundations on relatively soft soils,
almost ideal conditions are created for the pronounced effects of seismic soil-structure
interaction.
Contrary to traditional engineering approach and the established seismic code
procedure of soil-structure interaction analysis based on the appropriate modification
of fundamental period and the effective damping of the structure (BSSC 1995), the
present study attempts for the determination of a fictitious "equivalent input ground
motion" to be applied to fixed-base structure system in the fundamental mode without
necessarily modifying its dynamic properties. Thus, this alternate approach makes it
possible to develop "modified response spectra" applicable to first mode response of
soil-structure systems. Although such output requires the full analysis of SSI systems,
it may be justified if the systems under consideration can be standardised, as in the case
of above mentioned uniform shear wall building structures which may be characterised
by few independent parameters only.
The present paper does not have any claim in terms of originality of research
except that SSI effects are demonstrated in a different fashion. It is anticipated that
practical engineers who, in part, still experience some difficulty in understanding the
favourable and/or unfavourable effects of soil-structure interaction, may find it more
interesting to observe how SSI affects the building response as the effect is directly
read from the familiar pseudo-acceleration response spectra.

1 Professor, 2 Graduate student, 3 Dr.Eng., Bogazici University, Kandilli Observatory and Earthquake
Research Institute, Department of Earthquake Engineering, 81220 Qengelkoy, Istanbul - Turkey

15-1
EQUIVALENT GROUND MOTION AND PSEUDO-ACCELERATION

A systematic formulation of soil-structure interaction for building type structures is

presented in the Appendix. The following derivation, on the other hand, aims only to
give the definition of'"equivalent ground motion" and that of the "equivalent pseudo-
acceleration"'. Reference is made to the Appendix as necessary.
Referring to the soil-structure system shown in Fig.2, equations of motion of the
structural part of the system can be expressed in the frequency domain as,

([KJ+ icotCJ - co2 [MJK8J = - [MJ [Tso]{ u'0 } (1)

Being the first row of equation of motion of the soil-structure system given by Eq.Al
of the Appendix, in the above expression co denotes excitation frequency, [KJ, [CJ,
[Mss] represent structural property matrices and {6S} refers to structure's horizontal
displacement vector relative to the rigid foundation. Total acceleration vector of the
rigid foundation, {ii^}, includes sway and rocking degrees of freedom only, as
indicated in Fig.2:

(ut0} = <utx e>T (2)

In Eq.l, [Tso] is as given by Eq.A4 of the Appendix and represents the kinematic
transfer matrix for transmitting total motion of the rigid foundation to the structure.
Referring to the Appendix, relative structural displacements, {8S}, appearing in Eq.l
can be expressed in terms of modal coordinates of the fixed base structure. Resulting
modal equation of motion written for an r'th mode is as follows:

(Qr2 -co2 + 2ieD^r a)Ysr = -{<Dr}T [Mss][Tso]{u t0 } (3)

Since soil-structure interaction effect is dominantly pronounced in the first vibration

mode only, by considering Eq.2 above expression can be written for the first mode as,

(ni2 -o>2 + 2ia1 Q1) YS,=-<LX, M^xti 1* 6>T (4)

where Lxl represents the participation factor in the first mode (Eq.A20) and hi refers
to the centre of the effective seismic forces in the first mode of the fixed-base structure
measured from the level of soil-foundation interface (Eq.A21). By using variables yi
and iie defined in Eq.A25 and Eq.A30 of the Appendix, respectively, Eq.4 can be
rewritten as (Fig.2),

(Q,2 - a)2 + 2i co , Q,) Ysl = - Lxl (ii'x + y, ue ) (5)

Thus, "equivalent horizontal ground acceleration1' defined for the first mode of the
fixed-base structure to replace the horizontal component of the effective foundation
input acceleration, ii 8x , is determined as,

ex,eq =u tx + y, ue (6)

By the above definition, it can be stated that equivalent horizontal ground acceleration
corresponds to the total horizontal acceleration of the "rigid" structure defined at the

15-2
centre of modal inertia forces developed in the first vibration mode of the fixed-base
structure.
Under the above defined equivalent horizontal ground acceleration, equation of
motion of a single degree of freedom system with natural frequency of QI is expressed
in the frequency domain as,

(Q,2 - o>2 + 2i , Q,) us = - u gx,eq = - (ii'x + YI ue ) (7)

"Equivalent pseudo-acceleration", apsi, is then determined in the frequency domain

as,
aps, = Q,2 us = - (D, / p,2) (ii', + YI ue ) (8)

where P I = co / Q! and DI is as defined in the Appendix (Eq.A16b).

In practical applications it is appropriate to obtain the required response quantities
first in terms of transfer functions in the frequency domain. Since condensed equations
of motion given in Eq.A31 of the Appendix are also valid for accelerations, i.e., {u !0 }
and {ii go}, in lieu of displacements, transfer functions U gXjeq and Apsi corresponding
to equivalent horizontal ground acceleration and pseudo-acceleration, respectively, can
be obtained from Eq.A34 of the Appendix by imposing unit amplitude to the effective
foundation input acceleration, i.e., u gx = 1 and utilising Eqs.6,8 given above. Pseudo-
spectral acceleration for the first mode is then obtained in the time domain as,

Spa,, = [IFT (Apsl * FT (u gx))]mai (9)

where abbreviations FT and IFT denote Fourier Transform and Inverse Fourier
Transform operations, respectively, for which well known Fast Fourier Transform
algorithm is ideally suited. In the following examples complex, frequency dependent
stiffness influence coefficients of rigid surface foundations given in (Veletsos, Wei
1971) are utilised where equivalent radii are calculated for the sway and rocking
motions of the rectangular surface foundation considered in this study.
Condensed equations of motion in terms of rigid foundation degrees of freedom are
given in detail by Eqs.A31, A32, A33 and A34 of the Appendix. Nondimensional
parameters are, r^e^ne,, Tjee (defined by Eq.A9), e (defined by Eq.A6), X,i (defined by
Eq.A24), Yi (defined by Eq.A25) and three independent, nondimensional SSI
parameters defined by Eqs.A27. However in dealing with the above mentioned
uniform shear wall structures, it may be more appropriate to use the following
independent, nondimensional SSI parameters in lieu of those defined in Eqs.A27.

6A = Ms /(ps ABHn) (lOa)

Tg = 4Hn /Vsg (10b)

aA =Hn /A (lOc)

where A = 2a and B = 2b are the plan dimensions of the rectangular raft foundation
in the earthquake direction and in perpendicular direction, respectively. ps represents
the mass density of the superstructure material. (For reinforced concrete, ps 1.35 pg
may be assumed). Parameter 5A defined by Eq.lOa above may be taken nearly constant

15-3
for certain range of shear wall buildings. Parameters given in Eqs.A27 and Eqs.10 are
interrelated as follows:

= 2 aA (lla)

= (8KaA ps /pg)8A (lib)

= (ic/2)(Tg /T,) (lie)

where TI refers to the first natural vibration period of the superstructure (Ti = 2n /li)
and K is defined as
K =B/A = b/a (12)

NUMERICAL RESULTS

a) Equivalent ground acceleration applicable to first mode response

As an example of the above explained procedure, a five storey shear wall building is
analysed for Dinar Earthquake of October 1, 1995 (Ms = 6,0) by considering EW
component of a near-field strong motion record. Actually such buildings have been
built after the damaging earthquake occurred in the town of Dinar in western Turkey
for those people whose houses were either collapsed or demolished due to irreparable
heavy damage. Interestingly, buildings were built on relatively soft soil conditions with
mat foundations such that pronounced effects of soil-structure interaction can be
expected. A typical storey plan and vertical cross section of the building are shown in
Fig.l. Wall and slab thickness throughout the building are 0.15 m and 0.14 m,
respectively, whereas mat foundation is 0.40 m thick. SSI parameters defined in this
paper are calculated as, aA=0.728, K=0.927 (A=19.71m, B=18.27m, Hn=14.35m),
Tl,e=Tie,=0.568, Tjee^O.403, 6=0.271, 8A=0.1285, yi=0.6859, ^=0.8813. Fundamental
period of the fixed-base building is calculated as Ti=0.052 s, and average shear wave
velocity of the site is assumed to be Vsg = 200 m/s, leading to above defined fictitious
soil period uf Tg=0.287 s. Fig.3 shows acceleration time histories of actual and
equivalent (modified) ground motions applicable to fixed base structure in its first
vibration mode, respectively, which clearly demonstrate the pronounced response
amplification in such an extremely stiff building due to SSI effect. Contrary to the
beneficial effect of SSI which is usually the case in practice, this example shows that
constructing stiff, stocky buildings on soft soils may result in substantially higher
seismic response.

b) Modified response spectra applicable to first mode response

By inspection on above mentioned stiff buildings constructed by "tunnel formwork"

system, it can be concluded that parameters K, Tix6 = Tje, T[QQ, E and 8A may not vary
considerably in practice, whereas first-mode mass parameters K\ and yi can be
reasonably estimated by considering the nearly uniform mass distribution of such
buildings. Thus, by choosing building aspect ratio ceA and fictitious soil period Tg as
independent variables, sets of "equivalent pseudo-acceleration response spectra" for
the first mode response of the fixed-base structure can be generated in accordance with

15-4
Eq.9. Such response spectra shown in Fig.4 and Fig.6 are calculated for Dinar (1995)
earthquake EW component with the following constant parameters: K=l, T|xe='nex=0.5,
T|ee =0.33, e=0.175, 8A =0.15, X,i=0.75, -ft =0.75. The first independent parameter, i.e.,
fictitious soil period starts from Tg=0 (representing infinitely stiff soil) and increases up
to 1.40 s with 0.20 s increments. On the other hand, building aspect ratio (otA = Hn / A)
is taken as 0.75 and 1.50, respectively. In order to show the significance of SSI effect,
equivalent pseudo-acceleration response spectra amplitudes given in Figs.4,6 are
normalised with respect to those obtained for infinitely rigid soil and presented in
Fig.5 and Fig.7, respectively. Curves shown in those figures effectively correspond to
the first mode amplification (or de-amplification) spectra due to SSI effect.
Figs.4 through 7 clearly demonstrates favourable and/or unfavourable effects of SSI
for certain type of structures under a typical earthquake motion recorded on relatively
soft soil conditions. It is observed that peaks of the response spectra of the modified
ground motion generally shift to shorter periods with respect to no SSI situation. Thus,
as expected, response spectrum amplitudes for periods shorter than peak response
period tend to amplify, while those for periods longer than peak response period de-
amplify, indicating the well known effective period increase in the actual soil-structure
system. As a result of this trend substantial amplification occurs in very short-period
buildings, especially those with higher aspect ratios. In order to show the effect of
building aspect ratio, response spectra are presented in Figs. 8,9 for constant fictitious
soil periods of Tg = 0.4 s and Tg =0.6 s, respectively, under Dinar earthquake EW
component. In each case, building aspect ratio (OLA = Hn / A) is varied from 0.75 to
1.75 with increments of 0.25. Very high amplifications for short period buildings with
high aspect ratios are clearly observed. However amplifications are relatively less
pronounced as the soil softens.

CONCLUSIONS

Analysis of seismic soil-structure interaction is reformulated to result in an equivalent

ground motion and equivalent pseudo spectral acceleration for the first mode response
of a given fixed-base structure under a given earthquake. Equivalent ground
acceleration corresponds to the total acceleration of the "rigid" structure defined at the
centre of modal inertia forces developed in the first vibration mode of the fixed-base
structure. Expressions for equivalent ground motion and pseudo spectral acceleration
are obtained in the frequency domain in the form of transfer functions. In order to
obtain the desired output quantities in the time domain, convolution is applied to the
transfer function involved and Fourier transform of the earthquake input ground
motion through inverse Fourier transform. The well-known Fast Fourier Transform
Algorithm is efficiently used for the transform operations.
The first example dealing with an extremely stiff and stocky building actually
constructed in the city of Dinar (Turkey) after 1995 Dinar earthquake on relatively soft
soil conditions exhibited a considerable amplification in terms of equivalent (modified)
earthquake ground acceleration applicable to the first mode response of the building in
its fixed base condition. The second set of numerical results deals with the generation
of modified pseudo-acceleration response spectra under the same earthquake (Dinar
1995) for typical shear wall buildings for which certain parameters may be taken to be
constant. Response spectra ordinates are calculated and plotted by varying the soil
stiffness and the building aspect ratio, which clearly demonstrated the expected effects
of soil-structure interaction particularly on stiff, short-period buildings.

15-5
REFERENCES

Aydmoglu, M.N. (1980), Unified Formulations for Soil-Structure Interaction, Proc.

7th World Conference on Earthquake Engineering, Vol.6, pp. 121-128, Istanbul.

BSSC-Building Seismic Safety Council (1995), NEHRP Recommended Provisions for

Seismic Regulations for new Buildings, 1994 Edition, Part 1: Provisions and Part 2:
Commentary, Federal Emergency Management Agency, Washington, D.C.

Veletsos, A.S. and Y.T.Wei (1971), Lateral and Rocking Vibration of Footings,
ASCE, Journal of Soil Mechanics and Foundations, Vol.97, No.SM9, pp. 1227-1248.

Wolf, J, P. (1985), Dynamic Soil-Structure Interaction. Prentice-Hall, Englewood

Cliffs, N. J.

1 1
A
1I -I--
I
-

I _
1.

Figure 1. Plan and profile of the typical building

15-6
 Figure 2. Soil-structure system considered
DINAR EW (1995)
0.45

Fixed Base
Vsg=200 m/s

Figure 3. Equivalent (modified) earthquake ground acceleration for the typical building
to replace recorded ground acceleration in Dinar (1995) earthquake, EW component

15-7
 SSI amplification / deamplification factor Equivalent Spectral Acceleration (g)

CD
I"

^1 C/3
^ c/a
O h (
O aj
3
"si 1
xa>
a
- CD a D
Z
n
c m
i
(0 to
(V (0 to
o en en
'<>
I
TJ ro
p O (D
3 3 "n
5'
a
<3.S
r5 C
00 3
I3

HHHHHHHH HHHHHHHH ||
tatotototocQtoto tototoiouatoioio "
n n n n n n n n n n n n n n n n
_k_x_xoOOOO
4* ro o bo a* 4^ ro
(/)</)(/)(/)(/)(/)(/)
o'
a.
 SSI amplification / deamplification factor Equivalent Spectral Acceleration (g)
ro
'*>.

X
(D
a
er g
g
g P
(D 00
Tl
\
3
a. I
0) _x
<>
I_x
(O (O
(0 (O cn
3 en
o
(D

O
a
W

HHHHHHHH
H H H H H H H " " " "
CQ CQ CQ CQ IQ IQ IQ
II II II II II II II
_._._. O O O O
o oo ro
(/>
 Equivalent Spectral Acceleration (g)
Qa Equivalent Spectral Acceleration (g)
p
en
ro
en p
k> p ro
en en en

X
CD
Q. X
CD fD
fi> a
w p DO
CD CO g
Ul n 0)
I to
c fD
3 n
c m
0)a 3
a
01
(0
(O
(O
io en
ro
CD
5' o
ro
a 5'
a

Z! I I I I I M
X 3 3 3 3 3 O
n 111 111
X' 3 3 3 3 3 O
ffi >-. -. - - - ^
II II II II II
a > to
to ^ r* r* r* P CD || II II II II
o en
ro o en to -* r* 7* 7* p
* -o en ro o ^
ro en o en en
ro
APPENDIX : FORMULATION OF SOIL-STRUCTURE INTERACTION FOR
BUILDINGS

Referring to Fig.l, equations of motion of a building system with a rigid foundation

can be expressed in the frequency domain as (Aydinoglu 1980, Wolf 1985),

(Ala)
where,
' T
(u}=<{8s } (u'o}> (Alb)

([KJ+ fo[CJ - co2 [MJ) -co2 [MJ [Tso]

(Ale)
-co2 [TSO]T [Mss] ([Sg00] ~ co2 [Moo])

(P} T =<{0} T (P0} T> (Po> = [Sgoo]{u8o> (Aid)

in which o denotes excitation frequency, [KJ, [CJ, [MJ represent structural

property matrices, (ug0) refers to the effective foundation input motion and {8S }
represents the structure's horizontal displacement vector relative to the rigid
foundation (Fig.l). Since 2-D response of the soil-structure system is considered to
exclude vertical and torsional response, total displacement vector of the rigid
foundation, {u'o}, includes horizontal and rocking degrees of freedom only, as
indicated in Fig. 1:
(ut0} T = <u<I 6> (A2)

Dynamic stiffness matrix of the rigid foundation, [Sg00], is expressed as,

aSi
[Sgoo] = (A3)
aSe< a S 60

where Gg refers to shear modulus of the soil (ground) material, a denotes the half
length (or radius) of the rigid rectangular (or circular) raft foundation in the direction
of earthquake excitation, and SM, Sse=Sei, See are complex, frequency dependent,
nondimensional stiffness coefficients.
In Eq.Al, [Tso] represents kinematic transfer matrix relating the rigid part of the
total motion of superstructure to the degrees of freedom of the foundation which can
be expressed as (Fig.l),

1 1
(A4)
HI H2 . Hi . . Hn

where HI refers to the height of the i'th storey measured from the level soil-foundation
interface. Further in In Eq.Al, total foundation mass matrix, [M00], is expressed as,

[Moo] = [Mf00] + [TSO]T [Mss] [Tso] (AS)

15-11
in which [Mf00] denotes the mass matrix of the foundation itself:

1 0
[Mf00] = (A6)
0 a2/3

where thickness of the raft foundation is neglected and E denotes the ratio of the
foundation mass to the total mass of the superstructure, Ms. The latter is expressed as,

(A7)

in which m{ represents typical storey mass (Fig.l). By taking [MM] as a diagonal

(lumped) mass matrix, the second term of Eq.AS can be written in an open form as,

[TSO]T [MSS][TSO] = (A8)

or above expression can be simplified as,

[TSO]T [MSS][TSO] = (A9)

xHn TjeeHn2

It is obvious that for a single storey building with total mass concentrated at the top
TUe = ^ex = Tjee = 1, and in the special case of uniformly distributed mass along the
building height, Tjxe = Tjex = 1/2 and rjee =1/3.

Transformation to Modal Coordinates

{6S } in Eq.A.l can be expressed in terms of modal coordinates of the fixed base
structure as,
{6S}=[0]{YS} (A10)

in which {Ys} refers to the vector of modal coordinates and [O] represents mass-
normalised mode-shape matrix, i.e.,

(Alia)

Thus for a typical r'th mode,

(Allb)

(Allc)

15-12
in which Qr and r denote r'th mode natural frequency and modal damping factor of
the fixed base structure. Substituting Eq.AlO into Eq.Al and premultiplying the first
row by [3>]T yields the transformed equations of motion as follows:

[KR]{uR}= (P) (A12a)

where force vector given by Eq.A.ld remains unchanged and the displacement vector
takes the form,
{uR}T =<{Ys}T {u'.}T> (A12b)

with transformed dynamic stiffness matrix as,

[R] -V[L]
[KR] = (A12c)

where [L] represents the "modalparticipation matrix" which is expressed as,

[L] = [0]T [MSS][TSO] (A13)

and [R] is a diagonal matrix whose r'th mode element is,

Rr = nr2 -o)2 +2ia)r nr (A14)

Solving for {Ys} from the first row of Eq.A12 yields;

{Ys} = G>2 [R]-1 [L] {u'0 } = [D] [L]{u'0} (A15)

in which diagonal matrix [D] and its typical element corresponding to r'th mode are
defined as,
co^R]-1 (A16a)

Dr =pr2 /(l-pr2 +2ir |3r) (A16b)

where normalised frequency pr is defined as 0r = co / fir. Actually Eq.A16b represents

the complex frequency response of a fixed-based single-degree-of-freedom system with
natural frequency Qr subjected to a ground motion with unit displacement amplitude
and frequency of co.
Substituting Eq.AlS into the second row of Eq.A12 or in other words by
condensing out the modal coordinates of the superstructure from Eq.A12 yields;

[EooKu'o} = [Sg00]{ugo> (A17)

in which condensed dynamic stiffness matrix, [ <,], is expressed as,

[ ] = [SU - Q2 ([M ] + [L]T[D] [L]) (A18)

The term [L]T[D] [L] included in above can be expressed in an open form as,

15-13
 ZDr Lxr2
r=l r=l
[L]T[D][L] = (A19)

r=l r=l

in which Lxr represents the participation factor and Lxr2 the participating mass of the
fixed base structure for the r'th mode. The former can be expressed as,

(A20)

In Eq.A19, hr refers to the centre of effective seismic forces in the r'th mode of the
fixed-based structure measured from the level of soil-foundation interface:

hr - (All)

After Eq.A17 is solved for {u'o}, relative structural displacement vector can be
determined by utilising Eqs.AlO and A15 as,

{8s}=[<D][D][L]{u'0} (A22)

First Mode Approximation

It is well known that soil-structure interaction effect is dominantly pronounced in the

first vibration mode, and the higher modes can be taken into account without soil-
structure interaction considered. In this case Eq.A19 can be written as,

1 h,
[L]T[D][L]=D1 LI12 (A23)
hi h ,_

Since participating mass of the structure in the first vibration mode is a fraction of the
total mass, a new parameter KI can be defined as:
2 _
(A24)

On the other hand the location of the resultant of seismic forces in the first mode can
be expressed in terms of a new parameter yi as,

hi = yi Hn (A25)
Thus,

1 YiHn
[L]T[D][L] = (A26)
YiHn yi2Hn2

15-14
SSI Parameters

In addition to above defined system parameters, the following are the well known
independent, nondimensional parameters of soil-structure interaction:

a = Hn / a (A27a)

6 = Ms /(pg a3) (A27b)

s = n,Hn /Vsg (A27c)

where pg and Vsg represent the mass density and the shear wave velocity of the soil
medium (ground), respectively (Gg = pgV2sg). Nondimensional frequency parameter
ao = wa / Vsg can now be expressed in terms of above defined parameters as,

a0 = Pi s / a (A28)

For the special case of surface foundations and vertically propagating shear waves,
foundation input motion is expressed as,

{ue0} T = <ugx 0> (A29)

By defining a new variable, UG , as

ue = 6 Hn (A30)

Eq.Al? can now be modified as

[I00]<utx ue >T ={Qo} (A31)

and the elements of [!,] and (Q0) are then expressed in terms of the above defined
independent, nondimensional parameters, as follows:

Ixx = Sxx - 6 ac2 (1 + B + Ki D,) (A32a)

(A32b)

,,D1] (A32c)

Qx = SK ugx (A33a)

Qe = (Sex/ct)ugx (A33b)

For surface foundations, coupling term of the foundation stiffness matrix, S0x, may be
omitted and hence Qe = 0. In this case, solution of Eq.A31 yields;

u'x = SM Iee/(IxxIee-Ixe2)u8x (A34a)

ue = - S Ixe / (Ixx lee - k2) ug, (A34b)

15-15
 Study on Strong Ground Motions for the Application to Seismic Design of Structures
-Semi-empirical Method for Ground Motion Estimation and Non-linear Response Spectra-

Keiichi Tamura 1^ Yoshihiro Nakao2) and Ri'ki Honda3)

Abstract: The extremely strong ground motion generated by the Hyogoken Nanbu (Kobe) Earthquake
caused serious damage to many kinds of structures. The destructive ground motions in near field
generated by the intra-plate inland earthquake were attributed to the damage.
Recently semi-empirical method has been noted as the method synthesizing ground motions
including such strong motions in near field. Various earthquake source parameters have to be determined
for this synthesis, however it is very difficult to evaluate them with sufficient accuracy. The effects of
these parameters with expected variation on the estimated ground motions have to be evaluated for the
application of semi-empirical method to seismic design. In this paper firstly studied are the effects of the
source parameters on the estimated ground motions.
Under the extremely strong ground motions in near field such as generated by the Kobe Earthquake,
structures behave inelastically. The ground motion characteristics which affect the inelastic response of
structures should be considered when the seismic design force and input ground motion for seismic design
are discussed. Secondary studied in this paper are the effects of ground motion characteristics on the
inelastic response of structures.

1. Introduction described as JMA Kobe), whose peak acceleration

In the early morning of January 17, 1995, the was larger than 800 cm/sec2 and the peak of the
Hyogoken-Nanbu(Kobe) Earthquake occurred acceleration response spectra (damping h=5%) of the
causing serious damage to many kinds of structures. horizontal components was no less than 2g.
Although its magnitude was relatively moderate Ground motions generated by intra-plate inland
(Mj=7.2, Mj is Japan Meteorological Agency (JMA) earthquakes such as the Kobe Earthquake are near
Magnitude), extremely strong ground motions were field ground motions, which have different
generated. Its causative faults reached inland areas characteristics from those by inter-plate earthquakes
though the epicenter was located in the Akashi strait in strengths and frequency characteristics. From the
north of Awaji Island, which was unusual because serious damage to many structures caused by
most of the past large earthquakes in Japan were extremely strong ground motions from the Kobe
inter-plate earthquakes and therefor the causative Earthquake, it was recognized that 1) Near field
faults were in the ocean areas. The destructiveness of ground motion estimation techniques should be
the intra-plate inland earthquake was obviously improved and 2) Inelastic behavior of structures
attributed to its extremely strong ground motions in under extremely strong ground motions should be
near field, and records from the earthquake proved it considered for rational seismic design.
Many records show large peak acceleration and large 1) Ground motion estimation technique in near field
response spectra. One of such records was obtained Semi-empirical method has been noted as an
at Kobe Maritime Observatory of JMA (hereinafter effective technique for synthesizing near field ground
1) Head, Ground Vibration Division, Public Works Research motions. Various earthquake source parameters such
Institute, Ministry of Construction, DnEng as fault length, width and dislocation rise time have
2) Research Engineer, Ground Vibration Division, Public to be determined for this synthesis, however it is very
Works Research Institute, Ministry of Construction, MJEng difficult to determine these parameters with sufficient
3) Research Associate, Dynamics of Foundation Structures, accuracy. Effects of source parameters with expected
Disaster Prevention Research Institute, Kyoto UNTV, variation on the estimated ground motions should be
MEng.
16-1
evaluated when the technique is applied to seismic estimation point. Semi-empirical method can
design. In this paper the ground motion at JMA Kobe estimate short period component of ground motion,
from the Kobe Earthquake is estimated using source which are strongly affected by delicate underground
models which have fault parameters varied within structures on wave propagation path, because the
expected range, and the extent of the estimated Green's functions include complex effects of the
ground motion variation is evaluated. dynamic rupture process on the fault, heterogeneous
2) Seismic force and input ground motion for structures around the source and an estimation point.
seismic design considering inelastic behavior of
structures Ground motion estimation point

There are limits to the extent to which simply

strengthening structural members resists extremely
strong ground motions. Energy absorption capacity Ground motions generated at sub-faults
\\
should be improved by ensuring allowable
Fault Plane
appropriate displacement ductility.
Therefore it is important to study ground
motion characteristics which have effects on
nonlinear response of structures so that they can be
incorporated into seismic design force and input Fig. 1 Semi-empirical method
ground motion for seismic design.
This paper presents the effects of ground Temporal and spatial variations of dislocation are
motions on inelastic behavior of structures using considered by utilizing proper function Tmn and
strength reduction factor spectra and the ground weight function amn , respectively ( eqs. (1 ),(2)).
motion characteristics which contributes to the
effects.
m=l n=l fnn

2. Effects of variation of source parameters on (1)

estimated ground motions where,
(1) Ground motion synthesis method using Tmn : Function representing the difference in source
earthquake source model time function between small and large events.
Ground motions in short and long period ranges a)mn : Weight function representing spatial variation
are generated by semi-empirical and theoretical of dislocation
methods, respectively. Then, ground motions in short MO : Frequency component of ground motion from
and long period ranges are integrated into a ground small event
motion in wide period range. Rmn : Radiation pattern coefficient of the ground
(2) Semi-empirical method motion generated at subfault
In semi-empirical method ground motion RQ : Radiation pattern coefficient of the ground
records from small events such as foreshocks and motion from a small event
aftershocks with their hypocenters near the fault area tmn : Time delay due to fault rupture from
of a large event are utilized as Green's functions to hypocenter to subfaults and wave traveling
estimate a ground motion from the large event (Fig. from subfaults to the estimation point.
1). The fault plane is divided into subfaults as large N : Number of subfaults
as the fault size of the small events. The Green's
functions are summed up considering time delays
due to fault ruptures from hypocenter to subfaults
and wave traveling from each subfault to an

16-2
 COG
0 co<
1+ sid
1 . (n ln(fi>)-ln(flh)^ n
cor -^- coo
T XI
T,nn \)) = N exP ' - '' ' -
2 2 2 1 2 ln(n) '
2MV
\
2NNi
COT
2MV[ COG

D: Dislocation 0
dfc: Temporal valiation of dislocation at the time of /// (ftj) : High-pass filter
kAz (Fig. 2) : Low-pass filter
/Vl: Coefficient to avoid the synthetically generated : ft; component of the ground motion by
periodicity semi-empirical method
u-p (CD) : a component of the theoretical ground
motion
n, coG : Coefficients to determine the shape of filters
c
_o
"(0
o
_o D=NAD
en
Q

0 AT 2 AT (k-1)AT kAT T
Time
Fig. 2 Source time function
Frequency [Hz]
(3) Theoretical method
Long period component of a ground motion is Fig. 3 Pass filters
estimated by theoretical method in which the
solution for the elastdynamic Green's function in (5) Target ground motion and earthquake source
homogeneous, isotropic, unbounded medium is used parameters
as waves generated at subfaults. Ground motions Ground motions at JMA Kobe from the Kobe
from all subfaults are summed up considering time Earthquake with Mj=7.2 are synthesized. Basic
delay due to fault rupture from hypocenter to each earthquake source parameters such the location,
subfault and wave traveling from each subfault to the strike and dip angle of the event are determined on
estimation point. the basis of the source model proposed by Kikuchi
(4) Combination of long period and short period (1995) for the Kobe Earthquake, which are shown in
components of ground motions Fig. 4. Other source parameters such as fault length,
Ground motions estimated by semi-empirical width and dislocation rise time are determined by
and theoretical methods are combined by eq. (3), regressing the source parameters deduced for past
using high-pass and low-pass filters(Fig. 3). earthquakes whose magnitudes(Mj) were 6 or
u(a) = fH (a-uE (a))+fL (a))-uT ((o) (3) greater, which occurred after the 1923 Kanto
where, Earthquake. In the case that there are some source
models deduced for an earthquake, the most credible
source model is adopted to the regression analysis.
Standard deviations are given to source parameters

16-3
as expected variation ranges. Fig. 5 shows
relationship between magnitude and earthquake
source parameters which were deduced for past
earthquakes. Table 1 shows the source parameters
determined for the Kobe Earthquake (Mj=7.2) by the
678
regression analysis and standard deviations. Magnitude
(a) Magnitude - fault length
102
Estimation
point

10

6 7
Magnitude
(b) Magnitude - fault width

".2 1O2

Fig. 4 Earthquake source model for the Kobe Earthquake 12

Q

Table 1 Source parameters by regression analysis 6 7

Magnitude
Determined Standard deviations
Earthquake source (c) Magnitude - dislocation
Parameter
parameter
forMj=7.2 +0 -o
Fault length(km) 40.0 X1.68 X 1/1 .68
(67.2) (23.8)
Fault widthQon) 27.4 XI.71 X 1/1.71
10C
(47.0) (16.0)
Dislocation(m) 156 X1.67 X 1/1.67 ir
(2.60) (0.93)
XI.75 X 1/1 .75 10-
Risetime(sec) 2.28
(3.99) (1-30) Magnitude
Fault rupture velocity X1.29 X 1/1.29 (d) Magnitude - rise time
3.22
(km/s) (4.15) (2.50)
Dislocation - X22 X 1/2.2
velocity(m/s)
=2S. 5
-f 4
-2 3
Ground motions are computed with various
source models shown in Table 2. Casel is the basic
case using source parameters determined for an
Magnitude
earthquake with Mj=7.2 by regression analysis. In
(e) Magnitude - fault rupture velocity
this case the source time function proposed by
Irikura (1986) is assumed for semi-empirical method
and the ramp function is assumed for theoretical
method as its source time function(Fig. 6). Cases2 to
27 are cases in which source parameters are varied.
In contrast with Casel in which radial rupture 678
Magnitude
process from fault center is assumed, the focus and
(f) Magnitude - dislocation velocity
the rupture process are varied in Cases 12 to 18.
Case 19 is a source model which adds spatial Fig. 5 Relationship between magnitude and various
variation of dislocation to Casel on the basis of the parameters

16-4
 source model by Sekiguchi et al.(1996). Case20 varied from the source time function by Irikura
employs the source time function proposed by (1986). In Cases23 to 26 the fault plane is divided
Irikura(1986) for both semi-empirical and theoretical into A, B, C and D parts (Fig. 7). Temporal and
methods. Case21 and Case22 are derived from spatial variations of dislocation shown in Fig. 8 are
Case20 by changing the initial slope of the source adopted in Cases23 to 26. In Case27 the ramp
time function by standard deviation of dislocation function is assumed for both semi-empirical and
velocity (Table 1). Fig. 6 shows the initial slopes theoretical methods. Temporal variation of
dislocation is not considered in Case27.
Table 2 Source model cases
Model
Contents
name
Basic Case
The ramp function is assumed for semi-empirical
method (Fig. 6).
Casel
The source time function by Irikura (1986) is assumed
for theoretical method (Fig. 6).
Radial fault rupture extends from the fault center.
Case2 Variation of Casel in fault length by + O
Case3 Variation of Casel in fault length by O lrikura(1986)
Case4 Variation of Casel in fault width by + a Ramp function
CaseS Variation of Casel in fault width by O
Case6 Variation of Casel in rise time by + O
CaseV Variation of Casel in rise time by O
CaseS Variation of Casel in dislocation by + o
Case9 Variation of Casel in dislocation by o
CaselO Variation of Casel in rupture velocity by + O 0 1
Casel 1 Variation of Casel in rupture velocity by o Time [sec]
Variation of Casel in rupture start point. Radial rupture
Casel 2 Fig. 6 Source time function
extends from southwest upper comer.
Variation of Casel in rupture start point Radial rupture
Casel 3
extends from southwest lower comer.
Casel4
Variation of Casel in rupture start point Radial rupture Estimation point
extends from northeast upper comer.
Variation of Casel in rupture start point Radial rupture
Casel 5
extends from northeast lower comer.
Variation of Casel in rupture process. Unilateral
Casel 6
rupture extends from southwest
Variation of Casel in rupture process. Unilateral
Casel?
rupture extends from northeast
Variation of Casel in rupture process. Bilateral rupture
Casel 8
extends from fault center.
Casel 9
Spatial variation of dislocation is given to Casel on the Fault length
basis of Sekiguchi model.
The source time function by Irikura(1986) is given to Fig. 7 A, B, C and D parts of the fault plane
Case20 Casel (Fig.6). The source time function is assumed for
semi-empirical and theoretical methods.
Case21 + O dislocation velocity (Rg.6) is given to Case20.
A C
+ O dislocation -O dislocation
Case22 o dislocation velocity (Fig.6) is given to Case20.
+ o and o dislocations are given to A, D and B,
Case23 B D
C parts of Case 1 model, respectively. -O dislocation +O dislocation
o and + o dislocations are given to A, D and B,
Case24
C parts of Case 1 model, respectively.
(a)Case 23 (spatial variation of dislocation is
+ O and O dislocation velocities(Fig.6) are given
Case25 considered)
to A, D and B, C parts of Case 1 model, respectively.
o and+<3 dislocation velocities(Fig.6) are given Figs. 8(1) Cases 23 to 26
Case26
to A, D and B, C parts of Case 1 model, respectively
The ramp function (Fig.6) is given to Case20 as its
Case27
source time function.

16-5
 A C more than shorter period component by variation of
-O dislocation + O dislocation
fault length, rise time, dislocation and rupture
B D velocity (Fig. 10(a), (c), (d), (e)). Fig. 10(c) shows
+ O dislocation -O dislocation that variation of rise time has less effects on
estimated ground motion than another parameters.
(b) Case 24 (spatial variation of dislocation is 2) Rupture process
considered) In the cases that rupture extends unilaterally or
bilaterally, or focus is located at an upper comer,
A c acceleration response longer than 1.5 sec varies from
+ O dislocation velocity -O dislocation velocity
30 to 50 % of the another cases and long period
B D component is affected more than short period
- O dislocation velocity + O dislocation velocity component(Fig. 9(f)). Differences in rupture start
points and rupture process have large effects on
(c) Case 25 ( temporal variation of dislocation is
component longer than 0.7 sec(Fig. 10(f)).
considered.)
3) Temporal and spatial variation of dislocation
In contrast to Casel, spatial variation of
A C
-O dislocation velocity + O dislocation velocity dislocation is assumed in Case 19, 23 and 24.
Considering spatial variation of dislocation increases
B D ground motion component longer than 0.7 sec(Fig.
+ O dislocation velocity -O dislocation velocity
9(g)). Spatial variation of dislocation has larger
effects on ground motion component above 0.5 sec
(d) Case 26 (temporal variation of dislocation is
than component under 0.5 sec (Fig. 10(g)). The
considered.)
effects of spatial variation of dislocation on estimated
Figs. 8(2) Cases 23 to 26
ground motions are large as compared with fault
length, width, rise time, dislocation and rupture
(6) Effects of Source Parameters on estimated
velocity (Fig. 10(a)-(e),(g)).
ground motion
Different temporal variations of dislocation are
An aftershock ground motion (Mj=5.2) of the
assumed in Cases 1, 20, 21 and 22. Similar
Kobe Earthquake is used as a Green's function to
assumptions are adopted in Cases 25 and 26.
synthesize a main shock motion. Fig. 9 shows
Comparison of Case27, in which temporal variation
acceleration response spectra (damping h=0.05) of
of dislocation is not considered, and other Cases in
estimated ground motions. To study the extent of
Fig.9 (h) shows that temporal variation of dislocation
estimated ground motion variation, Fig. 10 presents
increases ground motion component in wide period
the ratio of the maximum spectral value to the
range, especially ground motion component shorter
minimum one for each natural period, in which
than 0.5 sec (Fig.lO(h)). Note that the temporal
source parameters are varied. The effects of source
variation is not assumed for the long period
parameters on estimated ground motion are
component of Casel. The source time function used
summarized as follows.
in Case27 is not thought to be proper practically
1) Fault length, width, rise time, dislocation,
because of the small short period component of the
rupture velocity
estimated ground motion. Among Cases 20 to 22
The ground motion component longer than 0.6
ground motion increases when initial slope of source
sec decreases as fault length, width or rise time is
time function increases. Consideration of the
increased by standard deviation, while it increases as
temporal variation has greater effects on ground
dislocation is increased (Figs. 9(a)-(d)). Ground
motion than the differences in initial slopes of source
motion component longer than 0.6 sec is affected
time function which are assumed in this paper (Fig.

16-6
 2) Spatial variation of dislocation has large effects on
(7) Conclusion ground motion, especially on long period
Ground motion at JMA Kobe from the Kobe component, as compared with fault length, width,
Earthquake is estimated. Effects of variation of rise time and dislocation.
various parameters on the estimated ground motions 3) Temporal variation of dislocation increases
are studied. The variation given to various ground motion in wide period range, especially
parameters are determined by past earthquakes. short period component When initial slope of
Following conclusions may be deduced from the source time function is increased, ground motion
study. is also increased. Consideration of the temporal
1) Long period component of ground motion is variation has greater effects on ground motion
affected more than short period component by than the differences in initial slopes of source time
variation of fault length, dislocation, rise time, function which are assumed in this paper.
rupture velocity and rupture process.

10000
[h-n n;t
10000
: ||,_nnc
10000 II, J' . Ifl
^ ..=L
a- V vA i
.O)
// \ to
OJ %
15
\f- ^k^
03
*4
1
0) / <B +*>
C 1000 c 1000

j
c 1000 -
o -fe o 0 n"
1-1
J\ CO
^ 0.
"V ^
Q) f5=J
elResf
eration 1 CC cc
V\\ k .V c
O
V\ \
*v\ "S
1
oo
\\ > 100 \ 100
J r
CD (O
V
U

CASE1 3. -^
* ^
V
CASE1 CASE1
CASE2 \> ^\ CASE4 V CASE6 1 J(
CASE3 CASES -- CASE?
10 ^ ' iu 10
0.1 0.2 0.5 1 2 0.1 0.2 0.5 1 2 5 0.1 0.2 0.5 1 2 5
Natural Period [sec.] Natural Period [sec.] Natural Period [sec.]
(a)Fault length (b)Fault width (c) Rise time
Fig. 9 (1) Acceleration Response Spectra

16-7
 1UUUU i ' r~r 10000L _._ 10000
It |h=0.05f
1 i! -;U

"to ; /V ! ! ! ' i

O) \l \ 4 I
; CO

\
/ der[gal]
Resaptionson e .O)
CD
^ j CD
w
c 1000 1000:^ 1000
O O
Q. i j\ 4- -* Q.
CO W
Q j >\J\* CD
CC o:
O
i\\
2 \M\
100 & 100
\I "t ^ 0 CASE13 -
8
CASE1
V
\
i\ 1
3 CASE14
CASE15
CASES \\ CASE16
CASE9 V^^
CASE17 -
CASE1B
,0
0 .1 0.2 0.5 1 2 5 0.1 0.2 0.5 1 2 5 0.1 0.2 0.5 1 2 5
Natural Period [sec.] Natural Period [sec.] Natural Period [sec.]
(d)Dislocation (e) Rupture velocity (f) Rupture process

10000 10000p=

CASE1
CASE20
CASE21 -
CASE22
CASE25
CASE26

0.1 0.2 0.5 1 2 5 0.2 0.5 1 2 5

Natural Period [secj Natural Period [sec.]
(g) Spatial Variation of Dislocation (h) Temporal Variation of Dislocation
Fig. 9 (2) Acceleration Response Spectra

16-8
 10 10

5
VS"
e< 2
O)
1
If
0.1 0.2 0.5 1 2 0.1 0.2 0.5 1 2
Natural Period [sec.] Natural Period [sec.]
(a) Fault Length (b) Fault Width
10c: T

CO
fi
1
0.1 0.2 0.5 1 2 0.1 0.2 0.5 12
Natural Period [sec.] Natural Period [sec.]
(c) Rise Time . (d) Dislocation

E< 5
flf
i

1
I
0.1 0.2 0.5 1 2 0.1 0.2 0.5 1 2
Natural Period [sec.] Natural Period [sec.]
(e) Rupture Velocity (f) Rupture Process

10

E< 5
c.

