HIGHWAY

RESEARCH RECORD
Number Soil-Structure Interaction:
413 A Symposium

9 reports
prepared for the
51st Annual Meeting

Subject Areas

23 Highway Drainage
63 Mechanics (Earth Mass)

HIGHWAY RESEARCH BOARD

DIVISION OF ENGINEERING NATIONAL RESEARCH COUNCIL
NATIONAL ACADEMY OF SCIENCES-NATIONAL ACADEMY OF ENGINEERING
Wethington, D,C, 1972
NOTICE
The studies reported herein were not undertaken under the aegis of the
National Academy of Sciences or the National Research Council. The
papers report research work of the authors done at the institution named
by the authors. The papers were offered to the Highway Research Board
of the National Research Council for publication and are published herein
in the interest of the dissemination of information from research, one of
the major functions of the HRB.
Before publication, each paper was reviewed by members of the HRB
committee named as its sponsor and was accepted as objective, useful,
and suitable for publication by NRC. The members of the committee were
selected for their individual scholarly competence and judgment, with due
consideration for the balance and breadth of disciplines. Responsibility
for the publication of these reports rests with the sponsoring committee;
however, the opinions and conclusions expressed in the reports are those
of the individual authors and not necessarily those of the sponsoring com-
mittee, the HRB, or the NRC.
Although these reports are not submitted for approval to the Academy
membership or to the Council of the Academy, each report is reviewed and
processed according to procedures established and monitored by the
Academy's Report Review Committee.

ISBN 0-309-02085-9
Library of Congress Catalog Card No. 72-9785
Price: $3.60
Available from
Highway Research Board
National Academy of Sciences
2101 Constitution Avenue, N. W.
Washington, D. C. 20418
CONTENTS
FOREWORD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

SUBSURFACE SOIL-STRUCTURE INTERACTION: A SYNOPSIS

Ernest T. Selig . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

lilSTORICAL DEVELOPMENT OF THE SOIL-STRUCTURE

INTERACTION PROBLEM
Don A. Linger . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

MATERIAL PROPERTIES AFFECTING SOIL-STRUCTURE

INTERACTION OF UNDERGROUND CONDUITS
Raymond J. Krizek and J. Neil Kay . . . . . . . . . . . . . . . . . . . . . . . . . . 13

EXPERIMENTAL STUDIES IN SOIL-STRUCTURE INTERACTION

F. Dwayne Nielson . . . . . . . . .. .. .· . . . . . . . . . . . . . . . . . . . . . . . . 30

BALANCED DESIGN AND FINITE-ELEMENT ANALYSIS OF CULVERTS

J. R. Allgood and S. K. Takahashi . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

STRUCTURAL DESIGN PRACTICE OF PIPE CULVERTS

Bernard E. Butler . . . . . . . • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

DESIGN OF CIRCULAR SOIL-CULVERT SYSTEMS

F. Dwayne Nielson and Natverial Dalal Statish . . . . . . . . . . . . . . . . . . . 67

BUCKLING OF CYLINDERS IN A CONFINING MEDIUM

C. V. Chelapati and J. R. Allgood . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

THE IOWA DEFLECTION FORMULA: AN APPRAISAL

Richard A. Parmelee and Ross B. Corotis . . . . . . . . . . . . . . . . . . . . . . 89
Discussion
M. G. Spangler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

SPONSORSlilP OF TlilS RECORD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

FOREWORD
Since the introduction of corrugated metal pipe in the late 1890s, there has been a con-
tinuing interest in proper pipe usage for maintenance purposes. Early methods of de-
sign were truly "rule of thumb" and were developed through "failure criteria." Refine-
ments of these methods resulted in a set of systems that were very effective in dealing
with pipe problems. However, the building of the Interstate Highway System brought
with it a need to develop new approaches that would incorporate the latest soil-structure
interaction theories and be applicable to the larger structures, higher embankments,
and new materials. Better inspection and construction control are considered requisite
to the utilization of more refined theories.
Practice in the design and construction of culverts continues to improve, but, because
many organizations have not been able to keep up with or apply the latest theories, the
Highway Research Board Committee on Subsurface Soil-Structure Interaction organized
a symposium to assess the state of the art and to delineate problems needing further
attention. This is considered the first step in an effort to provide an application bul-
letin. Committee Chairman Ernest T. Selig, in his paper, reviews the symposium and
sets forth some of the needs as established by the committee, one of which is to "de-
fine and categorize all of the important failure criteria for design, and indicate ex-
pected safety factors in current practice."
This RECORD will be valuable to those concerned with the proper design and con-
struction of culverts because it brings together the latest theories along with discus-
sions of recent experimental work and construction practices. A historical review sets
the stage for presentations describing material properties, balanced design, finite-
element analysis, structural design practices, and design charts.
In addition to the seven symposium papers, there are discussions of buckling of cyl-
inders and of the historic Iowa deflection formula.

iv
SUBSURFACE SOIL-STRUCTURE INTERACTION: A SYNOPSIS
Ernest T. Selig, State University of New York at Buffalo

•THE loads imposed on a structure buried in soil depend on the stiffness properties of
both the structure and the surrounding soil. This results in a statically indeterminate
problem in which the pressure of the soil on the structure produces deflections that in
turn determine the pressure. This soil-structure interaction is a subject that has been
of technical interest for many decades, and some of the basic concepts and design
methods in use today were initiated in the early 1900s. More recently, as a result of
an increase in the cost and the importance of buried structures, new information has
been accumulated through research, analysis, and testing. However, much of this
knowledge has not yet been adapted to design practice, and there are still many ques-
tions to be resolved.
The Highway Research Board Committee on Subsurface Soil-Structure Interaction,
which has responsibility for this general topic, has set as its immediate goal a com-
pilation of the essence of current knowledge with the purpose of improving design pro-
cedures. Although ultimately the publication of an applications manual may result, the
necessary first step is an assessment of the state of the art for the purpose of estab-
lishing the known facts, the areas of accomplishment, the subjects of uncertainty, and
the problems needing further research.
From the outset of the subcommittee effort, the scope and complexity of the subject
caused difficulty in establishing the best method of separating the subject into topics for
review. Historically, the design of underground structures has been subdivided into a
few general categories based on flexibility, configuration, and size. For example, a
structure would be classified as an arch or circular or box culvert based on shape, and
the design procedures would be classified as rigid or flexible. The former usually
represents corrugated steel pipe, and the latter represents reinforced concrete pipe.
Each of these subdivisions had empirical design methods associated with it. One of the
obvious limitations of this approach is that structures do not necessarily fit precisely
into one of the categories. Furthermore, the existing transition from one group to
another is not considered, and the limits of applicability are not clearly defined.
Current design should continue to involve available methods that are backed up by
field experience when these methods are applicable. At the same time, however, new
approaches are needed that incorporate the fundamental system parameters and that
are more suitable to larger structures, greater loads, new materials, and better in-
stallation tee hnique s.
At a symposium sponsored by the Committee on Subsurface Soil-Structure Interac-
tion the state of the art was reviewed and past accomplishments and future needs were
discussed. The subject was subdivided into the following categories:
1. Historical development: major past efforts and their chronological relation;
2. Material properties: basic properties of the structure and surrounding medium
and their influence on performance;
3. Experimental studies: laboratory and field experience that aids in understanding
or improving design procedures;
4. Analytic methods: finite element and other procedures for predicting perfor-
mance; and
5. Design philosophy: methods of design used in practice and their advantages and
limitations.

Sponsored by Committee on Subsurface Soil-Structure Interaction.

1
2

The paper by Linger briefly traces the history of the major accomplishments on the
topic of soil-structure interaction. The earliest work concerned conventional conduit
design and had as its focal point the contributions of Marston and Spangler at Iowa State
University. Most of the recent analytic and experimental contributions have come from
sponsored research dealing with protective construction. In the literature, frequent
mention is made of the term "arching," and numerous explanations of the arching phe-
nomenon are given. However, confusion and disagreement still exist as to the mean-
ing and cause of arching.
Arching should be considered as the transfer of load to or away from buried struc-
tures as a result of the difference in stiffness properties of the structure, with its ad-
jacent encompassing material, and the surrounding expanse of soil. The stress dis-
tribution around the structure is therefore different from that which would exist in the
same region of soil if the structure were not present. This latter condition is some-
times referred to as the free field. The paper by Allgood and Takahashi defines arch-
ing A as

where p 1 is the vertical pressure on the structure at the crown, and Pv is the free-field
vertical stress at the elevation of the crown. If the deformation characteristics of the
structure are the same as those of the soil, then p 1 = Pv and A= O; i.e., no change in
the state of stress occurs because of the presence of the structure. If the structure
is not as stiff as the soil it replaces, then p1 < p. and A > O, i.e., the arching is posi-
tive. Conversely, if the structure is stiffer than the soil, then p1 > Pv and A < O; i.e.,
the arching is negative. If the structure is surrounded by a zone of material that dif-
fers from the free-field soil, the same concept applies as long as the structural unit is
taken to be the structure together with the zone of material. For example, a rigid con-
crete structure encompassed in a layer of polyurethane foam or loose soil may have
positive arching rather than negative because the composite system is not as stiff as
the free-field soil.
In order to provide a quantitative definition of flexible and rigid structures, we must
consider both the properties of the soil and the properties of the structure. Allgood
and Takahashi recommend for circular culverts the use of the nondimensional term
M D3 /EI, where M. is the secant modulus of the soil in one-dimensional compres-
0

sion, D is the pipe diameter, E is the modulus of elasticity of the structure, and I is
the moment of inertia per unit length of the pipe wall. The classification proposed is
as follows:
1. Flexible: M. D3/EI > 10~,
2. Intermediate: 10 1 s M,D3/EI s 104, and
3. Stiff: M,D3 / EI s 10 1 •
Structures in the stiff category will experience negative arching, whereas structures
in the flexible category will experience positive arching. The transition occurs in the
intermediate category for which the most common design methods are least applicable.
The properties of the structure and the surrounding medium must be considered in
any rational design. Determination of the important structural parameters is relatively
straightforward, and accepted procedures are available for obtaining numerical values
with sufficient accuracy. In contrast, measurement of the appropriate parameters for
the surrounding medium is much more difficult. This is partly a result of the inherent
complexity of soil stress-strain relations, but it also results from a need to incorporate
the influence of bedding conditions and variations caused by construction procedures.
Most current design methods treat the system properties, particularly those associated
with the soil, by grouping them into several broad categories or by using empirical
parameters selected by experience and judgment. Few researchers rely on testing to
obtain quantitative values. However, the newest computer methods use rational mate-
rial properties that are more easily defined, and procedures are being prepared for
tests to provide direct determination of these properties.
A thorough discussion of the topic of material properties in relation to soil-structure
interaction is provided in the paper by Krizek and Kay. In addition, Parmelee and
 3

Corotis review the parameters required in the commonly used Iowa deflection formula
for flexible pipe design, indicating the empirical nature of these parameters and the
difficulty in relating them to measurable soil properties.
During the past two decades, a variety of laboratory model studies have been con-
ducted as part of research to better understand soil-structure interaction and to im-
prove on the theories. These studies have been very valuable in determining the key
parameters and in demonstrating their influence. Model tests are also useful for com-
paring the effect of new sets of conditions with those for which previous experience
exists. Examples are uncommon loading situations, new culvert shapes, or the altera-
tion in load on one culvert by an adjacent one in multiple installations. Model studies
on the other hand have serious limitations in quantitative prediction of full-scale per-
formance because of the difficulty in modeling field conditions. For example, the load-
ing is often caused by soil weight, which is not easily scaled, along with depth and stiff-
ness, and details in the construction process such as buildup of backfill in thin layers
are hard to represent correctly.
Field observations of buried structure performance are badly needed to prove new
theories and refine existing ones. However, few suitable data are available, and the
cost of obtaining needed information is substantial.
Papers by Nielson and Statish and by Nielson describe some of the past experimental
work on culverts. The major omissions in these papers are the results of studies in
protective construction research and studies using other experimental techniques, such
as photoelastic models, to investigate soil-structure interaction.
One of the major problems in developing a suitable analytic method for design of
buried structures is the difficulty in defining failure. For example, failure may be
based on either local or general buckling, seam or bolt rupture, substantial cracking
of concrete, or deflections sufficient to cause surface subsidance. The approach taken
to analyze for buckling failure, for example, is given in the paper by Chelapati and
Allgood.
The elasticity theory has been useful in providing some general trends, but the more
versatile finite-element analysis provides the most comprehensive analytic tool avail-
able for predicting load distribution on buried structures. For example, nonlinear soil
behavior, bedding details, slippage between the soil and the structure, and any desired
geometric shape can be analyzed. The paper by Allgood and Takahashi shows the bene-
fits of this method in relation to other methods.
By using the finite-element method we can carry out an analysis of a buried struc-
ture to any degree of detail desired. Of course, the greater is the refinement, the
greater is the cost. The limiting constraint then is the ability to define the real con-
ditions, particularly the soil properties and bedding conditions, in order to properly
simulate them analytically. It is feasible now to establish package computer programs
that can analyze common culvert situations at an economical cost.
Design practice in New York State is described in a paper by Butler to illustrate the
manner and degree to which past research has been applied. Not only must structural
design be considered to resist the soil loads, but handling qualities during construction
and durability to withstand adverse environmental conditions must also be incorporated
into the design factors. Recommendations for further research to improve design
methods are also suggested by Butler.
Following the symposium, a general discussion was held to review the achievements
and to establish the steps that should be taken to apply the new information to design
practice. The following tasks were identified as the most important steps to be ac-
complished under the direction of the subcommittee:
1. Define the basic terminology such as arching, backpacking, and bedding;
2. Determine the key parameters and groups of parameters that determine the per-
formance of the structure and the soil, giving recommended standard symbols for their
designation;
3. Define and categorize all of the important failure criteria for design, and indicate
expected safety factors in current practice;
4. Outline important aspects of installation techniques, and indicate the desired in-
spection procedures to ensure satisfactory results;
4

5. Define the requirements for suitable backfill;

6. Describe the important material properties for the soil and the structure, and in-
dicate how they should be measured;
7. List the requirements for field tests to verify design concepts;
8. Recommend steps to be taken for application of research results to design prac-
tice; and
9. Prepare educational plans for dissemination of available information.
The plan of action is to complete many of the aforementioned tasks through com-
mittee effort, drawing on available information. Needed research and more extensive
effort that may be required to develop design procedures will be recommended.
Based on the presentations at the symposium, it may be concluded that (a) informa-
tion exists from past research, and more is being generated that should be incorporated
into design practice; (b) current methods for design of small culverts must be modified
or replaced by methods that can accommodate large culverts and new structural mate-
rials with proper economy; and (c) agreement is needed on the best methods to describe
and measure the relevant properties of the structure and the surrounding media. Fur-
ther activity directed to the accomplishment of these tasks will be valuable to the pro-
fession.
HISTORICAL DEVELOPMENT OF THE
SOIL-STRUCTURE INTERACTION PROBLEM
Don A. Linger, University of Notre Dame

The term "soil-structure interaction" refers to the general phenomena in-

volved in the behavior of buried structures as a result of the properties of
both the structure and the surrounding medium in response to loading im-
posed on this system. This subject has been of technical interest for many
decades, and some of the basic concepts and design methods in use today
were initiated in the early 1900s. More recently, as the cost and importance
of buried structures have increased, new information has been accumulated
through research, analysis, and testing. However, much of this knowledge
has still to be adapted to design practice. This paper traces the historical
development of the subject of soil-structure interaction.

• THE subject of earth pressure and its application in engineering design have been dis-
cussed since the time of Rankine and Coulomb. Since that time considerable effort has
been expended in the determination of loads on retainment and underground structures.
The term "soil-structure interaction" is used because of the indeterminate effects of
the interaction between a structure and the soil. This indeterminancy is the result of
the distribution and magnitude of earth pressure varying with the amount and type of
deflection of the structure. The phenomenon of an earth pressure that is related to
soil deformation was recognized by Rankine and is referred to as the active and the
passive Rankine state in the analysis of horizontal earth pressures. It is, of course,
obvious that the general phenomenon is not adequately defined by this definition. The
soil properties and condition, the structural geometry and rigidity, and the character-
istics of the loading all affect the magnitude and distribution of earth pressure on a
structure. All of these characteristics are combined into the very complicated, in-
determinate problem of soil-structure interaction.
Until recently the design of buried structures was based primarily on the loading
produced by the overburden material on the structure, with only the shallow buried
structure receiving any significant live load. The advent of nuclear weapons and the
resulting need for protective structures brought a new dimension to the study of loads
on buried structures with loadings that are orders of magnitude greater than earlier
loadings. Almost simultaneously, the development of the Interstate highway program
began requiring more and larger highway culverts with greater fill heights and culvert
loadings than ever before.
The increase in highway construction and the national defense requirements renewed
the interest in underground structures. This interest has resulted in large-scale re-
search and development projects directed at the problem of soil-structure interaction.
It has also made us aware of the shortcomings and unknowns in the design of under-
ground structures.
Most important, however, this increased research effort has resulted in a corre-
sponding increase in the level of knowledge on the subject. Moreover, the subject has
received so much attention that it is difficult to keep abreast of the technical advances.
As a result of these great strides in research, development, and design knowledge of

Sponsored by Committee on Subsurface Soil-Structure Interaction.

5
6

soil-structure interaction, it is important to occasionally review the status of what we

know, or think we know, about the subject. The objective of this symposium is to re-
view the state of the art of soil-structure interaction knowledge in order to stimulate
current research and development on this subject.
The soil-structure interaction symposium has been broadly divided into the subjects
that generally provide the topical areas for design and research. Each of these broadly
defined subjects has been discussed in detail by the other authors contributing to this
symposium. However, it is the purpose of this paper to present a comprehensive
coverage of the historical background on the subject of the design of underground struc-
tures, a subject often referred to as the soil-structure interaction problem.
For convenience, the subject has been divided into two major areas: the development
of concepts in classical culvert design and the development of phenomenological concepts
in the response of buried structures. These two subject areas are intimately related
because both deal with the same subject. However, this division allows the reader to
follow the development of soil-structure interaction with a clearer perspective of the
research and development efforts.
The first area, classical culvert design concepts, traces the development of an ap-
proach that has attained a level of acceptance that is characteristic of traditional earth
pressure theories. The improvements and refinements in this approach are significant
and have formed the basis for the design of buried conduit. These theories are still
applicable and are used currently in design.
The second area, phenomenological concepts, traces the various studies that have
made significant developments in the understanding of the soil-structure interaction
problem. Many of these studies have been milestones in the understanding of the
phenomenon, but the application of the results of these studies is sporadic and often
lost in the confusion of technical advances.
The references discussed in this paper are not inclusive. An extensive bibliography
concerning the period 1900 to 1968 was compiled by Krizek, Parmelee, Kay, and El-
naggar (!).

DEVELOPMENT OF CONCEPTS
Not enough is known about soil-structure interaction to predict with any degree of
accuracy the ultimate load-carrying capacity of buried structures. In the design of
underground structures, the loading is usually based on empirical relations that are
not fully understood. If the loading on the underground structure is determined from
classical earth pressure theory, large variations can be expected between the actual
and the theoretical loadings. These variations can result from the underground struc-
ture deflecting more than the adjacent soil and thereby causing a reduction in the pres-
sure transmitted to the structure with a corresponding increase in the pressure carried
by the adjacent medium. Conversely, under load, the structure may not deform as much
as the adjacent soil, and the resulting redistribution can produce an increase in load on
the structure and a decrease in the pressure carried by the adjacent medium. These
two opposite conditions are similar to the active and passive earth pressure conditions
defined by Rankine more than 100 years ago. The difference in the conditions is deter-
mined by the direction of the soil stress produced by the soil movement along some
slippage plane. Because of the elegance of the classical earth pressure theory, it is
understandable that the earlie st approaches to the loading of buried structures should
take a form similar to the Rankine earth pressure theory. The two opposite conditions
of soil-structure interaction loading are characterized by the soil-structure systems
shown in Figures 1 and 2.
The amount of pressure redistribution is very difficult to quantify and depends on
the degree to which the relative deflection along the shearing plane has mobilized the
soil shear strength. From this it is apparent that identifying the location of the
shearing plane and the amount and type of stresses induced along the shearing plane
is an important part of defining the problem ..
In the case of a large underground structure deflecting under load, the soil at the
center of the roof span of the structure displaces with respect to the soil over the sup-
ports and also with respect to the adjacent soil in which it is buried. Because of the
 7

differential deflection of the various parts of the structure and the relative flexibility
of the soil and the buried structure, the soil-structure interaction phenomenon will
occur as a redistribution of pressure among various segments of the structure in
addition to the redistribution of load from the structure to the adjacent soil. This
simplification of a very complicated problem is shown in Figure 3.

Classical Concepts
Marston was the first to recognize that the loading on an underground structure is
dependent on the interaction of the structure and the surrounding soil. In 1913 he pub-
lished the Marston theory on soil loads on drainage pipes (£). This theory was based
on a prism of soil whose movement developed the forces shown in Figure 4 as it im-
posed a load on the underground structure. This theory clearly took into account the
relative deflection of the pipe and the settlement of the soil. However, the design of
the buried pipe was based on vertical loads only and was only applicable in the design
of buried rigid pipes such as clay tile or concrete pipes.
The earliest development of flexible conduit design criteria was based on empirical
equations developed by using the results of the 1926 American Railway Engineering
Association investigation ~). Tables were developed for the necessary pipe thick-
ness and diameter for various heights of fill. The design tables were based on the
assumption that failure occurred when the pipe deflection reached 20 percent of the
diameter. For design, the deflection was limited to 5 percent, thus providing a safety
factor of 4. It is interesting to note that no attempt was made to correlate the load-
carrying capacity with soil characteristics, and therefore there was little evidence of
any understanding of soil-structure interaction. It is also interesting to note that the
original fill height versus pipe requirement table was the forerunner of the gauge tables
commonly used today.
As highway construction increased during the 1930s, the use of larger and more
costly drainage structures also increased. The need for a more rational concept for
the design of flexible pipes was observed by Spangler, a former student of Marston.
Consequently in 1941, Spangler (1) published his Iowa formula for predicting the de-
flection of buried flexible pipe. Spangler introduced the first well-defined soil-structure
interaction concept (Fig. 5). This concept recognized that a passive type of soil pres-
sure is developed by the horizontal expansion of the pipe, which allowed the pipe to
carry more load with less deflection than in the unrestrained condition. Moreover,
he proposed that the deflection might be used as a basis for determining the magnitude
of the horizontal pressure developed on the sides of the pipe. He defined the propor-
tionality constant between the pipe deflection and the developed pressure and proposed
limiting values for use in design. This method was the first procedure that required
an evaluation of the necessary soil properties for application in design.
In 1960, White and Layer (Q) proposed the ring compression theory for the design
of flexible buried pipes as shown in Figure 6. This theory assumes that the ring de-
flection of the structure is negligible and that the failure occurs by the crushing of the
pipe walls. Model tests were conducted separately by Meyerhof (!!) and Watkins (1) to
evaluate the ring compression theory. The results of these studies showed that failure
could result from an additional parameter, that of the buckling of the culvert wall (Fig. 7).
Further studies by Watkins (!!), Meyerhof and Baikie (§), and Meyerhof and Fisher
(!Q) resulted in the further refinement of structural response in terms of the deflection,
crushing, and buckling aspects of the buried structure. However, in some of these
studies, loosely defined soil terms such as "good backfill," "compressible soil," and
"plastic soil" appear in the description of formulas and coefficients.
It is apparent that by the middle 1960s extensive studies had defined the problem and
isolated the important parameters, but a complete definition of the parameters did
not exist. In 1967, in an attempt to further clarify the interaction of the soil char-
acteristics and the deflection of the structure, Nielson presented a theory for de-
termining loads on buried conduit by an arching analysis (!!). The proposed method
used an adaptation of the Spangler deflection equation, but, despite the apparent good
agreement with the experimental data studied, little use has been made of this novel
8

diversion from the classical Marston procedure . The proposed arching condition is
shown in Figure 8.
An interesting aspect of the research and design procedures is the way in which
generally accepted methods treat either rigid structures or flexible structures, with
adequate procedures being available for each. However, only limited research has
been directed toward development of a comprehensive design procedure covering the
full range of pipe stiffnesses.

Phenomenological Concepts
Even though soil-structure interaction phenomena are still not completely under-
stood, it was recognized during the early studies that the overall compressibility of
the structure relative to the soil it replaces is important. Terzaghi treated this
phenomenon in considerable detail in his trapdoor tests (g) . This was one of the
first papers to comprehensively evaluate the stress distribution on a structure in a
fully buried condition. Terzaghi discusses the fundamental assumptions of the re-
searchers who contributed to an understanding of the problem (ll_): Engesser (1882),
Bierbaumer (1913), Coquot (1934), and Vollmy (1937). The principal contribution of
these studies was to delineate the formation of the soil surface along which the soil
arching stresses were mobilized (Fig. 9).
One of the next major milestones was a paper by Whitman, which reported on
the results of a buried dome study in which the soil was simulated by a uniformly
placed granular backfill (li). These tests were a part of a program that set an ex-
ample for many of the tests that followed in the study of buried structure responses.
The requirements for buried structure design criteria resulted in a unique con-
ceptual approach developed and presented by Newmark and Haltiwanger (1Q). This
publication advanced many new ideas and provided the impetus for much of the research
that followed .
In 1964, the state of the art of soil-structure interaction was reviewed at a sym-
posium held at the University of Arizona <1.!1_). The participants at this symposium dis-
cussed in detail the various aspects of the phenomenon. A paper by Triandafilidis et
al. (!.'V delineated the important variables of soil-structure interaction in a series of
tests performed on vertical cylindrical and disk structures designed to separate arch-
ing stresses from sidewall friction effects. The results of this study provided the
necessary quantitative data to enable researchers to make an analytical relation be-
tween structure stiffness and medium stiffness and the load on the structure.
At this symposium, Luscher and Hoeg(!§_) presented the results of a study that
discussed the uncertainty in the lateral pressures acting at the sliding surface. Con-
siderable attention was given by these authors to the assigned values of the at-rest and
active pressure coefficients used by other investigators . The results of a study by
Donnellan (19) were reported, which demonstrated the effect of depth of burial on the
load-carrying capacity of a cylinder. Donnellan's study defined the shallow and deep
burial conditions and the effect of burial depth on the deflection behavior of rigid and
flexible cylinders. An example of these results is shown in Figure 10.
Additional studies reported at this symposium by Selig ~ defined the methodology
for the measurement of soil pressures and deformations with great accuracy. This was
an important step forward in the research on soil-structure interaction. It was this
aspect that implied that soil tests could be used to evaluate and define the necessary
properties for the design of soil-structure systems. Researchers were quick to begin
studies of the effect of soil characteristics on the response of buried structures. A
notable example of this effort was one of the studies of Allgood (W. This research
presented a method for determining deflections and critical buckling loads based on
the one-dimensional confined compression modulus of the soil. From this research,
it was apparent that we had a handle on the problem of the interaction of the structure
and the soiJ. The next obvious step was to develop the means of modifying the pressure
on the structure by modifying the characteristics of the surrounding soil. This ap-
proach had been tried by Spangler ~) with considerable success but without any
quantified design criteria .
Figure 1. Active soil pressure condition. Figure 2. Passive soil pressure condition.

DEFORMED ORIGINAL
GROUND
GROUND
SURFACE
SURFACE
SHEAR EFFECT

UNDEFORMED
I, ,r DEFORMED
UNDEFORMED DEFORMED
STRUCTURE , Y STRUCTURE
STRUCTURE
} STRUCTURE

-------~ ~

Figure 3. Soil pressure redistribution. Figure 4. Marston theory of soil loads on

rigid buried pipes.

y = UNIT WT, OF SOIL IN

LBS. /CU. FT.
( A) 181 µ.= SOIL COEFFICIENT
OF FRICTION
REDISTRIBUTION OF REDISTRIBUTION OF
PRESSURE ON PRESSURE ON C = SOIL CONSTANT
STRUCTURE STRUCTURE AND SOIL BASED ON K , y, fl.
W = CyB 2 • MARSTON
LOAD ON BURIED PIP(

Figure 5. Loading assumptions for the Iowa Figure 6. Ring compression theory for design of
formula. flexible pipe embedded in dense soil.
LOAD C pD
LOAD
( DEVELOPS VERTCAL SOIL PRESSURE p

~
m·"'- :i
AT TOP OF PIPE l

:; ~

E'•~ "
!'.X . BY STATIC EQUILIBRIUM
BY DEFINITION pD
fc = 2A

l\X
T

Figure 7. Ring buckling curves for buried flexible pipes. Figure 8. Free-body diagram for arching
analysis.
~:
u
FILL SURFACE

~
~I
~ 5 0 t---- - , - -----r---...--,--,..------.,....--,---r---,--, FR~~~~ON
:r ANGLE
5 40 !
~ r - -; - --,;-o;::::-,....;.,.---t--+--"-'-f--'-"-""f'----ir--+ cj, = 90°
S yp
~ 30 ¢ 45° C

Q ¢ 30• C

E 201---- + --+----i--t - -
Nole Location of soil arch
support is of loco lion
ofmoximumshearstress
~ btlore 1,oH o,cr-.b, to,mtd
"
8 10
"ii'z
Cl)
o~-~-~-~--~-~-~---~~-~-
o OI o,7 o.e o.9
0 ,2 0 .3 0.4 0. 5 0.6
F::. RING FLEXIBILITY MODULUS •+(-ft 2
( INC H KIP-I)
10

Figure 9. Formation of soil arching Figure 10. Normalized radial displacement

stresses. of horizontal cylindrical structure
(diameter/thickness= 114).

--er
I r 11 ,,- REAL
SURFACE
+
0 ,2

to -
0 .2
+
to
:1
I/ /- ASSl)J,lf.O
SURFACE
0 ,2

Ort
0 .2

04
/.

8w, ot dcplf'l I ( b} 8urio1 O@lh 3

( o ) -- - - • -
°'omelc:r 16 Olometer "4

Figure 11. Effective density profile for standard Figure 12. Effective density profile for uncompacted
backfill. backfill over culvert.
151
- Roadway Embonkmenl

Slruclure
Backfill Structure
Bock rill

Figure 13. Effective density profile for straw backfill

adjacent to culvert.

Structure
Backfill
1
1-6" S1raw
Bock fill
 11

After the state of the art was advanced further, this procedure was explored once
again by the California Division of Highways in an extensive research program that in-
cluded the measurement of soil pressures on several large full-scale culverts with
various conditions of the surrounding media. These results were reported by Davis
and Bacher ~) and present an interesting characterization of the changes that oc-
curred during the development of further insight into the soil-structure interaction
phenomenon. The changes produced in the soil-structure interface pressure in this
study by using the soft "backpacking" material are evident in Figures 11, 12, and 13.

SUMMARY
The problem of soil-structure interaction is illustrated by the design requirements
for culverts to carry tremendous overburden fill heights and for complex buried struc-
ture systems to resist large surface loadings. The problem is further complicated by
the scarcity of failures attributable to design shortcomings and the difficulty in eval-
uating a failure when it does occur.
Current design practice is based largely on work conducted in the 1920s and 1930s,
and, despite the success of these practices, they are empirical in nature and depend
heavily on experience and engineering judgment. However, recent refinements in
these procedures have made possible much more daring uses of soil-structure inter-
action concepts.
Design engineers now have enough confidence to construct soil-structure systems
in which the soil pressures are controlled by the backfilling techniques or the back-
filling materials. This concept seems to have great potential, but the irony of this
"breakthrough" is that it was first presented by Spangler and Marston as a result of
their first tests on buried conduit, and it was called the "imperfect ditch method."

