You are on page 1of 128

M

M
a
a
n
n
h
h
a
a
t
t
t
t
a
a
n
n
C
C
o
o
l
l
l
l
e
e
g
g
e
e

C
C
e
e
n
n
t
t
e
e
r
r
f
f
o
o
r
r
G
G
e
e
o
o
t
t
e
e
c
c
h
h
n
n
o
o
l
l
o
o
g
g
y
y



Soil-Structure Interaction Research Project
-
B Ba as si ic c S SS SI I C Co on nc ce ep pt ts s a an nd d
A Ap pp pl li ic ca at ti io on ns s O Ov ve er rv vi ie ew w



Report No. CGT-2002-2

by

John S. Horvath, Ph.D., P.E.
Professor of Civil Engineering
Director/Center for Geotechnology



Manhattan College
School of Engineering
Bronx, New York, U.S.A.

August 2002

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
ii
This report plus others in the current Manhattan College Center for Geotechnology (CGT) and
former Civil Engineering Department geotechnical engineering program (CE/GE)
research report series are available in PDF format via the Internet at:

www.engineering.manhattan.edu/civil/CGT.html.



2002 by John S. Horvath. All rights reserved.

Prof. John S. Horvath, Ph.D., P.E.
Manhattan College
Civil Engineering Department
Bronx, New York 10471-4098
U.S.A.


email: john.horvath@manhattan.edu
personal webpage: www.manhattan.edu/~jhorvath/
voice: +1-718-8627177
fax: +1-718-8628035



Limitations

The information contained in this report is the result of original research and is made available
to the public solely as a contribution to the general state of civil engineering knowledge. As such,
this report does not constitute a design manual or standard, and should not be interpreted, used or
referred to as such. The author and Manhattan College assume no liability for the performance of
any structure constructed using the methodologies discussed in this report.
In addition, any reference, either direct or indirect, to a specific manufacturer, material or
product is made solely for the purposes of the study documented in this report and does not
constitute an endorsement or promotion of that manufacturer, material or product by the author or
Manhattan College.


Distribution and Use

Authorization is hereby granted by the copyright holder to copy, distribute and print this report
without restriction provided that this report is copied and printed in its entirety; distributed at no
cost; and used for personal education and individual reference purposes only. This report may not
be sold or used in whole or in part in any manner for commercial purposes of any type, including
manufacturer's product or marketing literature, without prior written consent of the copyright
holder.
Those wishing to make a paper copy of this report are advised that it is formatted for double-
sided printing. A gutter has been provided in the page formatting to allow space for binding along
the inner side if desired. It is suggested that the printer settings in Adobe Acrobat be set to
Print as image and Auto-rotate and center pages for optimum readability.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
iii
Contents



List of Figures............................................................................................................................... vii

List of Tables ................................................................................................................................. ix

The Manhattan College School of Engineering Center for Geotechnology and Its Mission . xi

Preface.......................................................................................................................................... xiii

Executive Summary
Introduction ................................................................................................................................ xv
Report Goals and Outcomes ....................................................................................................... xv

Section 1 - Background and Scope of Report
1.1 Soil-Structure Interaction ....................................................................................................... 1
1.2 Subgrade Models .................................................................................................................... 1
1.3 Report
1.3.1 Philosophy........................................................................................................................ 2
1.3.2 Scope ................................................................................................................................ 4
1.3.3 Organization..................................................................................................................... 4

Section 2 - Soil-Structure Interaction Modeling: Overview
2.1 Introduction ............................................................................................................................ 7
2.2 Mat Foundation Analysis
2.2.1 Parameters of Interest ....................................................................................................... 7
2.2.2 Problem Components ....................................................................................................... 7
2.2.3 Problem Solutions
2.2.3.1 Ideal............................................................................................................................ 8
2.2.3.2 Traditional ................................................................................................................ 10
2.2.3.3 Modern Alternatives................................................................................................. 12
2.3 Additional Comments
2.3.1 Site Characterization ...................................................................................................... 14
2.3.2 Computer Software......................................................................................................... 15

Section 3 - Soil-Structure Interaction Modeling: Structural Aspects
3.1 Introduction .......................................................................................................................... 17
3.2 Structural Analysis Notation
3.2.1 Introduction .................................................................................................................... 17
3.2.2 Displacements ................................................................................................................ 17
3.2.3 Forces/Stresses ............................................................................................................... 18
3.3 Structural Element Weight
3.3.1 Introduction .................................................................................................................... 18
3.3.2 Example: Effect of Mat Weight on Mat Behavior ......................................................... 18
3.3.3 Example: Effect of Mat Weight on Superstructure Behavior......................................... 19
3.3.4 Conclusions and Recommendations............................................................................... 19


iv
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
3.4 Flexural Behavior
3.4.1 Introduction .................................................................................................................... 20
3.4.2 Structural Mechanics Issues
3.4.2.1 Overview.................................................................................................................. 20
3.4.2.2 Traditional Solution (Euler or Simple Beam) .......................................................... 21
3.4.2.3 Shear Effects (Timoshenko Beam)........................................................................... 22
3.4.2.4 Non-Linear Effects
3.4.2.4.1 General............................................................................................................... 24
3.4.2.4.2 Beam-Column Equation..................................................................................... 25
3.4.2.5 Extension to Problems with a Subgrade................................................................... 27
3.4.3 Material Issues
3.4.3.1 Introduction.............................................................................................................. 29
3.4.3.2 Creep (Reduced Young's Modulus) ......................................................................... 29
3.4.3.3 Reduced Flexural Stiffness (Cracked-Section Behavior)......................................... 29
3.5 Dynamic Effects ................................................................................................................... 30

Section 4 - Soil-Structure Interaction Modeling:
Subgrade Aspects/Part I - Basic Concepts and Time-Independent Conditions
4.1 Introduction .......................................................................................................................... 31
4.2 Overview
4.2.1 Unifying Concept of Subgrade Models.......................................................................... 31
4.2.2 Three- Versus Two-Dimensional Models ...................................................................... 32
4.3 Three-Dimensional Models .................................................................................................. 32
4.4 Two-Dimensional Models
4.4.1 Overview........................................................................................................................ 33
4.4.2 Conventional Method of Static Equilibrium.................................................................. 34
4.4.3 Surface-Element Models
4.4.3.1 Background .............................................................................................................. 37
4.4.3.2 Overview.................................................................................................................. 38
4.4.3.3 Mechanical Models
4.4.3.3.1 Single-Parameter (Winkler's Hypothesis)
4.4.3.3.1.1 Basic Concept .............................................................................................. 39
4.4.3.3.1.2 Traditional Usage ........................................................................................ 40
4.4.3.3.1.3 Parameter Evaluation................................................................................... 40
4.4.3.3.1.4 Implementation into Structural Analysis Software...................................... 41
4.4.3.3.1.5 Deficiencies and Their Solution: An Overview........................................... 43
4.4.3.3.1.6 The Pseudo-Coupled Concept: Part I - Generic Concept ............................ 43
4.4.3.3.1.7 The Pseudo-Coupled Concept: Part II - The Discrete-Area Method........... 46
4.4.3.3.2 Multiple-Parameter
4.4.3.3.2.1 Overview ..................................................................................................... 47
4.4.3.3.2.2 Model Components...................................................................................... 48
4.4.3.3.2.3 The Beam-Column Analogy........................................................................ 51
4.4.3.4 Simplified-Continuum Models................................................................................. 52
4.4.3.5 Synthesis
4.4.3.5.1 Introduction........................................................................................................ 54
4.4.3.5.2 Implementation: Part I - General Concept ......................................................... 55
4.4.3.5.3 Implementation: Part II - Practical Application................................................. 57
4.4.3.5.4 Exceptions and Modifications............................................................................ 59



Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
v
Section 5 - Soil-Structure Interaction Modeling:
Subgrade Aspects/Part II - Time-Dependent Conditions
5.1 Introduction .......................................................................................................................... 63
5.2 Overview .............................................................................................................................. 63
5.3 Examples of Time-Dependent Effects in Surface-Element Models
5.3.1 Introduction .................................................................................................................... 64
5.3.2 Mechanical Models ........................................................................................................ 64
5.3.3 Simplified-Continuum Models....................................................................................... 65

Section 6 - Soil-Structure Interaction Applications: Overview
6.1 Introduction .......................................................................................................................... 67
6.2 Horizontal Structural Elements
6.2.1 Overview........................................................................................................................ 67
6.2.2 Foundations
6.2.2.1 Mats (Rafts) and Related Structures
6.2.2.1.1 Basic Application............................................................................................... 68
6.2.2.1.2 The Piled-Raft Concept ..................................................................................... 68
6.2.2.1.3 Anchored Foundations....................................................................................... 69
6.2.2.2 Slabs-on-Grade......................................................................................................... 69
6.2.2.3 Pavements ................................................................................................................ 70
6.2.2.4 Railways Track Systems .......................................................................................... 70
6.2.2.5 Tanks-on-Grade ....................................................................................................... 70
6.2.3 Geosynthetics
6.2.3.1 Introduction.............................................................................................................. 71
6.2.3.2 Reinforcement Function........................................................................................... 72
6.2.3.3 Compressible-Inclusion Function ............................................................................ 72
6.2.3.4 Combined Reinforcement and Compressible-Inclusion Functions.......................... 73
6.3 Vertical Structural Elements
6.3.1 Overview........................................................................................................................ 73
6.3.2 Two-Dimensional Applications...................................................................................... 74
6.3.3 Three-Dimensional Applications.................................................................................... 75

Section 7 - Soil-Structure Interaction Applications: Example for a Mat (Raft) Foundation
7.1 Introduction .......................................................................................................................... 77
7.2 Background
7.2.1 Introduction .................................................................................................................... 77
7.2.2 Structural Details
7.2.2.1 Overview.................................................................................................................. 77
7.2.2.2 Whitaker Laboratory................................................................................................ 78
7.2.2.3 Chemistry Building.................................................................................................. 79
7.2.3 Site Characterization
7.2.3.1 Overview.................................................................................................................. 80
7.2.3.2 Engineering Properties of the Boston Blue Clay ..................................................... 80
7.3 Analyses
7.3.1 Overview........................................................................................................................ 81
7.3.2 Subgrade Modeling
7.3.2.1 Models Used ............................................................................................................ 82
7.3.2.2 Model Coefficients
7.3.2.2.1 Single-Parameter (Winkler's Hypothesis).......................................................... 83
7.3.2.2.2 Multiple-Parameter ............................................................................................ 84
vi
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
7.3.3 Structural Assumptions .................................................................................................. 85
7.3.4 Computer Software
7.3.4.1 Overview.................................................................................................................. 86
7.3.4.2 SSIH Program Details .............................................................................................. 86
7.3.5 Results
7.3.5.1 Whitaker Laboratory
7.3.5.1.1 Observed Behavior............................................................................................. 87
7.3.5.1.2 Subgrade Modeling............................................................................................ 87
7.3.5.1.3 Structural............................................................................................................ 88
7.3.5.2 Chemistry Building
7.3.5.2.1 Observed Behavior............................................................................................. 89
7.3.5.2.2 Subgrade Modeling............................................................................................ 89
7.3.5.2.3 Structural............................................................................................................ 90
7.4 Conclusions
7.4.1 Subgrade Modeling ........................................................................................................ 90
7.4.2 Structural Assumptions .................................................................................................. 91
7.5 Suggestions for Practice ....................................................................................................... 92

Section 8 - References .................................................................................................................. 95

Section 9 - Notation.................................................................................................................... 103

Appendix A - Background Information re SSIH
A.1 Program Capabilities
A.1.1 Introduction ................................................................................................................. 105
A.1.2 Summary of Current Capabilities
A.1.2.1 General .................................................................................................................. 105
A.1.2.2 Structural ............................................................................................................... 106
A.1.2.3 Subgrade................................................................................................................ 107
A.1.2.4 Structure-Subgrade Coupling................................................................................ 108
A.2 Comments and Suggestions re Program Operation
A.2.1 Structural Modeling of Foundation Element ............................................................... 108
A.2.2 Subgrade Modeling ..................................................................................................... 109
A.2.3 Boundary Conditions
A.2.3.1 Structural ............................................................................................................... 109
A.2.3.2 Subgrade................................................................................................................ 109
A.3 Program Logic ................................................................................................................... 110

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
vii
List of Figures



Figure 2.1. Typical Mat Foundation Problem.................................................................................. 8
Figure 2.2. Problem Components: "Ideal" Analysis ........................................................................ 9
Figure 2.3. Problem Components: Traditional Analysis................................................................ 10
Figure 2.4. Problem Components: "Structural" and "Pseudo-Ideal" Alternative Analyses........... 12
Figure 2.5. Problem Components: "Geotechnical" Alternative Analysis ...................................... 13

Figure 3.1. Basic Components of a Beam Under Flexural Loading .............................................. 21
Figure 3.2. Beam Under Combined Transverse and Axial Loading.............................................. 24
Figure 3.3. Simple Beam and Beam-Column on a Subgrade ........................................................ 27
viii
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
This page left blank.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
ix
List of Tables



Table 4.1. Composition of Mechanical Models............................................................................. 49
Table 4.2. Hierarchy of Mechanical Models.................................................................................. 50
Table 4.3. Hierarchy of Simplified-Continuum Models ................................................................ 54
Table 4.4. Synthesis and Overall Hierarchy of Surface-Element Subgrade Models ..................... 56

Table 7.1. Subgrade Models Used in Horvath (1993c, 1993d)...................................................... 82
Table 7.2. Comparison of
o
W
k Values from Horvath (1993c, 1993d) .............................................. 83
Table 7.3. Values of k
W
(x) for the Whitaker Laboratory Using Liao's Pseudo-Coupled Method for
Cut-and-Cover Box Tunnel Base Slabs...................................................................................... 84
x
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
This page left blank.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
xi
The Manhattan College School of Engineering Center for Geotechnology
and Its Mission


The Manhattan College School of Engineering Center for Geotechnology (CGT) is a unique
organization that strives to be more than the typical academic research center or institute. It was
founded in 2001 at the personal initiative of Prof. John S. Horvath, Ph.D., P.E. of the Civil
Engineering Department who serves as its first Director. The CGT was the result of Prof.
Horvath's evolutionary realization after almost 30 years of geotechnical engineering practice that
the explosive growth in geotechnical and geoenvironmental engineering technology has made it
difficult for the engineering practitioner to keep abreast of new technical developments. The
traditional academic approach of simply publishing research results in narrowly disseminated
technical reports and papers (a philosophy of "if you print it, they will learn") has proven to be an
increasingly ineffective way of reaching practitioners and moving the state of art to the state of
practice. There is an ever-growing gulf between these two states of knowledge in geotechnical
engineering. The critical need for a total rethinking of how life-long continuing education is
achieved not only for engineering practitioners but academicians themselves is evidenced by the
appearance of "teach-the-teacher" training courses in drilled shaft foundations and geosynthetics
beginning in the late 1980s. If even academicians cannot keep up with new developments by
reading journal papers and conference proceedings, how can practitioners be expected to? The
stagnation of geotechnology also affects current engineering students and perpetuates the cycle.
The desirability of involving the practitioner in the process of formulating research programs so
that they may have a more direct and immediate benefit to practice is also something that is now
recognized more and more.
The CGT seeks to address the current need for effective, meaningful continuing education by
recognizing that the cycle of growth for any technology has three interdependent components,
what can be called the "trilogy of technology". Like a three-legged stool, each of these
components must be of equal length and strength if a given technology is to succeed. Thus the
CGT has adopted a holistic strategy of supporting geotechnology growth by recognizing the need
to concurrently address:

Technology advancement through research and development that involves not only the
engineering practitioner but also other end users of geotechnology such as construction
contractors and material manufacturers to the greatest extent practicable.

Technology transfer (T
2
) through education of engineers, contractors and manufacturers in a
multi-faceted, proactive way.

Technology documentation through standards development so that all end users
(practitioners, contractors and manufacturers) of a given technology work to a common set of
guidelines.

This trilogy-of-technology growth cycle is the cornerstone of all activities of the CGT. It is
important to note that the interaction of these three components is never completed but assumes a
constant cycle that leads to continuous growth of a technology. This concept is embodied in the
CGT logo of three interconnected arrows that is displayed on all CGT documents.
Another issue that the CGT seeks to address is the increasing specialization and
compartmentalization within civil engineering. This is a problem that affects numerous
professions and can only be expected to worsen as technology grows and a professional must
concentrate their education and practice ever more. Despite the inevitability of specialization,
xii
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
there are steps that can be taken to soften its effects by sensitizing and informing professionals of
the work of those in allied professions. Toward this end, the CGT seeks to involve other areas of
related professional activity such as structural engineering, environmental engineering and
science, and geology in both its research and, especially, technology transfer activities.
The CGT receives no direct financial support on a regular basis from Manhattan College. Thus
the success and growth of the CGT is totally a function of outside funding from individuals and
organizations whose philanthropic philosophies are consistent with the stated goal of the CGT to
treat technology growth in a more holistic fashion than is typically done in academia and
considers the entire process from research to standards with end-user input at all stages. In
addition, as part of its mission to promote technology transfer through education to the greatest
extent practicable the CGT is willing to partner with industry and other academic institutions not
only in research but also technology transfer and standards activities on any topic relevant to
geotechnical or geoenvironmental engineering. The new Manhattan College School of
Engineering William J. Scala Academy Room, which is located on the main floor of the Leo
Engineering Building and available for CGT activities, offers modern facilities for hosting
technology transfer activities. One benefit of Manhattan College's location on the northern edge
of New York City adjacent to both Interstate I-87 and mass transit is that it is quite accessible
(including free, off-street parking adjacent to Leo Engineering Building) from both within and
outside the City. When appropriate, the CGT will bring its technology transfer activities off
campus to meet the needs of a particular activity.
Regardless of financial support, the ultimate success and growth of the CGT will depend on its
being responsive to the needs of the engineering practitioner. Towards this end, the CGT
welcomes input from practitioners on an ongoing, continual basis. Suggestions for future research
and technology transfer activities based on perceived needs in practice are always welcome.
There is no topic that is too modest or simple for consideration. In fact, much of the research
conducted by Prof. Horvath since he came to Manhattan College in 1987 has been based on ideas,
large and small, that he developed as a result of his years in engineering practice. Additional
information about the CGT as well as access to published documents and other resources can be
found on the Internet at www.engineering.manhattan.edu/civil/CGT.html.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
xiii
Preface


The diversity and complexity of structures have increased continuously and significantly since
humans first created shelter for themselves thousands of years ago. A notable aspect of this
evolution is that design and construction changed from being based on precedence and individual
intuition and daring to the more rational and theoretically based endeavor of civil engineering.
Given the thousands of years of human existence, this change from art to applied science
occurred only in the relatively recent past. However, despite this fundamental change in how
humans design and build structures we still learn from what has preceded us. Thus insight into the
behavior of engineered structures, and the new or improved analysis, design and construction
methods that flow from such insight, always owes its roots to those who have researched, built,
occasionally failed and published before us.
In view of this, I believe that civil engineering research should be viewed as being
fundamentally evolutionary as opposed to revolutionary in its nature. The contributions contained
in this report, which are part of a long path of research I began in 1973 for my doctoral studies,
are no exception to this. In addition to the specific references used in my research and cited
throughout this report, there have been a few people who, over the years, provided insight or
assistance of a broad, often non-specific, but nevertheless crucial nature directly to me.
First and foremost, I acknowledge Mr. Stephen T. Mikochik, formerly Associate Professor of
Civil Engineering at the Polytechnic Institute of New York (now Polytechnic University) in
Brooklyn, New York, U.S.A. It was Prof. Mikochik who, from his knowledge of an obscure
technical paper written by a Dutch civil engineer named Loof, suggested the general topic of
improved subgrade models for mat (raft) foundation analysis that formed the basis for my
doctoral research and dissertation during the period from 1973 to 1979. Anyone who studied
geotechnical engineering at "Brooklyn Poly" during the 1960s to 1980s undoubtedly remembers
Prof. Mikochik as an intensely dedicated teacher with unique pedagogical habits and local
renown. So it is with no small pleasure that I note I was the first person to receive a doctoral
degree under Prof. Mikochik's tutelage and one of only a very few to do so before his sudden
early retirement from academia in the 1980s.
I also thank Dr. Samson S. C. Liao, P.E. for the many cordial hours of thought-provoking
discussion circa 1990 on the subject of subgrade models that helped me to focus, synthesize and
crystallize my diverse thoughts on the subject. Capable and knowledgeable researchers whose
interests parallel mine in the area of subgrade models and who have been willing to generously
share their knowledge without restraint have been few, so the input of Dr. Liao has been
particularly helpful to me. I am also grateful to Dr. Liao's employer at the time, Parsons
Brinckerhoff, Inc., for allowing me access to Dr. Liao's report of his original research into the
pseudo-coupled subgrade model concept as applied to box-tunnel base slabs.
Finally, I thank Dr. Nicholas F. Morris, P.E., Professor of Civil Engineering at Manhattan
College, for time spent discussing theoretical aspects of the finite-element method as applied to
structural analysis as well as several subtle analytical aspects of structural engineering. This
discussion, as well as his generous sharing of computer software developed by him, assisted me
in numerous general aspects of research and, in particular, aided in developing my SSIH finite-
element program for soil-structure interaction analysis. In addition, during his 1987-1995 term as
Chairperson of the Civil Engineering Department Prof. Morris was as supportive of my research
as the modest resources at Manhattan College would allow.

John S. Horvath, Ph.D., P.E.
Bronx, New York, U.S.A.
August 2002
xiv
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
This page left blank.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
xv
Executive Summary


Introduction

Soil-structure interaction (SSI) analysis using simplified mathematical models called subgrade
models to approximate the behavior of the ground under externally applied loads dates back to the
19
th
Century and the publication of the abstract mathematical expression we now call Winkler's
Hypothesis. This usage of subgrade models predates the evolution of modern geotechnical
engineering by decades and clearly indicates the interest in, and importance of, this topic to civil
engineers.
Unfortunately, the rapid advances in computational power that occurred in the late 20
th
Century
as a result of the personal computer have, in general, not been used to advantage to improve on
this now-ancient analytical methodology that is still used for SSI analyses in a wide variety of
applications. Geotechnical and structural engineers in practice still routinely use Winkler's
Hypothesis as the primary subgrade model in virtually all SSI applications even though it suffers
from the serious, fatal flaw of not replicating the vertical shearing that occurs within subgrade
materials. This shearing is the fundamental way in which load bearing and load transfer occur
within soils and other subgrade materials. Consequently, any mathematical model that does not
fundamentally duplicate this shearing even approximately has a significant, inherent
disadvantage.

Report Goals and Outcomes

While there was a time when crude, simplistic subgrade models such as Winkler's Hypothesis
were a pragmatic necessity in civil engineering practice, that time simply no longer exists and
never will again. Given the natural continuing evolution and improvement in technology, there
will, no doubt, come the day when subgrades can be routinely modeled as the true continua that
they are and using very sophisticated and accurate constitutive models as well. There has already
been decades of research and development along these lines using the finite-element and finite-
difference methods. However inevitable this "ultimate" subgrade modeling may be, it is still some
time in the future from the perspective of the civil engineering practitioner. Therefore, there is
still a need in the short- to medium term to improve upon the state of practice for SSI analysis.
This report is intended to be a contribution toward that goal.
This report begins with a thorough and fundamental review of both the geotechnical and
structural issues that affect the development of practice-oriented subgrade models. One of the
important points made is that the coefficient of subgrade reaction, k, is, theoretically, an observed
result or outcome of a SSI analysis. Thus this parameter is not a fundamental property of the
subgrade material and only becomes a required input parameter to a SSI analysis when Winkler's
Hypothesis is used as the subgrade model. This role reversal for the coefficient of subgrade
reaction, from calculated result to required input parameter, is the price paid for using the simple
but seriously flawed Winkler's Hypothesis. There is also a significant burden placed on the
engineer performing a SSI analysis as, in essence, the engineer must know the correct results
beforehand to be able to input the appropriate Winkler parameter values to be able to calculate the
correct results. The absurdity of this logic as well as the true meaning of the coefficient of
subgrade reaction and how it is distorted by Winkler's Hypothesis is something that has not been
properly discussed and emphasized in the published literature to date.
This report also summarizes research to date into what are called multiple-parameter subgrade
models that inherently include subgrade-shear effects and thus are fundamentally more advanced
and inherently more accurate than Winkler's Hypothesis. Multiple-parameter models can be
xvi
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
developed from both a mechanical approach using assemblages of springs, shear layers and other
physical elements, and a simplified-continuum approach that is based on the theory of a linear-
elastic continuum. Each methodology has its advantages and disadvantages. However, it turns out
that both approaches yield identical results so they can be synergistically synthesized into a single
methodology that combines the respective advantages. It is this synthesized approach to advanced
multiple-parameter subgrade modeling that offers the greatest promise for use in practice that is
the single most important outcome of this report.
The specific SSI application used to illustrate various concepts throughout this report is one of
the most common in practice, mat (raft) foundations. However, a section of this report is devoted
to a broad, comprehensive summary of actual and potential applications of subgrade modeling
and SSI analysis. This discussion is intended to serve as both a catalyst and resource document
for both the widest possible use of SSI analyses by practitioners as well as future research by
academicians.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
1
Section 1
Background and Scope of Report



1.1 SOIL-STRUCTURE INTERACTION

Many problems in civil engineering involve some type of structural element in direct contact
with the ground. Examples include building and bridge foundations; pavements and railway track
systems; earth retaining structures; and underground conduits and tunnels. When forces are
applied externally to the structural element and/or develop internally within the ground, both
problem components (structural element and ground) must deform and move in a compatible
manner. This is because neither the structural-element displacements nor the ground
displacements are independent of each other as a result of their intimate physical contact.
Therefore, these types of problems are broadly referred to as soil-structure interaction
a
(SSI)
problems, even when the ground (which is also referred to as the subgrade) is rock or a non-earth
material.
Strictly speaking, SSI is present to some degree in every problem where a structural element is
in contact with the ground. However, the current state of practice in geotechnical and structural
engineering still includes many instances where SSI is neglected (it was almost always necessary
to do so in the days before digital computers), and the structural element and ground are analyzed
independently of each other. This is done primarily for analytical simplicity because SSI analyses
are inherently statically indeterminate which means that, in addition to satisfying force and
moment equilibria, displacements must considered explicitly to rigorously solve any SSI
problem. While indeterminate analyses are easier than ever to perform given the current state of
computational hardware and software available to civil engineers, such analyses are still often
complex and time consuming, especially for combined structural and geotechnical systems (this is
illustrated in Section 2 of this report). Thus the pragmatic incentive to simplify the analyses and
not consider SSI for simple, routine applications is still very strong in practice.
In some cases, neglecting SSI is quite reasonable and is justified by experience based on
satisfactory performance of the resulting design. Examples include a footing foundation for a
low-rise building and simple rigid (gravity or cantilever) retaining walls. Unfortunately, there are
also cases where neglecting SSI is no longer justified based on the current state of knowledge and
computational technology yet is still done for any number of non-technical reasons (tradition,
ignorance, mental inertia). This can lead to designs that perform unsatisfactorily. Therefore, it is
important for a civil engineer to understand that SSI is always present for all types of subgrades
even if the engineer chooses to neglect it either explicitly or implicitly by virtue of an analytical
methodology that is employed. If SSI is neglected, there should always be sound theoretical and
practical bases for so doing.

1.2 SUBGRADE MODELS

In general, a SSI analysis requires a material (constitutive) model for the ground that can be
expressed mathematically. Such models are called subgrade models. Subgrade models represent a
unique type of constitutive model in geotechnical engineering because they are never intended to
be general or complete material models. Rather, the intent of a subgrade model is to strike a

a
Historically, some engineers limit the term soil-structure interaction to problems involving seismic
loading. This report adopts a broader definition that includes any type of loading as SSI is important to a far
broader spectrum of applications than simply those involving earthquakes.
2
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
balance between theoretical accuracy and ease of use in routine geotechnical engineering practice
when solving a particular SSI application. This goal is accomplished by crafting a subgrade
model that is very focused in its capability and only provides answers to what needs to be known
to solve a given SSI application. This derives from the fact that in most SSI applications, there are
a relatively small number of parameters that are required information in that application. So it is
quite acceptable to develop a material model for the ground that just provides the necessary
parameters, no more or less. For example, with mat (raft) foundations the design of the mat and
the superstructure it supports can usually be achieved solely by knowing the total settlements of
the mat (from which differential settlements are easily calculated) and bending moments within
the mat. Thus an acceptable subgrade model for mat problems is one that produces reasonably
accurate estimates of these two parameters (total settlement at the mat-subgrade interface and
bending moments within the mat) alone.
An additional requirement is that a subgrade model must be a pragmatic solution to a given
problem that can be used routinely by engineers in practice. Experience indicates that a
mathematically clever or elegant model that cannot be readily implemented into practice has no
immediate practical value and simply will not be used. SSI research during the past 30 years has
revealed numerous such models published throughout the 20
th
Century that have never progressed
beyond the technical paper or report in which they were first published. For example, as
discussed in detail in Section 2 one of the significant constraints of subgrade models in many SSI
problems, especially those dealing with mats and related structural elements such as the base
slabs of cut-and-cover box tunnels, is that they must be usable within the modeling constraints of
commercially available structural analysis software. These constraints turn out to be rather severe
with regard to the geotechnical aspects of the problem and, therefore, impact significantly on
subgrade modeling. For other types of structural elements such as anchored bulkheads or braced
excavations, application-specific software geared toward the geotechnical aspects of the
application can be and typically is developed which greatly relaxes the constraints imposed on
potential subgrade models. Therefore the practical utility of a subgrade model is often dependent
on the intended application. Thus it is perfectly reasonable to have a given subgrade model that is
practical in one application but not another.

1.3 REPORT

1.3.1 Philosophy

Considering further the two issues regarding subgrade models for SSI analyses in practice, i.e.
that they must produce the necessary output parameters yet have a practical solution capability, it
is of interest to note that the necessary parameters of a SSI application, e.g. total settlements and
moments for mats, generally remain the same over time. However, what constitutes a practical
solution for these parameters changes with time, often significantly. Thus the issue of practical
subgrade modeling is one that has and will continue to change with time concomitant with the
growth in computational power brought by the digital computer.
However, experience indicates that this issue is not so simple. The subgrade modeling aspect of
SSI is often one of the more controversial subjects of geotechnical and structural engineering
practice. This is because there is an inordinate tendency to resist change in this area and continue
using analytical methodologies that literally date back to the 19
th
Century. A consistent and
persistent argument for maintaining the status quo is that there is a real dearth of evidence that
using models that are now widely acknowledged as grossly incorrect has actually resulted in

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
3
structures that "failed"
b
. Experience clearly indicates that failure is generally the primary
motivator for civil engineers to improve the state of practice
c
. Consequently, it is important to
point out that an important philosophy adopted in this report and, indeed, for the Manhattan
College Center for Geotechnology's (CGT's) Soil-Structure Interaction Research Project in
general, is that past design without failure is not alone a justification for maintaining the status
quo. Expressed using an analogy, humans were able to survive thousands of years ago living in
rudimentary shelter and traveling only on foot. That does not mean that humanity should continue
to live the same way forever more. It is interesting to note that this sentiment was expressed in the
preface to a recent, practice-oriented special publication on mat foundations (Hemsley 2000):

"The design ofraft foundations undoubtedly represents one of the more difficult technical aspects of
civil engineering practiceUntil fairly recently, there was little alternative but to proceed on the basis
of greatly simplifying assumptions combined with rudimentary analysis. But although many such
designs were developed with remarkable success, the limitations of this traditional approach cannot be
disregarded and often are unacceptable in modern practice".

Nevertheless, the practical requirements of subgrade models are not forgotten in this report. The
intention here is not to publish yet another academic document containing results that will go
nowhere. Thus the approach to this report can be stated succinctly using two quotes found in
Wolf (1994). The first is:

"Make things as simple as possible but no simpler"

which is interpreted here as:

"Make things as simple as necessary but no simpler".