CO

1
0.1 0.2 0.5 1 2 0.1 0.2 0.5 1 2
Natural Period [sec.] Natural Period [sec.]
(g) Spatial Variation of Dislocation (h) Temporal Variation of Dislocation
Fig. 10 Variation of Acceleration Response

16-9
3. The effects of ground motion characteristics applied to the system when it behaves
on nonlinear response elastically.
(1) Evaluation of nonlinear response characteristics Fyp (ju) :Strength required to maintain the
of structures displacement ductility smaller or equal to
Extremely strong ground motion generated by the pre-determined allowable ductility ju .
the Kobe Earthquake, intra-plate inland earthquake, R/u can be calculated for arbitrary natural period
caused serious damage to various structures.
and therefore it makes spectrum, namely Strength
Through the investigation of the damage it was Reduction Factor spectrum ( Rju spectrum).
realized that seismic design force as large as the The concept of R/u is shown in Fig. 11 and
ground motion from the Kobe earthquake should be
according to energy constant rule it is approximated
introduced to seismic design. However there are still
as:
many unresolved points in the relation between
ground motion characteristics and inelastic response Rju = <j2ij.-\ (5)
of structures. In order to develop rational seismic
Eq. (5) is adopted in the Design Specifications for
design force and input ground motion, the effects of
Highway Bridges in Japan.
ground motion characteristics on inelastic response
of structures should be made clear. -i i

In this paper response of an inelastic single- Elastic Response

degree-of-freedom(SDOF) system is calculated for
ground motions from intra-plate and inter-plate Inelastic Response
earthquakes and the differences in inelastic response
due to ground motion characteristics are discussed
using nonlinear response spectra: Strength Reduction
Factor Spectra.
(2) Strength Redaction Factor Spectra
In order to evaluate the differences in the <5y Sye Smax
inelastic behavior of structures due to ground motion Fig. 11 Concept of RJU
characteristics, strength reduction factor Rp. is
calculated. Rfi indicates the possible reduction of
the strength of a structure when a certain (3) Effects of ground motions from intra-plate and
displacement ductility is allowed, and it is defined as: inter-plate earthquakes on inelastic response of
structures
(4)
A bi-linear system with damping ratio h=5%,
whose strength is assumed to be constant after
where, yielding, is adopted for an inelastic single-degree-of-
fj. Predetermined allowable displacement ductility,
freedom(SDOF) system, and its natural period is
defined as ju = 6max/6y
evaluated from the initial stiffness.
^max -Peak response displacement of the inelastic Fig. 12 shows Rfi spectra for Kaihoku Bridge
system exposed to an imput motion. record and JMA Kobe record. Kaihoku Bridge
Sy: Yielding displacement of the inelastic system. record is ground motion from the Miyagi-ken Oki
Earthquake, inter-plate earthquake, and JMA Kobe
Fye :Strength required to prevent the SDOF system record is ground motion from the Kobe Eartquake,
intra-plate earthquake. Both records were obtained
from yielding. It corresponds to the force on the stiff ground.

16-10
 Comparison of Rju spectra reveals that RJJ. 2) Amplitude of ground motion
of ground motion from the Kobe Earhquake is small To study the effects of ground motion
and independent of the allowable ductility p, in characteristics on Rfj. spectra, the amplitudes of
relatively short period range and becomes large in various frequency components of JMA Kobe record
the long period range especially for large fi . RJJ. and Kaihoku Bridge record are adjusted so that their
spectra of ground motion from the Miyagi-ken Oki 5% damped accelaration response spectra are close
Earthquake shows large dependence on allowable to the response spectra shown in Fig. 15. Fig. 16
ductility fj. in the wide natural period range shows the process of amplitude adjustment, in which
including long natural period range. phase characteristics are not changed. The spectra
(4) Ground motion characteristics having effects on shown in Fig. 15 are changed so that long period
Rju spectra components are increased in order of line number.
1) Ground motion time history
One of the uniqueness of the ground motion
characteristics of the Kobe Earthquake is that the a 2000
motion contains a few relatively long-period spike "1000
I 500
shaped waves and it may be attributed to the
I= NO:I
differences mentioned in (3), because similar 200 - No.2
No.3
tendency is observed in the Rfj. spectra of a
sinusoidal wave. Rfj. spectra of a single cycle of .1 0.2 0.5 I 2 5
Natural Period fsecl
sinusoidal wave with a period of 1.0 sec are shown in
Fig. 15 Target acceleration response
Fig. 13. It is independent of allowable ductility /j. in
the short period range, and it becomes large in the
long period range, converging Rfj. = fj..

20
18
16
14
12
=L 10
* 8
Natural Period [SBC] Natural Period (SBQ
6
4
Fig. 17 Variation of Rfj. spectra due to long
2 period component of ground motion
0.1 0.3 1 3
o
1 0.3 1 3
Natural Period [sec] Natural Period [sec]
Kaihoku Bridge Record (TR)
Rfj. spectra of these adjusted ground motions are
JMA Kobe Record (NS)
shown in Fig. 17. It can be seen that Rfj. decleases
Fig. 12 Difference in Rfj. spectra in the component shorter than 0.7 sec in the order of
line number. It indicates that when the long period
component of ground motion is increased, Rfj.
spectra are decreased in the short period range.

(5) Conclusion
Following knowledge is obtained from this study.
1) The effects of ground motion characteristics on
0.1 0.3 1 10
Natural Period [sec] inelastic response of structures are evaluated by
Rfj. spectra.
Fig. 13 A single cycle of wave with a period of 1.0 sec
2) Relatively long-period spike shaped waves in

16-11
 wave form have effects on R/J. spectrum Acknowledgments
characteristics. The authors would like to express their gratitude
3) Long period component of ground motion has to the personnel engaged in the ground motion data
effects on R{i spectra in short period range. observation for their efforts.
Authors would also like to express their cordial
Although further research is required, the above- appreciation to Dr. Shojiro Kataoka, Public Works
mentioned points should be taken into consideration Research Institute, for his informative discussions.
as the ground motion characteristics which have
effects on inelastic behavior of structures, when the Referencese
seismic design force and input ground motion for 1) Aki, K and P. G Richards, "Quantitative
seismic design are discussed. Seismology", theory and Methods, 1980, W.H.
Freeman and Company
2) Irikura, K., 'Prediction of strong acceleration
motions using empirical Green's function", 1986,
Proc. of 7th Japan Earthquake Engineering
Symposium.
Target spectrum SA (f) * 3) Kamae, K. et al., "A Technique for simulating
Original strong motion o Strong Ground Motion Using Hybrid Green's
Function, 1998, Bulletin of Seismological Society
of America
response
4) Kikuchi, M. , "Source Model For the Kobe
Earthquake", 1995, Chishitsu (Geology) News
Vol.486
5) Sato, R., "Japanese Source Parameters
Handbook", 1989, Kajima Institute Publishing
Co.,Ltd.
6) Sekiguchi, H., "Minute locating of faulting
beneath Kobe and the waveform inversion of the
NO source process during the Kobe, Japan,
earthquake using strong ground motion records",
ft - 1996. J.Phys. Earth, 44
7) Tamura, K et al, 1997, "Ground Motion
Characteristics of the Kobe Earthquake and
Seismic Design Force for Highway Bridges",
1997.
8) Proc. of Second National Conference on Bridges
and Highways

*Expressed by frequency t not by period

Fig. 16 Process of amplitude adjustment

16-12
 RELATIVE FLEXIBILITY OF A BUILDING FOUNDATION
by

M.D. Trifunac and M.I. Todorovska

Univ. of Southern California, Civil Eng. Dept., Los Angeles, CA 90089-2531

ABSTRACT

This paper analyzes deformations of the foundation in a seven-storey hotel building in

Van Nuys, California, from measurements of ambient noise and strong motion
recordings. This building has been instrumented by strong motion accelerographs, and
has recorded several earthquakes, including the 1971 San Fernando, 1987 Whittier-
Narrows, 1992 Landers, 1992 Big Bear, and 1994 Northridge earthquake and its
aftershocks. It suffered minor structural damage in 1971 earthquake and extensive
damage in 1994. Two detailed ambient vibration tests were performed following the
Northridge earthquake, one before and the other one after the 20 March aftershock.
These included measurements at a grid of points on the ground floor and in the parking
lot surrounding the building, summarized and analyzed in this paper. The analysis shows
that the foundation system, consisting of grade beams on friction piles, does not act as a
"rigid body" but deforms during the passage of microtremor and therefore earthquake
waves. For this geometrically and by design essentially symmetric building, the center of
stiffness of the foundation system appears to have large eccentricity (this is seen both
from the microtremor measurements and from the earthquake recordings). This
eccentricity may have contributed to strong coupling of transverse and torsional
responses, and to larger than expected torsional response, contributing to damage during
the 1994 Northridge, earthquake.

INTRODUCTION

Earthquake resistant design of structures must be based on analyses of realistic models of

the structure, foundation and soil system, considering wave propagation and all the
aspects of nonlinear response. Such analyses require solution of a complicated and
difficult to solve system of governing equations and boundary conditions. Hence, it has
been necessary to make various simplifications. In doing so, it is important to evaluate
the accuracy of the approximations and to define the range of the model parameters for
which the approximations are valid. This is best accomplished by careful experimental
verification using full scale tests of actual structures.

A common assumption in many models which consider soil-structure interaction effects

is that the foundation is rigid. This reduces the number of degrees-of-freedom of the
model, and gives good approximations for wavelengths long relative to the foundation
dimensions (Lee, 1979). For short wavelengths, this assumption can result in non
conservative estimates of the relative deformations in the structure and, in general, is
expected to result in excessive estimates of scattering of the incident wave energy and in

17-1
excessive radiation damping (Todorovska and Trifunac 1990a; 1991; 1993). The extent
to which this assumption is valid depends on the stiffness of the foundation system
relative to that of the soil, and also on the overall rigidity of the structure. For a nine-
storey reinforced concrete building, extensively tested during the 1970's, the foundation
could be represented by a "rigid" slab for NS vibrations (because of stiffening effects of
the end shear walls) but not for EW vibrations (Foutch et al. 1975; Luco et al. 1975;
1977; 1986; Moslem and Trifunac, 1986; Wong et al. 1987). The other extreme is to
neglect the stiffness of the foundation system and to assume that the wave energy is
transmitted from soil into the building according to the principles of wave propagation
(Todorovska and Trifunac 1989; 1990a,b,c; Todorovska et al. 1988). This approximate
approach underestimates the incident wave energy scattered by the foundation and
overestimates the energy transmitted into the building. The reality is somewhere between
these two approximations, and can be studied in detail only by means of numerical
methods.

In this paper, an instrumented seven-storey hotel building in Van Nuys, California, is

studied. Records of several earthquakes were available for the study, including the 1971
San Fernando (ML=6.6, R=22 km), 1987 Whittier-Narrows (ML=5.9, R=41 km), 1992
Landers (ML=7.5, R=186 km), 1992 Big Bear (ML=6.5, R=149 km), and 1994 Northridge
(ML=6.4, R=1.5 km) earthquake and two of its aftershocks (20 March: ML=5.2, R=1.2
km; and 6 December, 1994: ML=4.3, R=ll km). The building is supported by a friction
pile foundation. The Northridge earthquake caused severe damage, and the building was
declared unsafe. The damage was most severe at the fifth floor, where many columns
were damaged, just below the spandrel beam. The specific aspects of the response,
which caused this type of failure, have not been deciphered so far. One plausible group
of causes can be sought in the large relative deformations of the foundation system (pile
caps connected by grade beams), but the limited number of accelerographs, which
recorded the main event, is not sufficient to verify this hypotheses (Trifunac 1997,
Trifunac and Todorovska, 1997).

In this paper, a sumrnary of ambient noise measurements in the parking lot and on the
ground floor of this building is presented. The objective is to describe the deformations
of the foundation system during the passage of ambient noise waves (mostly Rayleigh
waves caused by surface traffic), and to speculate on how the foundation may have
moved during the Northridge earthquake. Further details about this experiment can be
found in Trifunac et al. (1998).

DESCRIPTION OF THE BUILDING

The building analyzed in this paper is a seven-storey reinforced concrete structure, in the
city of Van Nuys (Los Angeles metropolitan area), near the intersection of Roscoe Ave.
and 1-405 (Fig. 1). It was designed in 1965 (Blume et al., 1973) and served as a hotel
until 1994. Its plan dimensions are about 62 by 160 feet (Fig. 2a). The typical framing
consists of columns spaced at 20 foot centers in the transverse direction and 19 foot
centers in the longitudinal direction. Spandrel beams surround the perimeter of the
structure. Lateral forces in each direction are resisted by the interior column-slab frames
and exterior column spandrel beam frames. The added stiffness afforded the exterior

17-2
 3425'N-
San Fernando Valley
0 1 2 mi
S. Fernando, 1971:
Northridge, 1994: 5 10m
main event Dislocation amplitude
Newhall

210
*/

Landers, 1992 186km

Canoga Park

Northridge, 1994:
20 March aftershock
-j-3409'N
118 39'W W

Fig. 1 Geometrical relationship of the building site to the earthquakes causing strong motion.
Whirtier, Landers and Big Bear earthquake are outside the limits of this figure, at epicentral
distances 41, 186 and 199 km respectively.

frames by the spandrel beams, creates exterior frames that are roughly twice as stiff as
interior frames. With the exception of some light framing members supporting the
stairway and elevator openings, the structure is essentially symmetric. The contribution to
the overall stiffness and mass from the nonstructural brick filler walls and some of the
exterior cement plaster could cause some asymmetry for lateral motion in the longitudinal
direction, which is expected to be minor.
The first floor is a slab on grade over about 2 feet of compacted fill. Except for two small
areas at the ground floor, covered by one-storey canopies, the plan configurations of the
floors, including the roof, are the same. The floor system is a reinforced concrete flat
slab, 10 niches thick at the second floor, 8.5 inches thick at the third to seventh floors and
8 inches thick at the roof.

17-3
a)
D r7^" ^A1-.... A3--...-E1- ...... LI------!J-------I3-------1p. ----- JJ

c) N
: -i 1 I1 [1 t| ] |J1 13 T7TT41
il 0
T
c
u D C B A
B H1 t1 [] I1 13 1D 1 ] ^^: roof 8" slab
^
ii1 L
ja
\7 7th 8.5" slab a
ndl !' A
* 1------| ] [!]------{^"----|r"--'iJ--""1I-----T]
A
6th 8.5" slab
11 8 bays @ 18' - 9" = 150' - 0"
5th 8.5" slab
b) oc
1234 567 cc
4th 8.5" slab X

3rd 8.5" slab

2nd 10" slab

^
5
0
r
1st 4" slab
C\J
CD

20' -10" :it< 20' -10" 20'-10"

62>. 8"

8 bays 18'-9" = 150'.-0"-

Fig. 2 (a) Typical floor plan, (b) Foundation plan, (c) Typical transverse section.
The site lies on recent alluvium. A typical boring log shows the underlying soil to be
primarily fine sandy silts and silly fine sands. The average shear-wave velocity in the top
30 m is -300 m/s. The foundation system (Fig. 2b) consists of 38 inch deep pile caps,
supported by groups of two to four poured-in-place 24 inch diameter reinforced concrete
friction piles. These are centered under the main building columns. All pile caps are
connected by a grid of the beams. Each pile is roughly 40 feet long and has design
capacity of over 100 kips vertical load and up to 20 kips lateral load. The structure is
constructed of regular weight reinforced concrete (Blume et al., 1973).

The Feb. 9, 1971, San Fernando earthquake caused minor structural damage. Epoxy was
used to repair the spalled concrete of the second floor beam column joints on the north
side and east end of the building. The nonstructural damage, however, was extensive and
about 80% of all repair cost was used to fix the drywall partitions, bathroom tiles and
plumbing fixtures. The damage was most severe on the second and third floors and
minimal at the sixth and seventh floors.

The building was severely damaged by the 17 January, 1994, Northridge earthquake, and
was not in use on February 4, 1994, when we conducted the first ambient vibrations
experiment. The structural damage was extensive in the exterior north (D) and south (A)
frames, designed to take most of the lateral load in the longitudinal direction. Severe
shear cracks occurred at the middle columns of frame A, near the contact with the
spandrel beam of the fifth floor. Those cracks significantly decreased the axial, moment
and shear capacity of the columns. The shear cracks which appeared in the north (D)
frame on third and fourth floors, and the damage of columns D2, D3 and D4 on the first
floor caused minor to moderate changes in the capacity of these structural elements. No
major damage of the interior longitudinal (B and C) frames was noticed. There was no
visible damage in the slabs and around foundations. The nonstructural damage was
significant. Almost every guestroom suffered considerable damage. Severe cracks were
noticed in the masonry brick walls, and in the exterior cement plaster.

EARTHQUAKE RECORDINGS

The first known recorded strong motion in the building is of the Feb. 9, 1971, San
Fernando earthquake (Fig. 1). The sensors, three self-contained tri-axial AR-240
accelerographs, were located in the SE corner (near Chan. 14; Fig. 3), middle of the
fourth floor, and on the roof (near present Chan. 2; Fig. 3). During this earthquake, the
first strong motion waves started to arrive from N22E, having originated at depth ~9 to
13 km below epicenter (Trifunac, 1974). With rupture propagating up towards south at
about 2 km/s, the last direct waves were arriving from N 62E, 9-10 s later. The 1987
Whittier-Narrows, 1992 Landers and 1992 Big Bear earthquakes occurred at epicentral
distances of 41, 186 and 149 km respectively and caused strong motion arrivals from E
27S, East, and E 1.5S. During the 1994 Northridge earthquake, the first motions started
to arrive from the West, with the last arrivals coming from N 42W, about 7 to 10s later
(Fig. 1). These latter earthquakes were recorded by a CR-1 system; the sensor locations
are shown in Fig. 3.

17-5
 13.2m

Fig. 3 Location and orientation of 13 sensitivity vectors of CR-1 recording system. Chan. 14, 15
and 16 belong to SMA-1 accelerograph.

Table 1 summarizes selected parameters of the above earthquakes and accelerograms.

The San Fernando accelerogram was digitized manually, at a sampling rate of minimum
50 points per second. The accelerograms of the Whittier Narrows, Landers, Big Bear and
Northndge earthquakes were processed by the California Division of Mines and Geology.
Figure 4 shows the computed displacements for Chan. 1 (solid line) and 13 (dashed line)
(see Fig. 3) during the Landers and Northndge earthquakes. The other earthquakes which
triggered the instruments (Table 2) generally resulted in smaller displacement amplitudes
(Trifunac et al., 1998) and are not presented here.

Table 1 Selected earthquake and accelerogram parameters describing the data used in this work

Azimuth of Peak Peak

Magnitude Epicentral Arriving horizontal vertical
Earthquake ML Date distance strong morion acceleration acceleratio
km waves (g) n
IB)
San Fernando* 6.6 9 Feb., 71 22 22-62 0.25 0.17
Whittier 5.9 1 Oct., 87 41 117 0.16 -
Landers Ms = 7.5 28 Jun., 92 186 90 0.041 0.007
Big Bear 6.5 28 Jun., 92 149 91.5 0.01 0.007
Northndge 6.4 17 Jan., 94 1.5 240-350 0.44 0.27
Northndge aft. 5.2 20 Mar., 94 1.2 320 0.27 0.10
Northndge aft. 4.3 6 Dec., 94 10.9 40 0.06 0.03

* Not considered hi this analysis; provided for general background only.

17-6
Table 2 Other earthquakes which have triggered the instruments in the 7-storey hotel*
Aftershock of Whittier-Narrows 4 Oct. 1987 M=5.3 10:59:38 GMT
Pasadena Earthquake 3 Dec. 1988 M=4.9 11:38:26GMT
Malibu Earthquake 19 Jan. 1989 M=5.0 06:53:28 GMT
Upland Earthquake 28 Feb. 1990 M=5.2 23:43:37 GMT
Sierra Madre 28 June 1991 M=5.8 14:43:54 GMT

* V. Graizer, personal communication (1997).

Analyses of the displacement time histories for Chan. 1, 2, 3 and 13 (see Fig. 3) show
that during the larger peaks of the relative response, the torsion within the building
contributes 20 to 40% of the peak relative response, at the locations of Chan. 2, for
example (Whirtier-Narrows -23%; Landers -33%, Fig. 5 top; Big Bear -42%; and
Northridge -22%, Fig. 5 bottom). Comparison of displacement time histories of Chan. 1
and 13 shows that at the site of Chan. 1, the peak displacements were up to 10 to 20%
larger than at the site of Chan 13 (-20% during Landers earthquake, 30-35 s after trigger;
-10% during Northridge earthquake, 5 to 10 s after trigger, see Fig. 4). It is not likely that
these differences were caused by the nature of incident waves. The motions arriving from
the Landers earthquake had mainly long period surface waves, propagating from east to

Landers, 28 June, 1992 -

I 10 Northridge, 17 Jan., 1994

(D
0
(D
O
JO
Q_
0.1-23 Hz

0 10 20 30 40
Time - s

Fig. 4 Displacement time histories for Chan. 1 (solid lines) and 13 (dashed lines) located at
ground floor, at the west and east ends of the building respectively for strong motion shaking
during Landers, 1992, and Northridge, 1994, earthquakes.

17-7
Displacement - cm Displacement - cm
west (Fig. 1), while the waves generated by Northridge earthquake were mainly direct
near-field arrivals, propagating predominantly from west to east, with high phase
velocities associated with mostly vertical incidence, during the first 3 to 4 s of strongest
motion (between 4 and 8 s in Fig. 4). The amplitudes of the observed differences depend
somewhat on the choice of the band-pass filters (shown for each event in Fig. 4), but
occur systematically only during large motions. Furthermore, the differences should be
emphasized by EW wave arrivals, because the building is elongated in EW direction. If
the observed differences result from separation and relative displacement between the
soft soil surrounding the foundation and the piles, it is expected that these differences
would be large during Landers and Northridge earthquakes, which both caused large
motions at this site, and both arrived predominantly along EW direction. During two
aftershocks of Northridge earthquake the waves arrived from North-West and North-East
(Fig. 1), and so were less efficient in exciting the torsional response, but their motions
were also small.

AMBIENT VIBRATION EXPERIMENTS

General Overview and Objectives

Two ambient vibration experiments were conducted in the building, one on Feb. 4-5
(about two and a half weeks after the Northridge main event) and the other one on April
19-20, 1994 (about three months after the main event and one month after one of the
largest aftershocks, of March 20, M 5.2). Between the two experiments, the building
was temporarily restrained, as it was severely damaged by the main event.

The objective of the first experiment was to measure the dynamic characteristics of the
damaged building and to see whether the changes in stiffness due to the extensive
structural damage could be identified by small amplitude tests. The second experiment
was much more detailed. Besides detecting changes in stiffness due to new damage from
the 20 March aftershock, it also had as an objective to measure the motion of the ground
around the building. This was planed to be done by a series of measurements at a dense
grid of points in the parking lot of the building. Similar measurements were made during
a three-dimensional forced vibration survey of a 9-storey reinforced concrete building in
Pasadena (Foutch et al., 1975; Luco et al, 1975). The analysis of the amplitudes and
phases of the recorded motion confirmed that soil deforms as predicted by theoretical
models, and provided an experimental verification of various simplifying assumptions,
which usually accompany soil-structure interaction models (e.g., the rigid foundation
assumption, and the effects of embedment; Luco et al., 1986).

The aim of the parking lot measurements in the building was to detect ground
deformations associated with at least the fundamental transverse and longitudinal modes
of vibration. This would have been useful for characterization of soil-structure interaction
involving a complex pile foundation. However, no peaks associated with rocking or
translation at the apparent frequencies of the building-soil system could be found in the
Fourier amplitude spectrum, above the noise level. Nevertheless, the results came out to
be even more useful, revealing evidence of flexibility of the foundation and of wave
propagation through the first floor slab and the surrounding soil.

17-9
For the analysis of this paper, the parking lot measurements of the second experiment are
of interest, and are presented and analyzed. From the measurements in the building, only
the results on the apparent modal frequencies for both experiments are summarized.

Instrumentation and Methods of Analysis-Second Experiment

Four Ranger SS-1 seismometers and two Earth Sciences Rangers were used (Ivanovic
and Trifunac, 1995; Trifunac, 1972). The response was measured along frame C (Fig.
2a), at all columns and at each floor, for all three components of motion. Three of the
SS-1 Ranger seismometers were used to record motion at various locations (the location
and orientation were changed as required). The motion of the ground floor was measured
at each column and in all three directions (N, E and vertical).

Three reference points were used for calculations of the transfer functions (marked by
"R" in Fig. 6). Two of the Earth Sciences Ranger seismometers were placed on the
ground floor, at reference locations A5 and D5. Their orientation was always up. The
reference instrument for horizontal motions was at location B2 on the ground floor. It
was oriented either along the longitudinal (east) or along the transverse (north) direction,
depending on the measurement.
-150'

N Run no. 102 i

cj ]
10'

O , , "1 o
2 3 4 5 6 7 8 9
D R0
#0

30' - 30' C
~~<a
R2
B #0

A #1 R1
9.5' CD I , 72<ct
I <C1 d .-. 1 .3'
i-r i d: 30'
32' 30'

_-__.________c[ 3- - -----i""
30'
__._4._.

Fig. 6 Position of instruments during the recording of ambient noise in the parking lot of the
building.

17-10
The measurements in parking lot were carried out at 46 locations within 15 to 20 m from
the structure, and in three directions (north, east and up; Fig. 6). This was done during
daytime, when high direct sun could have contributed to the noise in the soft asphalt
surface. To abate this, the instruments were covered with towels. The experiment was
carried out continuously from 12 noon of April 19 (Tuesday), until 9 p.m., April 20
(Wednesday) 1994. Those were quiet sunny days (temperature was in the range of 12 to
25C). The building was not in use, and except for electricity, other facilities were not
available (no elevators, air-conditioning, or running water...). Each of the measurements
lasted about 3 minutes, and the sampling frequency was 400 points per second. The PC
computer used to record was located on the ground floor. The instruments were placed
either directly onto the concrete slab, ceramic tiles, or onto the asphalt, for the outside
measurements. Two calibration tests were performed for both horizontal and vertical
transducer orientation, one at the beginning and the other one at the end of the
experiment.

To describe the overall nature of microtremors in the area surrounding the building,
cross-correlation analyses were performed as follows. Measurements were performed by
"new" Ranger seismometers (#3, 4 and 5; e.g., during run No. 102, see Fig. 6), while the
reference instruments (#0, 1 and 2) were located inside the building at locations B2 (#2),
D5 (#0) and A5 (#1, as shown in Fig. 6). Two "old" Ranger seismometers recorded
vertical motion during all measurements, but only transducer #1 at A5 was used as
reference for the analysis of vertical motions in the parking lot.

Results on Modal Frequencies

The results of the measurements in the building are summarized here only in terms of
frequencies and mode shapes for horizontal motion. It was found that in the transverse
(N-S) direction the soil-structure system vibrates with frequencies/^ 1.4, 1.6, 4.2 and 4.9
Hz. In the longitudinal direction, the apparent frequencies were at/= 1.1, 3.7, 5.7 and
8.5 Hz. Detailed description of the mode shapes and of other aspects of the response is

Table 3 Mode shapes and apparent frequencies for EW and NS vibrations, measured during
Experiments 1 and 2.

f-Hz
Mode shapes
NS Expl. I Expl. II Af - %
Feb. 94 Apr. 94

1.4 1.4

1.6 1.6

3.9 4.2 10

4.9 4.9

17-11
outside the scope of this paper. Table 3 summarize the results on the apparent modal
frequencies for both experiments, for the longitudinal (EW) and transverse (NS)
directions, only to suggest the overall characteristics of the transfer functions for
horizontal motions and to provide a general background for the analysis of the foundation
response. It is seen that three out of the four identified frequencies in the longitudinal
direction were larger during the second experiment, while one (f= 5.7 Hz), reminded the
same. The increase in frequency most probably resulted from the wooden braces
restraining the building, placed at the longitudinal frames between the two experiments.
The frequency of the first longitudinal mode increased by 10%, and of the second and
fourth longitudinal modes by 6 and 5%. Apparently the restrainers did not affect the third
mode. It is also seen that the frequency of the first transverse mode and of the first
torsional mode are the same (apparently, the braces located along the longitudinal frames,
did not increase stiffness for those two modes), but of the frequency of the third
transverse mode had frequency larger by 10% for the second experiment.

Results of Motion of the Ground Floor and of the Surrounding Soil

General Characteristics

Figure 7 shows average Fourier amplitude spectra of NS, EW and vertical components of
motions in the parking lot (averaging was done to emphasize the predominant wave
motion and to reduce the local noise). The average spectra were obtained from three runs
at a group of three locations, located north, east, south and west of the building (i.e. total
of 12 locations, highlighted in Fig. 6 by cross-hatched schematic representation of the
recording instruments). It is seen that there are many large amplitude peaks in the
spectra. In most cases, these do not coincide with identified apparent modal frequencies
of the building (shown by solid, open and shaded bars in Fig. 7), and -were created by
strong periodic sources in San Fernando Valley (industrial sites with large moving
machinery). The overall large amplitudes for frequencies centered near 4 Hz were caused
by the NS traffic on 1-405 (-150 m west of the building) and by EW traffic on Roscoe
Blvd. (-50 m north of the building).

Cross-correlation Functions

The cross-correlation function, /?/, ref. (t), was computed for each location, defined by

f
/2
_r /2 /<*)/ (f+tfdt (1)

where ft) is the motion at the ith location and^eKO is the simultaneous record at the
reference point (B2 for EW and NS motions, and A5 for vertical motions, see Fig. 6) and
r~3 min. Then, the spatial distribution of the peak amplitude, ^i,ref('c)max> and
corresponding time lag, t, were plotted and analyzed. Figures 8-10 show contour plots
of the peak amplitudes of J?,-, ref. CO (solid line) and of the relative delay T (dotted lines).
The amplitudes are on an arbitrary scale, but consistent for the NS, EW and vertical
directions, and T is in seconds. In Fig. 8, 30 feet (~ 10 m) west of the building, T = 0.03 s.
This corresponds to apparent horizontal phase velocity of about 300 m/s, consistent with

17-12
 20
EW 01 4Q EW
TOR
15

10

0 6 8 10
20
o NS
CD NS Q1
~
co~ 15 -TOR
CD

10

CD
13
o
LL
"O
CD
0 6 8 9 10
CD
03
CD
< 20 h NS

15

10

0 34567 8 10
Frequency - Hz

Fig. 7 Average (over time and space: four groups of three points, shaded in Fig. 6) of Fourier
amplitude spectra of microtremor noise recorded in the parking of the building. Frequencies of
apparent NS, EW and torsional modes of building vibration are shown.

17-13
 amplitude
phase

Fig. 8 Contours of (fy ref. (T))maX for NS motion (arbitrary normalized amplitudes, shown by heavy lines), and (T) (in seconds) of (Rti ref.
tmax relative to the reference station at B2.
 ^ 0.20 '' ^
amplitude
phase

Fig. 9 Contours of (Rit ref. (T))max for vertical motion (arbitrary normalized amplitudes, shown by heavy lines), and (i) (in seconds) of (/?/, ref.
(t))max relative to the reference station at A5.
 amplitude
phase

-J
I

Fig. 10 Contours of (/fy ref. (i)) * for EW motion (arbitrary normalized amplitudes, shown by heavy lines), and (T) (in seconds) of (Rit ref. (t))max
relative to the reference station at B2.
the interpretation that microtremors are high frequency Rayleigh waves propagating
through shallow soil layers. The overall pattern of the time lag, T, implies wave arrival
from the west, and scattering and diffraction around the building foundation. The
corresponding contours for vertical motion (Fig. 9) imply wave arrival from west and
south-west, with apparent phase velocities between 250 and 300 m/s. The corresponding
contours for the EW motions are shown in Fig. 10.

The results in Figs 8-10 have been evaluated using unfiltered recordings. As it can be
seen from Fig. 7, this motion has most of its energy between 3 and 6 Hz, while the
recorded signal is small between 1 and 1.5 Hz. This is inconvenient, because the motions
of the soil driven by soil-structure interaction are expected to be seen at frequencies near
the horizontal and vertical apparent frequencies of the building (~1.0 Hz for EW and ~
1.4 Hz for NS and vertical motions). To analyze the motions near these frequencies, the
signals were first band-pass filtered, using a "cosine bell" function centered at 1.0 Hz for
EW motions and at 1.4 Hz for NS and vertical motions, and 0.6 Hz wide. The results
were consistent with the overall propagation of energy from west to east, but difficult to
interpret (Trifunac et al., 1998). The contours of the amplitudes of/?,-, ref. (T)max implied
strong warping of the building foundation and of the parking lot, different from slowly
decaying motions with same relative phase, away from the building, as would be
expected from soil-structure interaction effects based on rigid foundation modeling
(Foutch et al., 1975, Luco et al., 1975).

To eliminate the consequences of amplitude variations of the ambient noise with the time
of the day, the cross-correlation functions in Figs 8-10 were normalized by R\,\ (0) or R2,2
(0) for the unfiltered record and for band-pass filtered data, for all measurement runs
(runs 79-143). The variations in amplitude and phase caused by fluctuations in the
direction of approach of the ambient noise cannot easily be accounted for by simple
normalization, and the associated effects were not corrected for in the presented results.

This experiment was carried out about three months after the earthquake and many of its
aftershocks. Perhaps too soon after the earthquake for the "gaps" and "clearances"
between the vertical walls of the building and soil, and piles and the surrounding soil to
have been "recemented". It may be that what we see in the above measurements is the
response of the "disturbed" foundation system, with "minute cracks" and "gaps" in the
foundation soil, causing the wave motion in the parking lot to be so irregular. Of course,
this is further complicated by apparent arrival of wave energy from different directions,
though mainly from moving sources on a major freeway just 150 m west of the site (NS
vehicular traffic on 1-405).

DISCUSSION AND CONCLUSIONS

One of the more interesting results of this analysis is seen in Fig. 8, displaying
normalized amplitudes of the cross-correlation function of NS velocities for the complete
(unfiltered) recorded motions. It shows that during passage of microtremor waves,
mainly from west to east, the foundation essentially rotates about a point close to the
south-eastern corner of the building (near A9). The EW components of this motion,
shown in Fig. 10, are consistent with this interpretation if one allows for some in-plane

17-17
deformation of the foundation system, in the north and west ends of the building. If
present during strong motion, this would imply very large eccentricities of torsional
stiffness of the overall foundation, and consequently strong coupling of the NS and
torsional components of response.

We speculate that during very strong motion, the soil is pushed sideways by large relative
response of the foundation and piles, in this case more along the west end of the building.
The width of this separation probably closes partially during "dynamic compaction"
effected by many small aftershocks. Therefore, the pattern of recorded ambient noise
amplitudes, shown in Fig. 8, and our interpretation could also depend on the status of this
separation during the time of our experiment.

The above described mechanism acts as a powerful passive energy absorption system, but
analyses of its nonlinear, time dependent behavior and complex soil-structure interaction
analyses would be very difficult. Evaluation of its effects on the dynamics of the system
would require analyses in which the geometric characteristics experience large changes
during the duration of the excitation. Analyses of such problems are possible, but it is
helpful first to learn more about the expected nature of the changes with time from full
scale observations during actual earthquakes.

The above observations could result from nonuniform soil properties below the
foundation or from partial shear failure of several piles (probably as early as in 1971
during San Fernando earthquake), resulting in "softer" soil-pile system below the western
end of the building. Such variations in stiffness must be included in the response
analyses which should explicitly address the strong coupling between translation and
torsion. Several repeated full scale tests of the as-built structure would have detected the
range of actual variations of these centers of stiffness.

The measuring grid shown in Fig. 6 was not sufficiently dense to determine whether the
side walls of the building and the soil moved as a continuum or independently. Most
horizontal displacement contours are consistent with the assumption that these two are in
contact. The contours of vertical motions, however, suggest that some separation may be
present, for example, near the north-western corner of the building (Fig. 9).

All the contour plots of horizontal and vertical amplitudes of deformation of the ground
floor show that the foundation of this building did not act as a rigid body, but it deformed
with the passage of incident waves. The grade beams allowed differential vertical (Fig.
9) and horizontal motions (Fig. 8 and 10), which followed the deformation of the "body"
of soil with piles. In this example this "body" is stiffer than the surrounding soil by a
factor perhaps as large as two, because the velocity of NS displacements is roughly 40%
higher over the ground floor than outside the building (see the dashed lines representing
the relative phases of motion in Fig. 8). Consequently, in addition to the inertial forces,
the differential motions of the first story columns would contribute additional moments
and shears. The actual amplitudes of these additional effects can be calculated by
numerical modeling, but conservative estimates of their upper bounds can be obtained by
assuming that the soil-pile-foundation system has same stiffness as the surrounding soil,

17-18
and that there is no soil-structure interaction (Trifunac, 1997; Trifunac and Todorovska,
1997).

Albeit speculative and qualitative, the above considerations show that the state of the art
of strong motion instrumentation of buildings is not adequate to address most of the
aspects of the problems we considered. Additional recording channels on the ground
floor would have provided invaluable information. Analyses of the damage in this
building is outside the scope of this paper, but it could be shown that additional
instruments were called for at the upper floors as well. Therefore, with limited resources,
we should explore what is better: to instrument in more detail selected geometrically
simple structures, or to continue with the present programs which instrument many
buildings, but with limited instrumentation in each building. At present, while this
decision is made, the emphasis should be placed on processing and dissemination of all
recorded accelerations in structures, so that small (linear) and large nonlinear motions can
be analyzed and compared. For identification of the soil-structure system, all recorded
motions are valuable, even the very small ones, particularly when those contribute to the
database of simple and symmetric buildings, for which most of our analysis tools should
be applicable.

During the last 20 years, too much emphasis has been placed on laboratory experiments,
while the full-scale tests of structures have been neglected. The best and truly
informative experiments are the full-scale tests in actual buildings. In the laboratory, we
mimic imperfectly our hypotheses and expectations. Only the full scale observations of
the actual nature can unveil our misconceptions and occasionally provide a basis for
better understanding and for creation of new theories and ideas.

REFERENCES

Blume, J.A. and Assoc. (1973). Holiday Inn, Chapter 29 in "San Fernando, California
Earthquake of February 9, 1971," Volume I, Part A, U.S. Dept. of Commerce, National
Oceanic and Atmospheric Administration, Washington, D.C.
Foutch, D.A., J.E. Luco, M.D. Trifunac and F.E. Udwadia (1975). Full scale three-dimensional
tests of structural deformation during forced excitation of a nine-story reinforced concrete
building, Proceedings, U.S. National Conference on Earthquake Engineering, Ann Arbor,
Michigan, pp. 206-215.
Ivanovic, S. and M.D. Trifunac (1995). Ambient vibration survey of foil scale structures using
personal computers (with examples in Kaprielian Hall), Dept. of Civil Eng. Report No. 95-05,
Univ. of Southern California, Los Angeles, California.
Lee, V.W. (1979). Investigation of three-dimensional soil-structure interaction, Dept. of Civil
Eng. Report No. 79-11, Univ. of Southern California, Los Angeles, California.
Luco, J.E., H.L. Wong and M.D. Trifonac (1975). An experimental study of ground
deformations caused by soil-structure interaction, Proceedings, U.S. National Conference on
Earthquake Engineering, Ann Arbor, Michigan, pp. 136-145.
Luco, J.E., H.L. Wong and M.D. Trifunac (1977). Contact stresses and ground motion generated
by soil-structure interaction, Earthquake Eng. and Structural Dynamics, 5, 67-79.

17-19
Luco, I.E., H.L. Wong and M.D. Trifunac (1986). Soil-structure interaction effects on forced
vibration tests, Dept. of Civil Eng. Report No. 86-05, Univ. of Southern California, Los
Angeles, California.
Moslem, K. and M.D. Trifunac (1986). Effects of soil structure interaction on the response of
buildings during strong earthquake ground motions, Dept. of Civil Eng. Report No. 86-04,
Univ. of Southern California, Los Angeles, California.
Todorovska, M.I. and M.D. Trifunac (1989). Antiplane earthquake waves in long structures, J. of
Eng. Mech., ASCE, 115, 2687-2708.
Todorovska, M.I. and M.D. Trifunac (1990a). A note on excitation of long structures by ground
waves, J. of Eng. Mech., ASCE, 116, 952-964.
Todorovska, M.I. and M.D. Trifunac (1990b). Propagation of earthquake waves in buildings with
soft first floor, J. of Eng. Mech., ASCE, 116, 892-900.
Todorovska, M.I. and M.D. Trifunac (1990c). Analytical model for the plane building-
foundation-soil interaction: incident P-, SV-, and Rayleigh waves, Dept. of Civil Eng. Report
No. 90-01, Univ. of Southern California, Los Angeles, California.
Todorovska, M.I. and M.D. Trifunac (1991). Radiation damping during two-dimensional in-
plane building-soil interaction, Dept. of Civil Eng. Report No. 91-01, Univ. of Southern
California, Los Angeles, California.
Todorovska, M.I. and M.D. Trifunac (1993). The effects of the wave passage on the response of
base isolated buildings on rigid embedded foundations, Dept. of Civil Eng., Report No. 93-
10, Univ. of Southern California, Los Angeles, California
Todorovska, M.I., V.W. Lee and M.D. Trifunac (1988). Investigation of earthquake response of
long buildings, Dept. of Civil Eng. Report No. 88-02, Univ. of Southern Calif., of Los
Angeles, California.
Trifunac, M.D. (1972). Comparison between ambient and forced vibration experiments,
Earthquake Eng. and Structural Dynamics, 1, 133-150.
Trifunac, M.D. (1974). A three-dimensional dislocation model for the San Fernando, California,
Earthquake of February 9, 1971, Bull. Seism. Soc. Amer., 64, 149-172.
Trifunac, M.D. (1997). Differential earthquake motion of building foundations, J of Structural
Eng., ASCE, 123, 414-422.
Trifunac, M.D. and M.I. Todorovska (1997). Response spectra and differential motion of
columns, Earthq. Eng. and Structural Dynamics, 26, 251-268.
Trifunac, M.D., S.S. Ivanovic and M.I. Todorovska (1998). Experimental evidence for
flexibility of a building foundation supported by concrete friction piles, Soil Dynam. and
Earthquake Eng., (submitted for publication).
Wong, H.L., M.D. Trifunac and I.E. Luco (1987). A comparison of soil structure interaction
calculations with results of full-scale forced vibration tests, Soil Dynamics and Earthquake
Eng., 7, 22-31.

17-20
 SoiLrSTRUCTURE INTERACTION EEFECT ON AN NPP REACTOR BUIUDING
- ACTIVITIES OF NUPEC ; ACHIEVEMENTS AND THE CURRENT STATUS -

Yoshio Kitada", Masamitsu Kinoshita2), Michio Iguchi3), Nobuo Fukuwa4)

ABSTRACT: This paper presents activities of Nuclear Power Engineering Corporation (NUPEC) in the field
of the study on soil structure interaction (SSI). NUPEC had performed three projects related to the SSI study.
These are "Model Tests on Dynamic Interaction Between Reactor Building and Soil", "Model Tests on Base-
mat Uplift of Reactor Building", and "Model Test on Embedment Effect of Reactor Building". Through the
tests, the adequacy of the analytical methods applied to the current aseismic design was confirmed.
Nevertheless NUPEC is recognizing that the SSI phenomena is still problematic and there are many
unresolved problems such as handling of uncertainty in detailed soil structure, the nonlinear behavior of soil,
and structure-structure interaction etc. The dynamic cross interaction (DCI) between adjacent structures is one
of the biggest problems of this kind. NUPEC then had planned the project to investigate the DCI effect in
1993. The project has started as "Model Tests on Dynamic Cross Interaction Effects of Adjacent Structures".
The project consists of field and laboratory tests and starts in 1994 and will be completed in early 2002. In
this paper we describe firstly a brief review of the results of the past three test projects on SSI performed by
NUPEC. Secondly, we describe an outline and a summary of the current status of the on-going project,
"Model Tests on Dynamic Cross Interaction Effects of Adjacent Structures".