REFERENCES
1. Krizek, R. J., Parmelee, R. A., Kay, J. N., and Elnaggar, H. A. structural
Analysis and Design of Pipe Culverts. NCHRP Rept. 116, 1971, 155 pp.
2. Marston, A., and Anderson, A.O. The Theory of Loads on Pipes in Ditches
and Tests on Cement and Clay Drain Tile and Sewer Pipe. Eng. Exp. Sta., Iowa
state College, Bull. 31, 1913.
3. American Railway Engineering Association. Culvert Load Determination. Bull.
284, 1926.
4. Spangler, M. G. The structural Design of Flexible Pipe Culverts. Eng. Exp. Sta.,
Iowa state College, Bull. 153, 1941.
5. White, H. L., and Layer, J. P. The Corrugated Metal Conduit as a Compression
Ring. HRB Proc., Vol. 39, 1960, pp. 389-397.
6. Meyerhof, G. G., and Baikie, L. D. strength of Steel Culvert Sheets Bearing
Against Compacted Sand Back Fill. Highway Research Record 30, 1963, pp. 1-19.
7. Watkins, R. K. Failure Conditions of Flexible Culverts Embedded in Soil. HRB
Proc., Vol. 39, 1960, pp. 361-371.
8. Watkins, R. K. structural Design of Buried Circular Conduits. Highway Research
Record 145, 1966, pp. 1-16.
9. Meyerhof, G. G. Composite Design of Shallow Buried Steel structures. Proc.
47th Canadian Good Roads Assn. Conf., 1966.
10. Meyerhof, G. G., and Fisher, C. L. Composite Design of Underground steel
structures. Eng. Jour., Vol. 46, No. 9, 1963.
11. Nielson, F. D. Soil structure Arching Analysis of Buried Flexible structures.
Highway Research Record 185, 1967, pp. 36-50.
12. Terzaghi, K. stress Distribution in Dry and in Saturated Sand Above a Yielding
Trapdoor. Proc. Internat. Conf. on Soil Mechanics, Vol. 1, 1936.
13. Terzaghi, K. Theoretical Soil Mechanics. John Wiley and Sons, 1943.
14. Whitman, R. V., Getzler, Z., and Hoeg, K. static Tests Upon Thin Domes Buried
in Sand. Jour. Boston Soc. of Civil Engrs., Vol. 50, 1962.
15. Newmark, N. M., and Haltiwanger, J. D. Air Force Design Manual. U.S. Air
Force Special Weapons Center, Rept. AFSWC-TDR-62-138, 1962 .
12

16. Proceedings of the Symposium on Soil-structure Interaction. Univ. of Arizona,

Tucson, Sept. 1964.
17. Triandafilidis, G. E., Hampton, D., and Spanovich, M. An Experimental Evalua-
tion of Soil Arching. Proc. Symp. on Soil-structure Interaction, Univ. of Arizona,
1964.
18. Luscher, U., and Hoeg, K. The Beneficial Action of the Surrounding Soil on the
Load-Carrying Capacity of Buried Tubes. Proc. Symp. on Soil-structure Inter-
action, Univ. of Arizona, 1964.
19. Donnellan, B. A. The Response of Buried Cylinders to Quasi-static Overpressures.
Proc. Symp. on Soil-structure Interaction, Univ. of Arizona, 1964.
20. Selig, E. T. A Review of stress and strain Measurement in Soil. Proc. Symp.
on Soil-structure Interaction, Univ. of Arizona, 1964.
21. Allgood, J. R., Ciani, J.B., and Lew, T. K. Influence of Soil Modulus on the
Behavior of Cylinders Buried in Sand. U.S. Naval Civil Eng. Laboratory, Port
Hueneme , Calif., Tech. Rept. R-582, 1968.
22. Spangler, M. G. A Practical Application of the Imperfect Ditch Method of Con-
struction. HRB Proc., Vol. 37, 1958, pp. 271-277.
23. Davis, R. E., and Bacher, A. E. California's Culvert Research Program-De-
scription, Current status, and Observed Peripheral Pressures. Highway Re-
search Record 249, 1968, pp. 14-23.
MATERIAL PROPERTIES AFFECTING SOIL-STRUCTURE
INTERACTION OF UNDERGROUND CONDUITS
Raymond J. Krizek, Northwestern University; and
J. Neil Kay, University of California, San Diego

Notwithstanding current limitations of analytic procedures, the greatest

error in available techniques for the analysis and design of underground
conduits probably lies in the specification of material properties, especially
those for the soil surrounding the conduit. However, the mechanical prop-
erties of the conduit material and the conditions that exist at the soil-
conduit interface may be significant. Included herein is a brief discussion
of the material parameters that form a part of the classical procedures for
the analysis and design of pipe conduits, and the arguments against their
continued long-term use are given. The advent of the high-speed digital
computer and the finite-element method have provided the opportunity to
handle material properties in a more realistic manner, and soil-conduit
problems should be formulated to take this fact into account. Nonhomo-
geneity resulting from different materials being used for the underlying
soils, bedding, side fill, and backfill or embankment can be readily in-
cluded in the analysis, and incremental approaches allow nonlinear mate-
rial properties and the actual construction sequence to be incorporated in
a piecewise linear manner without too much difficulty; even three-
dimensional analyses by numerical methods have recently come into use.
Accordingly, appropriate soil properties must be specified to guide the de-
velopment of increasingly sophisticated analyses and computer programs.

•THE analysis and design of underground conduits are essentially a problem of soil-
structure interaction, and the solution of any problem of this type must give full cog-
nizance to the fundamental coupling phenomenon. Interpreted simply, this concept
states that the response of the conduit and the behavior of the surrounding soil are not
independent but intimately related in some complex manner. In general, the response
of a soil-conduit system depends on the characteristics (geometry and stiffness) of the
conduit, the characteristics (geometry, order of placing, and mechanical properties)
of the adjacent and overlying compacted fill, and the characteristics (compressibility)
of the in situ soil under and adjacent to the conduit. Notwithstanding the limitations of
analytic procedures, the greatest error in currently available techniques for analysis
and design probably lies in the specification of material properties, especially those
for the soil surrounding the conduit. However, the mechanical properties of the con-
duit material and the conditions that exist at the soil-conduit interface may also be very
significant. This paper will discuss briefly some of the ideas that can be used to de-
termine appropriate input information for material properties.

CURRENT DESIGN PROCEDURES

For the most part, buried conduits are constructed of either reinforced concrete or
corrugated metal, and the well-known work by Marston and Spangler and their co-
workers has exerted a significant influence on virtually all currently used design pro-
cedures (~). Generally accepted design methods treat reinforced concrete conduits as

Sponsored by Committee on Subsurface Soil-Structure Interaction.

13
14

a rigid structure and metal conduits as a flexible structure, and separate design pro-
cedures are available for each. In an effort to account for the soil-structure interac-
tion (or the relative stiffness of the soil and the conduit), these procedure involve a
variety of special parameters (such as settlement ratio, modulus of soil reaction, load
factor, and projection ratio) that are associated specifically with the buried conduit
problem. Although these parameters may achieve their intended goal when used with
good engineering judgment within limited ranges of applicability for which experience
is available, very often their use cannot be easily extended or generalized. Also of
considerable concern is the fact that no techniques are currently available to handle
conduits of intermediate stiffness. Despite the limitations outlined, these procedures
have served the profession well during the past 50 years, and few, if any, failures can
be attributed to the theory itself. As such, this work, including the special parameters
that attempt to account for the soil-conduit interaction phenomenon, represents an out-
standing example of engineering ingenuity, and the experience gleaned over the years
must not be treated lightly. Although good engineering practice dictates th.at currently
used design procedures should not be discarded until better ones have been provided,
this same good practice calls for a periodic appraisal of current procedures in the light
of recent advances in technology and theoretical developments.

CONTINUUM APPROACH
Despite the advantages and disadvantages that may be attributed to the Marston-
Spangler theories, it seems that the major advances in our knowledge of soil-conduit
interaction phenomena do not lie in modifying or improving the existing procedures and
the associated material parameters but rather in developing a different approach to the
problem. Pursuant to this idea, the most logical approach to the soil-conduit interac-
tion problem lies in treating all components (conduit, underlying soils, bedding, sidefill,
and overlying soils) as continua, each with its unique material properties. Although the
complexities associated with the geometry and material properties of a typical soil-
conduit system have in the past either precluded the use of this approach or necessitated
relatively crude computational procedures, a very versatile analytic tool has been made
available to the profession in recent years by the development of the finite-element
method and the advent of the high-speed digital computer. In addition to providing the
capability for describing the soil-conduit system as a nonhomogeneous, nonlinear con-
tinuum, such a treatment has the following advantages: The coupling or soil-conduit
interaction effect is inherently taken into account; input parameters would consist of
more fundamental characterizations of the soil and conduit material behavior; conduits
of intermediate stiffnesses can be analyzed; and the effects of the construction sequence
can be studied. Even three-dimensional analyses have recently been made. Accord-
ingly, the following discussion of material parameters is based on the premise that the
finite-element method offers the potential for significant improvement in our ability to
analyze complex soil-conduit systems.

INTERRELATED STEPS IN DESIGN PROCEDURE

Any design procedure consists, either implicitly or explicitly, of a synthesis of the
various steps shown in Figure 1; this diagram indicates that the design procedure is
intimately related to and dependent on the sampling and testing techniques, the interpre-
tation of the data, and the methods of analysis that are used. Accordingly, a change in
the design procedure is likely to bring about changes in one or more of the other steps
involved. In particular, use of the finite-element approach leads to a considerable
change in currently used design procedures for underground conduits because it requires
that the problem be formulated in terms _of material properties that are fundamental to
continuum mechanics. However, in certain cases these properties have been studied
for years, and many soils laboratories are currently equipped to conduct the required
tests.
 15

REQUIREMENTS OF VALID DESIGN PROCEDURE

As shown in Figure 2, three very important components a.re required to develop a
valid design procedure: a mathematical model, material properties, and field verifica-
tion. The mathematical model, which would probably include a computer program,
must be formulated such that it can describe the physical phenomenon under considera-
tion; in the past this component of the overall problem has attracted much attention,
and many sophisticated programs have been developed. The applicability of these pro-
grams, however, is limited by the assumptions on which they are based and the mate-
rial properties that are provided as input; very often these programs call for input data
that cannot reasonably be provided, and hence the engineering profession obtains little
benefit from their use. As has been frequently stated, the results obtained from any
computer program are only as good as the input information that is supplied. There is
considerable evidence to indicate that in recent years our ability to formulate mathe-
matical models and solve theoretical problems has far outstripped our ability to provide
appropriate input information concerning material properties. Finally, field verifica-
tion of theoretical predications is needed before any analytic procedure can be accepted;
however, the high cost of field instrumentation often impedes or totally precludes its
use, and the profession is therefore left with no reliable way to assess quantitatively
the validity of the combined mathematical model and material properties.

EMPHASIS OF STUDY
The principal objective of this study is to examine workable approaches that may be
taken immediately within the framework of currently available testing techniques to
interpret laboratory test results for use in obtaining the solution to a soil-conduit prob-
lem. The discussion is intended to be representative, not comprehensive, and the
purpose is to survey and compare various methods of testing and interpretation, not to
suggest one particular procedure. Although the properties of conduit materials and the
interface conditions between various zones are discussed briefly, the characterization
of the soil is considered to be of primary importance, and the principal thrust of the
presentation is therefore directed toward this end. Emphasis is centered around soil
testing procedures that are in common use, and the terminology and techniques of linear
elasticity are employed. The observed nonlinear behavior of virtually all soils may be
handled conveniently by a piecewise linear model, but no attempt is made to advance
more rigorous formulations and interpretations than may be realized within the capa-
bilities of most current laboratory test equipment. For example, consideration of all
three principal strains and stresses in a constitutive relation would require the conduct
of a true triaxial test; however, except for research purposes, the true triaxial test is
far too complex for widespread use at the present time, and a complete variation of the
properties in all three directions is therefore not treated herein.

SOIL PARAMETERS
It is convenient to consider two extremes of soil performance, which represent the
behavioral range of engineering soils: an ideal plastic, cohesive clay and an ideal clean,
coarse-grained, cohesionless sand. The major difference between these materials lies
not primarily in the ultimate shear stress available but rather in their stress-strain-
time characteristics for loads sustained over long periods of time. Loads on clay soils
will ordinarily cause time-dependent volume decreases, and mobilized shear stresses
may relax because of creep. The ideal cohesionless soil usually exhibits a relatively
low compressibility under added loads, and it responds with little time delay. Cohe-
sionless soils tend to develop and maintain a specific shear stress where differential
movements occur. In general, because of the time-dependent stress-strain charac-
teristics of most clayey soils, stress concentrations dissipate with time; consequently,
it is probable that the pressure normal to the conduit wall would tend to approach the
overburden pressure with the passage of time. No specific observations are available
to demonstrate the degree to which cohesionless soils can permanently sustain loads
transferred from the pipe, but it is likely that this relaxation phenomenon does not exist
16
to the same extent. Because most soils used in an actual conduit installation do not
fall into either of these extreme categories, it is difficult to predict the effect of time
on the response of a buried conduit. The type of soil (and most especially its degree
of saturation) and the time of interest in the problem will dictate whether the soil tests
to be discussed subsequently should be drained or undrained; however, for sake of
brevity, time considerations are not specifically included in this discussion, and the
manner in which they are handled is left to the engineering expertise of the designer or
researcher.

Importance of Modulus
As stated previously, the characterization of the soil is probably the most important
consideration in a soil-conduit system, and, more specifically, the modulus of the soil
is probably the single most important parameter that affects the response of the system.
In addition to the interaction between the conduit and the immediately adjacent soil,
there is interaction among the various soil zones; this is particularly important when
appraising the stiffness of the natural soil relative to the backfill in a trench installation,
and it further illustrates the importance of determining the moduli of the soils in the
various zones. Accordingly, it follows that considerable attention should be directed
toward describing the stress-strain behavior of the soil surrounding the conduit. Un-
fortunately, the modulus is intimately related to and significantly influenced by Pois-
son's ratio, and the latter is most difficult to quantify.

Assumed Isotropy of Modulus

The modulus discussed herein is assumed to be isotropic and dependent on the state
of stress in the soil at a point. The assumption of isotropy is very significant, espe-
cially in view of the nonlinear behavior of most soils, but currently available testing
procedures seem to justify no further refinement at this time. In effect, this means
that a change in stress in any given direction at a point causes a change in strain in that
direction such that the ratio of the two is a constant that is independent of the existing
stress in that direction; hence, this assumption is not strictly compatible with the non-
linear behavior of the soil. However, if this premise is accepted, the problem reduces
to one of defining the nature of the modulus and the state of stress on which it depends.

Nature of Modulus
When conducting a piecewise linear analysis that involves a nonlinear material, it
is possible to utilize three different definitions of modulus: the secant modulus (straight
line joining the origin and point on the stress-strain curve), the tangent modulus (deriv-
ative of the stress-strain curve at a point), and the chord modulus (straight line joining
two points on the stress-strain curve). Each has its inherent advantages and disad-
vantages, which are related in large part to the manner in which the numerical calcula-
tions are conducted. In view of the incremental nature of the applied load on a conduit
due to the normal construction sequence, the tangent ai1d cho,:d moduli seem to offer
some advantages; in such a formulation the problem is solved for any given state of
loading, and the stresses within each element are determined. Then, a modulus com-
patible with each state of stress is assigned to each element, and the response for the
next increment of load is determined, after which the procedure is repeated. Although
the foregoing procedure can, of course, be followed with the secant modulus, it involves
a somewhat less accurate approximation of the actual stress-strain curve.

Determination of Stress-Strain Response

Three laboratory tests may be conveniently employed to determine the stress-strain
response of a soil: the consolidation test, the conventional triaxial test, and the plane
strain test. Typical idealized states of stress, as well as typical stress-strain plots
and modulus-stress plots, for each of these tests are shown in Figure 3. As can be
readily seen, the specification of the three principal stresses in the consolidation test
and in the plane strain test requires the assumption of a value for the at-rest pressure
 17
coefficient, Ko, and Poisson's ratio, v, respectively; in the triaxial test the interme-
diate principal stress is considered to be equal to the minor principal stress. All three
of these tests represent very special states of stress, and each has deficiencies that
have been discussed at length in the literature. In all tests considered herein, the state
of stress is assumed to be homogeneous throughout the specimen; this is generally not
the case, and serious test errors can be introduced because of rough end platens and
sidewall friction. Once again, the methods for handling these conditions, as well as
other test errors, must be left to the judgment of the designer or researcher. A casual
glance at the drastically different character of the stress-strain plots obtained from
these tests suggests that a reconciliation must lie in the interpretation of the test re-
sults, and this is indeed the case. If we assume that the same drainage conditions exist
for all tests, the reconciliation of the indicated test results may be established by use
of Poisson's ratio if the soil is assumed to be linear elastic. Notwithstanding the ex-
tensive use of laboratory tests to determine soil properties, there are many engineers
who contend that any modulus determined from a laboratory test is subject to serious
error, and they advocate the use of a field test (such as a plate bearing test or a bore-
hol,e pressure meter test) to determine the soil modulus. However, these tests are
extremely expensive, and the assumption of linear elastic behavior is usually requirPrl
to interpret the results; consequently, they will not be discussed herein.

Relation Between Young's Modulus and Other Moduli

In any finite-element formulation based on a piecewise linear theory, two parame-
ters are required to characterize an isotropic material; these are the modulus of elas-
ticity or Young's modulus, E, and Poisson's ratio, v. By definition, Young's modulus
is the slope of the axial stress-axial strain curve in a uniaxial stress test, and Pois-
son's ratio is the ratio of the lateral strain to the longitudinal strain for a specimen
that is uniaxially stressed in the longitudinal direction. It is significant to note that the
definitions of both coefficients include a specification of the state of stress; this fact is
often overlooked in the interpretation of much test data. All too often one can find in
the literature a case where a conventional triaxial compression test was conducted on
a soil, and Young's modulus is defined as cr1/£1andPoisson's ratio as £3 /£1; both are in-
correct, as will be seen later. However, because the ratio ai/£1 constitutes a type of
modulus that is relatively easy to determine in a consolidation, conventional triaxial,
or plane strain test, this parameter will be termed M, Er, and EP respectively for each
of the three tests, and relations between E and each of these moduli will be determined
in terms of Poisson's ratio and the state of stress in the test specimen. Although con-
siderable discussion is given to the formulation of analytic expressions for these moduli,
graphic representations are also acceptable for use in computer programs. Most of
the subsequent discussion will center around the tangent modulus or the chord modulus
because these have the ability to approximate most closely the actual stress-strain be-
havior of the soil. Within the context of a piecewise linear formulation, the reference
state for each increment of loading may be taken as either the state of stress that exists
after the preceding load increment has been applied or the average of the stress states
before and after a given load increment (this latter choice leads to the use of an itera-
tive procedure), and a linear modulus is assumed to govern the response due to the
added load increment.

Determination of Modulus From Consolidation Test

From a consolidation test wherein the lateral strains are held equal to zero, 0'1 is
usually plotted versus £1, and a so-called constrained modulus, M, is thereby obtained.
This modulus would be appropriate for determining the settlement of a large uniformly
loaded area such that the lateral deformations are zero or negligible, but it is theo-
retically not appropriate for use in a two- or three-dimensional problem unless certain
modifications, which center around taking into account the Poisson effect, are intro-
duced. Based on the applicability of linear-elastic theory, the first invariant of the
stress tensor and the first invariant of the strain tensor can be related by
(1)
18
Because a2 = a3 = K,a1 and E:2 = E:3 = 0 in a consolidation test, Eq. 1 may be written as

(2)

which, upon substitution of the relation

Ko = v/(1 - v) (3)

and solution for E/M, becomes

E/M = (1- 2v){1 + v)/(1 - v) = (1 2

- v - 211 )/(l - v) (4)

where

(5)

As can be seen in Figure 4a, the variation of E/M as a function of v is considerable,

and this effect cannot be ignored when applying consolidation test results to multidi-
mensional problems.
One very convenient empirical expression for the constrained modulus, M, has been
reported by Janbu (!) as

(6)

where Pa is atmospheric pressure (introduced to maintain dimensional homogeneity),

and m and n are termed the modulus number and stress exponent respectively; in effect,
these latter two coefficients are empirical parameters to be determined experimentally.
However, it is very possible that certain broad correlations, as advanced by Janbu and
shown in Figure 5, may be established among these empirical coefficients and certain
types of soil or various conditions of a given soil type. As another example, limited
data, based on tests by Osterberg (6), have been interpreted by Krizek et al. (5) to sug-
gest that M is a function of dry denslty and overburden pressure for a large variety of
soils. This relation, which is shown in Figure 6, resembles the one proposed by Janbu.
Although it admittedly seems to be oversimplified, further study is certainly justified
to determine its range of applicability. The combination of Eq. 6 and Eq. 4 gives

E = [(1 - V - 21/)/(l - v)] m Pa (ai/p.) 1 -• (7)

Equation 7 deals exclusively with the interpretation of a laboratory consolidation test,

and it states specifically that Young's modulus, E, is a nonlinear function of the major
principal stress, a1. When used in conjunction with a finite-element model, the modulus
within each element for a given condition of loading would depend only on the major
principal stress in that element.
This approach might be improved by intuitively extending the Janbu relation, which
is developed in terms of a1 only, to an expression in terms of the sum of all three
principal stresses, as follows:

(8)

where m* and n* are empirical coefficients similar tom and n previously described.
Substitution of Eq. 8 into Eq. 4 yields

(9)

which is similar in form to Eq. 7 but more general in formulation. As far as the in-
terpretation of the consolidation test is concerned, the extension suggested previously
Figure 1. Interrelated steps in design procedure. Figure 2. Requirements of valid design
procedure.

Sampling - Testing .._ In terpretotion - Analy sis - Design

Field Verification

Figure 3. Typical stress-strain response from various laboratory tests. Figure 4. Modulus ratio as a
function of Poisson's ratio and state
of stress for various laboratory tests.

1.00
--....._
Consolidation Test Conventional Triaxial Test Plane Strain Test
..j..
2
0 .75

""
~ ~ ~
0
."
a:

""' \ \
050
(T - :i
<Tz =Ko
CTz=r;: 0'"3 11(~~. ~ 0.25
(o l Consolidation test
0"3= KoCT1 0'"3 I I
0
1.00

wl~ 0.75
·!a:
050
ll
. .
;
. I
:::f: 0.25
~ ~
ui (/)
ui
o
100

wl~ 0.75
Strain Strain Strain .'i
15

3
."
a:
0.50

"j 025

.
::,
.
::,

...,~ ...,
~
o., 0.2 o, o .4 o ,5
0
0
:E :E Poi sson 's Ratio, 11

Stress Stress Stress

20

is perfectly admissible; because 02 == CJ3 == K.01 in a consolidation test, Eq. 8 may be

written as

(10)

Another equally good procedure for incorporating stress-strain data into a finite-
element model is the use of the graphic representation shown in Figure 7. In this case,
one simply utilizes the chord modulus between any two arbitrary values of stress and
some representative constant value for Poisson's ratio in this same stress interval to
determine a value for Young's modulus, E, which is considered to be applicable within
this stress interval. Then, based on the state of stress (either CJ1 or 0 1 + 02 + 03, de-
pending on the formulation) of each element in the mathematical model for any given
loading, an appropriate modulus is assigned to each element for use during the next
increment of loading.
Some investigators have simply replaced Young's modulus, E, with the secant con-
strained modulus, M., which is associated with the anticipated final state of stress at
the elevation of the conduit; in other words, the effect of Poisson's ratio is completely
ignored. The apparent successes achieved with this approach indicate that various
unknown effects not taken into account tend to offset each other in certain cases. Despite
such limited successes, this approach should be used with extreme caution because it
is not on a sound theoretical basis. If, on the other hand, field verification is not ob-
tained, there is little justification to continue this theoretically incorrect practice.
Because the nature of a consolidation test precludes the determination of a failure
criterion, special provision will have to be made if this condition is approached in a
field problem, and this is, in fact, one of the disadvantages of the test. However, most
buried conduit problems are generally concerned with per.missible deformations rather
than catastrophic collapse, and for such conditions it is probably not necessary to be
overly concerned with the shear strength of a soil. The use of a consolidation test to
determine the soil modulus is particularly attractive in that the test is relatively easy
to conduct and most laboratories have the appropriate equipment. It can also be argued
(though not technically correct) that the uniaxial strain conditions of the consolidation
test are reasonably well duplicated in the free field and for many situations in a radial
direction near the conduit.

Determination of Modulus From Conventional Triaxial Test

In a conventional triaxial test on an unsaturated soil, the major principal stress, cr1,
the minor principal stress, CJ3, and the major principal strain, £ 1, are either measured
or controlled, and results, which exhibit a shape as shown in Figure 3, are usually
plotted as (01 - 03) versus £1. Based on such a plot, a special modulus, termed herein
the triaxial modulus, Er, may be defined as

(11)

This definition ignores the Poisson effect of 03 on e- 1, if considered in terms of absolute

values, and, if considered in terms of incremental values, it presumes that the Poisson
effect of CJ3 on E"1 is independent of the state of sh·ess within the specimen; actu ally, how-
ever, Poisson's ratio may· be expected to vary with the mean stress, a. == ¾(a1 + 2o3 ),
or the shear stress (01 - 03), or both. It should be noted that the often-used ratio
of (cr1 - 03) to £ 1 is in effect the ratio of a shear stress to a normal strain; to term this
ratio a modulus is inconsistent with most mechanics terminology, and such usage should
be avoided. With the assumption of linear elasticity, the constitutive relation

(12)

may be combined with the condition that 02 == o3 and rewritten to give the following re-
lation for Young's modulus, E:
(13)
 21

where Er is given by Eq. 11. A plot of Eq. 13 is shown in Figure 4b, and the range of
variation is indeed quite substantial.
Provided an acceptable value for v can be determined, triaxial test results could be
incorporated in the finite-element model as follows. The triaxial modulus, Er, can
conveniently be determined from the slope of the cr1 versus £1 curve; then, with a knowl-
edge of the 0"3/0"1 ratio and an estimate of v, Young's modulus, E, can be determined
from Eq. 13. In the finite-element formulation of the buried conduit problem, an E
value consistent with the interpretation of the laboratory test can be selected once the
values of cr1 and cr3 have been determined for a given state of loading.
As an alternative approach that has been demonstrated often in the literature, load-
deformation or stress-strain data that exhibit the shape shown in Figure 8a can be very
conveniently described by a two-coefficient hyperbola, which for a (cr1 - 0"3) versus £1
plot takes the form

(14)

where a and b are empirical coefficients to be evaluated by experiment. The limit of

Eq. 14 as £1 approaches infinity yields

(a1 - 0'3 )u1 t = 1/b (15)

However, because values of (cr1 - cr3) approach (cr1 - 03)u1t asymptotically as £1 goes to
infinity, it may be expected that values of (cr1 - 0'3) at failure, or (cr1 - 0'3 )r, will normally
be less than (cr1 - 0'3)u1t, and this will impose an upper bound on the validity of Eq. 14.
In general, we have

(16)

where R is another empirical coefficient that usually lies within the range of 0. 70 to
0.95 and very often between 0.8 and 0.9. Differentiation of Eq. 14 will yield a tangent
modulus, Et, which may be expressed as

(17)

because 0"3 normally is held constant in a triaxial test, and the evaluation of Eq. 17
where £1 equals zero will give an initial tangent modulus, E1, which may be written

E1 = 1/ a (18)

In order to avoid potential contradictions in the reference state (zero value) for stress
and strain, we should eliminate the strain parameter in Eq. 17; this can readily be ac-
complished by solving Eq. 14 for £1 and by substituting the result into Eq. 17 to give

Et = [1 - b(cr1 - 0'3)] 2/a (19)

Equation 19 is a quite general relation that may readily be used as described previously
in conjunction with a mathematical model, and the following procedure is suggested for
evaluating the empirical coefficients. When plotted in the conventional form of (cr1 - cr2)
versus £1, a typical set of triaxial test data will usually exhibit the shape shown in
Figure 8a. If Eq. 14 is rewritten in terms of transformed variables as

(20)

the data can be replotted as shown in Figure 8b and described by a straight line. Hence,
not only do the coefficients a and b have real physical significance, but they are ex-
tremely easy to obtain. The value of R for any given test is obtained simply by multi-
plying the actual failure value (cr1 - a 3 )r determined in the test by the empirical coef-
ficient b, as shown in Eq. 16.
22

Figure 5. Typical ranges for Janbu's Figure 6. Constrained modulus versus stress
coefficients. level for various dry densities.
1.0 orolne
\

n 0 .5 'Son~J
\s111 ,

-
1/)

C:
Q)
0
10•
\ Clo
1

-
-~
Q)
0
0
10 11

10 4
l
--,:

\ \\
,\
::,
m 10 5 ~
.J:l \ ~e>:\
C:
10 2 %
0
-:, ,c;.<>'
10
.,q--
~o,'-!.Ii
0 50% 100%

Porosity

Figure 7. Piecewise linear stress-strain Figure 8. Hyperbolic representation of stress-strain response.

response.

..!...
b
- - -+ - - - -- --i

Poisson's Ratio

(al Normal Stress-Strain Plot

S1ress

(bl Transformed Hyperbolic

Plot
I
 23

In general, the coefficients a and b will not be constant for different soils, nor will
they be constant for the same soil under different test conditions (such as strain rate,
confining pressure, and water content). Therefore, this formulation can be generalized
by performing a series of tests to determine the functional relation between a and b and
the other parameters of interest for a given problem. The test range should cover
those conditions that are expected in the field situation. Of particular interest is the
dependency of a and b on the state of stress. There is some indication that the initial
tangent modulus, E1 = 1/a, can be related to the confining pressure, 0"3, by

(21)

which is a modified form of the relation proposed by Janbu (4), and (a1 - 0'3)u1t = 1/b can
be replaced by the Mohr-Coulomb failure criterion -

(0'1 - 0'3)uu = 1/b = 1/R (0'1 - 0'3)r = (1/R)[2c cos ¢ + (0'1 - 0'3) sin¢] (22)

where c and ¢ are empirical coefficients that are determined from a series of triaxial
tests. The latter modifications to Eq. 19 have been proposed by Duncan and Chang (2),
and the resulting expression has been applied with some success to a variety of son=-
structure interaction problems. In view of this success, its application to the buried
conduit problem certainly appears justified. Also, this approach allows direct con-
sideration of large strains and a failure condition in the soil (this is in contrast to the
approach based on the consolidation test), and most laboratories have the appropriate
equipment to conduct the test.

Determination of Modulus From Plane Strain Test

In a plane strain test, E:2 is held equal to zero, and 0'1, E:1, and 0'3 are either con-
trolled or measured; 0'2 is then calculated by assuming a linear elastic constitutive
relation for the soil. Young's modulus, E, can be determined from a plane strain test
by combining
(23)

and
(24)

with the condition of E:2 = 0 to yield

(25)

where

(26)

The strong dependence of E/EP on v and cr3/a1 is shown in Figure 4c. Equation 25 can
be incorporated into a finite-element model in the same manner already described, and
the app1·opriate value of the modulus for each element would be selected on the basis of
the ratio of a3/a1. Alternatively, it seems very logical that a hyperbolic formulation
similar to that described for the triaxial test could be advanced. Although many buried
conduit problems may be considered essentially plane strain in nature, the determina-
tion of a modulus by means of a plane strain test has the serious disadvantage that rela-
tively few laboratories are equipped at the present time with plane strain test equipment.

Comparison of Methods for Determination of Modulus

The soil in the vicinity of an underground conduit is often subjected to considerable
confinement, particularly in situations where the horizontal dimensions greatly exceed
24

the vertical height of soil above the conduit. In such cases, the major principal strain
at most points throughout the system will usually be much larger than the other principal
strains. For example, in a typical embankment installation, the major principal strain
in the fill at some distance from the conduit will be primarily vertical, the horizontal
strain in the longitudinal direction of the fill will be essentially zero, and the horizontal
strain in the lateral direction will probably be very small. Near the conduit, the major
principal strain will be predominantly radial, especially where considerable conduit
deformation is involved; there may be a slight tensile strain parallel to the centerline
of the conduit; and the strain tangent to the conduit wall in a vertical plane will probably
be compressive. Although the foregoing reasoning is qualitatively correct, it is dif-
ficult to assess intuitively the quantitative relations that are involved. Nevertheless,
the predominance of the major principal strain indicates that the response of a soil-
conduit system is probably influenced more strongly by dilatational stresses than de-
viatoric stresses. Therefore, the consolidation test (or uniaxial strain test) may very
well provide the most reliable immediate source of input data for soil properties be-
cause it is concerned only with volume change characteristics; however, if the conduit
deformations are relatively small, as is the case for a concrete pipe, shear deforma-
tions in the soil adjacent to the conduit may be important, and the plane strain or tri-
axial tests may be more appropriate because they involve considerable deviatoric ef-
fects. The consolidation test has the very significant advantage that the required
equipment is currently available in almost every soils laboratory, whereas triaxial
test equipment is less common and plane strain equipment very rare.