This means that subgrade models that were pragmatic and necessary in pre-computer days should
and can be replaced by more accurate (albeit more complex) models that are possible for routine
use given the tremendous increase in computational power brought about since the 1980s by the
personal computer. Therefore, an important objective of this study will be to critically evaluate
what constitutes the minimum acceptable level of analytical simplicity in the current
computational environment.
Second, simplicity in general will be guided by following the philosophy:

"Simplicity that is based on rationality is the ultimate sophistication".

Thus an important thread throughout this report will be to find a rational (i.e. theoretical) basis as
much as possible not only for the subgrade models themselves but also for the input values of the
variables that are part of these models. It is well known in geotechnical engineering that a simple
model with accurate input will always provide superior overall results compared to an elegant
model with poor input. This highlights the importance of site characterization in modern
geotechnical engineering practice. The importance of coupling site characterization and
geotechnical analysis is the focus of the CGT's Integrated Site Characterization and Foundation
Analysis Research Project (Horvath 2000a, 2000c, 2002).
In summary, the central theme of this report is to present improved subgrade models that have
some basis in theoretical rigor and whose variables can be evaluated in a rational fashion in an

b
It is important to remember that there are two broad ways in which a structure can fail, i.e. reach its limit
state. One is by loss of serviceability, usually the result of excessive displacement, e.g. settlement (the
serviceability limit state, SLS), and the other is a partial or total collapse (the ultimate limit state, ULS).
c
As it was for humans in general in the millennia before civil engineering.
4
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
effort to get engineers to abandon use of traditional models that are needlessly simple (and
inherently inaccurate as a result of that simplicity) in today's computational environment. The fact
that these improved models need to be used by both geotechnical and structural engineers is also
an important consideration.

1.3.2 Scope

SSI is a very large subject and cannot practically be covered by a single report. Therefore, this
report is intended to be but one of several that discusses various aspects of SSI.
To begin with, this report is limited to applications where SSI modeling is now generally
recognized as being necessary for an acceptable solution of the problem. Thus methodologies
used to analyze a structural element or the ground independent of each other are not considered
except where necessary to illustrate the historical evolution of a particular type of SSI problem.
The three principal components of this report are:

a detailed discussion of the basic concepts of SSI and subgrade models for use in SSI
analyses. This essentially reflects an updated synthesis of work performed and published
previously in numerous venues over a period of years. The primary goal here is to provide a
single, updated discussion of the topic that will hopefully be of use by both practitioners as
well as academic researchers. A discussion of relevant structural engineering concepts is also
provided.

an overview presentation and discussion of a wide range of SSI applications to illustrate
where various subgrade models have been or might be used. The intent here is to both inform
readers as to the current state of knowledge as well as hopefully stimulate new research by
others into various SSI applications that will extend this knowledge.

a detailed presentation of how various subgrade models can be applied to a particular
problem involving a mat foundation. The primary goal of this is to illustrate at least
qualitatively how calculated results can be sensitive to the subgrade model chosen as well as
other details involving structural engineering aspects of the problem.

More-detailed consideration of the various SSI applications is left to future CGT publications as
well as research and publication by others.

1.3.3 Organization

The remainder of this report is broadly divided into two parts. The first deals with the various
aspects of SSI modeling. This discussion is divided into several sections:

Section 2 - Soil-Structure Interaction Modeling: Overview. An illustration of several key
concepts and practical issues in SSI modeling by using the common problem of a mat
foundation under static loading.

Section 3 - Soil-Structure Interaction Modeling: Structural Aspects. A review of relevant
structural analysis theory used for both the structural element (e.g. mat) in contact with the
subgrade as well as the subgrade itself.

Section 4 - Soil-Structure Interaction Modeling: Subgrade Aspects/Part I - Basic
Concepts and Time-Independent Conditions. A summary, critique and synthesis of the

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
5
various mathematical models that have been used or might be considered for use as subgrade
models. This section focuses on the basic time-independent condition.

Section 5 - Soil-Structure Interaction Modeling: Subgrade Aspects/Part II - Time-
Dependent Conditions. This is a continuation of Section 4 and focuses on time-dependent
conditions.

The second part of this report focuses on applications. This part is divided into two sections:

Section 6 - Soil-Structure Interaction Applications: Overview. A summary of actual and
potential applications for SSI modeling.

Section 7 - Soil-Structure Interaction Applications: Example for a Mat (Raft)
Foundation. A detailed illustration using case-history data for two mat-supported buildings
of how various assumptions related to subgrade models as well as structural aspects can
influence calculated results. The capabilities and assumptions incorporated in the computer
software used to produce the results discussed in this section is discussed in Appendix A of
this report. This software incorporates some of the advanced multiple-parameter subgrade
models discussed in this report.

In addition, there are two sections that contain reference material applicable to the entire report:

Section 8 - References. A list of all references cited in this report.

Section 9 - Notation. A definition list of all mathematical variables used in this report.
6
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
This page left blank.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
7
Section 2
Soil-Structure Interaction Modeling: Overview



2.1 INTRODUCTION

To provide an overview of the kinds of topics considered in this report as well as to introduce
several key issues, concepts and terms concerning SSI that are used throughout this report, the
problem of a mat foundation is considered in this section in an introductory, qualitative manner.
Mats are one of the most common SSI applications encountered in practice and thus serve well as
an example. This fact is made use of in other sections of this report as well. Additional issues of a
general nature are also discussed in this section.

2.2 MAT FOUNDATION ANALYSIS

2.2.1 Parameters of Interest

As discussed in Section 1.2, a subgrade model is typically not a general constitutive model for
materials encountered in the ground. Rather, a subgrade model is generally the simplest
mathematical model that will produce acceptably accurate and precise estimates of the key
parameters for a particular SSI application. Therefore, the first step in selecting or developing a
subgrade model for a particular application is to identify and rank in order of importance the
parameters of interest as this provides guidance as to what elements need to be included in a
subgrade model for that application.
With this in mind, the accuracy of subgrade models for mat analysis should be judged based on
consideration of the parameters that are important to this type of structural element. As noted in
Section 1.2, bending moments in the mat and mat settlements (both total and differential) are of
primary importance in design. The mat-subgrade contact stress, which will be referred to in this
report as the subgrade reaction stress or simply subgrade reaction
d
as suggested by Liao (1991,
1995), is generally of secondary interest, as is shear within the mat.
In conclusion then, a key premise of selecting an appropriate, modern subgrade model for
analyzing mat foundations is that an acceptable model for this application should be capable of
producing reasonably accurate and precise estimates of at least moments and total settlements
from a single analysis. From this, differential settlements are calculated in a simple and
straightforward manner so that all necessary parameters are produced. This is a reasonable
expectation given the computational capabilities available today with digital computers and is
consistent with other types of structural analyses where accurate estimates of both loads and
displacements of structural members are expected from the same analysis. Yet this requirement
was not always the case for mat foundations. This can be understood from a discussion of the
components of a typical mat-foundation problem.

2.2.2 Problem Components

To illustrate the key concepts of mat foundation behavior, consider the common application
where a mat supports a superstructure consisting of some type of building as illustrated

d
Note that the term subgrade reaction as used in this report has a general and generic meaning, and is not
limited to or associated with a particular subgrade model (usually Winkler's Hypothesis) as has generally
been the case in the past.
8
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
conceptually in Figure 2.1. Note that there are three principal components of the problem: mat,
subgrade and superstructure.




Some comments are required with regard to this figure:

For simplicity in both illustration and subsequent discussion, the building superstructure is
shown without below-ground space so that the top of the mat is flush with the ground surface.
This is by no means typical nor does it affect the generality of the subsequent discussion.

The depth coordinate, z, is shown as having its origin at foundation level (bottom of the mat).
This is for mathematical convenience in developing subgrade models for this application
e
.

In general, mat flexure is significant in both horizontal (x- and y-axis) directions and should
always be analyzed for this. However, for simplicity in presentation and discussion in this
section of the report mat flexure in only the x-axis direction is illustrated in the various
figures (although x-y behavior is indicated mathematically).

2.2.3 Problem Solutions

2.2.3.1 Ideal

More than for any other type of foundation, a mat-supported structure represents a situation
where SSI is important and should always be considered. The reason is that the load-displacement
behavior of any one component (mat, subgrade or superstructure) is physically linked to, and thus
dependent on the behavior of, the other two. This means that, ideally, the mat-subgrade-
superstructure system shown in Figure 2.1 should always be analyzed a single problem to achieve
maximum accuracy of results.

e
In geotechnical engineering, the origin of the z axis (i.e. z = 0) is usually placed at the ground surface for
convenience in calculating overburden stresses and other depth-dependent problem parameters.
Figure 2.1. Typical Mat Foundation Problem
mat
subgrade
superstructure
(building)
x
z
y
foundation level


Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
9
Figure 2.2 illustrates conceptually how the combined mat and superstructure together are
actually a single structural system (megastructure) that is in contact with the ground. As a result
of some system of loads applied to the megastructure, there will be displacements, including
vertical downward displacement (settlement) at foundation level, w(x,y), into the ground.




In concept at least, the ability to analyze the system shown in figures 2.1 and 2.2 exists at the
present time. Both the megastructure and subgrade would have to be discretized and solved using
some type of numerical method such as the finite-element method. A three-dimensional model
would be required for most problems and the modeled portion of the subgrade would have to
continue for some distances in both the x- and y-directions beyond the limits of the foundation as
well as extending to some depth z below foundation level. A single piece of computer software
would be required. No such software is known to exist currently although both the structural and
geotechnical components exist in various commercial and research forms. A supercomputer
would likely be required to execute such software given the huge number of simultaneous
equations to be solved. A team consisting of both geotechnical and structural engineering
specialists would be required to prepare the input as well as interpret results. Note that the
parameters considered necessary for any mat foundation problem (bending moments and total
settlements) would be inherently produced as part of the solution of the overall system. There is
no guesswork or approximation involved in obtaining the necessary results.
Although an "ideal" solution such as shown in Figure 2.2 is technically achievable at the
present time and will undoubtedly become reality in the future, even after such a solution
becomes possible it will be many years before it is feasible for use in routine practice. If nothing
else, this is because a team of geotechnical and structural specialists would be required to develop
the model and its material properties, and properly interpret the calculated results. Therefore, it is
clear that more-practical and pragmatic alternative solutions are required, at least for routine
practice, for many years to come. Although such solutions would be inherently less accurate than
the ideal solution shown in Figure 2.2, they are a necessity. The goal, of course, is to "make
things as simple as necessary but no simpler".
Alternative solutions to the problem shown in Figure 2.1 are discussed in the following
subsections. First considered is the traditional approach which has its origins in pre-computer
Figure 2.2. Problem Components: "Ideal" Analysis
w(x,y)
superstructure
mat
megastructure

10
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
days. This is followed by a discussion of what is achievable at the present time within the context
of currently available technology.

2.2.3.2 Traditional

The true mat foundation problem illustrated in Figure 2.1 is inherently highly complex and
statically indeterminate, so prior to the availability of digital computers the problem was
traditionally broken into its three basic components as shown in Figure 2.3. The analysis and
associated design is then conducted in the following sequence of separate steps to facilitate
solution of the overall problem:



A structural analysis and design of the superstructure is performed for some system of
applied loads as per usual (for most structures, more than one load case would be
investigated) and, most importantly, assuming no differential settlement of the column
supports. This produces a system of reactions at the base of the superstructure (which is also
the top of the mat) that are depicted and referred to qualitatively as q(x,y) in Figure 2.3. Note
that q(x,y) is not necessarily uniform in magnitude or even continuous (it usually is not).

A structural analysis and design of the mat is performed using the applied loads, q(x,y), from
the analysis of the superstructure as a known input. The key aspect of this step is that the
subgrade reaction, which is depicted and referred to qualitatively as p(x,y) in Figure 2.3, must
be assumed or accounted for in some way. As discussed in Section 4, a magnitude and
distribution of p(x,y) might be assumed beforehand or some mathematical relationship could
be incorporated into the analysis itself so that p(x,y) is calculated as part of the analysis. Note
that p(x,y) is not necessarily uniform in magnitude although it is generally continuous.

A geotechnical settlement analysis is performed to calculate the pattern of mat settlements,
w(x,y). Two different conceptual approaches have been used over the years for this. One
school of thought is to use the subgrade reaction, p(x,y), calculated from the mat analysis as
the input to some settlement analysis. This is logical in that it links the mat and subgrade
behavior as occurs in reality. However, experience indicates that the assumptions made
Figure 2.3. Problem Components: Traditional Analysis
w(x,y)
q(x,y)
p(x,y)
q(x,y)
p(x,y)
mat
superstructure
foundation level


Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
11
regarding the subgrade reaction for the preceding mat analysis and design can, in many cases,
produce grossly incorrect results when it comes to settlements (too small by up to an order of
magnitude). Therefore the other approach employed is to use a totally different set of
assumptions for the settlement analysis. The produces a paradox of having different
settlement patterns for the mat and subgrade that are simply ignored in practice. Regardless of
the approach taken, although bearing capacity of a mat-supported structure is generally not
critical it should be checked, especially if the subgrade consists of fine-grain soil.

For the purposes of this discussion, the exact analytical methodology used for each component
of the solution is not important although it is of interest to note that analyses were frequently two,
not three, dimensional in nature. This was a pragmatic necessity in the pre-computer age.
Therefore, the structural analysis and design of the mat, for example, would generally be done
first in one transverse direction (x axis) and then the other (y axis) with the results simply
superimposed.
The most important point to note for the present discussion, and one that has rarely been noted
and appreciated in the past, is that for this traditional solution approach the subgrade reaction,
p(x,y), appears explicitly in two of the three steps as an external force whose magnitude must be
assumed or postulated mathematically in some manner at the beginning of the analysis. However,
in the true problem as shown in figures 2.1 and 2.2 p(x,y) is actually an internal force whose
magnitude is calculated as part of the solution process
f
. This role reversal for p(x,y) from
calculated result to input parameter, which turns out to have enormous influence on SSI analyses
considered in this report, is solely a result of the artificial problem decomposition shown in
Figure 2.3.
Because p(x,y) has such a crucial role as a known input parameter in the simplified analyses
used in practice such as shown in Figure 2.3, it turns out to be very useful to define a new
subgrade stiffness parameter, k(x,y), that is called the coefficient of subgrade reaction
g
and is
defined as follows:


) , (
) , (
) , (
y x w
y x p
y x k = . (2.1)

Note that k(x,y) is completely general and generic, and thus independent of a particular
subgrade model. It is simply the ratio of the subgrade reaction to subgrade settlement at a given
point with coordinates (x,y). Note also that in the ideal problem as shown in figures 2.1 and 2.2,
k(x,y) is a calculated result because it is the ratio of two calculated results, the subgrade reaction
and subgrade settlement. Thus, strictly speaking, in the ideal case k(x,y) is not an input parameter
whose magnitude must be stated or postulated mathematically beforehand. This fact will be used
subsequently as one way of evaluating the accuracy of various subgrade models where it is
necessary to assume k(x,y) as part of the analysis. By comparing assumed values of k(x,y) to
calculated values where possible, an objective assessment of subgrade model accuracy is
possible.

f
For the sake of completeness, it should be noted that the superstructure reactions, q(x,y), would also be
calculated internal forces in an ideal analysis and not appear explicitly as they do in this traditional
approach.
g
The term coefficient, not modulus, is preferred here as the dimensions of k(x,y) are force per length cubed.
Use of the term modulus, which is frequently done with this parameter, should be discontinued as modulus
is generally reserved for stiffness parameters with dimensions of force per length squared, e.g. Young's
modulus, shear modulus, etc. This distinction in terminology is not trivial or academic as it there is a
distinctly different parameter called modulus of subgrade reaction, with dimensions of force per length
squared, that was defined and used by Prof. Aleksandar Vesic. This parameter is discussed in Section 4.
12
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
2.2.3.3 Modern Alternatives

Even with the current wide availability of digital computers in routine civil engineering
practice, the traditional approach of decomposing a mat foundation problem into three separate
analytical components as shown in Figure 2.3 and discussed in the preceding subsection persists
to this day. This is totally unnecessary and not recommended as more-accurate alternatives exist
within the capabilities of routine practice. Therefore, the traditional approach is rejected as a
viable alternative for continued use in modern practice.
That leaves three alternative approaches than can be taken to solving this problem:

A "structural" alternative in which the superstructure and mat are combined into one
structural model (megastructure) as shown in Figure 2.4. This captures the "ideal" behavior of
the structural components as noted in Figure 2.2 as it allows for extremely accurate modeling
of all structural components and correctly accounts for superstructure-mat interaction.
Numerous independent studies (e.g. Burland et al. 1977, Soil-structure interaction 1989,
Ulrich 1991) have found the latter to be important. Ignoring this interaction is one of many
weaknesses of the traditional method shown in Figure 2.3. Analysis of the megastructure can
be accomplished easily, even in three dimensions if desired, with the structural analysis
computer software that is commercially available and currently used routinely worldwide.
The primary shortcoming of the structural alternative is that the subgrade reaction, p(x,y),
must either be assumed or modeled mathematically (the latter is almost always the case
nowadays). As discussed in detail in Section 4, unfortunately only a very simple and
relatively crude mathematical model for p(x,y) called Winkler's Hypothesis has been used
historically as a subgrade model with the structural alternative. Perhaps more importantly, the
value or values of the necessary coefficient for the Winkler Hypothesis subgrade model is
(are) typically chosen in a way that generally has no rational basis. As also discussed in
Section 4, there are ways to implement more-accurate subgrade models whose coefficients
can be determined rationally but still remain within the capabilities of commercially available
structural analysis software. This has not been pursued to any significant extent to date but
certainly offers promise for the future. Nevertheless, the primary weakness of the structural
alternative is and will likely always be the approximations inherent in the subgrade modeling.


Figure 2.4. Problem Components:
"Structural" and "Pseudo-Ideal" Alternative Analyses
w(x,y)
p(x,y)
p(x,y)
mat
superstructure
megastructure
foundation level


Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
13
The "geotechnical" alternative is to rigorously model both the mat and subgrade in one
analysis as shown in Figure 2.5. This allows very sophisticated and accurate constitutive
models to be used for the subgrade material(s) as the subgrade continuum is modeled
explicitly. The problem is that commercially available computer software that can analyze a
three-dimensional continuum representing the subsurface beneath a structure is not widely
available in routine practice (only two-dimensional software is). Even if such software were
widely available, the reaction loads, q(x,y), from the superstructure would have to be
calculated in a separate structural analysis, presumably assuming no differential settlement.
As noted in the preceding paragraph, mat-superstructure interaction is very important to
consider in most cases if the correct mat settlements and bending moments as well as correct
forces and moments within the superstructure are to be calculated. Thus the geotechnical
alternative has at least two significant shortcomings, namely the lack of ability to modeling
superstructure interaction effects directly
h
and the difficulty in performing a three-
dimensional geotechnical analysis in the first place.



What will be referred to as the "pseudo-ideal" alternative is similar to the structural
alternative shown in Figure 2.4 in that an accurate analysis of the mat+superstructure
megastructure is performed using commercially available structural analysis software. The
key difference is that even though the widely used, but crude, Winkler subgrade model is still
retained, the process for determining the values of the necessary model coefficient is
significantly improved so that the accuracy of the final results are significantly improved.
However, to accomplish this improvement requires the iterative use of a separate but parallel
geotechnical analysis that is fairly sophisticated (the geotechnical computer software used
must be capable of modeling the subgrade behavior relatively accurately). The subgrade
reaction, p(x,y), is the common variable linking the structural and geotechnical analyses.
These separate structural and geotechnical analyses are then performed independently in an
iterative, interactive fashion until the solution for each analysis converges to the same values

h
There are and have long been approximate methods to account for mat-superstructure interaction. This
typically involves increasing the flexural stiffness of the mat using some empirical relationship that depends
on the flexural stiffness of each floor, for example, within the superstructure. This approach is discussed
further for the case history examined in Section 7.
Figure 2.5. Problem Components:
"Geotechnical" Alternative Analysis
w(x,y)
q(x,y)
q(x,y)
mat
superstructure
foundation level

14
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
of p(x,y) within some acceptable precision.

A logical initial reaction might be that the last (pseudo-ideal) alternative would be the one of
choice as it offers the potential for modeling both the structural and geotechnical components of
the problem quite accurately. In addition, this approach has actually been used at least since ca.
1970 and apparently produced good results on a number of major mat-foundation projects
involving high-rise buildings in the U.S.A. (Ulrich 1991). However, the fact is that the accuracy
of the geotechnical component of the analysis is constrained by the same shortcoming as for the
second (geotechnical) alternative, i.e. the lack of readily available and easily usable three-
dimensional continuum-analysis software in practice
i
. In addition, the fact that the pseudo-ideal
method does not appear to have achieved any significant use outside of a limited geographic area
within the U.S.A. and by those who developed the method despite its being introduced at least as
far back as ca. 1970 suggest that it is simply not appealing and pragmatic for widespread use in
practice. Consequently, it is ruled out as a candidate for routine use in practice although it is
technically a viable alternative for those willing to work within its constraints.
Therefore, of the two remaining alternatives (structural and geotechnical), it is clear that only
the first (structural) approach of working with and within the modeling limits of commercially
available structural analysis software represents a reasonable and practical approach to the
problem. Consequently, it is the only one considered in detail in the remainder of this report. The
key issue, then, for SSI analysis of mat-foundation problems is subgrade modeling. Specifically,
the challenge is how to strike a balance between theoretical accuracy and simplicity to model the
subgrade reaction, p(x,y), shown in Figure 2.4 within the constraints of commercially available
structural analysis computer software. This is a topic that is pursued qualitatively within the
contents of Section 4. The quantitative results of this are illustrated in detail in Section 7.

2.3 ADDITIONAL COMMENTS

2.3.1 Site Characterization

Although this report focuses on the mechanics of subgrade modeling, it is important to note that
no geotechnical model, no matter how sophisticated, is of any use if the input parameters are
quantitatively significantly in error. Therefore, for the sake of completeness it is important to
emphasize the need for an appropriate geotechnical site characterization study for each project
where subgrade modeling and a SSI analysis is to be performed. As discussed in detail in Section
4, the most important aspects of site characterization for SSI analysis of mats and related
structural elements is generally overall stratigraphy and the variation of Young's modulus with
depth. Therefore, site characterization should utilize appropriate laboratory and in-situ testing
technologies as well as modern soil mechanics concepts such as yield stress
j
for both coarse- and
fine-grain soils and the effect on the appropriate Young's modulus to use.



i
The software used by Ulrich (1991) and his colleagues is not known with certainty. It is believed to be
capable of solving the problem of a three-dimensional linear-elastic continuum with zones or areas of
different uniform vertical stresses applied to its surface. Such software is known to have been among the
first developed commercially for geotechnical applications and operating on mainframe digital computers in
the 1960s. One pioneering example of such software was the ICES SEPOL (Integrated Civil Engineering
System Settlement Problem Oriented Language) package developed at the Massachusetts Institute of
Technology in Cambridge, Massachusetts, U.S.A. (Schiffman et al. 1970).
j
This parameter was formerly referred to by various names such as preconsolidation pressure/stress, natural
prestress and maximum past vertical effective pressure/stress.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
15
2.3.2 Computer Software

Some clarification of terminology related to a specific aspect of computer software is required
as it relevant to the contents of this report. This relates to the term finite-element method.
Experience indicates that geotechnical engineers, at least in the U.S.A., often interpret and use
this term colloquially to mean software that rigorously solves some two- or three-dimensional
continuum for force-displacement, fluid seepage, or heat flow behavior (force-displacement is by
far the most common and the one relevant to this report). However, the term finite-element
method properly defines a mathematical solution technique that can be applied to solve many
problems, not just those involving continua and not restricted to geotechnical engineering or even
civil engineering. Strictly and simplistically speaking, the finite-element method is just a way to
solve an algebraic equation. What behavior or physical phenomenon the equation defines and
what the specific application is are separate issues. For example, the finite-element method might
be used to model weather or ocean systems in the applied sciences. To avoid confusion,
throughout this report the term finite-element method is used only in its proper context of
defining a mathematical solution technique. The particular application or equation solved will be
defined clearly in each case.
One final comment on the subject relates to structural analysis. As should be evident from this
section of the report, and as will be seen more formally in the following section, structural
analysis figures significantly into SSI analyses. Therefore it is of interest to note that the
computer software most often used nowadays for structural analysis makes use of the finite-
element method to solve structural systems. However, experience indicates that in practice it is
more common for structural engineers to refer to the solution as matrix analysis.
16
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
This page left blank.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
17
Section 3
Soil-Structure Interaction Modeling: Structural Aspects



3.1 INTRODUCTION

Although the focus of this report is the geotechnical aspects of subgrade models, it is useful to
first discuss several aspects of structural modeling and analysis for a number reasons:

Structural elements are often used to help conceptually visualize or explicitly model
components of subgrade models. Consequently, the implications of various structural
assumptions are important for understanding subgrade model behavior.

Various assumptions concerning the behavior of the structural elements used to model the
mat and superstructure can be made and each affects the overall calculated behavior in SSI
applications. For example, in mat foundations the assumptions made concerning structural
behavior of the mat affects the behavior of the entire superstructure-mat-subgrade system due
to the inherent interconnection of the problem components as shown in Figure 2.3. This will
be illustrated in Section 7 for a case history involving two mat foundations.

In general, it is important to understand the various assumptions, constraints and limitations
of the structural modeling capability of commercially available structural analysis software.

3.2 STRUCTURAL ANALYSIS NOTATION

3.2.1 Introduction

There are consistent differences between the notation used by geotechnical and structural
engineers, at least in U.S. practice. Because of the close connection and working relationship
between these two civil engineering disciplines in SSI analyses, and especially because of the
focus of this report on how to implement subgrade models within the context of commercially
available structural analysis software, it is important to discuss these differences to minimize the
potential for misunderstanding and error.
The differences can be divided into two categories:

displacements, and

forces/stresses.

3.2.2 Displacements

With regard to displacements, both geotechnical and structural engineers define positive
displacements (+u, +v, +w) as displacements in the positive direction of the relevant coordinate
axis (x, y, z respectively), with the axes defined using the "right-hand rule". The differences are
how:

the positive sense of the vertical axis is oriented and

18
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
the axes are labeled.

Figure 2.1 illustrates what can be defined as the "geotechnical" convention. The vertical axis is
defined as positive downward and the z direction is assigned to it. Note, however, the somewhat
atypical usage shown in Figure 2.1 and carried throughout this report is that the z-axis origin is
set at foundation level. As noted in Section 2.2.2, this is more convenient for the purposes of this
report. More commonly, the ground surface (or sometimes a free-water surface if it is above the
ground surface) is used as the origin to facilitate calculating overburden stresses.
Referring again to Figure 2.1, the "structural" convention would be to have the vertical axis
defined as positive upward and labeled the y-axis. The origin would be at the bottom of the
structure which is the bottom of the mat in this case (the same as shown in Figure 2.1).

3.2.3 Forces/Stresses

With regard to forces (and, by implication, stresses), structural engineers generally follow
traditional solid-mechanics convention and define tension as positive. On the other hand,
geotechnical engineers traditionally define compression as positive. This is because, in general,
soil has no tensile strength so tensile forces and stresses are relatively uncommon. Defining
compression as positive is simply easier as it removes the need to continually use negative signs.
This report will follow the geotechnical convention.

3.3 STRUCTURAL ELEMENT WEIGHT

3.3.1 Introduction

As discussed in Section 6, the structural elements in contact with the ground in most SSI
applications have either a horizontal or vertical orientation. One issue that arises with horizontal
elements, especially those that are relatively thick such as a mat foundation, is whether or not to
include their weight in calculations. For most problems, even mats, this is a relatively minor
issue. Nevertheless, it should be dealt with correctly, especially for the occasional project where it
may have a noticeable effect on the calculated results.
A careful assessment of the problem suggests that there is no single correct answer to this
question. Rather, an evaluation should be made on an application- and project-specific basis. In
fact, two sets of analyses, without and with weight, may be required in some cases to evaluate the
sensitivity of the final results to this assumption. This will be illustrated using the example of a
mat.
The issues regarding mat weight can be grouped into two categories:

effect of mat weight on mat behavior, and

effect of mat weight on superstructure behavior.

In both cases, interaction with the subgrade is involved. As a result, there can be an effect on both
of the parameters, total settlement and bending moments, that are critical for mat analysis and
design.

3.3.2 Example: Effect of Mat Weight on Mat Behavior

Mat weight causes subgrade settlement which, in turn, may induce bending moments within the
mat. However, these moments will only develop after the portland-cement concrete (PCC) of the

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
19
mat has set and achieved at least some of its strength. Thus the time rate at which settlement
occurs is the governing factor. Thus settlements due to compression of coarse-grain soil or the
undrained (initial) settlement component of fine-grain soil are likely to occur while the PCC is
still fluid and not cause moments within the mat. On the other hand, the primary consolidation
settlement component of fine-grain soil develops over time and after the mat PCC has set. Thus
such settlements can create moments within the mat.
Note that the secondary compression (creep) component of settlement is assumed to be stress
(and, therefore, weight) independent, at least for fine-grain soils. Therefore, such settlements will
occur regardless of how much the mat weighs and should be considered for their effect on mat
moments as appropriate. There is insufficient evidence at the present time as to whether or not
creep settlement for coarse-grain soils is also stress independent. This is generally small in
magnitude in any event and, as a result, usually neglected in routine practice.

3.3.3 Example: Effect of Mat Weight on Superstructure Behavior

There are two opposing considerations:

Mat weight causes settlement as discussed in the preceding subsection. This may, in turn,
cause the superstructure to settle as well. Again, the primary consideration is the time rate at
which settlement occurs. For coarse-grain soils and the undrained (initial) settlement
component of fine-grain soils, this can be assumed to happen instantaneously upon load
application for analytical purposes. Thus the mat will settle under its own weight before the
superstructure is constructed and have no impact on superstructure settlement. On the other
hand, the primary consolidation settlement component of fine-grain soils will occur over
some period of time after the weight is applied and, at least to some extent, after the
superstructure is constructed. Therefore, in this case mat weight would have some effect on
superstructure settlement
k
.

Mat weight resists post-construction uplift forces. This includes porewater pressures from
ground water that would typically be distributed over the entire width and length of the mat
underside as well as concentrated forces from the superstructure due to wind or seismic
loading. Clearly, the beneficial effect of mat weight should be considered in these cases.

3.3.4 Conclusions and Recommendations

A decision about including foundation weight in a SSI analysis should take all relevant issues
into account. Using the example of a mat foundation, it may be logical to both neglect and
include mat weight in different loading cases of the same analysis. For example, assume a mat on
a sand subgrade. Settlement due to mat weight would reasonably be assumed to occur before the
mat PCC hardens appreciably and certainly before the superstructure is erected so it would be
logical to neglect mat weight for a gravity-load-case analysis. However, a load case involving
post-construction uplift loads from groundwater or wind/seismic loads on the superstructure
should logically include the mat weight. Consequently, not only for this example but as a
guideline for SSI analyses in general it may be easiest to always define the foundation element as
weightless and simulate the effect of its weight where necessary or desirable simply as an
equivalent distributed applied load across the foundation element for a particular load case.


k
The same comments concerning the secondary compression (creep) component of settlement for both fine-
and coarse-grain soils apply here as well.
20
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
3.4 FLEXURAL BEHAVIOR

3.4.1 Introduction

Overall, flexural (bending) behavior has the greatest impact on the structural modeling aspects
of SSI applications. This is regardless of whether the structural element physically exists as the
structural element in contact with the ground or whether it is used in a fictitious context to
represent a component of a subgrade model. Thus it is important for the overall goal of SSI
analysis to discuss various issues that relate to flexural behavior.
These issues can be broadly divided into two categories:

those related to theoretical structural mechanics issues that exist independent of the specific
material used for the structural element although the type of material behavior assumed
(linear-elastic, elasto-plastic, etc.) usually must be specified, and

those related to a specific materials used.

