Seismic Behavior of

Large Panel Precast

Concrete Walls:
Analysis and Experiment
Ray W. Clough
Nishkian Professor Emeritus
Department of Civil Engineering
University of California
Berkeley, California

Faris Malhas
Former Graduate Student
Department of Civil Engineering
University of Wisconsin
Madison, Wisconsin

Michael G. Oliva
Associate Professor
Department of Civil Engineering
University of Wisconsin
Madison, Wisconsin

S tructures have traditionally been de-

signed for earthquake resistance by
providing a sufficient strength level to
dering the problem of how to provide a
usable product and caused design en-
gineers to become reluctant to design
resist a prescribed static lateral load. precast structures in seismic regions.
Though previous codes only defined the The competitive advantages of precast
required lateral loading, there was an systems in building construction have
implicit assumption that a certain created a demand for their use in seis-
amount of post-yield deforinability mic as well as nonseismic locations.
would be available within the structure Moreover, sonic areas of previous non-
while a yield level resisting capacity seismic use have recently been rezoned
was maintained. Recent codes have spe- as seismic regions and seismic resis-
cifically addressed the need for ductility tance design requirements will now
to be provided during the design in often control over wind design criteria.
conjunction with the specified design The problem in design is that precast
Ioading. It is unclear, however, what systems typically have weak connection
level of ductility might actually be regions. They cannot currently be easily
available in precast concrete construc- joined in a manner which would re-
tion. These recent changes have left semble monolithic concrete and still be
producers of precast components pon- assembled efficiently. Their behavior

42
under strong seismic excitation will
generally be substantially different from Synopsis
monolithic systems, but their actual
ductility level and failure sequence has The demand for economically
not been well documented yet. The re- competitive precast building sys-
search program described here investi- tems has been increasing around
gated the performance of one type of the world at the same time that con-
precast system, the panelized building cerns have grown regarding their
system, under replicated seismic motion seismic resistance capacities. Re-
to determine the demand for ductility cent codes have rezoned portions
which would he produced and the fail- of the United States to reflect higher
ure condition which might develop in a seismic hazard and to require seis-
wall of a prototype 13-story building. mic resistant design where wind
loading prevailed in the past. De-
signers have been left in the dilem-
Use of Precast Construction
ma of having to provide seismic re-
Precast concrete has been used in sistant design without having code
markets around the world to satisfy the provisions specifically addressing
tremendous demand for housing by pro- the unique characteristics of precast
viding a rapidly built system using fac- construction.
tory fabricated quality controlled com- The seismic resistant capacities
ponents. The competitive edge gained of one form of precast construction,
b y precast manufacturers in the United the large panel wall system, are
States and other countries has come described in this report. An inves-
from the development of refined modu- tigation consisting of shaking table
lar building systems with standardized tests with earthquake motion and
components and simple connections. subsequent analytical investigations
The panelized system is an example of the seismic response are in-
of a method which uses standardized cluded.
wall and floor/roof precast panels con- A set of conclusions and sugges-
nected together to form a complete box tions for improved performance of
type structure without a separate precast large panel construction are
framework., While individual panels in given, based on the tests, analytical
the system may be connected together simulations, and results from other
by a variety of means, all the connection researchers, along with a detailed
schemes tend to form points of weak- discussion of problems in precast
ness in the structure. concrete design and behavior of the
precast concrete system.
Seismic Resistance and Design
of Precast Structures
Many of the precast systems in use
throughout the world are not well suited
to resist the force and deformation de- systems for lateral resistance in strong
mands caused by earthquake loading. seismic areas has been precluded in
The systems were often originally de- most areas of the United States by code
signed for nonseismic regions, but their provisions which would require con-
advantages led to later use in low rise nections similar to monolithic cast con-
buildings in seismically active areas. crete. The simplicity of the connections,
Now, high rise buildings are being built which makes precast concrete econom-
with precast component systems in the ically viable, causes a lack of continuity
same active regions. The use of precast in stiffness and concentrated deforma-

PCI JOURNAt-1September-October 1989 43

 FElAs1IC
I 0 .i
^ I
Q 1 FYIELp
O
J ^

ALAS11C
AIaD OMAXMUM

DISPLACEMENT DISPLACEMENT

Fig. 1. Assuming that the earthquake transfers equal energy (area under curve),
elastic system (left) has small displacement, inelastic system (right) has larger
displacement.

tion demand in some locations may de- ductility) the joint must have available
velop during an earthquake. in order to perform without failure. If
Precast panelized wall systems have a the earthquake transfers a specific
lack of continuity in the horizontal con- amount of energy to the joint, that
nections between vertical wall ele- energy must be accommodated within
ments. The wall elements are relied the joint by either remaining elastic,
upon to provide both vertical load with a large force and small deforma-
carrying capacity and lateral load resis- tion, or by yielding, with a lower inter-
tance as shear walls. Yet, the shear walls nal force but larger deformation. This
are characterized by cantilever beam relation may be understood by viewing
type behavior with a lack of redundancy. Fig. 1.
Under lateral load the panel wall sys- The relation described above is the
tem's ability to carr y vertical loads may approach used in PCI Technical Report
be jeopardized by the wall's natural No. 5,' where a simplified approach to
tendency to yield and deform inelasti- seismic resistant design of precast
cally within a few weak horizontal structures is outlined. The PCI method
joints.
is not necessarily readily usable at this
Monolithic construction of joints may time, however, because it may be dif-
be appropriate in strong seismic zones; ficult to envisage a single-degree-of-
however, in zones of lower seismicity, freedom inelastic response mechanism.
structures with weak joints may still Thus, it could be hard to determine how
perfprm satisfactorily. The present diffi- much energy will be transferred into a
culty for design is in determining what structure, and more importantly, the
relation between strength and ductility actual yield capacity and available duc-
must be provided in such joints. Unfor- tility in most common precast connec-
tunately, there is not a single answer to tions is not known yet. The yield level
this design dilemma. For any given and available ductility must both be
yield strength which a designer might known to use the PCI approach or to as-
provide within a joint, the earthquake sign a proper force reduction factor (R)
energy level will dictate how much de- in design by current code equivalent
form nation capacity at the yield level (i.e., static load methods,