1. INTRODUCTION between buildings during earthquakes. NUPEC had planned

the project to investigate the DCI effect in 1993 after the
Nuclear Power Engineering Corporation (NUPEC) has preceding SSI investigation project, "Model Tests on
conducted a series of research projects on soil structure Embedment Effect of Reactor Building". The project has
interaction (SSI) under the entrustment of Ministry of started as "Model Tests on Dynamic Cross Interaction Effects
International Trade and Industry (MITI) of Japan to ensure the of Adjacent Structures". The project consists of field and
adequacy of aseismic design methodologies used in a nuclear laboratory tests and starts in 1994 and will be completed in
power plant (NPP) reactor building design. Up to the present, early 2002.
three projects have been completed. These are "Model Tests In this paper, we describe firstly a brief review of the results
on Dynamic Interaction Between Reactor Building and of the past three test projects on SSI performed by NUPEC.
Soil"(1980-1986), "Model Tests on Base-mat Uplift of Secondly, we describe an outline and a summary of the
Reactor Building"(1982-1987), and "Model Test on current status of the on-going project, "Model Tests on
Embedment Effect of Reactor Building"(1986-1994). Dynamic Cross Interaction Effects of Adjacent Structures".
Through the tests, many fruitful test data including earthquake
observation records and detailed information on the SSI 2. BRIEF REVIEW OF THE RESULTS OF THE PAST
phenomena have been obtained. As a consequence of TESTS ON SSI
detailed examination of the test data, the adequacy of the
analytical methods applied to the current aseismic design was NUPEC conducted three research projects on SSI
confirmed. phenomena from 1980 to 1994 to confirm the adequacy of
Nevertheless, to date, NUPEC has continued efforts to aseismic design processes for NPP reactor buildings. These
upgrade the methodology, because the SSI phenomena is still projects were "Model Tests on Dynamic Interaction between
considered to be problematic and there are many unresolved Reactor Building and Soil", "Model Tests on Base-mat Uplift
problems such as handling of uncertainty in detailed soil of Reactor Buildings", and "Model Tests on Embedment
structure, the nonlinear behavior of soil, and structure- Effect of Reactor Buildings". Numerous important results
structure interaction etc. The effect of adjacent structures on were obtained from these test projects. A brief review of
the earthquake response characteristics of the structure in each project follows.
consideration is one of the biggest problems of this kind. The
effect is termed here as "dynamic cross interaction (DCI)" 2.1 MODEL TESTS ON DYNAMIC SOIL STRUCTURE
which is defined as the dynamic interaction through the soil INTERACTION
1) Seismic Engineering Center, Nuclear Power Engineering
Corporation, Dr.Eng. The project was started in 1980 and completed in 1986. In
2) Seismic Engineering Center, Nuclear Power Engineering this test, three kinds of building model and two concrete block
Corporation. models were constructed in a field, representing an NPP site,
3) Professor, Dept. Architectural Engineering, Science to study basic phenomena related to SSL Two types of
University of Tokyo, Dr.Eng. reactor building model were used, namely BWR and PWR. A
4) Professor, Dept. Center for Cooperative Reserch in model building was also constructed adjacent to one of the
Advanced Science & Technology,Nagoya University, Dr.Eng. BWR type reactor building as a control. These building

18-1
models consisted of a reinforced concrete foundation and a Relationships between contact ratio and input motion
steel frame superstructure with reinforced concrete floors. magnitude, and between contact ratio and response of
Figures 1 and 2 show the layout of test models and outline of structure were examined using shaking table test results. In
each model respectively. Figure 3 shows a typical snap shot of addition, it was clarified that the influence of vertical input
test models, a BWR type reactor building and its adjacent motion on the response of structure models is negligible.
structure. The concrete block models, D, and D2, were This result was obtained from the comparison of structural
designed as cubic models with a side length of 4m. responses to the shaking table vibration test performed with
The purposes of the models are to investigate the SSI effect on and without vertical motion.
buildings, which have a low non-dimensional frequency. In (2) Field tests
addition, the block models are used to investigate the cross Relationships between contact ratio and distribution of
interaction through the soil between adjacent foundations. soil contact pressure as well as stiffness of soil contact
Vibration tests using an exciter and earthquake observations pressure were clarified.
were carried out using these models. These test results were We had obtained much valuable information on the
summarized and presented at several international conferences. phenomena related to the Base-mat uplift. These are
m-[6j. relationships between overturning moment and rotational
In the tests, many factors which affect SSI were studied using angle, between soil stiffness and the base-mat contact ratio,
vibration test data and earthquake observation data, e.g., the between soil damping and the base-mat contact ratio,
relationship between soil spring and base-mat size, soil layer between the dynamic structural response and base-mat uplift,
properties, soil contact pressure distributions at the bottom of and between the amplification in high frequency regions by
the base-mat. The simulation analyses were performed using base-mat uplift and vertical amplification induced by base-
the so called Sway-Rocking (S-R) models based on a 3-D mat uplift. The simulation analyses were also performed
(Dimension) wave propagation theory, 2-D FEM models using S-R models and 2-D FEM models. Simulation
including out of plane viscous boundary and lattice type SSI analysis using a method based on Green's function principle
analysis models. in the time domain was introduced as a detailed model.
Through these tests the following results were obtained:
(1) Interaction between the concrete foundation and soil 2.3 TESTS ON EMBEDMENT EFFECT OF REACTOR
Through excitation tests using four foundation models, it BUILDINGS
was clarified that the spring characteristics of soil could be
explained by the wave propagation theory. The project was started in 1986 and completed in 1995.
(2) Interaction between the reactor building and soil The objective of this project was to comprehend embedment
Basic dynamic interactions between the building and soil effect on the SSI characteristics of reactor buildings, because
as well as between buildings were grasped through vibration it is known that the seismic response of embedded structures
tests and earthquake observation. is greatly affected by the SSI phenomena. The backfill and
surrounding soil resulting from the building embedment
2.2 BASE-MAT UPLIFT TESTS OF REACTOR BUILDINGS makes the phenomena more complicated. The project
includes both laboratory tests and field tests. The laboratory
The project of "Base-mat Uplift Tests" was started in 1981 tests were performed using a shaking table to investigate the
and was completed in 1987. The tests were planned and spring characteristics of the soil along the side of a building
performed to confirm the base-mat uplift phenomena of an and to estimate the motions applied to the building models.
NPP reactor building, which is predicted through seismic The test models used consisted of building models made of
response analyses of the building. We also evaluated the aluminum and a ground model made of silicone rubber. The
base-mat uplift effect on the earthquake response building models were tested under different conditions with
characteristics of the building. Laboratory and field tests and without embedment in the ground model. The tests were
were carried out for this purpose. Laboratory tests were also carried out by changing soil stiffness, embedment depth
carried out using artificial ground models made of silicon and applied input motion. The field tests were performed to
rubber and two types of structure models. One structure investigate the spring characteristics of the actual soil layer
model, called a large-scale model, was made of concrete and and backfill soil around a structure by taking into account the
the other model, called small-scale model, was made of steel. actual soil conditions at NPP sites. Four reinforced
The large-scale model had dimensions of 1m square for the concrete building models representing a BWR and a PWR
base and 2m in height, as shown in Fig.4. In the test, the reactor building were constructed on two different kinds of
structure models were placed on the ground models, which rock fields at the site. In order to investigate the
were fixed on a shaking table, and artificial motions were embedment effect, each test building model was tested under
applied. The field tests were carried out using concrete conditions with and without embedment. Vibration tests and
blocks on two different actual fields representing NPP sites. earthquake observation were carried out using these models.
External forces were applied to the blocks using an exciter. Figure 6 shows a typical snap shot of the building models,
Figure 5 shows a typical snap shot of the field test site. The without embedment (model A) and with embedment (model
test results were presented on several papers at international B), used in the field test.
conferences. [7],[8]. The data obtained from the vibration tests and earthquake
Typical test results obtained in the study were as follows: observation were used to evaluate dynamic soil impedance,
(1) Laboratory tests foundation input motions, responses of structures, and the

18-2
 SSI characteristics of backfill and surrounding soils. The sway component ratio is not influenced by embedment.
evaluated results were then compared using Axisymmctric (4). The Fourier spectra of horizontal acceleration time histories
FEM models. The test results were presented in several obtained through earthquake observation of the test models
papers at many international conferences. [9J-[27]. show that the amplitude of the spectrum peaks due to SSI
Typical test results are summarized as follows; decrease and the predominant frequency becomes higher due
(1). Increments in the real and imaginary parts of the to embedment. (Four typical examples of superimposed
dynamic soil impedance due to the embedment were Fourier spectra of earthquake records obtained from models
observed. It was also observed that the dynamic A(without embedment) and B(with embedment) are shown in
characteristics became more complicated as the depth of Fig.7)
the embedment increased. (5). In the vicinity of the natural frequency of the subsurface
(2). The resonance frequencies of soil-structure systems soil, the foundation input motions for the embedded structures
increase with the embedment, but the resonance'amplitude were greater than those for non-embedded structures.
decreases. (6). Axisymmetric FEM models were confirmed as useful tools
(3). The rocking component ratio decreases and the ratio of for investigating the dynamic interaction of embedded
elastic deformation increases with embedment, but the structures.

50m 50m
N - 1 T

Model C Moclei A V 1
p1
-r-
\
\
AV
/
/
M odel [>2-^
I
i\ IK X c^ V
M ode! f3i

M 3del E t-^
"?

Fig.l
Layout of Building Models, A.B and C, and Block Foundation Models, D[ and EX. Fig.3 A Snap Shot of Building Models A and C
(Model Tests on Dynamic Interaction Between Reactor Building and Soil) (Model Tests on Dynamic Interaction Between Reactor Building and Soil)

(3) Model A

Fig.2 Outline of the Models Used in The Field Tesi Fig.4 An Outline of Laboratory Test Models
(Model Tests on Dynamic Interaction Between Reaclor Building and Soil)
(Model Tests on Base Mat Uplift of Reactor Building)
18-3
 Fig.5 A Snap Shot of Field Test Models
(Model Tests on Base Mat Uplift of Reactor Building)
fGal'sec.) (Gal-sec.)

i 5 IO(Hz)15 5 10(Hz)15
(a)EO.t l ' (b)EQ.2 V '
(Gal'sec.) (Gal'sec.)

0 5 10(Hzl15 5 10{Hz)I5
(c) EQ.3 (d) EQ.4
Fig7.
Comparison of Acceleration Fourier Spectra of Earthquake (EQ.) Responses Model B ( With Embedment)
Observed on RFs of buildings with and without embedment Fig.6 A Snap Shot of Field Test Models
(Model Test on Embedment Effect of Reactor Building) (Model Test on Embedment Effect of Reactor Building)
3. MODEL TEST ON DYNAMIC CROSS conditions, which are approximately the same as those at
actual nuclear power plant sites. The laboratory tests are
being carried out to supplement the field tests using a ground
An overview of the generic NPPs shows that there are two model made of silicone rubber and building models made of
aspects of the dynamic cross interaction (DCI) between aluminum. In these tests, building models are placed on the
adjacent buildings during earthquakes. One is the DCI effect ground model, which is set up on a shaking table. Thus the
between adjacent reactor buildings. The other is the DCI models are subjected to excitations for detailed testing to
effect between reactor buildings and adjacent buildings such evaluate the effect of the spaces between adjacent structures
as turbine buildings. In either case, the dynamic behavior of on the DCI effect among the building models. The tests are
reactor buildings during earthquakes is considered to differ ongoing (started in April 1994 and will be completed in
from that of isolated single buildings, and this condition is March 2002) and the results achieved to date are not
assumed in the current earthquake response analyses for sufficient to meet the final goal of this project. However,
aseismic design of NPP reactor buildings. Thus, the DCI valuable test data has been accumulating gradually. This
effect should be considered as one of the dynamic paper describes an outline and summary of the current status
characteristics of NPP reactor buildings if the effect is too of this project.
large to ignore. This is because the difference in dynamic
characteristics of reactor buildings affects not only the 3.1. BASIC CONCEPT
aseismic performance of the reactor building itself but also
the equipment related to NPP safety. At present the aseismic A system to investigate the SSI effect is shown in Fig.8.
design of such equipment is performed using the evaluated Figure 8(a) shows the method for single buiding and Fig.8
earthquake responses of reactor buildings without (b) shows that for adjacent structures,
considering DCI effects from adjacent structures. The The equation of motion for a single building as depicted in
following factors can be considered as a major cause of the Fig.8 (a) is given as:
effect: the dynamic impedance function, input motion to the
foundation and the dynamic characteristics of the building. ^BF
In order to evaluate this effect, NUPEC has been planning
and carrying out field and laboratory tests under a
commission from the Ministry of International Trade and where, UB and Uf are displacement vectors for the
Industry (MITI) of Japan using models of reactor buildings superstructure (B) and the foundation (F), respectively; KBBl
and adjacent structures. The field tests are being carried out KBF, and KFF, are dynamic stiffness matrices of the
to investigate the DCI effect under actual soil conditions. superstructure and S(C> is the dynamic impedance function
The field tests takes over the field and two building models matrix involving the DCI effects, and A <C) is the foundation
from the preceding SSI research project, "Model Tests on input motion vector.
Embedment Effect of Reactor Building". The site has soil

18-4
 different type buildings. The building models used in this
project are models of reactor buildings (BWR) and a turbine
Building 1 Building 1
building. The scale of these building models is about 1/10 of
the actual buildings. The space between adjacent building
Building 2 models for the two different type buildings (reactor building
D1
and turbine building) was determined by referring to the
D 1
closest example of these buildings, which was obtained by
B2 investigating actual plant construction conditions. The space
G.L. G.L between the identical building models was determined to
facilitate observation of the dynamic interaction effect
F1 F2 between the building models. The vibration tests of the
building models using an exciter were carried out to
evaluate accurate dynamic soil spring characteristics
(a)Single Building (b)Adjacent Building (dynamic impedance function and the foundation input
motion) including the adjacent effect on SSI phenomena of
the modeled buildings. Earthquake observation is also
Fig.8 Soil-structure Interaction System
carried out to investigate the interactions between two
adjacent building models under actual earthquake conditions.
Similarly, the equation of motion for adjacent buildings
Furthermore, in order to comprehend embedment effect on
(Bl and B2), as shown in Fig.8 (b), is described as:
the DCI of adjacent buildings, tests are carried out under
different construction conditions, i.e., with and without
embedment. Thus, the tests are divided into two parts. In
the first part, vibration tests and earthquake observation is
0 &BB2 ^ Ken carried out without building embedment. In the second part,
K BF1 0 KFFi+S tl S 12 building models are embedded and vibration tests and
^ K- BF2 $ 12 K earthquake observation are carried out.

3.2.2 LABORATORY TESTS

where suffixes "1" and "2" represent buildings Bl and B2,
respectively, SfOjy is the dynamic impedance function matrix The laboratory tests are planned to supplement the field
involving the DCI effects, and A (C>f is the foundation input tests. Because the building models used in the field tests are
motion vector involving the DCI effects. too massive and heavy to move, some parameters, which
Equations (1) and (2), respectively, are the equations of have a strong influence on the DCI effect, are difficult to
motion for a single building and for adjacent buildings in investigate in the field tests. The distance between adjacent
which the DCI effect between two buildings is involved. In buildings and/or the mutual adjacent effect between three
both cases, the dynamic interaction between the buildings closely constructed buildings are considered to be such
and surrounding soil is evaluated from the dynamic items. In order to investigate such effects, small-scale
impedance function and the foundation input motion. Thus, model tests are planned as the laboratory tests.
it is important in the present study to evaluate the dynamic Figure 10 shows a general outline of the laboratory tests.
impedance function of DCI in the building models and the The test model consists of a ground model made of silicone
input motion to their foundation through vibration tests and rubber and building models made of aluminum. The
earthquake observations. building models are made to represent reactor buildings and
a turbine building. The scale of the building models is about
3.2. TEST PLAN 1/260. These models are designed so that they have similar
SSI characteristics to those of the building models used in
Field tests and laboratory tests are planned to investigate the field tests. The ground models are fixed on a shaking
the DCI phenomena. Outlines of these tests are described table after placing building models on top. Several artificial
in the following sections. earthquake ground motions are then applied to both the
ground model and the building models to investigate the
3.2.1 FIELD TESTS DCI effect in detail. The parameters in the test are adjacent
building distance, building embedment depth and so on.
Field tests are carried out at the same testing site as the Furthermore, one reactor building model has exciters at the
field tests of the preceding test project, "Model Tests on top and base positions to apply eccentric centrifugal forces.
Embedment Effect of Reactor Building". Two building The building models are used to simulate the vibration tests
models are used from the preceding test project. performed in the field tests, which are carried out using an
Figure 9 shows an outline of the building models used in exciter.
the field test. Three kinds of model conditions are
introduced to investigate the effect of adjacent building on 3.2.3 OVERALL TEST PLAN
the SSI phenomena of the building in question, a single
building model, two identical building models, and two The " Model Test on Dynamic Cross Interaction Effects of

18-5
Adjacent Structures" is being carrying out as an 8 year test adjacent same type building models on their SSI
project from fiscal 1994 to 2002. The test term can be characteristics. In the third step, The turbine building model
divided into two parts at the end of fiscal 1997. In the first DF was tested as a single structure before constructing the
part, total planning and tests without building model reactor building model DA. Then, after the construction of
embedment were carried out. These test results are currently the DA model, the DF model was tested again together with
being studied in detail. In the second part, which began in DA as the final step in order to investigate the DCI effects of
fiscal 1998, the building models are embedded to investigate adjacent different type building models on their SSI
the influence of the building embedment on the DCI effects characteristics.
of reactor buildings. In each vibration test, the buildings were excited in three
directions, NS(north-south) direction, EW(east-west)
3.3. CURRENT STATUS OF TESTS direction and UD(up-down) direction independently. The
vibration force was applied firstly to the top of the building
Because the tests are being carried out over rather a using an exciter. Then the exciter was moved to the upper
long time span, we feel that the current status of the tests part of the base to apply force to the base of the building.
should be reported to researchers in this field to gather Items measured during the test were the displacement in the
many valuable comments on our project. Thus in the major part of the buildings and soil pressure at the bottom of
following section, we present the current status of the tests. the foundation. Observed data was processed into resonance
curves, and the soil pressure distribution and the data were
3.3.1 FIELD TEST used to evaluate vibration characteristics such as natural
frequencies and damping.
Figure 11 shows building model layout in the field tests. Figure 14 shows superimposed resonance curves of the
This figure represents our state of the art of the field test site BAs model, which were obtained in the vibration test before
in the end of fiscal 1997. Construction of all the building and after the construction of the same type adjacent building
models has been completed and earthquake observation is model BAn as an example of the test results. Though the
on-going. dominant frequency is nearly the same, the vibration peak
The site is located on a gentle hill. The building models after the construction of BAn forms two peaks and the peak
are placed in three kind of conditions; single reactor height becomes slightly lower. Figures 15 and 16 show
building model (AA), closely constructed twin reactor examples of the resonant curves of the turbine building
building models (BAn and BAs), and two different type model, DF, and one of the different two adjacent building
structures consisting of a reactor and a turbine building models. Figure 15 shows the resonant curves obtained by
model (DA and DF respectively). The building models are excitation in the NS direction, the direction of the buildings
made of reinforced concrete. Each building model is in a line. The resonant curves of the building model DF,
constructed on soil in a pit, which is excavated down to 4m before and after the construction of the adjacent building
from the surface level to prepare for the embedment of the DA, are superimposed. In addition, in Fig.16, two resonant
buildings in the next stage of testing. The scale of these curves obtained by excitation in the EW direction,
building models is about 1/10 of an actual BWR plant. perpendicular to the NS direction, are shown. In the figure,
Reactor building models all have the same dimensions of the resonant curves of the building model DF before and
8mx8m square in plan and 10.5m in height. The turbine after the construction of the adjacent building model DA are
building model has a rectangular plan of 6.4m in the north- superimposed. In Fig. 15, reduction of the dominant natural
south direction, 10m in the east-west direction and is 6.75m frequency of SSI in the NS direction can be seen after the
in height. construction of adjacent building model DA. On the other
The building models, BAn and BAs, are installed on a hand in Fig.16, the natural frequency of SSI is nearly the
soil layer of the same elevation. On the other hand, the same but a clear peak of 8.5Hz caused by the adjacent DA
building models, DA and DF are installed on different model is seen after the construction of the DA building
elevations in a pit reflecting actual plant conditions. The model. The dominant frequency of the DA model of 8.5 Hz
reactor building model, DA is installed on a lower ground is slightly higher than that of BAs and BAn whose dominant
level than that of the turbine building model by 1m as shown frequencies are about 7.0Hz because of embedment of the
in Fig. 12. Figure 13 shows a snap shot of the DA and DF base of DA model by 1m.
building models. Up to the present, we have complete the As has already been described in the previous section,
construction of building models and vibration tests using an earthquake observation of these building models has been
exciter for these building models under non-embedded carried out. Up to the present, acceleration records of nearly
conditions. Currently we are conducting earthquake one hundred earthquakes have been observed including
observation with these test models. small acceleration records whose maximum acceleration is
The vibration tests were carried out in four steps. In the less than several Gal. Unfortunately these observed
first step, the vibration characteristics of the reactor building records were not necessarily obtained under the same
model BAs was tested as a single structure before structural conditions because the building were not
constructing the same type reactor building model BAn. constructed at the same time. Therefore, the systematic
Then after the construction of BAn, the BAs model was comparison of earthquake observation data for the
tested again together with newly constructed BAn as the evaluation of DCI effects of adjacent building models
second step in order to investigate the DCI effects of become possible from October 1997 when construction of

18-6
 all building models was completed. As a typical example, shaking table test. In these figures, the dominant natural
we show an earthquake observation record of the earthquake frequency of SSI is observed at 5.8Hz (0.17sec. in period).
of November 1997. Figure 17 shows tri-axial acceleration The response spectra in Fig.21 are seen to have nearly the
time histories observed in a free field. Earthquake same characteristics in despite of the differences in the
observation in the free field is prepared at the site excitation direction (cross directional direction). We then
independently of the building models in order to estimate confirmed that the soil-structure model has no particular
the actual input motion to the building models during directional properties. Incidentally, the lower three natural
earthquakes. The acceleration record shown in Fig. 17 is frequencies of the ground model without the building model
observed at 3m below the ground surface level (GL-3.om). were observed at 2.4Hz, 7.3Hz and 13.0Hz. In addition, the
The record has a maximum acceleration of over lOGal. in natural frequency of the building model was observed at
the NS direction. 140Hz in a hammering test.
Figure 18 shows Fourier spectra of earthquake From now on, the vibration test using exciters installed in a
acceleration time histories observed on top of the building reactor building model, and the shaking table test by
models AA, BAs and BAn. Figure 18(a) shows the Fourier applying scaled earthquake motions, will be carried out
spectrum of the acceleration record of AA, a single structure together with other building models to be manufactured as
model. Figure 18(b) shows superimposed two Fourier adjacent building models. In particular, detailed
spectra observed in the acceleration records on BAs and Ban, investigation on the SSI effect under strong earthquake
the same type adjacent buildings. Although the Fourier motions, which cannot be realized in the field tests, will be
spectra in Fig.l8(b) show nearly the same spectral pattern, carried out systematically by taking into account the
the Fourier spectrum in Fig.l8(a) has a quite different distance between the adjoining buildings as a parameter. The
pattern. The decrement of the dominant peak height in the data obtained in the laboratory tests will be used to
Fourier spectra in Fig.l8(b) was observed as compared with supplement the test results obtained in the field tests.
the Fourier spectrum in Fig.l8(a). The cause of the
difference might be pointed out immediately as the adjacent 4. CONCLUDING REMARKS
effects of building model composition, and differences
between single and twin building compositions. However, The present paper describes the on-going study, "Model
because the detailed soil conditions under the building Test on Dynamic Cross Interaction Effect of Adjacent
models are slightly different, so it should also be checked Structures" which is being carried out by NUPEC under a
whether this difference is caused by the local soil conditions commission from the Ministry of International Trade and
or not. Industry. The purpose of the study is to evaluate the
influence of buildings, which are constructed close to
3.3.2 LABORATORY TESTS reactor buildings, on the earthquake response characteristics
of the reactor building. The influence can be regarded as one
For the laboratory tests, we manufactured a ground of the soil structure interaction (SSI) phenomena which are
model made of silicone rubber and a reactor building model still regarded as one of the most difficult problems in
made of aluminum. We then performed shaking table earthquake response simulation and/or design analysis of
vibration tests on the soil and building models. The ground structures. Up to the present, many studies have been
model has dimensions of 2.8m in diameter and 1.0m in conducted to investigate the SSI phenomena. As the results
height. In order to avoid vertical transformation in the of these studies, the effect of SSI on the earthquake response
boundary of the ground model due to its own weight, 180 characteristics of structures has been found to be described
brass bars, each pf which has a 3mm in diameter, were laid in analytical models with the soil springs, which are
around the circumference of the ground model in the axial assumed under the base of the buildings. However, SSI
direction. In addition, to investigate the detailed ground phenomena are still recognized as too complicate to allow
model motion, accelerometers were laid inside the ground proper analyses for earthquake response simulation and
model at 13 points as shown in Fig.19. Thus we were able to design of structures. Thus the success of the project will
measure 26 horizontal (13 cross directional components) depend on ways to evaluate soil-spring characteristics,
and 5 vertical components of acceleration. The building dynamic impedance function and the foundation input
model was made of aluminum with dimensions of motion properly.
30cmx30cm in plan and 40 cm in height. Total weight of the We are carrying out field tests using large-scale test
model was 25 kgf. The model was designed to be similar to building models on actual ground and a laboratory test using
the reactor building model used in the field test. small-scale building models with an artificial ground model.
In figures 20 and 21, we show the results of shaking table At present, the test is on-going and there are not enough data
tests. The tests were performed by putting the building for proper evaluation of the adjacent effect of the buildings.
model on the ground model, which is installed on a shaking However, the tests without embedment have been completed
table, then swept sine motion and/or scaled earthquake and important basic test data are gradually accumulating.
motion were applied. Figure 20 shows a resonant curve The embedment of the building models is scheduled for the
obtained in the shaker test by applying swept sine motion latter half of fiscal 1998.
using exciter installed in the building model. Figure 21 We will make much more effort hereafter to accumulate
shows acceleration response spectra of 5% damping basic test data for evaluating the embedment effect on the
obtained by applying scaled earthquake motion in the dynamic cross interaction (DCI) among nearby buildings.

18-7
 We will also conduct a detailed examination of the existing SMiRT, vol. Kl, K06/3: pp.l23-1278,Tokyo,1991.
test data to establish an evaluation method for the DCI 12. Kobayasi,T.,et al..: "Forced Vibration Test on Large Scale
effect. Model on Soft Rock Site (Embedment Effect Tests on
Soil-structure Interaction)". Trans.llth SMiRT, vol. Kl,
ACKNOWLEDGMENT K06/4: pp.!29-134,Tokyo,1991.
13. Inukai,T.,et al..: "Forced Vibration Test on Large Scale
The project is being carried out through consultation with Model on Hard Rock Site (Embedment Effect Tests on
the sub-committee on "Model Test on Dynamic cross Soil-structure Interaction)". Trans.llth SMiRT, vol.Kl,
Interaction Effects of Adjacent Structure" (Chair person: K06/5: pp.!35-140,Tokyo,1991.
Prof. Dr. M.Iguchi, Tokyo Science Univ., one of the authors). 14. Fujimori,T.,et al..: "Experimental Study on Effects of
In addition, the annual test results have been being Hardness of Supporting Ground Site (Embedment Effect
examined by the executive committee on "Verification Tests Tests on Soil-structure Interaction)". Trans.llth SMiRT,
for Seismic Analysis Codes" (Chairperson: Prof. Dr. vol.Kl, K06/6: pp.!41-146,Tokyo,l991.
A.Shibata, Tohoku Univ.). The authors would like to 15. Kurimoto,O.et al..: "Field Tests on Partial Embedment
express their thanks to everyone on the committee and the Effects (Embedment Effect Tests on Soil-Structure
sub-committee for their hearty encouragement and advice on Interaction)", Trans.l2th SMiRT, vol. Kl, K02/2, pp.43-
how to advanceing the project. 48, Stuttgart ,1993.
16. Fukuoka,A.,et al..: "Forced Vibration Tests on Three
REFERENCES Types of Embedded Structures (Embedment Effect Tests
on Soil-Structure Interaction)", Trans.l2th SMiRT, vol.
1. Iguchi,M.,et al...: "Model Tests on Interaction of Reactor Kl, K02/3, pp.49-54, Stuttgart ,1993.
Building and Soil", Trans.9th SMiRT, vol.Kl, pp.317-322, 17. Fujimori,T.,et al..: "Seismic Response of Embedded
Lausanne, 1987. Structures (Embedment Effect Tests on Soil-Structure
2. Odajima,M.,et al..: "Analytical Study on Model Tests of 18. Maeda,T.,: "Laboratory Tests on The Effect of Partial
Soil-structure Interaction", Trans.9th SMiRT, vol.Kl, Interaction)", Trans.l2th SMiRT, vol. Kl, K02/3,
pp.311-316, Lausanne, 1987. pp. 55-60,Stuttgart ,1993.
3.Iguchi,M., et al..,: "Large-Scale Model Tests on Soil- 19. Kurosawa,R.,: "Laboratory Tests on Soil Structure
Structure Building Interaction Part I : Forced Vibration Interaction with Backfill Soil Using Non-Linear Material
Tests ", Proc.9th-WCEE, vol.111, pp.697-702, Tokyo, 1988. Structures (Embedment Effect Tests on Soil-Structure
4.Iguchi,M., et al..,: "Large-Scale Model Tests on Soil- Interaction)", Trans.l2th SMiRT, vol. Kl, K03/6, pp.97-
Structure Building Interaction - Part II : Earthquake 102, Stuttgart ,1993.
Observation", Proc.9th-WCEE, vol.VIII, pp.315-320, 20. Inukai.T.et al..: "Dynamic Behavior of Embedded
Tokyo, 1988. Structure on Hard Rock Site", Proc.lOth-WCEE, vol.3,
S.Iguchi, M., et al..: "Model Test on Interaction of Reactor pp.1695-1700, Madrid, 1992.
Building and Soil (Part 1: Cross Interaction Tests)". 21. Ohtsuka,Y.,: "Embedment Effects on Dynamic Soil-
Trans.lOth SMiRT, vol. Kl: pp.211-216, Anaheim, 1989. Structure Interaction", Proc.lOth-WCEE, vol.3, pp.1707-
6. Iguchi, M., et al..: "Model Test on Interaction of Reactor 1712, Madrid, 1992.
Building and Soil (Part 2: Excitation by Earthquake)". 22. Fujimori,T.,: "Partial Embedment Effects on Soil-
Trans.lOth SMiRT, vol. Kl: pp.175-180, Anaheim, 1989. Structure Interaction", Proc.lOth-WCEE, vol.3, pp.1713-
7.Hangai,Y.,et al..: "Model Test of Base-mat Uplift of 1718, Madrid, 1992.
Nuclear Reactor Building (Part 1: Laboratory Test)". 23. Shohara,R.,: "Tests on Dynamic Interaction Between
Trans.lOth SMiRT, vol. Kl: pp.169-174, Anaheim, 1989. Foundations", Proc.lOth-WCEE, vol.3, pp.1879-1664,
8. Onimaru,S.,et al..: "Model Test on Base-mat Uplift of Madrid, 1992.
Nuclear Reactor Building (Part 2: Field Tests on Actual 24. Moriyama,K.,et al..,: "Partial-Embedment Test on Soil-
Ground)". Trans.llth SMiRT, vol. Kl, K08/2, pp.177-182, Foundation Interaction", Proc.lOth-WCEE, vol.3,
Tokyo,l991. pp.1911-1916, Madrid, 1992.
9. Nasuda,T.,et al..: "Embedment Effect Tests on Soil- 25. Kurimoto,O.,: "Input Motions for Rigid Foundations to
structure Interaction", Trans.llth SMiRT, vol. Kl, K06/1: Observed Seismic Waves", Trans.l3th SMiRT, vol.111,
pp.lll-116,Tokyo,1991. pp.13-18, Porto Alegre, 1995.
10. Shohara,R.,et al..: "Laboratory Model Tests with Silicone 26. Fukuoka A.,: "Dynamic Soil-Structure Interaction of
Rubber Ground Model (Embedment Effect Tests on Soil- Embedded Structure", Trans.l3th SMiRT, vol.III, pp.85-
structure Interaction)", Trans.llth SMiRT, vol. Kl, K06/2: 90, Porto Alegre, 1995.
pp.H7-122,Tokyo,199l. 27. Ohtsuka.Y.,: "Experimental Studiesy on Embedment
11. Moriyama, K.,et al..: "Comparison between Tests and Effects on Dynamic Soil-Structure Interaction",
Analyses for Ground-Foundation Models (Embedment Proc.llth-WCEE, Paper No.59, Acapulco, 1996.
Effect Tests on Soil-structure Interaction)", Trans.llth

18-8
 Q0

Single Reactor Building Model

( Without Embedment)

AAQ
U
Two Identical Reactor Building Models
(Without Embedment)

Turbine Building Reactor Building

Backfill Soil
2QQ
QQQB
QQQ
Two Different Building Models
( Reactor and Turbine Buildings With Embedment )

Fig.9 An Outline of the Field Test (Models and Test Conditions)

Exciter ( One-Direction )

Exciter ( Two-Direction ) B j Aluminum Plate

Reactor Building Model Detail

Control Building Reactor Building
Soil Model
(Silicon Rubber ) Turbine Building

r-2m

-Irn

^- 0

Shaking Table shaking Direction Specimen Base

Laboratory Test

Fig. 10 An Outline of the Laboratory Test

18-9
 Model D
( Two Different Buildings : Reactor and Turbine Buildings )

Model DA
( Reactor Building)

Model DF
( Turbine Building)

Model B
Model A Identical Buildings : Reactor and Control Buildings )
( Single Reactor Building : AA) \ ~*
\ **
-
5Ar i V
ild ing)
}^
i
Model 1 A
s
ild ing)"Tjjg <ii
P

Fig. 11 Layout of Test Building Models for the Field Test

Model DA

Model DF

-6400 -8000
1000 100 1700

Fig. 12 Cross Section of Building Models, DA(Reactor Bldg.) and DF(Turbine Bldg.)
in The NS Direction.

18-10
 Fig.13 A Snap Shot of Building Models, DA(Reactor Bldg.) and DF(Turbine Bldg.)
O Before the BAN construction
0 After the BAN construction
300.0

0.0
0.0 5.0 10.0 15.0 20.0
Frequency (Hz)
Fig. 14 Resonance Curves Obtained by The Forced Vibration Test of BAs Bldg.
(NS Component due to Excitation in The NS Direction )
O Before the DA construction
After the DA construction
100.0

o 80.0

^ 60.0
ci
< 40.0
d.
<S)

20.0

0.0
0.0 5.0 10.0 15.0 ZO.O
Frequency. (Hz)

Fig. 15 Resonance Curves Obtained by The Forced Vibration Test of DF Bldg.

(NS Component due to Excitation in The NS Direction)
18-11
 O Before the DA construction
After the DA construction
DU.U

g 40.0 ifvj^
o c

_b!
0 0

E
=1 30.0
.
DF l 0

a. DA
0 0 0
0>
< 20.0
S* / "f=
ci
if> \ .

^ 10.0 ^ s
A
'
V^*^axc^*sta.
* .
^^!

0.0
0 .0 5>.o 1().0 1 !i.O 2C
Frequency. (Hz)

Fig. 16 Resonance Curves Obtained by The Forced Vibration Test of DF Bldg.

(EW Component due to Excitation in The EW Direction )

xio Max. Ace.: -11.1 Gal

11 yVi;S'^'"irT.A1 ^F'WITnT'n-iTfi--r--"" -f - "

g -L2
mm i
fll -:

XI
o1 Max. Ace. : - 5.1'Gal
5 L2
*

L JW ^ ,.
ll j
& .0
- '"
< -L2
XI 0 l Max. Ace . : - 2.7Gal
5 L2
.
S .0 ftl.^.p.^
i< -L2 _
t
ii

.0 10.0 20.0
TIV ( sec. )

Fig. 17 Observed Earthquake Record at Free Field of The Testing Site.(1997.11.20)

18-12
 x:o xio
2.0
AA-RFA NSO BAS
-RFA-NSO

cz: M
~ L5 - L5
BAS
Single Building
- BAN
Two Identical buildings

.0 5.0 10.0 15.0 20.0 -0 10.0 15.0 20.0

Frequency ( Hz ) Frequency ( Hz )

(a) Single Structure (b) Adjacent Structure

Fig.18 Fourier Spectra of Earthquake Acceleration Time Histories Observed at The Building Tops

: Horizontal X Direction
- Vertical Direction
' Horizontal Y Direction
T-
8 i.x. Y
3 X. Y IX.Y
n
4
I X. Y. Z 4
' X. Y. Z
8 X. Y. Z
4-
'X.Y
rt -*- 4-X.Y
8 X. Y

A-A Section

Fig. 19 Location of Embedded Acceleration Sensors in The Soil Model

18-13
 _c? 0.30
<>
. C^)^
h^__^l
Shaken by A Micro Shaker Equipped
E in the Bldg. Model i
0
.g 0.20 o
o
"5. 0
s
K/ <->
< 0.10

n nn E^SnSSSmrmrr. }
5 10 15 20
Frequency (Hz)"

Fig.20 Resonant Curve of The Modeled Soil-Structure Obtained in the Shaker Test

Shaken in the X Direction

Shaken in the Y Direction
6000

h=0.05

~ra 4000
o

(73
-i
JD
2000
<D
U

J.02 0.1
Period ( sec )
Fig.21 Acceleration Response Spectra at The Top of Building Model
( Shaking Table Test Results of Cross Directional Excitation )

18-14
 DYNAMIC SOIL-FOUNDATION STRUCTURE INTERACTION ANALYSES
OF LARGE CAISSONS

By Chin Man Mok1, C.-Y. Chang2, Randolph Settgast3, Z.-L. Wang1,

Helge M. Gonnermann4, and C.-C. Chin3

ABSTRACT: A study was conducted to evaluate the important factors affecting the seismic
response of large caissons. This paper presents the results of several equivalent linear and non-
linear analyses performed for a typical case idealized based on the cellular caisson at Pier W3 of
the West San Francisco Bay Bridge subject to longitudinal excitation with a peak rock
acceleration of 0.6 g. This caisson is 127 ft long by 75 ft wide submerged in about 107 ft of
water. It is embedded in 110 ft of soil deposits and is founded on rock. Equivalent linear 3-D
analyses were conducted for the cases with and without the tower superstructure and suspension
cables. The results indicate that superstructures have small effects on the seismic caisson
response. The computed dynamic earth pressure and base stresses indicate that there will be soil-
caisson gapping, rock-caisson base lifting, interface sliding, and soil yielding. The results from
equivalent linear 2-D analyses in the direction of the short axis (longitudinal) are similar to those
from the 3-D analyses. When the soil embedment is removed from the model, the dynamic
stiffness and scattered motions at the caisson top only change slightly. However, the imaginary
part of the foundation impedance functions is significantly smaller. Non-linear analyses were
performed using a 2-D model. The results indicate that side gapping, base lifting, interface
sliding, and soil yielding reduce the earth pressure, base and caisson stresses, and caisson
motions. However, the frequency characteristics of the responses appear to be relatively
unaffected.

1 Senior Engineer, Geomatrix Consultants, Inc., 100 Pine Street, 10th Floor, San Francisco, CA 94111
2 Principal Engineer, Geomatrix Consultants, Inc., 100 Pine Street, 10th Floor, San Francisco, CA 94111
3 Staff Engineer, Geomatrix Consultants, Inc., 100 Pine Street, 10th Floor, San Francisco, CA 94111
4 Project Engineer, Geomatrix Consultants, Inc., 100 Pine Street, 10th Floor, San Francisco, CA 94111

19-1
INTRODUCTION

Large caissons have been used as tower and pier foundations for many major long-span bridges
(e.g., west and east spans of the San Francisco-Oakland Bay Bridge in San Francisco, California;
the Bronx Whitestone Suspension Bridge in New York City; the Tacoma-Narrow Bridge in
Seattle, Washington; and Benjamin Franklin Bridge in Philadelphia, Pennsylvania). Generally,
these caissons are deeply embedded in soft soil deposits overlying rock or in rock-like materials.
In relation to the seismic response and vulnerability evaluation of the bridges supported by large
caisson foundations, an important concern is the effects of soil-foundation structure interaction
(SFSI) on the superstructure response and the imposed load demands. Approaches used to
model the SFSI for large caisson foundations differ substantially in methodology and degree of
sophistication. There is little guidance for practitioners to follow in regard to choosing the
appropriate approach to incorporate important factors under various situations in their analyses.

This study was part of a research project sponsored by the Federal Highway Administration and
conducted by the National Center for Earthquake Engineering Research in Buffalo, New York to
investigate the seismic vulnerability of existing highway construction. In this study, parametric
sensitivity analyses were performed based on rigorous solution techniques to evaluate the
important factors that affect the seismic response of caisson foundations. It includes an
evaluation of soil yielding, gapping, slippage, sliding, and uplift in relation to potentially non-
linear inelastic seismic response of the superstructure and caisson foundation. The results of this
study will be used to develop guidelines on appropriate SFSI modeling requirements and
analysis procedures for seismic analysis of caisson foundations. These modeling requirements
include development of impedance functions (stiffness, mass, and damping matrices) to account
for caisson response, and effective scattered motions for foundations (i.e., foundation input
motions) for use in the response analysis of the superstructure using the substructuring approach.

This paper presents the results of dynamic equivalent linear and non-linear analyses performed to
evaluate the SFSI effects on the seismic response of a typical caisson foundation. The analyzed
example is based on the cellular caisson at Pier W3 of the West San Francisco Bay Bridge
subject to longitudinal excitation with a peak acceleration of 0.6 g at rock outcrop. Equivalent
linear finite element analyses were performed using the computer program SASSI (Lysmer et al.,
1988). Three-dimensional analyses were performed for the cases with and without the
superstructure to evaluate the effect of superstructure on the dynamic caisson response and to
identify the potential for soil yielding, gapping, sliding, and foundation uplift. Two-dimensional
equivalent linear analyses were performed to evaluate the appropriateness of using a 2-D model
to approximate the dynamic caisson response along the short axis (longitudinal direction). Two-
dimensional non-linear finite difference analyses were performed using the computer program
FLAG (Itasca, 1993) to assess the effects of soil gapping, sliding, and uplift on the response of
the caisson.