The discussion presented herein is divided along these lines. It is suggested that each of these
factors be considered as appropriate in practice.
For simplicity, most of the discussions here focus on uniaxial bending of a beam of unit width
and prismatic cross-section with a moment of inertia, I, that is composed of a linear-elastic
material with a Young's modulus, E. However, the concepts discussed and conclusions reached
are completely general so can be extended to biaxial bending of two-dimensional structural
elements such as plates. Although some general comments regarding plate behavior are included
in this discussion for the sake of completeness, the specifics of this extension are not covered in
this report as they are dealt with in other sources.

3.4.2 Structural Mechanics Issues

3.4.2.1 Overview

The behavior of a beam composed of a linear-elastic material and subjected to flexure is one of
the basic concepts in structural mechanics that is taught to all civil engineers as part of their
undergraduate education. However, there are a number of assumptions that can be made solving
this elementary problem that influence various aspects of SSI applications. For example, the
design of many mat-supported structures is governed by differential settlements of the
superstructure which, in turn, are influenced to varying degrees by the flexural behavior of the
mat. Thus anything that influences the flexural behavior and analysis of a mat should be clearly
understood by the design engineer so that the analytical capabilities of commercially available
structural analysis software are utilized to the fullest extent practicable and appropriate to a given
project.
The traditional method for portraying the mathematical load-deformation response of a beam in
uniaxial bending is a differential equation. While this will be used in this report, an additional
method of mathematical expression using matrix notation will be employed as well. This is
because matrix notation is particularly insightful into the subtle variations in beam behavior. The
basic form of the matrix formulation for beam flexure is

| |{ } { } q d S = (3.1)



Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
21
where:
[S] = stiffness matrix
{d} = displacement vector
{q} = load (force) vector.
The particular utility of Equation 3.1 is that all of the variations in beam behavior discussed in
this section of the report can be explained as variations solely in the formulation of the stiffness
matrix, [S]. This provides a concise and easily visualized methodology for understanding these
variations.

3.4.2.2 Traditional Solution (Euler or Simple Beam)

Figure 3.1 illustrates the basic components of a beam subjected to applied loads q(x). Note that
q(x) is not necessarily uniform or even continuous. Details regarding the beam supports (roller,
hinged, fixed or other) do not qualitatively affect the current discussion so are omitted for clarity.


In developing the traditional solution (called Euler or simple beam theory) for the displacement
of this beam in the z direction due to the transverse load q(x), three key assumptions and
approximations are always made:

The initial (undeformed) geometry of the beam is used. Nowadays in structural engineering
this is referred to as a linear analysis.

An imaginary vertical plane through the beam cross-section will remain planar (but not
necessarily vertical) after displacement of the beam. This is the classical "plane-sections-
remain-plane" assumption that is well known to civil engineers.

Vertical downward displacements (deflections) of the beam, w(x), are relatively small. This is
often referred to as small-deflection theory.

In addition, the flexural stiffness, EI(x), of the beam is usually assumed to be constant (= EI) for
simplicity although this is not required.
The resulting differential equation defining the behavior of a beam adhering to the above three
beam
Figure 3.1. Basic Components of a Beam Under Flexural Loading
x
z
q(x)

22
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
assumptions is the well-known equation

) (
) (
) (
2
2
2
2
x q
dx
x w d
x EI
dx
d
=
|
|
.
|

\
|
(3.2a)

which, assuming EI(x) = constant, becomes

) (
) (
4
4
x q
dx
x w d
EI = . (3.2b)

Using the stiffness-matrix concept stated in Equation 3.1, the flexural stiffness matrix, [S], for a
simple beam is


(
(
(
(
(
(
(
(
(
(

|
.
|

\
|
|
.
|

\
|
|
.
|

\
|
|
.
|

\
|
|
.
|

\
|
|
.
|

\
|
|
.
|

\
|
|
.
|

\
|

|
.
|

\
|
|
.
|

\
|
|
.
|

\
|
|
.
|

\
|
|
.
|

\
|
|
.
|

\
|
|
.
|

\
|
|
.
|

\
|
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
4 6 2 6
6 12 6 12
2 6 4 6
6 12 6 12
2 2
2 3 2 3
2 2
2 3 2 3
(3.3)

where l = beam length. Not surprisingly, the flexural stiffness of the beam, EI, dominates the
stiffness matrix.

3.4.2.3 Shear Effects (Timoshenko Beam)

One of the three key assumptions and approximations introduced during the formulation of
simple beam theory that has potential ramifications in SSI applications is the "plane-sections-
remain-plane" assumption. In reality, internal shear stresses develop within a beam during
bending. These stresses cause sections that are perpendicular to the longitudinal axis of the beam
and initially planar to warp as the beam displaces downward under load. This warp can be
visualized as horizontal displacements relative to a plane through the beam's longitudinal axis.
The result of this warping is that beam displacements are always somewhat greater than those
based on the traditional planar assumption. The additional component of displacement, i.e. the
magnitude of displacement over and above that estimated based on simple beam theory, is
referred to as shear deformations while the primary component of displacement is called bending
deformations (Timoshenko and Gere 1972). A beam analysis that considers both bending and
shear deformations (displacements) is sometimes referred to as a Timoshenko beam. Note that an
Euler (simple) beam only has bending deformations.
The theoretical influence of shear deformations can most easily be understood using the
flexural stiffness matrix, [S]. First, a new dimensionless parameter,
v
, that incorporates the shear
effects is defined as follows:


2
12
l A G
EI
v
v

= (3.4)


Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
23
where:
A
v
= shear area of the beam
l
,
G = elastic shear modulus of the beam
and the remaining terms were defined previously.
The flexural stiffness matrix, [S], incorporating shear effects can then be expressed as


( ) ( ) ( ) ( )
( )
( )
( ) ( )
( )
( )
( ) ( ) ( ) ( )
( )
( )
( ) ( )
( )
( )
(
(
(
(
(
(
(
(
(
(
(
(

+
+
(
(

+

(
(

+
(
(

(
(

+
(
(

(
(

+

(
(

+
+
(
(

+
(
(

+
(
(

(
(

+
(
(

+
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
l
EI
v
v
v v
v
v
v v v v
v
v
v v
v
v
v v v v
1
4
1
6
1
2
1
6
1
6
1
12
1
6
1
12
1
2
1
6
1
4
1
6
1
6
1
12
1
6
1
12
2 2
2 3 2 3
2 2
2 3 2 3
(3.5)

and Equation 3.1 still defines the behavior of the beam.
Comparing Equation 3.5 for a Timoshenko beam to Equation 3.3 which is for the traditional
simple beam, it is clear that the effect of shear stresses is to reduce the magnitude of most of the
coefficients in the stiffness matrix thus making the beam less stiff (more flexible) in flexure
which increases beam deflection under a given load.
It is important to understand that shear effects are always present in every beam and thus simple
beam theory is always approximate on the unconservative side (i.e. always underestimates) with
respect to beam deflections. However, both theory and experience indicate that simple beam
theory produces quite acceptable results for the majority of practical applications. Shear effects
become important primarily as the beam span-to-depth ratio decreases although the composition
and cross-sectional geometry of the beam influence results as well (Roark and Young 1975).
Although the focus here is on beams, it is important to note that there are many SSI
applications, including mat foundations, where the structural member is a plate which undergoes
bending in two directions. The same concepts regarding shear effects apply to plates under biaxial
flexure. In the case of plates, the terminology used is thin-plate theory (which is analogous to
simple beam theory in which shear effects are ignored) versus thick-plate theory (in which shear
effects are considered). Traditionally, thin-plate theory has been used for SSI analyses, in most
cases implied or by default as the issue of shear effects is not even discussed. The relatively few
studies to date of the effect of using thin- versus thick-plate theory for SSI analysis have involved
mat foundations. This work suggests that shear effects are not significant, at least for the
dimensions of mats used for buildings (Horvath 1993c, 1993d; Horvilleur and Patel 1995).
However, this should not be assumed for all mats or SSI applications in general. Recall that shear
effects are a function of not only the thickness of the flexural element but its span as well. Thus
there can be applications in which shear effects may become important in effectively increasing
mat flexibility. In any event, given the fact that it is relatively easy to specify "thick" plate
elements in commercially available structural analysis software it would seem reasonable to
simply always use such elements for all SSI analyses involving plates or, as a minimum, explore
the sensitivity of results of a given problem to the use of thin-versus thick-plate elements.

l
The shear area of a beam can be visualized or interpreted as the portion of the actual beam cross-sectional
area, A, that contributes to its rigidity in shear. Shear area is a function of the cross-sectional geometry of
the beam (Timoshenko and Gere 1972). For a rectangular cross-section as would be appropriate in many
SSI applications such as mat foundations, A
v
= 0.83A (Timoshenko and Gere 1972).
24
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
To conclude the discussion of shear effects, it is of interest to discuss the other limiting case
where bending deformations are insignificant and negligible in magnitude compared to shear
deformations. This would apply for a beam that was relatively very deep compared to its span
length, l. Such structural elements can actually exist in superstructures as what are often called
shear walls or shear beams. However, the primary reason for mentioning shear beams is that
shear-only structural elements have found significant use in advanced subgrade models used for
SSI analyses. It is in this latter context that shear beams will be discussed in detail in Section 4 of
this report.

3.4.2.4 Non-Linear Effects

3.4.2.4.1 General. The final structural mechanics issue discussed involves another of the
three key assumptions and approximations inherent in simple beam theory, namely that the
derivation of equations 3.2 is based on the initial, undeformed geometry of the beam
m
. The
implications of this can be illustrated by a simple example.
Consider the beam shown in Figure 3.2 which is identical to the one shown in Figure 3.1 but
with the addition of an axial force, P, that is defined as being positive in compression
n
. Under
simple beam theory, the axial force has no effect whatsoever on the flexural behavior of the beam
and only causes axial stress within, and concomitant axial strain of, the beam
o
. In fact, for a
simple beam the load P can be increased without theoretical limit (recall that linear-elastic
material behavior is assumed) and will never cause buckling of the beam
p
.

m
The third fundamental assumption and approximation of simple beam theory, small deflections, is not
discussed here. Experience indicates that this is acceptable in SSI applications used to date so there is no
need to use the more-complex large-deflection theory of flexural elements (beams and plates).
n
As would be expected, the same final overall results are obtained if P is assumed positive in tension.
Compression positive is chosen here as being consistent with the "geotechnical" sign conventions used in
this report.
o
Axial-only effects are sometimes called "truss" effects or behavior. This is because a concept analogous to
simple-beam theory (i.e. structural members that have flexural effects only) that is often used in structural
modeling is to have structural members (elements) that respond only to axial forces with no flexural effects.
This is the assumption made in basic or simple truss analysis for example. Flexural structural elements and
truss structural elements are the basic elements of linear analysis in structural engineering.
p
Note that the same concepts and result apply if the element in Figure 3.2 were oriented vertically as a
column.
beam
Figure 3.2. Beam Under Combined Transverse and Axial Loading
x
z
q(x)
P P


Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
25
In reality, an actual beam (or column) under combined transverse, q(x), and axial compressive,
P, loads will develop additional displacements, forces and moments not predicted by simple beam
theory and eventually buckle. This fact has long been recognized by structural engineers. In the
past, the behavior beyond that predicted by simple-beam theory was called the "P-" effect
because these additional displacements, etc. were caused by the axial compressive force, P, being
displaced transversely (vertically in Figure 3.2) through some distance (parallel to the z-axis in
Figure 3.2).
As noted in Section 3.4.2.2, a basic analysis performed using traditional simple-beam theory is
referred to nowadays as a linear analysis so a more-refined analysis performed considering the P-
effect is referred to as a non-linear analysis. Restated simply, linear theory is based on the
initial, undeformed geometry of a structure and non-linear theory takes into account the deformed
geometry.
It is of significant importance to this report to note that although most commercially available
structural analysis software is now capable of performing non-linear analyses (at least as an
option), this is generally not the default or basic version of such software. Much routine structural
analysis and design is still performed using only linear theory. Furthermore, the way in which
non-linear effects are typically considered in commercial software is by incrementally or
iteratively applying linear theory. There is, apparently, more than one solution algorithm for
dealing with non-linear effects (N. F. Morris, personal communication, ca. 1990). A discussion of
this topic is well beyond the scope of this report but it does indicate that the same problem
analyzed non-linearly by different software packages may produce different results.

3.4.2.4.2 Beam-Column Equation. Many civil engineers, especially geotechnical
engineers, call any structural member with both transverse and axial loads as shown in Figure 3.2
a beam-column. However, this broad, generic terminology is not strictly correct. This is because,
strictly speaking, the term beam-column only applies to a particular formulation of the problem
shown in Figure 3.2. Specifically, what constitutes a "true" beam-column and makes it unique is
not the combined transverse and axial loading per se. Rather, it is the fact that the formulation and
derivation of the differential equation defining the behavior of the beam-column is based on the
deformed geometry of the beam. This is a significant deviation from one of the three key
assumptions and approximations made for simple-beam theory. As a result, this makes the true
beam-column problem a non-linear problem in the context of the discussion in the preceding
subsection. Furthermore, if certain boundary conditions exist
q
it is possible to get the following
closed-form solution defining the flexural behavior of the beam in the true beam-column
problem:

) (
) ( ) (
2
2
4
4
x q
dx
x w d
P
dx
x w d
EI = + (3.6)

Comparing Equation 3.6 to the linear, simple-beam solution (equations 3.2), it can be seen that
the flexural effects of the axial force, P, on the deformed shape of the beam are reflected in the
second term on the left-hand side of Equation 3.6. Also, if P = 0 the beam-column equation (3.6)
reverts back to the simple-beam equations (3.2) as expected. However, the truly intriguing and
useful aspect of the true beam-column solution is that it turns what is fundamentally a non-linear
problem into a quasi-linear problem because it is possible to get a relatively simple, exact, closed-
form equation (3.6) defining the non-linear flexural behavior of the beam. Thus the inherently
approximate incremental or iterative numerical solutions normally required for non-linear

q
This primarily involves neglecting axial deformations of the beam due to the applied axial force, P. This
turns out to be quite acceptable for most SSI applications.
26
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
problems when linear theory is used as discussed in the preceding subsection are not required in
this case.
Another useful way to visualize the beam-column equation (3.6) is using the stiffness matrix
concept. If the shear effects that were discussed in Section 3.4.2.3 are neglected for simplicity, the
flexural stiffness matrix, [S], for a true beam-column is


(
(
(
(
(
(
(
(
(
(

|
.
|

\
|
|
.
|

\
|
+ |
.
|

\
|
+ |
.
|

\
|

|
.
|

\
|
+ |
.
|

\
|
|
.
|

\
|
+ |
.
|

\
|
+
|
.
|

\
|
+ |
.
|

\
|
+ |
.
|

\
|
|
.
|

\
|

|
.
|

\
|
|
.
|

\
|
+ |
.
|

\
|
|
.
|

\
|

15
2 4
10
6
30
2
10
6
10
6
5
6 12
10
6
5
6 12
30
2
10
6
15
2 4
10
6
10
6
5
6 12
10
6
5
6 12
2 2
2 3 2 3
2 2
2 3 2 3
Pl
l
EI P
l
EI Pl
l
EI P
l
EI
P
l
EI
l
P
l
EI P
l
EI
l
P
l
EI
Pl
l
EI P
l
EI
l
P
l
EI P
l
EI
P
l
EI
l
P
l
EI P
l
EI
l
P
l
EI
(3.7)

and Equation 3.1 can still be used to define the behavior of the beam-column.
Looking at Equation 3.7, note that from a practical perspective the effect of the axial force, P,
on the flexural behavior of the beam is to modify the flexural stiffness of the beam. A
compressive force (positive P in Figure 3.2) reduces all terms in Equation 3.7 and thus "softens" a
beam (i.e. makes it more flexible). On the other hand, a tension force (negative P in Figure 3.2)
increases all terms in Equation 3.7 and thus stiffens a beam (i.e. makes it more rigid).
Although not shown in Equation 3.7, shear effects as discussed in Section 3.4.2.3 and reflected
in the stiffness matrix for a simple beam (Equation 3.5) can also be incorporated into the stiffness
matrix for a beam-column. Note that shear has the same effect on a true beam-column as it has on
a simple beam, i.e. it effectively reduces the stiffness of the beam with regard to flexure. Thus
both shear and axial compression act in the same sense to reduce the overall apparent flexural
stiffness of a beam.
It is also of interest to note that the stiffness matrix for a beam-column provides an alternative
perspective for visualizing column buckling. Note that as P (defined as being positive in
compression as shown in Figure 3.2) increases in magnitude, at some point terms in Equation 3.7
become zero which means that the beam loses its flexural stiffness and becomes perfectly
flexible. As a result, the transverse displacements, w(x), will be infinite even for an infinitesimally
small load transverse load, q(x). In other words, the beam will undergo column buckling.
For the stiffness matrix shown in Equation 3.7, it can be seen that the critical magnitude, P
cr
, of
the axial force, P, that is required to cause buckling is


2
10
l
EI
P
cr

= . (3.8)

This compares favorably to the theoretical Euler buckling load of


2 2
2
9 . 9
l
EI
l
EI
P
cr

=

= (3.9)

for an axially loaded member with hinged ends which corresponds to the problem shown in
Figure 3.2 (Timoshenko and Gere 1961).
There are some interesting practical issues surrounding the beam-column equation. Although it
has historically received treatment in solid mechanics and structural engineering literature (e.g.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
27
Timoshenko and Gere 1961), it appears that the beam-column equation has seen more-extensive
use in practice (at least in the U.S.A.) not by structural engineers but by geotechnical engineers
who have used it for approximately 50 years to solve a wide variety of foundation problems that
involve SSI. Examples of applications include laterally loaded deep foundations
r
, flexible earth
retaining structures (primarily anchored and cantilevered sheetpile bulkheads), mat foundations
and subaqueous conduits (Haliburton 1971, Dawkins 1982, Horvath 1988a). More recently (ca.
early 1990s), the beam-column equation was used as a subgrade model (Horvath 1993e). This
latter application is discussed in more detail in Section 4 and used in the case history study
summarized in Section 7. Thus there is an interesting paradox that what is essentially a structural
mechanics solution is much more familiar to, and apparently used by, geotechnical engineers as
compared to structural engineers.

3.4.2.5 Extension to Problems with a Subgrade

Because this report focuses on structural elements bearing on a subgrade, it is necessary to
extend the beam and beam-column equations presented in the preceding subsections to
accommodate this. Figure 3.3 shows the general case of either a simple beam or beam-column
supported on a subgrade. The beam-subgrade contact stress (subgrade reaction) is denoted by
p(x). Note that p(x) is not necessarily constant although in most applications it will be continuous.
As before, the applied transverse load, q(x), is not necessarily constant or continuous.


The differential equations for a simple beam and beam-column for this problem are,
respectively,:

) ( ) (
) (
4
4
x q x p
dx
x w d
EI = + (3.10)

) ( ) (
) ( ) (
2
2
4
4
x q x p
dx
x w d
P
dx
x w d
EI = + + . (3.11)

r
The term "laterally loaded" is taken in this report to include both lateral (horizontal) and moment loading.
beam
Figure 3.3. Simple Beam and Beam-Column on a Subgrade
x
z
q(x)
P P
Note: P = 0 for a simple beam and P 0 for a beam-column.
p(x)

28
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
Alternatively, using the stiffness matrix formulation both the simple beam and beam-column
can be expressed using the same equation as only the stiffness matrix itself is different
s
:

| |{ } { } { } q p d S = + . (3.12)

where {p} is called the subgrade reaction vector.
At this point, the most important issue to appreciate concerning the subgrade reaction, {p}, in
Equation 3.12 is to reiterate that the focus of this report is SSI analyses that, for one reason or
another, need to be performed using commercially available structural analysis software and are
thus constrained by its capabilities
t
. Experience indicates that the primary difficulty in solving
Equation 3.12 within this context is that the subgrade reaction, {p}, represents a new problem
variable and, more importantly, one that current structural software is simply incapable of dealing
with explicitly
u
. Essentially, commercially available structural analysis software can only solve
Equation 3.1 and, at that, with certain limitations on the contents of the stiffness matrix, [S]. It can
certainly handle the basic linear (simple) beam (Equation 3.3) and a linear analysis with shear
effects (Equation 3.5) but reportedly not a true beam-column (Equation 3.7)
v
(N. F. Morris,
personal communication, ca. 1990).
Historically, this apparent dead end with regard to solving Equation 3.12 within the context of
commercially available structural analysis software has been circumvented rather cleverly by
using a very simple subgrade model, Winkler's Hypothesis. The attraction of Winkler's
Hypothesis in this context is that all the effects of {p} can be expressed in terms of the
displacements, {d}, alone. This means that {p} is eliminated as a variable from Equation 3.12 (or,
alternatively, p(x) is eliminated from equations 3.10 and 3.11) and Equation 3.1 can be used to
define the behavior of a simple beam or beam-column on a Winkler subgrade. In essence, all
subgrade effects appear simply as modifications to the various terms in the stiffness matrix, [S].
The details of how this is accomplished are discussed in detail in Section 4. But this ability to
essentially eliminate subgrade reaction explicitly from a problem certainly explains to a
significant extent why Winkler's Hypothesis enjoys such popularity to this day, at least for SSI
applications such as mat foundations which must be solved using commercially available
structural analysis software, even though it is widely known and acknowledged to be a
substantially inferior subgrade model
w
.
In summary, equations 3.1 and 3.12 can be used to succinctly state the primary goal of this
report. Specifically, the current challenge of subgrade modeling for SSI applications that use
commercially available structural analysis software is to be able to implement subgrade models
more accurate than Winkler's Hypothesis but within the form of Equation 3.1 and not Equation

s
Shear effects for either the simple beam or beam-column can be incorporated into the same equation as
well by using the appropriate stiffness matrix, [S].
t
This constraint applies to the majority of SSI applications in current practice and certainly the major ones
such as mat foundations where SSI analysis is a must. However, there are SSI applications, typically
involving laterally loaded deep foundations and earth retaining structures, where development of
application-specific software has either been done or is at least a feasible alternative. This report will also
serve the needs of these latter applications which do not have any constraint on the subgrade model that can
be incorporated in software development.
u
Fortunately, this rather significant limitation does not exist when application-specific SSI software is
developed.
v
Again, this is not a limitation when application-specific SSI software is developed.
w
Conversely, the continued use of Winkler's Hypothesis in application-specific SSI software that has no
restraints on subgrade models that can be used is even more curious and technically indefensible in this day
and age.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
29
3.12. This is the focus of Section 4
x
.

3.4.3 Material Issues

3.4.3.1 Introduction

Many, if not most, structural elements in contact with the ground in applications where SSI
analyses are performed are constructed of reinforced PCC. In SSI applications such as mats where
interaction with a superstructure must be considered reinforced PCC is often used for the
superstructure as well. There are significant material issues that involve the PCC itself as well as
the composite reinforced PCC. The issues impact on the flexural stiffness of the various
components of SSI applications so need to be discussed.

3.4.3.2 Creep (Reduced Young's Modulus)

It is widely recognized that PCC exhibits relatively significant creep after it has reached its
design strength, primarily as a result of continuing water-content changes within the material.
Creep of civil engineering materials is more the rule than the exception so this aspect of PCC
behavior is not surprising.
In general, a convenient way to both visualize creep as well as deal with it analytically is to
consider the Young's modulus, E, of a material undergoing creep as decreasing as a function of
time (Chambers 1984). For PCC, this equivalent modulus reduction is of the order of a factor of
two to three and will typically occur over a period of the first few years after construction. More-
exact details as to how to calculate this reduction for PCC can be found in Horvilleur and Patel
(1995).
While creep-related modulus reduction of PCC is routinely considered by structural engineers
for superstructure behavior, it appears to be much less often considered for mat behavior. There is
no reason why this should be as it easy to deal with (Horvilleur and Patel 1995). More
importantly, using the stiffness matrix concept for flexure it can be seen from Equation 3.3 that
the effect of reducing Young's modulus, E, is a reduction in each of the coefficients in the
stiffness matrix. This means that the structural element in contact with the ground, e.g. a mat
foundation, effectively becomes more flexible with time so for that reason alone differential
settlements of the mat and the superstructure it supports will concomitantly increase over time.
Consequently, ignoring creep of the mat PCC and the concomitant reduction in flexural stiffness
of the mat is potentially unconservative. This is because controlling differential settlement of the
superstructure is often the critical design criterion for a mat-supported structure.

3.4.3.3 Reduced Flexural Stiffness (Cracked-Section Behavior)

Another widely recognized material phenomenon is that reinforced PCC beams
y
will often
crack within the zone(s) of tensile-stress under service loads. As a result, the beam cross-section
has a reduced effective moment of inertia compared to its gross (uncracked) moment of inertia.
For a fully cracked section, a reduction in moment of inertia by a factor of approximately two is
common. Again, using the stiffness matrix concept for flexure it can be seen from Equation 3.3
that the effect of reducing the moment of inertia, I, is a reduction in each of the coefficients in the
stiffness matrix. Thus the primary result of a cracked beam is that it will be more flexible that an
uncracked one. This phenomenon is often referred to colloquially as cracked-section behavior.

x
The implementation of more-accurate subgrade models into application-specific SSI software that does
not have the same constraints concerning solution variables is also discussed and illustrated in Section 4.
y
The same issue discussed here applies to plates as well.
30
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
As with PCC creep, cracked-section behavior is assumed routinely in superstructure analysis
but apparently not in mat foundation analysis. There is no reason why this should not be done for
mats as well
z
. Horvilleur and Patel (1995) found that cracked-section behavior can have a
significant influence on calculated results for a mat, especially with regard to bending moments.
Consequently, ignoring cracked-section behavior of a mat is potentially unconservative as the mat
may be structurally underdesigned. In addition, using the stiffness-matrix concept it can be seen
from Equation 3.3 that the effect of a reduced moment of inertia, I, is a reduction in each of the
coefficients in the stiffness matrix which means that the mat is essentially more flexible so that
both mat and superstructure differential settlements will increase. Consequently, as with creep
ignoring cracked-section behavior of a mat is potentially unconservative for the superstructure as
differential settlement is often critical for its design.
In most cases, an analysis incorporating cracked-section behavior of the structural element in
contact with the ground would simply use the moment of inertia of the fully cracked section
which is easily calculated (Horvilleur and Patel 1995). However, it is worth noting that a more-
sophisticated analysis would be to model the progressive cracking of a reinforced PCC section
and the concomitant reduction in moment of inertia from its gross to fully cracked value. There
are empirical relationships such as Branson's equation to calculate the effective moment of inertia
as a function of developed moment. Such a progressive-cracking analysis would require
simulation of incremental loading as well as an iterative solution at each analysis step. Although
such an analytical effort is probably excessively exact for most SSI analyses in practice, it has
been applied to analyses involving mat foundations to at least illustrate the validity of the
technique (Horvath 1993c, 1993d).

3.5 DYNAMIC EFFECTS

The discussion in this section up to this point has focused on applications where the structural
element in contact with the ground is subjected to conditions where the applied loading is time-
independent or, based on experience, can be approximately assumed to be so. However, there are
SSI applications, typically involving seismic or other type of rapidly applied loading (which may
be either transient or steady state), where dynamic inertia effects on the structural element must
be considered. Depending on the stiffness of the structural element, it may be modeled as a rigid
or flexible element. This subject, especially for flexible structural elements, has received
considerable attention in the structural engineering and applied mechanics literature. However,
experience indicates that the dynamic behavior of structural elements has had little if any direct
application in subgrade modeling. Consequently, further consideration of this topic is beyond the
needs and thus scope of this report.

z
Two of the very rare documented applications of a cracked-section analysis to a foundation element, albeit
on a research basis, involved a spread footing (Kerr and Saxena 1977) and a mat (Horvath 1993c, 1993d).

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
31
Section 4
Soil-Structure Interaction Modeling: Subgrade Aspects/Part I -
Basic Concepts and Time-Independent Conditions



4.1 INTRODUCTION

This section of the report focuses on the various ways in which the subgrade reaction, p(x,y) or
{p}, can be modeled either by direct assumption or, more commonly, using a subgrade model.
Only time-independent conditions for both load application and subgrade response are considered
in this section. Overall, most SSI applications fall into this category. The discussion is extended
to time-dependent conditions for either the structural element or subgrade or both in Section 5.
For the reasons discussed in Section 2, the focus in this report is on subgrade models that are
feasible for the structural analytical alternative that was illustrated qualitatively for a mat
foundation in Figure 2.4. This means that any subgrade assumption or model used must be within
the modeling capabilities of commercially available structural analysis software. The primary
considerations are thus:

what variables can be solved for explicitly as discussed in Section 3.4.2.5 and

whether a linear or non-linear analysis is required as discussed in Section 3.4.2.4.

However, for the sake of completeness and to better understand the various approximations and
limitations of subgrade models proposed for use in practice, the entire spectrum of subgrade
methods is discussed. There are practical reasons for this:

One of the goals of the presentation and discussion in this section of the report is to show that
all subgrade models can be viewed or interpreted from a single, coherent perspective. One of
things that has created the very confused state of knowledge regarding subgrade models is the
perception that there are a multitude of unrelated models. In fact, every subgrade model that
is either known or has been hypothesized to date can be viewed as deriving from one,
unifying concept.

As noted in Section 3.4.2.5, there are numerous cases where application-specific computer
software has or might be developed for SSI analyses. In such cases there is no constraint on
subgrade-model variables as there is with structural analysis software. Thus there is no
constraint on potential subgrade models that can be used in such software.

4.2 OVERVIEW

4.2.1 Unifying Concept of Subgrade Models

As noted in the preceding subsection, one of the goals of this section of the report is to present
a unifying concept covering all subgrade models. Development of this concept is based on
recognition of the following key aspects of the problem:

The geotechnical behavior of most SSI applications is governed by the serviceability limit
state, SLS, of the subgrade as opposed to its ultimate limit state, ULS. Using the example of a
32
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
mat foundation, this means that behavior of the mat is governed by settlements as opposed to
a bearing capacity failure of the underlying ground.

In most SSI applications (e.g. see Figure 2.1 for a mat), the subgrade affecting the behavior of
the structural element is a three-dimensional continuum that can, with good approximation,
be taken to be a quasi-solid even though it is not a true solid.

Based on these two facts, it is reasonable to conclude that the subgrade in most SSI applications
can be viewed as an elastic continuum, although one that is not necessarily homogeneous,
isotropic or linear-elastic. Thus the unifying concept of subgrade modeling for SSI applications is
to view any subgrade simply as an elastic continuum (albeit one that can be highly complex in
behavior) with the various subgrade models each being a unique simplified approximation of that
continuum. The simplifications and approximations are introduced to render the problem solvable
with some reasonable level of effort but still capable of producing an acceptable level of accuracy
of calculated results. This logic is consistent with the philosophy of this report as stated in Section
1.3.1.

4.2.2 Three- Versus Two-Dimensional Models

Although the actual subgrade in SSI applications is always three dimensional, the first decision
with subgrade modeling is whether to model all three dimensions explicitly or only model two
dimensions (the horizontal x- and y-axis) explicitly, with depth (z-axis) variations accommodated
indirectly. Therefore, the first level of distinction between subgrade models is whether they are
two- or three-dimensional.