44
Objectives of Research Program formance of structures whose joints have
various yield strength-deformation
Shaking table tests would provide the
limits.
best means to study the seismic behav-
ior of precast systems, particularly the
relation between provided strength and Scope of Research Program
the demand for deformation induced by Three different 3-story assemblages of
the earthquake. Static tests of individual wall panels were tested under earth-
components or joints can provide in- quake motion on the shaking table.
formation on stiffness and strength, but Since the cost of shaking table testing
not the amount of deformation which limited the number of specimens which
will be needed. The actual forces trans- could be investigated, a few basic con-
ferred into the structure and deforma- figurations had to be selected which
tion developed during an earthquake, would represent the most common types
however, depend on the strength and found in panel wall structures. Each
ductility of the structure. The actual specimen was a single wall section built
forces and deformation can only be at one-third scale to represent a portion
determined by dynamic shaking tests of a 10 to 20-story precast shear wall.
or analyses. The first objective of this The measured test deformations in the
investigation was to quantitatively mea- walls were compared with computer
sure the response of a large scale panel aided predictions using various analytic
wall system during an earthquake. models. These correlation studies indi-
There have been no previous shaking cated which specific inelastic mech-
tests of large scale precast wall assem- anisms had to be accurately represented
blages. The shaking table tests con- in the analytic model. The analytic
ducted in this program were intended to model was then used to predict the be-
furnish a complete quantitative de- havior of a 13-story prototype wall sys-
scription of inelastic mechanisms and tem.
the effects of such mechanisms on the This report describes the basic re-
system's dynamic response. search and development work which
The second objective of the research has been completed on large panel pre-
program was to test analytic methods cast wall systems. The shaking table test
which have been proposed for use in program is outlined with an explanation
predicting the inelastic seismic re- of the model design, the testing system,
sponse of large panel precast wall as- the test procedures, damage observa-
semblages. Acceptable analysis tions and measured test results. The an-
methods could then be used to predict alytic correlation work is then summar-
the response of a complete prototype ized with a discussion of modeling
wall system which would be too expen- techniques. Finally, knowledge gained
sive and complex to test. Schrieker, Bec- from the tests and anal ytic work is used
ker, and kianotish 2 °3.4 developed com- to predict likely seismic limit states for a
puterized techniques based on static 13-story prototype wall system,
tests of wall connections and on as-
sumed forms of system deformation for
estimating inelastic response in precast LARGE PANEL PRECAST
wall systems. Shaking table experi- WALL SYSTEMS
mental results combined with correla-
tion studies would check the abilities of Large panel building systems are
the existing programs and aid in im- composed of vertical wall panels which
proving analytic techniques. Analytic support horizontal roof and floor panels
methods could then be used to develop to form a box like structure as dia-
improved designs by evaluating per- grammed in Fig. '? The vertical panels

PCI JOURNAL/September-October 1989 45

 are stacked and joined to create axial
loadhearing shear walls while the hori-
zontal panels act a diaphragms and
gravity load collecting roof and floor
s. stems, Martin, Patnian and Zcck57
have assembled reviews of various basic
large panel precast systems and the
typical joint configurations.
The primary difference between the
systems currently used is in the manner
in which the vertical and horizontal
panels are joined. Panel wall systems in
the United States typically use a "plat-
form" construction in which hollow- Fig. 2. Box like structure formed by a
core slabs form a horizontal layer simply large panel cross wall system.
supported on the wall panels below.
The tipper wall panels then rest on the
platform formed by the hollow-core menia was very similar to the system
floor slab as shown in Fig. 3. Panel sys- tested in this research project.
tems in Europe and elsewhere, how- Information from disaster inspections
ever, have connections which fre- has indicated that three of the hardest
quently use cast-in-place concrete to hit cities (Spitak, Leninakan and
form the joint between the wall and kirovakan) experienced strong shaking
floor panels. Nearly all systems suffer a over period ranges which would likely
similar weakness when used for resist- encompass the natural period of large
ing seismic loads, namely, economically panel buildings. Not a single large panel
efficient systems have only a limited building, however, was categorized as
amount of vertical reinforcement con- having collapsed or been damaged to a
tinuous across the joints and seismic degree requiring demolition though a
loading creates shear and flexural de- number of 4 to 9-story large panel
inands which can easily exceed the structures existed.* In the same area
capacity of that steel. numerous precast frame structures col-
lapsed and damage was even found in
Earthquake Performance and
the steel frame system of an industrial
building. Unfortunately, none of the
Existing Data
large panel buildings was instrumented
The existing use of large panel precast and little is known regarding their de-
systems was obviously accomplished sign strengths and likely level of seismic
with a considerable amount of testing induced shear force.
and analysis. Fortunately, very few large There were a number of Bulgarian
panel buildings have ever been sub- large panel buildings which were not
jected to strong seismic motions. The damaged by the 1977 Vrancea earth-
lack of records of performance, however, quake in Romania,' but the ground mo-
has created uncertainty regarding their tion was reported to have been pre-
ability to withstand the large forces or dominantly long period which would
energy levels which could result. The not excite short period shear wall struc-
best account of' large panel building tures. Shapiro 9 noted that significant
performance may be found in the after-
math of the 1988 Armenian earthquake
"Harris, J. R., "Precast Building Performance
near Yeravan in the Soviet Union. The During the Armenian Earthquake," presented
large panel building system used in Ar- during PCI Committee Days, April 13, 1989.

46
 F
'
Gavrilovic, and Suenaga. " •15• `6 Only two
4 p n'
reports described dynamic testing of
WALL PANEL
•a subassemblages; Harris' 7 has tested 1/16
DRY PACK n o'• scale models on a shaking table and
Polyakov 1 ' noted that vibro-platform
tests had been completed in the Soviet
Union but he did not provide any mea-
sured data. Though there is widespread
PRECAST SLAB use of precast large panel construction,
GROUT
UOUS there apparently have been virtually no
AL
full scale investigations to determine
the capacity demands which might be
made upon the systems during earth-
Fig. 3. Platform type horizontal quakes.
connection used in the United States

Methods for Predicting Response

damagedid develop in 2 and 4-story The precast walls act in a manner
large panel structures during the Gazli similar to monolithic shear walls in re-
1976 earthquake in the Soviet Union. sisting axial loads and shear forces in-
Some of the vertical panels had residual duced by wind or low amplitude earth-
lateral shifts of 4 to 6 in, (10-15 cm) and quakes. As long as the internal forces do
floor slabs had slipped off of walls. not cause nonlinear response, a precast
Damage to panels themselves was neg- wall could he analyzed like a monolithic
ligible. It was postulated that the panel shear wall. When precast large panel
shifting was due to the opening of hori- wall systems are used in low seismic re-
zontal joints during the initial motion. gions, it may be practical to design them
Then aftershocks caused gradual pro- for elastic linear behavior. Elastic be-
gressive displacements. Shapiro also havior may be achieved by ensuring that
noted that a portion of an actual 9-story the forces determined from an elastic
building had been tested with shakers at analysis, using a response spectrum
the top to a displacement of 1.8 in. (4.6 which has not been reduced, are less
cm), causing joint cracking and near than the panel and joint strengths.
doubling of the first natural period. Nonlinear response may be created
Other test work has determined the when either the shear force or over-
natural periods of in-situ full size turning moment surpasses a Iimiting
buildings through low amplitude shak- value. Horizontal slip starts in the wall
ing with vibration generators. Measured joint when the shear becomes too large.
periods of buildings with 4 to 12 stories The vertical reinforcing steel yields and
ranged between 0,17 and 0.52 seconds. rocking of the panels starts when the
Low amplitude tests, however, reflect moment becomes too high. Based on the
the natural periods in an "uncracked" limited existing test data, it appears that
state since axial loads generally keep slip and flexural distortion are the two
existing cracks closed. Polyakov 1O noted mechanisms of damage associated with
that large changes in period can develop nonlinear response in a simple wall unit
with large motion and damage- Many when called upon to resist strong seis-
tests ofjoints between panels have been mic motion. These two types of concen-
reported, such as those by Hanson, Vel- trated deformation are not common in
knv, Verbic" ' x ' a and others. Subas- monolithic shear walls because the
sernblage tests, with statically applied uniform continuous vertical and hori-
loads, have been completed by Borges, zontal steel tends to distribute the