Determination of Poisson's Ratio

As can be seen in Figure 4, the determination of Young's modulus, E, by use of the
preceding three tests is strongly dependent on a knowledge of Poisson's ratio, v. Un-
fortunately, v is a very illusive soil property to obtain, and it has provided a source
of frustration for many researchers. As a matter of fact, there is considerable sup-
port for the position that one can make an engineering estimate that is as good as or better
than any value that can be experimentally determined, and this may indeed be the case.
This is largely because of the fact that E, v, and the state of stress are intimately
related, and it is difficult to determine one parameter without a knowledge of the others.
This leads to a situation where an error in E causes an error in v, and vice versa.
Although a survey of the literature can provide substantial guidance in the selection of
a particular value for v, there is little quantitative justification to be found. Accord-
ingly, a continual effort must be advanced to improve our understanding of this param-
eter, and engineering ingenuity must be employed to find better ways of either measur-
ing v or offsetting its effect in a mathematical model.
Continuing with an interpretation of test results in terms of linear elastic theory,
we may write the following equations for both a conventional test and a plane strain test:

(27a)

(27b)

Multiplication of Eqs. 27a and 27b by £3 and £1 respectively, subtraction of the results,
and rearrangement lead to

(28)

For a conventional triaxial test, C12 == cr3, and Eq. 28 reduces to

(29)

whereas for a plane strain test, a2 equals v (a1 + C13), as given by Eq. 24, and Eq. 28
becomes
2
Av + Bv + C == 0 (30)
 25

where
(31a)

(31b)

and

(31c)

Solution of Eq. 30 by using the quadratic formula yields

v = (-B ± ✓ B 2 - 4AC)/2A (32)

Hence, if E3 is measured in either of the preceding two tests in addition to the conven-
tionally measured a1, 0'3, and E1, v may be theoretically determined by Eq. 29 for a
triaxial t est and by Eq. 32 for a plane strain test. However, in view of the lim itations
imposed by the basic assumptions (such as linear elasticity and s tress and str ain homo-
geneity) employed and the probable error associated with tile measur ement of Es (t hi s
is related to the assumption of strain homogeneity), it cannot be expected that these
equations will yield acceptable results unless extreme care is exercised in the test pro-
cedures. Values of E3 at various stress levels can be obtained either by direct mea-
surement or, for saturated samples, by measurement of the volume change; the latter
t echnique necess al'ily yields an aver age value for E3, whereas the form er may ver y
well give a maximum (an erroneous) value if measured at the midheight of the specimen.
However, for unsaturated samples direct measurement is the only recourse.
Another possible approach for determining v is to utilize the empirical relation (!., ~)

Ko = 1 - sin Cl) (33)

in conjunction with Eq. 3 to yield

v = (1 - sin cp)/(2 - sin c,o) (34)

where Cl) in this case can be determined for various stress levels prior to failure as well
as at failure. Although this approach is rather indirect, Eq. 33 is based on consider-
able experimental evidence, and the error associated with its use should lie within ac-
ceptable limits.
Although not quite so well founded as the hyperbolic stress-strain formulation, there
is some evidence to indicate that a hyperbolic equation of the form

(35)

can be used to relate the axial strain, E1, and the radial strain, E3, in a conventional
triaxial test, where the empirical coefficients r and s are determined in a manner
similar to a and b in Eq. 14. Then, analogous to the definition of the tangent modulus,
Et, given by Eq. 17, one might define a tangent Poisson's ratio, Vt, by solving Eq. 35
for E3 and differentiating with respect to E1 to obtain

(36)

The strain dependence of Eq. 36 can be changed to stress dependence by solving Eq. 14
for E1 and substituting the result in Eq. 36 to give
2 2
Vt = r[l - b (a1 - 0'3 )] /[l - (b + as) (a1 - <13)] (37)

The problem now reduces to one of determining the empirical coefficients as functions
of the state of stress and the other variables of the soil. From a theoretical point of
26

view, it is noteworthy to point out that the state of stress in a conventional triaxial test
is not consistent with that used in the definition of Poisson's ratio; hence, the definition
given by Eq. 36 is not strictly correct. Nevertheless, for all practical purposes, and
in view of the complexities introduced by the use of Eq. 29, the empirical formulation
given by Eq. 37 is probably justified for use in conjunction with Eq. 19. However, such
a formulation cannot satisfactorily account for dilatancy effects in soils. If the experi-
mentally determined value for v is equal to or greater than one-half, a value such as
0.49 is usually incorporated in the theoretical analysis because values of one-half or
greater are incompatible with classical linear elastic theory and the associated finite-
element method.
As can readily be appreciated, Poisson's ratio is an important soil parameter, but
it is most difficult to quantify. Accordingly, it seems that the best chance for success
with this parameter in the near future is to develop and employ a method of analysis
that essentially offsets its effect. This is one of the apparent advantages of using the
plane strain test to determine soil parameters. Because a plane strain laboratory test
essentially models the field situation for many buried conduit installations, it is likely
that the effects of Poisson's ratio can be minimized by interpreting the laboratory re-
sponse in terms of plane strain conditions and by using the result to analyze the field
problem.

Backpacking Materials
Considerable attention must be given to the mechanical properties of backpacking or
cushioning materials that may be used immediately adjacent to all or a portion of the
conduit. If uncompacted or lightly compacted soils are used, the modulus values (and
probably Poisson's ratio) will be strongly affected by the dry density, and laboratory
tests (similar to those previously discussed) must be conducted at densities appropriate
to the field installation. In addition, care must be taken to ensure that the dimensions
of these low-modulus zones in the field are consistent with those used in the finite-
element model. If nonsoil backpacking materials are used, special effort must be made
to quantify their mechanical properties, and this is often extremely difficult to do. If
this task cannot be accomplished with reasonable confidence, a parameter study may
be used to assess the relative importance of the value selected. Finally, it is very
likely that the time effects on a backpacking material, especially a nonsoil material,
cannot be neglected, but the manner in which they should be included is not at all clear.

INTERFACE CONDITIONS
Within the realm of material properties should be included the conditions that exist
along the soil-conduit interface and along the boundaries between different zones of soil.
In general, there are three situations that can be handled without too much analytic dif-
ficulty: full slip, no slip, and no slip until a prescribed stress has been reached. In
the absence of any information to the contrary, it seems most appropriate to utilize the
third condition. The upper bound of the prescribed stress would certainly be the shear
strength of the soil, and the extent to which the actual stress at the interface differs
from this upper bound would depend on the smoothness of the conduit. In general, it
seems quite reasonable to use the shear strength of the weaker soil at the interface be-
tween different soil zones. For cases where the conduit is very smooth, as with some
metal and plastic pipes, the full-slip condition may be more applicabl.e. However,
instead of attempting to characterize this condition exactly, one may examine the ex-
treme cases of full slip and no slip (5) for one particular type of problem and thereby
evaluate its influence. -

PROPERTIES OF CONDUIT MATERIALS

Although the mechanical properties of the surrounding soil probably exert the great-
est influence on the response of a soil-conduit system, the properties of the conduit
materials are also extremely important. In most, but not all, cases, however, the
properties of the conduit materials exhibit more limited ranges of variation and can be
 27

characterized with greater reliability than soils. A few of the key considerations regard-
ing conduit materials will be discussed in the following sections.

Metal Conduits
In general, steel and aluminum possess relatively well-known mechanical properties,
and their determination is not difficult; however, metal pipes have some characteristics
that are not readily and reliably evaluated. Among these are their susceptibility to
corrosion and abrasion, the effectiveness of protective coatings, the behavior of seams,
and the buckling strength. Various empirical equations and statistical correlations are
available to handle the durability problem, which indirectly affects the soil-structure
interaction problem by altering the pipe wall with time. The stiffness of a seam may
significantly influence the response of a conduit, particularly if the seam is less stiff
than the rest of the conduit wall or if there is an abrupt change in the cross section of
the wall at the seam. Directly related to the behavior of a seam are the characteristics
(not only strength, but durability, deformability, and brittleness) of the bolts that are
used. For large-diameter conduits, particularly those with relatively shallow cover
heights, stability against buckling must be checked. This leads to a relation that in-
volves relative values for the mechanical properties of both the conduit and the sur-
rounding soil as well as the cover height.

Concrete Conduits
The most important properties that influence the structural response of reinforced
concrete conduits are the compressive and tensile moduli of the concrete, the tensile
strength of the concrete, the amount and position of the reinforcing steel, and the quality
of the bond between the concrete and the steel. The role of the compressive and tensile
moduli are self-evident, whereas the tensile strength is important because it is directly
related to progressive cracking, which in turn alters the internal stresses and displace-
ments within the pipe wall. As a consequence of progressive cracking, the stiffness of
the conduit is continually modified as the load is increased, thereby introducing a de-
cidedly nonlinear response. The amount and position of the reinforcing steel of course
controls the stiffness of the conduit, and the associated analyses must be carefully re-
viewed in light of the final design. Although the amount of steel is relatively easy to
determine, accurate placement is largely dependent on fabrication procedures. The
quality of the bond between the concrete and the steel is important if the deformation
characteristics of the conduit are based on effective bond considerations. Bond failures
cause small local slips and some widening of the cracks with an associated increase in
the deformation of the conduit.

Plastic Conduits
The mechanical properties of many plastics used in the manufacture of conduits are
not sufficiently well understood, and much additional effort is needed before these prop-
erties can be reliably specified for use in a theoretical analysis. In particular, some
plastic conduits exhibit significant time-dependent effects, which influence their long-
term response. Thus far, however, the use of plastic conduits has been generally re-
stricted to the smaller diameter installations wherein mechanical properties play a
considerably lesser role in the overall behavior of the system.

CONCLUSIONS
Within the scope of the considerations discussed herein, the following conclusions
or summary statements can be advanced.
1. The more significant improvements in our ability to analyze soil-conduit inter-
action problems in the immediate future lies in the use of a continuum approach and a
finite-element formulation; however, the successful application of these procedures
will depend largely on our ability to provide appropriate input data for material prop-
erties.
28

2. The specification of reliable, quantitative values for material properties is

probably the single factor that limits most seriously our ability to predict the mechani-
cal behavior of soil-conduit systems.
3. The particular material properties to be used in the mathematical model, the
manner in which these properties are determined, and the methods of analyses and de-
sign are intimately related, and a variation in one step will generally precipitate
changes in one or more of the other steps.
4. Field verification is absolutely essential before a mathematical model and its
associated material properties can be accepted as a valid procedure for analysis.
5. Of all the material properties that affect the interaction of a soil-conduit system,
the characteristics of the soil present the greatest difficulty; and, of all the soil prop-
erties of interest, the stress-strain behavior is the most important.
6. Within the framework of currently available methods for testing and analysis,
the assumption of a stress-dependent, isotropic soil modulus is probably reasonable.
7. Both the stress dependency of the modulus and the incremental nature of the
construction sequence in a conduit installation can be conveniently analyzed by use of a
piecewise linear formulation.
8. Subject to proper interpretation, which involves consideration of Poisson's ratio
and the state of stress, an appropriate soil modulus can be determined from the results
of laboratory consolidation tests, conventional triaxial tests, or plane strain tests.
9. Time effects, which are not explicitly considered herein, must be taken into
account by conducting the laboratory test at an appropriate strain rate and with appro-
priate drainage conditions.
10. Dry density significantly affects the soil modulus, and moduli associated with
different densities must be determined in the laboratory and assigned to appropriate
zones in the field problem.
11. Although Poisson's ratio significantly influences the determination of a soil
modulus from any of the suggested tests, this parameter is most difficult to quantify in
a reliable manner; the use of linear elastic theory and experimental measurements to
back calculate Poisson's ratio leads to rather unreliable results in many cases, and
this situation suggests that the exercise of good engineering judgment to estimate Pois-
son's ratio may be equally useful.
12. The greatest hope for success in dealing with Poisson's ratio in the near future
centers around the development of a method of analysis that essentially offsets its effect;
this may be possible to some extent in all cases because Poisson's ratio is used both in
the interpretation of a laboratory test and in the analysis of the field problem.
13. Insofar as possible, it is desirable to describe the stress-strain behavior of the
soil in terms of a three-dimensional formulation, even if only a simplified one- or two-
dimensional version is utilized at present because three-dimensional analytic treat-
ments of soil-conduit problems are not too far in the future.
14. The interface conditions between the soil and the conduit and between various
soil zones should be specified by a no-slip condition with a limiting shear stress above
which slip occurs.
15. The mechanical properties of the conduit materials may exert a significant in-
fluence on the behavior of the physical system; particular examples include the prop-
erties of the bolts or welds in a metal conduit, the tensile strength of the concrete in a
concrete conduit, and the substantial time-dependent response of some plastic conduits.

REFERENCES
1. Brooker, E. W., and Ireland, H. 0. Earth Pressures at Rest Related to Stress
History. Canadian Geotechnical Jour., Vol. 2, No. 1, 1965, pp. 1-15.
2. Duncan, J. M., and Chang, C. Y. Nonlinear Analysis of Stress and Strain in Soils.
Jour. Soil Mech. and Found. Div., ASCE, Vol. 96, No. SM5, 1970, pp. 1629-1653.
3. Jaky, J. Pressure in Silos. Proc. 2nd Internat. Conf. on Soil Mechanics and Founda-
tion Eng., Rotterdam, Vol. 1, 1948, pp. 103-108.
 29

4. Janbu, N. Soil Compressibility as Determined by Oedometer and Triaxial Tests.

Proc. European Conf. on Soil Mechanics and Foundation Eng., Wiesbaden, Vol. 1,
1963, pp. 19-25.
5. Krizek, R. J., Parmelee, R. A., Kay, J. N., and Elnaggar, H. A. Structural
Analysis and Design of Pipe Culverts. NCHRP Rept. 116, 1971, 155 pp.
6. Osterberg, J. 0. Discussion of paper by R. R. Phillips. Proc. of the Conf. on
Soil Stabilization, M.I. T., Cambridge, 1952, pp. 167-168.
 '),t)
j

EXPERIMENTAL STUDIES IN SOIL-STRUCTURE INTERACTION

F. Dwayne Nielson, Colorado State University

This paper contains a literature review of certain experimental studies

that pertain to soil-structure interaction. It is divided into two parts:
model studies and field or full-scaletesting. The requirements that govern
model studies are reviewed briefly, and examples of applications are pre-
sented. A brief presentation is made of several studies that have used
model analysis. They cover topics such as effects of soil moisture and
density on culvert deflection, effects of differential soil compaction on cul-
vert stresses, imperfect ditch method of construction, stresses on multiple
pipe installations, pressure distribution on pipe, and soil properties. The
field study portion presents field studies of the imperfect ditch method of
construction, full-scale failure tests, and certain Canadian large pipe tests.
A circular culvert design method that talces into account most of the sig-
nificant variables that affect culvert performance is also presented.

•IN reviewing the literature on soil-structure interaction, I found that many researchers
had solved buried structure problems by using a field or model study. In spite of all
the work that has been done, the field of culvert design is still relatively new. There
is a great deal that is not known about the performance of culverts in various situa-
tions. Some design tools that have been available for many years have been applied
on an experiential basis only. Some of these will be discussed later in this paper.
Only those studies that have been done in the past few years will be presented here.
Many topics that are of considerable importance are omitted because of space limitations.
For purposes of discussion, experimental studies are divided into two groups: model
studies and field or full-scale tests.

MODEL STUDIES
Model tests have been successfully used to investigate the performance of full-scale
culverts. Other soil mechanics problems that can be treated as plane strain problems,
such as slope stability, have also been successfully modeled.
Models may be used to solve complex problems that cannot be readily evaluated by
using analytic means. The use of models in obtaining solutions related to performance
of underground structures is most attractive because of the complexity of the problems
involved. Time and money often prohibit the use of full-scale tests to investigate the
effect of different variabies on a system. Because uf the ease with which models can
be fabricated and tested, they yield much more information for a given amount of time
and money than do full-scale tests. Full-scale tests are useful in verifying results
obtained from model tests.
In order to establish reliable similitude requirements for a given model system, we
must define all variables that influence the phenomena. Defining the variables involved
requires that an investigator have considerable experience within the area of investiga-
tion. As experience is gained in an area, model analysis can be applied with confi-
dence to the solution obtained.

SIMILITUDE REQUIREMENTS
Before a brief review of similitude requirements is presented, it might be helpful
to give an example of how similitude and model analysis works.

Sponsored by Committee on Subsurface Soil-Structure Interaction.

30
 31

Most people interested in culvert design are familiar with the Iowa formula; there-
fore, it will be used to illustrate the modeling concept as follows:

(1)

where AX is the change in horizontal diameter; K is a bedding constant; W is the load

0

acting on the culvert and is equal to W = pD; P is the average vertical pressure acting
0

on conduit; E' is the modulus of soil reaction; r is the radius of the culvert = D/2; E
is the modulus of elasticity of the culvert wall; and I is the moment of inertia of the
pipe wall per unit length.
A discussion on the use of the Iowa formula and modulus of soil reaction is included
later in this report.
If pD is substituted for W and both sides of Eq. 1 are divided by D, we derive
0

AX:/D =KP r 3 /(EI + 0.061 E'r 3 ) (2)

If the numerator and denominator on the right side of the equal sign in Eq. 2 are
divided by EI,

AX/D K(Pr 3 /EI)/[1 + (0.061 x E 'r3 /EI)] (3)

or

7T1 = Kir:/(1+ 0.061 ira) (4)

Consider the two following pipes:

1. No. 1 (model): 4-in. diameter; modulus of elasticity, 10 x 106 psi (aluminum);
wall thickness (smooth wall), 0.050 in.; moment of inertia= t3/12 = 1.04 16 x 10- 5 in.'1/in.;
D3/EI = 0. 61443.
2. No. 2 (prototype) : 10-ft diameter; modulus of elasticity, 30 x 106 psi (steel); 8
gauge, 6- x 2-in. corrugation; moment of inertia= 1.15 in:1/ft = 0.0958 in.1/in.; D3/EI =
0.60106.
Assume that both culverts are embedded in a soil with a modulus of soil reaction of
5, 000 psi and that they are s ubjected to an average vertical pressure of 100 psi. Then,
in the case of No. 1, pD 3/ EI = 61.443 and E 'D3 / EI = 3,072 .15. In the case of No. 2,
pD3/ EI = 60.106 and E ' D3 / EI = 3,005.30.
By substituting the value for each pipe into Eq. 3 and by letting K = 0.083, we get the
following:
No. 1
AX/D = 0.083 (61.443)/[l + (0.061 x 3,072.15)]
AX/D = 0.02706 = 2. 706 percent

No. 2
AX/D = 0.083 (60.106)/[1 + (0.061 x 3,005.30)]
AX/D = 0.02706 = 2.706 percent

Thus Spangler's deflection equation predicts exactly the same percentage of deflection
for both the 4-in. pipe and the 10-in. pipe. This is the principle on which model anal-
ysis is based. If the individual pi-terms are the same, the results, regardless of size,
will also be the same.
The properties of the soil and pipe that govern soil-structure interaction under static
load have been defined, so models can be used to predict soil-structure interaction per-
formance.
32

In most soil-structure interaction problems, the form of the equation is not known.
If it were known, there would be little need to use model analysis. Therefore, one must
select the primary independent variables that influence the phenomenon and place them
in dimensionless pi-terms.

SIMILITUDE REQUIREMENTS FOR DEFLECTIONS UNDER HIGH FILLS

To illustrate the use of model analysis for a specific soil-structure interaction
problem, we will use an example for determining culvert deflection under a fill that
is high in comparison with the diameter of the culvert. The culvert is assumed to be
long enough such that end effects can be neglected and that maximum stress in the cul-
vert wall is below the yield point stress. The primary independent variables and di-
mensions involved are as follows:

Primary Independent Variable Dimension

Culvert diameter, D
Pipe wall stiffness, EI
Constrained modulus of elasticity of the soil or
modulus of soil reaction, M.
Pressure applied at the level of the pipe because
of the fill or other load above the pipe, P
Deflection of some point in the culvert, ~
(horizontal diameter used) L

It can be shown that these are all the variables that have a significant influence on
the amount of deflection that a culvert will experience. It may be argued that such
things as soil density, water content, and other soil properties must be included. The
influence of these variables is included in the constrained modulus of elasticity of the
soil. The constrained modulus is a function of soil density, water content, plasticity,
soil type, grain-size distribution, and all other soil variables plus boundary conditions.
According to the Buckingham pi-theorem, there must be three pi-terms (five pri-
mary independent variables minus two dimensions). These pi-terms may be formu-
lated as follows:

1T1 = AX/D

1Ts = M.D3/EI
or

If model analysis is to be used to predict the deflection of a full-scale pipe, all the
pi-terms must be the same for the model as for the prototype structure, as previously
illustrated in the example, or

where m refers to the model and prefers to the prototype or actual structure.
The model must be designed so that each pi-term for the model is the same as for
the prototype structure. Sometimes deviations are necessary, but then distorted model-
ing techniques are encountered that complicate the analysis and will not be considered
here. A more complete treatment of soil modeling is given elsewhere (19).
 33

Distorted modeling methods are discussed in various books on engineering simili-

tude. The modeling of soils by using distorted models may lead to unexpected prob-
lems because of the nonlinear nature of the soil culvert system.

SIMILITUDE REQUIREMENTS FOR WALL BUCKLING

Similitude requirements for wall buckling are not nearly as easy to satisfy as those
for deflection and pressure. The moment of inertia, area of the pipe wall, and the
yield point stress of the culvert material must be included. To get an exact model, not
a distorted one, is very difficult because the corrugations of a metal culvert preclude
the use of plane wall pipe as models. In deflection measurements, the area of the pipe
wall has been shown to make very little difference if moment of inertia is constant.
This can be seen by the elasticity solution of the problem and from model tests. How-
ever, in buckling problems, the area of the pipe wall or the radius of gyration of the
pipe wall must be included. Because of the difficulty in obtaining an exact model for a
corrugated metal culvert, model studies of wall buckling have not been very rewarding.

MODEL ANALYSIS OF DYNAMIC LOADING

Model analysis of dynamic loading becomes even more difficult in that the inertial
properties of the soil-culvert system must also be modeled. When this is attempted,
the density of the soil as well as the elastic properties must be modeled. A few in-
vestigators have modeled specific situations, but most of these studies have been as-
sociated with blast-shelter construction and will not be presented here.

EFFECTS OF SOIL DENSITY AND MOISTURE ON CULVERT DEFLECTION

As examples of what can be done with models, a few results will be presented. Fig-
ure 1 shows the results of a study to determine the effects of initial density of clay soil
on the deflection that a culvert undergoes (1). As can be seen, the soil density has a
considerable effect on the deflection. Figure 2 shows the effects that initial density of
sand has on the deflection of the culvert. A decrease in density shows a marked in-
crease in deflection for both sands and clays. Figure 3 shows the influence initial
moisture content has on the deflection of the culvert in clay soils. As the moisture
content increases, the deflection increases very rapidly. At a load of 50 psi, for ex-
ample, the deflection at 20 percent moisture is approximately 3 5 times the deflection
at 10.3 percent moisture. This is why density alone (even including grain-size distri-
bution and Atterberg limits) is a poor indicator of soil stiffness in cohesive soils.

EFFECTS OF BACKFILL DENSITY ON STRESSES IN PIPE

In 1964, Watkins (2) presented the idea of differential soil density to release stresses
in the pipe and pressures on the pipe. The soil around the pipe is compacted to a high
degree of density, and the soil next to the pipe is left in a rather loose state to form a
cushion around the pipe. Data were not presented at that time. Figure 4 shows the
method proposed by Watkins for achieving the differential density.
Figure 5 shows the results obtained from two different studies (3) that used models.
Both studies were made on the same 4-in. diameter 6061-T6 aluminum model. Both
studies used the same soil (Ottawa sand) compacted to the same density in the same
simulator. The only difference was in the manner in which the models were placed.
The soil in stankowski' s study was placed before the model, and then the model was
"jacked" into position. Insertion of the model appeared to compact the soil adjacent
to the model to a slightly higher density; however, no measurements of density were
made after model placement. The soil in Mohammed's study was "rained" around the
model pipe. As soil particles fell, those hitting the very edge of the model were deflected
away from it. The density of soil adjacent to the pipe was believed to be lower than the
soil mass. Therefore, density of the soil immediately adjacent to the model in Stan-
kowski' s study is assumed to be higher than in Mohammed's study. As can be seen,
there is a significant difference in stress because of the method of soil and culvert
placement. The difference in soil density and pipe stress shown points out that a very
Figure 1. Variation in horizontal deflection in clay. Figure 2. Variation in horizontal deflection in sand.

320
200

280
175

't;
p.
"'up. ~

240 m
150 m
m

,._o
,._o
·~ u
0.
~125 ·;;:: 200
p.
b

~
,._0
"
.
M

.
u
.3160

~
·•
~
> 75
..
'"
~120

50 Variables held constant 80

Soil: Bonneville Clay Variables held constant
EI 23.3 in lb. Soi 1: Wendover Sand
D 2,125 in. EI 23.3 in-lb
w • 15% D 2. 12S in
w - air dry
25 40

0 .1 0. 2 0. 3 0. 4 0. 5 0. 04 0. 06 0 .08 0 .10
Horizontal Deflection (Ax) in. llorizontal Deflection ( 6. x) in.

Figure 3. Effect of moisture on horizontal deflection. Figure 4. Watkins' method for reducing pressure on buried
circular conduits.

200

175

150

·;;:: 125
p.

¢1 1 Concept of Compacted Soil Arch with Cushion to Relieve

".., 100
~ Stress in Buried Flexible Pipe (Note Loose Soil Cushion)

~
.
·•
>~ 75

SD

Held Constant
Bonneville Clay
25 23.3 in lb.
2 . 125 in .
yO • 89 pc£
b, Method of Compaction by Which Conduit Shape is Maintained
and Ring Compression Load is Reduced by Loose Soil Cushion

D. I 0.1 0. 3 0. 4 0.5
Horizontal Deflection (4x) in ,
 35

easy and economical method of "backpacking" may be achieved by making use of dif-
ferential soil density; however, much more study is required.

IMPERFECT DITCH METHOD OF CONSTRUCTION

The imperfect ditch method of construction was one of the earliest methods used to
reduce the loads exerted on an underground conduit. Figure 6 shows the usual method
of installing the conduit by this method. The culverts that have been installed by using
this method have used a compressible layer that is the same width as the culvert and
placed almost directly above the culvert. Figure 7 shows the results of a study (3)
made to determine the effect of the width and height of the compressible layer above
the culvert on the stresses in the pipe wall.
The upper limit stress is the average of the absolute value of the stress plus two
standard deviations. In this model study, the upper limit pipe wall stress in almost
all cases was higher for an installation with a compressible layer than it was for no
compressible layer at all. The reason for this is that the pipe tended to deflect upward
into the compressible layer instead of flattening into an approximate elliptical shape.
Figure 8 shows a summary of the deflections obtained.
The model was a 6061-T6 circular aluminum tube, 4 in. in diameter. The compress-
ible layer was a common household sponge, with a thickness of 3/8 in. and widths of 4, 5,
6, and 8 in.
The data presented in Figure 7 are not intended to discredit the validity of the im-
perfect ditch method of construction. The figure does, however, serve as a warning
that, if the imperfect ditch is not properly designed, stress and deflection conditions
can be more severe in a pipe installed by this method than in one installed by normal
procedures. The compressibility of the sponge used in the model study was much
higher than most material used in actual field construction. The higher compress-
ibility of the imperfect ditch in the model would induce higher stresses because of the
vertical elongation that may not be experienced in most field structures.
Several other studies, in which different compressibilities of the compressible layer
have been used, have shown that the imperfect ditch is effective in reducing loads on the
structure.

MULTIPLE PIPE INSTALLATIONS

Model analysis can also be applied in other ways. Figure 9 (4) shows a photoelastic
model that was used to determine the stress concentration caused by multiple culverts.
Similitude requirements can be easily established for such studies. Figure 10 shows
a plot of the radial pressure-applied pressure versus spacing between the model cul-
verts. The numbers in parentheses are the values obtained for single culverts. This
type of soil-structure interaction model is extremely sensitive to "fit" of the culvert
model in the plastic. Figure 11 shows a plot of shear stress-applied pressure versus
model spacing. Solutions such as those shown in Figures 10 and 11 are plane stress
solutions, whereas most model solutions are probably closer to a plane strain solution.
The difference is usually insignificant when compared with the accuracy with which such
variables as soil properties can be determined.
Figure 12 shows a comparison (1) between stresses and deflections obtained from a
model analysis. The values were obtained from the elasticity solution of Burns and
Richard. Results shown in Figure 13 were obtained by the technique used to get the
data shown in Figure 12. It shows the results of a study to determine the pressure con-
centration (radial pressure and applied pressure) caused by multiple pipes.
The pressure acting on the outside of the pipe, the stress in the pipe wall, and the
deflection of the pipe were obtained by fitting a Fourier series to a general loading dis-
tribution as shown in Figure 14. The Fourier series coefficients were determined from
measured strains and displacements on the inside of the pipe. No pressure transducers
were placed on the outside of the model to disturb the pressure distribution.
Currently, most culverts are designed to carry considerably more load than is ac-
tually imposed on them. The designer should not worry about the strength of multiple
installation culverts. Even with the increased stress induced by multiple installations,
Figure 5. Effect of placement method on circumferential Figure 6. Imperfect ditch construction.
stress.

Top of hll ~
k
'/ (\\l\\w,!l/1/V,lllh!,W.\(\\\1\1\11\
0/
I
.
'" I
lo Excavate and Refill
wl th Co • prt!:S!i ible
"'6
0
0 I
-...-----r
f Mntcflnl
r---1
0

.
~
rz---:~~a~::ss Compact
Fill to This
I
I
I
I
I
I

I
0
___ .JI IL __ _ _ .JI

z
Level

~s
~

.,= 0
Stonkowsk
Maximum
,%/ Natural
Ground

C Stress {,Jii)l;i\\\\lili\ flllll)l,JhJJ} /1//lllllhM«,Q,li.\\«\\\\\\Wlli

...E4
// , k _ sunko~s~l
§
,g/ Average Stress
u
-~
u

.. ,._3 0/
/ ,/ Fi11ure 7. Variation of circumferential stress with
compressible layer parameters.
"'C I,/ ,,,,,, 200.----,-----.-----.-----.----....
1 ~
L V ~ Mohammed
,,;~ Average
;2 j .,., Stress

" ,/2'I ,,.« ,/ w

......
0
1so 1-----+-.,.,.---j----+--- - - + - - ---1
I J1 I ,,,"' Ill~

1/"' )':" µO.

I ,, A V,.
{,,,,," ~]
~~ 1001-- - -,tj,=:a'-=:::::'-'=-,-:j,---;;.,ir....==~-=---,I,
f,6 o,
""'
'" µ
20 40 60 BO 100 :g :; Layer
~...,
Surface Pressure - psi
'"
A
.
~ 5Dl-----+-- - - - j - - - - + - - ---+-- ---I
,g,

Figure 8. Effect of location of imperfect ditch on o._____._____.____..____...______,

deflection of horizontal and vertical diameters 1.0 1.25 1.50 1. 7 5 2. 0
(surface pressure= 60 psi). Width-Diameter Ratio W/D

Figure 9. Photoelastic model.

...
.,C
u

."
t,

0 0
1 6"

l
C
•
,:
u

..•
E
0
X I

0. 5 1.0 1.5 2, 0
Width Diameter Ratio W/D
Figure 10. Experimental radial interface pressure Figure 11 . Experimental shear stress concentration
concentration factors. factors.
1.9 :,,o

I. 8 i. 8

1. 7
•
1.5 2. 4

".
~

h
l.S

...,.
~
] •4 2, 0
h

I.J
-~
0.
0. l.2
• l.6
0

'.
<
h
LI
~
..
h 1.0 J. 2

.
~

-~.., 0.9
(0·6-1}
"' 0. 8 o. 8
0. 7

o. 0. 4
o. 25 0. 5 0. 75 l.0 0 . 25 0. 5 0. 75 1. 00
Q/D
Q/D

Figure 12. Comparison of predicted and Figure 13. Experimental radial interface pressure
recorded horizontal diameter changes for the concentration factors.
single cylinder system.

1.8
- - - - - Multiple Cylinder Systems
Single Cylinder Systems
Bonded Solution

Unbonded Solution 1. 6
Average of Bonded and
Unhanded Solutions
+ Experimental
-~~ 40 I. 4
.:'!:
.
h
~

...
~

h 30 1.2

·E
.......
u
20 1.0
(1)
w
(2)
•••• (1)
(3)
0. 8
10 OS•• (2)

••• • (3)

0 + .............................._..-I
0.6 ..................................
0 . 02 . 04 , 06 . 08 0 1.0 2. 0
Horizontal Diameter Change (in.) Q
2R
38

most culverts designed by current methods will carry the load with an adequate factor
of safety . However, as research and development continue and more refined and ac-
curate solutions of the loads on pipe become available, the increased stresses due to
multiple installations will have to be considered.