4.3 THREE-DIMENSIONAL MODELS

The obvious initial choice for a three-dimensional (3-D) subgrade model is a closed-form,
boundary value solution for an elastic continuum. Poulos and Davis (1974) provide a very
complete compendium of such solutions, most of them based on linear elasticity theory.
However, experience indicates that these solutions are based on problems that are too simplistic
and ideal to even approximate actual SSI behavior. More importantly, there is no mechanism for
incorporating such solutions into structural analysis software although it is feasible for
application-specific software.
The other alternative for true 3-D modeling is some type of computer-based numerical solution.
The simplest would involve dividing the subgrade into artificial layers or zones and then applying
theory of linear elasticity solutions to each layer or zone. By making use of the principle of
superposition that is valid for linear-elastic continua, the final result is simply the sum of the
individual results for each layer or zone. In this way, problems more complex than can be solved
in closed form can be handled. This was the basic concept of the legendary, landmark ICES
SEPOL software (Schiffman et al. 1970). With regard to SSI analyses, this has been done, for
example, for laterally loaded deep foundations (Poulos and Davis 1980).
However, nowadays it is more likely that a more-advanced numerical methodology would be
employed, typically using the finite-element method. This requires discretizing the continuum
into contiguous "bricks" (3-D elements); assuming that the stress-strain conditions within each
element are linear elastic but allowing the elastic parameters to vary between elements; then
solving the system of linear simultaneous equations generated. Such a numerical solution
probably represents the most accurate subgrade model practicable as relatively complex and
accurate rheological models can be used for the subgrade materials. However, as noted in Section
2.2.3.3 such software is difficult to come by and use in routine practice for a subgrade alone (this

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
33
is the "geotechnical" alternative discussed in that subsection) and would be even more difficult to
incorporate into structural analysis software (this is the "ideal" solution alternative discussed in
Section 2.2.3.1) although it remains the ultimate goal for the future.
In conclusion, it is clear that a true 3-D subgrade model is not feasible at the present time within
the context of structural analysis software. It is even at the edge of practicality for application-
specific SSI software which does not have the same constraints. This means that some type of
two-dimensional (2-D) model must be used.

4.4 TWO-DIMENSIONAL MODELS

4.4.1 Overview

By definition, 2-D subgrade models involve some mathematical expression that is stated only at
the interface (planar in the majority of SSI applications) between the structural element and
subgrade. Using the mat foundation shown in Figure 2.1 as an example, this planar interface is
foundation level which is the origin of the z-axis. In this case, the mathematical expression is
written as a function of the horizontal (x and y) axes. The primary challenge of 2-D models is to
incorporate the subgrade stratigraphy, material properties and their variations that occur with
depth (z axis) into the various terms of the mathematical expression.
Two-dimensional subgrade models can be separated into three broad categories. In order of
increasing approximation, they are:

the boundary element method (BEM) in which rigorous and relatively exact mathematical
functions are used. Fraser and Wardle (1976) provide an excellent example of the application
of BEM technology to mat foundations. Tomlinson (1986) also discusses application of the
BEM to mat foundations;

surface-element models (SEMs) which are similar conceptually to the BEM but involve use
of simpler and more-approximate mathematical functions to define the subgrade behavior;
and

the Conventional Method of Static Equilibrium (CMSE)
aa
. This is actually a direct
assumption and not a subgrade model in the strict sense of the term because only the
magnitude and distribution of the subgrade reaction, p(x,y) or {p}, is assumed. A separate,
theoretically uncoupled geotechnical analysis is required to estimate settlements. However,
for significant historical reasons the CMSE is considered a subgrade model in this report.

Experience indicates that the BEM is not a candidate methodology for routine use in practice
because the algorithms defining the subgrade behavior are very specific to the problem to be
solved. Consequently, such algorithms are not available in commercially available structural
analysis software and there are no indications that this will change in the foreseeable future.
Therefore, this leaves the SEM and CMSE as potential methodologies. In fact, these two
methodologies have historically formed the backbone of subgrade models used in SSI analyses.
Detailed discussion of the SEM and CMSE will begin with the CMSE. Even though it is the
least accurate of all 2-D subgrade analytical methodologies it is the one that was used most often
historically because of its relative simplicity. Therefore, it forms a useful basis of comparison for
all other subgrade models.
Much of the information on the CMSE and SEMs presented in this report was published

aa
Liao (1995) refers to the CMSE simply as the Conventional Method (CM).
34
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
previously in Horvath (1979, 1988d, 1989c). It is republished here both for the sake of
completeness as well as to present it with an updated, time-refined perspective.

4.4.2 Conventional Method of Static Equilibrium

The CMSE is the simplest and, for decades, was the most commonly used analytical method for
representing subgrade behavior in SSI analyses, especially for mat foundations. It is essentially an
extension of the assumptions that are made traditionally for footing foundations and will be
explained using the example of a mat foundation:

The vertical forces applied to the mat from the superstructure are assumed to be known
beforehand from a separate structural analysis. This analysis is almost always based on the
assumption of no differential settlement of the superstructure.

The mat is assumed to be rigid
bb
in the sense that it does not deform (change shape) but only
displaces as a rigid body and thus will never produce differential settlements of the
superstructure. In most cases, this means that the mat is assumed to settle uniformly. The
exception to this occurs when there is an eccentricity in the resultant of the vertical forces
from the superstructure with respect to the centroid of area of the mat in plan view. In this
case, the mat will rotate as a rigid body and thus settle non-uniformly but will still not
produce differential settlement of the superstructure.

The subgrade reaction, p(x,y) or {p}, is assumed to be completely independent of the flexural
stiffness of the mat and the composition of the subgrade.

The distribution of the subgrade reaction along the planar mat-subgrade interface is assumed
to be linear (in one dimension) or planar (in two dimensions). If the mat were loaded
concentrically, this means that the subgrade reaction would be uniform in magnitude beneath
the entire mat as is typically assumed for footing design. However, the resultant of all vertical
forces applied to a mat by the superstructure will generally not coincide with the centroid of
area of the mat in plan view (although it should be close to minimize tilting of the
superstructure) as it generally would for a column load on a spread footing. As a result, in
this case the magnitude of the subgrade reaction is not uniform across the bottom of the
foundation element. The magnitude and distribution of subgrade reaction is calculated in a
straightforward manner using classical structural mechanics concepts of static equilibrium for
an area loaded by both an axial load and moment, hence the name CMSE adopted many years
ago. The resulting subgrade reaction distribution in one dimension is thus trapezoidal in
shape.

Because both the applied forces and subgrade reaction magnitude and distribution are known,
the problem (at least with regard to bending moments within the mat) is rendered statically
determinate. As noted above, details concerning application of the CMSE can be found in
virtually every textbook on foundation engineering as well as in professional society committee
reports such as ACI Committee 336 (1988) and earlier publications (Horvath 1979, 1983b) so are
not presented here. The focus of this report is a critique of the CMSE methodology. However,
before discussing the CMSE further it is of interest to discuss several subtle aspects of its
interpretation that are generally not noted in texts and reports.
The above list of classical assumptions and interpretations of the CMSE is what might be called

bb
Consequently, many people refer to the CMSE simply as the rigid method.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
35
the "geotechnical" one as this is how geotechnical engineers routinely interpret the CMSE. This is
the interpretation that appears in foundation engineering textbooks. However, Liao (1995) noted
that, at least in some applications (he specifically identified box tunnels constructed by the cut-
and-cover method), there is an alternative "structural" interpretation of the CMSE that is
apparently used by structural engineers in at least some applications.
Using the base slab of a cut-and-cover box tunnel as an example, the specific assumptions of
this structural version of the CMSE are:

The subgrade reaction, p(x,y) or {p}, is always assumed to be distributed uniformly over
foundation level which in this case is the horizontal contact plane between the bottom of the
base slab and underlying subgrade. The magnitude of the subgrade reaction is simply the sum
of all the vertical forces acting at foundation level (tunnel dead and live loads, weight of soil
cover on the tunnel roof slab, surface live load if any above the tunnel roof slab) divided by
the foundation "footprint" (plan-view) area.

From a structural modeling and analysis perspective, the base slab is inverted and the
subgrade reaction treated as a uniformly distributed applied load. The forces and moments in
the tunnel box are then calculated using whatever structural analysis methodology is deemed
appropriate (typically computer software nowadays). In essence, the tunnel box is modeled as
a frame structure that is an inverted version of itself with a uniformly distributed load applied
across the top of the frame.

In summary, the primary difference between the geotechnical and structural versions of the
CMSE is that in the geotechnical version the superstructure loads are always calculated first and
used as input for the geotechnical portions of the analysis whereas in the structural version the
process is reversed. Also, in the geotechnical version the structural analysis is always rendered
statically determinate. This is not the case in the structural version of the CMSE. As a result, it is
possible that different structural designs for the same problem would be obtained from the two
versions
cc
.
Returning now to a critique of the much more commonly used geotechnical version of the
CMSE and again using the example of a mat foundation for descriptive purposes, there are a
number of important issues to point out:

As noted in Section 4.4.4.1, the CMSE is not a true subgrade model as it does not produce
any settlement estimate, although it does imply that there are no differential settlements.
Rather, the mat is viewed as a spread footing and settlement (there is only a single value as
there is for a footing) must be estimated separately using any one of several methods
developed for footings. Thus a SSI analysis using the CMSE alone does not provide all the
required results necessary to analyze or design a mat-supported structure, i.e. moments within
the mat are estimated but total settlements are not.

Virtually all mats exhibit at least some flexibility. Thus the fact that differential settlements
are always zero with the CMSE is always unconservative with respect to analysis and design
of the superstructure supported on the mat.

Experience indicates that the concept of a mat foundation essentially being a large footing has
created the unfortunate perception in many cases in practice that all that is required for design

cc
This difference in calculated results is what brought this situation of dual interpretations to Liao's
attention in the first place (S. Liao, personal communication, ca. 1990).
36
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
of a mat is some "allowable bearing pressure" as with footings. This is incorrect as footings
and mats are quite different in this regard. For a spread footing supporting a building column,
the footing dimensions in plan view are varied to match some "allowable bearing pressure"
that is based on consideration of several independent criteria:

o maximum allowable settlement under service loads,

o safety against bearing failure, and

o code requirements.

On the other hand, the plan dimensions of a mat are generally pre-determined by the
dimensions of the superstructure it will support. Thus the designer has relatively little control
over the plan dimensions of the mat and the magnitude of total settlement and subgrade
reaction. There is only modest control over differential settlements by virtue of varying the
thickness of the mat. Thus the designer must decide whether or not the total and differential
settlements are of acceptable magnitude for the particular superstructure in question. If not,
either the superstructure design must be modified, or a deep foundation or other alternative
such as a piled raft used.

Although virtually all mats exhibit some flexibility, in some cases it is modest and the actual
behavior is close to rigid. This was found, for example, for one of the buildings in the case
history discussed in Section 7. Nevertheless, even when a mat is relatively rigid with respect
to the underlying subgrade the assumption of a uniform or linear distribution of subgrade
reaction is simply incorrect. A broad spectrum of theory (Poulos and Davis 1974), numerical
modeling (Horvath 1993c, 1993d) and observation of actual foundations (Horvath 1989b,
Vesic and Johnson 1963) indicates that the subgrade reaction for a "rigid" foundation element
will always be non-uniform in distribution with well-defined maxima occurring near the
edges of the foundation element. Under such conditions, actual bending moments within the
mat will always be larger than those calculated on the basis of an assumed uniform or near-
uniform subgrade reaction as is done with the CMSE.

Increasingly, engineers acknowledge that bending moments calculated using the CMSE are
always incorrect. However, continued use of the CMSE is defended using the argument that
the calculated incorrect moments are at least always "conservative" and thus the error always
results in an overly "safe" design at worst. However, this argument itself is incorrect and thus
indefensible because mats always have both positive and negative moments. Experience
indicates clearly that CMSE-based moments are not "conservative" (i.e. larger in magnitude
compared to an actual or analytically more correct value) for both signs. While they are often
conservative for one sign they are not for the other. This means that a mat designed using
results from a CMSE analysis will be underdesigned from a building code perspective.

In conclusion, the CMSE is unacceptable for SSI analysis because it does not and cannot
provide estimates of total settlement. Furthermore, the CMSE always produces incorrect results
on the unconservative side for the two primary results required for mat design: differential
settlements and bending moments.





Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
37
4.4.3 Surface-Element Models

4.4.3.1 Background

The shortcomings of the CMSE were apparently recognized fairly early in the evolution of
modern geotechnical engineering which had begun in Europe in the 1920s and spread globally by
the 1930s
dd
. Surface-element models (SEMs), or at least the simplest one (Winkler's Hypothesis)
which had been postulated in the 19
th
Century
ee
, must have been sufficiently well known and used
as an alternative to the CMSE by the early 1950s to inspire Terzaghi to write his landmark paper
on how to estimate values of the parameters for such models (Terzaghi 1955)
ff
. This was long
before commercially available digital computers had reached even the most technologically
advanced academic institutions which meant that only simple closed-form solutions for
applications of Winkler's Hypothesis such as published by Hetnyi (1946) were available for use.
As discussed subsequently, Terzaghi's 1955 work is still cited and used in textbooks as well
engineering practice. However, it appears that few appreciate some of the important assumptions
made by Terzaghi concerning the use of Winkler's Hypothesis in SSI analyses:

He saw it solely as a methodology for producing estimates of bending moments that were
presumed to be closer to reality than those obtained using the CMSE because flexibility of
the structural element in contact with the ground was allowed for. In fact, in the conclusion to
Terzaghi (1955) he stated explicitly:

"the theories of subgrade reaction should not be used for the purpose of estimating
settlement or displacements."

This reflected the continued use of the traditional uncoupling of structural and geotechnical
SSI analyses that had begun with the CMSE, even if this leads to inconsistent results. There
are numerous examples in practice where a structural SSI analysis using classical (i.e.
Terzaghi) values for the Winkler's Hypothesis model coefficient yields estimated mat
settlements that are a fraction of one inch (several millimetres) while a geotechnical SSI
analysis using some other calculation methodology yields estimated settlements that are an
order of magnitude larger. Actual settlements tend to be much closer to the "geotechnical"
settlements than the "structural". The case history discussed in Section 7 illustrates just such
an occurrence for two actual mat-supported buildings.

Even with regard to bending moments, Terzaghi separated them into "local" and "global"
components
gg
. A local bending moment is one resulting from an individual, concentrated
"point" load acting on a mat and producing localized settlements of the mat in the vicinity of
the load. The horizontal portion of a mat affected by a single applied point load is relatively

dd
The first international geotechnical conference, which was held in 1936, is a useful benchmark.
ee
In an amusing example of claims and counterclaims so typical of the Cold War era in the latter half of the
20
th
Century, most "Western" researchers credit what is now known as Winkler's Hypothesis to E. Winkler
who reportedly presented his ideas in a book titled "Die Lehre von der Elastizitt und Festigkeit" that was
published in 1867 (Hetnyi 1946, Nair and Chang 1970, Rhines 1965). However, researchers from the
former Soviet Union (Vlasov and Leont'ev 1960) claimed that the concept was first proposed by the
Russian academician Fuss in 1801.
ff
This was not Terzaghi's first English-language published work on the subject however. He addressed the
topic but using a totally different, more-theoretical approach in his first English-language textbook
(Terzaghi 1943). Curiously, he did not mention his earlier work at all in his 1955 paper.
gg
There are corresponding local and global components of settlement even though this was not of explicit
interest to Terzaghi as noted above.
38
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
limited (Terzaghi called it the range of influence
hh
). However, because mats will have
multiple point loads their effects will overlap at some depth below foundation level due to the
inherent "load spreading" that occurs in the ground. This will produce a more-or-less uniform
vertical stress increase at that depth and below. This stress increase will produce additional
settlements that typically produce an overall concave-upward (dishing or sagging) settlement
pattern for the mat as a whole. The global bending moment component is due to this overall
pattern of settlement. Terzaghi clearly saw the role of Winkler's Hypothesis as being used
solely to estimate the local component of bending moment. A possible explanation for this
narrow usage is the apparent perception among at least some structural engineers that it is
more conservative to design for local moments only because dishing is assumed to increase
the overall radius of curvature of the foundation element which reduces the total moment
which is the sum of the local and global components (DeSimone and Gould 1972).

The position taken in this report is that this very narrow use of SEMs for calculating the local
component of moments only is not defensible within the context of modern SSI analyses. To
begin with, Landva et al. (1988) presented a well-detailed mat case history in which they discuss
the effect that the dishing-induced global moment component had on total moments. Their results
suggest that it is not always conservative to design based on the local moment component only. In
addition, the broader position adopted in this report is that an acceptable SEM must be capable of
producing accurate estimates of both the total bending moments in as well as total settlements of
the foundation element.

4.4.3.2 Overview

The traditional way of constructing or at least visualizing SEMs that dates back to the 19
th

Century is as assemblages of mechanical elements such as springs, flexural elements (beams in
one-dimension (1-D), plates in 2-D), shear-only layers and deformed, pretensioned membranes.
Such subgrade models will be referred to as mechanical surface-element models or, more simply,
mechanical models. As will be seen, the evolution of mechanical models through the 19
th
and 20
th

centuries started with the simplest and then become more complex with time.
On the other hand, in the latter half of the 20
th
Century an alternative approach to developing
SEMs evolved in which the starting point was the three complex sets of partial-differential
equations (compatibility, constitutive, equilibrium) governing the behavior of a linear-elastic
continuum. Simplifying assumptions were then applied to these equations to yield a SEM. These
will be referred to as simplified-elastic-continuum surface-element models or, more simply,
simplified-continuum models.
The evolution of mechanical and simplified-continuum models is discussed separately in
Sections 4.4.3.3 and 4.4.3.4 respectively. This is appropriate as these two families of SEMs
evolved surprisingly independently. Until the 1960s, the published literature in one category
typically did not mention the other. This presentation of mechanical and simplified-continuum
models is followed by Section 4.4.3.5 in which the advantages and disadvantages of each
approach to subgrade model building are discussed, and a synthesized, unified approach that
draws on the strengths of each is presented. This synthesis is an updated and improved
presentation of work presented originally in Horvath (1979). In many ways this synthesis is the
most important single contribution of this report.





hh
Terzaghi (1955) and others presented equations for calculating the magnitude of this parameter.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
39
4.4.3.3 Mechanical Models

4.4.3.3.1 Single-Parameter (Winkler's Hypothesis)

4.4.3.3.1.1 Basic Concept. Regardless of who actually first proposed it, what is now
universally referred to as Winkler's Hypothesis was initially expressed as an abstract
mathematical statement. Only later were physical models proposed to provide a visualization of
what was expressed mathematically.
In its basic form, Winkler's Hypothesis assumes that the settlement, w
i
, at an arbitrary point i on
the subgrade surface is caused only by the applied vertical normal stress (subgrade reaction) at
that point, p
i
. Furthermore, p
i
and w
i
are linearly related. Mathematically, this is expressed as


i W i
w k p
i
= (4.1)

where
i
W
k is defined as Winkler's coefficient of subgrade reaction at point i
ii
. Note that Winkler's
Hypothesis is what is called a single-parameter subgrade model because only one parameter (the
Winkler coefficient of subgrade reaction,
i
W
k ) is necessary to define its behavior.
This parameter is often referred to as the "soil spring constant" or a similar term because one
physical interpretation of the abstract behavior defined by Equation 4.1 is a spring (not
necessarily linear or elastic but usually assumed so) that is oriented perpendicular to the subgrade
surface. Another physical interpretations of
i
W
k is the unit weight (not density, as is often stated,
to be dimensionally correct) of a liquid
jj
. However the spring analogy is by far the more common
and enduring, and is the one adopted in this report. For an arbitrary number of contiguous points
over the subgrade surface, the general form of Winkler's Hypothesis is


Upon initial comparison, it might appear that Equation 4.2 is simply a different version of
Equation 2.1 which defined the generic parameter, k(x,y), the coefficient of subgrade reaction.
However, this is not correct as there are some subtle points implied by Winkler's Hypothesis in
general and Equation 4.2 in particular:

The settlement of any given point on an actual subgrade surface is influenced by the applied
pressure, p(x,y), at all points on the subgrade surface. For a Winkler subgrade, the settlement
at a given point is assumed to be caused solely by the applied pressure at that point as defined
in Equation 4.1.

ii
In most publications, Winkler's coefficient of subgrade reaction is referred to simply as the "coefficient of
subgrade reaction". The term "Winkler's coefficient of subgrade reaction" is used throughout the remainder
of this report to avoid confusion with broader, generic coefficient of subgrade reaction that was discussed in
Section 2.2.3.2 and defined in Equation 2.1. In addition, most publications use the notation of simply k for
Winkler's coefficient of subgrade reaction. However, in this report the subscript W will always be used
when referring to Winkler's coefficient of subgrade reaction to avoid confusion with the broader, generic
coefficient of subgrade reaction k. Finally, the terms "coefficient of subgrade reaction" and "modulus of
subgrade reaction" in any context are often used synonymously. Only the term "coefficient of subgrade
reaction" is correct here for the reasons given in Section 2.2.3.2.
jj
Curiously, references that use the liquid analogy almost always refer to it as a "dense liquid" as if there
were a concern that the foundation element would "sink" unless the liquid were sufficiently dense so that it
"floats". From a purely mathematical perspective, the unit weight of the fluid in this case is irrelevant as
long as it is greater than zero.
y) w(x y x k = y) p(x
W
, ) , ( , . (4.2)
40
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
As discussed in Section 2.2.3.2, for a foundation element supported on a real subgrade the
coefficient of subgrade reaction, k(x,y), is an observed or calculated result of the problem.
However for a foundation supported on a Winkler subgrade, the Winkler coefficient of
subgrade reaction, k
W
(x,y), is an input parameter whose magnitude as a function of x and y
must be defined, and thus known, beforehand. This role reversal for the coefficient of
subgrade reaction from problem result to problem input is the penalty for using such a simple
model to represent material as complex a soil. The key element of simplicity inherent in
Winkler's Hypothesis is the assumption stated in Equation 4.1 that there is a one-to-one
relationship between applied pressure and settlement at each point beneath a foundation. This
role reversal from an observed result to a known input is an issue that has not been
emphasized in the past and, as will be seen subsequently, turns out to be quite important.

4.4.3.3.1.2 Traditional Usage. Equation 4.2 represents the general form of Winkler's
Hypothesis. Note that there is nothing that requires the Winkler coefficient of subgrade reaction
to be constant in either the x or y direction. Nor is there any reason why the value of k
W
(x,y) at a
given point cannot vary nonlinearly and/or inelastically as a function of the applied stress.
However, a value of k
W
(x,y) that is independent of both position and load magnitude was assumed
traditionally, at least in problems involving horizontal foundation elements like mats. This
evolved in pre-computer days when the only solutions (closed-form) available for a Winkler
subgrade (Hetnyi 1946) were based on a constant value of the Winkler coefficient of subgrade
reaction.
For such an assumption, Equation 4.2 becomes

y) w(x k = y) p(x
o
W
, , (4.3)

where
o
W
k is a constant. This special case of Winkler's Hypothesis with a constant value of the
Winkler coefficient of subgrade reaction is what has been found historically in textbooks even to
this day as well as in reports of recommended practice by technical society committees (ACI
Committee 336 1988).

4.4.3.3.1.3 Parameter Evaluation. Historically, one of the most commonly asked
questions in geotechnical engineering is how to determine an application and project-specific
value for
o
W
k in Equation 4.3. In particular, there has always been in an interest in correlating
o
W
k with other geotechnical engineering parameters ranging from the theoretical (e.g. Young's
modulus) to the empirical (e.g. SPT N-value). Note that this question in all its parts implies that
the Winkler coefficient of subgrade reaction is a fundamental material property of subgrade
materials that can somehow be determined rationally.
The fact is that this implied assumption is simply incorrect and thus the search for the "correct"
value of
o
W
k is always, and will always be, an exercise in futility. This is because Winkler's
coefficient of subgrade reaction, even in its most general form as defined in Equation 4.2, is not a
fundamental soil property. As discussed in Section 4.4.3.3.1.1, what turns out to be the fatal flaw
in Winkler's Hypothesis is its fundamental assumption as expressed in Equation 4.1 that the
displacement at some point has a unique cause-and-effect relationship with the applied stress at
that point. This is contrary to all knowledge, both theoretical and observed, as to how actual
subgrades behave.
Additional proof as to why a "correct" value of
o
W
k simply does not exist is that Winkler's
Hypothesis with a constant value of the Winkler coefficient of subgrade reaction does a poor job
of replicating actual subgrade behavior. This has been demonstrated both for actual foundation

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
41
elements (Horvath 1988b, 1988c, 1993c, 1993d; Liao 1995; Vesic and Johnson 1963) as well as
the very simple idealized limiting cases of a perfectly flexible or perfectly rigid foundation
element (Horvath 1979, 1983a). In all cases, the "true" coefficient of subgrade reaction, i.e. the
observed result as defined by Equation 2.1, is found to vary horizontally beneath the foundation
element. In addition, the actual coefficient of subgrade reaction varies depending on the stiffness
of the structural element (if any) placed between the applied load and subgrade.
The extensive work by the late Prof. Aleksandar Vesic and his co-workers offers insight into
this futility of determining
o
W
k values from another perspective. In particular, Vesic's relatively
obscure published work with plate-type foundation elements
kk
(Vesic and Saxena (1970), with a
more accessible abstract in Scott (1981, pp. 161-162)) determined that:

In general, three separate values of
o
W
k are required in a given problem to match bending
moment, settlement and subgrade reaction between Winkler's Hypothesis and an isotropic,
homogeneous elastic half space directly beneath the applied load. The range in magnitude
between the largest and smallest
o
W
k values is of the order of six.

For the case of a perfectly elastic plate subjected to a point load, the match for any given
calculated results (settlement, bending moment or subgrade reaction) between a perfectly
elastic subgrade and Winkler subgrade with a constant value of the Winkler coefficient of
subgrade reaction was poor away from the applied load even if good directly beneath the
load. This implies that Winkler's Hypothesis with a constant Winkler coefficient of subgrade
reaction is not a good model for subgrades because it will produce results of variable
accuracy beneath the same foundation element.

4.4.3.3.1.4 Implementation into Structural Analysis Software. Despite these
obvious and extensive shortcomings, there are many reasons why Winkler's Hypothesis with a
constant Winkler coefficient of subgrade reaction is still used extensively in routine practice
despite its abysmal representation of actual subgrade behavior. Probably the most significant,
practical reason is that it allows the subgrade reaction, p(x,y) or {p}, to be eliminated as a variable
in the problem solution. As will be illustrated in this subsection, this allows easy implementation
into commercially available structural analysis software.
To illustrate the basic concepts involved in this, consider the 1-D problem and the behavior of
the beam shown in Figure 3.3 with an axial force, P, = 0. The flexural behavior of this beam is
given by

) ( ) (
) (
4
4
x q x p
dx
x w d
EI = + . (3.10)

The simplified version of Equation 4.3 for Winkler's Hypothesis with a constant coefficient of

kk
The obscurity and concomitant general lack of knowledge of Vesic's published work related to plate-type
foundation elements such as mats and rigid pavements has created some confusion in the published
literature. Numerous researchers and textbook authors have mistakenly used Vesic's much better known
work on beam-type foundation elements (Vesic and Johnson 1963), which involves a 12
th
-root equation, for
plate-type foundations which involves a cube-root equation. There are significant differences between the
two Vesic solutions as discussed in detail in Horvath (1988b, 1988c). This confusion has been all the more
unfortunate because plate-type shallow foundations such as mats, slabs-on-grade and "rigid" (PCC)
pavements are much more common in practice compared to beam-type shallow foundations such as
combined footings.
42
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
subgrade reaction in one (x-axis) direction only is

w(x) k = p(x)
o
W
. (4.4)

Combining equations 3.10 and 4.4 yields the behavior of the combined foundation element
(beam) and subgrade:
| | ) ( ) (
) (
4
4
x q x w k
dx
x w d
EI
o
W
= + . (4.5)

Note that the subgrade reaction variable, p(x), in Equation 3.10 has been eliminated from
explicit consideration in the final equation (4.5) because the entire subgrade contribution, at least
as modeled by Winkler's Hypothesis with a constant value of the coefficient of subgrade reaction,
is replaced by a constant coefficient,
o
W
k , times the settlement variable, w(x).
Given the extensive use of computer software in structural analysis and design nowadays, it is
even more insightful to view this alternatively using the stiffness matrix formulation. From
Section 3.4.2.5, the behavior of the beam shown in Figure 3.3 with P = 0 can be expressed as:

| |{ } { } { } q p d S = + (3.12)

where the flexural stiffness matrix, [S], of the beam is given by Equation 3.3 for a simple (Euler)
beam or Equation 3.5 for a Timoshenko beam as desired. Winkler's Hypothesis with a constant
value of Winkler's coefficient of subgrade reaction can be expressed using matrix notation as

{ } | |{ } d k = p
o
W
. (4.6)

Combining equations 3.12 and 4.6 yields

| |{ } | |{ } { } q d k d S
o
W
= + (4.7a)

which simplifies to

| |{ } { } q d S = ' (4.7b)

where [S'] is the modified or enhanced flexural stiffness matrix and is defined as follows:

| | | |
o
W
k S S + = ' . (4.8)

Thus all the subgrade effects are easily accounted for in the flexural stiffness matrix of the
foundation element. The modifications to the terms of the basic flexural stiffness matrix, [S]
(Equation 3.3 for a simple beam or Equation 3.5 for a Timoshenko beam as appropriate), that are
reflected in Equation 4.8 can be visualized as adding an external, independent spring at each node
of the foundation element (ACI Committee 336 1988, 1989; Bowles 1988). The orientation of
each spring is perpendicular to the foundation element-subgrade interface and thus perpetuates
the "soil spring" visualization of Winkler's Hypothesis. In the case defined by Equation 4.8, each
spring has the same, constant magnitude,
o
W
k . Commercially available structural analysis
software can easily accommodate such springs. It appears that this analytical simplicity and ease
of use more than anything else has perpetuated the use in practice of Winkler's Hypothesis with a

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
43
uniform, constant value of the Winkler coefficient of subgrade reaction,
o
W
k .
It is of interest to note that the discussion in the preceding paragraphs can be easily extended in
several ways. First of all, a true beam-column (the P 0 case shown in Figure 3.3) could be used.
Equations 4.7b and 4.8 still apply to, and govern the behavior of, the beam-column but in this
case the basic flexural stiffness matrix, [S], is the one defined by Equation 3.7. In addition, the
Winkler coefficient of subgrade reaction could be allowed to vary as a function of x as opposed to
being limited to a constant value. In such a case it simply means that the subgrade springs have
different values. Finally, it is also conceptually straightforward to extend this to a plate with two-
dimensional bending. In the case of a plate, the springs representing the Winkler subgrade are
added in both dimensions and can be constant or variable in magnitude as desired.

4.4.3.3.1.5 Deficiencies and Their Solution: An Overview. The deficiencies of
Winkler's Hypothesis as a subgrade model, especially when a constant value of the Winkler
coefficient of subgrade reaction is used as discussed in the preceding subsection, were apparently
recognized at least by the same ca. 1930s time frame as the beginning of modern geotechnical
engineering. This is evidenced by the fact that the results of applied-mechanics research into
improved subgrade models was published at least as early as 1940 (Horvath 1979, 1988d, 1989c).
However, it was not until ca. 1980s that definitive suggestions for improving on Winkler's
Hypothesis as used traditionally (i.e. with a constant value of the Winkler coefficient of subgrade
reaction) began to appear in the geotechnical and foundation engineering literature.
These recommendations for improvement within the overall context of mechanical models have
taken two very different conceptual paths. However, they both derive from the common fact that
the fundamental error in Winkler's Hypothesis as expressed in its fundamental definition
(Equation 4.1) is that it cannot replicate the mechanism of "load spreading" that develops within
an actual subgrade due to the development of shear stresses. Visualized using the ubiquitous
spring analogy for Winkler's Hypothesis, the "springs" of an actual subgrade are not independent
as Winkler's Hypothesis assumes but are coupled or linked together so that an applied load at
some point i produces settlement not just at point i but adjacent ones (i-1, i+1, etc.) as well.
Conversely, the settlement at some point i is the result of applied loads not just at point i but at
other points as well (which may or may not be adjacent). Thus it is convenient to state for the
purpose of discussion that the absence of "spring coupling" in Winkler's Hypothesis is its single
most significant shortcoming as a subgrade model. Thus any improvement to Winkler's
Hypothesis must incorporate spring coupling in some manner.
It is at this point that the two conceptual paths for improving on Winkler's Hypothesis diverge
fundamentally and completely. One path, which began in the 1930s if not earlier, seeks to
incorporate spring coupling inherently on a rational, theoretical basis within the subgrade model
itself. This approach, which produces what are generically called multiple-parameter models, is
discussed in detail in Section 4.4.3.3.2
ll
. However, it is better to first discuss the other theoretical
path that is frequently referred to as pseudo-coupling or the Pseudo-Coupled Concept when it is
applied to mats and related foundation elements (ACI Committee 336 1988, 1989; Bowles 1988)

mm
. This is because the Pseudo-Coupled Concept is, in reality, nothing fundamentally new but is
simply a repackaged version of the single-parameter Winkler Hypothesis.