PCI JOURNAL/September-October 1989 47

 flexural and shear distortions over a this research program to determine what
larger region. deformation demands and base forces
Several investigators have used the would be developed in such jointed
available test data and analytic tech- structures under earthquake motion and
niques to attempt to predict and investi- whether the structure could withstand a
gate the behavior of wall systems during strong motion without developing in-
strong ground shaking. Llorente, and stability. The specific design of connec-
Powell and Schricker 38.19 proposed tions in the test models was intended to
analytic means to investigate the sensi- be an improvement on previous forms of
tivity of panel wall response to various jointing for better seismic behavior. The
design parameters during strong earth- joints were not built in a "platform"
quakes. fashion as most connections in American
Llorente and Becker" examined the systems, but many facets of the response
effects of the postulated rocking motion of the tested joints could occur in plat-
in a series of analytic studies. They de- form or Other types as well. The joints
scribed rocking as an undesirable type were purposely designed to avoid a
of motion since acute shear concentra- shear slip mechanism, for the reasons
tions at the neutral axis and the severe noted above, while the platform type is
compressive strains at the closed end of expected to fail first in slip. 21 Only a
the rocking joint may induce failure and small portion of the description of the
lateral instability. Despite this possible tests and deformation mechanisms in
drawback, Llorente's analyses indicated the large panel systems can he pre-
that rocking may be helpful to the wall sented here; a detailed description of
system. It exhibits an isolation behavior the test program may be examined
which limits the force which can be elsewhcre.22
transferred into the wall, and softens the
wall system, moving the period to a
lower point in a response spectra. Test Models
Shear slip has been determined ex- The three large panel test specimens
perimentally to be a function of the total were one-third scale, 3-story high wall
vertical axial load being transferred segments axially loaded to represent a
throiigli the joint." Since a major portion portion of a wall near midheight of a 15-
of the shear may be resisted by pure story building. Each of the specimens
friction before slip starts, it is likely that was composed of three individual I-
slip might occur in joints near midheight story high wall panels. Every one of the
of the wall where the total axial force is subassemblies was 10i.6 in (2bl cm)
less than at the base but shear is still high overall. 'i rue scale modeling was
high. Llorente investigated the effects of employed at a size sufficient to allow
shear slip in analytic studies and con- use of normal concrete, though with
eluded that while it represents a source small aggregate. The overall mass of the
of energy dissipation and force isolation, model had to be artificially enlarged to
it should not be counted on as a reliable preserve the correct force ratios. The
resistance mechanism because ac- first model specimen was a simple as-
cumulated unrestrained slip could re- semblage of three wall panels. The sec-
sult in enough eccentricity to threaten ond model was similar to the first but
the stability and integrity ofa building. also included short perpendicular
"flange" walls at the ends of the main
SHAKING TABLE TESTS wall. The third wall included door
openings and strengthened lintel beams
A series of tests was conducted on but it will not be included in the tests
three large panel wall models as part of described here.

48
 OF

O
° WALL PANEL
0

GROUTED JOINT o °

o ° a a 'o; p 9
Q

v PRECAST SLAB

° p CONTMUOUS
VERTICAL
STEEL
Ij WALL PANEL i
I JOINT

KEY CONTINUOUS
VERTICAL
STEEL

PANEL

Fig. 4. The wall connection used in the test specimens.

The simple wall had a total vertical

reinforcing content across horizontal
joints of0.4 percent and the flanged wall
system had 0.7 percent of the wall cross
section area. In each case the vertical
reinforcement which continued across
the joint was concentrated at the ex- #
treme ends of the walls as detailed in
Fig. 4. The precast panels contained an
additional amount of well distributed Jj ^^ ►
horizontal and vertical reinforcement --==
throughout their interiors. Each test --
specimen had a steel platform and a set - _
r."
of mass blocks attached at its top to pro- . J R

vide the desired level of internal axial s - -^^

force and to induce lateral inertial forces `. IL' r`'
in the correct scale ratio. The walls were t
provided with an accessory lateral sup-
port frame which allowed vertical and
lateral motion parallel to the walls but
prevented out-of-plane motion. A test Fig. 5. Three-story model on the shaking
model and support frame are shown in table with a lateral support frame and
Fig. 5. mass blocks above.

PCI JOURNAL(September-October 1989 49

 Testing System subjected to a base motion of sufficient
The tests were performed in the intensity to cause appreciable damage.
Earthquake Simulator Facility at the The test sequence for the two speci-
University of California Earthquake mens is summarized in Table 1.
Engineering Research Center in Rich- The free vibration frequencies of each
mond, California. The 20 ft (6.1 m) test structure were measured before and
square shaking table was controlled to after each simulated earthquake to as-
reproduce the horizontal motion of a re- sess the degree of damage induced
corded earthquake. The wall models during the test. A low intensity "white
were attached to a special foundation noise" motion was applied through the
which was bolted directly to the shaking table and the acceleration response of
table. the structure was analyzed by Fourier
Instrumentation was selected to transform procedures.
monitor three types of dynamic re-
sponse; (1) shaking table motion, (2) ac-
celerations and displacements of the Damage Observations
models, and (3) local deformations and Observation of the specimens during
strains within the models. Horizontal and after the tests provides an indication
displacements and accelerations were of what damage and inelastic action oc-
measured at the base, at each floor level, curred, supplementing the instrumental
and at the top of each assemblage. Local data. There was no visible damage dur-
deformations measured within the test ing the low intensity shakes. A summary
specimens included the shear slip at of damage for the intense shakes fol-
vertical and horizontal joints, uplift at lows.
horizontal joints, and panel shear dis- Simple wall: The obvious visible re-
tortions. Strains were measured in se- sponse mechanism was rocking motion
lected reinforcing bars. The special associated with uplift at the lowest hori-
foundation included a set of force trans- zontal joint. No shear slip was noted. Up-
ducers to determine the magnitude of lift was accompanied by apparent com-
base shear transferred into the structure. pression damage at one end of the wall.
Two of the through joint reinforcing bars
at the damaged end of the wall buckled
Test Program
and the third had ruptured. Fig. 6 is a pho-
Each of the specimens was subjected tograph of the damaged area. Only minor
to a series of simulated earthquake mo- cracks existed at the opposite end.
tions. The applied motion was propor- Flanged wall: The visible response
tional to the N-S component of the was again dominated by rocking asso-
earthquake recorded at El Centro, ciated with alternate uplifting of the
California, in May 1940, the intensity wall ends and flange walls. The princi-
being modified by adjustment of the pal damage consisted of crushing or
table control system. The El Centro spalling at one end of the wall and in the
earthquake had considerable energy in adjacent flange; lesser but similar dam-
the period range near 0.5 seconds char- age occurred at the opposite end. Care-
acteristic of prototype structures. The ful examination of the damaged end re-
recorded earthquake record had to be vealed that all but one of the five verti-
time scaled to the ratio determined by cal continuous bars in the horizontal
the true scale modeling of the test connection below the flange and wall
specimens. In each case the test signal had ruptured; the remaining bar was
was first applied at a low intensity to buckled. The! opposite end had one
determine the system's elastic response flange tar ruptured and two other bars
behavior. Then each specimen was buckled.

50
Table 1. Test sequence Table 2. Measured natural frequencies.

Table Time of Frequency

Test acceleration Model measurement (Hz)
Model No. Earthquake (g's)
Simple Prior to test 5.20
Simple 1 EICentro 0.18 wall After Test 1 5.10
wall 2 El Centro 0,67 After Test. •.9{)
Flanged I El Centro 0.22 Flanged Prior to test 6.30
wall 2 El Centro 0.69 wall After Test 2 4.32

Test Results period of 0.16 seconds for a one-third

scale model, slightly lower than the ac-
The reduction in frequency (Table 2) tual measured period). Softening can
shows that the stiffness decreased by lengthen the period if joint slip or rock-
nearly 50 percent as a result of the dam- ing starts and could result in the struc-
age during the strong tests and verifies ture showing major amplification of the
concerns noted by Polyakov and ground motion.
Hawkins 1°' regarding the large period The instrumentation confinned the
changes which can occur in panel wall predominance of the rocking response
systems. Hawkins suggested that the described above during the strong
ground motion may not be significantly shakes. The structures remained elastic
amplified in panel structures if their during the low intensity tests. Peak
period remains at 0.3 seconds or less (a deformation quantities in Table 3 indi-

Fig. 6. Close-up view of damage at the end of the simple wall; the bar which is not
buckled had ruptured.