PRESSURE DISTRIBUTION ON PIPE

In 1941, Spangler (5) presented the Iowa equation for determining the deflection of a
flexible pipe. The pressure distribution used by Spangler was arrived at by measuring
the radial pressure on the pipe with friction ribbons. The steel ribbons were placed
on the pipe before the fill was placed. After completion of the fill, the ribbons were
pulled out from under the fill, and the pressure was assumed to be related to the coef-
ficient of friction and the force necessary to pull the ribbon out from under the fill.
The ribbons were calibrated in the laboratory before they were installed on the struc-
ture; however, before measurements are made, many things can change the calibration.
It is highly probable that the friction ribbons measured only the radial component of
pressure acting on the pipe. The shear component was neglected. This is a common
mistake with most field measurements that use some type of pressure transducer on
the outside of the pipe. Only the radial pressures are measured.
The pressure distribution used by Spangler is shown in Figure 15. This pressure
distribution is adequate for low-density soils that have a high value of Poisson's ratio;
however, for high-density granular soils with a low value of Poisson's ratio, the pres-
sure distributions appear to change somewhat. These conclusions are based on a
theory presented by Burns and Richard, which involves elastic media. Figures 16 and
17 show the pressure distribution obtained for two different conditions. Figure 16
shows the pressure distribution for a low-density soil (low value of constrained mod-
ulus) and a high value of Poisson's ratio. This condition would represent the low-
density clay that Spangler did most of his work on. For this case, the pressure dis-
tribution obtained from the elastic theory is approximately the same as the pressure
distribution measured by Spangler. Figure 17 shows the pressure distribution obtained
for a high-density soil (high value of constrained modulus) and a low value of Poisson's
ratio. This condition represents a well-compacted granular soil.
To determine whether the pressure distribution acting on the pipe changes for
granular soils, we compared data obtained by Stankowski (6) with the data shown in
Figure 17. Figure 18 shows the results obtained. They snow that the pressure dis-
tribution does change from that used by Spangler. The major change is in the hori-
zontal pressure distribution. Spangler' s pressure distribution has the horizontal pres-
sure acting only over a 100-deg section of the culvert. The measured pressures shown
in Figure 18 act over essentially the entire 180 deg. The measured pressures, there-
fore, present more resistance to horizontal movement of the pipe than that used in
Spangler' s pressure distribution. The measured vertical pressure also has a dip in
the pressure distribution at the center and at the edge of the pipe. This would reduce
the magnitude of the vertical load that is exerted on the pipe. This dip in pressure
seems to become more pronounced as the constrained modulus of elasticity of the soil
gets higher.
If the modulus of the soil reaction E' is determined by using the calculated values
from the theory of elasticity for pressure and deflection at the horizontal diameter,
the difference in pressure distribution makes the Iowa formula (Eq. 1) predict too much
deflection for soils with a high value of constrained modulus and low value of Poisson's
ratio. For example, the Iowa formula predicts approximately 40 percent more deflec-
tion than Burns and Richard's (full slippage) elastic theory for a soil with a high con-
strained modulus (10,000 psi) and low Poisson (0.1).
On the basis of the information presented here, it cannot be said that the Iowa for-
mula should be modified for dense granular soils. It does suggest, however, that more
study is needed in this area. Field studies that are made should be instrumented so
that shear stresses, as well as radial pressures acting on the pipe, can be measured
in order for the necessary verification to be obtained.
Figure 14. General loading components. Figure 15. Assumed distribution of pressure on
flexible culvert pipe.
V • !!E_
2r

v' • _W_c_ • _v_

2r SinA SinA

Figure 16. Pressure distribution for a low-density Figure 17. Pressure distribution for a high-density
soil. soil.

0
PY "' 148. psi

PY • 114

R ,. 60. in
Ms = 7S.
u • • 450
MU-HIGH, MS- LOW
P • 100. psi
NO SLIPPAGE

PX. M. PX• 146. psi

40

SOIL PROPERTIES DETERMINATION FROM MODEL STUDIES

The author, in another report (7), established the relation between the modulus of
soil reaction and the constrained modulus of elasticity. In establishing this approxi-
mate relation, he used the bonded shell equation of Burns and Richard's solution (8)
of a pipe embedded in an elastic medium. He concluded that the modulus of soil reac-
tion can be approximated by

E'= 1.5M. (5)

If the unbonded shell solution of Burns and Richard had been used, it would have been
shown that

E'=0.7M. (6)

It has been shown by Stankowski (6) and Nielson and Statish (9), who used model stud-
ies, that the actual modulus of soil reaction of soil is between the predicted modulus of
soil reaction by the bonded shell and the unbonded shell solution. Figure 19 shows the
results of a confined compression test used to determine the constrained modulus of
elasticity. Figure 20 shows how the modulus of soil reaction varies with pressure for
this same soil at the same initial density. If an average value of the modulus of soil
reaction is used, it can be shown that

E'= 0.8M. (7)

Because the modulus of soil reaction is directly associated with Spangler's pressure
distribution (Fig. 15), any question concerning the validity of the pressure distribution
on the pipe will also be directly applicable to the modulus of soil reaction. If the pres-
sure distribution acting on the pipe is different from that used by Spangler, modifica-
tion of the modulus of soil reaction may also be needed.

SPECIAL PROBLEMS
Model analysis has been applied to many soil-structure interaction problems. Wat-
kins has applied model studies to determine the minimum necessary height of cover
over a conduit for the safe crossing of construction equipment (10), determination of
pressures on culverts under stockpiles (2), and determination ofthe movement of soil
around a pipe (11). Linger (12) has applied model studies to determine how the flexi-
bility of aflat-top buried structure affects the redistribution of pressure on the structure.

FIELD OR FULL-SCALE TESTS

Many investigators have made attempts at field studies to determine or verify cer-
tain phenomena. Some of these studies have been associated with determining the
pressure on the culvert and the corresponding displacement. One variable almost in-
variably neglected is the shear stress acting on the culvert. Measurements of shear
stress are difficult to obtain on an actual field installation, whereas radial pressures
acting on the culvert are relatively easy to obtain. A significant part of the load is
neglected if only the radial pressures are measured.
Other studies have been made to determine the maximum load that a culvert can
carry and to determine how large a culvert can be installed without failure.
One of the earliest applications of the imperfect ditch method of construction re-
ported in the literature (13) involved a 48-in. concrete sewer pipe. Twenty years after
the construction of the sewer, the height of an embankment above a ¼-mile section was
increased from 60 to 78 ft without making any changes in the pipe.
Davis and Bacher (15) reported on California's culvert research program in which
field studies and observations were made on several culverts. Different backfill de-
signs were employed including an imperfect ditch construction and a variation there-
from in which a compressible layer of baled straw surrounded the culvert.
 41

In one case in which a layer of baled straw surmounted a rigid culvert, horizontal
and vertical pressures were found to be much less than in another in which no com-
pressible layer was used. When baled straw was placed above a flexible conduit, the
profile of the average induced pressure per foot of fill showed super-hydrostatic pres-
sure bulbs at the invert. Effective densities at the crown, sides, and midpoints of the
lower quadrants were about one-half that of the embankment. With increasing fill
heights, maximum effective densities were observed to decrease and minimal densities
to increase "so that some tendency toward a more uniform distribution is indicated."
The case where a layer of baled straw surrounded the conduit "showed the most prom-
ise, inasmuch as the lateral pressures were almost negligible and vertical (effective)
densities were about half that of the embankment."
An 18. 5-ft diameter structural plate culvert under 83 ft of cover was reconstructed
using the imperfect ditch method (16). The rebuilt culvert and the fill were instru-
mented such that pressure and deformation measurements could be taken. The be-
havior of the rebuilt culvert was observed to be significantly different from that pre-
dicted by the Marston-Spangler theory. Some of the most important conclusions of
this study were as follows:
1. The predicted average vertical pressure, by Marston' s theory, at the straw level
was almost double the average measured pressure.
2. The shear-plane method may be used between the straw level and the conduit
level to predict the average vertical pressure at the top of the conduit.
3. The maximum average vertical pressure on the conduit was approximately one-
half of the overburden pressure. This is comparable to the effective densities re-
ported by Davis and Bacher (15).
4. Lateral pressure on theside of the conduit was found to be substantially larger
than the average vertical pressure.
5. A reduction in both the horizontal and vertical diameters was observed; however,
at certain locations it appeared that the culvert may have deflected upward into the
compressible layer.

FULL-SCALE TESTS FOR DETERMINATION OF FAILURE

Watkins and Moser (17) conducted tests on full-size pipe to determine the failure
mechanism of pipes. Figure 21 shows a cross section of the test cell used. Spangler
stated that the test cell does not represent field conditions and that the data obtained
are a function of the test cell as well as the soil and pipe properties. Figure 22 shows
some of the data obtained by Watkins and Moser, which are plotted in dimensionless
form. If the three points at very low deflections and a PD 3 /EI of approximately 400
are neglected, most of the data appear to plot approximately as a smooth curve.
The effect that the test cell would have would be to increase the stress in the ui
direction shown in Figure 21. The increase in 0 1 should have the effect of increasing
the elastic properties (modulus of elasticity) of the soil. There seems to be a com-
pensating effect in that the length of the test section of the cell was quite short in com-
parison to an actual field installation. This short section would have the effect of re-
ducing the o2 stresses, which would have a decreasing effect on the elastic properties
of the soil (modulus of elasticity).
For pipe deflections of less than about 5 percent, the change in elastic properties of
the soil causes only small changes in the pressure at which failure occurred (approxi-
mately 15 percent change in pressure from failure and no deflection to failure at 5 per-
cent deflection). If the cell boundaries caused as much as 100 percent change in the
constrained modulus of the soil, the pressures at which failure occurred in the cell
would still be approximately the same as the failure pressure in actual field installa-
tion. Placement method, soil friction plus cohesion, moment of inertia, and area of
the pipe wall will probably shift the incipient failure line in Figure 22 up or down.
Because of the compensating effect of the stresses on the constrained modulus and
the low sensitivity of buckling to the constrained modulus, it is believed that the data
for buckling, at least at low deflections, are as good as any that are available.
Figure 18. Horizontal and vertical pressures on Figure 19. Confined stress-strain curve.
model culvert.
P• 11.0
P • 19. 7
PY 8. 100----------------.-----....-------
Applied Load
PY 16.
30, S PY 24.
P • 39, 3 PY • 31.

M • 2.
MS 3000 .

TEST 1-2 .0S0 IN ALU

2,5-in. diameter
MEASURED PRESS.
~ 40 1--- - - -+ - - - --+--.:;-¥1-l---- -+-- - - ---I
....
.;: 301 - - ----1------- • "',;l~;__.....Jl------l---- --1

PX • 9.
PX 16.
• 10S lb./ft. 3
PX 27.
PX • 34,

0. S I, 0 l.S 2. 0 2,5
Soil Strain (Per Cent)

Figure 20. Variation of the modulus Figure 21. Watkins and Moser's test Figure 22. Load deflection obtained
of horizontal soil reaction. cell. from data by Watkins and Moser.
Load Applied llere
)( :'lr,en.:cntal l'c-s1ol t~

000

40

.
");

100

18 20 22 24
E' (100 psi.) a)(/0

Figure 23. Large culverts in British Columbia Figure 24. Design chart for circular conduits.
(furnished by Chriss Fisher, Armco Corporation).

.
'b

,01 02 .o, .04 .o,

bX / 0
 43

LARGE-SPAN TESTS
Fisher (18) has installed several culverts with 40-ft spans. There have been no
failures, ancfthe measured deflections are usually less than 1 in. The secret of the
success of these large spans may be in the installation procedure. Fisher used what
he calls a thrust beam along the side of the culvert. Figure 23 shows the installation
of one of these large culverts.

PLASTIC PIPE
The Bureau of Reclamation in Denver, Colorado, is conducting field experiments on
plastic pipe. Because these studies are not complete at this time, no data are avail-
able yet.

DESIGN OF CIRCULAR CULVERTS

The way in which Watkins and Moser's (17) large-scale test data are plotted in
Figure 22 forms the basis for a possible design procedure. The M• fD 3 /EI lines shown
in Figure 22 are determined from the elastic theory for analysis of pipe presented by
Burns and Richard (8). Most of Watkins and Moser's data plot very close to calcu-
lated MlfD3 /EI values from the elastic theory (9). Watkins and Moser's data can be
used to delineate an upper boundary. The resulting design chart is shown in Figure 24.
As better failure data become available, which take into account all soil and pipe prop-
erties, the upper boundary can be adjusted accordingly. Future work may show that
one upper boundary is not sufficient to account for all variables. Examples for the use
of the design chart can be found elsewhere ~).

SUMMARY
In summary, a great deal of work has been directed toward a better understanding
of soil-structure interaction phenomena. Much more research is needed, but an or-
ganized approach needs to be made. At the current time, many individuals are con-
ducting research directed toward a better understanding of the performance of under-
ground structures. Much of this research would have been far more valuable if only
a few more dollars had been spent for proper instrumentation.

REFERENCES
1. Nielson, F. D., and Stankowski, S. Backfill and Multiple Installation Effects on
Culverts. Unpublished report.
2. Watkins, R. K. Structural Design Trends in Buried Flexible Conduits. Proc.
Symp. on Soil-Structure Interaction, Univ. of Arizona, Tucson, 1964.
3. Mohammed, A. I. Minimization of Stresses on Buried Conduits Constructed by the
Imperfect Ditch Method. New Mexico State Univ., Las Cruces, Master's thesis,
1970.
4. Tam, Y. B. Photoelastic Study of Multiple Pipe Systems. New Mexico State Univ.,
Las Cruces, Master's thesis, 1968.
5. Spangler, M. G. The structural Design of Flexible Pipe Culverts. Eng. Exp. Sta.,
Iowa State Univ., Ames, Bull. 153, 1941.
6. stankowski, S. An Analytical-Experimental study of Underground Structural
Cylinder Systems. New Mexico State Univ., Las Cruces, PhD dissertation, 1969.
7. Nielson, F. D. Modulus of Soil Reaction as Determined From Triaxial Shear Test.
Highway Research Record 185, 1967, pp. 80-90.
8. Burns, J. Q., and Richard, R. M. Attenuation of Stresses on Buried Cylinders.
Proc. Symp. on Soil-structure Interaction, Univ. of Arizona, Tucson, 1964.
9. Nielson, F. D., and Statish, N. D. Design of Circular Soil-Culvert Systems. Paper
presented at 51st Annual Meeting and published in this Record.
10. Watkins, R. K., Ghavami, M., and Lorghurst, G. R. Minimum Cover for Buried
Conduits Under Construction Loads. Paper presented at ASCE Transportation
Eng. Conf., San Diego, Calif., Feb. 1968.
44
11. Watkins, R. K. Characteristics of the Modulus of Passive Resistance of Soil.
Iowa state Univ., Ames, PhD dissertation, 1957.
12. Linger, D. A. Pressure Redistribution on Flat structures. Paper presented at
51st Annual Meeting and published in this Record.
13. Spangler, M. G. A Practical Application of the Imperfect Ditch Method of Con-
struction. HRB Proc., Vol. 37, 1958, pp. 271-277.
14. Larsen, N. G. A Practical Method for Constructing Rigid Conduits Under High
Fills. HRB Proc., Vol. 41, 1962, pp. 273-280.
15. Davis, R. E., and Bacher, A. E. California's Culvert Research Program-
Description, Current status, and Observed Peripheral Pressures. Highway Re-
search Record 249, 1968, pp. 14-23.
16. Willet, G. A., and Scheer, A. C. Behavior of the Rebuilt Cold Creek Culvert.
Dept. of Civil Eng. and Eng. Mechanics, Montana state Univ., Bozeman, 1968.
17. Watkins, R. K., and Moser, A. Response of Corrugated Steel Pipe to External Soil
Pressures. Highway Research Record 373, 1971, pp. 86-112.
18. Fisher, C. L. World's Largest Culverts. Proc., Highway Eng. Conf., New Mexico
state Univ., Las Cruces, 1969.
19. Young, D. F., and Murphy, G. Similarity Requirements for Underground Struc-
tures. Proc. Symp. on Soil-Structure Interaction, Univ. of Arizona, Tucson, 1964.
BALANCED DESIGN AND FINITE-ELEMENT
ANALYSIS OF CULVERTS
J. R. Allgood and S. K. Takahashi, U.S. Naval Civil Engineering Laboratories

This paper outlines a method for treating culverts in embankments. De-

sign is accomplished with approximate relations based on empirical deter-
mination of arching and maximum induced moment. Relations are provided
that permit determining the factor of safety against failure and collapse in
the various possible modes. This permits the adjustment of designs to
achieve a desired balance in these modes against failure and collapse. Dis-
cussion is provided regarding finite-element analysis of resulting designs.
A three-dimensional linear finite-element solution is provided for the design
example. It is shown that linear solutions predict deflections reasonably
well but that they overestimate thrusts and moments. The area of utility of
the approach as compared to other methods is discussed.

• A BODY of knowledge germane to the culvert problem has resulted from efforts to
develop design and analysis methods for buried shelters in order to resist the effects
of nuclear weapons. This paper adapts that information to the design of culverts. The
principal goal is to provide a method of solution for culvert problems excluding the
class of small culverts where handling and durability govern the design.
A classification of conduits, including a distinction of those governed by handling
and durability, may be found elsewhere (!). NCHRP Report 116 (!) reviews the older
design methods and delineates their limitations and deficiencies. Also, it proposes
two new methods: one based on foundation settlement considerations and the other
based on elastic theory relations ~' 1).
An approximate approach that circumvents some of the limitations of the elastic
theory has been formulated elsewhere (1). The use of this method is limited to fully
buried cylinders under high loads. It utilizes the elastic theory and an empirical
arching relation.
Among the contributions from weapons effects work is the development of relations
for modeling earth materials (Q, _§_). Some of these relations have been incorporated
in nonlinear finite-element programs that permit far better analysis of soil-structure
systems than has been possible in the past (1, ..!!, ~).
The approach recommended here is to use the approximate method to obtain approxi-
mate designs and then to use the finite-element method to analyze the resulting designs.
The importance of the soil and of the relative stiffness in controlling system behavior
is emphasized. It is control of these independent parameters that permits efficient
designs.
The following notation is used in this paper:
Ao = maximum active arching,
A. = area of section at springing,
b = width of base of embankment,
c = distance from midplane to extreme fiber of plate,
C = constant that depends on the mode of buckling,
C = constant that depends on the Poisson's ratio of steel,

Sponsored by Committee on Subsurface Soil-Structure Interaction .

45
46

Cxx, Cxy, ... Cys = material property constants,

C1, C2, C3, C4, C 5 = constants to account for different bedding conditions and flow
characteristics of the soil,
D = cylinder mean diameter,
d. depth to the plane of equal settlement,
E = Young's modulus of elasticity,
E' = modulus of soil reaction,
f( ... ) = function of several variables,
Gxy = shear modulus,
h = height of embankment above the invert,
I = moment of inertia,
N = thrust at the spring line,
P = surface pressure,
P1 0 = transitional buckling stress,
t: = equivalent thickness of culvert,
l:ii.x = horizontal diametral extension,
l:ii.y = vertical diametral extension,
£0 = unit vertical strain of culvert across diameter,
£8 unit vertical strain in the soil,
¢ = angle of friction,
Y = unit weight of soil,
O'allow = allowable stress,
cry yield stress, and
O arching coefficient.

BASIC CONSIDERATIONS
At the outset, it is worthwhile to establish a few basic definitions and to review the
dominant knowns and unknowns of the culvert problem. For present purposes, failure
will connote the occurrence of visible distress as indicated by wall crushing or exces-
sive cracking; buckling; plastic deformation (other than local); separation or rupture
of seams and joints; or excessive deflection that impairs the functional performance
or psychological acceptance of the installation. This definition requires the establish-
ment of suitable design criteria (!, _1). Designs should be evaluated in terms of their
failure and collapse loads.
Dimensional analysis and experiments on buried cylinders in a uniform granular
soil field (!Q) show that to a first approximation

i f(£i. ' ~ ' E~iD

= 3 ' ~
0

) (l)

in which y = radial deflection of the crown, D = cylinder mean diameter, p 1 = interface

load at the crown, M, = effective secant confined compression modulus at a stress equal
to the applied load, L = cylinder length, EI = cylinder wall stiffness, and do = depth of
cover over the crown.
For culverts, the effect of L/D is usually negligible, and the behavior can be ex-
pressed by the remaining four nondimensional pi-terms. Equation 1 may also be de-
duced from the elasticity theory on neglecting the influence of Poisson's ratio.
From past analyses, tests, and experience, we know the following:
1. EI/D 3 strongly influences the magnitude and distribution of the load on the culvert;
2. Soil stiffness, M., and variations of the soil field in the vicinity of the culvert
(bedding and backfill) strongly influence behavior;
3. Time effects are usually small except for clayey soils (with clayey soils, time
effects result in a shift in load to the pipe);
4. The possibility of shear failure in the soil in most circumstances is slight until
the onset of a culvert failure;
5. Live loads due to traffic are significant only for depths of cover of less than
about 5 ft;
 47

6. Extensional flexibility does not affect the behavior or design of steel, aluminum,
or reinforced concrete culverts; and
7. Use of the constrained (one-dimensional compression) modulus is advantageous
because it is relatable to dry density, vane shear strength, and certain other useful
indexes of field soil conditions and because it is a standard, widely used laboratory
test.
Principal uncertainties or deficiencies in knowledge of culverts are attributable to in-
sufficient data on the influence of variations in the adjacent field around the pipe-par-
ticularly the manner in which behavior is influenced by bedding, backfilling, and
backpacking; deficiency of measurements on actual pressure distribution, especially
in regard to interface shear; and inadequate criteria and definitions relating to failure
and factor of safety. Also, analytic methods available in the past were inadequate for
designing culverts under high fills or conduits with shallow overburden. Within limits,
the elastic theory is suitable as a basis for design.
A principal weakness of the elastic theory is that it does not account for arching as
it occurs in soils. According to the elastic theory, the minimum thrust possible at the
spring line in any culvert is equal to pr; but it is known from numerous experiments
that the thrust is usually less than pr (where r = D/2) (!Q, .!.!, 12, fl). A comparison
of thrust calculated from an empirical arching equation (4) and from the elastic theory
shows a 20+ percent divergence. From these results, it is clear that thrust cannot be
determined directly by the elastic theory. Moments from the elastic theory are also
in error, especially at high loads.
Elastic theory is adequate for determining deflection if (a) the boundary conditions
correspond to those in the theory, (b) the soil field surrounding the culvert is uniform,
(c) M. is taken as the secant modulus to the stress-strain diagram at a stress corre-
sponding to the applied load, (d) the shear stresses do not exceed the shear strength of
the soil, and (e) time effects are negligible. Conditions (a) and (b) often are not met in
practice; consequently, adjustments to the applied load and the effective modulus may
be necessary. Space does not permit an elaboration of how such adjustments are made;
however, if one recognizes that behavior is governed primarily by the relative strain
in the soil with respect to the average strain over the height of the inclusion, the method
of adjustment is fairly obvious.
Other considerations that aid in converging on a sensible design approach are as
follows:
1. Deflection and wall-crushing are the most likely failure modes where existing
seams are properly designed.
2. Moments are more variable than deflections and thrusts.
3. Deflections are the most predictable quantity and are the easiest to measure in
a completed installation. However, deflection is not the best criterion for the design
of a flexible culvert because, for such culverts, deflection has a weak dependence on
culvert stiffness. By elimination, then, wall strength and soil stiffness are the ap-
propriate principal bases for culvert designs where handling and durability do not
govern.
Ideally, one would like to achieve an equal probability of survival in all possible
failure modes. This means that, for a "balanced design," a higher factor of safety is
required against buckling than against wall-crushing and that a higher factor of safety
is required against wall-crushing than against deflection failure. Also, one might
logically have separate factors of safety against failure and against collapse.
A final item of background is desirable regarding the use of the elastic theory versus
the Iowa formula (!, 14). Actually, the Iowa formula conforms closely to the elastic
theory as may be seen by writing the deflection equations in a similar form (with Ax~
Ay) as follows:
Elastic theory (full slip):
(Ay /D) /(p/M,) (2)
48

Iowa formula:
(3)

These two equations differ appreciably only for M./(EI/D 3 ) < 10. For M./(EI/D 3 ) > 103,
the two equations give essentially identical results.
In its usual form, the Iowa formula contains a so-called modulus of soil reaction,
E', and coefficients to account for different bedding conditions and flow characteristics
of the soil. By letting M./(EI/D 3 ) go to infinity in Eqs. 2 and 3, it may be shown that
M. ,... 1.1 to 1.5E'. For a value of the bedding coefficient of 0.1, M. = 1.22E'. Use of
the bedding coefficient is an indirect means of accounting for different arching condi-
tions. It is desirable to determine arching directly and to use M. instead of E'. These
features and the foregoing requirements and reasoning are reflected in the design pro-
cedure that follows.
Because the main deficiency in our knowledge concerns the load-carrying capability
of the soil, a few words of clarification are in order.

SOIL PARAMETERS
The term "constitutive properties" (or laws) is used in continuum mechanics to mean
the set of equations that define the stress-strain properties of the media of concern.
Rather refined constitutive relations have been developed to represent earth materials
(.§_, .fil, and some of these have been incorporated in the computer programs that will be
discussed later.
Fortunately, the required constitutive relations for static loading are relatively
simple because unloading is not usually involved. For approximate calculations, the
following are required: an average value of Poisson's ratio, 11.; sometimes the cohe-
sion, c; the coefficient of lateral earth pressure, Ko; the angle of friction, ¢; and the
effective secant modulus at a stress equal to the applied load, M•. One should keep in
mind that, for embankments with large height-to-width ratios, the effective value of
Ko may be less than in fully buried installations. Of the preceding properties, the
secant modulus is of dominant importance.
Soil elements at different locations throughout an embankment are subject to differ-
ent confining stresses; thus it would be expected that the effective modulus would vary
throughout the cross section. For a properly designed and compacted embankment,
the effective modulus in the bottom central region should be nearly equal to the con-
fined compression modulus. The effective modulus at a given depth should correspond
to the overburden stress at that depth, except near the sides. Near the sides the ef-
fective soil modulus decreases, but so does the load; consequently, the conditions at
the midbottom region may be presumed to control the design. (stress conditions near
the ends are discussed later in the paper.) These comments are predicated on the
assumption that the width of the embankment is large (> lOD) compared to the diameter
of the pipe. If this is not the case, the effective soil modulus will have to be appropriately
reduced.
Designers of culverts should appreciate the wide range of values that M. may have
depending principally on the soil type, the placement methodology, and the applied
stress. M. may vary from essentially zero for saturated clays to several hundred
thousand for granular (locking) materials under high stresses. One should also be
aware of the large variability in M. attributable to placement. As an example, in lab-
oratory tests (1!) where care was taken to replicate placement of dry sand in a test bin
by using the sand-fall method, variation of M. from the mean value was ±20 percent;
certainly a greater variation must be expected in field installations.
Because of the wide variability in soil properties, the use of an unduly refined de-
sign procedure does not seem appropriate. The need is for a design method based on
the proper criteria that contain the principal parameters in correct relation to one
another.
 49

DESIGN PROCEDURE
A proposed design procedure for culverts in embankments is given by means of an
example. An example employed by others(!) is used so that comparisons can be made.
We are required to select a section for a 60-in. diameter steel culvert under a 20-ft
embankment of soil weighing 120 pcf. The basic steps in the design are as follows:
1. Determine the vertical stress at midheight, p., and the vertical stress at the
elevation of the crown, Pv•
Dead load: Pa = [h - (D/2)]y = [20 - (5/2)](120/144) = 14.6 psi
Pv = (h - D)Y = (20 - 5)(120/144) = 12.5 psi
Live load: 0
2. Determine M, corresponding to the stress at midheight of the culvert from con-
fined compression test results. For carefully placed granular fills compacted to at
least 85 percent AASHO T-99, one may use M, = 1,000 p.0 • 0 if no test results are avail-
able (!Q). In the present example, use M, = 1.22 E' = 1.22 x 700 = 854 psi to conform
to the comparison design (!).
3. Estimate the arching A and calculate the thrust from
N = Pv0 - A)(D/2) (4)
The relations A= 0.2 - 0.2 [1 - (do/D)] for do/D ~ 1.0 and A= 0.2 for do/D > 1.0 may
2

be used if better bases for the estimate do not exist (16). Bedding angle, projection
ratio, and certain other factors will influence A: -

N = 12.5 (1 - 0.2) (60/2) = 300 lb/in.

4. Calculate the stiffness required by the handling criterion D2/EI ~ 0.0433:

EI = 60 2/0 .0433 = 83,200 lb-in. 3/in.
Thus, M,/(EI/D 3 ) = (854 x 60 3 )/83,200 = 2,220.
5. Determine the moment using Figure 1 (experimental curve) with

P1 = Pv (1 - A) = 12.5 (1 - 0.2) = 10 psi

M = 0.005 p 1 D2 = (0.005 x 10.0)60 2 = 180 in.-lb/in.
6. Determine the equivalent flat plate thickness required to resist the thrust and
moment based on a factor of safety of 2 against yielding of the total section. Use the
approximate relation
O'auow = O'y/FS = (N/A.) ± (Mc/I)

where A.= t x 1 and I= t!/12. Substituting, 33,000/2 = 300/t. + (180 x 6)/t; then
0

t. = 0.265 in.
7. The corresponding stiffness is
EI= [(29 x 10 6 )0.265 3 ]/12 = 44,950 lb-in. 2/in.
which is less than 83,200 lb-in.2/in.; therefore, handling governs. Use a 12-gauge plate
with 22/2- by ½-in. corrugation (11). Spangler chose the same corrugation but a 10-gauge
plate (!). Calculations are given as follows for both gauges:
EI
1Y
I
12
6
= (29 x 10 )0.00343
60 3
= O 461 . EI
• , 1Y
I10
= 0 604
•

8. Determine M./(EI/D 3 ) and find the corresponding value of (ay /D)/(pv/M 1 ) 2e

r. /C r.. using Figure 2 with C = (1 + 11,)(l - 211.)/(l - 11.):

0
50

M,
""'I/'"'3
I = V.':t
854
n • 61 = 1,854
~ 3 I = 1,413
.i!, LJ i 12 J. EI/D 110

MiJJ_
Pv7M: I
12
:,! 2.46 tJ..y/D
Pv!M,
I
10
= 2.44

9. Calculate the depth to the plane of equal settlement (1) or use d 0 = D for h ~ 2D;
then, determine the arching coefficient

0 = (2d./D) [(£c/ £,) - 1]

and find the arching using Figure 3

012 = (2 X 60)/60 (1.83 - 1) = 1.66

A = 0.2 = Aa,sumed

therefore OK. When the predicted and calculated arching values do not agree, iterate
as necessary to bring the two values into agreement.
10. Determine the conformance with design criteria and the factor of safety for the
various possible modes of collapse. For deflection, from step 8,
tJ..x 12 :!e! tJ..y 1 2 = 2.46(pa/M.)D = 2.46(14.6/854)60 = 2.52 in,
/:J,.ylO = 2.38 in.
tJ..x/D = (2.52/60)100 = 4.2 percent < 5 percent
therefore OK.
FS I ~!vlng = 0.20/(tJ..x/D) = 0.20/(1.88/60) = 6.4
An alternate would be to calculate the buckling load corresponding to the second mode
and use FS I caving ""'P1cr(2)/P1- For wall crushing, by limit-equilibrium of the soil block
above the culvert,
FS Iwall = [2(1 - A)(2cd. + O'yt.)/ayt.(D - 2d.Kctan ¢)]
For granular soils, c = 0, d 0 = D/2Ko, and Ao,., tan¢

FS I wall = [2(1 - A)/(1 - Ao)] = [2(1 - 0.3)/(1 - 0.87)] = 10.8

For seam strength, load = 300 x 12 = 3,600 lb/ft and capacity (!1) = 23,400 lb/ft
FS Iseam = 23,400/3,600 = 6.5

For transitional buckling (1) ,

where
C 6~;
B = 0.75 for Lis = 0.3, D/b < 0.2; and
C = (1 + lls)(l - 21,1.)/(l - 11.).
For v. = 0.33, C = 0.742, and C = 4.5

Pier = 4.5 J 854 X 0.461 = 89 psi

 51

FS I
buokllnc = 89/10 = 8.9
11. Determine longitudinal deflection and tension, bending, and durability require-
ments in the manner suggested elsewhere (!). Other methods yield essentially the
same design(!). For large diameters, where handling no longer governs, resulting
designs are different.
One could argue that the minimum factor of safety of 6.4 is excessive for many in-
stallations; however, reducing the factor of safety for such small culverts is not pos-
sible wtless special handling provisions are instituted. The actual factor of safety in
an installation is probably greater than 6 because the value of M. used (854 psi) is low
for granular fills if reasonably good construction controls are maintained.
Principal advantages of the proposed method are that it permits treatment of
embankment-culvert systems of all materials, depths of cover, and sizes; incorporates
a rational method for accounting for arching; considers all potential modes of failure
and collapse and enables achievement of a "balanced" design where the factors of safety
in the different modes are in a desired ratio to each other; incorporates the means of
accounting for moments; and enables the logical design of systems with backpacking (4).
The principal deficiency of the proposed method is that resulting designs for large-
diameter culverts are strongly influenced by induced moments that are subject to
relatively large variations. Also, the method involves the use of empirical relations
with constants that are not as yet defined for cohesive soils.
Other deficiencies of the proposed design method are almost too obvious to mention.
Clearly, test data on large pipes and conduits are needed for Figure 1. No tests on
prototype culverts with the required measurements are known to exist although in-
strumentation exists to obtain the needed quantities. Empirical exponents are also
desired for cohesive soils (tests are being planned to obtain the data). The method
can be used with reduced accuracy without the data by assuming a conservative value
of arching. In spite of these limitations, the method represents an improvement over
prior ones because (a) it accounts for all principal variables, (b) all possible modes
of failure are considered, and (c) the factor of safety in all modes can be estimated
and, based on these values, adjustments can be made to the design to achieve a de-
sired "balance" among them.
Important designs developed with the preceding methodology can be analyzed by
using the finite-element method as is indicated in the following section.