4.4.3.3.1.6 The Pseudo-Coupled Concept: Part I - Generic Concept. Despite its
unique name, the Pseudo-Coupled Concept is simply a return to the general form of Winkler's

ll
In the context of subgrade models, the term parameter is taken to mean how many coefficients appear in
the differential equation that defines the behavior of a particular model.
mm
It has other names in other applications. For example, when used for laterally loaded deep foundations it
is known as the p-y method.
44
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
Hypothesis (Equation 4.2). It appears that the predominant reason for the Pseudo-Coupled
Concept is to retain the mathematical and modeling simplicity of Winkler's Hypothesis, and its
ability to be readily incorporated into commercially available structural analysis software as
discussed in Section 4.4.3.3.1.4.
In essence, the potential improvement that the Pseudo-Coupled Concept offers as a subgrade
model comes simply from allowing the Winkler coefficient of subgrade reaction to vary beneath
the foundation element and thus mimic the actual variable coefficient of subgrade reaction that
develops beneath foundation elements as was discussed in the preceding subsection. Because this
actual variation in the coefficient of subgrade reaction is due, in part, to the vertical shearing
(spring coupling) that develops in actual subgrade materials, using variable Winkler springs
beneath a foundation element creates the appearance or end-result of spring coupling without
explicitly doing so mathematically.
Upon initial consideration, the Pseudo-Coupled Concept sounds very attractive. It appears to
give the desired analytical improvement over the traditional use of Winkler's Hypothesis with a
constant coefficient of subgrade reaction without increasing the analytical complexity. Most
importantly, springs that vary in magnitude can easily be accommodated by commercially
available structural analysis software. However, published presentations of the Pseudo-Coupled
Concept rarely highlight several key facts:

Winkler's Hypothesis is still being used as the subgrade model. Whether the Winkler
coefficient of subgrade reaction is constant or variable doesn't change the fact that it is
fundamentally a poor subgrade model because it inherently neglects the shearing effects
(spring coupling) that occurs within all real subgrades. This means that the entire burden for
estimating the effects of spring coupling is borne by the engineer performing the analysis.
Note that this is not the case for multi-parameter subgrade models which inherently
incorporate subgrade shearing effects into their formulation.

The coefficient of subgrade reaction is still required to be an input parameter rather than a
calculated or observed result as it actually is. This is the tradeoff made for being able to use
such a simple subgrade model as Winkler's Hypothesis. Again, this is not the case for multi-
parameter models.

Most importantly, for the Pseudo-Coupled Concept to yield good results the input values of
the Winkler coefficient of subgrade reaction must match both the magnitude and variation of
the correct results (which are, of course, unknown beforehand). In essence, the answers to a
problem must be known beforehand so that they can be input to produce the correct answers.
The absurdity of the circular logic of this seems to have been overlooked or ignored by
proponents of the Pseudo-Coupled Concept. Once again, this situation does not arise when a
multi-parameter model is used.

Thus in reality, the utility of the Pseudo-Coupled Concept as an improved subgrade model for
general usage is questionable because each application is a unique combination of structural and
geotechnical components. For example, no two mat-supported structures are ever exactly the
same. As a result, the magnitude and distribution of the coefficient of subgrade reaction beneath a
mat is, in general, unique to a given application. Thus precedence may not offer much guidance
so it is unreasonable to expect an engineer to anticipate beforehand exactly what the spring-
coupling effects are for a given application so that the correct results can be calculated. This fact
is illustrated very well in Section 7 using a case history involving two mat foundations with the
same subgrade conditions.
The problems inherent in the Pseudo-Coupled Concept as an improved subgrade model for

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
45
general shallow-foundation applications have not deterred its promotion by its proponents,
especially in U.S. practice (ACI Committee 336 1988, 1989; Bowles 1988). In addition, there is
evidence that the Pseudo-Coupled Concept can be useful in very focused applications such as
laterally loaded deep foundations. For these reasons, further discussion of the subject is of
interest.
It appears the logic used to promote the general use of the Pseudo-Coupled Concept is that
although each foundation application is unique, it is possible to develop generic patterns or
distributions for the Winkler coefficient of subgrade reaction, k
W
(x,y), for very simple foundation
geometries and superstructure-load distributions. It is then argued that these results are of
sufficient accuracy for general use in practice. Typically, these patterns in subgrade reaction are
represented as some variation in k
W
(x,y) relative to a problem-specific base or minimum value of
k
W
(x,y). Of course, this still leaves the problem of how to rationally determine what this base
value should be.
The generic variations suggested to date have all been developed with mat foundations in mind
and assume an increase in k
W
(x,y) near the edges of the foundation element. This derives from the
fact that, for relatively simple loading, the actually observed subgrade reaction tends to increase
toward the edges of a foundation element, regardless of its relative stiffness
nn
. The simplest
suggestion is to merely double the constant, base value of the Winkler coefficient of subgrade
reaction along the edges of the foundation (ACI Committee 336 1988; Bowles 1988). The
arbitrary and ambiguous nature of this suggestion has been noted by Liao (1995). A somewhat
more sophisticated version of this is to use a variation in k
W
(x,y) that is based on the theory of
linear elasticity. Usually the solution for a perfectly flexible, uniformly loaded area on an
isotropic, homogeneous layer of finite thickness is used. Typically, this produces a Winkler
coefficient of subgrade reaction that is also about twice as large around the edges of a mat
compared to its base value at the center, but with a gradual, curved pattern of variation in
between. Such an approach is also discussed in ACI Committee 336 (1988, 1989) and Bowles
(1988).
Regardless of the actual method used, there is still the problem of how to rationally determine a
base value for k
W
(x,y) at the center of the foundation element. The various methodologies
available were discussed in some detail in Horvath (1979, 1988b). It appears that the most
rational method (keeping in mind that it is for an inherently irrational subgrade model, Winkler's
Hypothesis) is based on the simplified-continuum concept. As noted in Section 4.4.3.2, the
simplified-continuum concept in general is an alternative to mechanical models as a methodology
for developing subgrade models. This concept is discussed in detail in Section 4.4.3.4.
In summary, while some view the Pseudo-Coupled Concept as the long-sought improvement to
the traditional use of Winkler's Hypothesis with a constant Winkler coefficient of subgrade
reaction in applications such as mat foundation the actual improvement in a given problem is
subject to significant variability that can be difficult for an engineer to assess. The reason is that
the degree of improvement depends significantly on how closely the actual problem analyzed
matches the assumed problem that was used to develop the variation in the values of k
W
(x,y). In
addition, the selected base value for k
W
(x,y) will have an obvious influence as well. These issues
are discussed in detail in Horvath (1993c, 1993d), with a summary presented in Section 7 of this
report. The conclusion stated in Section 7 is that the Pseudo-Coupled Concept does not offer a
reliable, consistent methodology compared to the historical usage of Winkler's Hypothesis with a
constant value of k
W
(x,y), i.e.
o
W
k .
It is important to note, however, that there is evidence that the Pseudo-Coupled Concept can

nn
This presumes at least some embedment of the foundation element as would always be the case for a real
structure. This is necessary so that the subgrade, especially if it is a coarse-grain soil, will have the
necessary confining stress to be able to sustain elevated edge stresses.
46
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
have niche applications where it appears to offer reliable results. Its long-time historical use under
a different name for laterally loaded deep foundations was noted already. For mats and related
foundation elements, this potential acceptable use is for applications that have relatively simple,
repetitive geometry where a reasonable distribution pattern for k
W
(x,y) can be determined once
and for all using other, more-rigorous analytical methods. The only known application meeting
this criterion that has been studied in some detail to date is base slabs for box tunnels with a
rectangular cross-section that are constructed using the cut-and-cover method (Liao 1991, with a
more-accessible summary of his findings Liao (1995)). Liao found that for slabs that are stiff
relative to the subgrade, the type of loading (uniformly distributed or concentrated loads) was not
an important factor. Consequently, he developed a series of chart and tabular solutions for a
uniformly loaded, infinitely long strip resting on a homogeneous, isotropic, linear-elastic
continuum of finite thickness. This continuum problem was solved using the finite-element
method. These solutions provide relative values of the variation of k
W
(x)
oo
beneath the simulated
tunnel base slab. Liao's results provide a uniform value for k
W
(x) within the middle 60% of the
slab, with increasing values of k
W
(x) between there and the edge of the slab. However, as with the
generic distributions discussed previously there is still the problem of how to determine the
reference value of k
W
(x) at the center of the base slab. The simplified-continuum approach noted
above would appear to be the most appropriate although Liao offers his own suggestions on the
topic.

4.4.3.3.1.7 The Pseudo-Coupled Concept: Part II - The Discrete-Area Method.
There is a variation of the Pseudo-Coupled Concept called the Discrete-Area Method (Ulrich
1991) that warrants a separate discussion because it has significant differences from the generic
version of the concept that was discussed in the preceding subsection. The Discrete-Area Method
is an example of what was termed the "pseudo-ideal" alternative for SSI analysis as illustrated in
Figure 2.4. This method was actually alluded to and described briefly in Section 2.2.3.3 but is
explored in greater detail here. Although this method is known to have been applied only to mat-
supported buildings to date the methodology is sufficiently general that it could be used for a
wider variety of SSI applications.
In the Discrete-Area Method version of the Pseudo-Coupled Concept, both the magnitude and
variation of k
W
(x,y) are estimated on a project-specific basis. This is accomplished by having the
structural and geotechnical members of a design team conduct separate computer analyses for the
megastructure (mat + superstructure) and subgrade respectively. Experience to date indicates that
the structural analysis is performed using the usual commercially available software while the
geotechnical analysis uses software that can analyze a continuum (this is discussed further
subsequently). The structural plate representing the mat (in the structural analysis) and the
foundation level of the subgrade (in the geotechnical analysis) are each divided into the same
checkerboard-like pattern of "discrete areas" (hence the name), with the number, shape and size
of areas used arbitrary. In the structural analysis, each area on the mat is supported by an
independent spring but a different spring stiffness is allowed beneath each discrete area. Thus
from a structural perspective the general form of Winkler's Hypothesis is used for the subgrade
model. In the geotechnical analysis, each discrete area is subjected to a uniformly distributed load
(i.e. stress) but a different load magnitude is allowed for each discrete area. Spring stiffnesses (in
the structural analysis) and applied stresses (in the geotechnical analysis) of each area are
changed in an iterative, trial-and-error process until the displacement patterns from the two
analyses match within some error judged acceptable by the engineers involved. It is assumed that
this match provides the unique solution to the problem. Note that a major difficulty inherent in

oo
One of the additional simplifications of box tunnels is that because of their relatively great length
compared to width they are designed using a plane-strain analysis. This means that the Winkler coefficient
of subgrade reaction needs to be defined in one direction (x axis) only.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
47
the generic Pseudo-Coupled Concept is overcome because the variable Winkler coefficient of
subgrade reaction (spring stiffnesses) used in the structural analysis is determined specifically and
uniquely for a given problem by matching to a separate analysis of the subgrade and does not
depend on a match to some simple, generic solution.
While the Discrete-Area Method is perhaps the ultimate pseudo-coupled method and appears to
give good results consistently for building mats (Ulrich 1991), it does not appear to be an overall
attractive methodology for routine, widespread use in practice. From a technical perspective, the
weaker link is the geotechnical analysis. It is clear that a 3-D analysis is required. For the reasons
discussed in Section 4.3, neither closed-form elastic solutions nor a numerical analysis such as the
finite-element method (which would allow use of nonlinear constitutive models for the subgrade
materials) are feasible on a routine basis. This leaves a computer solution of an elastic continuum,
preferably a layered system. Such software is available and is apparently what has been used to
date with the Discrete-Area Method. While this is probably an acceptable approach for most
problems, it is the non-technical aspects that appear to have severely limited the use of this
method in practice despite its development as far back as the 1960s. Specifically, it is its inherent
iterative, trial-and-error nature requiring a team effort, not to mention the fact that a separate
series of analyses must be performed for every load case on a given project, that apparently have
discouraged its wider use in practice.

4.4.3.3.2 Multiple-Parameter

4.4.3.3.2.1 Overview. Because of the problems inherent in the Pseudo-Coupled Concept,
whether in its generic formulation or as the Discrete-Area Method, there is still a need for
fundamentally more accurate subgrade models that incorporate spring coupling inherently yet
would still be easy to implement and use in routine practice (Horvath 1989a). Of all the
alternative subgrade models investigated, the most-promising appear to be in the general category
of multiple-parameter models.
The fundamental and most significant attraction of multiple-parameter models as a group is that
a variation in the coefficient of subgrade reaction never has to be assumed beforehand as it does
when the Pseudo-Coupled Concept is used. The reason is that the soil shear (spring coupling) that
is inherently missing from Winkler's Hypothesis is incorporated explicitly in the derivation and
mathematics of even the simplest multiple-parameter model and is thus independent of the
engineering judgment required by the Pseudo-Coupled Concept. It is of interest to note that there
is a significant improvement in analytical accuracy just going from Winkler's Hypothesis to the
simplest multiple-parameter model. The degree of improvement continues as higher-order
multiple-parameter models are used. However, research to date indicates that the rate of
improvement decreases rapidly which means that it should not be necessary to use multiple-
parameter models of very high order to obtain results that are acceptable for routine practice.
An additional benefit of multiple-parameter models is that they are still surface-element
models. As a result, their governing equations do not require explicit consideration of the
subgrade depth (z-axis). The depth effects are built into the equation defining the behavior of a
multiple-parameter model when it is derived.
As might be expected, the benefits of multiple-parameter models come at a cost. However,
none of them turns out to be fatal to the practical use of these models:

Multiple-parameter models can be visualized as containing two or more physical components
compared to the single component (layer of springs) used to model Winkler's Hypothesis.
This means that when a multiple-parameter model is used with commercially available
structural analysis software modeling the subgrade is always more complicated compared to
Winkler's Hypothesis. In some cases, it may also mean that a non-linear structural analysis
48
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
needs to be performed. However, while these are complications they do not rule out use of
multiple-parameter models.

Boundary conditions must be dealt with. This is because spring coupling results in subgrade
displacements beyond the edge of the foundation element. This issue does not arise with
Winkler Hypothesis because of the lack of spring coupling so subgrade displacements always
end abruptly at the edge of the foundation element. The number of boundary conditions that
must be considered increases with the order and sophistication of the multiple-parameter
model chosen. Experience indicates that boundary conditions can be reasonably
accommodated for any multiple-parameter model that is likely to be considered in practice.

Depending on the particular model used, there are two or more model coefficients whose
values must be determined. In the traditional visualization and formulation of multiple-
parameter models as assemblages of springs and other structural elements (beams/plates;
shear layers; deformed, pretensioned membranes) this turns out to be a significant problem.
This is because the conceptual and numerical relationship between these mechanical elements
and subgrade properties such as Young's modulus is not the least bit obvious or intuitive, any
more than it was for the spring layer used to visualize Winkler's Hypothesis. However, this
impasse has been completely overcome in recent years by developing a novel method for
interpreting mechanical models in general. Presentation of this methodology is a major
contribution of this report that is presented in Section 4.4.3.5.

In conclusion, at the present time there is nothing that prevents reasonably simple multiple-
parameter models from being used both with commercially available structural analysis software
or a specially developed, problem-specific software.

4.4.3.3.2.2 Model Components. The evolution of multiple-parameter models, including
sketches of model visualizations and references to original publications, has been presented in
detail elsewhere (Horvath 1979, 1988d, 1989c) so is not repeated in this report. The focus here is
on summarizing and synthesizing concisely what has been published previously.
To begin with, it is useful to describe the three mechanical (physical) elements that are used to
either create mechanical models or visualize abstract mathematical expressions such as made by
Winkler, Pasternak and Loof. The common aspect of these elements is the relationship they have
with displacement, w(x,y), in a direction perpendicular to the subgrade surface and parallel with
the direction of the applied load p(x,y):

The basic element is one where the resistance to an applied load, p(x,y), is proportional to
w(x,y). This can also be viewed as being proportional to a zero-order derivative of w(x,y). As
discussed in Section 4.4.3.3.1.1, the most common physical element used to visualize this
mathematical behavior is a spring supporting a unit area of subgrade although unit weight of
a liquid is an alternative. The spring analogy is used exclusively in this report and the spring
stiffness, k, is the parameter of interest.

The next element is one where the resistance to an applied load, p(x,y), is proportional to the
first derivative of w(x,y), w'. The more-common physical element used to visualize this
mathematical behavior is an incompressible "shear layer" which is defined as a layer of
linear-elastic material of unit thickness that resists vertical shear forces only. This type of
element has been used to visualize the mathematical hypothesis made, apparently
independently, by Pasternak and Loof. The shear stiffness, g, of this shear layer is the model
parameter of interest. An alternative visualization of the Pasternak/Loof Hypothesis is a

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
49
deformed, pretensioned membrane which was used by Filonenko-Borodich where the
membrane tension, T, is the model parameter of interest. Note that the mathematical
formulation for the Filonenko-Borodich model is based on the deformed geometry of the
membrane which means that this is a nonlinear problem that can be solved in a closed-form,
quasi-linear fashion. This is conceptually identical to the formulation of the true beam-
column that was discussed in Section 3.4.2.4.2. In fact, the governing equation of the
Filonenko-Borodich model, at least for one-dimensional behavior, can be obtained by setting
the flexural stiffness, EI, equal to zero in the equation (3.11) for the true beam-column
supported on a subgrade and then recognizing that the membrane tension, T, is the same as an
axial force, -P. The fact that the Filonenko-Borodich model is an inherently nonlinear
problem based on a deformed geometry is emphasized here as this is something that does not
appear to have been fully appreciated in the past. This inherent nonlinearity has significant
implications for implementing this model within the context of commercially available
structural analysis software. As discussed in Section 3.4.2.4.1, although nonlinear analyses
are within the capability of most, if not all, structural software nowadays experience indicates
that such analyses are not the default mode in which the software is used.

The highest-order physical element defined to date for use in developing mechanical models
is one where the resistance to an applied load, p(x,y), is proportional to the second derivative
of w(x,y). The only physical element used to visualize this mathematical behavior is an Euler
flexural element. This would be a plate in 2-D or a simple beam in 1-D. The plate or beam is
assumed to be linear-elastic in its behavior so the flexural stiffness, D in 2-D and EI in 1-D, is
the parameter of interest.

With these mechanical elements in mind, Table 4.1 summarizes the composition of mechanical
models in order of their increasing mathematical complexity and, therefore, presumed accuracy as
a subgrade model. Models are grouped together when they have identical behavior even though
they may be visualized using different physical elements. Although Winkler's Hypothesis is a
single-parameter model it is included in Table 4.1 for comparison. The mathematically identical
Filonenko-Borodich model and Pasternak/Loof Hypothesis are the lowest-level multiple-
parameter models.


Table 4.1. Composition of Mechanical Models

Subgrade model
(in groupings of identical models)
Physical elements used to visualize model
(as viewed from top to bottom)
Winkler's Hypothesis springs
Filonenko-Borodich deformed, pretensioned membrane + springs
Pasternak's Hypothesis
Loof's Hypothesis
shear layer + springs
Modified Pasternak (see Note 1) springs + shear layer + springs
Haber-Schaim plate + springs
Hetnyi (see Note 2) springs + plate + springs
Rhines springs + plate + shear layer + springs
Notes:
1. The Modified Pasternak model is referred to as the Kerr model in later publications by Prof. Arnold D. Kerr
who first proposed this model.
50
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
2. The Hetnyi model should not be confused with his landmark published work (Hetnyi 1946) involving
Winkler's Hypothesis which is much better known. The development of his own multiple-parameter model was
actually a relatively minor component of Hetnyi (1946).


Note that only those multiple-parameter models that have been studied mathematically to any
extent are included in Table 4.1. Kerr and Rhines (1967) identified five additional, higher levels
of multiple-parameter models more complex than the Rhines model. However, other than to
illustrate the physical elements used to visualize each they did not investigate these higher-order
models mathematically. It is also of interest to note that for the Rhines model and higher there is a
physical non-uniqueness to the models
pp
. In general, an embedded plate produces the same
behavioral result as an embedded combined shear layer over spring layer.
Another way to organize and categorize mechanical models is to note that the behavior of all
mechanical models shown in Table 4.1 can be expressed using the same basic differential
equation

) , ( ) , ( ) , ( ) , ( ) , ( ) , (
4 2 4 2
y x w c y x w c y x w c y x p c y x p c y x p
3 2 1 2 1
w w w p p
+ = + (4.9)

where
i
p
c and
i
w
c are constant coefficients that vary depending on the model and may be zero in
some cases. These coefficients are composed of the various properties of the mechanical elements
used in that model, i.e. spring stiffness, k; shear layer stiffness, g; membrane tension, T; and plate
flexural stiffness, D. Table 4.2 shows a hierarchy of mechanical models based on the terms in
Equation 4.9 reflected in each model. A shaded cell in this table indicates that a particular term is
represented for the model.


Table 4.2. Hierarchy of Mechanical Models

derivative order of p(x,y) derivative order of w(x,y) Subgrade Model
(in groupings of identical models)
0 2 4 0 2 4
Winkler's Hypothesis
Filonenko-Borodich
Pasternak's Hypothesis
Loof's Hypothesis

Modified Pasternak (Kerr)
Haber-Schaim
Hetnyi
Rhines


It can be seen that adding a mechanical element in a kind of building-block approach as shown
in Table 4.1 (and carried beyond, at least in a conceptual context, by Kerr and Rhines (1967))
results in a higher-order model as reflected in Table 4.2. Note, however, that the increased order
is not necessarily consistent. For example, the Haber-Schaim and Hetnyi models omit some
intermediate order terms. Therefore, they are not necessarily more accurate than the Modified
Pasternak (Kerr) model
qq
.

pp
This is separate from the fact that a spring layer and liquid, and a deformed, pretensioned membrane and
a shear layer are identical mathematically as discussed previously.
qq
There is no known study of this issue to date.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
51
4.4.3.3.2.3 The Beam-Column Analogy. As discussed in Section 4.4.3.3.1.4, one of the
primary reasons for the enduring attraction of Winkler's Hypothesis in practice is that its
governing equation can be readily combined with the governing equation for a foundation
element on a subgrade so that the subgrade reaction, p, is eliminated as a variable. As a result, a
single equation (4.5 or 4.7b) can be easily obtained for the combined behavior of a foundation
element on a Winkler subgrade
rr
. Hetnyi devoted most of his 1946 monograph to developing
solutions to the 1-D problem of a beam that are of use in both civil and mechanical engineering.
For simplicity in presenting this concept of combining the behavior of foundation element and
subgrade into one equation, Section 4.4.3.3.1.4 focused on the basic 1-D case of a beam on a
Winkler subgrade with a constant coefficient of subgrade reaction,
o
W
k . However, as noted in
that subsection this concept can easily be extended to the more-general 2-D case as well as allow
for a variation in the Winkler coefficient of subgrade reaction. This can be accomplished by using
the classical equation for a plate on a subgrade:

( ) ) , ( ) , ( ) , ( ) , (
2 2
y x q y x p y x w y x D = + (4.10a)

which, assuming D(x,y) = constant = D, becomes

) , ( ) , ( ) , (
4
y x q y x p y x w D = + , (4.10b)

together with the general equation (4.2) for Winkler's Hypothesis. Details of this are omitted here.
It is of interest now to note that this same procedure of developing a combined foundation
element plus subgrade equation can be applied for the lowest-level multiple-parameter models
where all
i
p
c = 0 (see Equation 4.9 and Table 4.2). The version of Equation 4.9 appropriate for
such a model can be readily combined with equations 4.10 to produce a single equation for a
plate on a multiple-parameter subgrade. It follows that the simpler 1-D problem involving a beam
can be developed as well.
This concept was explored in some detail for the basic 1-D case in Horvath (1993e) using the
Pasternak/Loof Hypothesis
ss
(which is visualized as a shear layer over springs) and an Euler
(simple) beam. As can be seen in Table 4.2, the Pasternak/Loof Hypothesis is the simplest
mechanical model that inherently incorporates subgrade shear (spring coupling). The 1-D version
of Equation 4.9 for a Pasternak/Loof subgrade is


dx
x w d
g x w k x p
) (
) ( ) (
2
= (4.11)

where g = the shear stiffness of the shear layer and k = the spring stiffness of the spring layer.
Combining Equation 4.11 with that of a beam on a generic subgrade


rr
Historically and traditionally, most solutions of the combined foundation element-subgrade equation have
involved the 1-D version of this problem where the foundation element is a beam. This is usually called the
Beam on Elastic Foundation problem. This term is deprecated and thus not used in this report for a number
of reasons. First of all, this term creates needless confusion with a true elastic continuum that can be, and
has been, used as a subgrade model as discussed in Section 4.3. Second, a Winkler subgrade is not
necessarily elastic in its behavior. As discussed in Section 4.4.3.3.1.1, the springs used to visualize
Winkler's Hypothesis do not have to be either linear or elastic.
ss
For the sake of completeness, it should be noted that the Filonenko-Borodich model could be used
alternatively.
52
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
) ( ) (
) (
4
4
x q x p
dx
x w d
EI = + (3.10)

yields
) (
) (
) (
) (
2
4
4
x q
dx
x w d
g x w k
dx
x w d
EI = + (4.12a)

which, after rearranging terms, becomes

) ( ) (
) ( ) (
2
4
4
x q x w k
dx
x w d
g
dx
x w d
EI = + . (4.12b)

Equation 4.12b is recognized as the equation of a true beam-column supported on a subgrade
(Equation 3.11), in particular, a Winkler subgrade. As a result, this combined foundation element
plus Pasternak/Loof subgrade equation is called the Beam-Column Analogy (Horvath 1993e).
There are several interesting results contained in the Beam-Column Analogy. First, as expected
the beam and subgrade behavior are combined into a single equation. However, the spring-
coupling that is inherent in a Pasternak/Loof subgrade disappears from explicit consideration
leaving only a Winkler subgrade which, as discussed in Section 4.4.3.3.1.4, is easy to implement
within the constraints of commercially available structural analysis software
tt
. Examination of
equations 4.11 and 4.12b indicates that all the spring-coupling effects inherent in a
Pasternak/Look subgrade are replaced by a fictitious tensile force of magnitude g per unit width
of the beam applied to the longitudinal axis of the beam parallel to the horizontal x-axis. Thus g
acts opposite of the sense of P shown in Figure 3.3. As discussed in Section 3.4.2.4.2 and
illustrated in Equation 3.7, from the perspective of the flexural stiffness matrix of a beam a tensile
force makes the beam appear to be stiffer than it actually is. Thus it can be seen qualitatively that
the consideration of shear effects (spring coupling) in a subgrade, at least on a first-order basis
which is all a Pasternak/Loof subgrade provides, compared to a Winkler subgrade without such
effects has the result of reducing differential settlements of the foundation element.

4.4.3.4 Simplified-Continuum Models

Relatively much less well known compared to mechanical models is the family of surface-
element subgrade models (SEMs) that is based on simplifications and approximations applied to a
linear-elastic continuum. As opposed to mechanical models which started simple (Winkler's
Hypothesis) then have grown progressively more complex over time by adding layers of other
physical elements as reflected in Table 4.1, the development of simplified-continuum models
always starts with the most complex case (the complete set of partial-differential equations
defining the behavior of a linear-elastic continuum). Various assumptions are then made to these
equations to render the remaining equations fairly easy to solve in an exact, closed-form manner.
These assumptions typically involve certain stresses and strains being zero.
Development of simplified-continuum models to date has taken two, broad paths. An overview

tt
However it should be kept in mind that Equation 4.12b is the equation of a true beam-column and thus is
inherently nonlinear. As discussed in Section 3.4.2.4, commercially available structural analysis software
cannot, in general, solve a true beam-column explicitly and thus must treat it as an axially loaded beam
under nonlinear conditions. While this is within the capability of most software, the engineer using the
software must recognize the need to invoke a nonlinear analysis which is often not the default mode of the
software.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
53
of this development is presented in Horvath (1979, 1988d, 1989c):

Reissner (1958, 1967) pioneered a straightforward application of this concept to produce
what is hereinafter referred to as the Reissner Simplified Continuum (RSC) model. The RSC
was the model explored in detail in Horvath (1979) with a summary presented in Horvath
(1983a). The concept first proposed by Reissner was extended by Horvath (1979) to produce
two simpler models that are called the Pasternak-Type Simplified Continuum (PTSC) and
Winkler-Type Simplified Continuum (WTSC) models. Both the PTSC and WTSC were
explored in detail in Horvath (1979). The use of the PTSC as an alternative formulation of the
Beam-Column Analogy that was presented in the preceding subsection was discussed in
Horvath (1993e). A summary of development of the WTSC was presented in Horvath
(1983b) with further exploration of its application in Horvath (1988b, 1988c).

Vlasov and Leont'ev (1960) presented a less-direct application of the simplified elastic
continuum concept that has subsequently been used by others as well. This alternative
approach involves using variational calculus. The complication of this approach is that in
addition to making simplifying assumptions about an elastic continuum as Reissner and
Horvath did, an arbitrary function must be assumed to define how vertical displacements vary
as a function of depth (in the alternative approach first proposed by Reissner the depth-wise
variation of vertical displacements is simply a result of the overall solution). Thus the
variational calculus approach introduces yet another level of approximation and judgment
into the solution process that renders it less desirable to use as an approach to developing
practice-oriented subgrade models.

As illustrated in Horvath (1979, 1988d, 1989c), regardless of which of the above approaches is
taken all simplified-continuum models developed to date have the same basic equation defining
their behavior. This equation also turns out to be identical to Equation 4.9 which applies to
mechanical models and is repeated here for convenience:

) , ( ) , ( ) , ( ) , ( ) , ( ) , (
4 2 4 2
y x w c y x w c y x w c y x p c y x p c y x p
3 2 1 2 1
w w w p p
+ = + . (4.9)

One item to note is that simplified-continuum models have not yet been developed to the same
order or level that mechanical models have. Therefore, the
4
terms that appear on both sides of
Equation 4.9 have not been used to date for simplified-continuum models.
It is of interest to note that for simplified-continuum models, the constant coefficients
i
p
c and
i
w
c in Equation 4.9 are functions only of the elastic moduli, E and G, of the continuum as well as
H, the thickness of the continuum, i.e. the depth from the subgrade surface to the assumed rigid
base underlying the continuum. This fact turns out to be a significant benefit when developing a
synthesized approach to subgrade modeling in the next subsection.
Because all simplified-continuum models are defined by a common equation, a hierarchy of
simplified-continuum models is defined in Table 4.3 as it was for mechanical models in Table
4.2. Again, a shaded cell in Table 4.3 indicates that a particular term in Equation 4.9 is
represented for that model. Note that the lack of any models of higher order than the RSC is not
due to any limitation of the overall simplified-continuum methodology but simply the fact that
none is known to have been developed to date. However, it would be of interest to develop a
simplified-continuum model at least of an order that corresponds to the Rhines mechanical
model. If nothing else, such a simplified-continuum model would serve as a benchmark for
comparison with the RSC which is the most advanced and heavily studied simplified-continuum
model to date.
54
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
Table 4.3. Hierarchy of Simplified-Continuum Models

derivative order of p(x,y) derivative order of w(x,y) Subgrade Model
(in groupings of identical models)
0 2 4 0 2 4
Winkler-Type Simplified Continuum
Pasternak-Type Simplified Continuum
Vlasov and Leont'ev (1 layer)

Reissner Simplified Continuum
Vlasov and Leont'ev (2 layers)



4.4.3.5 Synthesis

4.4.3.5.1 Introduction

Having presented the key elements of the development and basic, time-independent forms of
SEMs, it is useful to summarize the pluses and minuses and pros and cons of both mechanical and
simplified-continuum models. This summary is from the perspective of their potential use in
practice, especially within the constraints of commercially available structural analysis software
which are much more severe than for application-specific software which can accommodate any
desired model.
First with regard to mechanical models:

On the negative side, in general it is difficult to interpret exactly what subgrade material
properties or characteristics are reflected in the various mechanical elements (springs, etc.).
This is especially true for the beam/plate in flexure and the tensioned membrane as there is no
behavioral pattern of actual subgrade materials that can be related intuitively to either of
these. As a result, it is difficult to assess the relative accuracy of some of the models. For
example, referring to tables 4.1 and 4.2 it is not clear that the Haber-Schaim or Hetnyi
models are inherently more accurate than the Modified Pasternak (Kerr) model. While the
former models have higher-order terms compared to the Modified Pasternak/Kerr, they are
missing intermediate-order terms that the latter includes. In addition, it is difficult to
conceptually develop a rational methodology for evaluating these various elements in a
particular problem.