PCI JOURNAJSeptember-October 1989 51

 Table 3. Peak deformation quantities.
imple Flanged wall
Loading conditions ' Test 1 Test _>
'fable acceleration (g's) 0.18 0.67 0.22 0.69
Acceleration at top of wall (g's) 0.28 0,80 0.4.2 1.08
Iiisplacementattop of wall (in.) 0.13 1.48 0.09 0.80
Uplift at end of wall (in.) 0.02 0.70 Not (1.36
available
Base shear (kips) 7.4 15.7 9.1 21.9
Base moment (in.-kips) 1525 3516 1688 4056

cate that an amplification of the ground displacement history as shown in Fig. 8.

motion occurred in the structure. The Overturning moment controlled the
rocking response during the strong mo- rocking response, but the shear was
tion tests was indicated by the uplift linearly related to the moment. The
gages at the base of the wall across the moment-shear ratio was 204 in. or 2.7
lowest joint. The nature of the uplift times the length of the wall. First yield
may he deduced from Fig. 7 which in the joint steel was measured at a mo-
shows the measurement at three points ment of 1790 in.-kips (202 kN-m). The
on one side of the simple wall. Gage U6 calculated yield strength of the wall was
is located at the extreme end; it clearly estimated as 2070 in.-kips (234 kN-m) or
shows that the wall uplifts a maximum of with a shear of 10.1 kips (45 kN) using
0.34 in. (0.86 ern) as the top displaces normal beam theory. The difference in
sideways in one direction. It is also clear predicted and actual first yield was a re-
that the uplift never completely returns sult of the plane section assumption
to zero. The displacement at Gage U8 used in nonnal beam theory, The behav-
shows similar but less uplift in synch- ior of the wall was essentiall y linear until
ronization with Gage U6 indicating that the shear exceeded –12.8 kips (-57
the point of rotation is near the opposite kN). On reversal, the stiffness decreased
end of the wall. The smaller peaks, and further damage occurred at +15 kips
which are out of synchronization with (67 kN), The sudden drop in shear cor-
Gage U6, occur when the opposite end responded with rupture of one of the
of the wall uplifts to a maximum of0.70 joint bars. Oil subsequent cycles the po-
in. (1.8 cm). sitive shear never reached the same
The second type of primary motion maximum value.
measured at the lower joint was slip, i.e., Similar rocking and slip response was
sliding of the two vertical walls relative measured in the tests of the flanged wall
to each other. The average sliding mo- though its stiffness and strength were
tion, though only reaching an amplitude obviously higher than in the simple
of 0.04 in. (0.1 cm), tends to be toward the wall. A very small amount of defonna-
south — the end which experienced con- tion occurred in the vertical joint be-
crete damage and spalling. tween the main wall and the flanges but
A final indication of the performance did not dissipate significant energy. A
of the simple waIl may be obtained from full description of the test results and
an examination of its base shear vs. top data measurements is given in Ref. 24.

52


C)
C-
0
G
as
Z

cn
CD
a
CD
U6 U1 U13
3 o t41 Uf2 0.
a
m
0
C)
LOCATION KEY

0
a-
CD
8.e 6.0 12.0

CD
time - seconds
CD

UPLIFT AT U6

0. 0.r

0. 0.

0. r e.
J J
0. Q
a

0.0 6.0 12.0 18.0 24.0

0.0 6.0 12.8 18.0 24.0

time - seconds lime - seconds

UPLIFT AT U7 UPLIFT AT U8
th
W Fig. 7. Uplift histories of the panel at the lower connection.
 16,

4a^

a.

Q
Q
w
I e.

w
a
-s.

-16.e--
1.5 -e.8 @8 a8 1.6

TOP DISPLACEMENT (inches)

Fig. 8. History of the base shear plotted with the top displacements.

ANALYTIC CORRELATION action) even though the command mo-

PROCEDURES tion specified only a horizontal compo-
nent. The analytic model included a
Extensive computer aided studies rigid table with rotational mass sup-
have been completed at the University ported on springs to simulate the
of Wisconsin using data from the shak- dynamic table pitching. This entire
ing table tests to verify the ability of an model is diagrammed in Fig. 9.
existing analytic method. The experi- Certain assumptions were initially
mental vs. analytic correlation studies made during the modeling: (1) the pre-
identified the essential response mech- cast panel elements were assumed to
anisms which had to be duplicated for maintain a linear stiffness and remain
successful prediction of the seismic re- elastic; (2) all inelasticity was assumed
sponse of the precast panel wall system to be concentrated within the connec-
at various levels of excitation. tion regions, and (3) the connection re-
gions were assumed to be precracked
Analytic Modeling due to likely shrinkage between the
joint grout and the precast panels.
The primary portion of the analytic
model consisted of the three stories of
precast walls and joints. A lumped Techniques for Joint Modeling
translational and rotational mass was Horizontal joints between vertical
placed above the wall elements and at- precast panels were assumed to be the
tached to the wall with rigid links to Iocations where all inelasticity would
simulate the mass blocks above the test take place and the analytic joint ele-
specimens. The wall elements were con- ments had to represent the inelastic
nected to a spring foundation which mechanisms likely to develop. Insuffi-
modeled the flexibility of the actual cient information exists at present to
foundation member and force trans- model the entire range of possible joint
ducers. Finally, the shaking table itself behavior with a single analytic 1. con-
exhibited a detectable amount of pitch nection element." Two approaches have
during each test (due to structural inter- been used to model this connection re-

54
C) TOP MASS

4
0
_czZI j4::::: RIGID
LINKS
Z
RIGID
LINKS TOP CAST
CD
D WALL
CD

2
O UBSTRUCIURE 0
0 o------- 6 PANEL WALL
Cr
CD
PRECRACKED
CD
JOINT
CD (ELASTIC)

---------------- 1
SPRING C SPRING 5d SPRING ST
(CONCRETE) (SHEAR KEY) (STEEL)
.----- ------ -^------ - A -------
INELASTIC DETAIL 'A'
NONLINEAR JOINT Detail A: bottom horizontal joint
SPRVNG-SYSTEM flexural spring-system
CAST WALL
FOUNDATON
RIGID SHAKING
TABLE
• ACTIVE NODE
Precast concrete panel wall
° PASSIVE NODE --• TABLE PITCH ^a
SPRINGS _R finite element model.

n Fig. 9. Diagram of the 3-story analytic model.