FINITE-ELEMENT ANALYSIS
Once a culvert is selected, two- or three-dimensional finite-element analyses can
be performed on the system. One may obtain two-dimensional linear or nonlinear
solutions that incorporate either small- or large-deformation theory ~' 18, ~' 20).
Also, the linear small-deformation code has been modified to permit accbunting for
interface slip and boundary separation~. Some of the referenced codes have been
modified to account for the presence of initial stresses and for dynamic loading. Ex-
amples of the use of some of these two-dimensional codes can be found elsewhere ~ '
21, 22).
- If other than stresses and deformations on a transverse section are desired, a
three-dimensional solution is required ~). The remainder of this paper is devoted
to presentation of a linear three-dimensional solution of the problem employed in the
example design (with 10-gauge plate) of the previous section. The geometry and ma-
terial properties used are given in Figures 4a and 4b respectively. Only one-quarter
of the soil-culvert system was analyzed because of symmetry. Note that different
moduli are used for the different layers because of differing dead load stresses at
different depths. The elastic modulus at the culvert corresponds to the M, used in
the design example.
One thousand, five hundred and sixty-eight 8-node hexahedron solid elements were
used to represent the soil, and 84 shell elements were employed to model the culvert.
The culvert was represented by a plate of equivalent thickness (0.387 in.) to properly
model the stiffness of its transverse section. No correction was made to account for cor-
52

rugations; thus, the longitudinal stresses in the cylinder would be expected to be greater
than those where circumferential corrugated plate is used. stresses in all elements
and deflections at all node points were obtained as output data; however, space limi-
tations permit visual display of only a small portion of these data.
Vertical soil stress contours in the Y-Z plane and in the X-Y plane are shown in
Figures 5 and 6 respectively. As may be observed, the interface stress at the crown
is about 15 psi as compared to 10 psi predicted with the arching relation in the example
design. At the invert, the normal stress is about 16 psi. The horizontal stress at the
spring line is about 15 psi. It is interesting to note from Figure 6 that the vertical soil
stress is greater at about 1 to 2 ft above the culvert than it is at the crown.
Horizontal stress contours in the Y- Z plane are shown in Figure 7. Perhaps the
most significant aspect of the horizontal stress is the rapid dispersal of the stress con-
centration adjacent to the spring line.
Stresses and forces in the culvert are shown in Figures 8 and 9. The contour plots
in Figure 8 show the longitudinal and circumferential stresses in the extrados of one-
half of the developed longitudinal section. At the center spring line the circumferential
stress in the extrados is about 11,100 psi.
The peak longitudinal stress is about 18,000 psi for the modeled plate; however, it
would be less for a longitudinally corrugated culvert.
Forces and deflections on the transverse sections at midlength and one-quarter of
the total culvert length from one end are shown in Figure 9. It is interesting to note
that the thrust at the spring line is about double the thrust at the crown and invert.
Moments at the crown and spring line are about equal in amplitude but of opposite sign.
Horizontal diametral expansion at the center section is 1.27 in., and the vertical
diametral shortening was 1.35 in. The corresponding vertical deflection determined
in the design was 2.38 in. The deflection by the Iowa formula, excluding deflection lag,
was 2.14 in. Absolute displacement of the invert at the center section was 4.77 in.
This is the amount of camber that should be provided initially to ensure that the longi-
tudinal axis of the culvert is straight when the embankment is completed.
In the example used, design deflections are in approximate agreement with values
from the finite-element solution. Peak thrusts and moments are about a factor of 2
larger than in the design. The reason for this is that the elastic theory does not prop-
erly account for the arching in granular soils. Two-dimensional elastic theory indicates
that, for M,/(EI/D~) = 1,850, the thrust at the spring line is 1.32 Pr• Experimental
data and the arching theory give the thrust as 0.8 Pr• The latter value is considered
correct for granular soils; however, elastic theory results may be more nearly cor-
rect for other than granular soils. Of course, knowledge of stress distribution from
elastic finite-element solutions is useful. Codes with constitutive relations that prop-
erly model soil behavior must be used to obtain correct amplitudes.
Incidentally, stresses and deflections are expected to be lower than the values from
the example for installations in granular soils. The reason is that the modulus of
properly compacted granular fill would be an order of magnitude greater than that used.
The low value was used to permit comparison of the results with existing designs for
the same problem.
Although the example design and analysis were for a steel culvert, the same method-
ology is applicable to concrete culverts. The principal difficulty in the design and
analysis of concrete cylinders is that the effective section modulus changes with load
as fine cracks develop around the perimeter. As a consequence, concrete cylinders
are not as "rigid" as is often presumed.

SUMMARY
This paper presents the bases for improved design and analysis methods for culverts.
The design method illustrated is the only known method that permits accounting for
arching and moment as an integral part of the design procedure. This is unimportant
for small metal conduits under moderate fill heights because handling and durability
usually govern designs under those circumstances. For culverts of cementitious
materials and for all culverts under high loads, induced moments become important,
Figure 1. Moments in buried cylinders. Figure 2. Deflections of buried cylinders
(corresponding to average of full• and
no-slip cases from the elastic theory and
V = 0.3).
0 .035 0 014 STIFF

'
0.030

0 025
" ELASTIC THEORY, NO SLIP
0 012

0010
o DORRIS
N ¢ ALBRITTON
Q 0020 0 .008
ir V NCEL

" 0 015
6. MARINO
o DORRIS & ALBRITTON 0.006
0 .__iL,._..__.._............,. ,_,_...........- ............
0010 ---- - 0.004 10- 1 1oD 10 1 10 2 1oJ 104 105 106
M,
0005 0.002
El /0 3

0
,o1 10 2 106

Figure 3. Plot for determining arching over Figure 4. Soil-culvert system model.
culverts in granular soil.
MATERIAL PROPERTIES FOR CULVERT !PSI)
NODAL POINTS - 2042 y Cxx = JJ,000,000
ELEMENTS - 1750 C11;y = 9,890,000 ELEV
CKS :: 0 (FTI
Cyy = 33,000,000
Cys = O 37.15
MAT. MODULUS
Gxy :. 11,500,000
NO. (PSI)
li.- -"""-'' -- ---tl2,5
15 464 2
1 - -- - -- - - - - t ~S
14. 515,8

~
~ 02 - l

r.
<

f-..CC
3 "'----
6 8_6c...1,__ _--t 10 O
·o ___J
0 0 ,2 04 06 08 X. 722 2
o,,:,/3 - 2
1 - - - - - - - -----is.o

-uf
a: =-(1 -1; )1-A

11 _ff ..ol!lta
-1 \% ~
·O
1. 766.6

POISSON'S RATIO = 0 3

fl =-2!!.a (~ -
D Es
1),
'
do 5 D UNIT WEIGHT "' ,06944 L8/1N ,3

A . QUADRANT USED 1N ANALYSIS. B, MATERIAL PROPERTY REGIONS

Figure 5. Vertical soil stress contours (Y-2 plane). Figure 6. Vertical soil stress contours (X-Y plane).

(
2
4
- ,,.. ST RESS IN

----
POUNDS PEA SQUARE INCH

14

6 14
8
0 14
2
4
6
~ SEE
ENLARGED
F.IGURE
....
--~
8
0
2 10
18
4 12
--.: 14
6
28 '-------'---- -- - -- - - ---.:__::..._.c,......:.:_....o,_..:::::,,._.....J 16
Figure 7. Horizontal soil stress contours (Y-Z plane) .

.--- -- - ~! ~ S Q U A R E INCH

10

11

Figure 8. Stress contours on developed longitudinal half-section.

- 10,000

---------------------- - - 5.000
10000 CEN'rFR l lNE

- ---- ---------------- 0

-10,000 • .000

A CIRCUMFERENTIAL STRESS AT EXTRADOS, /

>, STRESS IN POUNDS PER
SQUARE INCH

-~---==-------------- ;,;.; :;,.-\

-15,000 ~,000 CROWN:?,

[ .i ~--=:;-- . /-~ 8 LONGITUDINAL STRESS AT EXTAADOS,

Figure 9. Forces and deflections on transverse section.

A, THRUST ANO BENDING MOMENT. B HORIZONTAL ANO VERTICAL DEFORMATIONS~

 55

and economics require the use of arching. It is in these circumstances that the method
given is useful.
Important installations warrant detailed analysis, and the finite-element method is
useful for this purpose. A variety of two-dimensional codes may be used, including
one that accounts for interface slip and boundary separation. Linear three-dimensional
solutions, as illustrated in the text, may also be obtained; however, the magnitudes of
the resulting thrusts and moments will be markedly larger than those in actual installa-
tions in granular soils.
Key points in the design of culverts are as follows:
1. If adequate construction procedures are followed, handling and durability con-
siderations govern the required section properties of metal culverts less than about
5 ft in diameter under fills or equivalent loads less than about 30 ft in height (!).
2. For larger diameters or equivalent fill heights, good design and economics re-
quire consideration of moment and arching. With the larger diameters, achieving soil
control via quantitative measurements becomes essential.
The methods outlined here, together with the use of backpacking in accordance with
the relations found elsewhere (1), should permit more accurate and efficient designs
of culvert systems than has been possible heretofore.

ACKNOWLEDGMENT
The efforts of W. Erickson in preparing and running the finite-element problem is
gratefully acknowledged. The work on which this paper is based was sponsored by the
Defense Nuclear Agency.

REFERENCES
1. Krizek, R. J., Parmelee, R. A., Kay, J. N., and Elnaggar, H. A. Structural
Analysis and Design of Pipe Culverts. NCHRP Rept. 116, 1971, 155 pp.
2. Burns, J. Q. An Analysis of Circular Cylindrical Shells Embedded in Elastic
Media. Univ. of Arizona, Tucson, PhD thesis, 1965.
3. Hoeg, K. Pressure Distribution on Underground Structural Cylinders. U.S. Air
Force Weapons Laboratory, Kirtland Air Force Base, New Mexico, Tech. Rept.
65-98, April 1966.
4. Allgood, J. R. Structures in Soil Under High Loads. Proc. ASCE, No. SM3
Proc. Paper 8006, March 1971.
5. DiMaggio, F. L., and Sandler, I. Material Models for Soils. Defense Atomic
Support Agency, Rept. 2521, April 1970.
6. Nelson, I., and Baron, M. L. Application of Variable Moduli Models to Soil Be-
havior. Internat. Jour. Solids and Structures, Vol. 6, 1970, pp. 1-19.
7. Wilson, E. L. A Computer Program for the Dynamic Stress Analysis of Under-
ground Structures. Univ. of California, Berkeley, Rept. 68-1, Jan. 1968.
8. Farhoomand, I. Nonlinear Dynamic Stress Analysis of Two-Dimensional Solids.
Univ. of California, Berkeley, PhD thesis, 1969.
9. Wilson, E. L. SAP-A General Structural Analysis Program. Univ. of California,
Berkeley, Rept. UCSESM 70-20, Sept. 1970.
10. Allgood, J. R., Ciani, J. B., and Lew, T. K. Influence of Soil Modulus on the
Behavior of Cylinders Buried in Sand. U.S. Naval Civil Eng. Laboratory, Port
Hueneme, California, Tech. Rept. R-582, June 1968.
11. Allgood, J. R., and Herrmann, H. G., III. Static Behavior of Buried Reinforced-
Concrete Model Cylinders. U.S. Naval Civil Eng. Laboratory, Port Hueneme,
California, Tech. Rept. R-606, Jan. 1969.
12. Albritton, G. E. Behavior of Flexible Cylinders in Sand Under Static and Dynamic
Loading. Waterways Exp. Sta., Vicksburg, Mississippi, Tech. Rept. 1-821, April
1968.
13. Dorris, A. F. Response of Horizontally Oriented Buried Cylinders to Static and
Dynamic Loading. Waterways Exp. Sta., Vicksburg, Mississippi, Tech. Rept.
1-682, July 1965.
56

14. Spangler, M. G. The structural Design of Flexible Pipe Culverts. Eng. Exp. sta.,
Iowa state College, Ames, Bull. 153, 1941.
15. Luscher, U. Buckling of Soil Surrounded Tubes. Proc. ASCE, No. SM6, Proc.
Paper 4990, Nov. 1960.
16. Hendron, A. J., Jr., Gamble, W. L., Haltiwanger, J. D., and Newmark, N. M.
Design of Cylindrical Reinforced Concrete Tunnel Liners to Resist Air Over-
pressures. U.S. Naval Civil Eng. Laboratory, Rept. R 68.010, June 1968.
17. Handbook of steel Drainage and Highway Construction Products. American Iron
and steel Institute, 150 East 42nd street, New York, LC 66-21555, 1967.
18. Wilson, E. L. structural Analysis of Axisymmetric Solids. American Institute
of Aeronautics and Astronautics, Jour., Vol. 3, No. 12, Dec. 1965, pp. 2269-2274.
19. Wilson, E. L., Farhoomand, I., and Rukos, E. Dynamic Response Analysis of
Two-Dimensional structures With Initial stresses and Non-Homogeneous Damping.
U.S. Naval Civil Eng. Laboratory, Rept. CR 69.019, Nov. 1969.
20. Przemieniecki, J. S. Theory of Matrix structural Analysis. McGraw-Hill, Inc.,
1968, p. 384.
21. Nossier, S. B., Takahashi, S. K., and Crawford, J. E. stress Analysis of Multi-
component structures. U.S. Naval Civil Eng. Laboratory, Port Hueneme, Cali-
fornia, Tech. Rept. R-743, Nov. 1971.
22. Gates, W. E., and Takahashi, S. K. static and Dynamic Tests of Model Pressurized
Underground Fuel storage Containers. U.S. Naval Civil Eng. Laboratory, Port
Hueneme, California, Tech. Rept. R-728, May 1971.
STRUCTURAL DESIGN PRACTICE OF PIPE CULVERTS
Bernard E. Butler, New York State Department of Transportation

• A DISCUSSION of culvert design practice must include methods of correlating design

with actual construction procedures. Also pertinent are adaptations or extensions of
the most widely used design procedures (including durability), which satisfy the prac-
tical needs of both the designer and the installation.
This paper discusses these items, and reference is made to the methods used by a
large design and construction agency, the New York State Department of Transportation.
Various aspects of the New York approach may be useful and can be adapted and utilized
by any size organization if desired.
A great deal of research on culvert design has been done and is continuing. This
paper suggests that the results of significant theoretical studies be reduced to practical
terms so that they can be used routinely by all types of designers.
The structural design procedures of most organizations are based on the Marston-
Spangler formulas for rigid pipe and on adaptations of the "ring compression theory"
for flexible pipes. These methods have been thoroughly discussed, analyzed, and evalu-
ated in innumerable studies and will not be repeated here.
The results of all evaluations, however, generally support the same conclusion-that
current design procedures are satisfactory when properly applied. Proper application
must, of course, include engineering judgment, provision for durability, and consistent,
appropriate installation practices.
The most significant of the shortcomings commonly ascribed to the foregoing design
methods are as follows :
1. Interactions of the soil-structure system are not properly considered;
2. Some of the input parameters are difficult to ascertain and must generally be
assumed;
3. Results are usually conservative; and
4. Their applicability to extremely large structures under very shallow and ex-
tremely high fills is questionable.
Of all these criticisms, only the last one is not easily satisfied or provided for with
a satisfactory degree of confidence in current design practice. It is reasonable to
assume that the other shortcomings are sufficiently overcome such that practical re-
sults for routine installations are obtained. It is acknowledged, however, that the
resultant safety factor for structural criteria is usually greater than required. The
savings in cost that would accrue with the use of better-known formula input parameters
with either existing or new design methods, however, would not always be especially
significant for routine cases . For example, the minimum gauge of flexible pipe re-
quired to satisfy strength criteria after installation is often inadequate to also meet
handling and durability requirements.
Durability is not included as an integral part of the structural design process for
flexible structures. An initial attempt has been made in New York State to overcome
this shortcoming and is described later.
The number of designs made each year by any large agency such as the New York
State Department of Transportation is often staggering. Furthermore, design selections
are made by perhaps hundreds of individuals within a state (or within a large design
organization) who have limited expertise in soil mechanics and conduit structural

Sponsored by Committee on Subsurface Soil-Structure Interaction.

57
58

analysis. Consequently, the need exists for designs and installation details to be stan-
dardized for the most commonly used shapes of rigid and flexible pipes for heights of
cover up to about 100 ft. This concept is followed in varying degrees by many states.
The approach used in New York is, therefore, not new in all respects but is believed
to be quite comprehensive, easy to use, and successful in meeting the continuing
challenge of designing large numbers of culverts .
The section that follows discusses this approach and the factors involved in estab-
lishing uniform design and installation standards. It should not be assumed from this
presentation that current methods are considered to be entirely satisfactory or not in
need of improvement or change. A case is being made, however, for the workability
of current design methods when coordinated with construction practice and experience.

DESIGN APPROACH
The selection of a rigid or flexible pipe is not always made objectively. Sometimes
it is based on the personal preference of the designer. The most recent procedure now
being implemented in New York, however, is intended to be more objective; it is based
on anticipated performance and cost. This design selection pro cess first involves a
determination of the required cross section for rigid and flexible pipes (including hy-
draulic factors) . Structural analysis then follows, which must satisfy site require-
ments for fill height, foundation conditions, and durability considerations caused by the
structure's environment.
This procedure often requires that a design cost analysis be made for both rigid
and flexible pipes for many individual installations. Consequently, the desirability
of easy-to-use standardized designs becomes even more apparent.

Rigid Pipe
In order to reduce the design process for routine conditions to its simplest form,
i.e. , a fill-height table, we first established a minimum number of practical installation
conditions that cover all designs normally encountered in the field .
On New York State highway projects, for example, only two types of bedding are
considered to be consistently attainable in the field when using reasonable procedures
with varying qualities of inspection: class C or ordinary bedding and class A or con-
crete cradle bedding.
Only two design analysis loading conditions, positive projecting and imperfect trench,
are routinely needed for each bedding condition to cover all methods of pipe installation
commonly used on New York State projects. The positive projecting case covers the
embankment installation as well as the trench and negative projecting conditions because
the latter two types have been shown to require a trench width that is too wide for a
trench or negative projecting loading. This minimum trench width is based on a 2-ft
clearance between the pipe and the inside face of the trench plus the width of a sheeting
section, which is always required for safety with trenches 5 ft or more in depth . Sheet-
ing is seldom pulled incrementally as desired, and therefore the width of the sheeting
section must be considered as a part of the trench width for load analysis. With the
design analysis loading and bedding conditions thus established, the input parameters
for the Marston-Spangler formulas can be selected or assumed as required.
In the approach followed by New York State for reinforced concrete pipe, a safety
factor of 1 on the first crack strength and a soil unit weight of 125 pcf are used. The
variation in settlement ratio and load factor has been shown to have a limited effect on
the design results, providing reasonably representative constants are selected for these
parameters. Height of fill -gauge tables for all field installa tion conditions s uch as
embankment, trench (wide trench condition), and imperfect trench are therefore r ea dily
established by assuming values for settlement ratio and load factor and by making other
appropriate inputs into the load analysis formulas. The designer, using these tables,
needs only to select the "pipe strength-bedding type-installation method" combination
that most economically satisfies the field condition. New York State's current allow-
able fill-height tables for reinforced concrete pipe are shown in Figure 1. It is noted
that, in preparing these tables, design assumptions were modified and fill heights were
 59

Figure 1. Fill-height design tables for reinforced concrete pipe.

INSTALLATION METHODS

Method D Method A (Sheeted Trench)

(Imperfect Trench) Method B (Open Excavation)
Methods C-l&C-2 (Embankment)

MAXIMUM HEIGHT OF FILL (HI IN FEET MAXIMUM HEIGHT OF FILL (H) IN FttT
OVER TOP OF PIPE OVER TOP OF PIPE
PIP£
,. c~:~ ..
, ,.
IHIDJ.N','--._T 8EDDINO C0HC'lf.lr1 1c CRADLE
BEGOIHQ PIP[ 0A0INUY
,. , , , ,. ,
BEDDING
. .. .. ..
CONCRETE CftADU:
8[.0DING

.,.. .
IOIAMETEA Dt l. M:CTU
0 1411

...
m "fl" •~u
14,C 0
"" "" CU,H

.
liltche1l CLASS CLIJS CLASS (lrc11,,, C.U S!i CUS.t C:LAH CL' S, CLASS CLASS

.. ....
Cl,.ASS
:m

. u m: ur
l[
.,., Ill lI 1[

"I• ••,.
••••
76
•• II
.•••, •••• .....
17
60 110
It

'.
•• ll

... " "

H

. •• •• ••••
11

. ... .. •••• II

.,.
,o •• ••
II
t I
76
70
II 110
111 ., •• /J

,.II "' •• 14

,.. ,.. ....

11 JIJ
•• ,o• I0 ;4

••. ,.",.••
..
•• ,. ••
16
"'
..
27 I9

. ••. ••
2B

" . ... ..••

II

...•• •••• ,.,. •••• n ,,.

n
"' •• , ,. "
.
10 ll JI 14
113
"
.. . ." ..•••• .•...•
.....
•••• 50 17 II II

..••••,. . .... . .... .

., •• ,. 70
"' IT 41 10 II

•••• •• ,,. ,. ,,.". ."

00 17 IIJ 41 ,o JI 04

,.""
•• 00 74 111
.,.••
04 ,o
-
.. ..
II

..••.. .. ••••...
II
04
04

..•• ..,. ••.. -••. ...., ... ...

•• • ."'•
..!! It IO
•• •••• "
" ,. .
7l 74 f> j J 14 ZI I I

..•• --..,••••.
57 14 II u 04

10

IOt 30
41
41
74

74
OT
07
II

u
Ill
Ill

I I)
..••
00

101
",, .,. .
II

II
"
,.,. ,,
II

zo
u
.. II
11
17
17
24
14
'"

., •••• "'
...

.... ••.... •••••• .. ••••

,oe
.. ..."'
~
101 10 41 74 II 20 2B 17 24

,.
.
48

•• •• ...
114 73

...,. ...
114 21
e-!!L
,
,. . ••
IU u . 4•

48

47
n

72
7Z
72
07

..•• ..•••• "'...

41 n

I J
110

''"
144
,."" ..
II
ZI

111
ZI
u
•• . ••
20
-..- 24

14
ss-
H
14

Notes:
1. H-20 live loading effects are accounted for in the above
tables with a minimum 2 ft. cover.
2. All design selec tions are field const~ucted in accordance
with the corresponding Installation Methods and Bedding
Details on the 11 Standard Drawing 11 except where modified by
special requirements.

reduced where initially calculated allowable fill heights greatly exceeded 100 ft to ensure
that these installations are given special consideration.
Although it is acknowledged that many excellent manuals (1) and guidelines exist for
the application of the Marston-Spangler formulas, few if any-are intended to be directly
related to the actual construction practice of a given agency.
The key aspects involved in coordinating construction practice and design include
standards for bedding, backfill materials, compaction requirements, minimum tem-
porary cover and final cover, and construction procedures. Again, using New York
State as an example, the coordination of construction and design is accomplished on a
standard drawing that automatically becomes part of the plans and specifications when-
ever rigid pipe is used. Bedding details are shown that include additional procedures
to be used by the field engineer in preparing the foundation. For e xample, rock fow1 -
dations are 1·equired to be undercut below invert to as much as 75 percent of the pipe
diameter, and temporarily unstable wet soil is replaced during the bedding process.
Installation details a.re also shown for the trench and imperfect trench methods and
two conditions of embankment installation: one for placing the pipe before filling and
the other io1· building the fill to partial height and excavating a trench for pipe place-
ment. All pipes after bedding, whether in cut or fill, are backfilled with a specified
type of granular material placed to minimwn established limits and compacted to
between 95 and 100 percent of the maximum density as determined by AASHO T-99
Method C.
Durability design for rigid pipes must not be completely overlooked although it is not
so significant a consideration as for flexible pipes. A brief but informative guideline
on this subject for concrete pipe is given elsewhe1·e (~_).
60

Flexible Pipe
The most widely used design methods in current use are presented elsewhere (3, 4,
5). Each presentation is essentially the same and requires that the culvert be designed
fo meet seam strength, buckling, deflection, and flexibility criteria for the materials
and corrugation configuration being used. Height-of-fill-gauge tables are given in some
of the previously mentioned references, and similar ones have been prepared by many
design agencies, which also incorporate their experience.
The most frequently discussed problem concerning the use of the design method is
the inability to satisfactorily analyze the deflection aspect. This phase of the analysis
relies largely on [he determination or assumption of the modulus of soil reaction, E ',
for use in the Iowa formula developed by Spangler. Various approaches for determin-
ing a reasonable value of E' are summarized elsewhere (2). NCHRP Report 116 (2)
should be referred to when a deflection estimate is of specific interest. -
Extensive experience in New York highway work has shown that deflection of flexible
pipes has been insignificant, i.e., generally much less than the usual arbitrary 5 per-
cent criterion. This very favorable deflection performance is related to the compress-
ibility of the surrounding backfill and therefore is attributable to the installation
methods used. This has led to a recent decision to use round pipe exclusively .for
routine designs (except of course when pipe arches or other shapes are desired) and to
eliminate elongation and strutting as a design consideration and construction require-
ment. This modified design approach will not result in any practical sacrifice to the
value that E ' might attain if elongated pipe were to be used in place of round pipe. More
economical designs should be realized although any resultant savings may not be re-
flected directly in a contractor's bid prices for this work.
The design for durability is not so easily accounted for as was deflection. Use of a
New York State report (6) has been recommended for corrosion analysis. Durability
design in New York, however, now includes considerations for abrasion, flow condi-
tions, and service category as well as corrosion. In addition, the New York State cor-
rosion study has been continuing, and based on more data, refinements, and new find-
ings, the original report (6) is now used only as a reference for the corrosion part of
durability design. -
Attempts have been made to evaluate the corrosion aspect for a particular site by
a series of field tests. Unfortunately, no completely satisfactory field test program
has yet been developed that can be correlated to all of the factors that affect durability.
New York State is currently initiating an interim qualitative approach to durability that
relies on experience and past performance until more refined methods are developed.
Durability design for corrugated steel pipe (using the New York State method) pro-
vides for the use of plain galvanized material where feasible. It also requires, how-
ever, the coating and paving of inverts where indicated by using a rating system test
applied to the proposed installation. Height-of-fill-gauge tables have been prepared for
all corrugations of plain galvanized material by using the aforementioned design meth-
ods, but the tabulated gauges also contain an additional built-in consideration for dura..-
bility. More specifically, each gauge that is listed must provide a minimum safety
factor of 1 for seam strength and buckling at the end of a 40-year design life, assuming
a uniform rate of metal loss over this period. The gauge table for round steel pipe is
shown in Figure 2. Although the basis for durability assumptions is beyond the scope
of this presentation, it may be of interest to lmow that rates of metal loss used were
1 mil per year for diameters up through 48 in. and 2 mils per year for larger sizes.
These rates correspond to a statistical confidence level of about 85 percent. The re-
maining aspect of durability design for steel pipe requires that. a determination be made
for the need to completely asphalt coat and pave the invert of 2%- by ½-in. and 3- by
1-in. corrugated material or to install a reinforced concrete paved invert in structural
plate (6- by 2-in.) pipes. It should be noted that experience in New York State has
shown that asphalt coating alone is not significantly beneficial, and its single use is not
recommeded. When a pipe with an asphalt coating and a paved invert or a reinforced
concrete paved invert for plate is placed in an aggressive environment, the statistical
rate of metal loss is equal to or lower than the rate for plain galvanized material in a
nonaggressive environment.
 61

Figure 2. Fill-height design table for round corrugated steel pipe.

0 ROUND CDlllllleATED ITHL PIH:: ITEMIIX

RDUND CORRlleATED ITEEL PIPE· l'AYl!O INVERT• ITEII IIXP

..-. ...... .. ..
-•-·"·
.- .. ---. .--. -. .- .-.. . . . .•-•.• •. .• I •. ••. 'l --.• ..- ... ...- . . . J· •
1 - • u - 11 •- • • ·• • • • •-• •-• • • • •·• •· • •·• •-• •·• 11-n ,. • • •· • •·• •-• •-• •-•
__ __.. ~ 11 -:: :: ...
IIILn.t =i~ I• • • • ,. • • 9 • •
"'
II •
:: l. .
J - • ·I•
•1 • .. .

. .
. .. ..•
II " ,.. ,..'
1•
."." . '"".
. . ,..." . '"". .
1• II

........ ...." . ... . .. ... .. .. ".

Iii 1• II Ill II
II

... . ,. ,....
.... •
....
•• . •
14ft
.,. .... 10 II 1111
II 14

0
~

II

........ . ....
111
00 I
IO
I
I
I
I
I

• • ••••
.. I I I
I

••
I I
I r
I
I I

,
I '
r
I
I
I ',
'
..
IO 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
'

A rating system (durability index) was devised to determine the need to treat steel
culverts. The system is intended to supplement experience where it exists and pro-
vide a rational substitute where none is available. The durability index represents the
sum of the individual numerical ratings assigned to each of the following categories .
1. Surface water corrosiveness rating: The location of the project site on a refer-
ence map determines the relative corrosiveness rating, i.e., 1, 3, 5, 7, and 9. The
numerical ratings, which are assigned to geographic areas of the state (Fig. 3), are
in general accord with available field data and the relative influence of contributing
corrosive factors in these are as such as soil types, water hardness, and pH. The
designer i s cautioned at this s tage, however, that abnormally high corrosion rates
related to unique local surface water conditions, such as heavy agricultural and indus-
trial water or mine tailings, may necessitate the use of a different pipe material.
2. Abrasiveness rating: This rating category, which receives about half the emphasis
of the corrosion category, describes potentially abrasive and nonabrasive bed loads and
relates them to the gradient for the numerical rating. An abrasive or nonabrasive bed
load with a gradient 2 percent or less has a rating of 1. An abrasive bed load with a
gradient between 2 and 4 percent has a rating of 2, whereas the same bed load with a
steeper gradient has a rating of 5.
3. Flow rating: This category attempts to recognize the effect of frequency of flow
for various applications of pipe usage and is based largely 011 judgment. Ratings fo r
three conditions of flow and usage are as follows: (a) "highly intermittent" with a r ating
of 1 applies generally to storm and side drains that are no t' s ubject to long-standing
water; (b ) "intermittent" has a rating of 2 and generally applies to cross culverts carry-
ing streams or small flows that seasonally go dry; and (c) "continuous" with a 3 rating
usually applies to cross culverts with stream flows or any long-standing water condition.
62

Figure 3. Guide map for corrosiveness rating.