On the positive side, it is at least conceptually possible to model the various mechanical
elements within the context of commercially available structural analysis software. However,
as it turns out a moderate level of analytical sophistication, e.g. performing a nonlinear
analysis, is required for all multiple-parameter models.

In conclusion, although mechanical models can be feasibly implemented both in commercially
available structural analysis software as well as application-specific software, rational evaluation
of the various mechanical elements is, from a practical perspective, impossible
uu
.
With regard to simplified-continuum models:


uu
There have been suggestions in the applied-mechanics literature that the necessary coefficient values
always be determined experimentally. For civil engineering structures, this implies building a structure and
measuring its behavior so that the necessary coefficient values can be back-calculated and the structure
designed using the back-calculated values. This is obviously not a realistic approach.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
55
On the positive side, by using the general developmental procedure proposed by Reissner the
implications of the various simplifying assumptions are quite clear as is the physical meaning
of the remaining terms and coefficients in the differential equation that defines a model's
behavior. Thus there is insight into exactly what each model assumes and implies in terms of
subgrade behavior. Furthermore, evaluation of the equation coefficients is straightforward as
they are based solely on the subgrade elastic properties and problem geometry.

On the negative side, it is essentially impossible to visualize and model an elastic continuum,
simplified or otherwise, within the context of commercially available structural analysis
software.

In conclusion, the use of simplified-continuum models is limited to application-specific software
where there are no limitations on modeling capabilities. Experience to date has shown that this is
quite feasible (Horvath 1993c, 1993d). An example of such software produced the results
summarized in Section 7. Specific capabilities of this software are presented in Appendix A.
As can be seen, neither mechanical nor simplified-continuum models represent the "perfect"
subgrade model, especially when there is a need to work within the context of commercially
available structural analysis software. However, it turns out that the strengths of the two
categories of SEMs models complement each other so that the optimum solution is obtained by
using the best parts of each.

4.4.3.5.2 Implementation: Part I - General Concept

A synthesis of SEMs to produce an optimum combined result for use in practice has several
steps to it. To begin with, as noted in Section 4.4.3.4 all SEMs have the same basic governing
equation (4.9):

) , ( ) , ( ) , ( ) , ( ) , ( ) , (
4 2 4 2
y x w c y x w c y x w c y x p c y x p c y x p
3 2 1 2 1
w w w p p
+ = + . (4.9)

Therefore, tables 4.2 and 4.3 for mechanical and simplified-continuum models respectively can
be combined into to a single table (4.4) that shows the relative and absolute hierarchy between
SEMs.
Next, it is of interest to note that Rhines (1965) solved the problem of a stress, p(x), applied to
the surface of an elastic layer of finite thickness. He expressed the solution for surface settlement,
w(x), as an infinite series that had the general form of

...
) ( ) ( ) (
...
) ( ) ( ) (
6
6
4
4
2
2
6
6
4
4
2
2
3 2
+ = +
dx
x w d
c
dx
x w d
c
dx
x w d
c
dx
x p d
c
dx
x p d
c
dx
x p d
w w w p p
2 1 1
.(4.13)

The argument is made here that an elastic continuum can be viewed as the "ultimate" subgrade
model, and noting the similarity of form between Equation 4.13 which is the exact solution for an
elastic continuum subgrade model and the finite series represented by Equation 4.9 which is the
general equation of SEMs, it is clear that SEMs in general represent approximate solutions for an
elastic subgrade. This is because as with any infinite series such as Equation 4.13 a finite number
of terms of that series represents an approximate solution. While this conclusion that all SEMs
represent approximations of an elastic continuum is trivial for simplified-continuum models, it is
a novel conclusion for mechanical models. In essence, an elastic continuum can be visualized as
being made up of an infinite number of layers of springs, shear elements and flexural elements in
no unique combination.
56
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
Table 4.4. Synthesis and Overall Hierarchy of Surface-Element Subgrade Models

derivative order of p(x,y) derivative order of w(x,y) Subgrade Model
(in groupings of identical models)
0 2 4 0 2 4
mechanical
Winkler's Hypothesis

simplified continuum
Winkler-Type Simplified Continuum

mechanical
Filonenko-Borodich
Pasternak's Hypothesis
Loof's Hypothesis

simplified continuum
Pasternak-Type Simplified Continuum
Vlasov and Leont'ev (1 layer)

mechanical
Modified Pasternak/Kerr

simplified continuum
Reissner Simplified Continuum
Vlasov and Leont'ev (2 layers)

mechanical
Haber-Schaim

simplified continuum
[none identified to date]

mechanical
Hetnyi

simplified continuum
[none identified to date]

mechanical
Rhines

simplified continuum
[none identified to date]


Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
57
4.4.3.5.3 Implementation: Part II - Practical Application

The practical application of the concepts discussed in the preceding subsection is illustrated
here for a case that is likely to be quite common in practice. Assume that it is desired to use the
highest order synthesized SEM feasible as a subgrade model for a mat foundation. This means
that the mechanical model chosen must be used within the context of commercially available
structural analysis software. In addition, there must be equivalent simplified-continuum and
mechanical models to allow for evaluation of the coefficients in the equation governing the
behavior of the mechanical model.
The highest order simplified-continuum model that has been developed to date is the Reissner
Simplified Continuum. Referring to Table 4.4, it can be seen that the RSC and Modified
Pasternak (Kerr) models are theoretically equivalent as approximations for an elastic continuum.
The governing equation of the RSC for an isotropic, homogeneous linear-elastic continuum of
finite thickness H is

) , (
3
) , ( ) , (
12
) , (
2 2
2
y x w
H G
y x w
H
E
y x p
E
H G
y x p

(4.14)

where E and G are the elastic constants (Young's and shear moduli respectively) for the
continuum (Horvath 1979, 1988d, 1989c)
vv
. Noting from Table 4.1 that the Modified Pasternak
(Kerr) model consists of an incompressible shear layer of stiffness g sandwiched between two
spring layers, the governing equation for this model is (Horvath 1988d, 1989c)

) , ( ) , ( ) , ( ) , (
2 2
y x w
k k
k g
y x w
k k
k k
y x p
k k
g
y x p
l u
u
l u
l u
l u

=
+
(4.15)

where k
u
and k
l
are the spring stiffnesses of the upper and lower spring layers respectively.
Equating the constant coefficients in equations 4.14 and 4.15 results in three equations for three
unknowns (g, k
u
and k
l
). The results of solving these equations are:


H
E
k
u

=
4
(4.16a)


H
E
k
l

=
3
4
(4.16b)


9
4 H G
g

= (4.16c)

As an aside, note that equations 4.16a and 4.16b imply that the upper spring layer is three times
as stiff as the lower. In addition, because the two spring layers act in series, the equivalent overall
spring stiffness, k
eq
, is


vv
Additional RSC solutions for an elastic layer where Young's modulus varies linearly with depth and with
the square root of depth were given in Horvath (1979). In both cases, the equivalent constant coefficients in
Equation 4.14 are significantly more complex.
58
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2

H
E
E
H
E
H
E
H
k k
k
l u
eq
=

=
+
=
4
4
1
4
3
4
1
1 1
1
(4.17)

which is as expected. As discussed in Horvath (1979, 1988d, 1989c), the overall equivalent
"spring" stiffness in the three simplified-continuum models (Reissner, Pasternak-Type and
Winkler-Type) that have been identified and developed to date is always E/H. As discussed in
Horvath (1979, 1983b), the simplified-continuum concept has, after more than 100 years, finally
provided a sound physical explanation for what the Winkler coefficient of subgrade reaction
represents: it is Young's modulus divided by the thickness of a layer of isotropic, homogeneous
linear-elastic material where all stresses and strains within that material other than normal stresses
and strains in the vertical direction are assumed to be zero
ww
. Thus the absence of shear resistance
(spring coupling) in such a Winkler-type material (a Winkler-Type Simplified Continuum) is
clearly apparent.
The results given by equations 4.16 are still not in a final form that can be implemented in
commercially available structural analysis software. This is because available information
indicates that such software does not have the ability to model the incompressible shear layer that
is part of the Modified Pasternak (Kerr) model. However, as discussed in Section 4.4.3.3.2.2, a
deformed, tensioned membrane is mechanically equivalent to an undeformed, incompressible
shear layer. Consequently, the following would be modeled in the structural software to represent
the subgrade:

an upper layer of linear springs of uniform stiffness 4E/H;

a plate element with zero flexural stiffness
xx
, D, to approximate a membrane and under a
constant tensile force per unit width, T = 4GH/9, around all its sides;

a lower layer of linear springs of uniform stiffness 4E/3H.

In addition, it would be necessary to use a nonlinear analysis so that the proper behavior of the
membrane (plate) under tension is obtained.
There are two additional practical issues to be resolved that are beyond the scope of this report
to address in detail but are important to at least note. First, there is the question of how to
accommodate the two boundary conditions that are required to solve Equation 4.15. One
boundary condition involves the first derivative of p(x,y) and the other the first derivative of
w(x,y) which is the slope of the subgrade surface. The objective with respect to boundary
conditions is to not have to explicitly model the subgrade beyond the edge of the mat in this
example. Thus rational decisions must be made as to how the subgrade reaction, p(x,y), and slope
of the subgrade surface (first derivative of w(x,y)) change at the end of the mat. These changes
must then be expressed mathematically and built into the problem. An example of how this was
done for a lower-order subgrade model (the Beam-Column Analogy consisting of the synthesized
Pasternak/Loof Hypothesis and Pasternak-Type Simplified Continuum) where the only boundary
condition involved the slope of the subgrade surface is discussed in detail in Horvath (1993e).
The second issue is that appropriate, project-specific values of E, G and H need to be
determined by the geotechnical engineer. The key variable is H which, in the simplified-

ww
Solutions for a Winkler-type material where Young's modulus varies linearly with depth and as a square
root of depth were also given in Horvath (1979, 1983b).
xx
Alternatively, it may be necessary to use a very small stiffness value if the software does not execute
properly or provide meaningful results with D = 0.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
59
continuum concept, is the depth to the assumed rigid base or, more generally, the depth at which
the vertical strains that contribute to settlement for a problem can realistically be assumed to be
zero. Once H has been determined, the subgrade between the subgrade surface (which is
foundation level in a mat problem) and the depth H is divided into an arbitrary number of
artificial layers, each of arbitrary thickness and with an assumed constant value of E and
Poisson's ratio, . Then the equivalent homogeneous, isotropic value of E and Poisson's ratio, ,
can be determined using a strain-weighted average approach such as that developed by Fraser and
Wardle (1976). Once E and have been determined, G can be calculated easily using linear-
elastic theory.

4.4.3.5.4 Exceptions and Modifications

Traditionally, a subgrade model is used to represent the behavior of a stratum of one material.
This has also been the presumption throughout this report up to this point. In such a context the
synthesis of mechanical and simplified-continuum subgrade models presented in the preceding
subsections is theoretically correct and appropriate. However, there are some important
exceptions and modifications to this synthesis that require discussion.
Concomitant with the rapid development and growth in geosynthetics technology in the latter
decades of the 20
th
Century, there has been a trend in recent years to use a single subgrade model
to represent multiple and different components of a subgrade. There are at least three distinct
versions of this. Each involves the used of multiple-parameter mechanical models.
The first one considered involves a layer of geosynthetic tensile reinforcement (geotextile or
geogrid) placed over a subgrade consisting of soil or a non-earth material. For example, referring
to Table 4.1 the Filonenko-Borodich mechanical model, which consists of a deformed,
pretensioned membrane over a spring layer and has the following governing equation

) , ( ) , ( ) , (
2
y x w T y x w k y x p = (4.18)

can be used to model a layer of reinforcement (the membrane) over a relatively compressible
subgrade (the springs). In such an application the theoretical relationship (see Table 4.4) between
the Filonenko-Borodich mechanical model and Pasternak-Type Simplified Continuum (PTSC)
model, which has the following governing equation

) , (
2
) , ( ) , (
2
y x w
H G
y x w
H
E
y x p |
.
|

\
|
|
.
|

\
|
= , (4.19)

is not appropriate for evaluating the membrane tension, T, in the Filonenko-Borodich model. This
is because the synthesis between mechanical and simplified-continuum subgrade models
presented in the preceding subsections theoretically applies only when a single material is being
modeled. Thus the result that


2
H G
T

= (4.20)

where G and H are the shear modulus and thickness respectively of the subgrade is clearly not
applicable here. This is because the tensile force in the geosynthetic, T, is unrelated to the
properties of the subgrade material (G and H). However, the result that

60
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2

H
E
k = (4.21)

is of use because the spring constant, k, is in face related to the properties of the subgrade material
(E and H).
The second variation of subgrade modeling does not involve geosynthetics but two distinct,
relatively thin subgrade strata. The typical scenario to which this has been applied involves a
dense coarse-grain soil over a compressible fine-grain or organic soil. Referring to Table 4.1, the
Pasternak/Loof Hypothesis, which is visualized as an incompressible shear layer of stiffness g
over a spring layer of stiffness k and has the following governing equation (note that this is the
more-general, 2-D form of Equation 4.11)

) , ( ) , ( ) , (
2
y x w g y x w k y x p = , (4.22)

is typically used to model this situation. This is based on the assumption that the relatively
incompressible coarse-grain stratum will effectively provide all the shearing resistance of the
overall subgrade whereas the relatively soft fine-grain/organic stratum will contribute all
settlement. In such an application, the theoretical relationship (see Table 4.4) between the
Pasternak/Loof mechanical model and PTSC model (Equation 4.19) is not rigorously correct for
evaluating the shear-layer stiffness, g, or spring stiffness, k, in the Pasternak/Loof Hypothesis
(Equation 4.22). Again, this is because the synthesis between mechanical and simplified-
continuum subgrade models presumes modeling of a single material layer. However, this
synthesis can be approximately correct in this case provided that the constant coefficients in the
PTSC governing equation (4.19) are chosen appropriately. This means that the relationship


2
H G
g

= (4.23)

as well as Equation 4.21 can be used provided that:

the Young's modulus, E, in Equation 4.21 is that of the fine-grain/organic stratum;

the shear modulus, G, in Equation 4.23 is that of the coarse-grain stratum; and

the elastic layer thickness, H, in Equation 4.21 is based on the geometry of only the fine-
grain/organic stratum but H in Equation 4.23 is based on the geometry of only the coarse-
grain stratum.

The third alternative that can be envisaged is a combination of the preceding two. Consider the
problem of a layer of geosynthetic tensile reinforcement that is placed between a dense coarse-
grain soil and underlying compressible fine-grain or organic soil
yy
. This could be modeled as an
incompressible shear layer of stiffness g (the coarse-grain soil) over (or incorporating) a
deformed membrane under tension T (the geosynthetic layer) over a spring layer of stiffness k
(the fine-grain/organic soil). The governing equation of such a mechanical model is simply a
superposition of the Filonenko-Borodich and Pasternak/Loof models which are mathematically
equivalent (see Table 4.2) and thus addable in their overall, net effect:

yy
As a slight variation, the reinforcement layer might also be visualized as being placed within the coarse-
grain stratum as often occurs in practice with geogrids.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
61
( ) ) , ( ) , ( ) , (
2
y x w T g y x w k y x p + = . (4.24)

Synthesis of this equation with the PTSC model would again be approximately correct provided
that the constant coefficients in the PTSC governing equation (4.19) are chosen appropriately.
This means that equations 4.21 and 4.23 can again be used provided that

the Young's modulus, E, in Equation 4.21 is that of the fine-grain/organic stratum;

the shear modulus, G, in Equation 4.23 is that of the coarse-grain stratum;

the elastic layer thickness, H, in Equation 4.21 is based on the geometry of only the fine-
grain/organic stratum but H in Equation 4.23 is based on the geometry of only the coarse-
grain stratum; and

the membrane tension, T, in Equation 4.24 represents the contribution of the geosynthetic
reinforcement.

In conclusion, the use of a single subgrade model to represent multiple materials in an actual
application appears to be an area of growing research interest in geotechnical engineering,
especially in applications involving geosynthetic tensile reinforcement. A number of examples of
this are noted in Section 6. There is also some related discussion in the following section.
62
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
This page left blank.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
63
Section 5
Soil-Structure Interaction Modeling: Subgrade Aspects/Part II -
Time-Dependent Conditions




5.1 INTRODUCTION

Compared to time-independent conditions that were the focus of Section 4, there has been
relatively much less development for surface-element subgrade models that allow for time-
dependent behavior of the structural element, subgrade or both. This is perhaps surprising given
the increased interest in recent decades in time-dependent applications such as seismic loading
and the attention paid to time-dependent modeling of structural elements as was discussed in
Section 3.5. There appear to be a number of reasons for this lack of growth in time-dependent
subgrade models:

commercially available structural analysis software cannot accommodate time-dependent
subgrade models although the same software can usually model time-dependent structural
behavior;

time-dependent subgrade effects are very difficult to model if cyclic loading and/or inertia
effects are to be considered. Where these effects have proven to be important to model,
application-specific "geotechnical" computer software has generally been developed and
used. Such software is typically used for earthwork applications such as earth dams and not
SSI applications. The subgrade in such cases is typically modeled in a theoretically rigorous
manner as a 2-D or 3-D continuum that is solved using the finite-element or finite-difference
method.

The approach taken in this section of the report is to discuss time-dependent surface-element
subgrade modeling in a broad context. The intent here is to discuss the various categories of
applications rather than focus on specifics as was done in Section 4 for the more-common time-
independent case.

5.2 OVERVIEW

The subject of time-dependent effects in SSI applications has several distinct aspects, each of
which produces a set of unique conditions and requirements for modeling the structure and
subgrade. To begin with, the first question is whether the structure loading is time independent
(or at least reasonably so) or time dependent. If the loading is time independent, then there are no
dynamic effects on the structure. Only the time-dependent behavior of the subgrade needs to be
considered. This typically involves some rheological aspect of the material that comprises the
subgrade. A classic example of such an application is primary consolidation of a fine-grain soil
under a more or less constant applied load.
If the structure loading is significantly time dependent, then inertial effects on the structural
element are typically considered. As discussed briefly in Section 3.5, this is within the
capabilities of commercially available structural analysis software. However, things become
considerably more complex with regard to the subgrade material(s). This is because there are
several distinct phenomena that need to be considered:
64
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2

initially at least and all things being equal, there is usually increased material stiffness as the
rate of loading increases;

again, all things being equal material stiffness under cyclic loading is dependent on strain
level;

ultimately, there is usually a net decrease or (degradation) in overall material stiffness under
cyclic loading. Very often, pore pressure generation is a factor in this;

there are inertial effects within the subgrade due to the mass of the subgrade material;

the subgrade material exhibits material damping due to hysteresis under cyclic loading;

there is radiation damping through the subgrade; and

the overall behavior of subgrade materials is often dependent on the frequency of cyclic
loading.

Experience to date indicates that it is not possible to accommodate all or even most of these
behavioral aspects within the capabilities of the surface-element subgrade models that were
described in Section 4.4.3 and that form the backbone of time-independent SSI analyses. In fact,
there are numerous examples in the literature where the same SEMs (usually Winkler's
Hypothesis) used for time-independent analyses are used for dynamic SSI applications. Very
often the only concession made for the fact that it is a dynamic as opposed to static analysis is to
use a reduce value of the Winkler coefficient of subgrade reaction to account for the aggregate
degradation in subgrade stiffness under cyclic loading.

5.3 EXAMPLES OF TIME-DEPENDENT EFFECTS IN SURFACE-ELEMENT MODELS

5.3.1 Introduction

Despite the limited extension to date of SEMs to capture various time-dependent effects and the
limited utility such subgrade models would have within the context of commercially available
structural analysis software, it is still worthwhile to briefly discuss some of what has been done to
date. SEMs that are enhanced to better model time-dependent subgrade effects certainly have
significant potential use in application-specific software. Furthermore, the extensive discussion in
Section 4.4.3 of SEMs that are more advanced than Winkler's Hypothesis and the synthesis
between mechanical and simplified-continuum models will hopefully inspire future research in
this area.

5.3.2 Mechanical Models

The most common element added to mechanical models to address time-dependent subgrade
effects is the damper or dashpot. The key physical aspect of a damper is that its resistance (force)
is a function of velocity. It offers no resistance under static loading. This contrasts to a spring
whose resistance is solely a function of displacement, no matter how slowly or rapidly it occurs,
so that it offers the same resistance regardless of the duration of load application.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
65
Because springs and dampers complement each other behaviorally
zz
, their use in various
combinations forms the basis for numerous models of what is broadly called viscoelastic
behavior. The most basic, common use of a damper in SSI applications to date has been in
parallel with a single spring. This, together with a lumped mass, forms the physical model of a
single-degree-of-freedom system. When used as a subgrade model in traditional wave equation
analysis for driven piles, the combined spring + dashpot is called the Smith model.
As expected, the primary difficulty in using any of these models in practice is evaluating the
various model parameters. There is no universally successful method that has been developed to
date for accomplishing this.

5.3.3 Simplified-Continuum Models

Interestingly, in his original paper that introduced the simplified-continuum concept for
developing SEMs Reissner (1958) assumed a general viscoelastic subgrade. Consistent with the
way in which simplified-continuum models are developed using Reissner's basic procedure, no
mechanical elements are assumed. Rather, an overall behavior is postulated for the subgrade and
then a solution is developed based on various simplifying assumptions.
The specific assumption made by Reissner (1958) was that all the displacements within the
subgrade would have the following time-dependent behavior


t

+ 1 (5.1)

where is a constant coefficient that governs the time-dependent component of displacement.
Equation 5.1 means that all displacements have both time-independent and time-dependent
components that is broadly similar to a mechanical model of a spring and dashpot in series.
The solution of a Reissner Simplified Continuum that incorporates Equation 5.1 is

|
.
|

\
|

|
.
|

\
|

+ =
|
|
.
|

\
|

) , (
3
) , ( 1 ) , (
12
) , (
2 2
2
y x w
H G
y x w
H
E
t
y x p
E
H G
y x p (5.2)

which can be compared to that of the time-independent RSC (Equation 4.14). As also observed
by Reissner (1958), what would now be called a viscoelastic Winkler-Type Simplified
Continuum can be obtained by setting G = 0 in Equation 5.2 to produce

|
.
|

\
|
|
.
|

\
|

+ = ) , ( 1 ) , ( y x w
H
E
t
y x p . (5.3)

If this is coupled with the flexural equation for a plate with uniform stiffness resting on a generic
subgrade

) , ( ) , ( ) , (
4
y x q y x p y x w D = + (4.10b)

it produces a single equation defining the coupled behavior of the plate and viscoelastic WTSC:


zz
This is reflected in the fact that the suspension system for motor vehicles largely depends on springs and
dashpots used in parallel.
66
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
) , ( ) , ( 1 ) , (
4
y x q y x w
t H
E
y x w D =
(

|
.
|

\
|

+ |
.
|

\
|
+ . (5.4)

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
67
Section 6
Soil-Structure Interaction Applications: Overview



6.1 INTRODUCTION

This section of the report provides a mostly qualitative presentation and discussion of a wide
range of SSI applications. This includes applications that are already well known and used as well
as those less well known or even unexplored to date. The intention here is to illustrate, to the
broadest extent practicable, the actual and potential range of, and diversity of usage for, subgrade
models.
The primary purpose of this section is to provide a resource for both practitioners and
academicians who are interested in pursuing subgrade models for a particular SSI application in
greater detail. As noted in Section 1.3.2, some of the topics listed here are already planned for
further, more-detailed study by the Manhattan College Center for Geotechnology.
The focus in this section is on traditional applications where a subgrade model is used for a
single layer of material under time-independent conditions. However, where appropriate
references are made to both time-dependent analyses as well as those where multiple materials are
modeled using one subgrade model. An introduction to the latter applications, which have
become more common in recent years as a result of the growth in geosynthetics technology, was
presented in Section 4.4.3.5.4.
In terms of presentation organization, this section is broadly divided into applications where the
structural element in contact with the ground is oriented either horizontally (which includes a
wide variety of foundation elements such as mats as well as geosynthetics) or vertically (which
includes several types of earth retaining structures as well as deep foundations). This turns out to
be a useful division as there are broad theoretical issues unique to each category. Most SSI
applications clearly fall into one or the other of these categories which facilitates the discussion.

6.2 HORIZONTAL STRUCTURAL ELEMENTS

6.2.1 Overview

There are two categories of applications where the structural element in contact with the ground
is more or less horizontal in orientation:

foundations

geosynthetics.

Note that as used in this report, "foundation" has the very broad definition as any structural
element that transfers a fairly concentrated applied load, usually from a superstructure of some
sort, to the subgrade by spreading out or distributing the applied load in some manner, generally
by flexure of the foundation element. Thus foundations include not only the traditional shallow
foundations supporting buildings, etc. but also diverse elements such as tanks on grade,
pavements and railway track systems.
The predominant theoretical consideration for most subgrade modeling of SSI applications
involving horizontal foundation elements is that the analysis is performed using service, not
factored, loads. This is because deformations of the foundation element and displacements of the
68
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
subgrade for the serviceability limit state (SLS) govern the problem. Therefore, the synthesis of
mechanical and simplified-continuum subgrade models that was presented in Section 4.4.3.5 and
which was based on the assumption of a linear-elastic subgrade is appropriate. Regardless of
which synthesized subgrade model is chosen, the choice of the depth, H, to the assumed "rigid"
layer (better and more generally defined as the depth at which vertical strains cease to contribute
significantly to settlements) requires careful consideration on an application-specific basis. In
addition, boundary conditions for higher-order, multiple-parameter models require consideration.
Applications involving geosynthetics require careful evaluation on a case-by-case basis. This is
because the trend has been to use one subgrade model to define the behavior of multiple problem
components as was discussed in Section 4.4.3.5.4. Therefore, the synthesis of mechanical and
simplified-continuum models is not necessarily applicable, or straightforward in interpretation if
approximately applicable, in every case. In addition, there may be applications where the
ultimate, not serviceability, limit state governs. As a result, a subgrade model synthesis that is
based on an elastic layer may not be directly applicable unless bounds are put on calculated
results.

6.2.2 Foundations

6.2.2.1 Mats (Rafts) and Related Structures

6.2.2.1.1 Basic Application

Included in the category of foundation elements in addition to mats are combined and strip
aaa

footings as well as the base slabs of box tunnels constructed using the cut-and-cover method.
With relatively few exceptions, these types of problems are solved using commercially available
structural analysis software.
Historically, such foundations have been the most common "textbook" application for subgrade
models and SSI analysis. Winkler's Hypothesis with a constant coefficient of subgrade reaction
was used exclusively for many years until the development of the Pseudo-Coupled Concept that
was discussed in sections 4.4.3.3.1.5 through 4.4.3.3.1.7 inclusive. However, for the reasons
discussed in these subsections the Pseudo-Coupled Concept is not, in general, the answer to the
deficiencies inherent in using Winkler's Hypothesis with a constant coefficient of subgrade
reaction. The exceptions are applications such as tunnel base slabs where the geometry and
loading are very simple so that reliable pseudo-coupled models such as that developed by Liao
(1991, 1995) can be developed (Horvath 1993b, 1993f). The use of higher-order, multiple-
parameter subgrade models with "true" spring coupling is clearly superior, especially when a
subgrade model that is a synthesis of mechanical and simplified-continuum models is used to take
advantage of the benefits of each methodology. Clearly, this is an area of great promise for future
research.

6.2.2.1.2 The Piled-Raft Concept

Although the exact origin of what is now called the Piled-Raft Concept is not clear, there is no
doubt that most of its development in recent years occurred in western Europe, especially in
Germany where most of its refinements and extensions have originated. This concept makes
clever use of deep foundation elements to selectively supplement and load share with mats and
similar foundations. Very often, the deep foundation elements are only placed beneath portions of
a foundation and are intended to carry only a portion of the superstructure load. Thus this is

aaa
The applied loads on strip footings may be individual or continuous.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
69
fundamentally different from traditional deep-foundation application where the piles or shafts are
placed beneath the entire foundation and are assumed to carry all loads. An additional unique
aspect of the Piled-Raft Concept is that the deep-foundation elements are sometimes designed to
reach their ultimate geotechnical axial compressive capacity under service loads.
Because piled rafts represent a lesser known and underutilized geotechnology in many areas,
one activity of the Manhattan College Center for Geotechnology is to maintain a resource page
with a list of applicable references. This page can be accessed at:

www.engineering.manhattan.edu/civil/CGT/T2olrssi8.html

and represents a good starting point for any use of or research into the Pile Raft Concept.
With regard to the use of subgrade models for piled rafts, it appears that most of the state-of-art
work in Germany has made use of proprietary software, the theoretical basis of which is not
entirely clear from the published literature. Thus there appears to be an opportunity to develop
general subgrade models for use within the capabilities of commercially available structural
analysis software. One possible approach is that the subgrade models discussed in this report
could be supplemented with concentrated springs that each represent the load-settlement behavior
of a deep-foundation element.

6.2.2.1.3 Anchored Foundations

Anchored foundations are a well-known and long-used technology that is essentially the
reverse of the Piled-Raft Concept in that individual structural elements (ground anchors or deep
foundation elements) provide concentrated axial tensile resistance across all or a portion of a
foundation. However, the anchored-foundation problem has been treated historically as a simple
problem in static-force equilibrium as opposed to a SSI application that it actually is. Preliminary
research suggests that failure to consider the SSI aspects of anchored foundations can result in
anchors that are underdesigned for the desired application (Horvath 1990, 1993a, 2001b).
Therefore there appears to be an opportunity to improve the analysis and design methodology of
anchored foundations by more-extensive use of appropriate SSI analyses.