 gion: (1) as a continuous media, with joint configurations, without large shear
miiltinode inelastic rectangular contact keys, shear slip may he more significant
or interface finite elements, or (2) with and require special simulation.
discrete inelastic nondimensional
spring elements. Becker, Mueller and
Llorente 3 discuss the advantages of ANALYTIC CORRELATION
using interface or contact finite ele- RESULTS
ments in obtaining gradual opening of
the joint and avoiding impact problems Elastic Behavior
and the required discretization if spring
The earthquake motion reached a
elements are used. peak acceleration of 0.17g during the
Regardless of the modeling approach, Iow amplitude test of the simple wall;
the simulation of the joint concrete may this level of motion was not sufficient to
need inclusion of initial compressive cause visible cracking or yielding. Ex-
stiffness, crushing with stiffness and cellent correlation was achieved be-
strength degradation, and opening of a tween the measured experimental and
gap when under tension. The concrete predicted top displacements throughout
is represented with finite elements or a the time history when the correct sec-
series of discrete springs. The vertical tion properties were used in the analytic
tension reinforcement is usuall y ex- model. To obtain "correct" section
plicitly modeled with truss elements. properties, the lower panel's effective
Shear slip resistance may be included in cross section had to be able to change
the continuous concrete finite elements from gross to cracked when moments
or by special nondimensional slip exceeded the cracking level. It was then
springs. necessary for the model to regain its
The use of discrete springs in the form gross section stiffness as a condition for
suggested by Powell and Schricker 19 was matching the low amplitude response
applied in this study with a modified cycles after cracking had occurred. The
version of the DRAIN2D-MkII $ com- axial gravity load causes the wall to re-
puter program. Use of discrete nondi- develop its gross section when the
mensional springs makes it easier to de- overturning moment is low.
velop new elements to model specific
characteristics.
Our test results clearly indicated that Inelastic Behavior
successful modeling would require Simulation of the simple wall's in-
duplication of the rocking mechanism elastic response during strong ground
with attendant gap opening and closing. motion required an exceptionally com-
Moreover, it appeared as if the model plex modeling approach. The system's
should be able to reproduce rupture of response involved cracking of panels,
the steel reinforcement across the hori- yielding of steel in the joint, gap open-
zontal joint since rupture occurred very ing, buckling of steel rods, rupturing of
quickly once the bars started yielding. steel, and spalling and destruction of
As a consequence of these two require- unconfined concrete near the buckled
ments, particular attention had to be vertical bars. The effect of these various
given to the form of the timewise mechanisms in modifying the predicted
analysis algorithm since each of the pre- response of the wall system is discussed
ceding events would abruptly alter the in this section.
stiffness of the system. If a constant time The simplest inelastic model used bi-
step method were used, small time in- linear yielding truss elements to repre-
crements would be necessary, resulting sent the joint's steel reinforcement, The
in significant computation time. In other predicted response to the 0.67g earth-

56


1.5
SOLID = EXPERIMENTAL
H
DASH = ANALYTICAL
L
U 1.0
C

I- 5 1
z CI
LU 1
f 5
'II
w 0.0
U4
Cl

U) Ii

4
0
-1.0 RUPTURE 5„'

-1.5
0.0 1.0 2.0 3.0

TIME (seconds)
Fig. 10. Response history of the analytic mode( without rupture
capability (rupture occurs at 1.42 seconds in the test specimen).

1.5
SOLID = EXPERIMENTAL
DASH = ANALYTICAL
1.0
U

I-
2
LU
2 0.0
1.[/UI
W
U

-1.0
0
I-
-1.5

0 2 4 6 a

TIME (seconds)

Fig. 11. Response history of the analytic model which

included rupture (rupture occurs at 1.42 seconds).

quake using a time step of 0.001 seconds The inclusion of steel rupture was es-
is compared to measured response in sential for accurate modeling of the true
Fig. 10. The simulation gave good cor- behavior since the predominant rocking
relation until the joint reinforcement mechanism became active after rupture.
ruptured in the test specimen at 1.42 Fig. 11 shows the change in simulated
seconds. response which occurred when the truss

PCI JOURNAL(September- October 1989 57

 24 +4

12

IIF
^ ^• ^^

ROCKING

-24

-1.28 -0.32 0.32 1.28

TOP DISPLACEMENT (in.)
Fig. 12. History of the base shear and top displacement as
predicted by the analytic model.

element had a Iimiting value at which cross section.

rupture occurred. The predicted base 2. Yielding across the thickness of the
shear vs. top displacement plot, which is joint or over the unbonded length for
shown in Fig. 12, produces the same reinforcement which is made continu-
nonlinear elastic rocking response after ous across the joint.
rupture as was seen in the experimental 3. Rupture of steel crossing a joint.
data of Fig. 8. 4. Buckling of unruptured steel after
Many additional effects had to he con- it has undergone tension stretching and
sidered to achieve a satisfactory simula- is then reconipressed.
tion of the precast system's inelastic re- 5. Concrete cracking and gap open-
sponse. For instance, the wall panels do ing.
not bend in a manner consistent with fi. Degradation of concrete including
the simplified theory of "plane sections strength deterioration and loss of stiff-
remain plane." The gap at the tension ness.
end, during rocking, becomes larger 7. Decrease in concrete compression
than expected and the compression zone stiffness after a gap opening has oc-
at the opposite end becomes smaller curred,
than expected. Reliable modeling of the 8. Simulation of shear slip mech-
joint resistance requires use of numer- anisms if necessary.
ous closely spaced discrete concrete 9. Inclusion of closely spaced nodes
compression gap elements at the ends of at the interface between panels and joint
the wall due to the tremendous strain elements to avoid force concentration at
variation in a short distance. MalhasG5 particular points in the panel and to
has listed items which had to be specifi- simulate the rapidly varying joint forces
cally simulated or considered in devel- caused by the nonlinear strain variation.
oping the analytic model with the re- 10. Use of small time steps or variable
sponse shown in Fig. 10 including: time stepping to avoid integration errors
I. Nonlinear strain variation over a when stiffnesses abruptly change.

58
 RESPONSE PREDICTION high rise building. The individual pre-
FOR 13-STORY PROTOTYPE cast panels had very limited vertical
reinforcement (0.4 to 0.7 percent) made
]laving achieved successful correla- continuous across joints at each floor
tion between the measured response level. The reinforcement was concen-
and predictions for the test model, the trated at the ends of the wall with a pro-
analytic techniques were employed in totype spacing of 206 in. (5.23 in). The
estimating strong motion response and models experienced shaking from
defining certain limit states for the full ground motion proportional to the El
13-story prototype wall system upon Centro earthquake with their response
which the model sections in the experi- measured and compared to analytically
mental work had been based. Panel and predicted behavior.
joint stiffnesses were assigned on the
basis of experience with the one-third
Base Shear During
scale model, Because of the height of
the prototype system, the initial elastic Earthquake Testing
first mode natural period was nearly 0.€3 The 3-story simple precast wall suhas-
seconds. semblage examined here was designed
The limiting earthquake level for the to elastically resist a base shear equal to
wall system to avoid yielding was found 45 percent of the system's weight. If de-
to be at a maximum acceleration of 0.36g signed by the Uniform Building Code
if the ground motion was proportional to (UBC),28 the wall would be required to
the El Centro motion used in the wall resist a base shear of approximately 20
test program. The peak acceleration percent of the system's weight for con-
would vary considerably for other types struction in a high seismic area (Zone 4).
of motion since the yielding level was Thus, the test subassemblage was de-
found to be very sensitive to the match signed to resist a force of just over twice
between structural natural frequency the UBC's required minimum, A
and the earthquake spectra. A second moderate seismic motion with an accel-
limiting level of ground motion, again eration amplitude of 0.2g was success-
for a motion proportional to the El Cen- fully resisted by the system without per-
tro record, would be the amplitude ceptible damage or measured yielding
which would cause rupture of the verti- while peak base shears as high as 32
cal reinforcement. The predicted pro- percent of the system's weight were de-
totype rupture would occur when the veloped. When the ground motion was
ground motion reached an acceleration increased by a factor of 3 to 4, to a peak
amplitude of 0.9g though concrete acceleration of 0.7g, significant observ-
crushing initiates in the joint at an able damage developed with yielding,
amplihide of 0.5g and may cause deter- uplifting, and rocking at the lower hori-
ioration of the system's stability before zontal joint. At the higher level of mo-
rupture could occur. tion a base shear equal to 70 percent of
the weight was developed, indicating
that the force demand created by the
SUMMARY AND DISCUSSION earthquake was approximately equal to
the product of ground acceleration and
Project Description system mass. The base shear reached
Three precast panel wall subassem- 150 percent of the design base shear.
blages were tested under simulated The base shear which develops dur-
earthquake motion. Each of the speci- ing seismic shaking of a structure de-
mens was a 3-story one-third scale pends on the dynamic nature of the
model of walls from near midheight of a stricture, specifically its natural periods