....,.., """'""' ~!1!11••--

.., ...
A
I
C
D
[
-
tow

111th
worJhlth
3
II
T

4. Service rating: The relative importance of installations is given some recogni-

tion in this rating section, again based on judgment. Side drains and driveway pipes
are assigned a rating of 1. Cross culverts have a 2 rating.
The durability index is obtained by simply adding the preceding ratings that apply
to the site. A durability index of less than 14 indicates that plain galvanized material
can be used, whereas a higher rating indicates the need for coatings and invert paving.
Extensive experience in a project area, as stated earlier, would possibly modify the
decision indicated by the durability index.
The durability design aspects for aluminum material are different from those for
steel. The New York height-of-fill-gauge table covers 2%- by ½-in. and 9- by 21/2-in.
corrugation material and has no built-in allowance for metal loss. Currently, asphalt
coating and paved invert treatment are only to be considered for the extremes of pH
commonly mentioned (6) or where a high abrasion potential exists. This interim ap-
proach, which is based on much more limited data, does not necessarily imply that
we are not concerned with the durability of aluminum. This is especially true because
a recent finding at a few sites in New York has revealed soil-side corr0sion at the
crown. This phenomenon is thus far unexplained.
A problem that frequently confronts designers in considering the use of steel pipe is
the type of corrugation configuration to specify. There is an appreciable overlap of
sizes in the different corrugations, and the desirability of one over the other is usually
not apparent. This problem is overcome by using the gauge table arrangement shown in
Figure 2 and by allowing the contractor to supply any of the structural equivalents for
a given size where alternates exist. Economies could be expected with this procedure.
Supplying alternates by this method would not be so simple, however, where hydraulic
analysis for a given installation reveals the need for different pipe sizes for each cor-
rugation.
Installation requirements and procedures for flexible pipe are also presented on a
standard drawing that is automatically made part of the plans. This drawing shows
only ordinary bedding including treatment of an unstable soil foundation and rock under-
cut as well as a trench (sheeted or open cut) and two embankment methods of installation
similar to those for rigid pipe. Backfill and compaction criteria are also similar to
rigid pipe standards.
 63

Nonroutine Designs
Large pipe culverts of commonly used cross sections under very shallow fills, al-
though designed routinely from allowable fill-height tables, should not be treated as
routine installations. Their satisfactory performance under heights of cover between
2 and about 6 ft is credited in large measure to the installation procedures used rather
than the design methods, which are even less applicable to this condition. Every known
failure experienced on New York highway projects has been attributable to heavy con-
struction loads on pipes with inadequate cover and other substandard installation prac-
tices. For this reason, structures in this design situation should be installed in a very
closely controlled manner with special provisions made to protect the pipe from con-
struction loads. Warning the contractor is not necessarily sufficient, and consideration
should therefore be given to providing temporary ramping details, detours, or other
possible protective measures on the plans.
Special shapes, heavier live loads, or large sizes not covered on the fill-height tables
of course require individual attention. The most representative of the alternate design
procedures available for this situation should be used either for the design or as a check
of the design by the more routine methods. Another publication (2) discusses alternate
design methods and provides a basis with references for approaching this problem.
Pipes of all sizes and cross sections under very high fills not provided for on stan-
dard design tables are occasionally required. Rigid pipes of standard wall-strength
design for this case are seldom used because of questions concerning their ability to
resist high induced wall stresses and their compatibility with other devices such as
the imperfect trench and concrete cradle under very high fills. Davis, Bacher, and
Obermuller (7), for example, noted serious rupture of a concrete pipe on a 60-deg
cradle at a stage where adjacent sections installed under the same fill height, but with-
out a cradle, were not nearly so distressed. In addition, concern over possible detri-
mental long-term redistribution of loading effects with the imperfect trench method
suggests limitations on its use. Consequently, these cases require a thoroughly ana-
lyzed approach, including a possible special design of the pipe itself as well as special
installation details. The requirements of a special design for rigid pipe, however, are
often too complex and unfamiliar to the average designer, and an alternate such as a
box culvert (which is usually overdesigned) or a flexible pipe installation is often
seiecteci.
A number of large flexible pipes under very high fills (up to about 140 ft) have been
used without incident in New York State. These have been generally designed or re-
viewed on a project-by-project basis by a central unit that follows the same procedures
for structural design as were discussed previously with added emphasis given to soil
mechanics considerations.
Special designs should employ engineering judgment in evaluating the anticipated
performance of the structure. For example, the construction of a 22-ft span semi-
circular structural plate arch with abutments founded on rock beneath about 60 ft of fill
was recently analyzed. The design analysis provided marginally satisfactory results,
but there was concern as to whether a semicircular structure restrained at the abut-
ments could deflect laterally and mobilize adequate passive soil resistance. This was
resolved by using either of two alternates, a horseshoe-shaped arch or a round pipe
on a prepared bed.
Adverse foundation conditions that present stability and settlement problems for a
proposed culvert must of course be resolved prior to installation if problems affecting
cost and performance as well as failures are to be avoided. The evaluation and reso-
lution of these problems are based on an analysis of subsurface conditions, which
should be part of any important structure design. Such problems are varied and range
from removal and backfill of unstable soils to settlement analysis predictions for other-
wise stable foundations. Settlement analysis provides the basis for camber recommen-
dations and reveals the possible need for special joints for rigid types where definite
tendencies exist for the fill to spread. Standard-type joints used with rigid pipes on
steep gradients beneath high fills are also subject to opening, and this possibility
should be reviewed. All of the foundation problems that may be encountered cannot
be summarized here but can be identified if reviewed by a soils engineer.
64

Existing pipe culverts are frequently crossed by new highway embankments, which
presents a most perplexing problem to the designer. Most of these situations occur
with existing reinforced concrete sewer lines, and frequently insufficient data are
obtainable on the pipe strength class, bedding type, installation method, and backfill
materials, especially if the installation is not recent. This type of analysis, therefore,
most often involves many assumptions-beginning with a determination of how much
cover could be supported with the known or conservatively assumed pipe strength with
a positive projecting (or wide trench) condition and comparing this to the new height
of cover. The analysis can go anywhere from this point, but some of the directions
it should take are as follows.
1. A determination of the condition of the existing pipe is necessary to see if it is
worth saving or to provide a basis for adjusting the assumed or known strength of the
pipe. An internal inspection by trained personnel or a television survey may be suffi-
cient if the pipe cannot be occupied.
2. If the approximate analysis indicates that the pipe probably cannot support the
new load, some of the possible alternatives, short of replacing the line, include the
use of lightweight fill such as expanded shale that has a density of about 70 pcf or
cinders at 80 pcf, the imperfect trench used successfully by Spangler for this condition
(8), a combination of one or more of the preceding plus lowering the gradeline if feasi-
ble, and a protective structural relieving platform (which is usually not economical).
If these or other methods do not provide the desired degree of confidence for an
existing pipe in good condition and if the cost of replacement is significant, it may be
worthwhile to use load-reducing methods and proceed as planned but monitor the
immediate and extended performance of the pipe and assume responsibility for repairs
or replacement as necessary. A failure means different things to different people;
but in the case of a structurally sound concrete pipe handled in this way, the worst that
could be expected to occur where the computed overstress is not unreasonable would
be some vertical deformation, crack patterns, and spalling of concrete with possibly
some steel exposed. Therefore, repairs would be only a fraction of the total replace-
ment cost. Caution should be exercised before adopting this approach because other
factors such as settlement and a reduced diameter, if repairs become necessary,
affect the hydraulic capacity. In addition, a period of responsibility must be assumed.
This approach, however, has been used successfully on at least one New York State
project with large savings and possibly on others.

RESEARCH OBJECTIVES
Current Methods
An attempt has been made here to illustrate the utilization and workability of ac-
cepted design methods, particularly when coordinated with construction practice.
Deficiencies associated with these design methods were also discussed, and improve-
ments in this area would be of value to the practicing designer. The most notable of
these are believed to be the following.
1. A substitute for the 3-edge bearing test that relates more to field conditions
would be very desirable and is long overdue.
2. A review and evaluation of current and possibly new bedding methods should be
undertaken for rigid pipe. This should include an appraisal of any limitations of con-
crete cradle bedding as well as consideration of possible benefits from the use of
yielding types. The results of a study on this topic should include use criteria that,
if of value, can be economically and routinely incorporated into design.
3. A similar review and evaluation of the imperfect trench method is believed to be
important because it is so widely used. Its apparent success has been established, but
a more uniform methodology is needed to define the best materials to be used, their
locations and limits with respect to the pipe, and their possible limitations. Even
though this method is usually successful, a few installations in New York exhibited
cracking patterns that, although not considered serious, were generally objectionable.
 65

4. Backpacking studies have shown this to be a promising device for both reducing
s tress and m:iking its distribution more unifv:rrr1. around a s tructu1~e. A u1ei.hudoiogy
is required for design, materials, and construction practices; it should be incorporated
into current design methods, or a new one should be used.
5. It appears that available research results provide an adequate basis for corre-
lating deflection design and field performance, either by recognizing that the E' values
existing in the field are much higher than those suggested for use in design manuals
(4, 5) or by using other design methods available . An opinion is offered, therefore,
that7:oo much attention and concern are frequently given to the subject of deflection for
properly installed flexible pipes.
6. A need exists to improve load analysis procedures for large pipes under very
shallow fills where live loads are the principal forces to be resisted. As stated pre-
viously, current design methods do not adequately account for this situation, and now
that structures are becoming ever larger, these design deficiencies are even more
critical.
7. The design of large flexible structures under very high fills (i.e., greater than
about 100 ft) is generally accomplished without problems although refinements that
would give the vertical load more accurately are desirable.
Rigid pipes under very high fills are not so easily designed because the nonuniform
peripheral pressures developed cause large induced wall bending moments that must
be resisted. Although backpacking and possibly different bedding methods might assist
in overcoming this problem, it would appear desirable to develop a "flexible" re-
inforced concrete pipe. Pre stressed concrete pipe might also offer a greater range of
large sizes as well as design advantages under very high fills, and its wider use is
suggested where economical.

New Methods
The need for new or improved methods of design has not been generated by a rash of
failures. The failures that have occurred are generally attrib utable to poor installation
practices. Advances are desired and necessary, therefore, to obtain better methods
that include soil-structure interaction concepts. This objective is well stated else-
where (2): "The ultimate objective of anv culvert dei:dgn pro~P.clnrP iR to r,rPrli,..t
stresses and deformations at any part of the system at any point in time for a given
fill height. ... " The remainder of this statement details the desired aspects of the
theory in a very thorough manner. An added recommendation, however, is that thiR
new method, to whatever extent achieved, be adaptable or reducible to relatively sim-
ple terms, graphs, tables, or computer solutions so that standard designs can be es-
tablished for routine installation situations. The widespread acceptance and adoption
of any such new method would thereby be ensured.
New design methods must also place greater emphasis on durability. New York State
is continuing its durability studies toward the objective that the design methods described
here will become more quantitative, based on the development of a reliable field test
program.

ACKNOWLEDGMENTS
The portion of this paper dealing with durability is based on the latest recommen-
dations from the New York State Department of Transportation Engineering Research
and Development Bureau. Sincere appreciation and thanks are expressed to Peter
Houghton for the durability study and I. F. Rizzuto for his assistance in this area.

REFERENCES
1. Concrete Pipe Design Manual. American Concrete Pipe Assn., Arlington, Virginia,
1970.
2. Krizek, R . J., Parmelee, R. A., Kay, J. N., and Elnaggar, H. A. Structural
Analysis and Design of Pipe Culverts. NCHRP Rept. 116, 1971, 155 pp.
66

3. AASHO Standard Specifications for Highway Bridges. American Association of

State Highway Officiels, Washington, D. C., 1969.
4. Handbook of Steel Drainage and Highway Construction Products. American Iron
and Steel Inst., New York, 1967.
5. Townsend, M. Corrugated Metal Pipe Culverts: Structural Design Criteria and
Recommended Installation Practices. U.S. Department of Transportation, June
1966.
6. Haviland, J. E., Bellair, P. J., and Morrell, V. D. Durability of Corrugated
Metal Culverts. Report for the New York State Department of Transportation,
1967.
7. Davis, R. E., Bacher, A. E., and Obermuller, J.C. Structural Behavior of a
Concrete Pipe Culvert. California Division of Highways, Res. Rept. R&D 4-71.
8. Spangler, M. G. A Practical Application of the Imperfect Ditch Method of Con-
struction. HRB Proc. Vol. 37, 1958, pp. 271-277.
DESIGN OF CIRCULAR SOIL-CULVERT SYSTEMS
F. Dwayne Nielson, Colorado State University; and
Natverial Dalal Statish, New Mexico State University

In predicting the performance of a soil-conduit system, soil properties

must be included in the design procedure. It is suggested that the relation
between the constrained modulus of elasticity and the dry density of a soil
and the solution from theory of elasticity can be used to develop the desired
design procedure for circular flexible pipe. The design chart is developed
by using the solutions derived from the theory of elasticity and from the re-
lation between constrained modulus of elasticity and dry density. The upper
boundary of the design chart was approximated from the full-scale experi-
mental results on corrugation pipe as reported by Watkins and Moser ~).
The ring deflection of 5 percent was used as a limiting deformation of pipe.
It is shown that the actual experimental results of Watkins and Moser ~)
fall in the range of values of a dimensionless parameter. Two examples
are presented to illustrate the use of the design chart.

• THE development of a design procedure that includes soil properties is a problem of

considerable importance in the design of soil-conduit systems. During the past 60 years,
experience tables and rational procedures for the structural design of culverts have been
developed. For circular pipes, two major rational procedures are commonly used: One
is concerned with limiting deformation, and the other is concerned with limiting the
compressive load in the conduit wall. The proposed design procedure considers both
deflection and loads.
Watkins and Moser~), using full-scale corrugated steel pipes, have made experi-
mental studies in which they evaluated the effect of soil properties on the soil-conduit
system.
A solution based on the theory of elasticity has been developed by Burns and Richard
(1) for predicting the soil-structure phenomenon.
The object of this paper is to develop a design procedure by using (a) the solutions
from the theory of elasticity to predict the performance of the soil-conduit system and
(b) the experimental results of Watkins and Moser~) to develop upper limits on the
design. Soil properties must be included in the design procedure for a flexible soil-
conduit system.

LITERATURE REVIEW
The design and the installation of a flexible pipe are based on empirical and experi-
mental data. Marston (1), Spangler (l), and White and Layer (fil have attempted to de-
velop theories for the soil-culvert system.
Marston (!) presented his load theory in 1913. The choice of an abstract parameter,
settlement ratio, is important because it includes the effects of the deformation of
the fill, culvert, and natural ground under the culvert.
The pressure distribution, as shown in Figure 1, was used by Spangler in 1942 to
derive the Iowa formula, by which the horizontal deflection of a flexible conduit can be
computed. The pressure at the horizontal diameter is given by

Sponsored by Committee on Subsurface Soil-Structure Interaction.

67
68

(1)

where
Ph = maximum unit pressure at the extreme horizontal diameter,
E' = modulus of soil reaction of the side fill material,
t.X = horizontal deflection of pipe, and
D = normal diameter of the pipe.
The Iowa formula is
(2)

in which
t.X = horizontal deflection of the conduit in inches;
D1 = deflection lag factor-suggested values normally range from 1.25 to 1.50 for
design purposes;
K = bedding constant whose values depend on the bedding angle (0.083 for 90-deg
bedding);
We = vertical load per unit length of the pipe in pounds per linear inch;
R = mean radius of the pipe in inches;
E == Young's modulus of elasticity of pipe in pounds per square inch; and
I = moment of inertia of pipe wall.

ELASTICITY SOLUTION
The plane strain solution for an elastic cylindrical shell in an infinite linearly elastic
medium subjected to a uniformly distributed overburden pressure was presented by
Burns and Richard (J) . Equations are presented (Fig. 2 shows the notation) for the
following two cases:
1. In the bonded shell case, the shear stress exists at the interface between the shell
and the medium. For the bonded shell, the interaction pressures are

P = Pa[B(l - a!) - C(l - 3a! - 4bt ) cos 20] (3)

The shell displacement is

U = 1/2 (PoR/M.) {[1 + (B/C) at] - [1 + a! + (2/B) bfl cos 28} (4)

The shell thrust and moment are

Ne = PaR[B(l - ab) + C(l + at) cos 28] (5)

and

M9 = PoR 2 [(CUF/ 6VF) (1 - ab)+ ½ C(l - a! - 2bn cos 20] (6)

2. In the unbonded shell case, shear stress is zero at the interface between shell
and medium. For the unbonded shell, the interaction pressures are

(7)

The shell displacement is

U = ½(PoR/M,) {1 + [(B/C)ao] - [1 - a 2 + (2/B)b2 ] cos 20} (8)

and the shell thrust and moment are

(9)
 69

and

(10)

where the soil parameters are

2
M, = constra1nedmodulusof elasticity , FL- ;1.e., M. =[E*(l - µ*)/(1 + µ*)(1- 2µ*)];
K* = lateral pressure coefficient(-) ; i.e., K* = µ*/(1 - µ*);
E* = Young's modulus of elasticity , FL- 2 ; and
µ* = Poisson's ratio (-).
The shell parameters (culvert) are as follows:
R = mean radius, L;
D = diameter, L;
E 1 = plane strain Young's modulus, FL- 2 ;
A = area of shell wall per unit longitudinal length, L; and
I = moment of inertia of shell wall per unit longitudinal length, L 3 •
The load parameters are as follows:
B = ½(1 + K*), uniform pressure component;
C = ½(1 - K*), variational pressure component;
UF = M.DB/E 1 A, relative extensional flexibility; and
VF = M.D 3 2C/48E 1I, relative bending flexibility.
The dimensionless constants for the bonded shell are as follows:
a6 (UF - 1)/[UF + (B/C)] ,
a6 [C(l - UF)VF - (C/B)UF + 2B]/{(1 + B)VF + C[VF + (1/B)]UF + 2(1 + C)},
and
bt [(B + CUF)VF - 2B]/{(1 + B)VF + C[VF + (1/B)]UF + 2(1 + C)}.
The dimensionless constants for the unbonded shells are as follows:
ao (UF - 1)/[UF + (B/C)]
a2 = [(2VF - 1) + (1/B)]/[(2VF - 1) + (3/B)]' and
b 2 = (2VF - l)/[(2VF - 1) + (3/B)].
Watkins and Moser~), using corrugated steel pipes, performed interrelationship
studies between the properties of the pipe and the soil. The studies evaluated the sig-
nificant factors affecting the soil-culvert system. A test cell was constructed to test
a 5-ft diameter pipe. Hydraulic cylinders were used for loading. The results from
the studies indicated that the soil compression and pipe wall-crushing strength were
the important factors. Soil density was the main factor in predicting the soil modulus
and also proved to be the most important factor in predicting the performance of buried
corrugated steel pipes.

Constrained Modulus of Elasticity

Nielson (]_) established the relation between the modulus of soil reaction and con-
strained modulus of elasticity [E' = f(M.)]. In establishing this approximate relation,
he used the bonded shell equations of Burns and Richard's solution CT). Nielson con-
cluded that modulus of soil reaction can be approximated by

E' = 1.5 M. (11)

The relations for the unbonded equations of Burns and Richard's solution (1) are
as follows.
Using Eqs. 7 and 8 at 0 = 0 deg

(12)
70

and
(13)

Substituting Eqs. 12 and 13 into Eq. 1 gives

We used a computer to evaluate Eq. 14 for modulus of soil reaction. It can be concluded
that the modulus of soil reaction for the full slippage case can be approximated by

E' = 0 .70 M . (15)

It was shown by Stankowski and Nielson (!Q) , who used a model, that the actual mod-
ulus of soil reaction of a soil-culvert system is between the predicted modulus of soil
reaction by bonded shell and the unbonded shell equations of Burns and Richard 's (1)
solution. Therefore, the relation between the modulus of soil reaction and the con-
strained modulus of elasticity is assumed as

E' = 0.8 M. (16)

Constrained Modulus Versus Dry Density

Hsieh (1!) has determined the relation between the modulus of soil reaction and dry
density. He determined the modulus of soil reaction by using the Modpares device ~.
The experimental data of Hsieh (!l) are given in Table 1 for soils at optimum moisture
content.
The modulus of soil reaction is related to the constrained modulus of elasticity by
Eq. 16.
Watkins and Moser (1) reported their results in graphic form, relating percentage
of ring deflection to vertical soil pressure for a particular gauge, diameter, average
density, and percentage of standard density (using AASHO T 99 compaction test) .
To obtain M. values for the experimental results of Watkins and Moser (fil, we de-
veloped the relation of the constrained modulus and the ratio of dry density to the dry
density as det ermined by using t he AASHO T 99 compaction te st [ M, = f(yd/y99)] . We
used the values given in Table 1. The constrained modulus was obtained by applying
Eq. 16 to the data given in Table 1 and is shown in Figure 3.
Figure 3 also shows the correlation used to obtain a constrained modulus from
Watkins and Moser's soil density information. The relation of constrained modulus
of elasticity and ratio of dry density to the dry density as obtained by using the AASHO
T 99 compaction test may be significantly in error as shown by the scatter of data.
Changes in soil moisture content are not considered in this correlation. Other means
of determining M., such as the confined compression test, are superior to this method.

Similitude Analysis
The deflection of a flexible conduit in a fill that is long in relation to the diameter
of the pipe can be considered to be a function of the following primary quantities if the
pipe wall stress is below the yield point.
Primary Quantity Basic Dimension
i:lX = increase in horizontal diameter L (length)
D = original conduit diameter L
EI = conduit wall stiffness per unit length of conduit FL (force x length)
P = vertical pressure on conduit FL- 2
M, = constrained modulus of elasticity FL- 2
There are five primary quantities and two dimensions; therefore, according to Buck-
ingham's pi-theorem, the number of pi-terms should be 5 - 2 = 3.
 71

These three pi-terms may be formulated as follows:

1. 1r1 = aX/D, deflection term;
2. 1r2 = PD 3 /EI, load term; and
3. 1r3 = M,D 3 /EI, similar to soil term.
The pi-terms can be arranged into a functional equation form as follows:

Stankowski and Nielson (1Q) showed, by experimental analytic studies of underground

structural cylinders, that the horizontal diameter change of the conduit can be approxi-
mated by taking the average value of the bonded and unbonded equations, Eqs. 4 and 8
respectively, as predicted by Burns and Richard's solution (1).
The load-deflection diagram (Fig. 4) is obtained by plotting the load term versus the
ratio of average value of ax, Eqs. 4 and 8, divided by the original diameter. The di-
mensionless soil parameter is also plotted on the load-deflection diagram. The value
of Poisson's ratio of 0.25 is used for soil (.Q), and a Poisson's ratio of 0.33 is used
for the conduit material.
The experimental results of Watkins and Moser (9) at failure (Tables 2 and 3) show
varied failure condition. The failure criterion is considered as a percentage of de-
flection where one of the following conditions was first noticed: buckling of the pipe,
wall-crushing, or inverting of bottom of the pipe. The test data at failure are plotted
on the load-deflection diagram.
It can be seen from Figure 4 and Tables 2 and 3 that Watkins and Moser's experi-
mental data, at a failure condition, are generally within range of the M,D 3 /EI value
calculated from the theory of elasticity (once allowances are made for the error in-
volved in the determination of M,). The buckling data show some scattering. The
incipient failure line shown in Figure 4 is an approximate boundary that delineates the
failure zone of the conduit. The seam failure condition of the conduit in Test 19 falls
in the "safe" zone.
The design chart (Fig. 5) is developed from the load-deflection diagram by limiting
the deformation of the conduit to 5 percent of the original diameter. The incipient
failure line as shown in Figure 4 is used as the upper limit of the design chart. Fig-
ure 5 can be used for designing the soil-conduit system. The upper boundary of the
design chart may have to be adjusted as more failure information becomes available.

DESIGN PROCEDURE
The general form of a design procedure is as follows:
1. Select the diameter of the conduit according to the drainage requirement.
2. Determine the constrained modulus of elasticity by using an appropriate soil
test. If no other alternative is available, find the appropriate ratio of dry density of
soil to the dry density obtained by using AASHO T 99 standard compaction test and
apply it to Figure 3.
3. Select trial gauge of the pipe.
4. Determine the load parameter PD 3 /EI and the soil parameter M.D 3/ EI. Values
for D3/EI are given in Tables 4 and 5 for different corrugations and gauges.
5. M,D 3/EI and PD 3/EI values should be in the range of the design chart (Fig. 5).
If these values are out of range, then change section or gauge until it falls within the
range of the chart.
6. Enter in the design chart the PD 3/EI and M,D 3/EI values, and find the percentage
of ring deflection.
7. Check the ring deflection with allowable ring deflection (5 percent).
8. If the ring deflection is less, select the appropriate gauge or repeat with a
lighter gauge if feasible.
9. If the ring deflection is greater, revise the pipe gauge and repeat steps 3, 4,
5, 6, 7, 8, and 9.
 ----
Figure 1. Pressure pattern around flexible pipe. figure 2. Solution notation for buried pipe.

__
v_
SIH A

Table 1. Properties of soil samples.

Gradation
')'190
(lb/ft')
.
,,(lb/ft') ,,,
(lb/ft')
E(
2
(lb/in. )
,,Yf/;:--Y/ A'Yl&Yl&Yi&v/&YINIL'YI/

Sand 118.2 112.3 112.30" 2,105.4

115.25' 14,041.5 Figure 3. Correlation of Ms and percentage of standard
95.50' 1,474.3 compaction.
Mix 1 134.5 132.1 132.10" 14,287.4 110
133.30' 15,418.9
112.30' 1,711.1 ----
z
0
)<
""'~- )<)<
Mix 2 131.2 127.6 127.60"
129.40'
108.40'
12,313.1
14,389.3
1,866.3
."'
~ too
0
xx" X lf<fx

u
CORRELATIO~I USED
Mix 3 132. 5 128.0 120.00· 16,953.6 TO 0€TE ,"!!.ilN~
0
130.25' 24,157 .4 a: 90 CONSTl-iA1r-.;so t.mDULUS FOR

.""
108.00' WATKINS MI.O MOSERS DATA
1,883.1 0
z
Mix 4 126.7 117.5 117.50" 1,827. 5

..
U)

122.10• 10,993.5 80
98.10' 815.4 z I
Mix 5 133.0 123.9 123.90"
128.45'
3,766.0
9,734.5
."
a:
Q.
70
I
I
I
I

105.00' 1,212.1
I
Mix 6 130.5 121.8 121.80" 3,070.6 I
126.15' 5,273.1 60
I
103.30' 984.1 100 1000 1000 0

Mix 7 127.4 122.4 122.40" 2,475.9 CONSTRAINED MODULUS OF ELASTICITY (M

5
) PSI
124.90' 13,215.6
104.00' 1,179.0
Figure 4. Data plot on dimensionless
Mix 8 126.9 120.2 120.20" 15,015. 6
chart.
123.55' 28,678.1
102.20' 1,160.2

8
'Y99 • bh1eo + 'Yeg)/2. co.as 'Y99• 500

100
Table 2. Experimental results Vertical Ring Dry Standard
obtained by Watkins and Test Diameter Pressure Deflection Density Density
Moser. Number (in. ) (lb/ft') (percent) (lb/ft') (percent)

6 46.0 8,400.0 8.0 84.6 73 .0

12 60.0 8,600.0 3.2 94,0 77.0
17 60.0 6,000.0 17.2 84 . 6 69.0
18 48.0 14,000.0 7.0 91.2 75.0
19 36.0 13,600.0 5.0 97.3 80.0
21 36.0 4,200.0 10.0 74.9 61.0
52 60.0 12,800.0 1.20 101.7 83.0
61 36.0 14,000.0 1.40 120.1 93.1
62 60.0 14,000.0 0.60 129.0 106.0
63 48.0 4,200.0 0.20 119.7 98.0
68 60.0 12,000.0 0.40 118.4 97.0

Table 3. Data obtained by Diameter

Watkins and Moser. Test of Gauge Corrugation
Number (in.) (in.) Type of Failure PD'/EI M, D'/EI

6 16 3 by l Buckling 25.7 361

12 18 3 by 1 Wall-crushing 64.6 1,135
17 16 3 by I Buckling 35.8 585
18 16 3 by l 'Inversion in top 42.8 405
19 16 3 by I Buckling, seam failure 17.5 2,530
21 18 3 by 1 Sides buckling 6.81 118
22 16 3 by l Sides buckling 3.61 98
52 18 2il by ½ Sharp inversion, wall-crushing 414.0 6,760
61 16 21;' by½ Inversion rippling 78.2 1,820
62 18 2j1 by½ Wall-crushing 453.0 201,285
63
68
18
18
2i: by ½
21/, by½
Bottom inverting
Bottom inversion
69.5
388.0
7,370
13,420

Figure 5. Proposed design T

chart for circular corrugated

steel culverts.

.
.
,g

0 .0 1 .02 .oa .04 .OIi

6X/O

Table 4. Values of D3/EI Diameter (in.)

2 1
(inNlb) for 2 / 3 - by 3 /2 -in.
corrugations. Gauge 36 48 54 60 66 72 78 84

20 1.430 3.390 4.826 6.621 8.812 11.440 14.546 18.167

18 1.072 2.542 3.620 4.970 6.610 8.589 10.909 13.625
16 0.850 2.016 2.870 3.937 5.241 6.804 8.650 10.804
14 0.672 1.594 2.270 3.114 4.145 5.381 6.842 8.546
12 0.470 1.113 1.585 2.175 2. 894 3.758 4.778 5.967
10 0.355 0.8412 1.198 1. 643 2.187 2.839 3.610 4.508
8 0.281 0.6661 0.9484 1.301 1. 732 2.248 2. 858 3.570

Note: E = 29 x 106 psi; moment of inertia values taken from AISI Handbook, 1967.

Table.5. Values of D3/EI Diameter (in.)

(inNlb) for 3- by 1-in.
corrugations. Gauge 36 48 54 60 66 72 78 84

20 0.3123 0.7405 1.054 1.446 1.925 2.499 3.177 3.969

18 0.2334 0.5534 0. 7879 1.081 1.438 1.868 2.874 2.966
16 0.1858 0.4404 0.6271 0.8602 1.145 1.486 1.900 2.360
14 0.1478 0.3504 0.4989 0.6844 0.9109 1.183 1.504 1.878
12 0.1040 0.2467 0.3512 0.4818 0.6413 0.8326 1.058 1.322
10 0.0797 0.1890 0.2691 0.3691 0.4914 0.6379 0.8111 1.013
8 0.0641 0.1520 0.2165 0.2969 0.3952 0.5131 0.6524 0.8148

Note: E = 29 x 10 6 psi; moment of inertia values taken from AISI Handbook. 1967.
74

To illustrate this design procedure and the use of Figures 3 and 5, we present two
examples.
In example 1, the following are given: 48-in. diameter pipe, 22/2- by ½-in. corruga-
tion, height of cover is 10 ft, dry density of soil = 110.4 pcf, and dry density of soil
using AASHO T 99 compaction test= 120.0 pcf.
The following steps are taken to solve example 1.
1. To find the constrained modulus of elasticity, first determine the ratio of dry
density soil to dry density obtained by using AASHO T 99 compaction test:

-,,d/y 99 = 110.4/ 120.0 = 0.920

Then apply it to Figure 3:

M, 2,000 lb/ in.2

2. Try a 12-gauge pipe.

3. Derive from Table 4 the following:
D3/EI 1.113 in.2/lb,
M.D 3/EI 2,000 x 1.113 = 2,226,
P YH, and
PD 3/ EI (110.4 lb/ ft 3 x 10 ft) / (144 in.3/ft2) X 1.113 = 8.05.
4. Derive from Figure 5 the following: ~X/D = 0.5 percent. Change in horizontal
diameter is derived as follows : ~X = 48 x 0.005 = 0.24 in. With a factor of safety of 2.0,
PD 3/EI = 17.10 .
5. Derive from Figure 5 the following: ~X/ D = 1.8 percent. Change in horizontal
diameter is derived as follows: ~X = 0.01 x 48 = 0.48 in. Witha factorof safety of 1.0,
the ring deflection is 0.24 , and, with a factorof safety of 2.0, the ring deflection is 0.48.
Ring deflection is less than allowable; therefore, 12-gauge pipe is overdesigned. Try
a 14-gauge pipe as the ring deflection; the PD 3/EI value for a 12-gauge pipe is well
less than the allowable deflection.
In example 2, the following are given: 60-in. diameter pipe, 30 ft of fill , dry density
of soil = 106.0 pcf, and dry density of soil using AASHO T 99 compaction test = 110.0 pcf.
The following steps are taken to solve example 2.
1. To find the constrained modulus of elasticity , determine the appropriate ratio:

yd/y99 = 106.0/ 116.0 = 0.962

Then apply it to Figure 3:

M, = 2,500 lb/in. 2

2. Try a 14-gauge pipe with 22/2 - by ½-in. corrugation.

3. Derive from Table 4 the following:
D3/ EI 3.114 in.2/lb,
M, D3/ EI 2,500 x 3.114 = 7,800 ,
P yH , and
PD 3/ EI (106.0 lb/ ft x 30 ft)/(144 in.2/ ft2) x 3.114 = 68.
4. Derive from Figure 5 the following: ~X/D = 1 percent. The PD 3/EI value is on
the upper limit of the design chart; therefore, try a 12-gauge pipe, which would provide
an additional factor of safety. The ring deflection is less than the allowable 5 percent.
Therefore, a 14-gauge pipe (2 ½ - by ½-in. corrugation) will carry the load with a factor
of safety of about 1.
 75

CONCLUSIONS
The design of soil-culvert systems is concerned basically with predicting the ulti-
mate percentage of conduit deflection and the factor of safety against conduit wall
failure. The soil property must be included in the design chart.
The following conclusions can be drawn from the design chart and the relation be-
tween constrained modulus and soil density.
1. The relation between constrained modulus of elasticity and soil density indicates
that, if the soil has a high dry unit weight, then the constrained modulus is high. If
the soil has a low dry unit weight, then the constrained modulus is low.
2. The load deflection diagram indicates that, if M.D 3/EI is high, then the ultimate
deflection of conduit is low, but the load-carrying capacity of the soil-conduit system is
high. If M,D 3/EI is low, the ultimate deflection is high, but the load-carrying capacity
of the soil-conduit system is low.
3. A comparison of Watkins and Moser's (]) results with the design chart (Fig. 5)
and the error in evaluating the constrained modulus from yd/y99 indicate that an ap-
propriate factor of safety is required.