SUBSURFACE CONDITION

Figure 1 summaries the geologic condition at the site. It is interpreted based on the geotechnical
data provided by the California Department of Transportation. The site is covered by about 110
ft of soil deposits overlying interbeds of weathered sandstone and mudstone. The mudline is

19-2
located at a depth of 107 ft. The top soil consists of about 20 ft of very soft Bay Mud underlain
by about 30 ft of loose to medium dense sandy silt. Below these shallow soft layers is about 30
ft of medium dense to dense silty sand overlying about 10 ft of dense silty sand and gravel. In
between these granular soil layers and the weathered bedrock is about 20 ft of hard sandy
gravelly clay.

The measured shear- and compression-wave velocity profiles are also shown on Figure 1. The
shear-wave velocity increases approximately from 600 ft/sec at 30 ft below mudline to about
1000 ft/sec at 100 ft below mudline. The compression-wave velocity in this depth range is
almost constant at 5000 ft/sec. Below this depth range, the shear-wave velocity increases almost
linearly to about 4500 ft/sec, while the compression-wave velocity increases to about 11000
ft/sec at 140 ft below mudline. Below this depth to about 200 ft below mudline, the shear- and
compression-wave velocities of the rock are about 4500 ft/sec and 11000 ft/sec, respectively.
There is no measurement in the top 30 ft of soil. The shear-wave velocity in the very soft Bay
Mud is assumed to increase from 250 ft/sec to 300 ft/sec. The shear-wave velocity in the
underlying loose sandy silt is assumed based on extrapolation from geophysical measurements.
The compression-wave velocity in the top 30 ft of soil is assumed to be 5000 ft/sec.

DESIGN GROUND MOTIONS

The design rock motions were developed based on the ground motion study performed by
Geomatrix. The acceleration, velocity, and displacement time histories of the longitudinal rock
motion are shown on Figure 2. The corresponding 5% damped response spectrum is also shown
on Figure 2. The predominant frequency of this motion is about 3 Hz.

FREE-FIELD SITE RESPONSE

Free-field site response analyses were performed using the computer program SHAKE based on
an equivalent linear approach. The shear- and compression-wave velocity profiles used in the
analyses were idealized based on the geophysical measurement shown on Figure 1. The shear-
modulus reduction and damping ratio curves for clays were selected based on Vucetic and Dobry
(1991). Those curves for sands were selected based on Seed and Idriss (1970). The acceleration,
velocity, and displacement time histories of the computed longitudinal mudline motion are
shown on Figure 3. The corresponding 5% damped response spectrum is also shown on Figure
3. The results indicate that the site frequency corresponding to the design ground motions is
about 0.9 Hz. The strain-compatible shear-wave velocity and damping ratio obtained in the site
response analyses were used to obtain the dynamic parameters for use in the SFSI analyses.

CAISSON FOUNDATION

The west bay spans of the San Francisco-Oakland Bridge consist of dual suspension bridges
arranged back-to-back around a center anchorage. The general plan of Pier W3 is shown on
Figure 4. The cellular concrete caisson is submerged in 107 ft of water and is embedded in 100
ft of soil deposit. The caisson and the underlying tremie concrete seal penetrate 14 ft into rock.
The caisson is 127 ft long in the transverse direction and 75 ft wide in the longitudinal direction
with twenty-eight (4 by 7) 15 ft diameter circular openings. The openings are filled with water

19-3
and extend to 30 ft above the caisson bottom. The top of the caisson is located at 25 ft above the
water level.

THREE-DIMENSIONAL EQUIVALENT LINEAR ANALYSES

The analyses were performed using a quarter model to take advantage of the symmetry/anti-
symmetry conditions. The SASSI 'structure' finite element mesh is shown on Figure 5. The
mesh includes the caisson, superstructure tower, suspension cables, and two layers of soil/rock
finite elements surrounding the caisson. Rigid links were added at the top of the caisson
distribute the forces from the superstructure. The caisson is modeled by solid brick elements
whose dynamic properties are selected based on smearing the composite flexural and shear
rigidities of the caisson. The hydrodynamic masses accounting for the dynamic effects of water
surrounding the caisson and inside the circular openings are included in the model (Goyal and
Chopra, 1988). The hydrodynamic masses simulating the water in the internal openings are
smeared in the model and the hydrodynamic masses simulating the external water surround the
caisson are treated as lumped masses. The program SASSI was modified to include frequency
dependent springs for modeling the suspension cables. The springs are connected to the
superstructure on one end. Free-field rock outcrop excitation motions were prescribed at the
other end of the springs.

To account for SFSI in the seismic analysis of a global superstructure model of a long span
bridge, foundation impedance functions at the base of bridge piers or the top of caissons
generally are required as input to the analysis. Also required are the input scattered motions
incorporating the SFSI of the caissons at the same locations. To compute the foundation
impedance functions and scattered motions at the top of the caisson, a foundation substructure
model was created by removing the superstructure and cables from the mesh shown on Figure 5.

To study the effects of soil embedment on the seismic response of the impedance functions and
scattered motions, another model was created by further removing the soil elements in the
foundation substructure model.

Characteristics of Impedance Functions at Top of Caisson Foundation

The model shown on Figure 5 was analyzed using the dynamic finite element computer program
SASSI (Lysmer et al., 1988). Figures 6 and 7 shows the real and imaginary parts of the
impedance functions (6 x 6), respectively, at the top of the caisson at Pier 3 in the longitudinal
direction. The impedance functions are frequency-dependent. To account for the frequency-
dependent characteristics of the impedance functions (at least for the real parts) in the dynamic
structural analysis using conventional codes, the following idealizations were made.

The impedance functions are defined as follows:

(i,j = l,6) (1)

where ^ is the real part of the impedance, TCy is the imaginary part, and T is circular frequency.
The impedance functions expressed by Equation 1 are a 6 x 6 symmetric matrix. The diagonal
terms -, i = ,,3 are associated with translations, and -, i = 4,6 are associated with rotations. The
19-4
off-diagonal terms represent coupling between translations and rotations. For the piers analyzed,
only the off-diagonal terms associated with the horizontal translation and rocking G=/, ,=5; and 1=2.
1=4) are significant; the remaining off-diagonal terms are negligible.

An attempt was made to fit a polynomial function through the real and imaginary parts of the
computed impedance functions. The real parts of the impedance functions, ., were fitted by:

kij (co) = (ko)ij -co2 mj (2)

where (} is the static stiffness and (m) is the equivalent mass or mass moment of inertia. For the
imaginary parts, a third-order polynomial function was used to fit to the data.

= Aij + Bij co+djO*.2 +. Dtj

^
a> 3 (3)

The real part of the computed impedance was fitted reasonably well by Equation 2 (shown as
solid lines on Figure 6), indicating that the frequency-dependent stiffness can be reasonably
approximated by the use of a static stiffness and a mass (or mass moment of inertia) at
frequencies up to about 2.5 to 4 Hz for the horizontal translations and rotations. The imaginary
parts of the impedance functions shown on Figure 7 are strongly frequency-dependent, resulting
in dashpot coefficients that also are frequency-dependent.

Scattered Motions at Top of Caisson Foundation

The response spectrum (5% damped) of the acceleration time history (i.e., foundation scattered
motion) computed at the top of the caisson without the tower are shown on Figure 8. Also
shown are the response spectra of the rock motion and the free-field mudline motion. Generally,
the response spectra of the motions at the top of the caisson are amplified from the rock motion
and are lower than the mudline motions at periods longer than 1 second for the longitudinal
component, at periods longer than 0.8 second for the transverse component, and in the entire
period range for the vertical component. Comparisons of the response spectra (5% damped) for
the motions at the top of the caisson with and without the superstructure (tower) are shown on
Figure 9. The two sets of motions are similar in frequency content and spectral values,
indicating insignificant effects of the tower on the response of this caisson.

Effects of Soil Embedment on Impedance Functions and Foundations Scattered

Motions

Because caisson foundations generally are embedded in soft soil deposits, it is desirable to
examine effects of the upper soil deposits on the response of the caisson or specifically on the
impedance functions and input foundation motions at the top of the caisson. Impedance functions
and input motions were computed and compared for two caisson models at Pier W3, one with
and the other without soil embedment.

The real and imaginary parts of the impedance functions computed at the top of the caisson in
the longitudinal direction for the two cases (with and without soil embedment) are compared on
Figure 10 (solid lines are for the case with soil embedment; dashed lines are for the case without

19-5
soil embedment). Generally, the real part of the impedance functions for the case with soil
embedment is slightly higher than that for the case without soil embedment; the differences are
generally less than a few percent. However, the imaginary part is significantly higher for the
case with soil embedment than that for the case without soil embedment, reflecting greater
radiation damping associated with soil embedment. The effect increases with frequency.

Figure 11 shows the response spectra (5% damped) of the acceleration time histories of the
scattered motions for the two cases (with and without soil embedment). The frequency contents
of these motions are similar.

Seismically Induced Soil Stresses Surrounding the Caisson

Dynamic stresses in the soils surrounding the caisson (along the base and side of the caisson)
were calculated and compared with static hydrostatic stresses. The results indicated that
dynamic stresses calculated from the SASSI analyses (based on equivalent linear techniques) are
significantly higher than the static hydrostatic stresses, indicating a likelihood of separation (i.e.,
uplift along the base and gapping along the side of the caisson). Thus there is a need to perform
nonlinear response analyses to examine the effects of potential uplift and gapping on the
response of the caisson.

TWO-DIMENSIONAL EQUIVALENT LINEAR ANALYSES

A 2-D longitudinal model (short axis) was developed by considering a unit-width strip of the 3-D
model described above without the superstructure and cables. The impedance functions and
scattered motions obtained from the two-dimensional analyses were compared with the results
from the three-dimensional analyses. The results from 2-D and 3-D analyses are similar,
suggesting that a two-dimensional model can reasonably approximate the seismic response of the
caisson in the longitudinal direction. Figure 12 shows the comparison of the response spectra of
the scattered motions.

TWO-DIMENSIONAL NON-LINEAR ANALYSES

The primary objective of the non-linear analyses is to evaluate the significance of soil-caisson
gapping, rock-caisson uplifting separation, and near-field soil softening on the seismic motions
and stresses developed in the caisson. A viscp-elastic constitutive model is used to represent the
dynamic behavior of the soil and rock. The dynamic parameters of this model were calibrated to
those used in the equivalent linear analyses. The analyses were performed in time domain. A
Lagrangian approach is used to account for large-strain finite difference grid deformation. The
finite difference grid used in the analyses is shown on Figure 13. Interfaces were added to model
gapping, lifting, and sliding at the soil-caisson and rock-caisson contacts. It is developed based
on the finite element mesh used in the equivalent linear analyses. The grid boundaries were
extended sufficiently far away from the caisson to reduce the boundary effects on the caisson
response. Viscous dashpots were attached to the boundaries to simulate the wave propagation
through a semi-infinite medium. The input control motions are defined at the base and were
obtained as interface motions at the appropriate depth from the free-field site response analyses.

19-6
The cases analyzed are presented in Table 1. These cases include various soil-caisson and rock-
caisson interface properties as well as different levels of near-field soil softening.

Figure 14 shows the comparison of 5% damped response spectra of ground motions computed by
the 2-D SASSI and FLAG analyses at the top of the caisson assuming no interface gapping,
lifting, or sliding. The comparison of acceleration time histories at the center of the caisson at
the top, mudline, and base levels are shown on Figure 15. The shear and bending moment time
histories induced in the caisson at the mudline, above-tremie seal, and tremie seal levels are
compared on Figures 16 and 17. These comparisons show that the results from these two
programs are similar.

Figures 18 through 21 show similar comparisons of response spectra, acceleration time histories,
and caisson shear and bending moment obtained for the cases of smooth interfaces, moderate
interface strength, and glued interfaces (perfect contact). The results indicate that the seismic
motions and stresses developed in the caisson are sensitive to the interface properties. A softer
interface tends to reduce the peak response, but it does not significantly affect the frequency
characteristics of the response. The predominant frequency appears to be relatively insensitive.
It may be because a visco-elastic model was used to represent the dynamic rock behavior.

Similar comparisons of response spectra, acceleration time histories, and caisson shear and
bending moment obtained for the cases of different near-field soil softening were performed.
The results indicate that the responses are not sensitive to the properties of the soil because the
resistance provided by soft soil is small. This behavior is similar to the results obtained by
equivalent linear analyses without soil embedment. A comparison of the 5% response spectra of
the motions computed at various caisson levels is shown on Figure 22.

CONCLUSIONS

The results of this study indicate that soil embedment has significant effects on the imaginary
part of the impedance functions computed for the top of the caisson. However, the real part of
the impedance functions only decreases slightly when the soil embedment is absent. The
scattered motions at the top of the caisson are very similar for the cases with and without soil
embedment.

The lateral earth pressure, base contact pressure, and soil stresses computed by the equivalent
linear analyses indicate the possibility of soil-foundation separation (gapping and uplift). The
results of non-linear analyses indicate that motions and stresses developed in the caisson are
sensitive to the soil-caisson and rock-caisson interface properties. The peak responses are lower
for softer interface strength. However, the frequency characteristics are less affected.

ACKNOWLEDGMENTS

The research on which this paper is based was supported by the National Center for Earthquake
Engineering Research (NCEER), Highway Project No. 106, Seismic Vulnerability of Existing
Highway Construction, under the sponsorship of the Federal Highway Administration (FHWA
Contract No. DTFH61-92-C-00106).

19-7
APPENDIX I. REFERENCES

Itasca Consulting Group, Inc. (1993). "FLAG, fast Lagrangian analysis of continua, Version 3.2."

Lysmer, J., Ostadan, R, Tabatabaie, M., Vahdani, S., and Tajirian, F. (1988). "SASSI - A system
for analysis of soil-structure interaction, user's manual", Geotechnical Engineering Division,
Civil Engineering Department, University of California, Berkeley, and Bechtel Power
Corporation.

Seed, H.B. and Idriss, I.M., (1970). "Soil Moduli and Damping Factors for Dynamic Response
Analyses", Report No. EERC 70-10, Earthquake Engineering Research Center, University of
California, Berkeley, California.

Vucetic, M and Dobry, R. (1991). "Effect of Soil Plasticity on Cyclic Response", Journal of
Geotechnical Engineering, ASCE, Vol. 117, No. 1, pp. 89-107.

Goyal, A, and Chopra, A.K. (1988), "Simplified Evaluation of Added Hydrodynamic Mass for
Intake Towers", Journal of Engineering Mechanics, Vol. 115, No. 7.

19-5
 Mudline
Compressional-wave velocity (ft/sec)
(El.-107)
0 5000 10000 15000
0
^ I '
Shear-wave Velocity -120 Very Soft Bay Mud
-20
Compressional-wave Velocity
-140
-40 Loose to Medium Dense Sandy Silt PIER
-160
CAISSON
-60

-180 Medium Dense to Dense Silty Sand

TJ
3 -200 Dense Siltv Sand and Gravel
-100 Caisson Bottom
JO Hard Sandy Gravelly Clay (El. -224 ft)
T3
ffl -220 zone). V
f -120
Shale Tremie Concrete
-240
-140
El.-231 ft
-260
-160

.280 Figure 1: Idealized Geologic Profile

-180

-300
-200
0 2000 4000 6000
Shear-wave Velocity (ft/sec)
 Displacement (cm) Velocity (cm/sec) Acceleration (g)
Spectral Acceleration (g)
0000000000004^04^00 p o
o
CD
to

I
M
O >
o
o
X

GO
I___-
 I I I I I I 11 I I I I I I II

a
o

1 ' o
o
OH
00

I I I I I I I 11____I I I I I I 11 I I I I
0.01 0.1 1 10
Period (sec)
i i i i i i i i i

0.5

0
A C
o -0.5
o
i i i i I i i i i I i i i i I i i i i I l I l i I i i i i

160 1 i i i i i i i i i i i i r
80

8 -80
I I
a>
> -160 i i i i i i i i i I i i i i I i ii i I i i i i i i i j i

80
40
<D

I
-40
ICO

-80 i i i i i i i i i l i i i i l i i i i I i i 1 I J L
0 10 20 30 40 50 60
Time (sec)

Figure 3: Computed Longitudinal Free-Field Mudline Motions

19-11
 El. = 25'

Steel Casing
Steel Casing

vo
I
Tremie
Concrete

El. = -195'
Tremie
Concrete
El. = -224'
Tremie
El. = -231' Concrete
75' Seal

Elevation

Figure 4: General Plan and Elevation

 Figure 5:
SASSI Finite Element Quarter Model

19-13
 l.OOE+6 l.OOE+8 T | I | I | T
Computed by SASSI
Approximated by K - w M_
O.OOE+0 5.00E+7
PH
3

X
-l.OOE+6 O.OOE+0
Q
-2.00E+6 -5.00E+7

-3.00E+6 i I i -l.OOE+8

l.OOE+8 2.00E+10
1 ' [
i \ \ i i
5.00E+7 1.50E+10

O.OOE+0 l.OOE+10
Q
5.00E+9

-l.OOE+8 i i i I . I O.OOE+0 I . I i I i
1 2 3 1 2 3
Frquency (Hz) Frequency (Hz)

Figure 6: Real Part of Foundation Impedance Functions at Caisson Top

 2.00E+6 O.OOE+0
Computed by SASSI
Apprx. by 3-deg polynomial
1.50E+6 -2.50E+7
OH

l.OOE+6

5.00E+5 -7.50E+7

O.OOE+0 -l.OOE+8
I
I-1
en
O.OOE+0
1 i ' r i r 8.00E+9

6.00E+9

4.00E+9
Q

-l.OOE+8 O.OOE+0 i . i
1 2 3 1 2 3
Frquency (Hz) Frequency (Hz)

Figure 7: Imaginary Part of Longitudinal Foundation Impedance Functions at Caisson Top

 8 T i I I I 111 I I i i i I 111 I I T i i i i
Scattered Motion at Top of Caisson (El. 25 ft)
Free-field Mudline Motion
Input Rock Outcrop Motion

I o
M
CTi o

OH
C/D

0
0.01 0.1 10
Period (sec)

Figure 8: Comparison of 5% Damped Response Spectra of

Substructure Scattered Motions with Computed
Free-Field Mudline and Input Rock Outcrop Motions
 8 I I I I I iI

Motion from Complete SSI

Scattered Motion

o
iai

ctf
s
CD
2
^

I I I I I I iI I i I I I 111
0
0.01 0.1 1 10
Period (sec)
Figure 9: Comparison of 5% Damped Response Spectra of
Substructure Scattered Motions with Motions at
Caisson Top from Complete SSI Analyses
 l.OOE+6 l.OOE+8

With Soil Embedment

O.OOE+0 5.00E+7 Without Soil Embedment"/
/_|
/
-l.OOE+6 O.OOE+0

X
-2.00E+6 -5.00E+7

-3.00E+6 I I I I I I I -l.OOE+8

l.OOE+8 2.00E+10
M
I
00
i I 'i 1r i r i i ii r i I '

5.00E+7 1.50E+10
5/3 i
5/3
i i
^1
O.OOE+0 v^ l.OOE+10
fi
I
J?-5.00E+7 5.00E+9

-l.OOE+8 O.OOE+0 i i i i i i
1 2 3 1234
Frquency (Hz) Frequency (Hz)

Figure 10: Comparison of Real Part of Foundation Impedance Functions

for Cases With and Without Soil Embedment
 2.00E+6 I I I O.OOE+0 i i' i r
I f
With Soil Embedment _
Without Soil Embedment
1.50E+6

t
en

l.OOE+6
CO
PH

-5.00E+7

X
fe 5.00E+5 -7.50E+7

O.OOE+0 -l.OOE+8 J_i I i i i

Mi O.OOE+0 8.00E+9
<& i i r i r i i

^-2.50E+7 6.00E+9
i
PH

^-5.00E+7 4.00E+9
Q

j|-7.50E+7

-l.OOE+8 O.OOE+0 i i i I . I
0123 1 2 3
Frquency (Hz) Frequency (Hz)

Figure 11: Comparison of Imaginary Part of Foundation Impedance

Functions for Cases With and Without Soil Embedment
 Spectral Acceleration (g)

o
 Elevation (ft)
_ 100
_ o
-100
_-400
_-500
300 ^200 -l6o O1 ibO 2dO300
Horizontal Distance from Centerline of Caisson (ft)
Figure 13: Finite Difference Grid
19-21
 Spectral Acceleration (g) Spectral Acceleration (g) Spectral Acceleration (g)
1

CD

f" 3
w
I-H o
O
p
C CfQ
i-a. a
e HI
o
D-
C/5
fD
O

' ^
o|
O g
B Cfl
-5 ^
a
CD
 2 I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I
1 _ Top of Caisso
0
o -1
o
-2 i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i

1 i i i i I i i i i i i ii i^ r r i r Ti \\ \ \\ ir i i i r i i i i i
).5 I_ Mudline Level of Caisson
0
o ).5
I
o
N>
CO
< -1 i i i i i i i i i i i i i i i i i i i i_i i i i i i i i i i i i i i i i i i

i i i i i r i i r r i i i i i i i i i i i i i i r i i i i i
| 0.5
Bottom of Caisson 2DSASSIJI
LAC -
1 0

o
-1 i i i i i i i i i i i_i i i i i i i i i i i i i i i i i i iii i I i i i i
0 10 15 20 25 30 35 40
Time (sec)

Figure 15: Comparison of Longitudinal Acceleration Time Histories

from 2-D SASSI and FLAG Models
 ^20
I i r i i i i i i i r i i i i i i i i i i r i i i
10 I_ Mudline Lev<!. of Caisson
0

-20 i i i i I i i i i I i i i i I i i i i I i i i i i i i i i i i i i i i i
^ 40
i i i i i F i i i i i i i i i i i i i i i i i i i i i i i r i i i
Above Tremie Seal n 2D SASSI.

fi 0
-20
.8 -40

i i i I i i i I I I I I I I I I I I I i I I I I I I I I I I I I I I i i i

10 Tremie Seal
1/5

o
<D
-20 i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i
0 10 15 20 25 30 35 40
Time (sec)

Figure 16: Comparison of Longitudinal Shear Stress Time Histories

from 2-D SASSI and FLAG Models
 ^ 200000 I I I I I I I I I I I I I I I I i i i i I i i i i
Mudline Level of Caisson 2DSA
SASSI _
^ 100000
2D FLAG
^
g -100000
G
* -200000 i i i i i

I I I I Al I I I I I I I I I I I I I I I

) 100000

i i i i i i i i i i i i i i i i i

i i i i i i i i i i i i i i i i i i i i i i I I I I I I I I I I I I I I I I
Tremie Seal . * ' ' '

i i i i i i i i i i i i i i
15 20 25
Time (sec)

Figure 17: Comparison of Longitudinal Bending Moment Time Histories

from 2-D SASSI and FLAC Models
 8 T I I I I I I 11 I I I I I I III i I i i i M
Smooth Interface Top of Caisson
Ctf u - Moderate Interface Strength
V-t
JQ Glued Interface
"to A
8 4
<
i 2

o
CD
o< o
C/3

i i i i i i 11 i i i i i 11 i i i i i i 11

-H
Mudline Level of Caisson
ta
CD 0
8 2

CD
I I I I I I M
0

4 I I I I I I M i i i i i 111 i i i i i 111
Bottom of Caisson
ctf 3

8 2

CD
I I I I I I M I I I I I I M
0
0.01 0.1 10
Period (sec)

Figure 18: Comparison of 5% Damped Response Spectra of

Longitudinal Motions Computed by FLAG
Models with Various Interface Properties

19-26
 1.6 i i i i i i i i i i i i i i i M
i 'i 'm/' 'i I i i i i i i i i i i i i
1_ Top of Caisson i i

''-' ^ '/ Fj i M | j y 9 j< u- M ' If (j v H f W 1/ 1 } II (' '"?' f ' '

i/kiiVA/V^^1*
8-0.8
o
I I I I I I I I I I I I I I I I I I I_ IIIJIIIIIIIlJlllI

0.8 i i i I i i i i i i i i i | i i i i i i i i i i i i i i
0.4 Mudline Level of Caisson
I
CO 0
0)
"8-0.4
o
i i i i i i i i i i i i i i i i i i i i i i i j_i i i i i i i i i

0.8
0.4
Bottom of Caisson
0
0) rface
"8-0.4 Moderate Interface Strength"
o I I I I I |-r T --i- r - 1 Glmedlnferfacj i i i i
I I I I I I I I I I I i I I I I

0 10 15 20 25 30 35 40
Time (sec)

Figure 19: Comparison of Longitudinal Acceleration Time Histories

from FLAG Models with Various Interface Properties
 ^-H
i i i i i i i i i Srnodth Interne ' '
Mudline Level of Caisson Moderate Interface Strength
t/3

a-
VI

-20 I I I I I I I I I I I I I I I I I I I I I I I I I I I I
CO

20 i i i i i i i i i i i i i i i i i TIM i i i i i i \ i i i i i i i i
_ Above Tremi^ Seal ' i h ' ' h '
t/3

a-10
CO

I
M
oo
s ?n
^c
CO"^U
i i i i i i i i i i i i i I I I I I I I I I I I I I I I I I I I

i i i i i i i i i i i i ] i i i [ i i i i i i i i i | i i i i i i i i
Tremie'Seal ' ' i . i i i
C/3

VI

CO
-20
o 10 15 1 20 25 30 35 40
Time (sec)

Figure 20: Comparison of Longitudinal Shear Stress Time Histories

from FLAG Models with Various Interface Properties
 Sg 100000
g) jjx 50000
^
C5 O" n^
& g -50000
Jj-100000 I I I I I I I I 1 I I I I I I I I I I I I I I I I I I I I I I I

^ 200000 I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I
Smooth Interrice
g) i. 100000 Above Tremie Seal /, Moderate Interface Strength

? ^ 0

i -100000
to
S -200000 I I I I I I I I I I I I I I I , , , , , I I _l I J I I I I I I I I

^ 200000 l l l l I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I

g) j^ 100000 Tremie Seal

W g-100000
^ -200000 i i i i i i i i i i i i i i i i i i' i i i i i i i i i i i i i i i i i
0 10 15 20 25 30 35 40
Time (sec)

Figure 21: Comparison of Longitudinal Bending Moment Time Histories

from FLAG Models with Various Interface Properties
 i i n i mi i i i i i111 i i i i i i 11
without Reduction Top of Caisson
a

8 4
2
o
<D
0

Mudline Level of Caisson

I I I I I I I I I I I I I I I I I I T I I I I I I

Bottom of Caisson
a
<L> ~
8 2

o
(L)
P.
0 I I I I I I M I_____I I I I I I I I i i
0.01 0.1 1 10
Period (sec)

Figure 22: Comparison of 5% Damped Response Spectra of

Longitudinal Motions Computed by FLAG
Models with and without Soil Stiffness Reduction

19-30
 SEISMIC DESIGN PROCEDURE OF BUILDING STRUCTURES

INCLUDING SOIL-STRUCTURE INTERACTION EFFECT

BY

Yoshikazu KITAGAWA0. Yutaka MATUSHIMA25. Yutaka YAMAZAKI3'

Soichi KAWAMURA4>, and Yoshio INDUE5'

1) Department of Structural Engineering . Hiroshima University

2) Institute of Engineering Mechanics , University of Tsukuba

3) Building Research Institute , Ministry of Construction

4) Technology Research Center , Taisei Corporation

5) Housing and Urban Development Corporation(HUDC)

ABSTRACT

The objective of this paper is to present a procedure where design values of buildings are estimated by taking

into account the dynamic soil-structure interaction produced during earthquakes. The building structure is

represented by the mass spring system having swaying and rocking springs and dashpots located at the

foundation. The subsoil layers are transformed to the uniform medium with an equivalent shear wave velocity and

Poisson's ratio. Generally, this is applied to the reinforced concrete and steel framed reinforced concrete building

structures. The numerical results indicate the reasonable feature of the soil- structure interaction effect.

INTRODUCTION structure.

On the other hand, building structures have been

Since the seismic design method for the building constructed on rather soft unstable soil, along with

structures in Japan was amended in 1981, research focus efficient uses of the land in Japan. In order to obtain

has changed from static analysis to dynamic analysis. more rational structures, it becomes all the more

Dynamic analysis, as a technology aided by rapid progress, necessary to evaluate the seismic design for building,

in recent year, in both theory and computers appears to taking into consideration the characteristics of the softr

be the most precise design method for the dynamic soil ground motion and the soil-pile-structure interactions.

phenomenon of an earthquake. In the frequency domain, Taking into account these aspects above, we carried out

the response level at the structure roof is determined by this research for building structures of HUDC while

the products of the following five characteristics; (1) considering dynamic soil-structure interaction effects

epicentral characteristics, (2) seismic wave propagation theoretically, analytically, and experimentally. 1). 2) The

characteristics, (3) ground characteristics surrounding main purpose was to set up the standards for evaluating

the structure, (4) interaction characteristics between seismic inputs for design.

structure and surface layer, and (5) characteristics of On the premise of these conditions, this paper presents a

20-1
 procedure where seismic design values of building rs = 4
structure are estimated by taking into account the soil-

structure interaction and an example of numerical results T s = 0 .05 (l og 2 A + I .0 )2 + 0 .95 (3)

obtained by applying the procedure to a real building. where, A is ratio between the length and the width of the

building (=2c/(2b)). 2c is the width of the building

MODELING OF BUILDING AND DESIGN PROCEDURE perpendicular to the direction of vibration and 2b is the

width of the building of the direction of vibration. _/r& .

L

The building structure is represented by the mass- KK are swaying and rocking spring constants of the

spring system having swaying and rocking and dashpots embedded part of the foundation and calculated according

located at the foundation as seen in Fig.1 . to the equations below:

At that time, the subsoil layers are transformed to the E K s = F K s -(G 0 /G.)-( E H/rs )

uniform medium with an equivalent shear wave velocity E K R = 2.5 F K K -(GJG.)-( E H /rR )

and Poisson's ratio. Design procedure in the proposal is (4)

shown in Fig.2. where, G , G are the average shear rigidity at the soil

touching the side wall, and effective shear coefficients of

EVALUTION OF SPRING CONSTANTS the soil. G0 /Ge ~ 1. // 's ti"16 depth of the base

foundation (bottom of the base beam) from the ground

The static swaying and rocking spring constants of the surface. However, when compression/settlement of the

soil. K .1r . Kn are calculated according to the following soil is expected, the settlement shall be subtracted.

equations: P KS. P KR are swaying and rocking spring constants of

KS ~ Ms '(r^s + E^s + p^s) the pile and calculated acceding to the equations below:

K R =MR -(rK R + E K R+P K R ) (D = a f -N.

where, /a ^ are coefficients for modification
=Z n K (5)
regarding cumulation of swaying and rocking constants of

the soil and the pile. r Ks , r KR are calculated where, j is the serial number of the pile and y is the

according to the equations below; distance between the center line of the foundation and

the center of the j th pile- ap , the reduction

coefficient of the swaying constant of piles due to the

(2) group-pile effects, is calculated according to the following

where, y , v , g , and V are. respectively, effective equations:

unit of the soil, effective Poisson's ratio of the soil. ap - exp \- (a pl ^/w, -1 + aP2 ^jn2 -l)}
acceleration of gravity, and shear wave velocity of the soil. an = Q.5/JSJB
rs , rR are effective radii of the foundation swaying and aP2 = Q.3/JS2 /B (6)
rocking motions, respectively. y/ . y/ are coefficients where, St is the interval of piles in the vibration direction.

for modification of the effective radii and calculated 5 is interval of piles perpendicular to the direction of

according to the equations below; vibration, B is the diameter of the pile and N is the

20-2
total number of piles (Fig.3). Jf is the swaying spring

constant of a single pile and in case pile-head is fixed, it is

calculated as follows:

P Ks = l2 p -lP -/3 P3 /fo + flPlP )3 + 2} P^

K K - P aR
d -PK
KR u<y
(12)

flp=^Kh -B/(4-EP -IP ) (7) where, g ds , ,,ds , s dR . p dR are the ratio of dynamic
where, is Young's modulus of the pile / is the rigidity reduction and calculated as follows:
2nd inertial moment of the pile and f(.h is horizontal s dK =(l-0.05aOR )2 ;aOR <\0
reaction of the soil. ^ is the spring constant of the s dK =0.25 -aOK>\Q

j th single pile in the vertical direction and calculated

according to the equation below: where, aOR is a non-dimensional frequency regarding the

P K Vl =a-Ap -EP lt (8) rocking spring and equal to Q-T^/F Q is the 1st

where, A is the net area of the cross section of the pile, natural circular frequency of the interaction system based

(. is the length of the pile and a is the coefficient for on the static spring constant(=27r/7])- The j th

each of pile. In case of a supported pile which is natural period \T I and the mode of the interaction

reinforced concrete pile in site, it is calculated as follows system is calculated by and use of weight of each story,

a = 0.022(l/B)- 0.05 (9) spring constants, weight of the foundation, and dynamic

where, ^^ fj, are calculated as follows: swaying and rocking spring constants.

/<, =(!-,)"+,"
//i, =1.0 00) EVALUATION OF DAMPING FACTOR

where. # is the ratio between the swaying spring

constants of the pile and the total of swaying spring The damping factor of the j-th mode (h ) for

constants of the soil and pile and calculated as follows: calculating the response story base shear

<XS = P KS /( F KS + K KS + P KS ) (1D coefficient^ C,). is calculated by the following equation:

The natural period of the interaction-system is rteq,j~BfI/ J /7-J-

Bn I/
^S/SJ S'/?l ^^s/R
f -J- V t j -/7n-J-V
S riR ^P/S,j p'/Jl *>t-

calculated by the use of weight of each story, spring +PrK.J -P h'R d4)
constants, weight of the foundation and static soil swaying where, the damping factor of each part is weighted

and rocking spring constants calculated in that step. The according to the strain energy, y is the ratio between

dynamic swaying and rocking spring constants of the soil, the associated strain energy and the whole strain energy

Ks , KR are respectively calculated according to the of the vibration system at the j th vibration mode.

following equations: B h is the damping coefficient of the building and is

normally 0.02. ./jL Jj are damping coefficients of the

soil for the swaying spring and rocking spring, respectively,

modified according to the layered soil and calculated as

ft, =// (= 1.0) follows:

S"S = S"S'S S'S K S'S"S~*~S"I

20-3
where, 8 is the coefficient for modification of the v K = 1 .0

damping of the soil due to the layered soil, s is the s Ks =Jl-A'/(2A)

coefficient for modification of the soil due to the (18)

embedment of the foundation. K is the coefficient for where, A is area of the opening at the foundation and

modification of the soil due to the opening of the A is the area of the foundation (=2b*2c). ^ <,hR are

foundation and h is the damping coefficient of the half- calculated respectively as follows:

space elastic soil replaced with the equivalent. o

h/ is h=0.30-a

the constant of visco-damping of the soil and is normally ;aOK <\

0.03. v A are coefficients for modification of the s hK =Q.\5-aOK -Q.Q5 ;\<aOK <5

swaying spring and rocking spring of the soil, respectively, .A =0.70 ;aOK >5 (19)

due to the embedment of the foundation and calculated The damping coefficients of pile for the rocking spring,

as follows: respectively, modified to the layered soil. hs ,, hl( are

calculated as follows:

;aox <c aos / s ~ P s ' p MS 'F"I

phR = P SR -,,hK + P h, (20)

s ss =1- fc.2,^ -aos )-Q.9(yjyL, -i)3 ]/
where, ; /j h are respectively the damping

l\2s aos -c aos )\VjVc, -l)3 +0.36J] coefficients for the swaying and rocking of pile located at

;c aos <aos <l.2g aos the equivalent hatf-apace elastic soil and are normally 0.03.

s 6S = 1 .0 ;1 .2g aos < aos <l.2g aos h is damping coefficient due to the visco-damping of

^=0.05 \aOR <c aOR the pile and is normally 0. A . K are coefficients for

S 8R = Q.Q5 + 0.95(ZOR -c aOK )/(\.8g aOK -c aO modification of the clamping coefficient of the pile for the

;c aOK <aOR <l.8f; aOK swaying and rocking due to the layered soil is 1 , except in

S SK =LQ ;L8g a0,<aOR (16) the case of special investigations.

where,
RESPONSE STOREY SHEAR FORCE COEFFICIENT

FOR DESIGN

The response story shear force coefficient for the

design for the i-th story is calculated as follows:

fi c,=z.A, -*,(T;)^ -c0 (2D

Q is the 1 st natural circular frequency of the interaction where. Z, /?, (ij ). Q, are seismic hazard zoning

system based on the dynamic soil spring constant and coefficient vibration property at the 1st natural period

H is the thickness of the equivalent two layered soil's VT ). standard layer shear force coefficient as given in the

1st layer. e K K are calculated, Building Standard Act respectively. ^ is the coefficient
00 O/\ ^ O O A

respectively, as follows: at the i-th story showing the vertical distribution of the

response story shear force coefficient 1= (f> /$) A,

20-4
is the coefficient for modification of the spectrum at the building has fourteen stories and are made of reinforced

1 st natural period and calculated as follows: concrete (Rgs. 4 and 5). The underlying ground consists

mainly of sitty soil down to depth 18m from the ground

i ;heq , <0.05 surface (Rg. 6). The constants of the soil are shown in

(22) table!. The building is supported by reinforced concrete

where, /( is unity in the usual range of the natural period piles, 21 m in length placed in site. The diameter of the

\T } and is calculated by the use of 7" in accordance piles is from 1.6m to 1.9m.The constants for analysis are

with the equations below: shown in Table 1. In table 1. A!, is the spring constant of

f,=0 ;7^ <0.05. T; >10.0 each story of the building and K,

A
K. i\r, are dynamic soil

/, = (f og 7; +1.30)/0.3 ;0.05 < T; < 0.1 spring constants of the swaying and the rocking,

{,=1.0 ;0.1<7^<2.5 respectively. The design story shear force coefficients

/, = (l - l^ )/0.6 ;2.5 < 7^ < 10.0 (23) in a longitudinal direction of the building evaluated by this

where, ^', is .the storied shear force coefficient for proposal is shown in Rg. 7. It is less than required by the

calculating at the /' th story of the building and is code owing to the effect of soil-structure interaction.

calculated as follows:

(24) CONCLUDING REMARKS

where, N is the total number of stories above the A procedure where seismic design values of building

foundation and W is the weight of the m th story. structure are estimated by taking into account the soil-

B [/is i-th participation function of the m-th story. structure interaction is presented. The numerical results
rj mj J

/I is the coefficient for spectrum modification due to obtained by applying the procedure to real buildings

damping at the j th natural period and is calculated indicate the reasonable features of the soil-structure

similarly by the equations for %. ^(r J is the design interaction effect

spectrum coefficient at the j th natural period \f J

as given in the Building Standard Act. k is the highest ACKNOWLEGEMENTS

mode number to be considered to be more than 2. In order to execute this research, a research committee (chairman, the late Y

However, when the base shear force coefficient of the 1 st Osawa. professor emeritus. University of Tokyo) was crgarized in the Japan

story falls below the shear force coefficient Z ' R~C Association for Bulding Research Promotion, and the scheme of the research and

given in the Building Standard Act its lower limit is 75% of studying resdts were carried out The authors wisfi to express there sroere

7. ' RI ' C, and the response story shear force coefficient thanks to members of the committee and persons concerned.

for each story is modified at the same rate.

REFERENCES

EXAMPLE OF APPLICATION FOR PROPOSAL 1), 2)lnoue.Y. and Osawa. Y. et al," Study on Seismic Input to Apartment Houses

of HUDC (part-6.7jBJ " . Proc. of Annual Meeting of AU Oct 1987 / " A

Various types of building structure founded on different Proposal for Seismic Design Procedure of Apartment Houses including SoJ-

grounds are analyzed by following this procedure. The Structure Interaction Effect " . Proc. of 9th WCEEvol VIII. PP.365-370. Aug. 1988

20-5
 Mode Ii ng
Swaying Spring

I
V
Rocking Spring

(a) Buildin&Foundation and Soil (b) Swaying-Rocking Model (SR Model)

Fig.1 Interaction System Model

START
Direction Analyze
Condition of Building
and Soil
Calculation of Static-Soil
Spring Constants nt

Calculation of the 1 st Natural

Period of the Interaction System
(using static soil springs)

Calculation of the Natural Period

and Mode of Interaction System
Fig.3 Total Number of Piles n )(n2
I Setting up Damping Coeflc.lents

Calculation of the Design

Response-Storey Shear Force Coefficent
*
-n cng cr .a
.zrs^a ci:?s =::zo
U -l L__ __. C:;D

r ^>
Design Base Shear Coefficient e i cri.-.r) crrro CZ~Z^3 K==7=l C=1^J crnza
at the 1st Storey
Nl e-^r=.i crrra cr^-^3 CT- J

S.S50
37.JCO
Fig.2 Design Procedure

2.CSQ
*IT Fig.4 Plan of Building
2i3- 2XM
2.TM
- L depth
ZH-
2J30
(TO*
urlftce
olU
" Hit
01
.( "/"i
, toi. m^P'\i
29- 2.KO , 2.9M 135 ' "-'-
- 730 70
2a- rr
: *;
. 2.590 130
3.0 III Vi
26- - , 1.633 100 'S| U^-
110
Zjffl I.CCO 105 ?-^|
IW
24- Z) . .r,dy "' '. ill
2-eo L IO.MO 250 ;-..*
Z3- 2SJJ
isa
250) .,., 39.260 450
v
-299

s^
ZQ~~~ i .'
12M r
M Hi

39.0

- I6.S30 300
**""

';*"
t? -5QD

Fig.5 Section of Building Fig.6 Constants of Soil

20-6
 Tab.1 Constants ,KR

loaiitwllal tranmrn
1 4 5. 01 4 3. 55 7
13 6. 298 5. 320
12 7. 329 6. 80 1
I 1 7. 949 8. 299
Ki 10 8. 180 9. 7'07
9 9: 108 11. 385
(I/O.) 1 10. 156 13. 107
7 10. 897 IS. 368
6 12. 198 17. 801
5 13. 132 20. 679
4 14. 105 24. 976
3 IS. 503 30. 513
2 18. 195 40. 495
I 28. 171 43. 865
"K (l/u) 1. 8S9 3. 068
0.1 O.I 0.1 0.4 Oi
ECI itt Uo/r) 8. 156X10" 0. 185X10"

Fig.7 Design Story Sheer Force Coefficients,

20-7
 EXPERIMENTAL FACILITIES IN THE UNITED STATES FOR
SOIL-STRUCTURE INTERACTION RESEARCH
By Scott A. Ashford1

ABSTRACT: This paper provides a brief summary of experimental facilities in the United States that can
be used for Soil-Structure Interaction (SSI) research. The objective is to provide the participants of this US-
Japan SSI Workshop with a description of existing and upcoming facilities in order to promote discussions
on future collaborations at the workshop. The paper focuses on dynamic soil-structure interaction, though
some facilities are included that are without dynamic capabilities. The first three sections discuss centrifuge
testing, 1-g shake table testing, and full scale testing. A brief description of the National Geotechnical
Experimental Site program is included within the full scale testing section. The final section introduces the
NHPS concept in light of the review of existing facilities.