PCI JOURNAUSeptember-October 1989 59

 and damping. Thus, the characteristics a age. The amount of top displacement
designer gives to a structure can influ- which accompanied the extensive
ence the base shear or strength demand yielding was dependent on the design
and displacements which an earthquake yield force level and reached more than
will create. In the test structure the de- five times the yield displacement before
sign shear capacity at yield was less than the strength capacity became reduced
the earthquake induced force and (displacement ductility = 5). In fact, the
yielding ensued, The yielding resulted maximum top displacements, with
in a 50 percent reduction of the system's rocking after joint reinforcement had
stiffness causing a significant change in broken, became as high as ten times the
the natural period. The yielding and yield displacement. Seventy-five per-
change of period has a direct influence cent of the top displacement was due to
on what level of force demand the the rocking of the bottom joint.
earthquake will tend to create within
the structure.
Connection Capacity
Demand — Rocking Motion
Structural Capacity Demand and
Design Strength A rocking mechanism effectively iso-
lated the test walls from the ground mo-
Minimum design forces such as tion and limited the amplitude of base
specified in UBC are lower than the ac- shear which could be transferred into
tual forces which would be induced in a the structures. During rocking the sys-
structure by a design earthquake if the tem's weight, acting along a path near
structure were to remain elastic. The the center of the wall, provided the only
codes allow design for reduced forces restoring moment to counteract the
under the assumption that during a de- overturning moment caused by inertial
sign earthquake the structure will yield effects. This constant resistance capacity
when the reduced design force level is is clearly evident in the negative dis-
reached and plastic deformation will placement cycles of Fig. 8. This mech-
take place until the reversing nature of anism isolated the wall above the joint
the earthquake causes the forces to de- from receiving any greater moment or
crease. It is implicitly assumed that the shear. The load limiting effect, initiated
structure will be able to withstand this by opening of the lower joint, however,
yielding and plastic deformation with- prevented the spread of inelasticity to
out failing. Unfortunately, our under- any other locations in the wall system.
standing of the dynamic mechanisms The reinforcement which was continu-
and available plastic deformation iii ous across the lower horizontal joint ex-
precast concrete structures is very in- perienced tremendous elongation and
complete, making it difficult to take ad- rupture since nearly all the deformation
vantage of reduced force design. demand created by the ground motion
The connection below the lowest pre- was concentrated within the single joint.
cast panel was forced to sustain large Strains in the limited cross joint rein-
deformations in the wall system tested. forcing bars surpassed 4 percent, the
The connection yielded because the level at which strain gages became de-
base shear which the earthquake motion fective. The average bar strain calcu-
would have created, if the structure had lated from the joint uplift if the bar had
remained elastic, was more than three not ruptured would have been as high as
times the yield strength of the connec- 32 percent, far beyond the bar's strain
tion. Since the connection was not able capacity.
to remain elastic, it was forced to A maximum uplift of 0.7 in. (1.78 cm)
undergo plastic deformation with dam- occurred in the model test wall. The

60
equivalent uplift in the 13-story proto- flexural or rocking type failures and
type building would have been 2.1 in. shear failures showed that both types of
(5.3 cm). Under these conditions the failure could lead to instability,
closed end of the joint is under tre- The shear failure mechanism (shear
mendous compressive stress since the slip) is capable of dissipating energy
axial load and flexural compression is very efficiently so that the energy
resisted within a very small concen- transferred into the structure by the
trated compression zone. This zone ground motion does not create large
must be able to resist high compression forces or large displacements. Llorente
forces without brittle crushing. The found nevertheless that shear slip may
compression zones in the test specimens be undesirable because resistance de-
exhibited limited crushing of the joint pends largely on friction and when
concrete. The crushing naturally started sliding starts it is liable to lead to ac-
at the outer fiber and proceeded inward cumulated unrestrained displacements
as material was lost. Crushing only oc- under certain earthquakes when the slip
curred over a limited distance near the occurs predominantly in one direction.
ends of the walls in the test specimens. There is certain danger in having unre-
The amount of crushing which occurred strained displacement because large
appeared to be limited by the short secondary (P-Delta) moments develop
length of time within which the wall and eccentricity will occur in perpen-
was at a high uplift. Reversal of the up- dicular walls.
lift, caused by the reversing ground mo- A rocking mechanism dissipates little
tion, reduced the compression force and energy as could be seen in Fig. 8, and
limited the degree of concrete crushing, creates severe force concentration in the
Connections play the most important compression region. As the wall rocks
role in controlling the behavior of a pre- open, all the axial load and the compres-
cast large panel wall system during sion force of the flexural couple has to
seismic loading regardless of the design be resisted in a small compression stress
approach used, Though current prac- zone at one end of the wall. Compres-
tice in aseismic design is aimed at de- sion crushing of the concrete may occur,
veloping strong connections and forcing leading to instability. Rocking's main
inelastic behavior away from connec- advantage is that it should not result in
tions, just the opposite behavior occurs accumulated displacements.
in precast systems. Weak connections The importance of connection design
can, however, operate successfully if in precast large panel wall structures
their design explicitly provides for the with the two mechanism alternatives,
inelastic demands of earthquake motion. decisions regarding design strength and
The level of yield strength has to be bal- associated deformation demand, and
anced with sufficient deformability and Iimited knowledge of available ductility
failure strength to allow repeated cycles leave a designer in a quandary when
of shaking without collapse. attempting to provide an efficient and
The connections between panels of safe system. Both of the mechanisms
the wall system described in this paper noted above create softening of the
were designed to yield and fail in flex- structural system with increasing dis-
ure before shear, Large shear keys pre- placement and may act to isolate the re-
vented a premature shear failure. Panel mainder of the wall from increased force
wall systems using the platform con- transfer, however, they also prevent the
nection of the United States are likely to spread of inelasticity in the wall. Overall
fail first in shear. Llorente and it appears that the flexural or rocking
Becker's`" ° investigations of the bene- type of mechanism is preferable in its
fits and disadvantages inherent in resistance to developing accumulated