RECOMMENDATIONS
In developing the design procedure, it is found that additional studies in several
areas should prove valuable. They are as follows:
1. The design procedure as developed in this paper should be verified on field in-
stallations.
2. Investigation of the relation of the constrained modulus to other standard soil
tests is necessary. This should include well-known tests such as the California bear-
ing ratio test and the Hveem stabilometer test.
3. The experimental results of Watkins and Moser (]), as plotted on the load de-
flection diagram, indicate that the incipient failure might vary with the radius of
gyration of the pipe. The relation between radius of gyration and the failure line
should be investigated.

REFERENCES
1. Marston, A. The Theory of External Loads on Closed Conduits in the Light of
the Latest Experiments. Eng. Exp. Sta., Iowa State Univ., Ames, Bull. 96, 1930.
2. Spangler, M. G. The Structural Design of Flexible Pipe Culverts. Eng. Exp.
Sta., Iowa State Univ., Ames, Bull. 153, 1942.
3. Barnard, R. E. Design and Deflection Control of Buried Steel Pipe Supporting
Earth Loads and Live Loads. Paper presented at 60th Annual Meeting, ASTM,
June 17, 1957.
4. Watkins, R. D., and Nielson, F. D. Development and Use of the Modpares De-
vice in Predicting the Deflection of Flexible Conduits Embedded in Soil. Jour.
Pipeline Division, Proc. ASCE, Vol. 90, Proc. Paper 3782, Jan. 1964.
5. Nielson, F. D., Bhandhausavee, C., and Yeb, K. Determination of Modulus of
Soil Reaction From Standard Soil Tests. Highway Research Record 284, 1969,
pp. 1-12.
6. White, H. L., and Layer, J. P. The Corrugated Metal Conduit as a Compression
Ring. HRB Proc., Vol. 39, 1960, pp. 389-397.
7. Burns, J. Q., and Richard, R. M. Attenuation of Stresses for Buried Cylinders.
Proc., Symp. on Soil-Structure Interaction, Univ. of Arizona, Tucson, Sept. 1964,
pp. 378-393.
8. Nielson, F. D. Modulus of Soil Reaction as Determined From Triaxial Shear Test.
Highway Research Record 185, 1967, pp. 80-90.
9. Watkins, R. K., and Moser, A. P. The Structural Performance of Buried Corru-
gated Steel Pipes. Eng. Exp. Sta., utah State Univ., Logan, Sept. 1969.
10. Stankowski, S., and Nielson, F. D. An Analytical-Experimental Study of Under-
76

ground structural Cylinder Systems. Eng. Exp. sta., New Mexico state Univ.,
Las Cruces, Oct. 1969.
11. Hsieh, H. Determination of Modulus of Soil Reaction From standard Soil Prop-
erties. New Mexico State Univ., Las Cruces, Master's thesis, May 1968.
12. Manual of steel Construction. American Institute of steel Construction, Inc.,
New York, 1967.
BUCKLING OF CYLINDERS IN A CONFINING MEDIUM
C. V. Chelapati, California State College at Long Beach; and
J. R. Allgood, U.S. Naval Civil Engineering Laboratory,
Port Hueneme, California

A large-deflection theory is presented for defining the critical radial pres-

sures that result in the buckling of elastically supported homogeneous
cylinders with simple ends. Solutions are given for radial support around
the extrados and for radial support only on the outward acting lobes of the
circumferential waves that form on loading. It is shown that the difference
between the energy load and the Euler load is a maximum for cylinders
with no radial support and that the energy load approaches the Euler load
as the length-to-radius ratio and the foundation coefficient decrease and as
the radius-to-thickness ratio increases. Critical pressures are greater
for cylinders supported around the entire extrados than for cylinders
supported only on the outward displacing lobes. Practical utility of the
theory is found in its application to conduits and cylindrical protective
structures buried in soil fields. For such applications, the equation defin-
ing the lower bound of buckling is modified to permit one to account for the
influence of the surface boundary and to include soil parameters that are
readily obtainable in the laboratory. A few tests were performed to check
the theory. The results of these tests and other available data agree rea-
sonably well with the theory.

•SOLUTIONS to the problem of defining the strength of conduits, tunnel liners, and
cylindrical shelters buried in soil depend on the ability of engineers to predict buckling
loads. The general purpose of this paper is to provide the needed solutions. Specific
objectives were to define and compare the Euler and energy loads, study the effect of
all-around support as compared to support only on outward-deflecting lobes, determine
the influence of the dominant parameter on buckling resistance, and compare the results
of the elastic theory solutions with experimental results. These goals were pursued to
provide improved criteria and methods for the design of buried cylinders.
The configuration analyzed (Fig. 1) consists of a radially supported cylinder with a
bonded interface and simple ends. The cylinder is subjected to uniform radial pres-
sure. Two types of radial support are considered: elastic support around the com-
plete perimeter and elastic support only on the outward-deflecting portions of the lobes
that form during loading. Parallel solutions are developed from the theory for both of
these cases.
In studying the buckling of shells, von Karman and Tsein (1) showed that some sys-
tems have states of stable equilibrium at loads less than the Euler load. These states
of equilibrium are separated by "energy barriers" such that work must be expended to
pass from the unbuckled to the buckled configuration. Such work might be supplied by
perturbations in the load or by other means; consequently, the critical load may be less
than the Euler load as shown in Figure 2.
Subsequent to the work of von Karman and Tsein, Friedrichs (2) pointed out that at
state B' (Fig. 2) the potential energy, V, is greater than at B; consequently, buckling is
not likely at a load less than Pm where the potential energy in the two states is equal.
The latter condition is indicated qualitatively by points D and D' (Fig. 2). The load Pm
will be referred to as the transitional buckling load or the energy load.

Sponsored by Committee on Subsurface Soil-Structure Interaction.

77
78
A number of investigators, the earliest of whom was Link (3), have developed plane
strain solutions for elastically supported cylinders by using the classical small-
deformation theory. Later, Luscher endeavored to fit such relations to data from
tests of buried cylinders (!).
Langl1aar and Boresi (5) derived an expression for the energy load for cylinders
subjected to hydrostatic loading and found that the energy to cause a transition from
the unbuckled to the buckled state is small and could easily be supplied by accidental
shocks. Their work was subsequently compared with the results of experiments by
Kirstein and Wenk (6) who observed both elastic snap-through and recovery.
Forrestal and Herrmann (7) showed that the Poisson's ratio of a confining medium has
less than a 15 percent effect on the critical load but that the presence or absence of inter-
face shear has a relatively large effect for confining media with low Poisson's ratios.
In studies of other geometries, Gjelsvik and Bodner (8) found that the energy load
is relatively insensitive to the assumption (number of terms in the series) for the de-
flected shape.
In the following sections, the Euler and energy loads are investigated by modifying
the methodology of Langhaar and Boresi to include terms that contain the strain energy
of the support. The solutions are compared with the results of experiments on cyl-
inders buried in sand.

DEVELOPMENT OF BUCKLING EQUATIONS

The general procedure for the theoretical development is to express the strain
energy and the potential energy in terms of displacement by using the large-deflection
theory. Then the various energy components are summed to obtain the increment of
total potential energy, which is set to zero to obtain the critical pressures.

FORMULATION OF DISPLACEMENT RELATIONS

Rectangular and cylindrical coordinates for a cylinder are shown in Figure lb, where
the origin of coordinates is taken at the center of the section at midlength. Let u, v,
and w represent the axial, circumferential, and radial displacements respectively of
the point P due to buckling. The components of strain may be expressed in terms of
these displacement co ponents and the coordinates as ~)

(1)

In these expressions, the subscripts on displacements denote partial derivatives, fx is

the axial strain, €8 is the circumferential strain, and 'Yxe is the shearing strain. The
€t
strains at incipient buckling are denoted by €~ > and 0 >.
0

Because u, v, and w are unknown, the shape after buckling has to be assumed so that
we can compute the changes in total potential energy. They are assumed to be repre-
sented by the terms containing up to 3n0 of the Fourier series as

u Uo + u1 cos n0 u2 cos 2n0 + U3 cos 3n0 }

V v1 sin n0 + v2 sin 2n0 + V3 sin 3n9 (2)

w Wo + W1 cos n9 + W2 cos 2n9 + W3 cos 3n8

where u1 , v1 , and w1 are functions of x alone and n denotes the number of complete waves
around the perimeter.
If we also assume that the incremental circumferential strain, .ll.€ 8 , is equal to zero,
Eq. 1 yields
 79

(3)

By substituting the expressions for v and w in Eq. 2 in Eq. 3 and by regrouping terms,
we can show that all the coefficients v1 and w1 can be expressed in terms of w1 alone as
follows:

-(w1/n)
n[n - (1/n)] 2 wi/[2a(4n2 - (4)
V3 0

Wo = - {[n - (1/n) J'wD/4a }

W1 = W1 (5)
W2 - Un - (1/ n) J2 wD / [ 4l!,(4n2 - 1) J
W3 =0

On the basis of experimental data, it is assumed that

wi/a = [W 0 cos(7rx/L)J/fn - (1/n)J (6)

where Wo is a constant, n is the number of circumferential waves around the perimeter,

and L is the length of the cylinder. These displacement relations permit one to express
the components of the total potential energy in terms of W~ as exemplified in the follow-
ing section.

strain Energy Due to Radial Support

AU-Around Support-The change in strain energy due to radial support is given by

AU. = 2a
0
/2
0
L
2
/
211'
(k.w /2) d0dx (7)

where ~ is the coefficient of soil reaction in lb/in. 2/in. or other consistent units.
Substitution of Eqs. 2 and 6 in Eq. 7, integration, and elimination of dimensions by
EahL give

(8)

where
d1 = 7rn2/[4(n2 - 1)2], and
d2 = [37r(32n1 - 16n2 + 3)J/ [256(4n 2 - 1) 2]
Outward Lobe Support-In deriving Eq. 8, it was assumed that the radial support is
effective for both positive and negative values of the displacement w, the positive direc-
tion of w indicating the outward radial displacement.
With radial support only for positive values of the radial displacement, strain energy
is stored in regions of positive displacement, w, and is zero for portions with negative
displacement. It is assumed that the strain energy stored in the media is equal to the
strain energies because the components of the displacement ware imposed separately.
Thus, recognizing that W3 is zero by using Eq. 5, we may write the strain energy as
80

0/
2'1T
(wo/a) 2 d9 + n
0
J 2'1T
2
2n [(wi/a)cos n8] d9

+ 2n 2, / : : : [(w,/a) cos 2n9l' d9 dx (9)

8n

The first term does not contribute to AU. because wo is negative; therefore,

(10)

If we use Eqs. 4 and 6 (eliminating dimensions by EahL) and simplify, the change in
strain energy is

AU./EahL = ~a/[E(h/a) 3 ] (h/a) 2 (cliW! + d2W;) (11)

where
d1 = 1rn2 /[8(n2 - 1) 2 ), and
d2 = 31r/[512(4n2 - 1)2 J.

Other Components of Total Potential Energy

Expressions for the strain energy components, AU, due to membrane stresses and
bending stresses and the increment of potential energy, AO.,, due to external pressure
(as developed by Langhaar and Boresi, 5) are given below. AU due to membrane
stresses is as follows: -

(12)

where
b1 = f 1 + Ksr 2,
2
b2 = f2 - (K7 - fs) r + K1r4 - C/h,
bs = f4 + (Ks - fs) r 2 + (Ks + f6 ) r 4 - cp2 - g, and
r = a/L
The f1 's, 'Pi' s, and g1 's are functions of the ratio a/L and n, and the Kt' s are functions
of n only. Expressions defining these functions are given in the Appendix. AU due to
bending stresses is as follows:

(13)

where
C1 = 1r/[ 48(1 - v2 )] (n2 + 21r 2r 2 ( 1 + [v/ (n 2 - 1))} + [1r 4 n 2r 4 / (n 2 - 1) 2 )),
2
C 2 = 1r/[256(1 - v2 )] (n 2 + 1r 2 r 2 (1 + [v/(n - l)J} + ¾-rr 4 r 4 (2 + [1/(4n 2 - 1) 2 ]}), and

C 3 = 1r/[12,288(1 - 2
V )] (5n 2 + 21r 2 r 2 (5 - 17v + [8v/(4n 2 - 1)] J).
AO., due to external pressure is as follows:

(14)
 81

where
a1 1Tn 2/ [4(n 2 - 1)], and
a2 (37T/ 128) ( 1- {1/ [2(4n 2 - 1rn).

Increment of Total Potential Energy

The increment of total potential energy, ~V, which results from buckling, is given by

(15)

Substituting Eqs. 12, 13, 8, and 14 in Eq. 15 yields ~V for the case of all-around sup-
port as

(16)

where
A = p/[E(h/a) 3],
D = k,a/[E(h/a) 3J,
t = h/a,
B1 = b1 + C1t2 + Dd1t2,
B2 = b2 + C2t2 + Dd2t2, and
B3 = b3 + C3t2.
For the case of support on the outward acting lobes only, Eqs. 12, 13, 11, and 14 in
Eq. 15 yield the same relation as Eq. 16 except that d1 and d2 are replaced by d1 and d2
as given in Eq. 11.

Euler and Energy Loads

According to Tsein' s criterion, the increment of total potential energy from the
unbuckled to the buckled configuration is equal to zero, that is,

~v = o (17)

Therefore, by setting Eq. 16 to zero, the expression for the critical pressure coef-
ficient A= Por/ E(h/ a)3, is

(18)

Minimizing Eq. 18 with respect to W! gives

(19)

The smallest positive value of w; from Eq. 19 substituted in Eq. 18 gives the energy
load. Note that the root W! of Eq. 19 must be positive . If the roots are either negative
or imaginary, the energy load does not exist. In some cases, a positive root gives an
energy load greater than the corresponding Euler load. For such cases, the energy
load and Euler load merge, and the phenomenon of snap-through does not occur.
The Euler load can be obtained simply by setting w; = 0 in Eq. 18. Thus,

(20)

METHOD OF SOLUTION
Uniform Surrounding Field
Equation 18 is solved by calculating the critical pressure for a given set of indepen-
dent variables corresponding to several values of A (9). The minimum value is the true
critical pressure. A computer code was written to obtain solutions for A for permuta-
tions of the length-to-radius ratio, L/a, thickness-to-radius ratio, h/ a, and foundation
82

coefficient, fi. All calculations were based on a Poisson's ratio, v, of 0.33 . Plots of
the computer output are shown in Figures 3 and 4.
Differences between the Euler load and the energy load (Figs. 3 and 4) are largest
for small L/ a, large a/ h, and small D. The difference is as large as 25 percent in
some cases. The energy load approaches the Euler load as L/ a increases, a/ h de-
creases, and D increases. With short, thin cylinders, the Euler load at small values
of D is nearly the same for both types of support considered. The same is true for the
energy load. At large values of D, type of support is much more important as is evi-
dent from the following considerations. For infinitely long cylinders, it can be shown
that the Euler and the energy loads are identical and that for all-around support they
may be expressed by the relation

(21)

where
A = p0 ./[E(h/a}3],
D = k, a/[E (h/ af' J,
oc: = (n 2 - 1)/( 12 (1 - v 2 ) ], and
(3 = 1/(n2 - 1).
The lower bound for the critical pressure is

XEL = 2 'a: {3D = 0.6116 t'B for V = 0.33 (22)

Likewise, for infinitely long cylinders with lobar support (support only on the outward-
deflecting lobes)

XL 0 =o::+0.5/fD (23)

The lower bound for lobar support is given by

J\LO = ;2oc:(3D = 0.4325 tt for V = 0.33 (24)

From the ratio of Eqs . 22 and 24, the influence of type of soil support is

(25)

or about 40 percent difference.

The foundation coefficient, k., in Eq. 22 can be expressed in terms of the one-
dimensional (confined) compression modulus by using the theory for soil-surrounded
tubes (4). For concentrically surrounded tubes, the modulus of elastic support, k, = k,a,
may be- expressed in terms of the modulus of elasticity of the soil, E., by the relation
plotted in Figure 5. Further, E, is related to the confined compression modulus, M,,
by the expression

E, = CM, (26)

where C = [ (1 + v, )(1 - 2v,) ]/ (1 - v,).

From the theory of a soil-surrounded tube (4), it has been shown that the ratio of
k, to E, is -

B = k, / E, = [1 - (a/ ¾) 2 ] / ((1 + v,) [1 + (a/ a 0)

2
(1 - 2v,)J} (27)

where a and a are radii as shown in Figure 5.

0

It follows that

k,a BCM 0 (28)

 83

Figure 1. System geometry.

(a) Section {b) Coordinate system

Figure 2. Pressure, energy, and displacement interrelations.

Pu
0..

f pm
~
0..

Pl

Displacement, w Displacement, w

Figure 3. Variation of critical pressure with D for elastic support.

100
70 L/a 1, a/h = 500

40

.., 20

. .~1~ 10
7
I<
L/a 10, a/h = 500
4
Energy load
2 10, a/h = 100 Euler Load

I .__ _.__.._.._~_._........__ _ ..__.__.__._...LJu...o.L..- - ' ---'----'-_._.__._.....,

JO 20 40 70 100 200 400 700 1000 2000 4000 10000
k a
%
ii =
~3
Figure 4. Variation of critical pressure with D for lobar support.
100
70 1, a/h = 500

40

I<(
- - Energy load
- --- Euler Load

1
10 20 40 70 100 200 400 700 1000 2000 4000 10000

D a

Figure 5. Modulus of soil support for elastic ring (4).

(a - a)/a
0

1.5 0.67
4 0. 25
0.080 . - - - -~ - - - ~- - - -. - - - - -- - -~
---

8 0.40

0. 20

0
0 0.20 0.40 0.60 o. 80 1.0

Figure 6. Buckling data.

3
100 X 10 , -- -- - - - - -- - - , - - - - -- - - ----,--- - - - - - - , 1041

a NCEL Data

a Albritton 1s Data

C = 4.2
0

JO X
3
10 104.1 !
A

Note: Data corrected to d /D = I, and

0
v = 033. pe r = p (1 -A)

3
1 X J0 ' - - - -- - - - - -- - - ' ' - - - -- - -- - - - ' - -- -- - - - - ' 10.41
3
100 X 10 1000 X 103 M/,J, ---- 10,000 x 103
365 3,652 36,522

o-
 85

Substituting Eq. 27 in Eq. 22 gives

(29)

where
C = 6 fBC,
3
'11 = EI/ D
D = mean diameter of cylinder, and
EI = stiffness of section.
In the strict mathematical sense, Eq. 29 is correct only for concentrically surrounded
tubes. It is approximately correct, however, for cylinders buried beneath the earth's
surface, which are loaded by the soil cover and possibly a uniform surface loading.

THEORY VERSUS EXPERIMENT

To facilitate a valid comparison of theory and experiment, we performed tests on
thin metal cylinders in a segmented soil tank where the confined compression modulus,
M., and the at-rest coefficient of lateral earth pressure, k, could be determined as an
integral part of each test (10). All cylinders were 5 in. in diameter, 21 in. long, and
either 6, 12, or 18 mils thick. Sand type and density were varied to get a wide range
of confined compression moduli. A dry, rounded sand and a sharp-grained sand were
used, each at three different densities. Measurements included surface pressure,
cylinder deflections, and strains in the soil, tank, and cylinder. Details of the experi-
ment may be found elsewhere (10).
The experimental data are plotted together with Eq. 29 in Figure 6. The buckling
load was taken as Per = p, (1 - A) where Pr is the surface pressure at failure, and the
arching was calculated from the corresponding strains at the haunches. The soil
modulus used was the secant modulus at the failure load. All data plotted were cor-
rected to one-diameter depth of cover.
Representative data from Albritton' s tests (11) (SDAl, SDA2, SE 1, SE2, SFl, and
SF2) are also plotted in Figure 6. In these data, the moduli were from confined com-
pression data. No other data are known to the authors in which all the necessary pa-
rameters are available. As may be seen, however, the Naval Civil Engineering Lab-
oratory and the Albritton data agree reasonably well with the theory. There is, of
course, the expected spread common to buckling and soil test data.
In general, the experimental buckling loads are greater than those from Eq. 29.
This is considered to be due to greater compaction of the soil in the vicinity of the
cylinder than at corresponding depths in the free-field. As a consequence, elastic
support gives a better approximation of actual buckling loads than does lobar support.
Soil moduli back-calculated from deflections were in close agreement with mea-
sured values. This adds to confidence in the validity of the experimental data.
Predicting the surface and overburden pressure to cause buckling of a buried cyl-
inder involves a determination of the arching over the structure (12). The percentage
of the applied load that reaches the interface may vary by a factorof 20 or more de-
pending on the relative stiffness of the inclusion and the confining soil.

CONCLUSIONS
The theoretical analysis and the comparison of the theory with test data substantiate
the following conclusions:
1. The energy load and the Euler load are significantly different only for conditions
(low length-to-radii ratios, low values of the foundation coefficient, and large radii-
to-thickness ratios) that are not commonly encountered in practice. The theory agrees
reasonably well with applicable experimental results.
2. The critical load with elastic support around the total perimeter is considered
the best model of actual buried cylinders.
3. The buckling resistance of long buried cylinders is principally dependent on the
bending flexibility, the secant value of the confined compression modulus, and Poisson's
86

ratio of the confining media as defined by Eq. 29. Arching is extremely important in
governing the percentage of the applied load that reaches the interface.
4. The theory is considered adequate for obtaining conservative estimates of the
elastic buckling load of buried cylinders.

REFERENCES
1. Von Karman, T., and Tsein, H. S. The Buckling of Spherical Shells by External
Pressure. Jour. Aeronautical Sciences, Vol. 7, 1939, pp. 43-50.
2. Friedrichs, K. O. On the Minimum Buckling Load for Spherical Shells. California
Institute of Technology, Theodore von Karman Anniversary Volume, 1941, pp. 258-
272.
3. Link, H. Beitrag zum Knickproblem des elastich gebetten Kreisbogentragers.
Stahlbau, Vol. 23, No. 7, July 1963.
4. Luscher, U. Buckling of Soil-Surrounded Tubes. Jour. Soil Mechanics and Founda-
tions Div., Proc. ASCE, Vol. 92, No. SM6, Proc. Paper 4990, Nov. 1966.
5. Langhaar, H. L., and Boresi, A. P. Snap-Through and Post-Buckling Behavior of
Cylindrical Shells Under the Action of External Pressure. Eng. Exp. Sta., Univ.
of Illinois, Urbana, Bull. 443, 1957.
6. Kirstein, A. F., and Wenk, E., Jr. Observations of Snap-Through Action in Thin
Cylindrical Shells Under External Pressure. The David W. Taylor Model Basin,
Washington, D. C., Rept. 1062, Nov. 1956.
7. Forrestal, M. J., and Herrmann, G. Buckling of a Long Cylindrical Shell Sur-
rounded by an Elastic Medium. Presented at ASCE Struct. Eng. Conf., Preprint 108,
Oct. 1964.
8. Gjelsvik, A., and Bodner, S. R. Non-Symmetrical Snap-Buckling of Spherical Caps.
Jour. Eng. Mechanics Div., Proc. ASCE, Vol. 88, No. EM5, Oct. 1962.
9. Chelapati, C. V. Critical Pressures for Radially Supported Cylinders. U.S. Naval
Civil Eng. Laboratory, Port Hueneme, Calif., Tech. Note N-773, Jan. 1966.
10. Allgood, J. R., Ciani, J. B., and Lew, T. K. Influence of Soil Modulus on the Be-
havior of Cylinders Buried in Sand. U.S. Naval Civil Eng. Laboratory, Port
Hueneme, Calif., Tech. Rept. R- 582, June 1968.
11. Albritton, G. E. Behavior of Flexible Cylinders Buried in Sand Under Static and
Dynamic Loading, Waterways Exp, sta,; Vicksburg, Miss., Rept. 1-821, April
1968.
12. Allgood, J. R. Structures in Soil Under High Loads. Presented at the ASCE
structural Eng. Conf., Portland, Ore., April 6-10, 1970.

APPENDIX
EXPRESSIONS FOR VARIO US TERMS IN THE
THEORETICAL DEVE WPMENT

2 2 2
256(1 - U )(n - 1)

5
5ln

~+
2 2
TTS~ 5o + 7n -
2 2 2 2n~
( 4n - 1) 4n
 87

!T2(n2 + 1)
2 2
8(n - 1)

C 2 2
8(1 + u)(n - l)

TT:3
32(1 + II
) [
(4n
;n2
_ 1)
2 + 2 2n2 - 12 - l ) + 4~::
2(n - 1)(4n
~ i)~~

rr
512(1
3
+ 11)
[ 1
( 4n2 _ l)2
2
( 4n2 - 1)
+ JJ

K "' 2
9
16(n2 - 1) (4n - 1)

2
!T(l - 11)n(8n - 3)
2
32(4n - 1)

(1 - u)nn
2
2 (n - 1)

2 2
TT(l - u)'n· (2n + 1)
2 2
4(n - 1)(4n - 1)

. 4(n
2
- 1)

2
8(4n - 1)

y =
88

f =
1

f3 "'

f C
4

g C
THE IOWA DEFLECTION FORMULA: AN APPRAISAL
Richard A. Parmelee and Ross B. Corotis, Northwestern University

The Iowa deflection formula is based on the assumption that the supporting
strength of buried corrugated metal pipe installations arises through the
lateral reactive soil pressures induced at the sides of the pipe. Since its
introduction more than 30 years ago, the Iowa deflection formula has served
as a basic criterion for the design of buried flexible pipe systems. The
formula is based on an assumed distribution of loading around the pipe, and
contains three parameters that are empirical in nature. The present study
examines the significance and possible variation of these parameters in the
deflection equation. Limited field studies have indicated that realistic
variations in these parameters can lead to an almost threefold change in
design requirements. One of the most influential parameters in the for-
mula is the modulus E'. Values of E' as determined from the measured
response of full-scale installations exhibit a thirtyfold variation. Yet
these data form the basis for establishing recommended values of E' for
design use. Extrapolation of the observed deflections from field tests
shows that in most cases use of the suggested design criteria yields uncon-
servative fill heights. The present study shows no strong correlation be-
tween E' and percentage of standard Proctor density for the soil adjacent to
the pipe, the pipe diameter, or the ratio of fill height to pipe diameter. A
statistical analysis compares observed values of E' with common proba-
bility distributions, and a log-normal distribution is fitted. Probabilities
associated with various ranges of E' confirm that there is no rational basis
for recommending a design value that is much greater than the median, or
central, value.

•THE use of corrugated metal pipe was originally confined to small culverts in which
the size rarely exceeded 36 in. in diameter. The height of fill over the early culverts
was usually very low, too low in many cases to protect the pipe from damage by traffic
loads. As their use developed, corrugated metal pipes were made in larger diameters,
and the heights of the fills constructed over them were increased markedly.
During these early days, no successful attempt was made to develop a rational means
for designing this type of structure according to the principles of mechanics. Rather,
reliance was placed almost wholly on service experience and intuition in the determina-
tion of safe heights of fill.
The predominant source of supporting strength for a flexible pipe is the lateral pres-
sure of the soil at the sides of the pipe. The pipe itself has relatively little bending
strength, and a large part of its ability to support vertical loads must be derived from
the passive pressures induced as the sides move outward against the soil. The ability
of a flexible pipe to deform readily and thus utilize the passive pressure of the earth on
each side of the pipe is its principal distinguishing structural characteristic and ac-
counts for the fact that such a relatively lightweight pipe can support earth fills of con-
siderable height. Because so much of the total supporting strength depends on the side-
fill material, any attempt to analyze the structural behavior of this type of pipe under
a fill must consider the soil at the sides to be an integral part of the structural system.

Sponsored by Committee on Subsurface Structures Design.

89
90

IOWA DEFLECTION FORMULA

During experiments on circular flexible pipe at Iowa State College, Spangler ob-
served that, as soil loads are applied, the flexible ring is flattened down into an ap-
proximately elliptical cross section with a decrease in vertical diameter l::.y and an
increase in horizontal diameter Ax. In the now classic report (11) on the design of
flexible pipe design, Spangler noted, "Preliminary observationshad led to the hy-
pothesis that, as a fill is built over a flexible culvert, the horizontal pressure on the
pipe at any point bears a nearly constant relationship to the horizontal movement of the
point. This constant ratio has been called the modulus of passive pressure, e, of the
fill material and may be expressed as units of pressure per unit of movement."
The loci of horizontal movement of all points within the top and bottom 40 deg of the
pipe ring are assumed to be small, and for mathematical convenience a simple para-
bolic curve embracing only the middle 100 deg of the semicircle is used to define the
distribution of horizontal pressure.
Next, it was necessary to make some assumptions concerning the distribution of
pressure around the remaining portions of the pipe. The Marston load theory (6) was
used to evaluate the total vertical load on the pipe, Wo• For flexible pipe, Spangler as-
sumed that this total load is uniformly distributed across the breadth of the conduit.
The vertical reaction on the bottom of the pipe is equal to the vertical load, We, and
is distributed uniformly over the width of bedding of the pipe. The half-width of this
bedding and vertical reaction is defined in terms of the angle a as measured from the
vertical centerline of the conduit.
The distribution of pressures around a flexible pipe under an earth fill, determined
in accordance with this hypothesis, is shown in Figure 1.
Once this fill hypothesis was established, it was a simple matter to develop mathe-
matical expressions for the moments, thrusts, shears, and deflections of the pipe. The
rP.sulting expression for the horizontal deflection of a flexible pipe is known as the Iowa
formula:
4
l::.x = K We R 3 /(EI+ 0.061 e R ) (1)

in which R is the radius of the pipe, E is the modulus of elasticity of the pipe material,
1 1s the moment 01 merua 01 the cross section of a i-m. length 01 pipe waii, anct JS.. 1s a
dimensionless bedding constant that results from the necessary mathematical manipula-
tions and is a function of the bedding angle a.
Equation 1 gives a value of deflection that may be called the immediate deflection.
For the deflection that will develop over a long period of time, the equation must be
multiplied by a deflection lag factor, D1.
In 1957, Watkins and Spangler (18) examined the Iowa formula dimensionally. They
had hoped to design some small-scale model tests to measure the soil modulus, e. To
their surprise, however, they found that e could not possibly be a property of the soil
because its dimensions were not those of a true modulus. The dimensionally correct
modulus is

E' = e R (2)

and is known as the modulus of soil reaction.

Consequently, the modified Iowa formula is given as

(3)

If the Iowa formula is rearranged, the following terms can be introduced to describe
three separate factors that affect the pipe deflection: load factor = D1 K We; ring stiff-
ness factor = EI/R3; and soil stiffness factor = 0.061 E'. Thus, the Iowa formula equa-
tion (,!) can be represented as
t:.x = load factor (4)
(ring sillfness factor + soil stiffness factor)
 91
In the following sections, the range of variation and influence of each of these fac-
tors will be examined in detail.

Ring Stiffness Factor

The ring stiffness factor is a function of the modulus of elasticity of the pipe mate-
rial and the moment of inertia of the particular corrugation configuration and the radius
of the pipe. The variation of this factor with respect to different corrugations and radii
is shown in Figure 2. In most cases, the ring stiffness factor has very little influence
on the deflection of a particular pipe because the soil stiffness factor is much larger.