INTRODUCTION capacities an order of magnitude larger than the rest: the 1256
g-ton centrifuge at the US Army Corps of Engineers
The purpose of this paper is to provide a brief summary of Waterways Experiment Station in Vicksburg, Mississippi; the
experimental facilities in the United States that can be used for 440 g-ton centrifuge at the University of Colorado, Boulder;
Soil-Structure Interaction (SSI) research. The objective is to the 400 g-ton centrifuge at the University of California, Davis,
provide the participants of this US-Japan SSI Workshop with a and the 100 g-ton centrifuge at Rensselaer Polytechnic
description of existing and upcoming facilities in order to Institute in Troy, New York. The other ten centrifuges located
promote discussions on future collaborations at the workshop. in the United States have a capacity of between 5 and 15 g-
Contact information is provided for most facilities to direct the tons. The four medium and large centrifuges are discussed in
participants to further information if they are interested. more detail below.
Though I have tried to include most of the larger facilities, and
a sampling of smaller facilities, I am certain that some
valuable facilities have been inadvertently left out of this TABLE 1. Description of US Centrifuges.
paper. Much of the information contained in this paper is the Max.
result of a request through the United States Universities for Centrifugal Max. Max.
Geotechnical Engineering Research (USUGER) e-mail Acceleration Payload Capacity
membership list. In addition, their have been several reviews Location (g) (Ibs/kg) (g-ton)
of earthquake engineering research facilities in recent years, (1) (2) (3) (4)
most notably the Assessment of Earthquake Engineering USACEWES 350 8800/4000 1256
Research and Testing Capabilities in the United States (EERI, CU, Boulder 200 4400/2000 440
1995) which is still available to interested workshop UC, Davis 100 8000/3640 400
participants from EERI. This critical review of our existing
capabilities has led to much discussion among the earthquake Boeing 600 400/180 120
engineering community on the urgent need for a National RPI 200 1000/454 100
Network of High Performance Seismic Simulation (NHPS), CU, Boulder 100 300/136 15
which will be discussed in more detail later in this paper. Tyndall AFB 100 300/136 15
This paper focuses on dynamic soil-structure interaction, 15
U. of Maryland 200 150/68
though some facilities are included that are without dynamic
capabilities. The paper is separated into 4 parts. The first three MIT 200 150/68 15
sections discuss centrifuge testing, 1-g shake table testing, and UC, Davis 100 200/91 10
full scale testing. A brief description of the National UC, San Diego 100 200/91 10
Geotechnical Experimental Site program is included within U. of Florida 120 125/57 7.5
the full scale testing section. The final section introduces the
Caltech 175 80/36 7.5
NHPS concept in light of the review of existing facilities.
U. of Florida 100 100/45 5
CENTRIFUGE FACILITES
The centrifuge at the US Army Corps of Engineers
There are currently 14 geotechnical centrifuges located in Waterways Experiment Station has the largest capacity of
the United States as shown in Table 1 (Dobry and Elgamal, centrifuges in the United States. This new facility (opened in
1989; Marcuson et al, 1995). Eight of these have shakers that 1995) has a radius of 6.5m, a centrifugal acceleration range of
can be used for earthquake testing. The largest four have

1 Assistant Professor of Geotechnical Engineering, Division of Structural Engineering, University of California at San Diego, 9500
Gilman Drive, La Jolla, CA 92093-0085. Telephone: 619-822-0431. Fax: 619-822-2260. E-mail: [email protected]
21-1
10 to 350g's. The maximum capacity of 1256g-tons is reached SHAKE TABLES
at 143g's. Though primarily used for geotechnical research, it
is also used extensively for coastal, cold region, There are currently 6 "medium" size shake tables in
environmental, hydraulics, and structural applications. In operation in the United States, located at the University of
addition to modeling earthquake loading for soil-structure California at Berkeley (6.0m x 6.0m), the State University of
interaction experiments, it is also has the capability to model New York at Buffalo (3.7m x 3.7m), the University of Illinois
blast loading, frozen environments, and low pressure at Urbana-Champaign (3.7m x 3.7m), the US Army
environments (even vacuums that would exist on other Construction Engineering Research Laboratory (3.7m x 3.7m),
planets). More information can be obtained at the the University of Nevada, Reno (3.7m x 3.7m), and the
http://www.wes.anny.mil/centrifuge. University of California, San Diego (4.9m x 3.0m). These are
The University of Colorado at Boulder has two centrifuges "medium" size table compared to Japanese standards, and
in their Geotechnical Engineering Laboratory, the largest have payload capacities ranging from 6,800 to 54,500kg. Two
being a 400 g-ton capacity Wyle centrifuge capable of additional tables will be in operation shortly: the 3.7m x 3.7m
dynamic simulation. This centrifuge is can carry a model 1m table at the University of California, Irvine, and the 4.9m x
in dimension. More information can be obtained at 3.7m table associated with the Seismic Retrofit Modification
http://bechtel.colorado.edu/Labs/Centrifuge.hrml. Device (SRMD) testing system at the UC San Diego. The 6
The University of California, Davis, also has 2 centrifuges. degree-of-freedom SRMD table at UCSD will be one of the
The effort to install the largest of the two centrifuges started in most powerful shake tables in the US, with a longitudinal
1978, with the centrifuge being operational at UC Davis in force capacity of nearly 9,OOOkN, a longitudinal stroke of
1988. This centrifuge currently has a 9-m radius, a maximum 1.2m, and a maximum velocity of 1.8m/s. In addition, there
payload of 4500 kg and an available bucket area of 4m2. This are over 10 additional small tables distributed around the US.
centrifuge is limited to 53g acceleration due to limited drive Of the medium sized table, one has a flexible container for
torque, but has the potential to nearly double it's capable conducting soil-foundation-structure interaction experiments
acceleration. Currently 64 channels of data acquisition are (UC Berkeley), and two have containers planned (UN Reno
available with sampling at 5000Hz. The shaker, installed in and UC San Diego).
1995, is capable of accelerating a rigid mass of 2700kg up to The container at UC Berkeley was designed for soil-
14g's, but for flexible models accelerations up to 30g's can be foundation-structure interaction experimentation in cohesive
obtained. For this shaker, 2 flexible shear beams and a rigid soils. The container consists of a thin-walled cylindrical
container are available with plan dimensions of well over 1 rubber bag, reinforced with aramid fibers placed as horizontal
meter. A great deal of additional information can be found at hoops as well as in independent vertical strips. The container
http://www.engr.ucdavis.edu/%7Ecgni/FACILITIES.html, the is approximately 2.4m in diameter and 2.1m high. The top and
website for UC Davis' Center for Geoteclinical Modeling. bottom of the bag are connected to rigid rings, supported by
The 100 g-ton centrifuge at Rennesslaer Polytechnic eight axial load-bearing columns with ball-and-socket joints at
Institute is the only medium sized centrifuge in the US, and is the rings. This container, when shaken on the UC Berkeley
the largest on the eastern coast of the US. The nominal radius shake table can provide appropriately scaled response of up to
of the centrifuge is 2.5 m, and has a container of 0.8 m in plan 0.8g and can accommodate two-directional shaking (Reimer,
dimension. A 0.5 ton payload can reach a centrifugal 1996).
acceleration of 200g's and a 1 ton payload lOOg's.

Figure 2: UC Berkeley's SSI Testing Container

Figure 1: Centrifuge at US Army Corps of Engineers The University of California, San Diego, is planning to
Waterways Experiment Station. construct a laminar soil box for soil-foundation-structure
interaction testing of bridge columns. Though currently in the
planning stages, this laminar box will be similar in design to
the large laminar box at the NIED in Tskuba, Japan. It will be

21-2
used on both the medium-sized shake table at UC San Diego
as well as the larger SRMD shaking table under construction.

Figure 3: UCSD's Mobile Structures Testing Laboratory.

The University of California, Davis, has constructed an in-

ground large diameter soil container for investigating plastic
hinge formation in reinforced concrete piles. The corrugated
Figure 2: UCSD 's Seismic Retrofit Modification Device steel container has a 6.7m diameter and is 5.5m deep. A large
capacity concrete block (1.4m xl.5m x 6.1m) is used to
The University of Nevada, Reno, is currently constructing provide lateral reaction for the test pile, and axial loads are
a laminar soil box for soil-foundation-structure interaction provided by hydraulic jacks and tie rods connected to a strong
testing in sands. The box will have plan dimensions of 3,2m x beam at the base of die container. These features allow for the
3.2m and a height of 1.9m. The walls of the box will be investigation of combined axial/lateral loading of the piles.
composed of alternating layers of aluminum and neoprene. In More information can be found in Chai and Hutchinson
order to simulate overburden pressure in the granular soils to (1998).
be tested, the laminar box is equipped with a scalable top plate A full scale column/shaft is being installed for testing at
to allow the box to be pressurized up to 30 psi (Sherif Elfass, the University of California, Los Angeles. The column/shaft,
personal communication). 2.4m in diameter, 21m deep and 15m high, is heavily
instrumented with soil pressure meters, inclinometers,
FULL SCALE TESTING FACILITIES extensometers, strain gauges, displacement sensors, gamma
ray sensors, as well as sensors for time domain reflectometry.
There is a move in SSI research in the United States The column/shaft will be subjected to a series of slow cyclic
towards full scale testing and instrumentation of full scale lateral load tests, with loads being applied to the top of the
structures in high seismic areas. Many of the full scale tests column through cables connected to ground reactions. More
that take place are "one-of-a-kind" due to the high cost and information can be found in Janoyan and Selna (1998).
space requirements. In this section, a few of the full scale test Other examples of large scale testing facilities include a
facilities that do have some opportunity for repeated use are 6.7m diameter, infinitely deep sand box -at the University of
discussed. In addition, special opportunities for cooperative Michigan. The box is located adjacent to the structural
full scale testing, including the National Geotechnical engineering reaction wall to allow testing with a variety of
Experimentation Sites program, are presented. hydraulic actuators and aboveground structures (R. Woods,
The University of California, San Diego, is developing a personal communication). Another example is the Louisiana
Mobile Structures Testing Laboratory for conducting full scale Transportation Research Center Test Wall at Louisiana State
lateral load testing of deep foundations. It will also be used for University (M. Tumay, personal communication). This 47m x
monitoring behavior of full scale bridge superstructures. The 16m x 6m high embankment is being used to test 7 different
mobile laboratory will consist of a 196-channel high-speed types of slope reinforcement and retaining wall systems,
data acquisition system capable of sampling rates of SOOHz, a including a large diameter culvert. On a smaller scale, Utah
2000-kN hydraulic actuator with a portable hydraulic power State University has developed a soil container for conducting
supply capable of driving the actuator at velocities greater than lateral load testing of piles and pile groups in their
50 mm/sec, and a digital controller. The laboratory is being geotechnical laboratory. The steel ribbed container measures
used in Fall of 1998 for conducting lateral load testing of over 3.05m long by 0.91m wide by 1.5m deep, and is being
drilled piers up to 2m in diameter at the UCSD campus, and used to study cyclic loading of piles in soft clay (J. Caliendo,
also lateral load testing of drilled piers and driven pile groups personal communication).
sand liquefied by controlled blasting at the Treasure Island
NGES. More information can be found in Ashford and The NeTI Ground Motion Generation Project
Elgamal(1998).
An innovative SSI project is underway by the Nevada
Testing Institute (NeTI), a non-profit Nevada corporation (P.
Mote, personal communication). NeTI has developed a
consortium of university and non-university research
21-3
institutions to design, test, and construct a ground motion test University of Houston Foundation Test Facility (Houston
bed at the Nevada Test Site. The site will be used for large- Texas).
and full scale soil-foundation-structure interaction testing The theme research areas that were deemed a priority for
under earthquake loading. Earthquake excitation for the site the NGES Program are geotechnical earthquake engineering
will utilize the Repeatable Earth Shaking by Controlled (liquefaction, site amplification, and permanent deformations),
Underground Expansion (RESCUE) techniques. Using these calibration of new equipment, proof-testing of site
techniques, ground motions are generated by rapidly improvement techniques, geo-environmental problems,
expanding a buried neoprene rubber bladder using gasses expansive clay problems, and foundation prototype testing.
generated by timed propellant ignitions. Information for each of the Level I and II sites is available to
Successful preliminary testing at a site in California was potential users, including information on soil type,
completed some time ago, and Phase 1 testing at the Nevada stratigraphy, previous test results, as well as results of
Test Site are ongoing. Tests have been completed on a 1.5m previous research efforts conducted at the sites. Additional
by 6m test bed, with source/soil test pressures on the order of information can be obtained at the NGES website
550 kPa. Ultimately, up to a 60m by 60m test bed is http://unhinfo.unh.edu/nges/.
anticipated, with the capability to repeatably produce ground
displacements of 500 mm and accelerations as high as 4g's. NATIONAL NETWORK FOR HIGH PERFORMANCE
Universities participating in the initial testing include the SEISMIC SIMULATION
University of Nevada, Las Vegas; University of Nevada,
Reno; University of California, Irvine; as well as UC's Over the last 25 years, there have been at least 8
Lawrence Livermore National Laboratoiy. More information assessments of the experimental facilities in the US for
can be obtained by contacting the Nevada Testing Institute by conducting earthquake engineering related research (EERI,
fax at (702) 895-0512. 1995), the most recent being the Assessment of Earthquake
Engineering Research and Testing Capabilities in the United
States prepared by EERI in 1995 as required by law in the
1994 reauthorization of the National Earthquake Hazard
Reduction Program (NEHRP). The assessment project was
sponsored by NSF and the National Institute of Standards and
Technology (NIST). This report concluded that the that the
state of research and testing capabilities had not kept up with
the needs of the greater community in reducing losses from
earthquakes. Among the recommendations contained in the
report were to upgrade existing experimental facilities, to
develop new moderately-sized regional centers with unique
and complementary facilities, to study the feasibility of a
single national test facility, and to continue and expand
programs where facilities and research results can shared
among many users (including international users).
An outgrowth of this latest assessment is the proposed
National Network for High Performance Seismic Simulation
(NHPS). The NHPS has been proposed within NSF as a Major
Figur 4: Preliminary test site for NeTI Research Equipment (MRE) initiative in order to develop
integrated experimental research facilities to improve of
fundamental understanding of earthquakes and help avoid
National Geotechnical Experimentation Sites Program catastrophes caused by lack of knowledge of the behavior
engineering materials, soils, and construction during
The National Geotechnical Experimentation Sites (NGES) earthquakes. The intent of the NHPS is to develop a network
Program is a system of multiple user, well-documented of research facilities open to everyone that would provide a
geotechnical test sites with easy access. The NGES Program means for rapid exchange of ideas and information and offer
was funded by the National Science Foundation and the immediate access to research data generated anywhere in the
Federal Highways Administration. The NGES Program also network.
has a central repository of geotechnical data collected at each A draft report from an NHPS workshop held in June 1998
NGES to promote exchange of information and reduce at UC Davis (http://www.engr.ucdavis.edu/%7Ecgrn/NHPS/)
research costs. states that the NHPS would include three major categories of
In all, over 40 sites have been designated in the NGES new and enhanced equipment:
Program. Of these, 5 have been classified as Level I and II
sites; site that meet most or all of the research needs identified 1. Advanced-design earthquake simulation facilities
in national workshops, as well as having favorable site including simulation systems for structural testing,
conditions. These site are Treasure Island Naval Station (San geotechnical centrifuges with shakers, and wave
Francisco, California), Texas A&M University Riverside generators for the study of tsunamis.
Campus (College Station, Texas), Northwestern University 2. Large scale testing systems for testing structural
Lake Fill Site (Evanston, Illinois), University of elements, assemblies of elements, and response
Massachusetts-Amherst, Amherst, Massachusetts), and modification devices at full scale.
21-4
 3. Field simulation and laboratories including field testing
installations, mobile laboratory units for monitoring
constructed facilities before, during and after an
earthquake.

The NHPS would greatly enhance our ability to conduct

SSI experimentation in the US. Some specific facilities hi the
NHPS include development of new instrumented field sites,
development of high-energy mobile seismic wave forces,
development of sites for ground motion simulation with high
explosives, as well as the upgrading of existing centrifuge
shaker capabilities. Much more detail on the development of
the proposed NHPS is available at
http://www.engr.ucdavis.edu/%7Ecgm/NHPS/.

REFERENCES

Chai, Y. H., and Hutchinson, T. C. (1998). "Revereved Cyclic

Response of Reinforced Concrete Piles in Cohesionless
Soil," Proceedings, The 5th Caltrans Seismic Research
Workshop, June 16-18, Sacramento, CA.
Dobry, R. and Elgamal, A.-W. (1989). "The New RPI
Geotechnical Centrifuge," Geotechnical Centrifuge
Newsletter, Cambridge, England, April 1989.
EERI (1995). Assessment of Earthquake Engineering
Research and Testing Capabilities in the United States.
Earthquake Engineering Research Institute Publication No.
WP-01, September 1995, 23 p.
Janoyan, K., and Selna, L. (1998). "Design and Construction
of Large Scale Column/Shaft Test Project," Proceedings,
The 5th Caltrans Seismic Research Workshop, June 16-18,
Sacramento, CA.
Marcuson III, W. F., Ledbetter, R. H., Green, R. A.,
Steedman, R. S., Franklin, A. G., and Hynes, M. E. (1995).
Problems in Geotechnical Earthquake Engineering that
Demand Experimental Test Facilities, Commissioned
Paper by EERI for Assessment of Earthquake Engineering
Research and Testing Capabilities in the United States.
Reimer, M. (1996). "Physical Modeling of Seismic Soil-Pile
Interaction," EERC News, Earthquake Engineering
Research Center, University of California, Berkeley, 17(2),
April 1996, pp. 1-3.

21-5
 DENSE INSTRUMENTATION IN BRI BUILDING AND SURROUNDING GROUND
BY

IZURU OKAWA", TOSHIHTOE KASHIMAZ) AND SHIN KoYAMA2)

ABSTRACT: The Building Research Institute (BRI) had conducted earthquake recordings mainly in buildings
since 1957. Several recorded motions from our earthquake recording are used for seismic design of buildings.
However, there had been regret that these motions recorded in buildings are contaminated with the vibratory
characteristics of building itself. The consideration of the dynamic property of building is important when
those motions are applied for other buildings. In the next step, we had started earthquake recording only in
ground of Sendai area in the northern part of Japan focussing on the effect of surface soil conditions on seismic
motions. Although we had obtained only records with small amplitudes, we recognized that the amplification
characteristics are varied depending on site-specific nature of the recording sites. In 1993, we had a big
earthquake in Hokkaido. Extremely large amplitude of 700 or 900 cm/s/s was obtained in the free field during
the earthquake. However, damage to buildings was minor. Since then, we had kept in mind what the
effective input motion to building is. The minor damage had been partly explained by the fact that most of the
buildings in Hokkaido keep large foundation with deep embedment because of its cold weather in winter. We
had devastating earthquake damage in Kobe. Considerable, not large, number of strong motion records had
been obtained during this earthquake. Comparing free field motions with those from inside of buildings, it
might be said that the former shows larger amplitudes than the latter. This difference might be partly caused by
the soil-structure or soil-foundation-structure interaction. With these in background, we had started other
earthquake recording systems after the Kobe earthquake. One is the strong motion seismograph array network
deployed around the greater Tokyo metropolitan area. The other is the dense instrumentation in BRI campus.
When we look the structural damage during the recent large earthquakes in Japan, the evaluation of effective
input motion to buildings are very important. In this paper, our attempt on the earthquake recordings is
introduce focussing in above matters.

INTRODUCTION believed to be earthquake-free area by residents around there.

We have instead obtained record in Osaka several tens km
With rapid progress and advances in recording technology away from Kobe.
and reduction in cost for installing the instruments, BRI seismograms had generally been installed in the
tremendous number of strong motion records have been buildings with some exceptions since its beginning.
obtained and released with recent large earthquakes in Japan. Considering the large amplitude recorded by the free field
Strong motion recordings were started with seismometers sites, we came up to bear idea in mind that the free field
installed in structures. The Building Research Institute ground motion is not identical to the input motion to the
(BRI) published many epoch-making records such as the structure. Here, we should define what is the input motion.
Kawagishi-cho apartment, for 1964 Niigata earthquake, That is the motion to be used for the analysis of superstructure
recording the tilting of the apartment building due to soil with the assumption of fixed-base model. It might be
liquefaction. This was the first recording of liquefaction appropriate to call it effective input motion to structure.
phenomena in the world. After that, we had also obtained With the terminology, it is said that we should seek for the
large earthquake record in the campus of Tohoku University, effective motion or how to distinguish effective motion from
during 1978 Miyagiken-oki earthquake. The earthquake recorded ones.
was recorded at the 1st and 9th floor of the building in the Recently we set up two kinds of earthquake recording systems.
campus. The amplitude in the first floor was large at the time. One is a system with 2 to 4 three component sensors in one
Attention was drawn from many researchers that the 9th floor site installing in first or basement floor, top floor and nearby
accelerogram showed amplitude more than 1G. We had also free filed. We installed seismometers at 20 sites around
large amplitude records during 1993 Kushiro-oki earthquake, Tokyo area. It is not easy; however, to find places for free
and 1994 Hokkaido Toho-oki earthquake and 1994 Sanriku field, since there are few free grounds with little influence of
Haruka-oki earthquake. Unfortunately, we had not installed buildings. The other is the dense instrumentation to
seismographs in Kobe. Because Kobe had long been building and surrounding ground including deep boreholes.

^ Head, Building Engineering Division, International Institute of Seismology and Earthquake Engineering, Building Research Institute,
Tatehara-1, Tsukuba-shi, Ibaraki-ken, 305-0802, Japan
2) Senior Research Engineer, International Institute of Seismology and Earthquake Engineering, Building Research Institute, Tatehara-1,
Tsukuba-shi, Ibaraki-ken, 305-0802, Japan
22-1
We set up this system in newly constructed 8-story steel awakened us again to the importance of disaster prevention
reinforced concrete (SRC) building next to the main building measures for large-scale urban areas. It is important to predict
in our institute campus. Totally 66 channels record the the probability of a future earthquake and its impact, and
event when certain level of motion is triggered. We hope the make as many preparations as possible in anticipation of such
recording system of this kind will be very directly utilized for an event. It is also very essential to grasp the damage
evaluating the SSI phenomena. condition immediately to put in effect the necessary
countermeasures. BRI has established twenty new observation
sites placed radially in the Tokyo metropolitan area. This
BRI EARTHQUAKE RECORDING ACTIVITIES
project aims to investigate the characteristics of the seismic
N?.ticnwk<i' Strong Motion Observation h motion affecting the whole Kanto Plain through observation
BRI has installed strong-motion instruments in major cities records. The system immediately collects information on the
throughout Japan. There are now 47 observation sites in seismic intensity at the time of an earthquake occurrence. The
operation using the digital strong-motion instrument. The site location is shown in Fig.5. We had tried to install
observation points are shown in Fig.l. The objects of seismometers both in the buildings and free field. However,
observation are mainly buildings, and the measuring point is it is very difficult to find free field point near the building
usually placed both on the top and in the foundations of the especially densely populated Tokyo area. The recording
building. We are increasing the free field measuring points. sites in the suburbs have both, and therefore comparative data
However, the objective points had long been within the for SSI problem can be obtained.
building. Most of the observation sites are connected to BRI
Dcnsu scisr oribtL-r 7 my oart'iLjuakc- roccrdhig in varbus soil
via telephone line in order to reduce maintenance work and to
conditions of Sendai
retrieve data directly and promptly.
We also installed seismographs in grounds. Eleven
This observation network has obtained many noteworthy
recording stations are deployed around Sendai area, Japan as
records. As was mentioned previously, the Kawagishi-cho
shown in Fig.6. The location and geology of observation
apartment record for 1964 Niigata earthquake, in Fig.2
sites are listed in Table 1. The recording stations were
recording the tilting of the apartment building due to soil
selected so that typical surface soil conditions of urban areas
liquefaction. This was the first recording of liquefaction
in Japan are contained. Sendai was severely damaged
phenomena in the world. The earthquake record in the
during the 1978 Miyagiken-oki earthquake. There is result
campus of Tohoku University, during 1978 Miyagiken-oki
of damage investigation. The damage distribution had clear
earthquake was also from BRI recording system. The
correlation with the soil conditions. Therefore, we will get
earthquake was recorded at the 1st and 9th floor of the
the difference of strong motion characteristics when large
building in the campus as shown in Fig.3. The amplitude in
earthquake database is completed. Each station has three
the first floor was larger one at the time. Attention was drawn
seismographs with three component sensors. They are
from many researchers that the 9th floor accelerogram
installed from surface to so-called engineering bedrock.
showed amplitude more than 1G. There had been time with
These records have provided data to evaluate the effect of
little significant ground motions. Then, we had also large
surface geology on seismic motions. Unfortunately,
amplitude records during 1993 Kushiro-oki earthquake, 1994
earthquake data that are large enough to examine the effect of
Hokkaido Toho-oki earthquake and 1994 Sanriku Haruka-oki
nonlinear behavior of soils are still very few in number. The
earthquake. During the 1993 Kushiro-oki (Off Kushiro)
specifications for recording instruments are listed in Table 2.
Earthquake, 711 gal was recorded as the peak acceleration on
In Table 3, the recorded earthquakes with moderate
the ground surface at Kushiro Local Meteorological
amplitudes are listed.
Observatory. Whereas, JMA seismometer installed in the
same observatory recorded 920 gal. After this earthquake, BRI DENSE SEISMOMETER ARRAY IN BUILDING AND
BRI installed additional seismometer under the ground of the GROUND
site. These records are shown in Fig.4. In 1994 Hokkaido
Toho-oki earthquake, simultaneous recordings in surface and p* 'ical position of t^e hstryne^tatioi i
underground were obtained. In the 1994 Sanriku-haruka- Our institute (BRI) is located with approximately 60
oki (Far off Sanriku) Earthquake, a large acceleration kilometers to the northeast from downtown Tokyo. The site
amplitude record was obtained in the building next to the is situated 30 meter above sea level on the diluvial heights
severely damaged old Hachinohe municipal office building. between Sakuragawa-river flowing into Kasumigaura Lake
and Kokaigawa-river, a branch flow in the greater
Stron3rMotio" I* stain uj it Network in the Mbtropol'rta- Area Tonegawa-river water system.
After 1995 Hyogoken-Nanbu (Kobe) earthquake, considerable The external views are shown in Photos 1 and 2. As is shown,
number of strong motion seismographs were installed the new building is connected with the main building. The
nationwide. Its typical example is the Kyoshin-Net (K-Net) geological investigation shows that clay and fine sand are
by National Research Institute of Earth Science and Disaster main contents up to 90 meters depth underground, inserting
Prevention, Science and Technology Agency (NTED). The sandy gravel. We confirmed a sandy gravel layer at 88-meter
Japan Meteorological Agency (JMA) and many local depth. Its shear wave velocity, however, is not obtained.
governments also deployed wide coverage of strong motion Transfer functions for shear waves between ground surface
seismograph installation nationwide as well as the STA. and depths of 42 meter, 68 meter and 88 meter are shown in
The 1995 Hyogo-ken-Nanbu Earthquake (Kobe Earthquake) Fig.7. We assume here that the shear wave velocity of the
22-2
layer beneath 88-meter depth is about 500 m/s. The figure
shows common predominant periods of 3-5 Hz and 9 Hz. And
the transfer function with up to 88-meter depths includes peak
at 1 Hz. This might be the peak relevant to the fundamental
frequency of the surface soils. In addition, the layer under 88
meter depth up to approximately 250 meter is called upper
Kazusa formation group, the underlying layer is called lower
Kazusa formation group, the Tertiary layers.

Building Characteristics
The outline of the building and ground is shown in Fig.8.
The building for seismometer installation is a newly
constructed as the Urban Disaster Prevention Research Center
building completed in March 1998. The building has 8 stories
with single basement floor. Total building area is
approximately 5,000m2 and supported by mat foundation on Photo 1. Distant View of the Earthquake Disaster
clayey layer of 8.2 meters underground. Prevention Research Center Building with the Main
Building of BRl in Tsukuba.
Recording; System an J Observation
The sensors are installed with 11 locations (33 channels) in
new building, 7 locations in surrounding ground, and 4
locations (12 channels) in main (older) building. Since the
installation was very recent, the records obtained to date are
very few. Here, a set of recorded motions of June 24th,
1998; Southwest of Ibaraki prefecture earthquake is
introduced. The depth of hypocenter was about 70 km and
its magnitude was estimated 5.1. In Fig.9, records of ground
surface levels are compared. In Fig.10, the acceleration
response spectra for these motions are shown. It is seen that
free field ground surface records are larger than that for inside
the building, in this case. In addition, comparisons are made
between waveforms and response spectra for the records from
underground. The results are shown in Figs. 11 and 12.

CONCLUSIONS
Photo 2 Front view of the building of Earthquake
The earthquake observation project with dense accelerometer Recording
array configuration has just started. High quality records will
be accumulated year by year. We are ready to make these data
open to the public via Internet, hoping the research of ground
motion prediction becomes more active and the seismic
design methodology for buildings is more upgraded in the
future.

REFERENCES

1) Building Research Institute, "Report on the Damage by

1978 Off-Miyagi Prefecture Earthquake" (in Japanese),
Research Report of BRI, No.86, pp.75-81, 1978.
2) T.Kashima et al., "Strong-Motion Observation Program in
the Metropolitan Area", (in Japanese) Proceedings of the
AIJ Annual Meeting, Vol.B-2, pp.275-276, Kinki, 1997
3) T.Kashima et.al., "Earthquake Motion Observation in and
around 8-Story SRC Building", (in Japanese) Proceedings
of the AIJ Annual Meeting, Vol.B-2, pp.213-214, Kyushu,
1998
4) I.Okawa et al., "Dense Array Observation and Analysis of
Strong Ground Motions at Sites with Different Geological
Conditions in Sendai", BRI Research Paper 139, 1994

22-3
 Table 1 Location and geology of observation sites and de pths of accelerometers
Latitude Longitude
Soil
Year
vs
Site name Abr. type' /(Hz) (m/s) Depth (m)
Miyagino MIYA 3815'24"N 14055'I6"E 2 2.40 1984 680 22 54
Nakano NAKA 3815'14"N 14100'26"E 3 1.28 1985 720 30 61
Tamagawa TAMA 3819'03"N 14100'34"E 1 7.50 1986 1400 2 11 33
Oridate GRID 3815'26"N 14048'39"E r 2.06 1987 1050 57 76
Tsutsujigaoka TSUT 3815'30"N 14053'36"E 2 2.08 1988 950 36 59
Tsurumaki TRMA 3815'38"N 14058'15"E 3 1.26 1988 660 25 79
Okino OKIN 3813'26"N 14055'05"E 3 2.14 1988 820 I 17 62
Shiromaru SHIR 38H'29"N 14054'53"E 2 2.70 1988 830 1 20 76
Tsurugaya TRGA 3817'16"N 140541 53"E 2 1.78 1988 1000 2 37 62
Nagamachi NAGA 3813'45"N I4053'01"E 2 1.44 1989 700 1 29 81
Arahama ARAH 3813'11"N 14059'00"E 3 1.22 1989 750 1 31 76
ft : Theoretical first predominant frequency, Vs: Shear wave velocity of lowermost layer.
* Soil types according to Japan Building Standard. I: hard, 2:medium, 3: soft.
** Although the soil condition for GRID site was presumed hard soil, it was confirmed by the soil survey
that the surface soils were so heavily weathered that the soil condition type might be assigned to be the
second classification.

Table 2 Specifications of the array observation

system Table 3 List of observed earthquakes
Instrument Specification h A
# Date Time M (km) (km) Max. Ace. /
Type: Tri-axial velocity-
Accelerometer feedback type 8608 1986/10/14 06:11 5.0 53 135 26.2 (NAKA) 3
Frequency range: 0.05 to 30 Hz for IG 8615 1986/12/01 5:15 6.0 51 108 29.6 (TAMA) 3
8701 1987/1/9 15:14 6.6 72 191 43.5 (NAKA) 4
Amplifier and Resolution: 16 bit
8702 1987/1/14 20:04 7.0 119 505 11.1 (NAKA) 3
Dynamic range: 96dB
AD Converter
Sampling rate: 1/100 or 1/200 sec. 8704 1987/1/21 8:36 5.5 50 112 48.8 (NAKA) 3
Pre-event Delay device: 1C memory 8708 1987/2/06 21:24 6.4 30 178 47.4 (NAKA) 3
Memory Delay time: 5 sec. 8709 1987/2/6 22:16 6.7 35 167 94.1 (NAKA) 4
Oscillator: Quartz with accuracy of 8717 1987/3/10 12:24 5.6 29 166 15. 7 (NAKA) 3
io-7, lo-8 8719 1987/4/7 9:41 6.6 37 136 74.5 (NAKA) 4
Clock
Precision: 0.01 sec. 8721 1987/4/17 4:23 6.1 45 151 45 J (NAKA) 3
Calibration by: NHK(JBC) 8724 1987/4/23 5:13 6.5 49 145 75.0 (NAKA) 3
Medium: Digital magnetic tape 8739 1987/9/4 13:55 5.8 42 185 25.4{NAKAL 3
Digital Data with 9 track, half-inch 8740 1987/10/4 19:27 5.8 51 125 60. 1 (NAKA) j
Recorder in width and I600BPI 8911 1989/4/28 0:27 4.9 53 98 29.0 (OKIN) 3
in recording density 8915 1989/6/24 4:59 4.1 14 11 35.0 (TRMA) 3
8926 1989/11/02 3:26 7.1 0 256 22.8 (TRMA) j
9217 1992/7/18 17:37 6.9 0 269 11.2 (TRMA) 3
9234 1992/12/18 1:21 5.9 34 160 4 1.0 (TRMA) 3
9305 1993/1/15 20:07 7.8 101 594 36.3 (TRMA) 3
9325 1993/11/11 9:06 5.5 36 154 18.5 (TRMA) 3
9327 1993/11/27 15:11 5.9 112 51 104.2 (NAKA) 4
9409 1994/8/14 18:06 6.0 42 136 45.3 (NAKA) j
9410 1994/8/16 19:09 6.0 22 154 22.0 (ARAH) j
9413 1994/10/4 22:24 8.1 23 806 59.7 (OKIN) 3
9414 1994/12/10 18:26 5.1 51 111 28.4 (NAKA) 3
9417 1994/12/28 21:20 7.5 0 343 35.3 (NAGA) 3
9502 1995/1/7 7:37 6.9 30 261 28.0 (OKIN) 3
9602 1996/2/17 0:23 6.5 6 178 106.4 (NAGA) 4
9604 1996/4/23 13:08 5.2 76 117 75.0 (OKIN) j
9701 1997/2/20 5:22 5.3 88 99 46.7 (SHIR) j
M: JMA (Japan Meteorological Agency) magnitude, hi focal
depth, A- averaged epicentral distance, /: JMA seismic
intensity at Sendai

22-4
 Observation Sites

f ""': Specific Observation Projects

Hirosaki

f
, Gendai Dense Strong-Motion Instrument
Array Observation (11 sites)
*"Sendai

Kanazawa

/ .:--*;ws:.-
- Maizurti ;: : ^ i.
Yonago^^ iV^ Ng ^
1 "^^-Strong-Motion Observation Network in
lama Tokyo Metropolitan Area (20 sites)

/
l Shingu

0 100 200 400 600km

Fig. 1 Locations of Strong Motion Observation Network of Building Research Institute

22-5
 200 -i 1000 n
500-
100- 0-
0- -500-
-1000-J
-100- 1 n
<8 -200 -J
15 20 25 30 35 40 45
\ time (s)
10 15 20 25 30 100-
time (s)
200- -55 -100-
100- I I I I I I
15 20 25 30 35 40 45
0-
time (s)
-100-
Fig.4 Recorded Motion on Ground at Kushiro BRI
-200- Recording Sit in Kushiro JMA Observatory during 1993
Kushiro-oki Earthquake
0 5 10 15 20 25 30
time (s)
Fig.2 Recorded Motion in Kawagishi-cho Residential
Apartment during 1964 Niigata Earthquake. The building was
titled due to liquefaction.

1000- 1000-
500- 500-
0 0
-500 H -500-
-1000 -1000-
I \ r
0 10 20 30 40 0 10 20 30 40
time (s) time (s)

Fig.3 Accelerograms recorded in Tohoku Univ. building during 1978.6.12 Miyagiken-oki earthquake
Upper left: 9th floor, EW, Upper right: 9th floor, NS, Lower left: 1 st floor, EW, Lower right: 1 st floor, NS

22-6
 Observation sites
Sites operated in
other project

Contours represent depth of the upper

boundary of the pre-Neogene systems 50km
(after Suzuki, 1996)

Fig. 5 Strong Motion Instrument Network in Tokyo Metropolitan Area

22-7
 " '' '. "
Contour line represents ..-
thickness .of Alluyjurn in rriejer "3

"3$
' .3
_38-lON]
' 3
. 1
i0;krm;

Fig. 6 Locations of Earthquake Recording in Underground of Sendai

68m Model
42m Model

Fig.7 Transfer Function of the S-wave in Surface Ground Fig.8 Cross-sectional configuration of recording points
(Surface/lncident*2) for dense instrument array in the Building Research
Institute Campus, Tsukuba

22-8
 30 40 50 20 30 40 50
time (s) time (s)
C/J U
-52 10-
1 0- JiiUiui*
"tW^ JtakJL^ktfhjaJUU^
o -10-
% -20-
I I I I
20 30 40 50 10 20 30 40 50
time (s) time (s)

-52 10-
j[ 0- wj*
o -10-
% -20-]

10 20 30 40 50
time (s)
40 c-V
g 10-
0-
o -10-
< -20
20 30 40 50 10 20 30 40 50
time (s) time (s)
Fig.9 Recorded Motions in 1st Floor, C01, B01, A01 from
the above. Fig.11 Recorded Motions in underground, GL-1m,
GL-14m, GL-43m, GL-89m from the above.
80-
80-i-
A01 EW
B01 EW
60- C01 EW - - - A14 EW
1FEEW 60- - - A43 EW~
- A89 EW
40-
40-

20--
20-

468 468
0.1 1 10
Period (s) 0.1 1
Fig. 10 Comparison of Ace. Response Spectra between Period (s)
Recorded Motions in the Ground Surface Level including
1st Floor inside the Building (EW component) Fig.12 Comparison of Ace. Response Spectra between
Recorded Motions in the Ground (EW component)
 PLANNING AND IMPLEMENTATION OF A SOIL-STRUCTURE
INTERACTION EXPERIMENT

.1
Mehmet elebi

ABSTRACT: The significant parameters in developing a soil-structure interaction experiment are

reviewed. These parameters were described within five major recommendations determined during
a workshop held in 1992. Recommendations of that workshop are presented. A project following
the guidelines is currently underway, and its current status is discussed.