PCI JOURNALISeptember-October 1989 61

displacements. If the flexural housing, tends to make it easier to
mechanism is selected, then it remains achieve high elastic strength capacities.
for the designer to insure that the joint In apartment buildings many of the
has a sufficient balance of strength and interior walls can be load resisting panel
ductility or deformability to survive the walls. With numerous walls available,
seismic demands, The best balance of the lateral load capacity may become
those quantities has not been deter- quite high. The horizontal joints be-
mined but the model test walls have tween panels need reinforcement
sufficient capacities to withstand the which is continuous hetwen panels, but
effects of a major seismic motion. a small amount of reinforcement which
is connected at the end of each wall can
Toughness of Large Panel System provide a considerable moment resist-
ingeapacity due to its Iarge moment arm
The ability to survive the effects of an combined with the axial compressive
earthquake, through strength, energy stresses in the wall from gravity loading.
dissipation, ductility and defbrmability,
has often been referred to as toughness.
Though the walls examined in this study Evaluation of Analytic Techniques
exhibited little energy dissipation and A set of criteria which needs to be
only moderate ductility in their rocking included in an inelastic analysis to cor-
mechanism, they did endure a strong rectly simulate response has been de-
earthquake test. The combination of term rued for wall rocking mechanisms.
force isolation, varying stiffness and Particular attention must be given to the
period, and ability of the system to special conditions which exist during
undergo large displacements associated the rocking type of response: nonlinear
with rocking while maintaining its sta- cross section, opening and closing of the
bility allowed it to maintain vertical load joint gap, rapid changes in stiffness, and
carrying without collapse. The tough- degrading material characteristics, New
ness exhibited in the walls appears to be elements had to be added to an existing
primarily a result of the deformability panel wall analysis program to simulate
and the force isolation effect. all the required joint stiffness char-
The large panel building system can acteristics. The integration time step in
also he designed to withstand a major a step-by-step analysis had to be care-
earthquake by providing toughness in fully chosen to avoid errors caused by
the form of a large elastic energy ab- abruptly changing stiffness. Beam mod-
sorption capacity. The precast panel eling of the cantilever wall system was
wall buildings which survived the De- not successful once the joint crack
cember 1988 Armenian earthquake opening reached a stage where non-
were very similar in construction to the linear strains exist across the section.
walls described here and appeared to Even in the elastic range (before yield-
have been provided with such a capac- ing) the wall's response is particularly
ity. The very nature of the application of sensitive to changes between gross and
many large panel systems, to provide cracked section stiff iesses.

62
 CONCLUSIONS AND RECOMMENDATIONS

Based on the three specimens tested bility in a three-dimensional structure

in this program and the corresponding and must be provided when inelastic
analytic simulations, certain conclusions behavior is anticipated.
can be drawn. 5. Analytic techniques exist which
1. Shear walls of large panel precast are capable of predicting the behavior of
construction can be designed to resist large panel wall systems but are u.un-
loads induced by moderate earthquakes practical or unreliable for normal use
while remaining elastic and being easily because modeling is complex and the
constructed. Even though the systems response of the system is very sensitive
have weak connection regions, with to the changing stiffness of the connec-
only a small amount of continuous rein- tion region.
forcement through the joints, significant The design of precast structures to
capacity can be achieved. When the de- have acceptable seismic resistance is a
sign is based on full elastic loads, it is perplexing problem at this time because
not necessary to provide connections of their weak jointed nature. Recent
which resemble monolithic concrete. changes in design codes will require
2. When shear wall systems, and par- buildings to be designed for seismic
ticularly the joints, are designed to resist load in portions of the United States
forces which are lower than the likely where wind Ioading once controlled.
elastic force which would be induced by Additionally, the demand for economical
the design earthquake, such as design buildings, particularly for housing, has
forces which are often given in codes, created a need for means of seismic re-
inelastic action will likely result during sistant design in regions of strong mo-
a design earthquake. If the construction tion. Certain steps may he taken to treat
is similar to the test specimens, then the the current design problems in regions
joints may undergo extreme deforma- of low seismicity and to develop designs
tions, such as the rocking noted in the in the future for regions of high seismic-
tests, and joint reinforcement ma y nip- ity.
ture, but there is a good likelihood that A. Large panel buildings should he
the system will he able to survive the designed to elastically resist the seismic
ground motion without collapse. A forces in regions of low seismicity. Cur-
rocking mechanism will result and rent technology allows such design
would need special provisions to main- when seismic forces are not high. When
tain stability. codified equivalent static loading
3. The particular large panel system methods are used, the design base shear
tested in this program appears to have should not he a "reduced" force. The
the potential to he able to survive ATC 21 approach for defining the base
strong seismic motions without collapse shear could be employed without the
though experiencing serious deforma- load reduction factor "R". Since it is
tion and damage. often possible to use many walls in the
4. Energy dissipation, which is very building as Iateral load resisting struc-
low with a rocking mechanism, may not tural walls, the actual force developed in
need to be a prime objective in the de- each will often be relatively small as in
sign of the large panel precast seismic the panel wall buildings surviving the
resistant systems if sufficient deform- Armenian earthquake.
ability is provided while maintaining B. The rocking mechanism, which oc-
stability. Minimum ties as suggested by curs when the elastic capacity is passed,
PCA27 are essential to maintaining sta- appears to be the more desirable of the

PCI JOURNAL^September-October 1989 63

two likely inelastic mechanisms possi- would solve three of the basic problems
ble in a horizontal joint between wall in inelastic panel wall response. First,
panels (i.e., shear slip or rocking). The since the vertical connection would not
rocking mechanism is unlikely to de- be an essential link in the gravity load
velop unrestrained motion or accumu- bearing system, its loss would not en-
lated deformation which could Iead to danger the stability of the building sys-
instability. tem. Secondly, the joint could be a
It appears that it would be desirable to source of energy dissipation since it
modify the platform system of construc- would serve primarily as a shear transfer
tion used in the United States in a man- mechanism. This would complement
ner which would limit slip and create the lack of energy dissipation and duc-
flexural motion. This might be achieved tility of the existing rocking mechanism.
by providing grouted keys or shear re- Third, having a vertical coupling joint
sisting links between the stacked wall between walls would create redundancy
elements. If the platform system can be in the wall system. It is well recognized
modified to force the inelastic mech- that redundancy is a very desirable fea-
anism to become a flexural one, then ture in any structural system which may
steps must also be taken to provide an he loaded beyond its elastic limit.
increase in the compressive strength in E. Large scale tests and analytical
the joint region. Once rocking starts, simulations, except for the Armenian
high compressive forces become con- earthquake results, have involved only
centrated near the ends of the walls. The two-dimensional assemblages. It is un-
platform system would probably not clear how the complete three-dimen-
have sufficient compression capacity to sional building system will act, Either
resist the necessary loads in a ductile the slip or rocking mechanism, occur-
manner and development of a modified ring in walls in one direction, will affect
joint would be necessary. the strength and stability of other walls
C. Particular attention must be running in a perpendicular direction,
given to tying the entire building sys- since those walls will presumably be
tem together as a means of preventing bent about their weak axis. Analytical
accumulated deformations from de- approaches will have to be developed to
veloping and for maintaining stability. As model the three-dimensional behavior
a minimum, the PCA recommendations and tests should be used to verify pre-
for ties around the periphery and dictions since stability effects are dif-
through the diaphragms of the system ficult to simulate in an inelastic system.
must be provided. F. Large deformation in the hori-
D. Attention should be aimed at the zontal joints between vertical panels,
use of vertical connections between either slip or particularly rocking, will
stacks of panels as the first location of exert serious deformation demands
inelasticity rather than rocking or slip in upon the floor or roof diaphragms.
the horizontal joints between panels. Rocking in a three-dimensional struc-
Vertical joints could be used to form a ture may literally tear the floor dia-
coupled shear wall system analogous to phragms apart. Special reinforcement
the system recently developed for and ties may be necessary to ensure the
monolithic construction_ A limited integrity of such diaphragms but insuffi-
amount of research has been dedicated cient information is available at present
toward this end 2A but apparently has not to provide any design guidelines. The
been successful yet. bending deformability of precast floor
A mechanism within the vertical joint diaphragms should be substantiated.