Load Factor
The load factor incorporates the parameters that have to do with the magnitude and
distribution of the soil pressures on a buried pipe. The pipe deflection is directly pro-
portional to the load factor, and yet less is known about its components than any other
value in the Iowa formula.
In a discussion of the deflection lag factor, D1, Spangler (11) has stated, "The deflec-
tion lag factors observed in the experiments ranged from 1.3810 1.46, and in no in-
stance was equilibrium completely attained. Therefore, 1. 5 is suggested as a conserva-
tive value for design use for standard corrugated-pipe culverts installed without strut-
ting or predeforming."
In the second edition of his textbook (14), Spangler states, " The deflection lag factor
cannot be less than unity and has been observed to range upward toward a value of 2.0.
A normal range of values from 1.25 to 1.50 is suggested for design purposes."
The load on the pipe depends on whether the pipe is installed in a trench condition
or in an embankment condition; however, in an attempt to obtain information of direct
use to the design engineer, we may "simplify" the Iowa formula by assuming that the
settlement ratio is zero. Then the total load on the pipe can be expressed as

We= Hy Bo (5)

It is now possible to substitute this value in the Iowa formula and solve for the fill height.
The deflection can be expressed as a percentage of the nominal diameter (i.e., let
P = ~x/Bo), and the equation can be expressed as

H = 144 P / (D1 Ky) [(EI/R3 ) + 0.061 E'] (6)

in which the constant 144 was introduced for dimensional consistency to yield a field
height in feet.
Equation 6 provides a basis for making a parametric study of the various factors in-
fluencing the height of fill as predicted by using the Iowa formula. From this equation
it can be seen that the design fill height is inversely proportional to the value of the
deflection lag factor. Thus, any variation of the lag factor can have a very significant
effect on the design height of fill, ·and care should be exercised in selecting an appro-
priate value for this parameter. In addition, it should be noted that a change in the
bedding constant, K, over its range of values can modify the design fill height by as
much as 30 percent.

Soil Stiffness Factor

Spangler regards E' as a semi-empirical constant and has stated that the properties
of the soil that influence this factor are somewhat obscure although qualitatively it is
certain that texture and density characteristics are of prime importance.
The influence of the soil stiffness factor can be illustrated by means of Eq. 6. It will
be assumed that K = 0.100. For purposes of comparison three design criteria will be
established: design criterion I assumes E' = 700 psi and D1 = 1. 50; design criterion II
assumes E' = 1,400 psi and D1 = 1.50; and design criterion III assum es E' = 1,400 psi and
D1 = 1.25.
92

The variation of fill height for design criterion I is shown as the lowest curve in
Figure 3. The effect of doubling the soil stiffness factor is reflected by the middle
curve in Figure 3 (i.e., design criterion II). The corresponding change in fill height
is not linear but is dependent on the value of the ring stiffness factor.
The variation of fill height for design criterion III is shown as the highest curve in
Figure 3. It should be noted that the only distinction between this curve and the curve
for design criterion II is a change in the value of D1 from 1. 50 to 1.2 5. The modified
fill height as reflected by this change is of significance but is not as large as the varia-
tion produced by doubling the value of E'.
For an R3 / EI value of 50, the ratio of fill height s for design criterion III and design
criterion I is 2.01. This difference is due to the changes in the D1 and E' parameters.
If the bedding coefficient is also included in the variation, an even greater difference
in fill heights is obtained. Assuming the bedding angle o: to be O deg for design criterion
I yields a design fill height of 22.84 ft; assuming a value of o: = 90 deg for design crite-
rion III yields a design fill height of 60.21 ft. The resulting ratio of fill heights is 2.64.
Thus, an almost threefold variation of design fill heights is possible, depending on
which set of "realistic" values are assigned to the various parameters in the Iowa de-
flection formula.
The values of the parameters for design criterion I are equivalent to those suggested
by the Bureau of Public Roads in 1966 (17), whereas the values of the parameters for
design criterion Ill are equivalent to thevalues recommended by the Federal Highway
Administration in 1970 (20).

ANALYSIS OF TEST DATA

From the foregoing discussion it can be seen that the most influential parameter in
the Iowa deflection formula is the modulus of soil reaction E'. Unfortunately, this
modulus is not a fundamental soil property that can be evaluated by means of a stan-
dard laboratory test. Researchers (13, 19) have attempted to determine E' by direct
laboratory measurements but withoutsuccess. Recently, several investigators (1, 5,
7, 8) attempted to correlate E' with basic soil properties, but these correlations have
not yet been widely tested.
Consequently, the design engineer must rely 011 past experience and "1·ecommended
values" of E' for design purposes. Spangler has suggested that E' be considered as a
semi-empirical constant whose values may best be determined by observation and mea-
surements on actual culvert installations. Recently, Spangler published a table showing
values of E' for various kinds of soil in various states of compaction as deduced from
18 actual culvert installations (15). These data are given in Table 1. This limited
study reveals that the modulus E' varies over a wide range, from as little as 234 psi
to as much as 7,980 psi, a 34-fold variation.
These data are from two experimental programs and three actual installations. The
original sources of the information are as follows:

Item Date Report Reference

Number Test Dates Published Number
1 to 10 1927 to 1936 1938 10, 11
11 to 15 1924 to 1926 1929 2, 12
16 1952 1956 3
17 1952 1956 16
18 1966 1969 9

Thus 83 percent of the data are from test installations that were constructed more than
35 years ago. Spangler (15) has stated that these data are the result of his efforts to
accumulate information onvalues of E' and to correlate them with soil properties.
For purposes of later identification, the culverts have been divided into three groups.
Group I includes items 1 through 10 and consists of the pipe in the Iowa State College
tests. Group II includes items 11 through 15 and consists of the pipe in the North Caro-
lina tests. Group III consists of the three large-diameter installations having high fill
 93

heights. Additional data from the original test results are given in the first five col-
umns and the eighth column of Table 2.
The reported values of e and E' for items 1 through 17 (as shown in the last two col-
umns of Table 1) were evaluated by means of the basic definitions. The test data used
for these calculations consisted of the measured deflection and the side pressure, h,
which were measured for items 1 through 10 and estimated for items 11 through 17.
As an alternative approach, the modulus can be evaluated from the Iowa deflection
formula. Thus, if Eq. 3 is rewritten in the form

E' = 16.39 [(D1 K We/Ax) - (EI/R3 )] (7)

and assumptions are made concerning values of D1, K, and We, then E' can be evaluated
on the basis of the measured values of the deflection and the ring stiffness factor.
It has been suggested that the modulus of passive resistance and/or the modulus of
soil reaction is dependent on the density of the soil adjacent to the pipe. To check
this hypothesis, we made a linear regression analysis of the e and E' data given in
Table 1 with the percentage of standard Proctor density data for group I pipe given in
Table 2. The prediction equations, as obtained from the linear regression analysis,
are given in Table 3 together with the correlation coefficient for each case. From
Table 3 it can be seen that there is a higher degree of correlation between e and PPD
than there is between E' and PPD. This analysis cannot be considered as conclusive
because only 10 data points are available for consideration.
Regression analyses were also made to explore the dependence of the moduli e and
E' on the parameter D and the nondimensional ratio H/D. By considering these param-
eters, it was possible to analyze various combinations of the 18 items given in Tables
1 and 2. The resulting prediction equations and the associated correlation coefficients
are given in Table 3.
The results given in Table 3 do not establish a strong enough dependence of E' on
PPD, D, or H/D to warrant the use of any of these equations for design purposes. For
example, if the data for item 17 are considered, the prediction equations at the bottom
of each group in Table 3 yield the following values:

Parameter Considered E' Value

Percentage of Proctor density = 100 916 psi
Diameter = 84 in. 2,432 psi
Ratio of H/D = 19.6 2,834 psi

All of these values fall short of the value of 7,980 psi as given in Table 1.
The measured horizontal and vertical deflections for 16 pipe culverts are given in
Table 2 together with the ratio of these deflections. The mean value of the ratio
Av.rt./Ahoriz. is 1.13, and the standard deviation is 0.18. Thus, it can be seen that the
vertical and horizontal deflections are not equal, and the vertical deflection is generally
the larger.
All of the test pipe in groups I and II (items 1 through 15) were studied prior to the
introduction of the E' concept, and the "exact" values of e as reported in the original
publications (2, 11) are given in Table 2. Using these values of e, we calculated the
theoretical horizontal deflection of the pipe by means of the Iowa formula as given by
Eq. 1. These values are given in Table 2 as (Ah)eaie., together with the ratio of this
deflection to the measured horizontal deflection. The mean value of this ratio is 0.96,
and the standard deviation is 0.38, indicating a fairly wide scatter of the data.
As an alternative to evaluating E' by means of the basic defining equation, we ex-
plored the possibility of utilizing more of the test data. This was accomplished by as-
suming D1 = 1.00 and K = 0.100 and by rewriting Eq. 7 in the form

E" = 16.39 [0.20 We/(Ahortz. + Avort.) - (EI/R 3 )] (8)

The measured values of the ring stiffness factor, We, Ahor1z., and Avort. as reported in
the test results were utilized to evaluate this alternative form of the modulus. The
Table 1. Values of E' for 18 flexible pipe culverts.
Flll
Height/
Pipe Flll Pipe Side
Diameter Heii;ht Diameter Pres - Defl ec - Back- E'
Item Group (in.) (ft) (ft/ ft ) sure lion Load Type of Soil fill (psi/in.) (psi)

1 42 15 4.28 M M Loam topsoil u 14 294

2 42 16 4.56 M M Well-graded
gravel u 32 672
3 36 15 5.00 M M Sandy clay loam u 13 234
4 42 15 4.28 M M Sandy clay loam u 15 315
5 48 15 3.74 M M Sandy clay loam u 14 336
6 60 15 3.00 M M Sandy clay loam u 12 360
7 36 15 5.00 M M Sandy clay loam T 28 502
8 42 15 4.28 M M Sandy clay loam T 25 525
9 I 48 15 3.74 M M Sandy clay loam T 29 696
10 I 60 15 3.00 M M Sandy clay loam T 26 780
11 II 20 12 7.20 E M Sand NA 35 350
12 II 21 12 6.86 E M Sand NA 82 861
13 II 30 12 4.80 E M Sand NA 25 375
14 II 30 12 4.80 E M Sand NA 80 1,200
15 II 31.5 12 4.58 E M Sand NA 56 882
16 Ill 66 170 30.90 E M Clayey sandy silt C 40 1,320
17 Ill 84 137 19.60 E M Crushed sand-
stone C 190 7,980
18 Ill 216 83 4.60 M M Graded crushed
gravel C 58 6,300
Note: e = modulus of passive resistance, E'= modulus of soil reaction, M = measured, U"" untamped, T = tamped, E = estimated, NA= not available, and
C • compacted.

Table 2. Summary and analysis of test data for 16 flexible pipe culverts.

Data From Teat Results

Horizontal Deflec-
Percent- lion Calculated E"
age of Measured Deflections Using e Value (based
Standard on "F:/ E'
Proctor w, lhorll. ~Hrt. t:..,.,,. / e (t:..,) .. 1<. (t:..,),.,,./ t:...,,) (E' from
Item Density r,, (lb/in.) (in.) (in.) bihorts• (psi/in. ) (in.) (t:..,)..... (psi) Table 1)

1 87.7 443 1.38 1.43 1.04 14.0 1.23 0.89 226 0.77
2 93.2 443 0.75 0.80 1.07 32.0 0.78 1.04 699 1.04
3 86.8 -0.38 300 1.27 1.21 0.95 12.93 1.17 0.92 241 1.03
4 89.2 -0.26 375 1.39 1.31 0.94 14.82 1.41 1.01 332 1.05
5 91.4 -0.38 408 1.62 1.56 0.96 14.39 1.56 0.96 338 1.01
6 89.0 -0.67 508 1. 79 1.82 1.02 11.60 1. 72 0.96 402 1.12
7 93.5 -0.18" 342 0.67 0.71 1.06 28.30 0.83 1.24 657 1.31
8 91.5 -0.14" 425 0.77 0.84 1.09 24.67 0.94 1 .22 747 1.42
9 92.6 -0.17" 458 0. 86 1.02 1.19 29.30 0.97 1.13 716 1.03
10 90.5 -0. 76" 483 1.01 1.26 1.25 25.71 0.87 0.86 640 0.82
11 98 0.48 0.543 1.13 33.3 0.479 1.00 296 0.85
12 173 0.212 0.233 i.10 63 .2 0. 176 0. 83 al3 0.36
13 162 0.631 0.812 1.29 26.3 0.711 1.13 352 0.94
14 233 0.234 0.261 1.12 70.0 0.255 1. 09 1,030 0.86
15 224 0.322 0.379 1.18 50.3 0.324 1.01 673 0.76
18 9,800 1.4 2.4 1.71 8,366 1.33

Note: Item 11 was a smooth iron pipe with t = 0.076 in. Item 13 was a smooth iron pipe with t = 0.109 in. Item 14 was a steel tube with t • 0 ,349 in,
All others were corrugated metal pipes,
'Adjusted value.

Table 3. Prediction
Correlation
equations. Group Characteristic Value Coefficient

Percentage of Proctor density•

I e 0.718
I E' 0.557
Pipe diameter'
I C1, C2 25.0, -0.09 0.086
I Cl, Cf 96.4, 8.2 0.363
I, Il combined C1, C2 66.0, -0.09 0.572
I, U combined Cl, C2 955.0, -10.1 0.431
I, II, II1 combined Cl, C~ -550.0, 35.5 0. 721
Fill height-pipe diameter ratio'
I C3,' C,a 14.4, 1.56 0.120
1
I C3 1 C11, 810.7, -83.1 0.305
I, II combined C3/ C4 -4.8, 7.6 0.468
1
I, II combined C3, C.,. 492.0, 14.6 0.061
I, Il, Ill combined c;, c~ 521.0, 118.0 0.385

'e = -240 5 + 2,886 and E'= -3,790 + 47.06.

'• • c 1 + c2 (0} and E' = Ci + C2(D},
'e = C3 + c,(H/ D} and E' = C3 + C.(H/0}.
 95

resulting values are given in Table 2 together with the ratio of this modulus to the value
of E' as reported by Spangler and given in Table 1. The mean value of this ratio is
0.98, and the standard deviation is 0.25, indicating that approximately two-thirds of the
data lie within the range of 0. 73 to 1.23.
The quantity We as used in Eq. 8 was that value reported in the test results and
given in Table 2. In order to investigate the validity of the design assumption that the
settlement ratio is zero, we evaluated the weight of the total prism of soil above the
culvert by means of Eq. 5 (listed as the quantity We), The values of W0 are compared
with the We values. Perfect correspondence does not exist between these two quantities;
however, the data are very well correlated (the correlation coefficient is equal to 0.953),
and a regression analysis yields the equation

W. = 10 + 1.32 We (9)

The mean value of the ratio W0 /We is 1.38, and the standard deviation is 0.20.
From the foregoing it can be seen that the value of the modulus of soil reaction as
evaluated from test results is very sensitive to the method of calculation. In addition,
the assumed value of the bedding coefficient will have a large influence on the result.
The design fill height as predicted by Eq. 6 and the values of the parameters as given
by design criterion I (K = 0.100, E' = 700 psi, and D1 = 1.50) and design criterion m
(K = 0.100, E' = 1,400 psi, and D1 = 1.25) ai·e shown in Figure 4, and it can be observed
that only seven of the points for design criterion I lie in the conservative region of the
figure. A linear regression analysis was made of both sets of data and yielded the fol-
lowing results:
1. Design criterion I:

Hdeelgn = 11.4 + 0.47 Hextrapolated (10)

The correlation coefficient is 0.67.

2. Design criterion III:

Hdealgn = 33.4 + 0.52 Hextrapototed (11)

The correlation coefficient is 0. 66.

Thus, within the range of values considered, there is a fairly good correlation of the
data; however, both criteria tend to yield unconservative results when compared to the
observed behavior of actual culvert installations. Consequently, it appears that a more
complete definition is required for the various parameters used in the Iowa deflection
formula.

STATISTICAL EVALUATION OF E'

There are really insufficient data to do a truly meaningful statistical evaluation of the
data concerning E' as given in Table 1. However, because no other data are available,
the most one can do is to provide a statistical analysis of the 18 observed values.
The first case is group I, the second is group II, the third is groups I and II com-
bined, and the fourth is groups I, II, and III combined. The first three cases indicate
a fairly symmetric clustering of the data around the 300- to 600-psi range but with a
slight tendency for the number of observations to decrease more gradually on the
higher side than on the lower side. The histogram in such a case is said to be skewed
to the right. The fourth case produces a much greater skewness. This skewness leads
one to suspect that the normal, or Gaussian, distribution, which is symmetric, will
not represent all four cases well. Three common probability distributions that exhibit
skewness toward higher values are the Rayleigh, exponential, and log-normal distri-
butions.
A useful statistical aid is a graphic display of the observed "cumulative" made by
plotting the data on a special graph paper. In Figure 5, the observed "cumulative" for
Figure 1. Pressure pattern Figure 2. Equivalence of corrugation configurations and R3 /EI values.
around flexible pipe (1.Q).
w, 0 20 40 60 80 100 120
Po =-2r

l.orrugat1on
I ! ! ! ! ! llall
Confill!retion ' 'Dlickn•,

0-18" 24n 30" 36"

0.064 11

((//
0 . 079"

0.109"
2 2/ J " 1/2
O. l38"

0 . 168"

:0-30" 36 11 42" 48" 54" 60 11

0 . 064"

I(@
0 . 079"

J X l 0.109"
0 . 138"

0.168"

84" 96 11 108 11 120 11

0 . 109"
0.138"

0.168"
6 X 2 • 0.188 11

0 . 218"

0 . 249"
0.280"

0 20 40 60 80 100 12
3
R /EI, in~/kip

Figure 3. Fill height as a function of assumed criteria. Figure 4. Fill heights for 5 percent long-time deflection.

100 60 - , - - -- - - - - - - - -- - - - -- . . - - - - -

Design Criteria III:

E' • llt00 psi
70
80
o • 1.25
1 Unconservative
Region

60 )( /
E' • 1400 psi /
o • 1.50
/
1
)(
/
~ so /
"' X X X /
0 /
".
-"
t 40
/x -#, X
XX /
/

........,= /
/
"' / 0
0
20
E' • 700 psi
.!ho
. /
/ 0
Cl
Cl
D • 1. 50 0 /
1 /

0 -+---+- -1----<1--~I--- +--+---+---+---+-

w ~y
/
~ Conservative
Region
0 20 40 60 80 100
3 /
R /EI, in~/kip 10 /
/0 Design Criteria I: E' • 700 psi, D = 1.50
1
/
X Design Criteria III: E'= 1400 psi, D ~ 1.25
/ 1
Ooof-----1-----t-- -- - 1 - - --+----+-----+-
0 10 20 30 40 50 60
Extrapolated Fill Height - feet
 97

groups I, II, and III combined is graphed by using four different types of probability
scales: normal, Rayleigh, exponential, and log-normal. The more colinear are the
observed points on a particular scale, the better is the assumption that the data follow
that particular distribution.
It is observed that graphing the observed "cumulatives" on a log-normal probability
scale produces a result significantly closer to a straight line than with the other dis-
tributions. It should be noted that there are not sufficient data to expect exact fit for
any proposed distribution.
Some important statistics of the data are given in Table 4. The observed mean, m,
is found from the data as

n
m = (1/n) I: E~ (12)
i=l

where E~ is one observed value of E' and n is equal to the number of observed values.
The observed standard deviation, a, a measure of dispersion in the observed data,
is found as

a=
n
{ (1/n) ~ [(E~ - m)2]
}½ (13)
1=1

The observed coefficient of variation, V, is simply a/m. The observed median, m, is

the central value of the observed data.
Using the observed mean and standard deviation, we fitted a log-normal distribution
to the data for all four groupings of the data. From this fitted distribution, the model-
calculated median and standard deviation of the logarithm of E' were found. The appro-
priate formulas are

m* = m exp [(-½) atn (14)

ai = [Ln (V 2 + 1)//2 (15)

If we consider the number of observations, the observed and model-calculated medians

given in Table 4 agree rather well.
In order to substantiate the adoption of the log-normal distribution, we performed a
chi-square test on the observed data. This test is a measure of the difference between
the observed values of the histogram and the values predicted by the assumed distribu-
tion. The difference, d1, is compared with a table value of the chi-square distribution
for a given level of significance. If d1 is less, the deviations are not considered signif-
icant (at the given level), and the assumed distribution may be accepted.
Because of the limited amount of data at hand for this problem, the results of a chi-
square test must be used cautiously. The test was carried out at the common 10 per-
cent significance level. The results are as follows: group I, d1 = 1.64 and chi-square =
2.7; group II, d1 = 0.66 and chi-square= 4.6; groups I and II, d1 = 0.97 and chi-square=
4.6; groups I, II, and III, d1 = 70.31 and chi-square = 12.0. The chi-square value has
been found as a function of the number of degrees of freedom in the deviations.
For groups I and II separately and I and II combined, the deviations are well below
the chi-square level, and the log-normal distribution is acceptable. For groups I, II,
and III combined, however, the chi-square test indicates rejection of the assumed dis-
tribution. Data often fail a chi-square test because of the uncertainty surrounding the
tail areas of a distribution. This is especially true when insufficient data are available.
In this case, a graphic·Kolmogorov-Smirnov test can be performed. Such a test defines
an acceptable region, at a given significance level, for observed data surrounding the
assumed distribution. It is based on the fact that the maximum difference between the
"cumulative" of the histogram and the "cumulative" of the assumed distribution is a
variable whose distribution is independent of the assumed distribution. Figure 6 shows
98

Figure 5. Probability "cumulatives" for groups I, 11, and Ill .

1-FE , l.0
_g
.,
-7
.,
.5

....

•3

.l
m
...
.ur
10
E' ·"'
a) Normal (Gaussian) b) Rayleigh
1-FE I , .

0 ,9

na
0.1
o.,

J r1
O.t FE '

J
60
~

40
"!,I)

I
:IO

Jl)

-
E'
E'
.1
0 JOOO 20<l0 5000 i()OO fOOO 0000 1000 ll(OQ
,,.,, 400 1J<IO 8000

c) Exponential d) Logoorma 1

Table 4. Observed and calculated statistical parameters for E'.

Model
Observed Observed Calculated
Observed Standard Coefficient Standard Observed Calculated
Mean Deviation of Deviation Median Median
Group (psi) (psi) Variation of Log (psi) (psi )

I 471.4 182.42 0.3870 0.3736 368 439.63

II 733.6 326.04 0.4444 0.4246 861 670.37
I, II combined 558.8 269.99 0.4832 0.4581 502 503.15
I, II, III combined 1,332.33 2,094.15 1.5718 1.1155 600 715.18
Figure 6. Kolmogorov- ~-
Smirnov test for groups I, FE '
II, and Ill. 99.'J
99.8 10'7. Significance Level

" 98

!iS

90

BO
70
,d
.ro
40
lO Upper K·S limit

ID

10

Figure 7. Probability 0.002

ranges for data in groups ;., median = 503 psi.
I and II. 'M
0 337. limits: 411 and 764 psi.
g 807. limits: 271 and 935 psi.

,,;... 0 . 00 1
~

..."
m
A

0
1500 20 0 25 30 0 3500 E'

Figure 8. Probability 0 . 002

ranges for data in groups
I, II, and Ill. 'M
0
33% limits : 437 and 1170 psi.
807. limits : 161 and 3184 psi.
g
t 0 .001
!J:j
...
A
m

0
2000 2500 3000 3500 E'
100

a Kolmogorov-Smirnov test on groups I, II, and m combined at the 10 percent signif-

icance level. Because all of the observed data points are contained between the upper
and lower Kolmogorov-Smirnov limits, the fitted log-normal distribution may be ac-
cepted at the 10 percent significance level.
It is desirable to determine ranges of probable levels for E'. To do this, the log-
normal distribution was used with the parameters given in Table 4. A standardized
normal variable, u, was determined from the general log-normal variable, E', by the
following relation:

u = (1/crLn) Ln (E' /m) (16)

Design levels were chosen from the student t distribution. This distribution is
similar to the normal, but it reflects the uncertainty associated with having only a finite
number of data observations. Bot)l 80 percent and 33 percent ranges were found. The
results are shown in Figures 7 and 8 respectively for groups I and II combined and
groups I, II, and III combined.
The 80 percent range is defined as the range in which there is an 80 percent proba-
bility that an observed value will lie. This means there is a 20 percent probability that
an observed value will lie outside of the range. The limits of this range have been
chosen so that there is a 10 percent probability that an observed value will be below the
range and a 10 percent probability that an observed value will be above the range. Notice
that such a definition produces unsymmetric limits with respect to the median. This is
because of the skewed nature of the log-normal distribution.
Based on the information shown in Figures 7 and 8, it can once again be observed
that the wide spread of the data as given in Table 1 makes it extremely difficult to justify
the use of E' values in excess of the median values. Thus, from a statistical point of
view, it is essential that additional data be recorded from full-scale culvert installations
in order to obtain a statistically meaningful and realistic value of the modulus E' for
design purposes.

CONCLUSIONS
Since its introduction more than 30 years ago, the Iowa deflection formula has been
utilized as a fundamental design critPrion for r.orrne;::it,;,rl m,;,t?.l. plpe. 'T'ho 1,..,.,., f"""""l"
is based on an assumed distribution of loading around the pipe, including the total ver-
tical load, We, as predicted by the Marston load theory. In addition, the Iowa formula
contains three parameters, D1, K, and E', which are empiric~J quantities . To date, no
extensive study has been made to establish a sound basis for estimating or evaluating
these very important parameters. Realistic variations of these parameters can yield
an almost threefold variation in the design requirement for a particular installation of
pipe.
Spangler has collected data concerning experimentally determined values of E'; how-
ever, as given in Table 1, these 18 values vary over a very wide range. The modulus
of soil reaction E' is not a basic material property whose value can be determined from
field or laboratory tests. The values of E', which have been suggested for design use,
are based on empirical considerations, and this study has shown that no strong correla-
tion exists between this modulus and the parameters of the soil-culvert systems whose
properties are given in Tables 1 and 2. A statistical study of these data confirms the
fact that there is no rational basis for assigning any value for E' that is greater than
the median value.
This study has shown that the Iowa deflection formula per se can be an effective de-
sign tool. However, there are insufficient data available in the open literature to assist
the designer in making realistic and rational decisions concerning the establishment of
appropriate and reliable values of the parameters r,d, We, Di, K, and E'. Until such
time that sufficient quantities of additional field data can be obtained and analyzed in a
statistical sense, the designer should exercise extreme caution and discretion in assign-
ing values to all of these parameters in the Iowa deflection formula.
 101

REFERENCES
1. Allgood, J. R. Influence of Soil Modulus on the Behavior of Cylinders Buried in
Sand. U.S. Naval Civil Engineering Laboratory, Port Hueneme, California, Tech.
Rept. R-582, June 1968.
2. Braune, G. M., Cain, W., and Janda, H. F. Earth Pressure Experiments on
Culvert Pipe. Public Roads, Vol. 10, No. 9, 1929, pp. 153-176.
3. Castes, N. C., and Proudley, C. E. Performance Study of Multi-Plate Corrugated-
Metal-Pipe Culvert Under Embankment-North Carolina. HRB Bull. 125, 1956,
pp. 58-169.
4. Howard, A. K. Laboratory Load Tests on Buried Flexible Pipe-Progress Report
No. 2. Bureau of Reclamation, Rept. REC-OCE-70-24, June 1970, 74 pp.
5. Luscher, U. Buckling of Soil-Surrounded Tubes. Jour. Soil Mechanics and Founda-
tions Division, Proc., ASCE, Vol. 92, No. SM6, Nov. 1966, pp. 211-228.
6. Marston, A. Theory of External Loads on Closed Conduits in the Light of the
Latest Experiments. Eng. Exp. sta., Iowa State Univ., Bull. 96, Feb. 19,
1930.
7. Nielson, F. D. Modulus of Soil Reaction as Determined From the Triaxial Shear
Test. Highway Research Record 185, 1967, pp. 80-90.
8. Nielson, F. D., Bhandhausavee, C., and Yeb, K. Determination of Modulus of Soil
Reaction From Standard Soil Tests. Highway Research Record 284, 1969, pp. 1-12.
9. Scheer, A. C., and Willett, G. A., Jr. Rebuilt Wolf Creek Culvert Behavior.
Highway Research Record 262, 1969, pp. 1-13.
10. Spangler, M. G. The Structural Design of Flexible Pipe Culverts. Public Roads,
Vol. 18, No. 12, 1938, pp. 217-231.
11. Spangler, M. G. The Structural Design of Flexible Pipe Culverts. Eng. Exp. Sta.,
Iowa state Univ., Bull. 153, Dec. 24, 1941, 84 pp.
12. Spangler, M. G. Stresses and Deflections in Flexible Pipe Culverts. HRB Proc.,
Vol. 28, 1948, pp. 249-259.
13. Spangler, M. G., and Donovan, J. C. Application of the Modulus of Passive Re-
sistance of Soil in the Design of Flexible Pipe Culverts. HRB Proc., Vol. 36, 1957,
pp. 371-381.
14. Spangler, M. G. Soil Engineering, 2nd Ed. International Textbook Co., Scranton,
Pennsylvania, 1960.
15. Spangler, M. G. Discussion of Rebuilt Wolf Creek Culvert Behavior by Sheer, A.
C., and Willett, G. A., Jr. Highway Research Record 262, 1969, pp. 10-13.
16. Timmers, J. H. Load Study of Flexible Pipes Under High Fills. HRB Bull. 125,
1956, pp. 1-11.
17. Townsend, M. Corrugated Metal Pipe Culverts-Structural Design Criteria and
Recommended Installation Practices. Bureau of Public Roads, U.S. Department
of Commerce, 1966, 26 pp.
18. Watkins, R. K., and Spangler, M. G. Some Characteristics of the Modulus of
Passive Resistance of Soil: A Study in Similitude. HRB Proc., Vol. 37, 1958,
pp. 576-583.
19. Watkins, R. K., and Nielson, F. D. Development and Use of the Modpares Device.
Jour. Pipeline Division, Proc., ASCE, Vol. 90, Jan. 1964, pp. 155-178.
20. Wolf, E. W., and Townsend, M. Corrugated Metal Pipe-Structural Design Criteria
and Recommended Installation Practice. Bureau of Public Roads, Federal Highway
Administration, U.S. Department of Transportation, 1970, 26 pp.

DISCUSSION
M. G. Spangler, Iowa State University
The paper by Parmelee and Corotis is a very welcome addition to the literature that
deals with the structural design and performance of flexible conduits under earth fill
102

loads. The authors have reviewed the development of the Iowa deflection formula and
have pointed out thP. empirical and semi-empirical nature of the deflection lag factor,
D1, and the modulus of soil reaction, E'. They have skillfully applied methods and
principles of the science of statistics to a determination of the adequacy of currently
available data on which to base selections of workable values of these parameters. The
writer is in complete agreement with the authors' conclusion that there is insufficient
knowledge available at this time to enable a designer to select realistic and economical
values of the needed parameters.
At the time the deflection formula was developed, experimental evidence clearly in-
dicated the important influence of the modulus of soil reaction, but the range of the ex-
periments was relatively narrow. Since then, applications of the equation to actual
situations has revealed that E' appears to vary over a very wide range, from as little
as 234 psi to as much as 8,000 psi. These facts indicate and the authors' statistical
analyses confirm the need for a massive program of field measurements to establish a
working body of data for use in this area of design. The paper provides a valuable
background and guidance material for any future study directed toward fulfilling this
need.
SPONSORSHIP OF THIS RECORD
GROUP 2-DESIGN AND CONSTRUCTION OF TRANSPORTATION FACILITIES
John L. Beaton, California Division of Highways, chairman

GENERAL DESIGN SECTION

F. A. Thorstenson, Minnesota Department of Highways, chairman

Committee on Subsurface Structures Design

Kenneth S. Eff, Department of the Army, chairman
Roger L. Brockenbrough, T. F. DeCapiteau, Paul D. Doubt, W. B. Drake, C.R. Hanes,
Paul M. Heffern, John G. Hendrickson, Jr., L. R. Lawrence, F. Dwayne Nielson,
E. F. Nordlin, Richard A. Parmelee, A. J. Reed, M. G. Spangler, R. S. Standley,
Harold V. Swanson, James D. Washington, Reynold K. Watkins, Howard L. White

SOIL MECHANICS SECTION

Carl L. Monismith, University of California, Berkeley, chairman

Committee on Subsurface Soil-Structure Interaction

Ernest T. Selig, State University of New York at Buffalo, chairman
Jay R. Allgood, Mike Bealey, Bernard E. Butler, T. Y. Chu, T. F. DeCapiteau, Kenneth
S. Eff, Lester H. Gabriel, Delon Hampton, Raymond J. Krizek, Don A. Linger, F.
Dwayne Nielson, Harry H. Ulery, Jr., Reynold K. Watkins, Howard L. White

Lawrence F. Spaine and John W. Guinnee, Highway Research Board staff

The sponsoring committee is identified by a footnote on the first page of each report.

103