1. INTRODUCTION Since 1978, during several workshops and

technical meetings, specific recommendations
The objectives of this paper are (a) to have been repeatedly made to instrument a
introduce the background information in building for soil-structure interaction studies
establishing a special purpose array in a (e.g. Lee, 1978; Iwan, 1978; Iwan 1981). As
seismically active region of the United States recently as November 4 5, 1991, during the
specifically for studies of the effect of NSF workshop on ^Experimental Needs for
soil-structure interaction, (b) to review and Geotechnical Earthquake Engineering," held
define the parameters and details of a soil- in Albuquerque, New Mexico, strong-motion
structure experiment, and (c ) to describe the instrumentation for soil-structure interaction
current state of implementation. was given a high priority (Higgins, 1992).
U.S. Geological Survey (a) Circular 947
In the past, during design/analysis processes describes a general SSI scheme (elebi et al,
of engineered structures, it was assumed that 1987) and (b) Circular 1079 spells out
the foundation of a structure was fixed to a priority recommendations for special purpose
rigid underlying medium. In the last four arrays including those that will facilitate soil-
decades, however, it has been recognized that foundation interaction studies (Page et al,
soil-structure interaction (SSI) alters the 1992).
response characteristics of a structural system.
In important engineered structures, detailed A workshop held in 1992 resulted in a set of
numerical and closed-form-solution methods recommendations (upon which this paper is
have been applied to SSI analyses. To date, based) to define the needs for and the
the strong-motion data from instrumented parameters essential for implementation of a
buildings are insufficient to confirm the soil-structure interaction experiment. During
validity of the soil-structure interaction that workshop, beneficial and adverse effects
analysis methods and procedures as applied to of soil-structure interaction were discussed
structures other than nuclear power plant
structures. Soil-structure interaction analysis
procedures are now included in various codes Research Civil Engineer, Earthquake
(e.g. ATC--3, NEHRP-1985). Hazards Team, U.S. Geological Survey,
Menlo Park, Ca., USA 94025

23-1
(elebi, Lysmer and Luco, 1992). Prior to carry out studies related to vertical spatial
this workshop, there have been no meetings variation of motions to calibrate convolution
that directly addressed the detailing of a and deconvolution processes and
soil-structure interaction experiment except applications. (The only building with a
the ones related to the critical structures of tri-axial downhole instrument is in Norwalk,
nuclear power industry [e.g. the Lotung California. However, the downhole
Array] (Tang, 1987, Tang et al, 1987a, instrument is within a caisson (of a cluster of
1987b, 1990, 1991). caissons) only 30 feet below the basement
level. Recent data shows that its motion is
2. MOTIVATION same as the basement of the building; elebi,
1993a and b). The cluster of caissons has
Although, currently, there are over 200 altered the soil condition by making it stiffer
instrumented structures in the United States, than it was. Therefore, the foundation and the
there is no instrumented structure that will caissons have very similar motions.
allow detailed calibration and/or confirmation
of the validity of the soil-structure interaction (c) There are no horizontal spatial arrays in
analysis methods. The significant sets of data the vicinity of a building to specifically study
acquired during the 1987 Whittier, 1989 free-field motions and how these motions are
Loma Prieta and 1994 Northridge altered by interaction with the foundation of a
earthquakes provide insight into structural building structure. Specific question as to at
responses and clearly show that soil-structure what distance from a building the ground
interaction took place in several instrumented motion is unaffected by the interaction of a
buildings; however, the data set is insufficient building has not yet been answered.
to calibrate soil-structure interaction methods
or to quantify the significant parameters 3. IDEAL SOIL-STRUCTURE
related to it. That is, to date, we do not have INTERACTION EXPERIMENTAL
strong-motion response data from SCHEME
instrumented structures complete enough to
carry out detailed studies of the methods and An ideal layout of arrays that includes
procedures used in soil-structure interaction soil-structure interaction instrumentation is
analyses, and, in turn, assess their impact on provided in Figures 1 and 2 (elebi et al,
design codes and related analysis procedures. 1977; elebi and Joyner, 1978). Such a
Examples of deficiencies in existing layout should have four main arrays:
instrumented building systems are as follows:
1. Superstructure array
(a) The strong-motion instrumented structures 2. Soil-structure interaction array
do not have pressure transducers and 3. Vertical Spatial array
accelerometers around the periphery of the 4. Horizontal Spatial array.
foundation system (1) to check the horizontal
and vertical dynamic pressures and the These arrays are depicted schematically in
variation of the forces, and (2) to quantify both Figures 1 and 2.
rocking and uplifting during strong-motion
events. 4. LOTUNG AND HUALBEN
EXPERIMENTS
(b) There are no downhole arrays below the
foundation or in the vicinity of a building to The most detailed soil-structure interaction

23-2
 (SSI) experiment to date was implemented in soil-structure interaction (SSI) effects for a
1985 by EPRI at Lotung. The purpose of the typical (and regular) building (hereinafter
Lotung experiment was to facilitate the study referred to as typical building) during
of SSI for a 1/4- and 1/12-scale, strong-motion earthquakes. This principal
reinforced-concrete, cylindrically-shaped recommendation is motivated by the fact that
nuclear power plant containment models there is still great uncertainty as to the
under strong ground motion earthquakes significance of seismic soil-structure
(EPRI, 1989; Tang , 1987 and Tang et al, interaction (SSI) for typical structures. There
1987a, b 1990). The Lotung experiments may be both beneficial and adverse effects of
provided insight into the SSI response of a soil-structure interaction. However, in many
very stiff structure (fixed-based frequency on cases, SSI is simply ignored in design without
the order of 7 10 Hz and SSI frequency of establishing whether it will increase or
2.7 Hz) on an extremely soft soil condition decrease the response of the structure. The
(shear wave velocity of the top layer between additional detailed recommendations to
300-1000 ft/sec. (100-330 m/s). The results follow provide guidelines for the design of an
of the Lotung experiment showed that the experiment, which, if activated by a strong
response of the structure was mainly in the earthquake, will remove some of the above
rocking mode (rigid-body rotation) and that uncertainties.
the SSI effect in structural deformation and
seismic wave spatial variation under stiffer It is necessary to consider what is currently
soil conditions were not addressed. To known about SSI effects and what can
remedy those shortcomings, another realistically be observed and analyzed by
experiment at a stiffer soil site, Hualien, has current methods. For example, it is known
been implemented (Tang et al, 1991). The that a major manifestation of SSI is a
shear wave velocity of the top layer at this site contribution to the rocking motion of the
is approximately 1200 ft./sec. (-400 m/s). structure and perhaps to local deformations of
Some of the lessons learned from the Lotung the foundation of the structure. Thus, the
experiment and from the instrumentation instrumentation should be designed to
schemes of both the Lotung and Hualien observe these effects. Observations which can
arrays can be used in the study of be checked against the results of numerical
soil-structure interaction for regular building calculations are much more valuable than
structures. However, the natural frequencies observations for which such comparisons
of the containment structures of both the cannot be made. Thus, the building, its
Lotung and Hualien experiments are much foundation system, and the site configuration
higher than those of regular buildings, the should be relatively simple thus the need
subject of the SSI experiment discussed for a typical and regular building.
herein.
The motivations for an SSI experiment can be
5. RECOMMENDATIONS OF THE itemized as:
1992 WORKSHOP
(a) To improve the state-of-the-art of
5.1 Recommendation 1: (Needs and formulations and procedures for the
Motivation) evaluation of SSI effects.

A field experiment should be implemented to (b) To provide a clear and useful guidance as
observe the structural behavior of and the to when SSI should be incorporated in the

23-3
analysis of a building, and, when necessary, upper 50 feet of the site.
how it should be done.
(c) The site should be level and essentially
(c) To check the accuracy of numerical horizontally layered. This is a critical
prediction of SSI and, in particular, of the requirement if observations are to be
rocking of the foundation since there is not compared with analytical results.
yet great confidence in specific numerical
predictions of the amount of rocking a (d) The site should not be liquefiable and
major contributor to SSI. should have a stable ground water level.

5.2 Recommendation 2: (Site Location and (e) A detailed site investigation should be
Soil Conditions) performed before the site is selected. The
investigation should include several borings
The test site should be located in an area with to establish stratigraphy, { in situ} shear-wave
relatively high seismicity, and should be velocity measurements, laboratory tests on
easily accessible for installation and undisturbed samples and ground water
maintenance of the instrumentation. observations.

The following areas are identified by the (f) Permanent open space around the building
USGS as having the highest earthquake must be ensured for long-term observation of
probabilities (WGCEP, 1988, 1990): free-field motions. This requirement is a
~must" and the chances of it being satisfied
(i) The San Francisco Bay Area [ Faults: San are probably highest if a public building is
Andreas, Hayward and Rogers Creek], chosen for the experiment.

(ii) Southern California (Upland, Redlands, 5.3 Recommendation 3: (Foundation)

San Bernardino) [Faults: San Jacinto and San
Andreas]. The foundation system of the candidate
structure should be as simple as possible and
In order for the SSI effects to be significant should not inherently minimize SSI effects.
the test site should be a soil site rather than a Thus:
rock site. Also, the geometry and ground
water conditions of the site should be (a) The preferred foundation type is a stiff
relatively simple such that the incident wave box or mat foundation. The contact surface
field can be well-defined and analyzed. This with the underlying soil should be
leads to the following recommendations: approximately plane.

(a) The site should not be too shallow. Rock (b) A 1- or 2-story basement is acceptable.
should be located at an appreciable depth However, the foundation system should not
(e.g. more than 50 feet below the foundation be fully compensated since this will tend to
level of the candidate structure). minimize the inertial SSI effects, one of the
effects that is desirable to observe. (A fully
(b) A firm alluvial site is preferable. Such a compensated foundation system is one for
site would consist of sands and gravels with which the weight of the displaced soil is equal
shear-wave velocities Vs in the range of to the weight of the entire structure including
500-1000 fps (-150-300 m/s) within the the basement).

23-4
(c) The initial experiment should exclude pile structure can be clearly defined and
supported structures. instrumentation can be more easily installed.
This is especially important if pressure cells
5.4 Recommendation 4: (Superstructure) or other instruments are to be installed on the
external basement walls or in the backfill.
It is preferable that a new building (before
construction starts) can be identified for 5.5 Recommendation 5: (Instrumentation)
instrumentation as part of the SSI experiment
rather than using an existing building. It is Several types of instrumentation should be
further recommended that the building (to be employed to record forces, motions and local
instrumented for an SSI experiment) have the deformations in the structure and the
following general characteristics: surrounding soil.

(a) The -geometry and load-carrying system 5.5.1. Superstructure Instrumentation:

of the structure should be as simple and
regular as possible. A building which is The main instrumentation in the
symmetric about two axes is preferable. The superstructure should be digital
design of the building should fall within the accelerometers with a common time base.
scope of current seismic design codes. It Enough instruments should be installed to
should also be amenable to accurate analysis. determine the translational, torsional and
rocking motions at least at three levels of the
(b) It is desirable that the structure have structure, including the base level and the top
different stiffnesses in its two principal floor. The exact location of the instruments
directions. However, the aspect ratio of its should be determined only after extensive
plan dimensions should not exceed 3 to 1 analytical response studies and ambient and
(preferably 2 to 1). Furthermore, to insure that forced vibration tests of the structure.
there is reasonable radiation damping, the Additional sensors should be installed within
building should not be too slender. On the the structure to measure story drifts and slab
other hand, The in-plan dimensions of the deformations at several levels.
building should not be large to cause
horizontal wave length interactions. 5.5.2. Foundation Instrumentation:

(c) The structure should not be too light, since In addition to accelerometers, other sensors
this would minimize SSI effects. A reinforced (linear variable displacement transducers
concrete structure or a steel structure with [LVDTj or other instruments) should be
concrete walls is preferable. installed to record local deformations of the
foundation system. This is especially
(d) The fixed-base natural period of the important if the foundation mat is flexible or
superstructure should be of the order 0.5 if shear walls are founded on independent
seconds. This corresponds to a 5- to 10-story foundations. It is also desirable to be able to
building, depending on the building type. record dynamic contact pressures on
basement walls and the foundation slab.
(e) If at all possible, a new, Unfortunately, currently available pressure
yet-to-be-constructed, building should be cells are not reliable for observations that
chosen. With access to the structure during extend over several years. Also, they are
construction, the load-carrying system of the virtually impossible to install in an existing

23-5
backfill. Direct recording of contact pressures downhole accelerometers have been
may therefore not be practical. It may, selected and purchased. The intent is to
however, be possible, and it is certainly deploy these immediately below the
desirable, to install rugged instruments that foundation of the building at least at two
can record wall/soil separation or foundation but preferably at three vertical locations.
uplift. In addition, at a distance away from the
building, another downhole array
5.5.3. Free-field Instrumentation: containing 2-3 downhole accelerograps
will be deployed.
A minimum of three boreholes should be
instrumented to record free-field motions. (b) Pressure Transducer Systems: In selecting
The boreholes should surround the pressure transducer system, consultations
instrumented building and should be located with technical staff of USGS and other
far enough away from all existing and institutions led to the concept of
planned structures to ensure that the records combinations of flatjack and differential
obtained are not contaminated by SSI effects. pressure transducer system (Kilgore,
However, the boreholes should not be so far Johnston and Warrick, personal
away from each other that incoherency effects communication, 1996). Figure 3 depicts a
destroy the coherency between the motions conceptual schematic of the deployment
observed in the different boreholes. At least of the flatjack and the differential pressure
three triaxial accelerometers should be transducer combination system. Several
installed in each borehole: at the surface, at flatjacks will be buried between sand
mid-depth, and at a depth deeper than the layers below the foundation system and
foundation level of the candidate building. If outside of the side walls. Each flatjack
the bedrock is within a depth of 300 feet will be connected to a valve inside the
(-100 m) an additional instrument should be building. The connection will lead from
installed at the soil/rock interface in each the valve to a differential pressure
boring. transducer (DPT) and a dummy flatjack.
Thus differential variation of the pressure
The surface instruments in the three borehole below the foundation and on the side
sets will double as a surface array. However, walls of the building will be realized.
it is recommended that additional surficial With the use of flatjacks, it will be
instruments be deployed closer to the building possible to record the average differential
to detect any changes in motion due to SSI pressure over a larger area than the
and/or due to the presence of the backfill. usually smaller area that pressure
transducer covers.
6. CURRENT STATUS
(c) Structural Array Hardware: Currently, we
6.1 Selection of Hardware plan to deploy only accelerometers
throughout the superstructure. However,
In selecting hardware, priority was given to laser technology allows deployment of
those that will be deployed below and in the displacement transducers although, at
periphery of the foundation and basement. present, these are very costly to acquire
These are: and deploy.

(a) Downhole accelerometers: Triaxial (d) Recording Systems: We intend to use a

23-6
 standard digital recording systems that of USGS, provided valuable input. Finally,
works on 2.5 volt signal. The DPT, the author thanks the officials of County ansd
accelerometers and downhole City of San Bernardino for their willingness
accelerographs work with this signal. to assist in developing the experiment.

6.2 Selection of Site 10. REFERENCES

We are in contact with the officials of City elebi, M., Lysmer, J., and Luco, E.,
and County of San Bernardino. These (compilers), 1992, Recommendations for a
officials will assist us in identifying a project Soil-Structure Interaction Experiment (Report
that is on the drawing table and meets our based on a workshop held at San Francisco,
requisite parameters. We expect this to occur Ca on February 7, 1992), USGS, OFR: 92--
within the next 12 months. 295, April 1992.

7. MANAGEMENT AND OTHER M., and Joyner, W., 1987,

BENEFITS OF THE EXPERIMENT Instrumentation for spatially varying ground
response integrated with structural response
When implemented, the experiment will be in a seismically active region, PROC., Joint
managed and maintained by the USGS IASPEI/IAEE Working Group on vvEffects of
National Strong Motion Program (NSMP). Surface Geology on Seismic Motion," IUGG
The data acquired through the experiment XDC General Assembly, Vancouver, B.C.,
will be open to all investigators. It is August 1987.
anticipated that the data will be used as key
research material related to soil-structure elebi, M., etal , 1987, Integrated
interaction methods. Future workshops may instrumentation plan for assessing the seismic
be held to discuss the data and related response of structures a review of the
researches. current USGS program, USGS Circular 947.

8. CONCLUSIONS
elebi, M., 1993a, Seismic response of two
This paper present requisite parameters for a adjacent buildings with downhole and free
soil-structure-interaction experiment. The -field recordings (Part I: Data and Analysis),
parameters were established during a 1992 Journal of Structural Division, American
workshop. Current status of the project is Society of Civil Engineers, v. 119, no. 8,
described. pp.2461-2476, August 1993.

9. ACKNOWLEDGMENTS elebi, M., 1993b, Seismic response of two

adjacent buildings with downhole and free
This paper is based on input from the -field recordings (Part II: Interaction),
participants of the 1992 workshop (USGS - Journal of Structural Division, American
Open File Report No. 92-295 [elebi, Lysmer Society of Civil Engineers.v. 119, no. 8, pp.
and Luco, 1992]). Guidance and advice 2477-2492, August 1993.
provided by J. Lysmer, E. Luco and J.
Roesset and others are appreciated. In EPRI, 1989, Proceedings: EPRI/NRC/TPC
designing the pressure transducers systems, Workshop on Seismic Soil-Structure
B. Kilgore, M. Johnston and R. Warrick, all Interaction Analysis Techniques Using Data

23-7
from Lotung, Taiwan, Volumes 1 and 2, seismic experiment and soil-structure
EPRINP-6154. interaction method validation, Nuclear
Engineering and Design, 123, 397 412.
Higgins, C. J., 1992 (editor), PROC. NSF
Workshop Experimental Needs for Tang, Y. K., etal , 1991, The Hualien
Geotechnical Earthquake Engineering, large-scale seismic test for soil-structure
Albuquerque, New Mexico, November 4 5, interaction research, SMIRT 11,
1991 (in preparation). Transactions, Vol. K, August 1991, 69-74.

Iwan, W. D., ed., 1978, Proceedings of the WGCEP (Working Group on California
International Workshop on Strong Motion Earthquake Probabilities), 1990, Probabilities
Instrument Arrays, May 1978, Honolulu, of Large Earthquakes in the San Francisco
Hawaii. Bay Region, California, U.S. Geological
Survey Circular 1053.
Iwan, W. D., ed., 1981, Proceedings of the
U.S. National Workshop on Strong-Motion WGCEP (Working Group on California
Earthquake Instrumentation, April 1981, Earthquake Probabilities), 1988, Probabilities
Santa Barbara, California. of Large Earthquakes Occurring in California
on the San Andreas Fault, USGS OFR:
Lee, K. L., W. R Marcuson, K. H. Stokoe, 88-398.
and F. Y. Yokel, editors, 1978, Research
needs and priorities for geotechnical
earthquake engineering applications,
Workshop at the University of Texas, Austin,
June 1978.

Page, R., Boore, D. M., Bucknam, R. C., and

Thatcher, W. R., 1992, Goals, Options, and
Priorities for the USGS Earthquake Hazards
Reduction Program: 1991-1995, USGS
Circular 1079, February 1992.

Tang, H. T., 1987, Large-scale soil-structure

interaction, EPRINP-5513-SR.

Tang, H. T., etal , 1987a, A large-scale soil

structure interaction experiment: Part
I Design and construction, SMIRT 9, Vol.
K2, 177-182.

Tang, Y. K., etal , 1987b, A large-scale

soil-structure interaction experiment: Part
E-EPRI/NRC Research Program on Method
Validation, SMIRT 9, Vol. K2, 183-188.

Tang, H. T., etal , 1990, Lotung large-scale

23-8
 ^stance to epfcenter Buj^in? Instrumentation
-^ _
Instrumentation nth floor
^
around perimeter (translatron* torsion)
of foundation ^ J
^,IHHJ
.
1 I-VUI
Horizontal Free \
CtransFation, torsion?
ip f

i foundation
Strong Motion f ""T^Q V x' <transJatfon torsion)
" 0 jc L
d i uuAiny
DEnse Arrays f %p
C ! C3 J. L -i

CD
~~ o ) u vertical ' ^L

specialized array o D spatial I1

<e*g liquef action> o d arrays (1

^^ /^ ^<\ , d
*
cf
p
^/^
> ,
f* V^C *^V j^O\. v' ff\>

Source /^ -^ >N: N
* (a) VERTfCAL CROSS SECTION V BE(fFO k

{
<>
<>
<>
<>
Horizontal Array
o o o o ^ Pran of
I 11 Foundation
PLAN VIEW OF HORIZONTAL FREE HELD ARRAY

Figure 1. General Description of an Integrated Soil-Structure Interaction Experiment

23-9
 Approximately
twice height
of the structure
^
-M
i
A
Vertical
acceleration
3-D
"* > 5 *
H^ ,r w acceleration

=L
B
r

BuE Wins. o Pressure

w iflv. B-^ Horizortt^E
^ 1 acceleration
Surficial Specific
/
/ Free Field array
CHorlzorttaD
i
Instrumentation
4
J ' O O p^ ,i
C

O i 4 it 4 o *
4fr
H * f
0
< O o ^

tt I

Vertioal spatial array

t' y
//

Figure 2. Details of Instrumentation in the Vicinity of the Building, Below and Around the
Foundation of Soil-Structure Interaction Experiment

23-10
 FFEREmiAL
FRESSLRE
TRANSPUTER

SANP APOVE &

PELOW FLATJACK

Figure 3. Flatjack and Differential Pressure Transducer Configuration around the Foundation of the
Building

23-11
Figure 4. Actual Flatjack (50 cm in diameter) and Differential Pressure Transducer

23-12
 EQUIVALENT 1-DOF MODEL FOR STRUCTURAL CONTROL
INCLUDING SOIL-STRUCTURE INTERACTION EFFECTS

J. Enrique Luco

Department of Applied Mechanics and Engineering Sciences

University of California, San Diego, La Jolla, California 92093-0411.

ABSTRACT

A simple model for the seismic response of a one-storey structure subjected to active
control in the presence of soil-structure interaction effects is presented. The approach is based
on the successive use of equivalent 1-DOF oscillators which account for the effects of control
and soil-structure interaction. Simple expressions for these oscillators based on exact analytical
solutions of the control equations and approximate solutions of the interaction equations are
presented. The study includes an evaluation of the effects of soil-structure interaction on the
seismic response of actively controlled structures in which the control gains have been determined
with and without inclusion of soil-structure interaction effects. A simple procedure to include
the interaction effects on the control gains is also presented.

24-1
 INTRODUCTION

In the last few years there has been increased interest in the study of the effects of
soil-structure interaction (SSI) on the response of structures subjected to active control [Wong
and Luco (1990, 1991, 1992), Sato et al (1991), Sato and Toki (1992), Alam and Baba (1991,
1993), Smith et al (1994), Wu and Smith (1995)]. These studies follow two lines of inquiry:
the first line is concerned with the evaluation of the effects of soil-structure interaction on the
response of structures with active control systems designed on the basis of conventional analyses
which do not include the interaction effects; the second line considers the incorporation of the
soil-structure interaction effects into the design of the control system and algorithms. These two
issues were discussed by Wong and Luco (1991, 1992) for a structural model consisting of a
shear beam supported on a rigid foundation embedded in a viscoelastic half-space. Control, in
this case, was achived by an absorbing boundary located at the top of the building. Recently,
Smith et al (1994) and Wu and Smith (1995) have considered the effects of SSI on the response
of a one-storey activelly-controlled structure supported on a rigid rectangular foundation resting
on an elastic half-space. The analysis was based on the use of external control forces obtained
by application of the linear optimal control theory to equations of motion which excluded or
included the SSI effects.

Here, we reconsider the problem studied by Smith et al (1994) of the effects of SSI on
the response of a one-storey structure subjected to active control. The approach differs from
that of the previous authors in that elementary methods are used to derive equivalent oscillators
which exactly account for the effects of control and approximately account for the effects of soil-
structure interaction. By combination of these solutions it is possible to obtain simple expressions
for the seismic response of the one-storey structure including the effects of control and SSI.

We start by summarizing approximate solutions of the interaction equations for a one-

storey structure supported on a rigid surface foundation resting on a viscoelastic half-space when
subjected to seismic excitation and to external and internal forces. The solution for seismic

24-2
excitation was initially presented by Bielak (1971) and Jennings and Bielak (1973) while that
for external forces was presented by Luco et al (1986, 1987). The well known result of these
approximate solutions is that the deformation of the one-storey structure including SSI effects
can be calculated from the response of an equivalent or replacement oscillator supported on a
rigid soil and characterized by a modified natural frequency and damping ratio. The equivalent
oscillator is subjected to an effective ground acceleration and to an effective force at the top of
the structure. The extension of the approximate solution to internal forces (or, more precisely,
to internally reacted forces) reveals that the effective force acting on the equivalent oscillator
depends on whether the actual forces correspond to internal or external forces. If the forces are
internal (such as those on a diagonal tendon system) then they affect the equation controlling
the deformation of the superstructure but do not appear in the global equations of motion of the
structure-foundation system. On the other hand, external forces appear in the equation reflecting
the deformation of the superstructure and in the global equations of motion for the system.

As a second preliminary step we consider the response of a damped 1-DOF system

subjected to active control in the absence of SSI effects. In particular, we consider the optimal
control of the free-vibration response of the oscillator for a quadratic performance index. An
analytical solution for the control gains is obtained by elementary means which do not require
consideration of Riccati's equation. In the particular case of an undamped oscillator, the resulting
gains coincide with those obtained by Meirovitch and 6z (1980) and Meirovitch and Baruh (1982)
by analytical solution of Riccati's steady-state equation. Clearly, the response of the oscillator
including the effects of control can be assimilated to that of an equivalent (uncontrolled) oscillator
with a modified natural frequency and damping. Simple expressions for the frequency and
damping ratio of the equivalent oscillator are presented.

In a third stage we consider the seismic response of a one-storey structure supported on

a flexible soil and subjected to active control by internal control forces with gains determined
without inclusion of SSI effects. The solution stans by replacing the one-storey superstructure by
a modified structure which exactly incorporates the effects of control. The modified one-storey
24-3
structure is then subjected to the seismic excitation in the presence of soil-structure interaction
effects and an approximate solution to the interaction equations is obtained on the basis of an
equivalent 1-DOF oscillator which includes both the effects of SSI and control.

Finally, a simple procedure to include the effects of soil-structure interaction on the deter-
mination of the control gains is presented. The procedure relies on representing the uncontrolled
structure including SSI effects by an equivalent 1-DOF structure on a rigid soil. The optimal
control gains for this equivalent structure are then obtained analytically. The original super-
structure is then represented by an equivalent one-storey structure which includes the effects of
control and the response of this modified structure to seismic excitation in the presence of SSI
effects is obtained by the approximate solution summarized in the first part of the study.

24-4
 EQUIVALENT OSCILLATOR FOR SOIL-STRUCTURE
INTERACTION EFFECTS

In here we summarize an approximate solution for the response of a structure subjected to

internal and external forces and to seismic excitation when the effects of soil-structure interaction
are included in the analysis. The case of seismic excitation was considered by Bielak (1971)
and Jennings and Bielak (1973) while the case of an external force at the top of the structure
was considered by Luco et al (1986, 1987).

We consider the in-plane vibrations of an elastic one-storey structure supported on a flat

rigid foundation resting on an elastic half-space. The structure is characterized by its mass mi,
stiffness ki t damping constant c\ and height h\. The foundation is characterized by its mass m0
and by its equivalent radius a. The underlying elastic half-space is determined by the density
p, the shear-wave velocity /?, the hysteretic damping ratio a and Poisson's ratio v. The system
is subjected to a vertically-incident plane SH-wave with total particle motion ug on the ground
surface in absence of the structure, i.e. for free-field conditions. With respect to forces acting on
the structure we consider two cases. In the first case, a horizontal external force FI is applied
at the top of the structure. In a second case, internal horizontal forces FI and F0 = FI are
applied at the top and base of the structure together with an internal moment M0 = F\h\
acting on the foundation. This self-balancing set of forces and moments may represent the case
of an internal control system such as a tendon system.

The global equations of motion for the structure-foundation system and the equation of
motion for the top mass for harmonic excitation with time dependence eta;t are

4- (mi

I00 i) 4- KR()O MR (2)

>i(u0 + hi60 4- iii) + ciui + (3)

24-5
in which u0 is the horizontal displacement of the foundation relative to the free-field ground
motion, 00 is the rocking rotation of the foundation and u\ is the deformation of the structure.
The terms KR and KR represent the frequency-dependent horizontal and rocking (complex)
impedance fuctions for the foundation and J0 is the sum of the moments of inertia of the
foundation and the top mass with respect to horizontal axes through their centroids. The terms
FR and MR correspond to the resultant force and the resultant moment (with respect to the base)
of all external forces (other than soil reactions) acting on the structure-foundation system. If only
an external force FI acts at the top of the structure, then FR = FI and MR = FI/II. For a system
of self-equilibrating internal forces and moments, FR = FI +F0 = 0 and MR = FI/II H-M0 = 0.
In Eqs. (1) and (2) the effects of the coupling impedances KHR = KRH have been neglected.

At this point, we introduce the notation

(4a)

(46)

KH =mio (1 + 2t[& + (wMy)fjf ]} (5a)

KR =/i?mic4 {1 + 2t[& + (wMO&i]} (56)

where

(6a)

,7 ,
(7a)
mi

(76)

24-6
 LJR = t (80)

-% - (86)
J

The terms u\ and fi correspond to the fixed-base natural frequency and the fixed-base damping
ratio for the structure, respectively. The frequencies WH and UJR correspond to the characteristic
frequencies for horizontal and rocking vibrations of a rigid structure on the flexible soil. The
damping ratios # and R reflect radiation damping in horizontal and rocking vibrations. The
hysteretic material damping in the soil is represented by the terms containing fs in Eqs. (5a) and
<5b).

By use of Eqs. (4) and (5), the equations of motion for the complete system can be
rewritten in the symmetric form

[e +w (9)

where e = 1 for the case of an external force FI acting at the top of the structure. In the case
of a self-equilibrating system of internal forces e = 0. The elements of the matrix [Z}(u;)] are
given by
2 . \

(10a)

(106)

33 = (lOc)

24-7
 (10d)

(lOe)

#23 =D32 = - (10/)

In writting Eq. (9), the terms m0 and J0 have been ignored when compared with mi and mi/if,
respectively.

The solution for the case of a rigid soil (i.e. in the absence of soil-structure interaction)
can be obtained by considering the limiting case UH oo and UR > oo. In this case,

= 00 = 0 (lla)

and

1- (116)
+

An approximate solution of Eq. (9) for the case of an external force FI acting at the top
of the structure (e = 1) can be obtained by neglecting certain terms involving 1, H and R
and assuming that the system frequency o>i is close to wi, /.e. assuming that the soil-structure
interaction effects are small. The resulting solution is given by

+
(12)

where
11
_ _ I 11
i'
-1 2 '
(13)

24-8
and
- \ 3
t*/i \ f (jj\
~ ) fl (14)
Q/1 /

Equations (13) and (14) give approximate expressions for the system frequency and for the
system damping ratio fi.

If we compare the form of the solution for the relative displacement u\ in Eq. (12) with
the solution for an oscillator on a rigid soil given by Eq. (lib), we observe that the relative
displacement response for an oscillator on a flexible soil can be calculated from the response of
an equivalent oscillator on a rigid soil with frequency o>i and damping ratio f i subjected to the
effective ground motion ug and to the effective force at the top FI given, respectively, by
2
Ua U, (15a)

and
Fi = -^ (156)

In the case of a system of internal forces (Fi at the top, F0 = JFi at the base, and
M = the parameter e = 0 and the approximate solution is given by

____ I
\
Of
aa '
_l_
Fi
(16)

in which cDi and fi are also given by Eqs. (13) and (14). In this case of internal forces (e = 0),
the equivalent oscillator on a rigid soil is subjected to the effective ground motion ug and to the
effective force at the top FI given, respectively, by
/- \ 2
ug = [ I UJ 1\
Ug (I7a)
/T7 \

and
(176)

24-9
It must be noted that the equivalent force at the top FI given by Eq. (17b) differs from that
given by Eq. (15b).

For later use, we note that the impedance functions KH and KR are usually normalized
in the form

=pj32 a(kn + iooCn) (ISa)

KR =p(32 a3 (kR + ia0cR) (186)

where a is the equivalent radius of the foundation, a0 = a;a//3 is a dimensionless frequency,

kn and kR are the normalized horizontal and rocking stiffness coefficients, and CH and CR are
the normalized horizontal and rocking damping coefficients. In general, the coefficients fc#, c#,
kR and CR depend on OQ, the soil damping ratio s and Poisson's ratio v, To determine o>i and
|i by use of Eqs. (13) and (14) these coefficients should be calculated at a0 = o>ia//3 and,
consequently, an iterative process is required to determine o>i.

Substitution from Eqs. (16) into Eqs. (7) and (8) leads to

(mi/pa?) (19a)
0 I

. OC !. /
(196)

\ P \aV kR

# = "" ~s"jt/~o (2Q6j

where (mi/pa3) is the mass ratio and /ii/a is the slendemess ratio.

The variations of the system frequency o>i normalized by the fixed-base frequency u\
and of the system damping ratio l\ as a function of the relative stiffness parameter ai = u)ia/(3

24-10
are shown in Fig. 2 for two values of the slenderness parameter h\/a. The results are based on
a structure with a mass ratio mi/pa3 = 1.0 and with a fixed-base damping ratio 1 = 0.02. The
circular foundation of radius a rests on a viscoelastic half-space characterized by the shear-wave
velocity (3, Poisson's ratio v 1/3 and material damping ratio fs = 0.02 (it was assumed that
the damping ratios for P- and S-waves were equal to fs). The impedance functions for a circular
foundation resting on an elastic half-space presented by Luco and Mita (1987) were used in the
analysis after modification to account for the effects of material damping.

The information in Fig. 2 reflects the well known results that the system frequency is
reduced and the system damping is generally increased as the soil becomes softer, i.e. as ai
increases. The reduction of the system frequency is larger for the more slender structure while
the increase in the system damping is smaller. (Indeed, if fs 0, the system damping ratio ?i
may be smaller than 1 if the structure is very slender).

24-11
 OPTIMAL CONTROL OF A SINGLE-DEGREE-OF-FREEDOM OSCILLATOR

We consider the optimal control of the free vibrations of a single-degree-of-freedom

oscillator characterized by the mass mi, the frequency u\ and by the damping ratio 1. The
equation of motion of the oscillator subjected to the control force F\(t) is given by

4- u\m = Fi(t)/mi (21)

with appropriate initial conditions on ui(0) and iti(O). The control force is selected in such a
way that the quadratic performance index J, defined as

r[Fi(t)/mi] 2 } dt (22)
o
is minimized subject to the constraining Eq. (21). In Eq. (22), 71, 72 and r are weighting factors.
In particular, if 71 = 72 then the first two terms in the integrand are proportional to the sum
of the kinetic and strain energy of the oscillator. The factor r is a measure of the cost of the
control force.

Next, we write the control force F\(t) in the form

FI (t) = -mi [ulgui 4- uihui] (23)

where g and h are the normalized gains. Substitution from Eq. (23) into the performance index
leads to
J= i I u\dt +u;?(72+/fy2) f u\dt - ^f3ghu\(G) (24)
0 0

in which /3 = rcj2 .

Also, substitution from Eq. (23) into Eq. (21) allows us to write the equation of motion
for the controlled oscillator in the form

w2cwi = 0 (25)

24-12
which corresponds to the equation of motion for free-vibrations of an equivalent oscillator of
mass mic = mi, frequency
(26)

and damping ratio

(27)

Multiplying Eq. (25) by ui(t) and Ui(t) and integrating leads to

CO

ju\dt= [u\(0) + w?X (28)

and
CO CO

u^dt= I u\ dt + >l (29)

0 LO
which upon substitution into Eq. (24) results in the following expression for the performance
index

(0)^(0) + -

(30)

where
G(g) = (31)

The optimal values for the gains g and h which minimize J(p, h) are then obtained by
setting the partial derivatives dJ/dg and dJ/dh equal to zero. It is found that the optimal gains
satisfy the conditions
rlH 9\Rn ^v J- fin 2
=0 (32)
dg 1+g (1+0)2

24-13
and
*Yl ' PM ' ^-*\9) m.
2(2g1+ &) /?" = f\ ' ^oo\
(33)

The resulting optimal gains are given by

(34)

h= 2 + a7i + 2</l + a72 - 2 - % (35)

where a = 1//3 = l/^Jr) is the control parameter (The uncontrolled case corresponds to
a = 0). For the special case of 71 = 72 = 1 and f i = 0 the present results coincide with
those obtained by Meirovitch and 6z (1980) and Meirovitch and Baruh (1982) by a procedure
involving the solution of a Riccati equation.

The value of the performance index for the optimal gains is given by

J = (hul(Q) + 2pwnn(0)ui(0) + [h + $(2& + h)] w?u?(0)} /aui (36)

which can be confirmed to be a positive definite form of iti(O) and iti(O).

Finally, substitution from Eqs. (34) and (35) into Eqs. (26) and (27) leads to the following
expressions for the frequency cjlc and damping ratio f ic of the equivalent oscillator

(37)
(1-ori/2)^ 1/2
(38)
\/l + #72

Some of the characteristics of the equivalent oscillator defined by Eqs. (37) and (38) need
to be stated here:

(i) in the uncontrolled case (a = 0) cjic = wi and ic = i>

(ii) the equivalent frequency u\ c is independent of 71 the weighting factor for the velocity

24-14
 (iii) if 72 = 0, i.e. if the displacement ui is not included in the performance index, then
= ^i and only the effective damping ratio changes to fic = [fj -4- (CK71/4)] 1/2, and

(iv) if 71 = 0, i.e. if the velocity ui does not appear in the performance index, then ic
l/\/2 as a oo (i.e. as the cost r of control tends to zero).

The variations of UIC/MI and f ic against the control parameter a = \/(LJ\T) are shown
in Fig. 3 for different values of the weighting factors 71 and 72. The results in Fig. 3b correspond
to the case 1 = 0. The controlled frequency LJIC remains unchanged (for 72 = 0) or increases
slightly with a while the equivalent damping ratio LC increases strongly with a. For example,
for 71 = 72 = 1.0, ?i = 0 and a = 2, the frequency u\c increases to 1.32o;i while the damping
ratio fic becomes equal to 0.707.

24-15
 RESPONSE OF A CONTROLLED 1-DOF OSCILLATOR WITH GAINS
DETERMINED WITHOUT SSI EFFECTS

In this section we consider the effects of soil-structure interaction on the seismic response
of an activelly-controlled structure in which the gains have been determined without consideration
of the interaction effects. For this purpose we consider a one-storey structure characterized by
the parameters mi, u?i, fi and hi and subjected to an internal control force Fi(t) acting on the
top mass. It is assumed that the control force is given by Eq. (23) with the control gains g and
h determined by minimizing the performance index given by Eq. (22) for free-vibrations of the
structure with the foundation kept fixed. The resulting gains are given by Eqs. (34) and (35).

When the effects of soil-structure interaction are included, the equation of motion for the
top mass is given by Eq. (3) in which the control force FI is given by Eq. (23). The resulting
equation of motion for harmonic vibrations can be written in the form

m\ug (39)

in which ujic and fic are the effective controlled frequency and damping ratio given by Eqs. (37)
and (38), respectively.

The global equations of motion for the structure-foundation system are given by Eqs. (1)
and (2) with FR = MR = 0 (e = 0) since the control forces in this case are internal forces. The
system of interaction equations for u0, 00 and u\ corresponds to

(40)

in which the matrix [Dc ((jj)} is given by Eq. (10) after the substitutions MI u\c and fi . fic.

24-16
 The approximate solution to Eq. (40) is given by
U0
u.
(41)

where
^2~ + 7T" + 73 (42)

and
(43)

The peak amplitudes of the transfer functions v^/ugt h\Q0/ug and u\/ug at u = &\c are
given by

I u0/ug | = (44a)

\hi90/ug \ = (446)
lie

\ui/ug \ = (44c)
;ic
The corresponding amplitude of the control force at u = u>ic is given by

v2l
/i2 (45)

To illustrate the effects of soil-structure interaction and control on the seismic response
of a one-storey structure we consider a structure characterized by mi /pa3 = 1.0, hi/a = 1
or hi/a = 2 and fi = 0.02. The foundation is modeled as a flat rigid disk foundation of
radius a placed on a uniform half-space characterized by the shear-wave velocity /?, density p,
Poisson's ratio v = 1/3 and hysteretic material damping ratio s = 0.02. The control gains
are determined for weighting factors of 71 = 72 = 1.0 and do not include SSI effects. The

24-17
results obtained are shown in Figs. 4 and 5. In particular, each figure shows the apparent system
frequency o>ic including the effects of SSI and control and the corresponding system damping
ratio fic. Also shown are the amplitudes of the transfer functions u\/ug , Fi/<J{m\ug ^ U0/ug
and hi00/Ug at the frequency &\c. The results in Fig. 4 and 5 are presented versus the relative
stiffness parameter ai = Uia//3 where a?i is the fixed-base natural frequency of the uncontrolled
superstructure, a is the radius of the foundation and 0 is the shear-wave velocity of the soil. The
results are shown for several values of the control parameter a = (^Jr)" 1 ranging from a = 0
for the uncontrolled case to a = 2 in which there is a large control force.

The results in Figs. 4 and 5 indicate that as control increases (i.e. as a increases) the
system frequency >ic, the system damping ratio fic and the control force FI also increase while
the deformation of the structure, the relative displacement of the base and the rocking of the
base decrease significantly.

The effects of soil-structure interaction tend to reduce the system frequency &\c and the
control force FI but the reductions are significant only for values of ai > 0.25. The interaction
effects clearly increase the relative displacement and rocking motion of the base. The effects of
SSI on the deformation of the structure depend on the amount of control acting on the structure.
For a small amount of control (a < 0.02), the system damping ratio fic increases and the
deformation of the structure u\ decreases significantly as the soil becomes softer. On the other
hand, if the amount of control is large (a > 0.02), the system damping ratio f ic decreases with
a\ and a slight increase in the deformation u\ of the structure can be obtained as the soil becomes
softer. For fixed values of a\ = o?ia//3, mi/pa3 and a, the effects of interaction appear to be
stronger as the slenderness ratio h\/a increases.

24-18
 RESPONSE OF A CONTROLLED 1-DOF OSCILLATOR WITH GAINS
DETERMINED INCLUDING SSI EFFECTS

We consider now the seismic response of a one-storey structure subjected to active control
in which the gains have been determined by approximately including the effects of soil-structure
interation. To start, we recall from Eq. (16) that the relative displacement ui for a one-storey
structure supported on a flexible soil and subjected to an internal force FI and to the seismic
excitation ug can be approximately obtained as the response of an equivalent 1-DOF structure
on a rigid soil satisfying the equation of motion

u, (46)

in which the equivalent frequency u>i and damping ratio fi are given by Eqs. (13) and (14),
respectively. The corresponding effective input ground motion ug and the effective force FI are
given by Eqs. (17a) and (17b), respectively.

To derive the optimal control parameters for the equivalent oscillator we use again, for
the purpose of comparison, the same performance index J
oo

J=
= JI {|7ii + Tawfwf+ dt (47)
o
which we rewrite in the form

J= dt (48)

in which
72 = 72 (49)

and
Q

f = (o>i/d>i) r (50)

To simplify the process of obtaining the optimal gains, the displacement HI is constrained to
satisfy the equation of motion for the equivalent oscillator in the absence of seismic excitation.
24-19
 Following Eqs. (23), (34) and (35), the optimal effective control force is given by

FI = mi \u\gm + uihui\ (51)

in which g and h are the effective gains

672-1 (52)

+ 071 + 2^/1 + 672 - 2 - 2|i (53)

where a = \/(Cj\f) = (u)i/ui)6a. The actual control force FI = (o;i/o>i)4Fi is given by

FI (t) = -mi [o;^i + uihui] (54)

where the actual gains, including SSI, are:

g =(ui/&i) 2g (55)

h =(ui/u>i)*h . (56)

The effects of SSI on the gains g and h are illustrated in Fig. 6 for the case mi /pa3 = 1.0,
hi/a = 1 (a, b) and hi/a = 2 (c, d), & = 0.02, fs = 0.02, v = 1/3, 71 = 1 and 72 = 1. In this
figure, the values of g* = (2g/a) and h* = (/i/x/2a) are shown versus the relative stiffness
parameter ai = uia/j3 for values of the control parameter a = (^i?')~ 1 of 0.002, 0.02, 0.2 and
2.0. It is apparent that the effects of SSI on g and h are significant only if ai > 0.5. The effects
appear to be stronger for the more slender structure and when the control force is small.

To calculate the relative response of the controlled structure we have several possibilities.
One approximation is to substitute from Eq. (51) into Eq. (46) leading to

Hi + 2icictti + &lcui = -ug (57)

in which now
i(l + a72) V4 (58)

24-20
and
1/2
1- (59)
+ Q72 2

In the frequency domain, the peak amplitude of the transfer function ui/ug for the oscillator
defined by Eqs. (58) and (59) would occur at u; = Cb\c and would be given by

f,~, I, , \2
(60)
u

The amplitude of the internal control force FI at LJ = u>ic corresponds to

h* (61)

where g and /i, are given by Eqs. (52) and (53), respectively.

A second and preferable approximate approach to obtain the response of the controlled
structure including soil-structure interaction effects is to substitute the internal control force
FI = (u>i/u;i) 4Fi in which FI is given by Eq. (51) into Eq. (3), the equation of motion for the
top mass. The resulting equation of motion is given again by

4- = m\ug (62)

in which now
(63)

and
(64)
2

Thus, the superstructure subjected to an internal control force FI can be represented by an

equivalent 1-DOF structure with frequency u\c and damping ratio ic . The response of this
equivalent structure to the seismic excitation in the presence of soil-structure interation effects
can then be obtained by use of Eq. (41) in which u\c and ic are calculated from Eqs. (42)
and (43) with LJIC and fic given by Eqs. (63) and (64). The values of the transfer functions
\Uo/ug \, \h\60/ug \ and \u\/ug \ at the controlled system frequency u; = o>ic are again given by

24-21
Eqs. (44a), (44b) and (44c). The amplitude of the control force at u>ic is also given by Eq. (45)
where g, h, o;ic, fic, >ic, and fic are now calculated by use of Eqs. (55), (56), (63), (64), (42)
and (43), respectively. This second approach is used in the calculations that follow.