64
 ACKNOWLEDGMENTS
The test program described here was the wall models which were supplied by
funded by the National Science Foun- RAD Construction Company of Bel-
dation. Drs. M. Velkov and P. Gav- grade, Yugoslavia. Bahrain Shahrooz, Ali
rilovic of the Institute of Earthquake Belhadj, and Cherif Baleh, graduate
Engineering and Engineering Seismol- students at the University of Wisconsin,
ogy (IZIIS) of Skopje, Yugoslavia, pro- provided help in data reduction and an-
vided aid in obtaining and assembling alytic studies.

REFERENCES
1. Clough, D., "Design of Connections for 8. Brankov, G., "Effects of Vrancea — 1977
Precast Prestressed Concrete Buildings Earthquake on the Prefabricated Struc-
for the Effects of Earthquake," Technical tures in Bulgaria," Bulgarian-American
Report No. 5, Prestressed Concrete In- Seminar on Seismic Safety of Prefabri-
stitute, Chicago, Illinois, 1985. cated Concrete Buildings, Bulgarian
2. Schricker, V., and Powell, G., "Inelastic Academy of Sciences, Sofia, Bulgaria,
Seismic Analysis of Large Panel Build- 1984, pp. 8-19.
ings," Report 80-38, Earthquake En- 9. Shapiro, G., and Ashkinadze, G., "Ul-
gineering Research Center, University of timate Stresses in Large Panel Buildings
California, Berkeley, California, Septem- Exposed to Seismic Load," Proceedings,
ber 1980. 7th World Conference on Earthquake En-
3. Becker, J., Mueller, P., and Llorente, C., gineering, V. 5, Istanbul, Turkey, 1980,
"Nonlinear Dynamic Analysis of Precast pp. 351-358.
Concrete Walls," Nonlinear Design of 10. Polyakov, S. W., et. al., "Investigations
Structures, Study No. 14, University of into Earthquake Resistance of Large Panel
Waterloo, Ontario, Canada, 1980, pp. Buildings," Proceedings, 4th World Con-
507- 543. ference on Earthquake Engineering, V. 1,
4. Kianoush, M. R., and Scanlon, A., "Inelas- Santiago, Chile, 1969, pp. 165- 180.
tic Seismic Response of Precast Concrete 11. Hanson, N. W., "Design and Construction
Large Panel Coupled Shear Wall Sys- of Large Panel Structures, Supplemental
tems," Structural Engineering Report 134, Report C; Seismic Tests of Horizontal
Department of Civil Engineering, Univer- Joints," Portland Cement Association,
sity of Alberta, Edmonton, Canada, March Skokie, Illinois, January 1979.
1986. 12. Velknv, M., "Large Panel Systems in
5. Martin, L. DJ., and Korkosz, W. J., "Con- Yugoslavia: Design, Construction and Re-
nections for Precast Prestressed Concrete search for Improvements of Practice and
Buildings Including Earthquake Resis- Elaboration of Codes," Proceedings,
tance," PCI Technical Report 2, Pre- Workshop on Design of Prefabricated
stressed Concrete Institute, Chicago, Concrete Buildings for Earthquake Loads,
Illinois, March 1982. ATC-8, Applied Technology Council,
6. Patman, P., et al., "Industrialized Building Berkeley, California, December 1981,
— A Comparative Analysis of the Eur- pp. 81-120.
opean Experience," Department of Hous- M. Verhic, B., "Nonlinear Behavior of Large
ing and Urban Development, Washington, Panel Connections," Research Confer-
D.C., April 1968. ence on Earthquake Engineering, Skopje.
7. Zeck, U. I., "Joints in Large Panel Precast Yugoslavia, June 1980, pp, 219-234.
Concrete Structures," Report R76-16, 14. Borges, F., and Ravarra, A., "Structural
Department of Civil Engineering, Mas- Behavior of Panel Structures Under Earth-
sachusetts Institute of Technology, Cam- quake Actions," Report Laboratorio Nac-
bridge, Massachusetts. January 1976. ional de Civil Engenaria, Lisbon.

PCI JOIJRNALSeptember-October 1989 65

 Portugal, 1966. ence on Earthquake Engineering, Skopje,
15. Gavrilovic, P., and Velkov, M., "Ex- Yugoslavia, 1980.
perimental Testing of Three Story Models 22. Oliva, M. C., and Shahrooz, B., "Shaking
and Connection Systems of Modified Table Tests of Large Panel Wall Struc-
RAD-Balency System," Institute of Earth- tures," Proceedings, 8th World Confer-
quake Engineering and Engineering Seis- ence on Earthquake Engineering, V. 6,
mology, University Kiril and Metodij, San Francisco, California, July 1984, pp.
Skopje, Yugoslavia, 1981. 717-724.
16. Suenaga, Y., "On Box-Frame-Type Pre- 23. Hawkins, N., "Seismic Resistance of Pre-
cast Reinforced Concrete Construction of stressed and Precast Concrete Structures
Five Storied Multiple Houses," Concrete (Part 2)," PCI JOURNAL, V. 23, No. 1,
Journal, V. 12, No. 7, 1974, pp. 1-26. January- February 1978, pp. 40-59.
17. Caccese, V., and Harris, H., "Seismic Be- 24. Oliva, M. G., and Clough, R., "Shaking
havior of Precast Concrete Large Panel Table Tests of Large Panel Precast Con-
Buildings Using a Small Shaking Table, crete Building System Subassemblages,"
Report 2 --- Sinai! Scale Tests of Simple Report 8 3-14, Earthquake Engineering
Precast Shear Wall Models Under Earth- Research Center, University of California,
quake Loading," Report D84-01, Depart- Berkeley, California, June 1985.
ment of Civil Engineering, Drexel 25. Malhas, F., "Seismic Response of Large
University, Philadelphia, Pennsylvania, Precast Panel Walls: Analytical Correla-
December 1984. tions with the Shaking Table Tests Re-
18. Llorente, C., "Inelastic Behavior of Pre- sults," PhD Thesis, Department of Civil
cast Concrete Shear Walls," PhD Thesis, and Environmental Engineering, Univer-
Department of Civil Engineering, Mas- sity of Wisconsin, June 1988.
sachusetts Institute of Technology, Cain- 26. Uniform Building Code, International
bridge, Massachusetts. 1981. Conference of Building Officials, Pasa-
19. Powell, G., and Schricker, V., "Ductility dena, California, 1985.
Demands of Joints in Large Panel Struc- 27. Schultz, D., Burnett, E., and Fintel, M.,
tures," ASCE Fall Convention Preprint "Design and Construction of Large Panel
3022, San Francisco, California, Ameri- Structures. A Design Approach to General
can Society of Civil Engineers, October Structural Integrity," Report 4, Portland
1977. Cement Association, Skokie, Illinois,
20. Llorente, C., Becker, J. M., and Roesset, J. 1977.
M., "The Effect of Non-Linear-Inelastic 28. Applied Technology Council, "Tentative
Connection Behavior on Precast Panel- Provisions for the Development of Seis-
ized Shear Walls." Session on Mathemati- mic Regulations for Buildings," Applied
cal Modeling of Reinforced Concrete Technology Council, Publication ATC-3-
Structures — ACI Committee 442, 06, June 1978.
American Concrete Institute Convention, 29. Wiss, Janney, Elstner & Associates, "Cy-
Toronto, Canada, April 1978. clic and Monotonic Shear Tests on Con-
21. Becker, J. M., and Mueller, P., "The Role nections Between Precast Concrete
of Connection in Aseismic Design of Panels for MIT." WJE Report No. 77578,
Large Panel Buildings," Research Confer- Northbrook, Illinois, July 30, 1981.

NOTE: Discussion of this paper is invited. Please submit

your comments to PCI Headquarters by June 1, 1990.

66