6.2.2.2 Slabs-on-Grade

Historically, slabs-on-grade constructed of reinforced PCC, especially for various commercial
and industrial applications, have been one of the more-common applications for subgrade models
and SSI analysis. Winkler's Hypothesis with a constant coefficient of subgrade reaction has been
used almost exclusively for foundation problems still to the present so this is clearly an
application where improved subgrade models are needed.
It should be kept in mind that although slabs-on-grade are in many ways geotechnically similar
to mat foundations, there are some important differences. Chief among them is the fact that slabs
typically have unique combinations of both highly concentrated as well as distributed loads of
varying shapes and dimensions. Thus slab settlement patterns may be very localized, of a larger
scale or some combination of the two. In addition, unlike mats which are typically constructed to
form a monolithic element covering the entire plan dimensions of the superstructure, slabs-on-
grade typically have several construction joints. The load-carrying capacity of the slab differs
near and at its joints compared to within an interior portion of a slab segment. Therefore, the
specific location of loads on a slab can have a significant influence on both the calculated and
actual behavior. This should also be considered during the design process. There are a number of
individual papers in Hemsley (2000) that address the unique analysis and design issues involving
slabs-on-grade in detail, albeit with an emphasis on U.K. codes and practice.
70
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
6.2.2.3 Pavements

Historically, "rigid" pavements constructed of either unreinforced or reinforced PCC for use in
transportation applications have been one of the more common applications for subgrade models
and SSI analysis. It appears that Winkler's Hypothesis with a constant coefficient of subgrade
reaction was used almost exclusively for PCC pavement design for decades beginning in the early
part of the 20
th
Century (actually predating the appearance of modern geotechnical engineering)
and this subgrade model is still used extensively to this day
bbb
. Clearly, this is an application
where improved subgrade models are needed.
PCC pavements are in many ways geotechnically similar to slabs-on-grade which were
discussed in the preceding subsection. However, unlike slabs PCC pavements are typically
designed only for relatively concentrated loads from vehicle or aircraft tires. Like slabs, PCC
pavements typically have construction joints although there may be some structural continuity
provided at the joint using steel dowels. In any event, the load-carrying capacity of the pavement
differs at its joints compared to that within a pavement section. Therefore, the specific location of
loads on a PCC pavement can have a significant influence on both the calculated and actual
behavior. This should also be considered during the design process.

6.2.2.4 Railway Track Systems

Another long-used application for subgrade models and SSI analysis is railway track systems
consisting of steel rails bearing on crossties of timber or other material that are embedded in the
ground which usually contains a surficial layer of ballast. In fact, this may very well be the oldest
application of subgrade models for SSI analysis. An infinite or semi-infinite "beam" (interpreted
in this application to be the combined rail-tie system) resting on a Winkler subgrade with a
constant coefficient of subgrade reaction is one of the few SSI problems that was solvable in a
closed form in the days before numerical methods and digital computers. In the earliest years of
the 20
th
Century and years before modern geotechnical engineering was launched by Dr. Karl
Terzaghi, Dr. Arthur Newell Talbot performed measurements on railway track systems to, in part,
determine what is still called the track modulus. Although track modulus reflects the overall net
stiffness of the rail-tie-ballast-subgrade system, subgrade stiffness is clearly an important
component. Consequently, railway track systems are another SSI application where improved
subgrade models are clearly needed.

6.2.2.5 Tanks-on-Grade

Storage tanks bearing directly on the ground are a particularly interesting category of
structures. They are quite common in both urban and rural areas alike worldwide as they are used
for a wide variety of petrochemical liquids as well as potable water. In the former application,
they are made of steel plates but in the latter application they are often reinforced PCC.
Regardless of the application, the bottom of the tank generally rests more or less directly on the
subgrade and acts as its own foundation. Especially in the case of steel tanks where the bottom
plates are generally less than one inch (25 mm) thick this is as close to a "perfectly flexible"
foundation as is even found in practice.

bbb
The historical importance of subgrade models and SSI analysis in the design of PCC pavements is
reflected in what is called the Westergaard Problem which consists of a point load applied to an elastic
plate resting on a Winkler subgrade with a constant Winkler coefficient of subgrade reaction. This is the
most-famous solution for what Timoshenko and Woinowsky-Krieger (1959) call the Plate on Elastic
Foundation family of problems. Note that the combined plate + Winkler subgrade is what Haber-Schaim
used as a subgrade model (see Table 4.1).

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
71
Note that for the purposes of this report, tanks-on-grade are defined as a shell that is right-
circular-cylindrical in shape with a bottom and roof. The roof may either be connected
structurally to the tank shell or be free-floating on the tank contents. Thus circular or rectangular
tanks or basins without a roof such as used in water or wastewater treatment facilities are not
considered to be tanks on grade. In practice, the base slabs of such structures are treated as a type
of mat foundation. The reason for this distinction in terminology is that experience indicates that
tanks-on-grade have unique behavioral issues and failure modes because of the interaction
between the roof, shell and bottom.
Despite the common worldwide use of tanks-on-grade, their treatment in traditional
geotechnical and structural engineering textbooks is curiously non-existent. As a result, students
in civil engineering rarely get formal educational exposure to this type of structure. Because
tanks-on-grade represent a lesser known technology, one activity of the Manhattan College
Center for Geotechnology is to maintain a resource page with a list of applicable references. This
page can be accessed at:

www.engineering.manhattan.edu/civil/CGT/T2olrssi9.html

and represents a good starting point for any use of or research into tanks-on-grade. In particular,
these references collectively summarize and define all the possible failure modes, both
serviceability and ultimate, structural and geotechnical, of tanks-on-grade. This is important
because most of these modes are unique to these structures and thus unfamiliar to civil engineers
not already familiar with them.
With regard to the use of subgrade models for SSI analyses of tanks-on-grade, based on the
relatively limited geotechnical published literature to date there does not appear to be widespread,
routine use of subgrade modeling and SSI analysis for this application. However, there is
anecdotal information that suggests that SSI analysis is used by structural engineers for such
tanks. Given the importance of the interaction between a tank's superstructure
(roof+shell+bottom) and its subgrade this would appear to be an area where SSI analyses would
be most appropriate. In addition, because tanks-on-grade are such a specialized type of structure it
seems likely that application-specific software would be appropriate. This would allow much
greater flexibility in the choice and implementation of subgrade models into such software.

6.2.3 Geosynthetics

6.2.3.1 Introduction

Although the materials and products we now call geosynthetics have been available for
decades, the use of subgrade models and SSI analyses in the analysis and design of geosynthetics
applications is still a relatively recent and still-evolving occurrence. This is because geosynthetic
analysis and design methodologies have traditionally been based on presumed rigid-plastic
material behavior and limit-equilibrium concepts.
Design by function is the precept by which any geosynthetic application should be designed.
Therefore, the first issue to address with regard to discussing the specific use of subgrade models
for SSI analyses involving geosynthetics is to discuss the functions that have been investigated to
date:

reinforcement using geotextiles and geogrids, and

compressible inclusion using resilient (elasticized) expanded polystyrene (EPS) geofoam.

72
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
Note that there are some unique aspects to SSI modeling of geosynthetics in a more or less
horizontal orientation. First, in many cases one subgrade model is often used to represent more
than one material component. This was discussed conceptually in Section 4.4.3.5.4. To date, this
approach has been quite common in reinforcement applications. Second, because the use of
geosynthetics is very application specific, it seems likely that the development and use of
specialized software is appropriate in many cases. This allows much greater flexibility in the
choice and implementation of subgrade models into such software.

6.2.3.2 Reinforcement Function

There was a burst of publication on this topic that began in the mid-1990s (Douglas 1995;
Ghosh and Madhav 1994a, 1994b, 1994c; Horvath 1994; Shukla and Chandra 1994a, 1994b,
1994c, 1995; Yin 1997a, 1997b). As can be seen, many of these papers
ccc
were by the same
authors and each individual paper tended to represent one narrow aspect of the overall problem of
embedding geosynthetic tensile reinforcement within a coarse-grain soil stratum over a fine-grain
soil stratum. One of the more interesting aspects of this work was the development and use in
some cases of time-dependent, multiple-parameter mechanical subgrade models with the time-
dependency representing the primary consolidation phenomenon of the fine-grain soil stratum. As
discussed in Section 5, the use of mechanical subgrade models in time-dependent SSI analyses
has been relatively rare.
It does not appear that the results of the above-cited work have encouraged greater use of
subgrade modeling and SSI analyses for the analysis and/or design of geosynthetic reinforced
structures in routine practice. This is still largely done using well-established limit-equilibrium
methodologies. Nevertheless, this cited work is significant because it highlights the fact that
geosynthetic-reinforcement applications are truly SSI problems by their nature and that SSI
analysis, as opposed to a simpler approach based on limit-equilibrium concepts, is most likely a
more-accurate representation of the overall geosynthetic-reinforcement problem that deserves
continued and expanded research in the future.

6.2.3.3 Compressible-Inclusion Function

This is one of the newer geosynthetic functions identified to date. It is one of several functions
that were defined in the 1990s as part of the growing recognition and acceptance of the fact that
geosynthetic materials and products are not limited to the traditional planar (two-dimensional)
product families such as geotextiles and geogrids that were discussed in the preceding subsection.
It is now recognized, even if not fully appreciated, that geosynthetics include materials and
products that are truly three-dimensional. Predominant in this latter category are cellular
geosynthetics such as geofoams and geocombs.
The basic concept of the compressible-inclusion function is that a relatively compressible
geofoam product that is generically called a compressible inclusion is placed between a relatively
rigid and/or unyielding (non-moving) structural element and the adjacent ground. The
compressible inclusion compresses sacrificially so that the loads exerted on the structural element
by the ground are reduced. Expanded polystyrene (EPS), frequently subjected to an additional
manufacturing step that renders it resilient or elasticized, is the geofoam material of choice for
this. An overview of the compressible-inclusion function can be found in Horvath (1998a) with
additional, more-detailed information in Horvath (1998b). A detailed bibliography of published
material on the subject through December 31, 2000 can be found in Horvath (2001a).

ccc
As long as this list of cited references is, it is by no means complete and is not intended to be so. In
particular, there are numerous additional references cited in these papers that should be reviewed by anyone
performing a comprehensive review and assessment of the subject.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
73
The particular geofoam compressible-inclusion application of interest here is where the
geofoam layer is placed horizontally (applications with a vertical orientation are covered in
Section 6.3). These fall into two broad categories:

beneath a structural slab overlying a subgrade where various types of natural phenomena can
produce volume changes within the subgrade that result in surface heave. The most-common
example of this involves subgrades containing expansive soils ("swelling clays"); and

over the crown of an underground conduit or roof of a "small"
ddd
tunnel to allow positive
arching to develop in the vertical direction within the soil overlying the conduit or tunnel.
The result of this induced arching is a significantly reduced vertical load on the conduit or
tunnel. This is basically a modernized version of a concept used since the early 20
th
Century
over conduits
eee
and embodies the imperfect ditch concept discussed in detail by Spangler and
Handy (1982).

Both of these applications are already well proven both in terms of their general concept and
specifically with regard to the viability of using EPS-geofoam products to achieve the desired
goals. A detailed bibliography of published material on the subject through December 31, 2000
can be found in Horvath (2001a). However, the primary issue retarding growth in the use of these
applications in practice is the current lack of appropriate analytical methods. At present, the only
simple analytical methods for use in routine practice for both applications are based on limit-
equilibrium only. In reality, a SSI analysis is more appropriate for both applications. This is
because there is, in general, no unique solution for these applications but an infinite number of
"correct" answers that depend on matching the compressibility of the subgrade to the
compressibility of the geofoam product to provide an estimate of the load acting on the structural
element. Therefore, this is an area where subgrade model development could have an immediate
benefit in practice.

6.2.3.4 Combined Reinforcement and Compressible-Inclusion Functions

For the sake of completeness, it is useful to note that an application that is largely unexplored to
date is to combine both of the geosynthetics described in the preceding subsections in a
synergistic manner. This involves embedding one or more horizontal layers of geosynthetic
tensile reinforcement in the ground above or below a horizontal layer of an appropriate geofoam
compressible-inclusion product. As the geofoam compresses vertically, the tensile-reinforcement
layer(s) strain and develop resistance against vertical forces acting either upwards or downwards
depending on the specific application. There appear to be a number of potential applications for
the use of this concept. The use of subgrade models to perform an appropriate SSI analysis
appears to be a key tool to use to develop this concept for practical use.

6.3 VERTICAL STRUCTURAL ELEMENTS

6.3.1 Overview

There are two categories of applications where the structural element in contact with the ground
is more or less vertical in orientation:

ddd
Small in this context refers to the relative ratio of the overall tunnel width compared to the thickness of
soil overlying the tunnel. Unless this ratio is sufficiently small in magnitude, vertical arching, which is
crucial to the successful use of a compressible inclusion in this application, cannot develop.
eee
Bales of straw were used as the compressible-inclusion material in early applications of this concept.
74
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
those where a two-dimensional analysis (usually assuming plane-strain conditions) involving
a slice of unit width through the structural element and adjacent ground produces satisfactory
results. This approach is typically used for a wide variety of earth retaining structures; and

those where a three-dimensional analysis (at least in concept) is necessary because of the
geometry of the structural element involved. This typically applies to deep foundations under
lateral and/or moment loading.

The predominant theoretical consideration for subgrade modeling of SSI applications for
vertical structural elements is that the geotechnical ultimate limit state (ULS), i.e. soil "failure", is
generally reached over at least a part of the structure even under service loads. This is in marked
contrast to applications involving horizontal structural elements where the SLS usually governs.
This means that the synthesis of mechanical and simplified-continuum subgrade models that was
presented in Section 4.4.3.5, which was based on the assumption of a linear-elastic subgrade,
must be used with care in vertical applications. Although quasi-elastic conditions may exist over
a part of the load-deformation range, at some point the load is capped at some maximum value
that typically corresponds, at least in a broad conceptual sense, to the active or passive lateral
earth pressure state. Decades of experience doing this with laterally loaded deep foundations
using p-y curves has shown that this requirement of "capping" load-deformation behavior is not
an undue burden for subgrade model development and use of SSI analysis for vertical structural
elements.
Regardless of which synthesized subgrade model is chosen, the choice of the depth, H, to the
assumed "rigid" layer requires careful consideration on an application-specific basis. In general,
this is somewhat more difficult for vertical as opposed to horizontal structural elements because
the distance H in this case is a horizontal distance. Thus a true "rigid" layer will rarely be
encountered in practice which means that H is the distance from a wall or pile, for example,
beyond which the effects of the wall or pile undergoing horizontal displacement have no practical
importance. In addition, the selection of boundary conditions for higher-order models require
careful consideration when vertical structural elements are involved as the appropriate conditions
may be different at the top and bottom of the element.

6.3.2 Two-Dimensional Applications

Historically, most 2-D applications where there has been interest in applying subgrade
modeling and SSI analysis involve various types of "flexible" earth-retaining structures such as
anchored and cantilevered sheetpile bulkheads and, to a lesser extent, braced excavations.
Flexible underground conduits and mined tunnels with a circular cross-section have also seen
some interest. However, all of these structures have been and still are traditionally analyzed and
designed based on a limit-equilibrium approach using active and passive lateral earth pressure
theory. There is no explicit consideration of deformations and displacements, although the ability
of the structure to deform/displace does (or at least should) enter into the process implicitly by the
selection of appropriate lateral earth pressure state(s). The overall methodology typically used in
current practice is identical to that used for "rigid" earth retaining structures (gravity and
cantilever retaining walls) and, more recently, various types of geosynthetic walls (mechanically
stabilized earth walls, MSEWs, and segmental retaining walls, SRWs).
Despite the generally acceptable performance of "flexible" structures designed using limit-
equilibrium-based methodologies, the fact remains that these are, as a group, structures that
deform and displace under service loads. Therefore, a SSI analysis is a conceptually better way to
analyze and design such structures and something that should be pursued. Certainly the
philosophy expressed in Section 1.3.1 applies here, i.e. that the status quo is "good enough" is not

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
75
a per se argument for not pursuing change.
Work to date along the lines of making greater use of subgrade models and SSI analysis for the
above-described applications has generally been limited to the use of Winkler's Hypothesis as the
subgrade model (Haliburton 1971, Horvath 1988a). This work suggests that higher-order,
multiple-parameter models are required to accurately replicate actual behavior. This is because
the lack of spring coupling in Winkler's Hypothesis results in its inability to capture the
horizontal arching that develops behind a deforming and/or displacing earth retaining structure. It
is now well known that the development of limiting (i.e. active or passive) lateral earth pressures
starting from the at-rest state does not occur as "rigid" soil wedges as classical lateral earth
pressure theory (Coulomb and Rankine) assumes. Rather, there is considerable deformation
within the soil wedges so that the phenomenon of arching plays an important role in both the
magnitude and distribution of lateral earth pressures (Handy 1985; Harrop-Williams 1989b,
1989c).
A more-recent application that has great potential for subgrade models and SSI analyses
involves the use of geofoam compressible inclusions behind relatively rigid and non-yielding
(non-moving) earth retaining structures. Recognition of and research into this concept dates back
to the 1980s and involves the same soil mechanics concepts that were discussed in Section
6.2.3.3. However, in this case the compressible-inclusion is oriented vertically between the
structure and retained soil. The overall goal is to allow the retained soil to displace and mobilize
its inherent strength in the absence of structure movement. This results in a reduction in lateral
earth pressure acting on the wall from the at-rest to approximately the active state
fff
. This is
referred to as the Reduced Earth Pressure (REP) Wall concept.
As an extension of the REP-Wall concept, if the retained soil contains multiple horizontal
layers of geosynthetic tensile reinforcement (geogrids, geotextiles or metallic elements) the lateral
earth pressure reduction can be even larger than from at-rest to active and even approach zero.
This is because the reinforcement supports the soil in the same way as in a MSEW application.
Consequently, the synergistic use of a geofoam compressible inclusion and tensile reinforcement
(broadly similar to that noted in Section 6.2.3.3) is called the Zero-Earth Pressure (ZEP) Wall
concept. Both the REP- and ZEP-Wall concepts are discussed in Horvath (1998a, 1998b) with
some updated information in Horvath (2000b). A detailed bibliography of published material on
the subject through December 31, 2000 can be found in Horvath (2001a). Finally, the efficacy of
using a geofoam compressible inclusion to accommodate the movements of integral and semi-
integral abutment bridges is discussed in Horvath (2000b).

6.3.3 Three-Dimensional Applications

The primary use of subgrade models and SSI analyses for 3-D applications involving vertical
structural members has been and still is for deep foundations under lateral and/or moment
loading. This is one of the oldest uses of SSI analyses in civil engineering that dates back to the
1950s. As a result, in many ways this SSI application represents a mature technology.
However, there is still room for improvement. This is because Winkler's Hypothesis is still used
as the subgrade model. Even though the soil "springs" used nowadays are quite sophisticated in
that force-displacement nonlinearity and soil yield are accounted for through the use of depth-
variable p-y curves, the basic phenomenological shortcoming of Winkler's Hypothesis, i.e. the
complete lack of inherent spring coupling which represents soil shear, still remains.

fff
Because the development of horizontal arching is involved, the resulting earth force is actually somewhat
less than that from the theoretical Coulomb active state. In addition, the geometric shape of the actual
lateral earth pressure distribution is closer to parabolic as opposed to triangular as is typically assumed with
classical earth pressure theory. Nevertheless, for simplicity in routine practical the classical active earth
pressure state and triangular distribution are typically assumed.
76
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
It is of interest to note that in this context the absence of inherent spring coupling in Winkler's
Hypothesis is particularly problematic. This is because of the true three-dimensionality of
laterally loaded deep foundations. As a result, the soil shear approximated by spring coupling
occurs both vertically parallel to the longitudinal axis of the deep foundation element as well as
horizontally in a direction transverse to both the longitudinal axis of the deep foundation element
and the applied lateral load. Conceptually, this latter mechanism represents the shearing that
occurs on vertical planes between the sides of the deep foundation element and adjacent soil as
the deep foundation element moves horizontally into the soil in a direction parallel to the
direction of the applied lateral and/or moment loading.
As noted in Horvath (1984, 1989d), higher-order, multiple-parameter subgrade models are well
suited to explicitly model spring coupling/soil shear in this application. This offers a more-
rational way to address this issue as opposed to the continued use of Winkler's Hypothesis and p-
y curves. This is because the effects of spring coupling/soil shear are reflected in p-y curves but in
an empirical way that makes it impossible to isolate their effects
ggg
. This makes it difficult to scale
p-y curves to deep foundation elements of different diameter as well as deal with things such as
group effects and cyclic loading.
Another deep foundation application where subgrade models and SSI analyses appear to be
potentially useful is for the rational analyses of the axial load-displacement behavior of driven
piles, typically steel, that are bent during installation. Such piles are often referred to colloquially
as dog-leg piles. Because of their initially bent geometry, when a downward vertical load is
applied the pile resists that load in what is hypothesized to be a rather complex mechanism that
primarily involves bending of the pile into the surrounding soil as opposed to mobilization of the
traditional axial-capacity mechanisms of side friction and end bearing. If this hypothesized
behavior is correct, the axial load-displacement and ultimate capacity (ULS) of dog-leg piles is
more closely related to how deep foundations resist lateral and moment loads. This means that
dog-leg piles may reach the ULS either geotechnically or structurally (through yield in flexure).
However, the concern about dog-leg piles is generally not so much their ultimate axial
compressive capacity but their axial stiffness under service loads. Again, it is hypothesized that a
dog-leg pile will have lower axial stiffness compared to a straight pile and thus not carry its
presumed equal share of load within a group or cluster of piles. This means that the straight, or at
least straighter, piles in a group may be relatively overloaded.
Attempts to develop a rational analytical procedure for dog-leg piles date back at least to the
1950s. However, most methods have utilized a limit-equilibrium approach based on ultimate soil
resistances. This does not capture the important force-displacement behavior that can only be
obtained from a SSI analysis using subgrade models. Thus there remains an unrealized
opportunity to develop a workable method for analyzing the load-settlement behavior of dog-leg
piles.

ggg
As noted in Section 4.4.3.3.1.6, p-y curves can be viewed conceptually as a version of the Pseudo-
Coupled Concept but for vertical, not horizontal, structural elements. This is because all the soil shear
(spring coupling) that exists in a real laterally/moment loaded deep foundation application has been
empirically incorporated into the Winkler springs based on decades of research that included back-
calculation from numerous actual deep foundations subjected to lateral loading.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
77
Section 7
Soil-Structure Interaction Applications:
Example for a Mat (Raft) Foundation



7.1 INTRODUCTION

To conclude this report, it is useful to illustrate the practical application of a number of the key
concepts that were presented in earlier sections. This will be accomplished using a case history
involving two mat foundations. Not only are mats a widely used and classical SSI application that
has been used to illustrate numerous concepts throughout this report but this particular case
history provides and supports a surprisingly comprehensive range of results and conclusions
concerning SSI analyses.
The material presented in this section represents a summary of a comprehensive analytical
study that was reported in greater detail in Horvath (1993c, 1993d). Unfortunately, neither of
these references is still available. Consequently, the paper by DeSimone and Gould (1972) in
which the basic data for this case history was presented originally is suggested as a source for
many of the factual details beyond those provided herein.

7.2 BACKGROUND INFORMATION

7.2.1 Introduction

The case history involves two mat-supported buildings, the Whitaker Laboratory and Chemistry
Building, that were constructed adjacent to each other at the Massachusetts Institute of Technology
(MIT) campus in Cambridge, Massachusetts, U.S.A. in the 1960s. An unusually detailed presentation
of important structural and geotechnical data, design assumptions and procedures, and observed
settlements over a period of several years after construction was published previously by DeSimone
and Gould (1972) who were involved in the design of these mats. Although the design procedures
represent 1960s technology to a certain extent in that only manual-calculation methods were used, the
subgrade model used in the original design (Winkler's Hypothesis with a constant Winkler coefficient
of subgrade reaction) would still be considered state-of-practice today by many engineers.

7.2.2 Structural Details

7.2.2.1 Overview

The two buildings are structurally and geotechnically similar in many respects:

underlain by essentially identical subsurface conditions;

relatively low-rise, reinforced-PCC frame superstructures;

relatively deep basements that extend to approximately 30 feet (9 m) below the final outside
grade. Although construction was performed in the "dry", approximately half of that depth is
below the permanent groundwater table after construction dewatering was stopped. Thus the
long-term design loads included a substantial uplift water pressure on the bottom of the mat;

78
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
building footprints that are significantly longer than wide in plan view by a ratio of the order of
four to one;

overall megastructures (superstructure plus mat) that are orders of magnitude stiffer in the
longitudinal direction compared to the transverse. This is the result of relatively thick (2 foot (610
mm)) reinforced-PCC exterior below-grade walls that are connected structurally to both the mat
and ground-floor slab. Therefore, in the longitudinal direction the below-grade portion of each
building acts as a very deep (approximately 30 feet (9 m)) box beam. As a result, mat flexure in
both buildings was essentially limited to the transverse direction, similar to the behavior of a cut-
and-cover box tunnel base slab. This was used to advantage as only plane-strain analyses in the
transverse direction were performed in the post-construction study that was reported originally in
Horvath (1993c, 1993d). This one-dimensional flexural behavior of the mats also simplified
presentation of calculated results and the assessment of the relative accuracy of the subgrade
models considered; and

full moment connections between the exterior below-grade walls and mat play a significant role
in the flexural mat behavior in the transverse direction in two ways:

o bending moments are transmitted to each end of each mat from the lateral earth and water
pressures acting on the exterior below-grade walls, and

o rotation of each end of each mat is resisted.

Despite their numerous similarities, there are significant differences in the structural scheme
incorporated in the superstructure and resulting mat loading of the two buildings. This resulted in a
surprisingly wide difference in observed behavior and calculated results which ultimately resulted in
an interesting range of conclusions concerning subgrade modeling.

7.2.2.2 Whitaker Laboratory

The Whitaker Laboratory is the simpler of the two buildings in terms of the overall superstructure
scheme and loads applied to the mat. The below-grade portion of this building is also remarkably
similar to the geometry and loading on a typical base slab of a cut-and-cover box tunnel. This
happenstance allowed for some additional comparisons of subgrade modeling alternatives.
The Whitaker Laboratory has eight stories above and two levels below the finished outside grade.
The mat, which is founded approximately 34 feet (10.4 m) below grade, is 60.0 feet (18.3 m) by
219.7 feet (67.0 m) in plan dimensions and 3.75 feet (1140 mm) thick. In the transverse direction, all
vertical loads are transmitted to the mat via the two exterior below-grade walls plus an interior
column line (actually a continuous wall
hhh
) near the centerline of the mat. Overall, this is typical of a
classical loading on a building mat in that these three column lines carry their proportionate share of
the combined dead and live service loads from the superstructure. The column service loadings are
equivalent to placing a uniformly distributed stress of approximately 2,700 lbs/ft
2
(130 kPa) across the
top surface of the mat.
The gross (uncracked) flexural stiffness, EI, of the mat alone in the transverse direction was
estimated to be 1.9x10
6
kip-ft
2
/foot width of the mat (2600 MN-m
2
/m). Using the method discussed in
ACI Committee 336 (1988), the gross superstructure stiffness alone was estimated to be 1.7x10
5
kip-
ft
2
/foot width (230 MN-m
2
/m). Thus in the analytically important transverse direction the mat is

hhh
This feature adds to the overall flexural rigidity of the structure in the longitudinal direction that was noted in
the preceding subsection. This feature also adds to the physical similarity to a cut-and-cover box tunnel.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
79
relatively much stiffer than the superstructure, by approximately one order of magnitude. However,
the settlement data reported by DeSimone and Gould (1972) that was used as the basis for
comparison with calculated results reported by Horvath (1993c, 1993d) was obtained approximately
six and one-half years after the superstructure was completed. Thus one-third of the theoretical gross
stiffness values for both the mat and superstructure was used in all comparative analyses performed
by Horvath (1993c, 1993d). This was to account approximately for assumed time-dependent (creep)
effects on the Young's modulus and, therefore, overall stiffness of the PCC. The impact of this
assumption is discussed subsequently.
Note that these reductions in flexural stiffness are separate from additional stiffness reductions that
may occur due to cracking of the mat and/or superstructure. Consideration of this latter aspect is
discussed further in Section 7.3.3. The cracking moment
iii
for this mat was estimated to be 150 kip-
ft/foot width of mat (670 kN-m/m). The fully cracked stiffness was assumed to be one-half of the
gross stiffness.

7.2.2.3 Chemistry Building

The Chemistry Building is similar to the Whitaker Laboratory in overall appearance. It has five
stories above and two levels below grade. The Chemistry Building mat, which is founded
approximately 30 feet (9.1 m) below grade, is 65.5 feet (20.0 m) by 279.5 feet (85.2 m) in plan
dimensions and 2.5 feet (760 mm) thick. Note that the mat thickness is only two-thirds that of the
Whitaker Laboratory. However, in terms of the structural scheme incorporated into the superstructure
and the resulting loads applied to the mat the Chemistry Building is far more complex that the
Whitaker Laboratory. This is because the superstructure frame of the Chemistry Building spans
almost the entire width of the mat. As a result, all the superstructure dead and live load is transmitted
to the mat through the exterior below-grade walls plus two column lines located within 10 feet (3 m)
of each exterior walls. Relatively little load is applied within the center two-thirds of the mat. When
this fact is combined with the approximately 15 feet (4.6 m) of uplift water head that acts on the
underside of the mat in the long term the result is a net downward service load in the transverse
direction that is concentrated almost totally at the edges of the mat.
The gross flexural stiffness, EI, of the mat alone in the transverse direction was estimated to be
5.6x10
5
kip-ft
2
/foot width of mat (760 MN-m
2
/m). The gross superstructure stiffness was estimated to
be 8x10
4
kip-ft
2
/foot width (110 MN-m
2
/m), again, relatively much less than that of the mat. Note that
the Chemistry Building is overall much more flexible than the Whitaker Laboratory in the transverse
direction. This fact combined with the relatively unusual loading distribution applied to the Chemistry
Building mat and a thinner mat to begin with are the reasons why these overall similar structures
exhibited significantly different settlement behavior in their respective transverse directions.
Because the settlement data used to compare actual and calculated behavior for the Chemistry
Building was obtained approximately two years after the superstructure was completed, one-half of
these theoretical gross stiffness values was used in all comparative analyses performed by Horvath
(1993c, 1993d). This was to account approximately for assumed time-dependent effects on the
Young's modulus and stiffness of the PCC. The impact of this assumption is discussed subsequently.
Note again that this reduction in stiffness is separate from additional stiffness reductions that may
occur due to cracking of the mat and/or superstructure. Consideration of this aspect is discussed
further in Section 7.3.3. The cracking moment for this mat was estimated to be 70 kip-ft/foot width of
mat (310 kN-m/m). The fully cracked stiffness was assumed to be one-half of the gross stiffness.





iii
Defined here as the theoretical bending moment that will just initiate tensile cracking of the PCC.
80
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
7.2.3 Site Characterization

7.2.3.1 Overview

For all practical purposes, the subsurface conditions at the adjacent sites of the Whitaker Laboratory
and Chemistry Building are geotechnically identical and were assumed so for the purposes of the
study reported in Horvath (1993c, 1993d). Below foundation level of each building is approximately
70 feet (21.3 m) of clay that is known locally as the Boston Blue Clay (BBC) although its actual color
varies depending on the degree of post-depositional desiccation that has occurred within a given
portion of the stratum. The BBC stratum is underlain by Pleistocene glacial kame (outwash) and basal
till (ground moraine) strata that are predominantly coarse-grain in texture and were assumed to act as
a rigid base in all analyses reported in Horvath (1993c, 1993d). Thus in the context of the
synthesized subgrade models discussed in Section 4.4.3.5, the depth below foundation level to the
rigid base, H, = 70 feet (21.3 m).

7.2.3.2. Engineering Properties of the Boston Blue Clay

Only the engineering properties of the BBC stratum impact on subgrade modeling and SSI analyses
of the mat foundations of these two buildings. In particular, the appropriate undrained and/or drained
Young's moduli and Poisson's ratios of the BBC are required for developing values of the various
subgrade-model coefficients as discussed in Section 4.4.3.5. In this case, bearing capacity was not an
issue so it was not necessary to develop an appropriate profile of undrained shear strengths.
There were several steps in logic to go through to arrive at the final values of the elastic parameters
of the BBC that were used in the analyses reported in Horvath (1993c, 1993d):

Piezometric data in DeSimone and Gould (1972) indicate that for each building full heave of the
underlying BBC had occurred during the construction time required for excavation, i.e. 100% of
the primary consolidation associated with unloading had occurred.

It was assumed that for the purposes of calculating settlements, only 100% of the primary
consolidation (reconsolidation in this case) under the full net vertical effective stress caused by
each buildings would occur within the time frame of interest after completion of each building.
Settlements due to creep (secondary compression) were thus assumed to be negligible within the
time frame of interest.