The effects of SSI on the seismic response of a one-storey structure subjected to active
control by internal forces with gains determined including SSI effects are shown in Figs. 7
and 8. Again, the characteristics of the structure, foundation and control system correspond to
mi/pa3 = 1.0, hi/a = 1 (Fig. 7) and hi/a = 2 (Fig. 8), & = 0.02, fs = 0.02, v = 1/3 and
7i = 72 = 1-

The results in Figs. 7 and 8 for the case of gains determined including SSI effects follow
the same trends as those shown in Figs. 4 and 5 for the case of gains determined without SSL For
a\ > 0.5, the peak values of the transfer fuctions ui/ug, uQ/ug and hi00/ug are slightly larger
and the control force is slightly smaller when the SSI effects are included in the calculation of
the gains.

24-22
 CONCLUSIONS

A simple model for the seismic response of a one-storey structure subjected to active
control in the presence of soil-structure interaction effects has been presented. The resulting
representation in terms of a modified 1-DOF oscillator which includes the effects of control and
soil-structure interaction offers insight into the problem and allows us to evaluate the effects of
interaction on the seismic response of actively controlled structures in which the control gains
have been determined with and without inclusion of SSI effects.

It has been found that control reduces not only the internal deformation of the structure
but also the relative horizontal displacement and the rocking motion of the base. If the control
forces are small, the effects of SSI tend to reduce the deformation of the structure and the
control forces while increasing the relative displacement and rocking motion of the base. For
large control forces, the SSI effects may lead to deformations of the structure slightly larger than
those obtained when the interaction effects on the response are ignored. The effects of ignoring
the interaction between the structure and the soil in the calculation of the control gains are small
and result in an slightly lower response of the structure and the foundation at the expense of a
slightly larger control force.

The present results for internal control are slightly less favorable than those found by
Wong and Luco (1992) for an instance of external control. Similar advantages have been obtained
by Wu and Smith (1995) in their comparison of externally and internally controlled systems.

ACKNOWLEDGMENTS

The work described here was supported by a grant from the Ohsaki Research Institute of
Shimizu Corporation. The author wishes to acknowledge the contributions of Dr. F. C. P. de Bar-
ros to the calculation of the numerical results presented here.

24-23
 REFERENCES

1. Alam, S. M. S. and S. Baba (1991). "Active Optimal Control for Earthquake Excited
Structures Using Optimal Observer," Second Joint Japan/US Conference on Adaptive
Structures, Nagoya, Nov. 1991, 211-226.

2. Alam, S. M. S. and S. Baba (1993). "A Robust Active Optimal Control Scheme Including
Soil-Structure Interaction," Journal of Structural Engineering, ASCE, 119(9), 2533-2551.

3. Bielak, J. (1971). "Earthquake Response of Building-Foundation Systems," Report EERL-

71-04, Earthquake Engineering Research Laboratory, California Institute of Technology,
Pasadena, California.

4. Jennings, P. C. and J. Bielak (1973). "Dynamics of Building-Soil Interaction," Bulletin

of the Seismological Society of America, 63, 9-48.

5. Luco, J. E., H. L. Wong and M. D. Trifunac (1986). "Soil-Structure Interaction Effects on

Forced Vibration Tests," Report 86-05, Dept. Civil Engineering, University of Southern
California, Los Angeles, California, Sept. 1986, 112 pp.

6. Luco, J. E., M. D. Trifunac and H. L. Wong (1987). "On the Apparent Change in Dynamic
Behavior of a Nine-Storey Reinforced Concrete Building," Bulletin of the Seismological
Society of America, 77, 1961-1983.

7. Luco, J. E. and A. Mita (1987). "Response of a Circular Foundation on a Uniform Half-

Space to Elastic Waves," Earthquake Engineering and Structural Dynamics, 15, 105-118.

8. Meirovitch, L. and H. Oz (1980). "Active Control of Structures by Modal Synthesis." In

Structural Control, H. H. E. Leipholz (Editor), North-Holland Publishing Co., Amsterdam,
505-521.

9. Meirovitch, L. and H. Baruh (1982). "Control of Self-Adjoint Distributed Parameter

Systems," Journal of Guidance, Control and Dynamics, 5, 60-66.
24-24
10. Sato, T., K. Toki and H. Matsushima (1991). "Optimal Control of Structures Taking
into Account the Dynamic Soil-Structure Interaction," Colloquium on Control ofStruc-
tures,Tokyot July 1991, 257-263 (in Japanese).

11. Sato, T. and K. Toki (1992). "Predictive Control of Seismic Response of Structures Taking
into Account the Soil-Structure Interaction/' Proceedings First European Conference on
Smart Structures and Materials, Glasgow, Scotland, 245-250.

12. Smith, H. D., W-H. Wu and R. I. Borja (1994). "Structural Control Considering Soil-
Structure Interaction Effects," Earthquake Engineering and Structural Dynamics, 23, 609-
626.

13. Wu, W-H. and H. A. Smith (1995). "Comparison of SSI Effects on Externally and Inter-
nally Controlled Systems," Smart Mater. Struct., 4, A158-A168.

14. Wong, H. L and J. E. Luco (1990). "Active Control of the Seismic Response of Structures
in the Presence of Soil-Structure Interaction Effects," Proceedings U.S. National Workshop
on Structural Control Research, Oct. 25-26, 1990, University of Southern California, Los
Angeles, California, 231-235.

15. Wong, H. L and J. E. Luco (1991). "Structural Control Including Soil-Structure Interaction
Effects," Journal of Engineering Mechanics, ASCE, I, 2237-2250.

16. Wong, H. L and J. E. Luco (1992). "Effects of Soil-Structure Interaction on the Seismic
Response of Structures Subjected to Active Control," Proceedings Tenth World Conference
on Earthquake Engineering, Madrid, Spain, 2137-2142.

24-25
Figure 1. Model of One-Storey Structure.

24-26
 is

Figure 2. Normalized System Frequency o>i/u>i and System Damping Ratio fi as a Function
of the Relative Stiffness Ratio ai = u^a//?. [mi/pa3 = 0.5, /ii/a = 1 and 2, fi = 0.02,
& = 0.02, i/ = 1/3].

2.0 I i i i i 1.2
h (a)
1.5

1.0

.4
.5

i i i i i i
.0
i i i i ill

a a

Figure 3. Normalized Frequency u>lc/u>i and Damping Ratio fic of the Equivalent Oscillator
as a Function of a = (u^r)" 1 for Different Values of the Weights 71 and 72 (fi = 0).

24-27
 1.5 I I I T 1.00 I I 1
(a) _ (b) -
.75
1.0
3 ".50
u
IB
.5
.25

Figure 4. Effects of SSI on the Response of an Actively Controlled One-Storey Structure for
the Case of Gains Determined without SSI. Results include: (a) u>ic/o>i, (b) |ic and (c) \u\/ug \,
(d) \Fi/wlmiug \ t (e) \u0/ug \ and (f) \hi00 /ug \ at LJ = wic. Slenderness ratio /ii/a = 1.0.

24-28
 1.5 i I I r 1.00 i i i i r
(a) _ (b) -
.75
1.0
3 .50
u J_P
13
.5
.25

.0 .00
30 i i i i 2.0 i 1 l r
(c) _ (d) -
1.5
20
CO
0.002
I
-0.02 J 1.0
-0.2
I
' 10 3
2.0
- .5

J____I____I____I
0 .0

2.0 I I i i I 8 i i
(e) -
1.5

00
3
1.0
O CD
=9
j=r
.5

i i I I
.0
.0 .5 1.0 1.5 .0 .5 1.0 1.5
a, = a, =<0j a/p

Figure 5. Effects of SSI on the Response of an Actively Controlled One-Storey Structure for
the Case of Gains Determined without SSI. Results include: (a) >ic/k>i, (b) fic and (c) \u\/ug \,
(d) \Fi/u\miug \i (e) \u0/ug \ and (f) \h\60/ug \ at o> = cDic. Slenderness ratio hi/a = 2.0.

24-29
 l.J iiill 1.3 IIIII
h,/a=l (a) _ hj/a=l (b) _

1.0 1.0
^^^ -~ ~"~~--*^
*

.5 .5

.0 i i i i i .0 iiiii

1.5 11111 1.5 iiiii

. /a
h l/d 0
z, ff*\ _
^c; hj/a = 2 (d)
a
..---"""
1.0 -^ 0.002 - 1.0
~-^s. 0.02 "*^^
^^^ c\ o *
- "V "-0.2
----.....X. -.- 2.0
V.

.5 .5
\ x --^^

n iiiii r\ 1 1 1 1 1 "
.5 1.0 1.5 .0 .5 1.0 1.5

Figure 6. Effects of SSI on the Normalized Control Gains g* = (2g/a) and h* = (h/V2a) for
hi/a = 1 (a, b) and hi/a = 2 (c, d).

24-30
 1.5 I I I I r 1.00 i i i i r
(a) _ (b) -
.75
1.0
3
u
]3
.5
.25

.0 .00
30 i i i I r 2.0 i i I i
(c) _ (d) -
1.5
20 i
CC 0.002
I
- 0.02 J 1.0
-0.2
I
' 10 3
2.0
- .5

L 1 I
0 .0
2.0 I I \ \ 2.0 1 I I I

(e) - (f) -
1.5 1.5

00
i 3

o
1.0 ^1.0
a
jcT
.5 .5

i i i
.0
.0 1.0 1.5 .5 1.0 1.5

Figure 7. Effects of SSI on the Response of an Actively Controlled One-Storey Structure for
the Case of Gains Determined with SSI. Results include: (a) tDic/^i (b) fic and (c) \ui/ug \,
(d) \F\lu\miug \, (e) \u0/ug \ and (f) \hi00/ug \ at u = Cj\c. Slenderness ratio hi/a = 1.0.

24-31
 1.5 i I I I T 1.00 1 1 I i T
(a) _ (b) -
.75
1.0
3
u
jj>Ji.50
13
.5
.25

.0 J____I____I____I____I .00

30 2.0 I I T I

(d) -
_ L5
20 p
00

DO
I
-0.02 _ f 1.0
-0.2
r' 10 2.0
3

- .5

....... +......H........^.-.....^......H........ i i
o .0
2.0 i l I r 8 I i I I

(e) - (f) -
1.5

1.0 05

.5

i i i___i
.0
.0 .5 1.0 1.5 .0 .5 1.0 1.5
a, = co, a/p

Figure 8. Effects of SSI on the Response of an Actively Controlled One-Storey Structure for
the Case of Gains Determined with SSL Results include: (a) LJIC/UI, (b) fic and (c) \u\/ug \,
(d) \Fi/UiiniUg\ t (e) \u0 /ug \ and (f) \hi00 /ug \ at u; = u)\c. Slenderness ratio hi/a = 2.0.

24-32
 SEISMIC RESPONSE OF A PIPELINE EMBEDDED
IN A LAYERED MEDIUM

J. Enrique Luco(1) and Francisco C. P. de Barros(2)

f 1) Department of Structural Engineering, University of California, San Diego, La Jolla,

California 92093-0411, USA.

(2) Departamento de Ciencias Fundamentals, Radiagoes e Meio Ambiente, Instituto Militar

de Engenharia, Pra$a General Tiburcio 80, CEP 22290-270, Rio de Janeiro, Brasil.

ABSTRACT

A method to obtain the three-dimensional harmonic response of a infinitely long pipeline

of circular cross-section embedded in a layered viscoelastic half-space and subjected to harmonic
plane waves impinging at an oblique angle with respect to the axis of the shell is presented. The
procedure combines an indirect integral representation for the field in the exterior half-space with
a model of the pipeline or tunnel based on Donnell shell theory. The integral representation for

the soil is based on the use of moving Green's functions for the layered viscoelastic half-space.
The accuracy of the formulation is tested by comparison of results obtained by use of different
discretizations. Extensive comparisons with previous two- and three-dimensional results for the
case of a shell embedded in a uniform half-space and some new numerical results for a pipeline
embedded in a multilayered half-space are also presented.

25-1
 INTRODUCTION

In this paper we consider the three-dimensional response of an infinitely long cylindrical

shell of circular cross-section embedded in a horizontally layered half-space (Fig. 1 a, b). The
shell representing a pipeline or a tunnel and the soil are excited by plane waves impinging at an
oblique angle with respect to the axis of the shell (Fig. 1 a, c). Although the geometry of this
model is two-dimensional, the response is fully three-dimensional.

Most of the previous work on the seismic response of pipelines or tunnels has been
concentrated on two-dimensional models. In particular, the anti-plane shear case of a cylindrical
shell of circular cross-section buried in a uniform elastic half-space and subjected to plane SH-
waves normal to the axis of the shell has been considered by Lee and Trifunac (1979) and
Balendra et al (1984). The solution is obtained by expanding the fields in both the exterior
and interior region in series of cylindrical wave-functions. The coefficients of the terms in
these series are obtained by truncating an infinite system of equations in the infinite number
of unknown coefficients. The plane-strain case for P-, SV- and Rayleigh-wave excitation has
been considered by El-Akily and Datta (1980, 1981), Datta et al (1983, 1984), Wong et al
(1985) and Chin et al (1987). El-Akily and Datta (1980, 1981) considered a cylindrical shell
of circular cross-section buried in a uniform half-space. The external field was represented by
a series expansion in cylindrical wave-functions while the shell was modelled using Fliigge's
bending theory. Matched asymptotic expansions and a successive reflection technique were used
to determine the coefficients in the expansion. Datta et al (1983,1984) considered a cylindrical
pipe of circular cross-section lying in a concentric cylindrical region of soft soil buried in a
uniform half-space. The fields within each of the three regions were expanded in series of
wave-functions. The unknown coefficients in the expansion were obtained by truncating an
infinite set of linear equations in the infinite number of unknown coefficients. Wong et al (1985)
considered the two-dimensional response of a lined tunnel of non-circular cross section buried in
a uniform elastic half-space by use of a technique involving wave-function expansion in the half-
space combined with a finite element representation of the tunnel and its immediate cylindrical

25-2
vicinity. The same approach was used by Chin et al (1987) to study the response of a pipe of
circular cross-section buried in a back-filled trench embedded in a uniform elastic half-space.

Three-dimensional models of infinitely long pipelines similar to those considered in this

study have been considered previously by Datta and Shah (1982) for a full space and by Wong et
al (1986a, 1986b) and Liu et al (1988) for a half-space. In particular, Wong et al (1986a, 1986b)
have considered the response of a cylindrical pipeline of circular cross section buried in a uniform
elastic half-space and subjected to obliquely incident plane P-, SV-, SH- and Rayleigh-waves.
The solution was obtained by expansion into wave-functions in both the half-space and the shell
and by truncation of the resulting infinite system of equations. Liu et al (1988) have obtained the
three-dimensional response of an infinitely long pipeline buried in a backfilled trench embedded
in a uniform half-space when subjected to obliquely incident P- and SV-waves. These authors
have used a hybrid approach in which an internal region including the pipeline is modelled by
finite elements while the exterior region is modelled by use of a boundary integral representation
in terms of Green's functions for a uniform half-space.

In the present paper, a method of solution which combines an indirect integral represen-
tation for the exterior soil with a simplified shell theory (Donnell, 1933) for the internal pipeline
or tunnel is presented (Luco and de Barros, 1994). The integral representation for the exterior

domain is based on the moving Green's functions for a layered viscoelastic half-space obtained
by de Barros and Luco (1992, 1994b). In this way, the physical three-dimensional problem is
reduced to an essentially two-dimensional mathematical problem. The integral representation for
the exterior domain, the solution for the shell and a discussion of the accuracy of the formulation
is presented next. Detailed comparisons with previous two- and three-dimensional solutions for
a shell in a uniform half-space and new numerical results describing the response of cylindrical
shells embedded in layered media and subjected to obliquely incident P-, SV- and SH-waves are
also presented (de Barros and Luco, 1994b). An extensive bibliography on the seismic response
of pipelines and tunnels is included at the end of the paper.

25-3
 FORMULATION

The geometry of the model is illustrated in Fig. 1. The external region representing the
soil consists of (N 1) horizontal viscoelastic layers overlying a viscoelastic half-space. The

infinitely long shell of circular cross-section is parallel to the free surface of the half-space and

is located at a depth H. Perfect bonding is assumed to exist between layers and between the
shell and the exterior medium. Each of the media in the exterior half-space is characterized,
for harmonic vibrations, by complex P- and S-wave velocities G.J = &j(l + 2iaj ) 1 /2 and

(3j = (3j(l + 2ipj ) 1/2 where i ^/ l, and by the density PJ (j 1,7V). The terms aj
J3j represent (approximately) the real parts of the P- and S-wave velocities, and aj and ^
represent the small hysteretic damping ratios for P- and S-waves, respectively. The shell is
characterized by the centerline radius a, tickness h, Young's modulus E0 , Poisson's ratio v0 and
density p0 .

In what follows, the excitation and the response will have harmonic time dependence of

the type e'lu}t where u is the frequency. For simplicity, the factor elu}t will be dropped from all
expressions.

Free-Field Ground Motion.

As a first step in the formulation, it is necessary to determine the ground motion and
the stress components for free-field conditions, i.e., in the absence of the shell. The seismic
excitation is represented by homogeneous plane P-, SV- or SH-waves, such that the normal to
the wave front in the underlying half-space forms an angle #v with the vertical axis (0V = 0 for
vertical incidence). The projection of the normal to the wavefront on the horizontal plane forms
the angle #h with the axis of the shell (x-axis).

To calculate the free-field ground motion it is convenient to consider the coordinate

system x', y', z' (z1 = z) shown in Figs, la and Ic. Referred to this coordinate system, the

25-4
incident motion within the underlying exterior half-space is represented by

z> (1)
where A is the amplitude of the incident displacement, k' = (o;//3;v)sin$v for S-waves and
kf = (id/aw) sin#v for P-waves. The term v'N is defined by

( u \
i I I cos #v for S-wave excitation
\PN J
(2)
i ( w \ cos 0V for P-wave excitation

In Eq. (1), ZN is the depth of the last interface with the underlying half-space and {[/'} is the
vector
' (sin #V5 0, cos #V ) T for P-wave excitation

(cos #v , 0, sin#V ) T for SV-wave excitation (3)

(0,1,0) T for SH-wave excitation

The total free-field ground motion satisfying all the continuity, free-surface and radiation
conditions for the layered geometry shown in Fig. Ic can be calculated by the approach described
by Luco and Wong (1987). The resulting free-field displacement and stress fields in the x'y'z' -
coordinate system are denoted here by

-ik'*'
= {V'ie(z')}e (4)

(5)

where the elements of U'\^(zr ) and S/ iff(z/ ) are independent of y' and depend only on z' = z.

To impose boundary conditions at the interface between the layered half-space and the
shell it is necessary to introduce the rotation of coordinates
x' cos #h sin Oh 0 x X
sin #h cos #h 0 y}= y (6)
0 01 z z

25-5
leading to the free-field displacement and stress fields in the xyz-coordinate system:

= {U lff(x0 )}e-ikx (7)

= p lff(x0 )]e-^ (8)

where x0 = (O,?/,^), k (U/PN) sm#v cos0h for S-wave excitation and k =

sin0v cos#h for P-wave excitation. In Eqs. (7) and (8),

{Ulf[(x0 )} = [C] {t/' lff(z)}e- ifc ^ineh (9)

Piffle)] = [C] [V(z)] [Cfe-ifc'" sineh (10)

in which the y-dependence enters only in the last exponential factor.

V
The free-field traction vector {t\ff(x)} on the area that will be in contact with the shell
can be written in the form
Oiff(aO} = {Tiff(x0 )}e-ifcc (11)

where

{Tlff(x0 )} = piff(a?o)]M*o)} (12)

in which {v} = (0, 1/^,1/2 ) T is the unit normal to the shell's boundary at x0 pointing into the
shell. It is noted that the variable x appears only in the exponential factors exp( ikx) affecting
i/
, [criff] and {t^}.

Finally, it is convenient to recall for future reference that the incident ground motion in
the underlying half-space referred to the (x, y, z) coordinate system is represented by the plane
wave
{UIN }inc = A{U} exp < i [x sin 6V cos #h + y sin Ov sin 0J + v'N z > (13)

where A is the amplitude of the incident displacement and the vector {U} is given by

( sin #h i cos #h, 0)T for SH-wave excitation,

(cos #h cos 0V , sin^h cos 0V , sin 0V )T for SV-wave excitation, and (14)

(cos #h sin 0V , sin ^h sin 0V , cos 0V ) T for P-wave excitation,

25-6
in which 9V and #h are the vertical and horizontal angles of incidence. The velocity v appearing in
Eq. (13) corresponds to the (complex) velocity a^ in the underlying half-space for P-excitation
and to PN for SV- and SH-excitation.

It should be noted that if #h = 90 or if #v = 0, then, the incident displacement becomes

independent of the coordinate x and the problem becomes two-dimensional.

Contact Problem.

In the presence of the shell, the total displacement vector {u\(x)} and the total traction
vector {t\(x)} in the exterior region are written in the form

{u\(x)} - {itiff(f)} + {uis (x)} (15)

{t} (x)} = {tlff(x)} + {tls (x)} (16)

where {ui s } and {t\ s (x)} are the scattered displacement and traction vectors, respectively. The
exterior field satisfies the conditions of vanishing tractions on the free surface (z = 0) and
the continuity conditions at layer interfaces. The exterior scattered field must also satisfy the
radiation conditions at infinity. At the interface S between the shell and the exterior medium,
the continuity conditions

= {u2 (x)} (17)

, xeS (18)

IS

apply, in which {u2 (x)} and {t2(x)} are the displacement and traction vectors for the shell. In
i/
here we assume that {u2(x)} and {^(a)} can be related in the form

{u2 (x)} = I [G22 (x,x')] {t2 (x')}dS (19)

Js

where [(^22(^5 xr )] is the 3x3 matrix of Green's functions for the shell.

25-7
 To solve the boundary-value problem, the exterior scattered field is represented as result-
ing from the action of a distribution of concentrated loads moving in the direction of the x-axis

with velocity c = u>/k. These moving loads act in the layered exterior half-space (without the

shell) on the surface Si (Fig. Id) located within the region to be occupied by the shell. The
scattered displacement field {uis (x)} is, then, written in the form

{uis } = {Uis (x0 )} e (20)

where

{Uis (x0 )}= f [Gn(x0 ,y0 )}{F(y0 )}dl(y0 ) , (21)

J Li

in which [Gu(x0 , y0 )] is the 3 x 3 matrix of moving Green's functions (Barros and Luco, 1992,
1993). The first, second and third columns of the matrix G correspond to the displacement
vector at x0 = (0, ?/, z) for a unit point load acting in the x, y and z-directions, respectively,
and moving with velocity c = u/k along a line parallel to the re-axis passing through the point
ff0 = (0,3/1,2:1) on Si. The contour LI corresponds to the intersection of Si with the plane

x = 0. The 3x1 vectors {F} represent the unknown amplitudes of the j-th moving loads. The
V
corresponding traction vector {ti s (x)} for the scattered field at the interface S with the shell can

be written in the form

(22)

where
= [H n (x0 ,y0 )}{F(y0 )}dl(y0 ) (23)

in which rx
xx
a:xy ayxy
0 vy vz 0 0 0 x
a:xz
00 0 vy vz 0 .x
(24)
00 0 0 vy vz yy yy
X
a.yz
.x
zz

25-1
In Eq. (24) (v>y(x0 ), vz (x0 )) are the direction cosines of the normal to the surface S of the shell,
and cr^rr 0,2/0), <jy.x (x0 ,y0 ],..., are the stresses at x0 = (Q,y,z) induced by the moving unit
point loads acting in the x, y and z-directions.

The continuity conditions (17), (18) and Eqs. (7), (11), (20) and (22) indicate that {u.(x)}
V
and {tj(x)} can be written in the form

= {I7j(ay} e~ik* (3 = I, 2) (25)

(26)

The relation (19) is then given by

) di(x'0 ) (27)

where the contour L corresponds to the intersection of S with the plane x = 0, and

|- -, fOO

G22 (^0 ,4) = / 22(*,*') (28)

J 00

By use of Eqs. (15), (16), (21), (23), (25), (26) and (27) the displacement fields {Ui(x0)}
and{t/2(^0 )} at the interface S between the shell and the soil can be written in the form

}^i(i7o) (29)

{U2 (x0 )} = - dl(x'0 ]

[<522 (x0 , <)] [Hn&, 2/o)] (30)

At this point, we use a weighted version of the displacement continuity condition and
require that
T x0 )})dl(x0 ) = () . (31)

25-9
Substitution from Eqs. (29), (30) into Eq. (31) leads to

7'01 ?7
yojln TFY?7 H dh
\ r \yo)f ULi (if
\yoj} I\ -Ly
D(i7'}
Vi/o/ f!> (19}
voz/

where

(33)
JL JL L JL JL J /

and

2//!)] {Uiff(x0 )} dl(x0 ) +

'f _ -\ _^ 1
(34)

Eq. (32) represents an integral equation for the unknown distribution of forces {F(y0 )}.

The kernel [B] and the right-hand-side {D} of Eq. (32) depend on the moving Green's
function matrix 622(^o,^) for the shell. These Green's functions are derived in the next

section.

MOVING GREEN'S FUNCTIONS FOR A CYLINDRICAL SHELL

The equations of motion for harmonic vibrations of the cylindrical shell are given by

-u2Po h{u'2 } - K0 [L]{u'2 } = {t2e'} (35)

where {u2 } (u,v,w) T represents the midsurface displacement vector in the local coordinates
shown in Fig. 2, {^e7 } is the effective traction vector also referred to the local coordinates, p0 is

25-10
the density of the shell, h the tickness, K0 = E0 h/(l v2 ) in which E0 is the Young's modulus
and v0 is the Poisson's ratio. The operator matrix [L] has for elements (Donnell, 1933)

dx2 2a2
(l-i/0 ) d2 1 d2
2 ~dx2+ ~^2 ~d62

_h^
~12 ( <94 2 + -T-
a2 dx2d02 a4 80*
(36)
i/
Z/12 = Z/21 =
2a

^A
a <9x

-L/9S = -L/39. = ^
^r -^-r
^d<9

j^ *^
The effective traction vector {t^1} is related to the actual traction vector {2'} on the
outside surface (r = r0 = a + h/2) of the shell through the relation

0 0
v i
AI a_
_ 2 dx
A^_
2a 96
{ *'} ' (37)

To solve Eq. (35) we make explicit the exponential x-dependence and write

(38a)

^ ikx
(386)

Applying the Finite Fourier transform

Wn = \e~m6 dO (39a)

00
inO
W(U) y wn (396)
n= oo

25-11
to Eq. (35) leads to
E0
(40)

where

[I] (41)

in which [I] is the 3 x 3 unit matrix. The elements of the matrix [Ln] are given by

/T \9 / J- "0 \
(ka) 2 + ^ rr

^)(fca) 2 +n5

= -^(^j [(H4 + 2(/ca) 2 ^2 + 4] - 1

(42)
/1 _j_ ^o \
= ^21 = ( IT (fca) n
V z /
= L3 i = i i/o (fca)

32

The matrix [Bn] is given by

0 0

f 7^ 0 (43)

Solving Eq. (40) and inverting the Finite Fourier transform leads to
-27T

[G^(e,ff)}{T2'(ff) ( ^
(44)

where
(45)

25-12
 To impose the continuity conditions at the interface between the shell and the external
medium we must consider not the midsurface displacements {U2 (0)} but the displacements
{U2 (0}} on the outside boundary r r0 = a + h/2. These displacements are related by
h d
1 0 2 dx
r\ -1 I h _ h 8
U L "*" 2a 2a 86 (46)
00 1

The resulting expression for the outside displacement vector in terms of the tractions is

aJ [&m (e,ff)] {t2'(ff)} dff (47)

where
[D 1 [Bn] (48)

in which
1 0 i
01 + (49)

0 0

Some of the stresses of interest are given by

dx a 86 a
Ko , d 1 d 1
h 0 dx a 86 a

_
dx

d^ , t/o d^ \
dx2 "*" a2 d0 2 y
2 Q2 \
Z nU ! o
a2 86
'- l/n-X ? }
0 8x 2 I {u'2 (x,0)} . (50)
2 52
1 ~o
d
0 a dx a 8x86

where z r a [z = h/2 on the outside wall ( r r0) and z h/2 on the inside wall

(r = TI) ]. By writting

^x
^6 >.e zfca: (51)
2axe/(l - z/0 ) 2Exfl /(l - i/0 )

25-13
it is found that

(52)

where

(53)

in which

0 iv0n (ka) 2 + z/0n2

iv0 (ka) in 1 0 in v0 (ka} 2 + n2 (54)
m i(ka) 0 0 ika 2(ka)n

The relation between displacements and tractions given by Eq. (47) involves the dis-
placement and traction vectors referred to the local (cylindrical) coordinates. To connect the
displacements and tractions on the shell with those of the soil it is necessary to refer these vec-
V
tors to the global cartesian coordinates (x,2/,z). Denoting by {U%} and {T^} the vectors
V
and {Ti} when referred to the global coordinate system we have

= [C0 ] {0J (55a)

and
= [C0] T {f2 } (556)

where
0 0
[C0]= sin0 cos9 (56)
cos 9 sin 9

The relation between the displacement and the traction can now be written in the form
"2-rr
{U2 (0)} = a [G22 (6>, 07 )] { (57)

where

22(0,0')] [C0(o')} (58)

25-14
 NUMERICAL APPROACH

The integral equation (32) is discretized by replacing the unknown distributed forces
{F(y0 )} by a set of unknown concentrated forces {Fsj } (j'< = 1,7VS) acting at 7VS source points

2/o = ffsj n L\ and by imposing Eq. (32) at the same set of discrete points y^ = ysj- (j' = 1, 7V"S).
In addition, the integrals over L appearing in Eqs. (33) and (34) are discretized by use of
numerical integration formulae involving a set of 7V0 observation points x0 = xoi (i 1,7V0) on
L. The resulting set of linear algebric equations can be written in the form

[S] {F} = {D} (59)

where the 3x3 blocks of the 3NS x 3NS matrix [B] correspond to [B(ySijySj)], {F} T

} T ,...) and {D} T = ({D(ysl }T ,

To reduce the possibility of ill-conditioning it is useful to write {F} in the form of a

finite Fourier expansion with respect to the angular coodinates Oj = arctan [(H zsj)/xsj] of the
source points ysj (j = 1,7VS). In this case,

{F} = [M] {F0} (60)

where the i row of 3 x 3 blocks of [M] is given by ([7], cosOi [7], sm6i [7], cos 2^ [7],
sm26i [7], ..., cos ^f-Oi [I]) in which [7] is the 3x3 identity matrix, 7VS is assumed to be even
and 0i = 0. The coefficients {-F0 } in the expansion are obtained from Eq. (59) in the form

Once the forces {F} have been calculated by use of Eqs. (60) and (61), the displacement fields
in the external medium and in the shell can be calculated from Eqs. (29) and (30).

CONVERGENCE OF THE NUMERICAL APPROACH

The numerical results obtained by the procedure described in the previous section depend
in principle on the location and number of source points (7VS ) and on the number of observation

25-15
points (7V0) used. The first step is to test the convergence of this approach. For this purpose
we consider a concrete circular shell (p0 = 2,240%/ra3 , E0 = 1.6 x 10 10 7V/ra2 , VQ = 0.2,
h = O.ln = 0.0909r0) buried to a depth H = 5.0rj = 4.545r0 in a uniform half-space
(pi = 2,664%/m3 , EI = 7.567 x 109 N/m2 , i/i = 0.333, a = fy = 0.001). The half-space
is subjected to non-vertically incident (#v = 30) P- and SV-waves propagating in the direction
of the shell (#h = 0). The frequency of the excitation is such that 77 cjr0 /7r/3i = 0.105.

Numerical results for the normalized displacement components at a few points on the
ground surface (z = 0) and at a few points on the external wall of the shell (r = r0) are presented
in Table 1 for different numbers of sources and observers (7VS , 7V0). Also shown are some values
for the normalized hoop stress on the centerline (r = a) of the shell. In all cases, the sources
are equally spaced on a circle of radius a' = r0 3(27rr0/7V0 ) (7V0 > 20). Thus, as the number
of observation points increases, the sources move closer to the actual boundary r = r0 .

The displacement amplitudes Ui \Ui/A\ are normalized by the amplitude A of the

incident displacement vector on the ground surface. The normalized hoop stress is given by
00 = \&ee(a)/up\fi\A\ . The numerical results presented for source/observer combinations
(7VS , N0} of (20, 40) and (40, 80) show that the procedure is very stable as the number of source
and observation points increases. It appears that 20 source points and 40 observation points are
sufficient for most applications.

VALIDATION AND CRITICAL COMPARISONS

In the comparisons that follow the displacements are normalized by the amplitude A of
the incident displacement field. All of the stress components with the exception of <JXQ = <JQX
are normalized by upflA = (ajr0/J3)p,(A/r0) where p, J3 and p, = J32 p are the density, shear
wave velocity and shear modulus of reference and r0 is a length of reference corresponding to
the external radius of the circular shell. The shear stress VQX is normalized by (1 v0)up/3A/2
where i/0 is the Poisson's ratio of the shell. The reference quantities p, J3 and JJL are taken to

25-16
correspond to those of the underlying half-space (which correspond to pi, /?i and /^i in the case
of a uniform half-space). In judging the comparisons it must be kept in mind that the present

results include a small amount of attenuation in the soil Q = p = 0.001 and no attenuation

in the shell (QO = ^0 = 0) while the results by other authors typically do not include any
attenuation.

Finally, the present results have been calculated using Ns = 20 source points equally
spaced on a circle of radius rs r0 6irr0/N0 where N0 = 40 is the number of observation

points equally spaced along the external shell boundary (r = r0).

Two-Dimensional Anti-Plane Shear Cases.

A first comparison is made with results presented by Lee and Trifunac (1979) for a
circular cylindrical shell of external radius r0 , internal radius r\ = 0.9r0 , thickness h = O.lr0
and embedment depth H 1.5r0 subjected to a vertically incident SH-wave with particle motion
along the axis of the pipeline (#h = 90, 0V = 0). The shell is characterized by shear modulus
/io, shear wave velocity J30 and density p0 (j!0 = ffipo) and the surrounding uniform half-space
is characterized by p,\, fi\ and p\. Lee and Trifunac (1979) present results for p,0 /P>i 3 but
do not state the value for p0 /pi or J30 /Pi- m nere we assume (Lee, personal communication)

that PO/PI = 3 and, consequently, /30 /A = 1. The present results were calculated by assuming
a small amount of attenuation ^1 = 0.001 in the half-space and no attenuation in the shell
00 = 0. The results of Lee and Trifunac correspond to purely elastic media. Finally, the
comparisons were made for the dimensionless frequency 77 = Ljr0/irJ3i = 0.5.

Figs. 3a and 3b show the comparisons for the amplitudes of the normalized displacements
Ux = | uz /A | on the surface of the half-space (z = 0) and on the external boundary of the
shell (r r0). These displacements are normalized by the amplitude A of the incident SH-
wave at z = 0. Fig. 3c shows the comparisons for the amplitude of the normalized stress
^rx = |0VzA*>PiA^I on the external wall of the shell (r = r0). The results in Figs. 3a and

25-17
3b show excellent agreement between the present results and the results of Lee and Trifunac
(1979) for the surface displacements (z 0) and for the displacements on the external pipe
wall (r = r0). The results of Lee and Trifunac (1979) for the normalized contact stress Erx on
r = r0 differ in shape and amplitude from the present results (Fig. 3c). The peak value for |Erx |
obtained by Lee and Trifunac is about 6 times larger than our result.

As a second comparison we consider the results presented by Balendra et al (1984) for

a concrete circular shell (r0 = 3m, r\ = 2.7m, h = 0.3m, p,0 = SAGPa, J30 1, 870m/ sec,

Po = 2,410 kg/m3 , v0 = 0.2) buried to a depth H = 2.5r0 = 7.5m in a uniform half-

space (//i 0.111 GPa, fi\ 260m/sec, p\ = 1,640%/m3 ). The medium is subjected
to a non-vertically incident (#v = 30) SH-wave propagating normal to the axis of the shell
with a frequency of 10.61 Hz. In this case, the dimensionless frequency 77 takes the value

77 == tjr0 /7r/3i = 0.245. Comparisons for the amplitudes of the normalized displacement Ux =
\ux (ri)/A\, and normalized shear stresses Era = \(7rx (r0 )/upi^ia\ and E0X = |2<70x (ri)/[(l
v0 )Ljpi/3ia]\ are presented in Figs. 4a, 4b and 4c, respectively. The figures show excellent
agreement between the present results and those obtained by Balendra et al (1984). The small
differences for Y>QX can be attributed to the fact that the present calculation is based on a thin
shell theory which in the 2-D case (ka = 0) leads to a shear stress a0x which does not vary
across the tickness of the shell.

As a final comparison for the two-dimensional anti-plane shear case we consider the
results presented by Liu et al (1991) for a concrete circular shell (p0 = 2.24 x IQ3 kg/m3 ,

E0 = 1.6 x 1010 7V/m2 , i/0 - 0.2, h = O.ln = 0.0909r0) embedded to a depth H = 5n =

4.545r0 in a uniform half-space (pi = 2.664 x 103 kg/m3 , E} = 7.567 x 109 7V/m2 , ^ =
0.333) and subjected to a vertically incident SH-wave (Fig. 8d in paper in reference). The
corresponding comparison for the amplitude of the normalized displacement Ux = \ux (r0 )/A\
for a dimensionless frequency 77 cjr0/7r^i = 0.105 is shown in Fig. 5. The agreement
betweeen the two sets of results is excellent.

25-18
Two-Dimensional Plane-Strain Cases.

As a first test for the plane strain case we consider the results presented by Datta et al
(1983) [see also Datta et al (1984)] for a concrete circular shell ( p0 = 2.24 x 103 kg/m3 , E0 =
1.6 x 1010 W/ra2 , i/o = 0.2, h = O.ln = 0.0909r0) buried to a depth H = 8.33n = 7.573r0
in a uniform half-space (pl = 2.665 x W3 kg/m3 , EI = 6.9 x 108 JV/ra2 , i/, = 0.45). The
medium is subjected to vertically incident P- and SV-waves with a dimensionless frequency
77 = wr0/7r/3i = 0.132.

Comparisons for the normalized amplitudes of the radial displacements Ur = \ur (rQ)/A\
on the external wall of the pipe are presented in Fig. 6a and 6b for vertically incident P- and
SV-waves, respectively. There is a excellent agreement between the present results and those
presented by Datta et al (1983). To make these comparisons it was necessary to renormalize
the results presented by Datta et al (1983) for P- and SV-waves by multiplying these results by
factors of 1.362 and 2.0, respectively. These factors correspond to the ratio of the peak free-field

radial displacements at r = r0 to the amplitude A of the incident waves. These peaks occur at
9 = 90 and 6 0 for P- and SV-waves, respectively.

The corresponding comparisons for the normalized hoop stresses E00(r0 ) for P- and SV-
waves are shown in Figs. 6c and 6d, respectively. The agreement between the two sets of results
is very good. The results of Datta et al (1983) shown in Figs. 6c and 6d were renormalized by
multiplying by factors of 5.434 and 0.629 which correspond to the peak values of the normalized
hoop stresses 00(r0 ) in the free-field. Also shown in Figs. 6c and 6d are the normalized hoop
stresses E#0(a) on the centerline of the shell (segmented lines). The significant differences
between the stresses at r = r0 and r = a indicate a significant amount of bending of the shell.

As a second test for the plane strain case we consider the results presented by Wong et
al (1986) for a concrete circular shell ( p0 = 2,240%/m3 , E0 = 16 x 109 N/m2 , v0 = 0.2,
h = O.lri 0.0909r0) embedded to a depth H = 2.On 1.818r0 in a uniform half-space
(pi = 2,665%/m3 , EI = 0.69 x 109 N/m2 , vi = 0.45). Values for the hoop stress E00(r0 )

25-19
and for the longitudinal stress Exx (r0) at r r0 were presented for nonvertically incident
(l9v = 10) P- and SV-waves for 77 = wr0/7r/3i = 0.132.

Comparisons for the normalized hoop E00(r0 ) and longitudinal Exx (r0 ) stresses at r = r0
for both P- and SV-waves are shown in Fig. 7. The agreement between the present results and
those of Wong et al (1986) for E#0 is very good. Some small differences for Exx (r0 ) (Figs. 7c
and 7d) can be attributed to the present use of a simplified shell theory in which Err is considered
to be much smaller than E00 and Exx . The results of Wong et al (1986) shown in Fig. 7a, 7b,

7c and 7d were renorrnalized by multiplying by factors of 2.026, 1.856, 1.840 and 0.5764,

respectively. These factors correspond to the peak values of the corresponding normalized
stresses in the free-field.

Also shown in Figs. 7a, b, c, d are the normalized stresses (segmented lines) calculated

by the present approach on the centerline r a of the shell. It is apparent, particularly for

SV-excitation, that the bending effects are significant.

As a third test for the plane strain case we consider the results presented by Liu et
al (1991) for a concrete circular shell ( p0 = 2.24 x W3 kg/m3 , E0 = 1.6 x 1010 7V/m2 ,
v>0 = 0.2, h = O.lri = 0.0909r0) buried to a depth H = 5.On = 4.545r0 in a uniform half-space
(pi = 2.664 x 103 kg/m3 , EI = 7.567 x 109 7V/m2 , 1/1 = 0.333). The medium is subjected to
vertically incident P- and SV-waves with dimensionless frequency 77 = ur0 /-Kft\ 0.105.

The amplitudes of the normalized radial displacements Ur = \ur (r0 )/A\ on the external
wall of the pipe are compared in Figs. 8a and 8b for vertically incident P- and SV-waves,

respectively. To compare both sets of results it was necessary to renorrnalize the results presented
by Liu et al (1991) by multiplying by the factors 1.667 and 0.424 which correspond to the peak
values of the normalized amplitudes of the free-field radial displacements at r = r0 for P- and
SV-waves, respectively. In this case, the peak values of ur (r0 ,9) in the free-field occur at
9 = 135 and 0 = 90 for P- and SV-wave, respectively.

25-20
 Comparisons for the normalized hoop stresses 00 (a) on the centerline of the shell are
shown in Figs. 8c and 8d. For the purpose of the comparison the results of Liu et al (1991)
for 00 have been renormalized by multiplying by 2.766 and 1.974 corresponding to our peak

values for 00 (a) in the free-field. It is appparent from Fig. 8 that excellent agreement exist
between the two sets of results. We note that Liu et al (1991) present two sets of results, one
labeled "analytic" and a second set calculated by a hybrid approach. The comparisons in Fig. 8

refer to the "analytic" results. The agreement with the hybrid results of Liu et al (1991) is also
good but not as close as that shown in Fig. 8.

Three-Dimensional Case.

To test the results in the three-dimensional case of waves impinging on the shell at

angles other than 90 we consider first the results presented by Wong et al (1986) for a concrete
circular shell ( p0 = 2,240^/m3 , E0 = 16 x 109 7V/ra2 , i/0 = 0.2, h = O.ln = 0.0909r0)
embedded to a depth H = 2. On = 1.818r0 in a uniform half-space (pi = 2,665fcg/ra3 ,

EI = 0.69 x 109 AT/m2 , v\ - 0.45). Values for the hoop 00(r0 ) and longitudinal xx (r0 )
stresses on the pipewall (r = r0) were presented for obliquely incident P-, SV- and SH-waves
characterized by 6>h = 30 and 0V = 10 for 77 = u^/nfa = 0.132.

Comparisons for the amplitudes of the normalized stress 00(r0 ) for P-, SV- and SH-wave
are presented in Figs. 9a, 9b and 9c, respectively. The corresponding comparison for xx (r0 )
are presented in Figs. 9d, 9e and 9f. Clearly, there is good agreement between the two sets of
results. The results of Wong et al (1986) in Figs. 9a, b, c, d, e and f have been renormalized by
factors of 1.941, 1.082, 1.466, 1.834, 1.076 and 0.284, respectively, corresponding to our peak
values for the normalized stresses on r = r0 in the free-field.

Figs. 9a to 9f also show with segmented lines the normalized stresses 00 (a) and xx (a)
on the centerline r = a of the shell. In is apparent that large differences exist between 00 (r0 )
and 00 (a) for SV- and SH-waves indicating the importance of bending of the shell.

25-21
 As a second test of the results in the three-dimensional case we consider the results
presented by Liu et al (1991) for a concrete circular shell ( p0 2.24 x 103 kg/m3 , E0 =

1.6 x 10 10 7V/m2 , i/o = 0.2, h = O.ln = 0.0909r0) buried to a depth H = 5.0n - 4.545r0 in
a uniform half-space (pl = 2.664 x 103 kg/m3 , EI = 7.567 x 109 7V/m2 , ^ = 0.333). The

medium is subjected to non-vertically incident (#v = 30) P- and SV-waves impinging in the
direction of the pipeline (#h = 0). The dimensionless frequency corresponds to 77 = ur0/-K{3\
0.105.

The amplitudes of the normalized radial Ur (r0 ) and longitudinal Ux (r0 ) displacements
at r = r0 and of the normalized hoop stress 00 (a) on the centerline r = a are compared in
Fig. 10. The results of Liu et al (1991) for P-waves were renormalized by multiplying by factors
of 1.428, 0.573 and 2.325 which correspond to the peak values of Ur , Ux and 00 on the free-
field. The corresponding results for SV-waves were renormalized by factors 0.937, 2.482 and
1.621, respectively. Significant differences can be seen between the present results and those of
Liu et al (1991). The discrepancies are smaller for the dominant displacement components [Ur
for P-waves, Fig. lOa and Ux for SV-waves, Fig. lOe] than for the secondary displacements [Ux
for P-waves, Fig. lOb and Ur for SV-waves, Fig. 10d]. The discrepancies between the two sets
of hoop stresses 00 (a) for SV-waves (Fig. lOf) are particularly large.

The differences shown in Fig. 10 between our results and those of Liu et al (1991) for
the three-dimensional case are somewhat surprising considering the excellent agreement found
between the two sets of results for two-dimensional cases (Figs. 5 and 8). To confirm our
results we have recalculated our three-dimensional results by use of a hybrid approach which
combines a finite element model for the shell with an indirect boundary formulation for the
external half-space (FE/IBF, Luco and Barros, 1993b). In this case the shell is represented by
four concentric layers including 160 triangular elements in each layer. The displacements on the
contact area r = r0 calculated by the hybrid FE/IBF approach coincide almost exactly with the
results obtained by use of the present approach (DT/IBF). The moving Green's functions (Barros
and Luco, 1992, 1993) which have been used in both the DT/IBF and FE/IBF approaches have

25-22
been carefully tested. Calculations for the three-dimensional response of a cylindrical canyon
embedded in a uniform half-space and subjected to obliquely incident waves (Luco et al, 1990)

based on the use of the same Green's functions have been validated by subsequent calculations

by Zhang and Chopra (1991). Also, our three-dimensional results appear to agree with the earlier
results of Wong et al (1986) (Fig. 9). These considerations tend to reinforce the validity of our
present three-dimensional results. We note that the comparison with Liu et al (1991) involves

a case in which the three-dimensional effects are much stronger than in the comparison with

Wong et al (1986). The apparent horizontal speed of the excitation along the shell for SV-waves
in the case considered by Liu et al (#h = 0, Ov = 30) is c/fii = 2.0 while the corresponding
apparent speed for the case considered by Wong et al (0h = 30, #v = 10) is 6.65.

NUMERICAL RESULTS FOR A LAYERED MEDIUM

As an example we consider the response of a cylindrical concrete shell of circular cross

section embedded in a layered viscoelastic half-space. The shell of external radius r0 = 2.5m
and thickness h = 0.25m is characterized by E0 = 2.646 x 1010 7V/m2 , z/0 = 0.167 and

Po 2,500A;g/m3 . The soil is represented by four viscoelastic layers overlying a viscoelastic

half-space. The properties of the model are listed in Table 2. Two locations of the shell
are considered. In the first and second cases the centerline of the shell is located at depths
H = 11.5m (H/r0 = 4.6, first layer) and H 42m (H/r0 = 16.8, third layer), respectively.
Excitations in the form of non-vertically incident P-, SV- and SH-waves (#v = 30) impinging
normal (#h = 90) and along (0h = 0) the axis of the shell are considered. All calculations were
performed for a frequency of 10 Hz corresponding to a dimensionless frequency 77 uro/Trfis =
0.098. The response is normalized by the amplitude A of the incident displacement field at an
outcropping with the same properties as the underlying half-space. The normalized amplitudes
Ux = |ita;/j4|, Uy = \Uy/A\ and Uz = \uz /A\ on the free-field ground surface (x = y = z = 0)
in absence of the shell are listed in Table 3.

25-23
 The response in the two-dimensional case of P-, SV- and SH-waves impinging normal to

the axis of the shell (#h = 90) is illustrated in Fig. 11 for Ov = 30. The results shown include
the amplitudes of the normalized displacements Ur (r0 ), U0(r0 ) and Ux (r0) on the interface
between the shell and the soil (r = r0) and the amplitudes of the normalized hoop H00(a) and
shear Y,Qx (a) stresses on the centerline r = a.

Results for the three-dimensional case of non-vertically incident (9V = 30) P, SV and
SH-waves impinging along the axis of the shell (0h = 0) are shown in Figs. 12 and 13. The

results in Fig. 12 include the amplitudes of the normalized longitudinal Ux (r0 ), tangential U0(r0 )
and radial Ur (r0 ) displacements on the soil-shell interface (r r0). The results in Fig. 13 include
the amplitudes of the normalized logitudinal Hxx (a), tangential (hoop) 00 (a) and shear D0x (a)
stresses on the centerline of the shell (r = a)

CONCLUSIONS

A procedure has been presented to calculate the three-dimensional response of a cylin-

drical pipeline of infinite length embedded in a layered viscoelastic half-space and subjected to
obliquely incident waves. The procedure combines an indirect integral representation for the
field in the exterior half-space with a simplified Donnell shell theory for the pipeline or tunnel.
The procedure has been tested by comparison with previous solutions for a shell embedded in
a uniform half-space. The effects of layering have been illustrated by a set of new numerical

results for the two- and three-dimensional response of shells embedded in multilayered media
and subjected to P-, SV- and SH-waves.

Comparisons for the particular two-dimensional case of excitation impinging normal to

the axis of the shell indicate that the present results are consistent with earlier results of Lee
and Trifunac (1979), Balendra et al (1984) and Liu et al (1991) for SH-waves and with those of
Datta et al (1983, 1984), Wong et al (1986) and Liu et al (1991) for P- and SV-waves. These
comparisons confirm the accuracy of the present approach in the two-dimensional case. In the

25-24
three-dimensional case the situation is more controversial. The present results do agree very
closely with three-dimensional results for the stresses within the shell presented by Wong et
al (1986) but do not agree with the three-dimensional results of Liu et al (1991). However,

the present results for the three-dimensional case considered by Liu et al (1991) agree very
closely with a second set of results obtained by the authors by use of an hybrid approach
(Luco and Barros, 1993b) in which the shell was represented by a finite element model while
the exterior region was accounted for by means of an indirect boundary formulation based on

moving Greens's functions. The comparisons with the work of Wong et al (1986) and the

confirmatory results obtained by a second method suggest that the present approach is also valid
in the three-dimensional case.

ACKNOWLEDGMENTS

The work conducted here was supported by a Grant from California Universities for
Research in Earthquake Engineering (CUREe) as part of a CUREe-Kajima Project.

25-25
Table 1. Normalized Displacement Components for Non-Vertically Incident (6V = 30) P- and
SV-Waves Propagating in the Direction of the Axis of a Shell (#h = 0) for Different Numbers

of Source (7VS) and Observation (7V0) Points.

P-Wave SV-Wave
(N0 ,NS ) (AT0 ,7VS)
Variable Location (20, 40) (40, 80) (20, 40) (40, 80)

ux y/r0 = 0 1.0347 1.0347 3.4015 3.4015

(* = o) y/r0 = 1 1.0303 1.0303 3.4013 3.4013
2//r0 = 3 1.0036 1.0036 3.4007 3.4007

Uy y/r0 = 1 0.0520 0.0520 0.0087 0.0087

(* = 0) y/r0 = 3 1.0216 1.0215 0.0167 0.0167

uz y/r0 = 0 2.0071 2.0070 0.0846 0.0846

(* = 0) y/r0 = l 1.9884 1.9884 0.0874 0.0873
2//r0 = 3 1.8881 1.8881 0.1093 0.1093

ux (9-0 0.3565 0.3565 2.0488 2.0487

(r = r0 ) (9-45 0.6606 0.6606 2.4959 2.4959
(9 = 90 0.7847 0.7847 2.6699 2.6699

Ue (9 = 0 1.2737 1.2737 0.8990 0.8990

(r = r0 ) 0 = 45 1.1403 1.1404 0.5430 0.5430

Ur (9-0 0.0239 0.0238 0.2444 0.2446

(r = r0 ) (9 = 45 1.2153 1.2153 0.7654 0.7656
(9 = 90 1.8732 1.8733 0.8865 0.8865

00 9 = G 10.141 10.140 2.9166 2.9204

(r -a) (9 = 45 5.6199 5.6198 3.4289 3.4314
(9-90 1.6686 1.6684 5.4410 5.4436
0 - 270 2.0929 2.0921 5.1771 5.1800

25-26
Table 2. Properties of the Layered Soil Model.

Medium H
m m/sec
P a
m/sec
P
kg/m3
.-,
1 23 185 827 1,900 0.01

2 11 305 1,304 1,800 0.01

3 16 370 1,428 1,800 0.01

4 28 443 1,478 1,800 0.01

5 oo 510 1,634 1,800 0.01

Table 3. Normalized Amplitudes of Free-Field Displacements on the Ground Surface

(x = y = z = 0) for Non Vertically Incident P-, SV- and SH-waves (0V = 30) for Oh = 90 and
0 - 0.

[0h = 90] [Oh = 0]

Wave Ox Uy uz ux Uy uz
P 0 0.473 3.036 0.473 0 3.036
SV 0 3.366 0.073 3.366 0 0.073
SH 2.870 0 0 0 2.870 0

25-27
 BIBLIOGRAPHY
Buried Pipelines and Lined Tunnels

1. Mow, C. C. and W. L. McCabe (1963). Dynamic Stresses in an Elastic Cylinder, Journal

of the Engineering Mechanics Division, ASCE, 89, EMS, 21-41.

2. Mente, L. J. and F. W. L. French (1964). Response of Elastic Cylinders to Plane Shear

Waves, Journal of the Engineering Mechanics Division, ASCE, 90, EMS, 103-118.

3. Kuesel, T. R. (1969). Earthquake Design Criteria for Subways, Proc. ASCE, 95, ST6,
1213-1231.

4. Sakurai, A. and Takahashi (1969). Dynamic Stresses of Underground Pipelines During

Earthquakes, Proc. 4th World Conference on Earthquake Engineering, Santiago, Chile,
January, Vol. 2, B4-81 to B4-96.

5. Duke, C. M. (1971). Damage of Water Supply Systems, The San Fernando, California
Earthquake of Feb. 9,1971, Geological Survey Professional Paper 733, U.S.G.S. and

NOAA, 225-240.

6. King, P. V. and J. M. Betz (1972). Earthquake Damage to a Sewer System, Journal of

Water Pollution Control Federation, (WPCF), May, 859-867.

7. Aoki, Y. (1973). Seismic Design Spectra for Trench Type Tunnel, Proc. Japan Society
of Civil Engineering, No. 211.

8. Kubo, K. (1973). Behaviour of Underground Waterpipes During an Earthquake, Proc. of

the Fifth World Conference on Earthquake Engineering, Roma, Vol. 1, 569-587.

9. NOAA/EERI Subcommittee on Water and Sewerage Systems (1973). Earthquake Dam-

age to Water and Sewer Facilities, San Fernando Earthquake of 1971, Vol. 2 - Utilities,
Transportation and Sociological Aspects, (Editor: C. M. Murphy), National Oceanic and

Atmospheric and Administration, 73-193.

25-28
10. Okamoto, S. and C. Tamura (1973). Behavior of Subaqueous Tunnels During Earthquakes,
Earthquake Engineering and Structural Dynamics, J_, 253-266.

11. Pao, Y. H. and C. C. Mow (1973). The Diffraction of Elastic Waves and Dynamic Stress
Concentration, Crane and Russak, New York.

12. Richardson, C. B. (1973). Damage to Utilities, The Great Alska Earthquake of 1964 -
Engineering, Written by Committee of Alaska Earthquake, National Academy of Science,

1034-1073.

13. Toki, K. and S. Takada (1974). Earthquake Response Analysis of Underground Tubular
Structures, Bulletin of Disaster Prevention Research Institute, Kyoto University, Vol. 24,
June, 107-125.

14. Yeh, G. C. K. (1974). Seismic Analysis of Slender Buried Beams, Bulletin Seismological
Society of America, 64, 1551-1562.

15. Katayama, T., K. Kubo and N. Sato (1975). Earthquake Damage to Water and Gas
Distribution Systems, Proc. of U.S. National Conference on Earthquake Engineering, Ann
Arbor, Michigan, EERI, 396-405.

16. Newmark, N. M. and W. J. Hall (1975). Pipeline Design to Resist Large Fault Dis-
placement, Proc. of U.S. National Conference on Earthquake Engineering, Ann Arbor,

Michigan, EERI, 416-425.

17. Parmelee, R. and C. Ludtke (1975). Seismic Soil-Structure Interation of Buried Pipelines,
Proc. of U.S. National Conference on Earthquake Engineering, Ann Arbor, Michigan,
EERI, 406-415.

18. Patel, Y. A., F. L. Cho and A. P. Dimopoulos (1975). Pipeline Designs for Earthquake
Zones, Pipeline and Gas Journal, December.

25-29
19. Tamura, C, S. Okamoto and M. Hamada (1975). Dynamic Behavior of Submerged Tunnel
During Earthquakes, Report of the Institute of Industrial Science, 24, 169-246.

20. ASME Guide for Gas Transmission and Distribution Piping Systems, New York, 1976.

21. Katayama, T. (1976). Effect of Ground Conditions on Seismic Damage to Buried

Pipelines, Proc. of the U.S.-Japan Seminar on Earthquake Engineering Research with
Emphasis onLifeline Systems, November, Tokyo, 197-210.

22. Narita, K. (1976). Study on Pipeline Failure due to Earthquakes, Proc. of the U.S.-Japan
Seminar on Earthquake Engineering, (Editors: K. Kubo and P. C. Jennings), 157-176.

23. Popelar, C., A. S. Rosenfield and M. F. Kanninen (1976). Steady State Crack Propagation
in Pressurized Pipelines, ASME Paper No. 76-PVP-7, September 1976.

24. Kennedy, R. P. and R. A. Williamson (1977). Fault Movement Effects on Buried Oil
Pipeline, Transportation Engineering Journal, ASCE, 103(TE5), 617-633.

25. Kennedy, R. P., A. C. Darrow and S. A. Short (1977). General Considerations for Seismic
Design of Oil Pipeline Systems, Lifeline E. E., The Current State of Knowledge, Proc.
of ASCE Specialty Conference, Los Angeles, 252-266.

26. Kubo, K., T. Katayama and A. Ohashi (1977). Present State of Lifeline Earthquake
Engineering in Japan, Lifeline E. E., The Current State of Knowledge, Proc. of ASCE
Specialty Conference, Los Angeles, 118-133.

27. Shinozuka, M. and H. Kawakami (1977). Underground Pipe Damages and Ground Char-
acteristics, Proc. of ASCE Specialty Conference on Lifeline Earthquake Engineering, Los
Angeles, August, 293-307.

28. Shinozuka, M. and H. Kawakami (1977). Ground Characteristics and Free-Field Strains,
Technical Report No. CU-2, Columbia University.

25-30
29. Wang, L. R. L. and M. O'Rourke (1977). State of the Art of Buried Lifeline Earthquake
Engineering, Proc. of the Current State of Lifeline Earthquake Engineering, ASCE, 252-
266.

30. Nelson, L and P. Weidlinger (1978). Dynamic Seismic Analysis of Long Segmented Life-
liness, Trans. ofASME, Jopurnal of Pressure Vessel Technology, Paper No. 78-WA/PVP-4,
December, 1-11.

31. Novak, M. and A. Hindy (1978). Dynamic Response of Buried Pipelines, Proc. of the 6th

European Conference on Earthquake Engineering, Dubrovnik, Yugoslavia, September,

1978 2, 533-540.

32. Sanchez-Sesma, F.J. and J. A. Esquivel (1978). Difraccion de Ondas SH por Tunels,

Institute de Ingenieria , UNAM, Mexico.

33. Wang, L. and K. Cheng (1978). Seismic Response Behaviour of Buried Pipelines, ASME
Annual Winter Convention, San Francisco, California, December, 11-15.

34. Ariman, T. and G. E. Muleski (1979). Recent Developments in Seismic Analysis of Buried
Pipelines, Proc. of the U. S. National Congress on Earthquake Engineering, Stanford
University, 643.

35. Ariman, T., S. C. Liu and R. E. Nickell (Eds.) (1979). Lifeline Earthquake Engineering
- Buried Pipelines, Seismic Risk and Instrumentation, AMES, New York, 285.

36. Constantopoulos, I. V. et al (1979). Dynamic Analysis of Tunnels, Proc. 3rd Int. Conf.
Num. Methods in Geomechanics, Aachen, Germany, 2, 841-848.

37. Hindy, A. and M. Novak (1979). Earthquake Response of Underground Pipelines, Earth-
quake Engineering and Structural Dynamics, 7, 451-476.

25-31
38. Lee, V. W.. and M. D. Trifunac (1979). Stresses and Deformations near Circular Un-
derground Tunnels Subjected to Incident SH-Waves, J. of the Engineering Mechanics

Division, ASCE, 105, 643-659.

39. Muleski, G. E., T. Ariman and C. P. Auman (1979). A Shell Model of a Buried Pipe in a
Seismic Environment, Journal of Pressure Vessel Technology, Transaction of the ASME,
101(1), 44-50.

40. Nelson, I. and P. Weidlinger (1979). Dynamic Seismic Response of Long Segmented
Lifelines, Journal of Pressure Vessel Technology, Transactions of the ASME, 101(1),
10-20.

41. Novak, M. and A. Hindy (1979). Seismic Response of Buried Pipelines, Proc. of the

Third Canadian Conference on Earthquake Engineering, June, Montreal.

42. O'Rourke, M. and L. R. L. Wang (1979). Earthquake Response of Buried Pipeline,

Proc. of the ASCE Specialty Conference on Earthquake Engineering and Soil Dynamics,
Pasadena, California, June, 720-731.

43. Constantopoulos, I. V. et al (1980). Seismic Analysis of Buried Tunnels, Proc. 7th World
Conf. on Earthquake Engineering, Istanbul, Turkey, 8, 193-200.

44. Datta, S. K. and N. El-Akily (1980). Seismic Response of a Buried Pipe in an Infinite
Medium, paper presented at the Century 2 ETC Pressure Vessel and Piping Technology
Conference, August 12-15, San Francisco, California.

45. El-Akily, N. and S. K. Datta (1980). Response of Circular Cylindrical Shell to Distur-
bances in a Half-space, Earthquake Engineering and Structural Dynamics, 8, 469-477.

46. Hindy, A. and M. Novak (1980). Response of Pipelines to Random Ground Motion,
Jornal of the Engineering Mechanics Division, ASCE.

25-32
47. Lee, L. H. N., T. Ariman and C. C. Chen (1980). Axisymmetric Buckling of Buried
Pipelines by Seismic Excitations, Technical Report No. 5, ERADUPS Project, University
of Notre Dame, December 1979 and ASME paper No. 80-C2/PVP-75.

48. Novak, M. and A. Hindy (1980). Seismic Analysis of Underground Tubular Structures,
Proc. of the 7th World Conf. on Earthquake Engineering, Istanbul, Turkey, 8, 287-294.

49. Takada, S. (1980). Seismic Response Analysis of Buried PVC and Ductile Iron Pipelines,
Recent Advances in Lifeline Earthquake Engineering in Japan, (Editors: H. Shibata, T.

Katayama and T. Ariman), ASME, 23-32.

50. Ariman, T. and G. E. Muleski (1981). A Review of the Response of Buried Pipelines
under Seismic Excitations, Earthquake Engineering and Structural Dynamics, 9, 133.

51. Datta, S. K., A. H. Shah and N. El-Akily (1981). Dynamic Behavior of a Buried Pipe in

a Seismic Environment, Journal of Applied Mechanics, 49, 141-148.

52. Datta, S. K., A. H. Shah and N. El-Akily (1981). Dynamic Behavior of Embedded
Pipelines, in Seismic Risk Analysis and Its Application to Reliability Based Design of
Lifeline Systems, Proc. of Review Meeting of the U.S.-Japan Cooperative Research, Hon-

olulu, Hawaii, 143-172.

53. El-Akily, N. and S. K. Datta (1981). Response of Circular Cylindrical Shell to Dis-
turbances in a Half-space: Numerical Results, Earthquake Engineering and Structural
Dynamics, 9, 477-489.

54. Muleski, G. E. (1981). Seismic Response and Fracture of Buried Pipelines, Ph. D. Dis-
sertation, submitted to the University of Notre Dame.

55. Parnes, R. and P. Weidlinger (1981). Dynamic Interaction of an Embedded Cylindrical

Rod under Axial Harmonic Forces, Int. Journal of Solids and Structures, 17, 915-924.

25-33
56. Wang, L. R. L. (1981). Seismic Evaluation Model for Buried Lifelines, Proc. of the
Second Specialty Conf. on Lifeline Earthquake Engineering, ASCE, 335-347.

57. Wang, L. R. L., T. Okubo, E. Kuribayaski, T. Iwasahi and O. Ueda (1981). Lifeline
Earthquake Engineering Literatures in Japan, Technical Report CEE - 8025172, School of
Engineering and Environmental Science, University of Oklahoma, Norman, Oklahoma.

58. Datta, S. K. and A. H. Shah (1982). Dynamic Response of Buried Pipelines and Tun-

ne\s,Earthquake Ground Motion and Its Effects on Structures, (Editor: S. K. Datta),

ASME, New York, 53, 181-197.

59. Christiano, P. P. et al (1983). Response of Buried Tunnels to Earthquake Excitation,

Trans. 7th Int. Conf. Struct. Mech. Reactor Technology, Vol. K, 287-294.

60. Datta, S. K., T. Chakraborty and A. H. Shah (1983). Dynamic Responses of Pipelines to
Moving Loads, Earthquake Behavior and Safety of Oil and Gas Storage Facilities, Buried
Pipelines and Equipment, (Editor: T. Ariman), ASME, New York, 77, 246.

61. Datta, S. K., A. H. Shah and K. C. Wong (1983). Dynamic Amplification of Stresses
and Displacements Induced in a Buried Pipe in a Semi-Infinite Medium, Technical Re-
port CUMER-83-3, Dept. of Mechanical Engineering, University of Colorado, Boulder,
Colorado.

62. Muleski, G. E. and T. Ariman (1983). Improved Seismic Input and Soil Models and
Quasi-Static Approximation in Buried Pipes, Earthquake Behavior and Safety of Oil and
Gas Storage Facilities, Buried Pipelines and Equipment, (Editor: T. Ariman), ASME,
New York, 77, 202-208.

63. Wang, L. R. L. (1983). Role and Development of Soil Parameters for Seismic Response
of Buried Lifelines, Earthquake Behavior and Safety of Oil and Gas Storage Facilities,

Buried Pipelines and Equipment, (Editor: T. Ariman), ASME, New York, 77, 312-323.

25-34
64. Balendra, T., D. P. Thambiratnam, C. G. Koh and S-L Lee (1984). Dynamic Response of
Twin Circular Tunnels due to Incident SH-Waves, Earthquake Engineering and Structural
Dynamics, 12, 181-201.

65. Datta, S. K., P. M. O'Leary and A. H. Shah (1984). Dynamic Responses of Buried
Pipelines to Incident Longitudinal and Shear Waves, Technical Report CUMER-84-3,
Dept. of Mechanical Engineering, University of Colorado, Boulder, Colorado.

66. Datta, S. K., A. H. Shah and K. C. Wong (1984). Dynamic Stresses and Displacements
in Buried Pipe, Journal of Engineering Mechanics, 110(10), 1451-1466.

67. Franssens, G. R. and P. E. Lagasse (1984). Scattering of Elastic Waves by Cylindrical

Obstacles, Journal of the Acoustical Society of America, 76, 1535-1542.

68. Gomez-Masso, A. and I. Atalla (1984). Finite Element versus Simplified Methods in
the Seismic Analysis of Underground Structures, Earthquake Engineering and Structural
Dynamics, 12, 347-367.

69. Wong, K. C., A. H. Shah, S. K. Datta and P. M. O'Leary (1984). Dynamic Amplification
of Displacements and Stresses around Buried Pipelines and Turu\els,Earthquake Source
Modeling, Ground Motion and Structural Response, (Editor: S. K. Datta), ASME, New

York, 80, 147-161.

70. Datta, S. K., P. M. O'Leary and A. H. Shah (1985). Three-Dimensional Dynamic Re-
sponse of Buried Pipelines to Incident Longitudinal and Shear Waves, Journal of Applied
Mechanics, 52(4), 919-926.

71. Muleski, G. E. and T. Ariman (1985). A Shell Model for Buried Pipes in Earthquakes,
Soil Dynamics and Earthquake Engineering, 4(1), 43-51.

72. O'Leary, P. M. and S. K. Datta (1985). Dynamic Response of a Buried Pipeline at Low
Frequencies, ASME Journal of Pressure Vessel Technology, 107, 44-50.

25-35
73. O'Leary, P. M. and S. K. Datta (1985). Dynamics of Buried Pipelines, Soil Dynamics
and Earthquake Engineering, 4, 151-159.

74. Wong, K. C., A. H. Shah and S. K. Datta (1985). Dynamic Stresses and Displacements

in a Buried Tunnel, Journal of Engineering Mechanics, 111(2), 218-234.

75. Wong, K. C., S. K. Datta and A. H. Shah (1986). Three Dimensional Motion of a Buried

Pipeline I. Analysis, Journal of Engineering Mechanics, 112(12), 1319-1338.

76. Wong, K. C., A. H. Shah and S. K. Datta (1986). Three Dimensional Motion of a Buried
Pipeline H. Numerical Results, Journal of Engineering Mechanics, 112(12), 1338-1345.

77. Chin, Y. K, R. K. N. D. Rajapakse, A. H. Shah and S. K. Datta (1987). Dynamics

of Buried Pipes in Back-Filled Trench, Soil Dynamics and Earthquake Engineering, 6,
158-163.

78. Singh, V. P., P. C. Upadhyay and B. Kishor (1987). On the Dynamic Response of Buried
Orthotropic Cylindrical Shells, Journal of Sound and Vibrations, 113(1), 101-115.

79. Singh, V. P., P. C. Upadhyay and B. Kishor (1987). A Comparison of Thick and Thin
Shell Theory Results for Buried Orthotropic Cylindrical Shells, Journal of Sound and

Vibrations, 119(2), 339-345.

80. Liu, S. W., K. R. Khair and A. H. Shah (1988). Three Dimensional Dynamics of Pipelines
Buried in Back-Filled Trenches due to Oblique Incidence of Body Waves, Technical
Report CUMER-88-4, DepL of Mechanical Engineering, University of Colorado, Boulder,
Colorado.

81. Upadhyay, P. C. and B. K. Mishra (1988). Non-Axisymmetric Dynamic Response of

Buried Orthotropic Cylindrical Shells, Journal of Sound and Vibrations, 121(1), 149-160.

25-36
82. Upadhyay, P. C. and B. K. Mishra (1988). Non-Axisymmetric Dynamic Response of
Buried Orthotropic Cylindrical Shell Due to Incident Shear Waves, Journal of Sound and
Vibrations, 125(2), 227-239.

83. Dwivedi, J. P. and P. C. Upadhyay (1989). Effect of Imperfect Bonding on the Axisym-
metric Dynamic Response of Buried Orthotropic Cylindrical Shells, Journal of Sound
and Vibrations, 135(3), 477-486.

84. Hadley, P. K., A. Askar and A. S. Cakmak (1989). Scattering of Waves by Inclusions in
a Nonhomogeneous Elastic Half Space Solved by Boundary Element Methods, NCEER-
89-0027, National Center for Earthquake Research, State University of New York at
Buffalo, Vol 1.

85. Chen, H. L., S. P. Shah and L. M. Keer (1990). Dynamic Response of Shallow-Buried
Cylindrical Structures, Journal of Engineering Mechanics, 116(1), 152-171.

86. Dwivedi, J. P. and P. C. Upadhyay (1990). Effect of Fluid Presence on the Dynamic
Response of Imperfectly Bonded Buried Orthotropic Cylindrical Shells, Journal of Sound
and Vibrations, 139(1), 99-110.

87. Balendra,T., C. G. Koh and Y. C. Ho (1991). Dynamic Response of Buildings due to

Trains in Underground Tunnels, Earthquake Engineering and Structural Dynamics, 20,
275-291.

88. Dwivedi, J. P. and P. C. Upadhyay (1991). Effect of Imperfect Bond on the Dynamic

Response of Buried Orthotropic Cylindrical Shells under Shear-Wave Excitation, Journal

of Sound and Vibrations, 145(2), 333-337.

89. Liu, S. W., K. R. Khair and A. H. Shah (1991). Three Dimensional Dynamics of Pipelines
Buried in Backfilled Trenches due to Oblique Incidence of Body Waves, Soil Dynamics
and Earthquake Engineering, 10(4), 182-191.

25-37
90. Cheung, Y. K. and J. X. Zhu (1992). Dynamic Interaction Analysis of a Circular Cylin-
drical Shell of Finite Length in a Half-Space, Earthquake Engineering and Structural
Dynamics, 21, 799-809.

91. Chua, K. H., T. Balendra and K. W. Lo (1992). Groundborne Vibrations due to Trains
in Tunnels, Earthquake Engineering and Structural Dynamics, 21, 445-460.

92. Dwivedi, J. P., B. K. Mishra and P. C. Upadhyay (1992). Non-Axisymmetric Dynamic

Response of Imperfectly Bonded Buried Orthotropic Pipelines Due to an Incident Shear
Wave, Journal of Sound and Vibrations, 157(1), 81-92.

93. Navarro, C. (1992). Seismic Analysis of Underground Structures, Proc. Tenth World Conf.
on Earthquake Engineering, Madrid, Spain, 3, 1939-1944.

94. Penzien, J., C. H. Chen, Y. J. Lee and W. Y. Jean (1992). Seismic Analysis of Retangular
Tunnels in Soft Ground, Proc. Tenth World Conf. on Earthquake Engineering, Madrid,
Spain, 3, 1619-1624.

95. Luco, J. E. and F. C. P. de Barros (1994). Seismic Response of a Cylindrical Shell

Embedded in a Layered Viscoelastic Half-Space. I: Formulation, Earthquake Engineering
and Structural Dynamics, 23, 553-567.

96. Barros, F. C. P. de and J. E. Luco (1994a). Seismic Response of a Cylindrical Shell

Embedded in a Layered Viscoelastic Half-Space. II: Validation and Numerical Results,

Earthquake Engineering and Structural Dynamics, 23, 569-580.

25-38
97. Luco, J. E. and F. C. P. de Barros (1993). Three-Dimensional Response of a Cylindrical
Tunnel Embedded in a Layered Medium, (to be submitted for publication}.

Additional References

1. Barros, F. C. P. de and J. E. Luco (1992). Moving Green's Functions for a Layered Vis-

coelastic Half-Space, Report, Dept. ofAppl. Mech. & Engng. ScL, University of California,
San Diego, La Jolla, California, 210 pp.

2. Barros, F. C. P. de and J. E. Luco (1994b). Response of a Layered Viscoelastic Half-Space

to a Moving Point Load, Wave Motions, 19, 189-210.

3. Donnell, L. H. (1933). Stability of Thin Walled Tubes under Torsion, NACA Report,

No. 479.

4. Luco, J. E. and F. C. P. de Barros (1993). On the Appropriate Depth Dependence for

Plane Waves Reflected in a Viscoelastic Half-Space, (accepted for publication).

5. Luco, J. E. and H. L. Wong (1987). Seismic Response of Foundations Embedded in a

Layered Half-Space, Earthquake Engineering and Structural Dynamics, 15(2), 233-247.

6. Luco, J. E., H. L. Wong and F. C. P. de Barros (1990). Three-Dimensional Response of

a Cylindrical Canyon in a Layered Half-Space, Earthquake Enginnering and Structural
Dynamics, 19(6), 799-817.

7. Zhang, L. and A. K. Chopra (1991). Three-Dimensional Analysis of Spatially Varying

Ground Motion around a Uniform Canyon in a Homogeneous Half-Space, Earthquake
Engineering and Structural Dynamics, 20, 911-926.

25-39
 (a) (c)

(b) (d)

Sources Observers

Figure 1. Schematic Representation of Cylindrical Shell Embedded in a Layered Half-Space,

(a) Top view showing horizontal angle of incidence, (b) Cross section, (c) Free-field model
showing vertical angle of incidence and (d) Location of sources and observers.

25-40
 - cr.

Figure 2. Schematic Representation of Shell Showing Local Coordinates.

25-41
 "1II1IIT~ I I I I I I I
Ux (a) Uv (b)
3-

GO

i i i i i i i
0 -202 90 180 270 360 0 90 180 270 360
y/r0 0 (degrees) 0 (degrees)

Figure 3. (a) Normalized Longitudinal Displacement Ux on the Ground Surface z = 0, (b)

Normalized Displacement Ux on the Soil-Shell Interface r = r0 and (c) Normalized Longitudinal
Shear Stress Erx on r = r0 for a Vertically Incident SH-Wave (#h = 90, 0V = 0) Impinging on
a Shell (h = O.lr0) Embedded to a Depth H = 1.5r0 in a Uniform Half-Space. Present results
are shown with solid lines; the results of Lee and Trifunac (1979) are shown with open circles.
In Fig. 2c the results of Lee and Trifunac have been divided by a factor of six (77 = 0.5).

.6

.4 -

90 180 270 360 0 90 180 270 360 90 180 270 360

0 (degrees) 0(degrees) 0 (degrees)

Figure 4. (a) Normalized Longitudinal Displacement Ux at r = r0, (b) Tangential Stress Srx
at r = r0 and (c) Shear Stress 0X at r = n for a Nonvertically Incident SH-Wave (0V = 30)
Impinging Normal (0h = 90) to a Shell (h = O.lr0) Embedded to a Depth H = 2.5r0 in a
Uniform Half-Space. Present results are shown with solid lines, those of Balendra et al (1984)
are shown with open circles (77 = 0.245).

25-42
 0 90 180 270 360 0 90 180 270 360 0 90 180 270 360
6 (degrees) 6 (degrees) 6 (degrees)

Figure 5. (a) Normalized Longitudinal Displacement Ux at r = r0, (b) Tangential Stress Erz at
r = r0 and (c) Shear Stress 0X at r = n for a Vertically Incident SH-Wave (0V = 0) Impinging
Normal (0h = 90) to a Shell (/i = 0.0909r0) Embedded to a Depth H = 4.545r0 in a Uniform
Half-Space. Present results are shown with solid lines, those of Liu et al (1991) are shown with
open circles (rj = 0.105).

25-43
 P-WAVE SV-WAVE
I I I I I I ! i i i i i i i
- Ur (a) _Ur (b)

20

0 i i i i i i i
0 90 180 270 360 0 0 360
0 (degrees) 6(degrees)

Figure 6. (a), (b) Normalized Radial Displacements Ur at r = r0 and (c), (d) Normalized Hoop
Stresses E#0 at r = r0 (solid lines) and r = a (segmented lines) for Vertically Incident (0V = 0)
P- and SV-Waves Impinging Normal (0h = 90) to a Shell (h = 0.0909r0) Embedded to a Depth
H = 7.573r0 in a Uniform Half-Space. Present results are shown with solid lines or segmented
lines, results of Datta et al (1983) are shown with open circles (77 = 0.132).

25-44
 P-WAVE SV-WAVE
25

0,
90 180 270 360 V0 90 180 270 360
6 (degrees) 6(degrees)

Figure 7. (a)^ (b) Normalized Hoop Stresses 00 at r = r0 (soh'd lines) and r = a (segmented
lines) and (c), (d) Normalized Longitudinal Stresses EIX at r = r0 (solid lines) and r = a
(segmented lines) for Nonvertically Incident (0V = 10) P- and SV-Waves Impinging Normal
(0h = 90) to a Shell (h = 0.0909r0) Embedded to a Depth H = 1.818r0 in a Uniform Half-
Space. Present results are shown with solid lines or segmented lines, results of Wong et al (1986)

for r = r0 are shown with open circles (77 = 0.132).

25-45
 P-WAVE SV-WAVE
i i i i i i i
Ur (b)
3 -

I I I I I I I
See (d)

0.
0 90 180 270 360 0 0 360
0 (degrees) 0 (degrees)

Figure 8. (a), (b) Normalized Radial Displacements Ur at r = r0 and (c), (d) Normalized Hoop
Stresses 00 at r = a for Vertically Incident (0V = 0) P- and SV-Waves Impinging Normal
(0h = 90) to a Shell (h = 0.0909r0) Embedded to a Depth H = 4.545r0 in a Uniform Half-
Space. Present results are shown with solid lines; the results of Liu et al (1991) are shown with
open circles (77 = 0.105).

25-46
 P-WAVE SV-WAVE SH-WAVE

0 90 180 270 360 0 90 180 270 360 0 90 180 270 360

6 (degrees) 6 (degrees) 6 (degrees)

Figure 9. (a), (b), (c) Normalized Hoop Stresses and (d), (e), (f) Normalized Axial Stresses Exx
for Nonvertically Incident (<9V = 10) P- ,SV- and SH-Waves Impinging Obliquely (ft, = 30)
on a Shell (h = 0.0909r0) Embedded to a Depth H = 1.818r0 in a Uniform Half-Space. The
present results at r = r0 are shown with solid lines while those at r = a are shown with
segmented lines. The results of Wong et al (1986) at r = r0 are shown with open circles
(77 = 0.132).

25-47
 I I I I ~i i i i i i r~
U P-WAVE (b) _ Z00 P-WAVE (c)

1I I I I I I i I i I i I
Ur SV-WAVE (d) Ux SV-WAVE (e) SV-WAVE (f) _

0
0 90 180 270 360 "0 90 180 270 360 "0 90 180 270 360
0 (degrees) 0 (degrees) 6 (degrees)

Figure 10. (a), (d) Normalized Radial Displacements Ur at r = r0 , (b), (e) Normalized Axial
Displacement Ux at r = r0 and (c), (f) Normalized Hoop Stresses ## at r = a for Nonvertically
Incident (0V = 30) P- and SV-Waves Impinging with Angle 0h = 0 on a Shell (h = 0.0909r0)
Embedded to a Depth H = 4.545r0 in a Uniform Half-Space. Present results are presented by
solid lines (DT/ffiF) and solid dots (HE/DBF). The results of Liu et al (1991) are shown with

open circles (77 = 0.105).

25-48
 P-WAVE SV-WAVE

2 -

o0 90 180 270 360 0 90 180 270 360 0 90 180 270 360

e (degrees) 0 (degrees) 0 (degrees)

Figure 11. Normalized Radial C/r, Tangential Ue, and Longitudinal Ux Displacements at r = r0
and, Hoop Stresses Eee and Shear Stresses T,0x at r = a for Nonvertically Incident P-, SV- and
SH-Wave (0V = 30) Impinging Normal (0h = 90) to a Shell (h = O.lr0) Embedded to Depths
H/r0 = 4.6 (segmented lines) and 16.8 (solid lines) in a Multilayered Half-Space (77 = 0.098).

25-49
 P-WAVE SV-WAVE SH-WAVE
1.2 ~i i r I I I 1.2 1 1 1 1 1 1 1 .4 Tiiiir~
Uv (a). -Ux (d)_ U (g)

- .2
/"N-
1^' "xx^ f/ \

.0
i i i i i i
.0 1 1 1 1 I 1 1 .0

1 I I I I I I
Ur (f)

oo 90 180 270 360 0 90 180 270 360 0 90 180 270 360

0 (degrees) 0 (degrees) 0 (degrees)

Figure 12. Normalized Longitudinal Ux, Tangential Ue, and Radial Ur Displacements at r = r0
for Nonvertically Incident (0V = 30) P-, SV- and SH-Wave Impinging with Angle 0h = 0 on
a Shell (h = O.lr0) Embedded to Depths H/r0 = 4.6 (segmented lines) and 16.8 (solid lines)
in a Multilayered Half-Space (T? = 0.098).

25-50
 P-WAVE SV-WAVE SH-WAVE
20 IIIIIF 40 I T "I I T~ 10 1 I I I I T~
(a) (d) x (g)
15 30

10 20

10

i i i i i i i i i i i i
0

50 i i i i 10 i i i i i r 20 i i i i i i
(e)
40
15
30
10
20
5
10

i i i i i i i I I I I I t I
0 0 0
20 I i r i 50 I I I I I I 30 i i i i
(c)
40
(f) (i) _
15
20
30
10
20
10
10

i i i i i i i
0 0
90 180 270 360 0 90 180 270 360 0 90 180 270 360
6 (degrees) 6 (degrees) 6 (degrees)

Figure 13. Normalized Axial Exx , Hoop 00, and Shear 0X Stresses for Nonvertically Incident
(0V = 30) P-, SV- and SH-Wave Impinging with Angle 0h = 0 on a Shell (h = O.lr0)
Embedded to Depths H/r0 = 4.6 (segmented lines) and 16.8 (solid lines) in a Multilayered
Half-Space (77 = 0.098).

25-51