The net post-construction vertical effective stress at foundation level imposed by each building
was less than the vertical effective overburden stress that existed prior to construction. Thus both
buildings represent what is sometimes referred to as a "compensated" or "floating" foundation.
Stated another way, the weight or mass of the soil removed from within the building footprint
exceeded that of the building ultimately constructed plus the live loads applied to the building.
Consequently, each building stressed the BBC within its recompression (unload-reload) range
only.

Overall, in consideration of the above facts it was judged that one analysis performed using the
drained Young's modulus and Poisson's ratio of the BBC within its recompression range best fit
the conditions under which the comparison between observed and calculated settlements was to
be made. It is worth noting that in design it is generally desirable to perform separate analyses
using both the undrained and drained Young's moduli to be sure that the entire range in
conditions is accounted for. In addition, some consideration of creep (secondary compression)
during the design life of the structure would also be made.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
81
The subgrade models discussed in Section 4.4.3.5 require single, average values of both Young's
modulus and Poisson's ratio. The way in which this was obtained was as follows:

The 70 feet (21.3 m) of BBC was divided into 14 artificial layers, each 5 feet (1.5 m) thick.

The drained Poisson's ratio was assumed to be 0.25 for all layers.

A constant value of the recompression drained Young's modulus was estimated for each artificial
layer using a procedure that involved two theoretical relationships. First, the drained constrained
modulus, D
jjj
, was calculated using a theoretical relationship in Lambe and Whitman (1969) that
requires the estimated average vertical effective stress level in the soil under the full building
load; the coefficient of compressibility of the soil, a
v
, in the recompression range; and the
recompression index, C
r
, of the soil. In this case, information concerning site-specific values of
both a
v
and C
r
were provided by DeSimone and Gould (1972). The calculated value of D was
then used with the assumed value of the drained Poisson's ratio in a theoretical equation based on
linear-elastic theory to calculate the recompression drained Young's modulus for that layer. As
information, Stark and Vettel (1991) provide a summary of alternative techniques for estimating
the drained Young's modulus of fine-grain soil.

A weighted-average methodology based on linear-elastic theory that is described in Horvath
(1988b) and is conceptually similar to that used by Fraser and Wardle (1976) was used to
calculate an equivalent average recompression drained Young's modulus = 800 kips/ft
2
(38 MPa)
for the entire BBC stratum beneath foundation level. The general validity of calculating
equivalent values of elastic parameters for a layered system is discussed in Burland et al. (1977)
and Tomlinson (1986). No correction (i.e. increase) in Young's modulus was made to account for
embedment of the mat below the surrounding grade. This is consistent with a recommendation by
Christian and Carrier (1989) that the embedment effect is small and can be ignored.

Assuming linear-elastic conditions exist, the shear modulus was calculated to be 320 kips/ft
2
(15
MPa).

7.3 ANALYSES

7.3.1 Overview

The primary purpose of the study reported originally in Horvath (1993c, 1993d) and
summarized herein was to perform a very comprehensive suite of analyses for both the Whitaker
Laboratory and Chemistry Building. These different analyses can be envisaged as a matrix
formed by making various structural assumptions as well as using several different subgrade
models, including variations within a given model in some cases. The purpose of investigating so
many different combinations of variables was to evaluate the absolute and relative accuracy of
various subgrade models as well as the sensitivity of results to various structural assumptions. It
was fully appreciated beforehand that this study was not definitive in the absolute sense that a
wide variety of buildings and subsurface conditions was not studied. Nevertheless, as it turned
out there was a surprisingly large behavioral difference between the Whitaker Laboratory and
Chemistry Building so that a reasonable range of results and useful conclusions were obtained.
The results that were calculated and plotted were:


jjj
This should not be confused with the flexural stiffness of a plate which has the same notation.
82
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
total mat settlement,

bending moments within the mat, and

the subgrade reaction stress acting on the underside of the mat

even though only measured settlements were available for comparison. As noted in Section 1.2,
only settlements and bending moments are required results for mat analysis and design in
practice. However, for the research purposes of the study determination of the subgrade reaction,
p, allowed easy calculation of the generic coefficient of subgrade reaction, k, using Equation 2.1.
As noted in Section 2.2.3.2, this parameter is, theoretically, a calculated result. This is certainly
true when higher-order, multiple-parameter subgrade models are used. It is only when Winkler's
Hypothesis in all its variations, which includes the Pseudo-Coupled Concept, is used that the
coefficient of subgrade reaction becomes an assumed input parameter. Therefore, calculation of k
and plotted comparisons between various subgrade models provides invaluable insight into how
assumed values from Winkler's Hypothesis compare to calculated values obtained using the
higher-order, multiple-parameter models.

7.3.2 Subgrade Modeling

7.3.2.1 Models Used

Table 7.1 summarizes the subgrade models used in Horvath (1993c, 1993d) and specifies the
methodology for quantifying the applicable coefficients in the differential equation for each
model. Details for this are provided in the following subsection.


Table 7.1. Subgrade Models Used in Horvath (1993c, 1993d)

Conventional Method of Static Equilibrium (for bending moments only)

Single-Parameter (Winkler's Hypothesis)
constant coefficient of subgrade reaction:
original design value from DeSimone and Gould (1972)
Winkler-Type Simplified Continuum (WTSC)

variable coefficient of subgrade reaction (Pseudo-Coupled Concept):
WTSC value doubled at edges only
WTSC value at center of mat and increased gradually to approximately doubled at edges
Liao's Method for cut-and-cover box tunnels (Whitaker Laboratory only)

Multiple-Parameter
Beam-Column Analogy (coefficients evaluated using Pasternak-Type Simplified Continuum)
Reissner Simplified Continuum


Note that Liao's (1991, 1995) pseudo-coupled method was used for the Whitaker Laboratory
only. Strictly speaking, Liao's Method is applicable to the base slabs of cut-and-cover box tunnels
only (Horvath 1993b, 1993f). However, as discussed in Section 7.2.2.2 the physical construction,
overall geometry and loading of the below-grade portion of the Whitaker Laboratory is very

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
83
similar to a box-tunnel base slab. Consequently, the use of Liao's Method was judged to be
reasonable under these application-specific circumstances. It also allowed evaluation and
assessment of what was then a newly proposed methodology. The results obtained in Horvath
(1993c, 1993d) subsequently justified the decision to include Liao's Method.

7.3.2.2 Model Coefficients

7.3.2.2.1 Single-Parameter (Winkler's Hypothesis)

In general, the most important aspect when using Winkler's Hypothesis as a subgrade model is to
determine the application-specific constant value of Winkler's coefficient of subgrade reaction,
o
W
k .
This is used both in traditional analyses where this constant value is applied across the entire
subgrade as well as for some of the simpler pseudo-coupled applications such as those
recommended in ACI Committee 336 (1988) where k
W
is varied across the subgrade relative to
some base value of
o
W
k .
As discussed in Section 4.4.3.3.1.3, numerous methods have been proposed over the years for
estimating
o
W
k . Table 7.2 contains a summary of various values that were calculated for the Whitaker
Laboratory and Chemistry Building using the site characterization results discussed in Section 7.2.3.
Except for the "design" values, all were determined as part of the study reported in Horvath (1993c,
1993d). The design values were taken from DeSimone and Gould (1972) and were reportedly
developed solely on the basis of their experience and judgment, and, as a result, cannot be derived or
otherwise rationally determined from any known analytical method.


Table 7.2. Comparison of
o
W
k Values from Horvath (1993c, 1993d)

o
W
k , in kips/ft
3
(kN/m
3
)
building
design WTSC elastic (see Note 1) Vesic (see Note 2) Terzaghi
Whitaker 86 (14000) 11 (1700) 21 (3300) 25 (3900) 5 (800)
Chemistry 150 (24000) 11 (1700) 20 (3100) 38 (6000) 5 (800)
Notes:
1. Based on collocation with a uniformly loaded, perfectly flexible loaded area. See Horvath (1988b) for details.
2. This is the value to be used only for calculating moments. For estimating settlements, use 50% to 100% of this
value. See Vesic and Saxena (1970) for details.


Note that several of the
o
W
k values in Table 7.2 (elastic, Vesic and Terzaghi) are presented for
information and informal comparison only. As noted in Table 7.1, the analyses performed for the
study reported in Horvath (1993c, 1993d) used only the design and Winkler-Type Simplified
Continuum (WTSC) values of
o
W
k . The WTSC has been found to provide both a consistent and
theoretically based method for estimating
o
W
k compared to other methods. However, this statement
must always be kept within the absolute context that there simply never is a "correct" value for
o
W
k .
As discussed in Section 4.4.3.5.3 and derived in detail in Horvath (1979, 1983b), the WTSC
equation for estimating
o
W
k is

84
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2

H
E
k
o
W
= . (7.1)

As discussed in Section 7.2.3, for the MIT case-history site E = 800 kips/ft
2
(38 MPa) and H = 70 feet
(21.3 m) which yields the value shown in Table 7.2.
The final item to discuss involves the additional use of Liao's pseudo-coupled method for the
Whitaker Laboratory. Table 7.3 contains the values of Winkler's coefficient of subgrade reaction,
k
W
(x), as a function of the distance x as measured in the transverse direction from the centerline of the
mat (the k
W
(x) results for Liao's method are always symmetrical relative to the centerline of a box-
tunnel base slab). Note that b is the half-width of the slab (mat in this case). Details concerning the
use of Liao's method can be found in Liao (1991, 1995). Note that the primary outcome from using
Liao's method is that the Winkler coefficient of subgrade reaction increase rapidly at the edges.
This mimics the theoretically infinite increase predicted by linear-elastic theory which was used
to calibrate his method.


Table 7.3. Values of k
W
(x) for the Whitaker Laboratory Using
Liao's Pseudo-Coupled Method for Cut-and-Cover Box Tunnel Base Slabs

k
W
(x), in kips/ft
3
(kN/m
3
)
0.0b to 0.6b 0.7b 0.8b 0.9b 1.0b
16 (2500) 16 (2500) 19 (3000) 21 (3300) 53 (8300)


7.3.2.2.2 Multiple-Parameter

As stated in Table 7.1, two multiple-parameter models were included in the study reported
originally in Horvath (1993c, 1993d). One was the Beam-Column Analogy that was discussed in
Section 4.4.3.3.2.3. Referring to Table 4.4, it can be seen that there are at least five different ways
to interpret the coefficients in Equation 4.12b that defines the Beam-Column Analogy. Of these,
the Pasternak-Type Simplified Continuum (PTSC) is clearly superior for the reason that the
simplified-continuum approach in general is superior when it comes to evaluating surface-element
model coefficients (see the discussion in Section 4.4.3.5).
Referring to Equation 4.12b, as derived in Horvath (1979) the fictitious axial force, g, acting per
unit width of the beam-column (mat in this case) is

MN/m) (164 kips/ft 11200
2
ft 70 ksf 320
2
=

=
H G
g . (7.2)

The subgrade spring stiffness, k, is the same for all simplified-continuum models that have been
developed to date:

) kN/m (1700 kips/ft 11
ft 70
ksf 800
3 3
= = =
H
E
k . (7.3)

The other multiple-parameter model studied was the Reissner Simplified Continuum (RSC). The
required coefficients are defined by Equation 4.14 and can be evaluated directly. The details are
omitted here.


Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
85
7.3.3 Structural Assumptions

The basic details and assumptions regarding the structural aspects of the analyses reported
originally in Horvath (1993c, 1993d) were stated in Section 7.2.2. Note that this means that all
analyses used gross (uncracked) structural stiffnesses that were empirically reduced from their
theoretical values to account for time effects on the Young's modulus of PCC. The following
additional assumptions were made for all analyses by virtue of items built into the computer software
used:

If axial (horizontal) loads were applied to the mat, it behaved like a beam-column. Per the
discussion in Section 3.4.2.4, this is the same as saying that a non-linear analysis was performed
using commercially available structural analysis software. However, considering the magnitude
of the axial forces involved this was expected to have a significant impact only on analyses
employing the Beam-Column-Analogy (Pasternak) subgrade model (which was intentional).

Shear effects were considered in formulating the flexural stiffness of the mat. Per the discussion
in Section 3.4.2.3, this is the same as saying that "thick-plate" elements were used in
commercially available structural analysis software. This likely had little effect compared to
traditional simple-beam assumptions which ignore such effects.

There was some variation in structural assumptions for each building. These variations were
grouped into three "levels" of analytical sophistication to evaluate the relative importance of some of
the interaction effects. The assumptions in each level are as follows:

The basic (Level 1) case was meant to represent the simplest traditional analysis where no mat-
superstructure interaction was modeled explicitly, although the effect of superstructure stiffness
was approximated by adding it to the mat stiffness (it had a very small effect as discussed in
Section 7.2.2). Thus the applied loads were assumed to be independent of differential settlement
of the superstructure. Note that although the moment loading caused by the earth and water
pressures acting on the exterior below-grade walls was included, the rotational restraint provided
by the walls was not as this detail is often either intentionally omitted or overlooked in mat
analysis when it exists.

The Level 2 analysis was similar to Level 1, although the rotational restraint provided by the
exterior below-grade walls was modeled as a simple rotational spring at each end of the mat
kkk
.
The rotational spring constant per unit length of the wall was calculated assuming that each wall
was a beam of unit width and using the theoretical unit-rotation equation (4EI/L) from structural
matrix-analysis theory. Note that in this case EI represents the calculated wall stiffness and L the
actual wall height.

The Level 3 analysis was an attempt to allow a redistribution of column loads as a result of
differential settlement of the mat. This was accomplished by applying the superstructure dead and
live loads to a single-story elastic frame that was supported on the mat. Both the gross stiffness of
the superstructure (frame) and loads were applied incrementally to simulate construction of each
building. Given the relatively low ratio of superstructure-to-mat stiffness for each building, it was
anticipated beforehand that this relatively sophisticated analysis would have a minor effect for
each buildings. However, the analyses were performed anyway for the sake of completeness and

kkk
These rotational springs were actually placed one foot (300 mm) in from each end so that the spring was
centered on where the wall rested on the top of the mat.
86
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
academic investigation. The Level 3 analyses were actually divided into two sub-levels, "a" and
"b". In the Level 3a analyses, no cracked-section behavior of the mat was assumed (this was the
same as for the levels 1 and 2 analyses). Cracked-section behavior is rarely considered when
analyzing shallow foundations in general (Kerr and Saxena 1977), but was investigated as part of
the Level 3b analyses to evaluate its influence. The cracked-section behavior of the mat was
defined using the classical Branson equation in which the operational stiffness of a beam
undergoing cracking is a function of its gross stiffness, fully cracked stiffness, the theoretical
cracking moment and the actual bending moment. The first three parameters were discussed in
Section 7.2.2. The applied moment came out of the calculations. Note that this required an
iterative analysis as there is an interdependency between calculated moments and flexural
stiffness of the mat. Finally, the significance of assuming a reduced uncracked stiffness to
account for time effects on the Young's modulus of PCC was also evaluated as part of the Level
3b analyses.

7.3.4 Software

7.3.4.1 Overview

The computer program used for the study reported originally in Horvath (1993c, 1993d) is named
SSIH (Soil-Structure Interaction Analysis of Horizontal Foundation Elements). It is a FORTRAN
code developed for academic-research purposes only and for use on microcomputers capable of
executing software in DOS mode. The differential equations defining the behavior of both the
foundation element and subgrade are solved using the finite-element method.
Prior to the start of the study reported in Horvath (1993c, 1993d), it was recognized that more-
accurate models of the building superstructures could have been accomplished using other software
(SSIH can only accommodate a one-story elastic frame). However, this would have precluded
evaluation of the RSC model which was felt to be crucial to this study (at the time, the synthesis and
implementation concepts discussed in sections 4.4.3.5.2 and 4.4.3.5.3 had not been fully developed).
Thus it is recognized that even higher "levels" of analytical sophistication are possible using
commercially available structural analysis software.

7.3.4.2 SSIH Program Details

Appendix A of this report contains excerpts from the current SSIH user manual that detail the
logic and capabilities of this program. Of particular importance is the discussion in Section A.2.3
concerning boundary conditions for the multiple-parameter models. As noted in Section
4.4.3.3.2.1, whenever a multiple-parameter surface-element subgrade model is used one or more
boundary conditions must always be considered at each end of the structural element. As it turns
out, these boundary conditions do not offer unique choices. Experience to date indicates that there
can be a relatively significant difference in calculated results depending on the particular
boundary condition alternative chosen for a given model. Thus this is a topic that requires careful
consideration whenever higher-order, multiple-parameter subgrade models are used for SSI
analyses.
With regard to the specific application of SSIH for analyzing the behavior of the Whitaker
Laboratory and Chemistry Building, the mats were modeled using 31 and 36 elements of
approximately equal lengths, respectively.





Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
87
7.3.5 Results

7.3.5.1 Whitaker Laboratory

7.3.5.1.1 Observed Behavior

All settlements discussed here were measured in early 1971, approximately six-and-one-half years
after the building superstructure was completed. The measured settlements indicated that this building
settled as a rigid body in its longitudinal direction. This was expected in view of its structural rigidity
in that in direction as discussed in Section 7.2.2.1. However, the building experienced a slight rigid-
body tilt or rotation of approximately 1/3500 or 0.016 toward the west which was the result of the
west end of the building settling approximately 0.75 inches (19 mm) more than the east end. This was
notable considering that the average total settlement of the building was of the order of 1.5 inches (40
mm). This differential movement appears to be the result of a combination of factors related to
structural details of the building (it abutted an existing building on its west end) and the BBC stratum
being somewhat (approximately 5 feet (1.5 m)) thicker going from east to west. The study reported in
Horvath (1993c, 1993d) was based on average structural conditions and the average thickness of
the BBC stratum. Therefore, calculated settlements are compared only to the average measured
settlements.
Of greater interest for the purposes of the study reported originally in Horvath (1993c, 1993d)
was the fact that even in the narrower transverse (width) direction the mat foundation for this
building behaved in a relatively rigid manner. Maximum differential settlement in the transverse
direction was only about one-eighth of an inch (3 mm) between the center and edges. The overall
settlement pattern in the transverse direction was the classical "dishing" or "sagging".
Overall, the Whitaker Laboratory was very close to the idealized case of a relatively rigid mat with
simple loading. Thus it was expected to be a test case for simple subgrade models which are
traditionally assumed to provide accurate results for this simple case.

7.3.5.1.2 Subgrade Modeling

The absolute and relative accuracies of the various subgrade models were evaluated primarily by
comparing the calculated and measured settlements. Bending moments could only be compared
among the calculated values. In addition, it was useful to compare the calculated coefficient of
subgrade reaction for the RSC (the most-accurate subgrade model used) to the assumed values for the
various versions of Winkler's Hypothesis that were used.
As expected, the RSC model provided the best agreement with observed settlement behavior and
that agreement was quite good on an absolute basis. With regard to Winkler's Hypothesis, there was
no significant difference between the three variations listed in Table 7.1 that used the WTSC value of
o
W
k . Winkler's Hypothesis with the original design value underestimated settlements by an order
of magnitude. However, in fairness to the original designers they followed the traditional (i.e.
Terzaghi) philosophy regarding the use of Winkler's Hypothesis that was discussed in Section 4.4.3.1
and did not use the results from their analyses using Winkler's Hypothesis as an estimate of total mat
settlements. The results from the Beam-Column Analogy (Pasternak subgrade) are similar to the
Winkler subgrade results, but with a flatter settlement pattern that more closely matched the observed.
As is typical, the relative variation in bending moments as a function of subgrade model is
significantly less than the relative range in calculated settlements. Two items were of particular
interest:

There was very good agreement between the RSC model results (which were assumed to be the
benchmark for the moment comparisons) and those from Winkler's Hypothesis with the original
88
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
design value of
o
W
k . Considering the engineering judgment that went into the latter value (the
discussion in DeSimone and Gould (1972) is most interesting) the agreement is a testament to
both the experience and judgment of the original designers and Terzaghi's original intention
that Winkler's Hypothesis be used to calculate local bending moments only
lll
.

The results from the CMSE ("rigid") method were not universally conservative (i.e. too large)
as is frequently assumed in practice.

Finally, with regard to the coefficient of subgrade reaction, the RSC model results, which were
again taken as the benchmark for accuracy, indicated a fairly simple pattern of a relatively constant
value within the center two-thirds of the mat that gradually increased to about double the value at the
edges. Overall, there was excellent agreement with Liao's pseudo-coupled method for tunnel base
slabs.

7.3.5.1.3 Structural

The various structural assumptions evaluated by the different levels of analysis revealed the
following. Note that these should be taken as overall trends rather than as absolutes:

Both calculated settlements and moments were sensitive to the assumption of rotational restraint
provided by the exterior below-grade walls. Neglecting the restraint is not necessarily
conservative as it can underestimate moments in some portions of the mat.

Simulating the superstructure presence as a loaded frame as opposed to using constant loads and
a constant value of flexural stiffness resulted in some load shifting from the center to the outer
columns even though the differential settlement in the transverse (width) direction was quite
small. This load transfer had a surprisingly noticeable effect on calculated bending moments in
the mat.

There was a slight but noticeable effect of using a two-thirds reduction in the stiffness of the
foundation and superstructure as an approximation of time-related effects on the Young's
modulus of PCC. The reduced modulus resulted in more-flexible behavior as expected and, in
this case, slightly better agreement with observed settlements.

No conclusion was reached as to whether or not cracked-section behavior of the mat was
important. This is because using the RSC model calculated moments exceeded the assumed
theoretical cracked-section value only in the vicinity of the column line near the center of the
mat.

The overall conclusion that can be drawn basically confirms observations made previously by
others. There is significant interaction between a mat and the superstructure it supports, and all
reasonable efforts should be made to model that interaction, even for relatively low-rise
superstructures. Furthermore, all the material issues such as stiffness reduction due to modulus
reduction and cracking that are routinely considered for reinforced-PCC structures should be
considered for mats as well.



lll
The global component of bending moment for the Whitaker Laboratory was probably quite small given
the small differential settlement in the transverse (width) direction.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
89
7.3.5.2 Chemistry Building

7.3.5.2.1 Observed Behavior

All settlements discussed here were measured in early 1971, approximately two years after the
building superstructure was completed. The measured settlements indicated that this building also
settled as a rigid body in its longitudinal direction. This was expected in view of its structural rigidity
in that in direction as discussed in Section 7.2.2.1. No rigid-body tilt or rotation was evident for this
structure.
Overall, the average settlement of the Chemistry Building was slightly less than that of the
Whitaker Laboratory, of the order of 1.25 inches (30 mm). Of greatest interest for the purposes of the
study reported originally in Horvath (1993c, 1993d) was the fact that in the transverse (width)
direction the mat foundation for the Chemistry Building behaved in a relatively much more flexible
manner compared to the Whitaker Laboratory. Maximum differential settlement in the transverse
direction is difficult to estimate precisely because settlement measurements did not extend to the full
width of the mat but they appear to approach 0.5 inches (12 mm). In addition, the overall settlement
pattern was the relatively uncommon "hogging" as a result of virtually all of the superstructure load
being applied at and near the edges of the mat.
The combined greater relative flexibility plus fundamentally different settlement pattern of the
Chemistry Building compared to the Whitaker Laboratory produced significantly different results.
These differences provided for interesting differences despite the overall similarity of the structures
and geotechnical conditions.

7.3.5.2.2 Subgrade Modeling

The absolute and relative accuracies of the various subgrade models were again evaluated primarily
by comparing the calculated and measured settlements. Bending moments could only be compared
among calculated values. In addition, it was useful to compare the calculated coefficient of subgrade
reaction for the RSC (the most accurate model used) to the assumed values for the various versions of
Winkler's Hypothesis that were used.
As expected, the RSC model provided the best agreement with observed settlement behavior. That
agreement was fair on an absolute basis, with somewhat greater differential settlements calculated
compared to those observed. In this case, the Beam-Column Analogy (Pasternak subgrade) results
were similar to those from the RSC. With regard to Winkler's Hypothesis, there was again no
significant difference between the three variations that used the WTSC value of
o
W
k . All grossly
overestimated the differential settlement. Winkler's Hypothesis with the original design value
again significantly underestimated settlements and even predicted uplift in the center of the mat.
Note again, however, that the original designers followed traditional philosophy and did not use the
results from their analyses using Winkler's Hypothesis as an estimate of total mat settlements.
In this case, the relative variation in calculated bending moments as a function of subgrade model
was quite large due to the relatively flexible mat. Two items were of particular interest:

There was fair agreement between the RSC model results (which were assumed to be the
benchmark for the moment comparisons) and those from Winkler's Hypothesis with the original
design value of
o
W
k .

The results from the CMSE ("rigid") method were grossly different from those of the RSC
model and were again not universally conservative (i.e. too large) as is frequently assumed in
practice.
90
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
Finally, with regard to the coefficient of subgrade reaction, the RSC model results, which were
again taken as the benchmark for accuracy, indicated a much more complex pattern of essentially
continuous variation from a minimum at the center of the mat to maxima at the edges. The edge
values were more than 10 times that at the center. In this case, an assumption of anything close to a
uniform value of the coefficient of subgrade reaction as would typically be made with Winkler's
Hypothesis would be grossly incorrect and nowhere near reality. In addition, the edge-to-center
difference was much more than the simple Pseudo-Coupled Concept rules of thumb would tend to
predict. This clearly illustrates that the coefficient of subgrade reaction is, in general, highly problem-
dependent. It also emphasizes the fact that assumed simple, generic variations of this parameter, e.g.
doubling at the edge, do not provide reasonable results in all cases.

7.3.5.2.3 Structural

The various structural assumptions evaluated by the different levels of analysis revealed the
following. Note again that these should be taken as overall trends rather than as absolutes:

Both calculated settlements and moments were again sensitive to the assumption of rotational
restraint provided by the exterior walls. Neglecting the restraint is not necessarily conservative as
it can underestimate moments in some portions of the mat.

Simulating the superstructure presence as a loaded frame as opposed to using constant loads and
a constant value of flexural stiffness had little effect in this case. Perhaps this was because the mat
was already quite flexible.

There was a slight but noticeable effect of using a one-half reduction in the stiffness of the
foundation and superstructure as an approximation of time-related effects on the Young's
modulus of PCC. The reduced modulus resulted in more-flexible behavior as expected and, in
this case, slightly poorer agreement with observed settlements.

With regard to cracked-section behavior of the mat, there were some interesting outcomes. On
one hand, the RSC model calculated moments that exceeded the assumed theoretical cracked-
section value over virtually the entire mat. On the other hand, the reduced flexural stiffness due to
cracking had relatively little effect on the overall behavior. Again, it is suggested that because the
mat was inherently quite flexible additional factors that increased the flexibility somewhat had
only modest overall effect.

7.4 CONCLUSIONS

7.4.1 Subgrade Modeling

Based on the interpreted results of the study reported originally in Horvath (1993c, 1993d)
together with other research that is referenced in this report, the primary conclusions drawn with
respect to subgrade models are:

The Conventional Method of Static Equilibrium, which implies a perfectly rigid mat, poorly
approximates observed behavior. Virtually all mats exhibit some flexibility relative to the
subgrade. Even if a mat is relatively rigid, there is an increases in mat-subgrade contact stress
(subgrade reaction) at the edges of the mat, a phenomenon that is not duplicated by the
trapezoidal contact stress assumption of the CMSE. Thus for any mat-subgrade stiffness, at least

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
91
some of the moments calculated using the CMSE will generally always be in error on the
unconservative side.

Use of Winkler's Hypothesis with a constant Winkler coefficient of subgrade reaction does not
produce accurate estimates of moments and settlements from a single value of
o
W
k . For example,
the original design values for
o
W
k , which were chosen solely on the basis of engineering
experience and judgment, provided surprisingly good agreement with the RSC model for
moments but underestimated settlements by an order of magnitude. Conversely, the WTSC
values for
o
W
k produced relatively better estimates of settlement, but poorer moment agreement
compared to the RSC results.

In general, results from using a variable Winkler coefficient of subgrade reaction (the Pseudo-
Coupled Concept) are inconsistent. The simplest approach of doubling the value of
o
W
k along the
edges of the mat (either abruptly or gradually; the calculated results for the two buildings
investigated for this case history were essentially the same) produced only modest improvement
compared to using Winkler's Hypothesis with a constant value for
o
W
k , with overall poor
comparison to both observed behavior and results from the RSC model. On the other hand,
Ulrich (1991) demonstrated that by uniquely determining the magnitude and variation in
Winkler's coefficient of subgrade reaction for a given project, good results, at least in terms of
matching calculated and observed settlements, can be obtained. The conclusion is that variations
in the Winkler coefficient of subgrade reaction that are based on simple rules of thumb, e.g.
doubling the values at the mat edges, cannot be expected to be accurate for the infinite range in
problem variables that occur in practice. Simply stated, the accuracy of results from the Pseudo-
Coupled Concept in general is directly related to how well the Winkler coefficient of subgrade
reaction assumed matches the actual.

The results obtained using the Beam-Column-Analogy subgrade model, which incorporates the
Pasternak/Loof Hypothesis, are slightly to significantly better than Winkler's Hypothesis. The
degree of improvement appears to be variable and problem-dependent, increasing with
decreasing relative mat-subgrade stiffness. Thus, the Beam-Column Analogy shows promise as
an improved subgrade model compared to Winkler's Hypothesis but solely on an interim basis
until consistently more-accurate subgrade models such as the Reissner Simplified Continuum are
fully implemented in practice.

Of the subgrade models considered, the Reissner Simplified Continuum consistently provided the
best agreement between observed and calculated settlements. This is consistent with conclusions
based on previously published theoretical work (Horvath 1979, 1983a).

7.4.2 Structural Assumptions

Consideration of structural effects, particularly mat-superstructure interaction, is also important in
mat analysis. This conclusion is consistent with the findings of many others, e.g. Burland et al.
(1977), Soil-structure interaction (1989) and Ulrich (1991). For example, it is believed that even
better agreement between observed settlements and calculated results could have been achieved for
the two buildings discussed if the superstructure were modeled more accurately than using the simple
single-story frame model in the SSIH program that was used. This is especially true for the Chemistry
Building where the hogging pattern of settlement would have transferred more load to the column line
near the center of the mat, thus flattening the calculated settlement profile to more closely match the
92
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
observed. Although other structural effects such as change in concrete modulus with time and cracked
section behavior did not appear to be major variables for the two mats considered, a different
conclusion might result for other structures. In an event, these are known phenomenon that are
relatively simple to consider in routine design.

7.5 SUGGESTIONS FOR PRACTICE

The conclusions reached in the study reported originally in Horvath (1993c, 1993d) and
summarized in this report support earlier recommendations (Horvath 1979, 1983a, 1989a) that
implementation of improved subgrade models in mat design practice in particular, and SSI analysis in
general, is both highly desirable from consideration of computational accuracy and feasible from
practical considerations. The overall recommendation is that a single, reasonably accurate subgrade
model be used to calculate both of the parameters, total settlements and bending moments, that are
required for the design of a mat foundation.
Within this context, the following specific suggestions are made regarding subgrade models:

Use of the Conventional Method of Static Equilibrium ("rigid method") should be discontinued.

Use of the traditional form of Winkler's Hypothesis with a constant Winkler coefficient of
subgrade reaction,
o
W
k , should be discontinued.

Use of the general form of Winkler's Hypothesis with a variable coefficient of subgrade reaction,
k
W
(x), (the Pseudo-Coupled Concept) can produce acceptable results if and only if the reference
analysis used to produce the values of k
W
(x) matches the problem of interest in terms of geometry,
loading and mat stiffness. Thus the simple methods of doubling
o
W
k at the edges or using a
generic variation based on an elastic solution should not be used. Unfortunately, the choices of
reasonably accurate pseudo-coupled methodologies are limited, especially for mats and similar
foundation elements. The Discrete Area Method is conceptually sound and apparently produces
good results consistently (Ulrich 1991) but it requires close coordination between structural and
geotechnical engineers. Experience to date indicates this is too cumbersome for routine practice,
especially on smaller projects, and will likely continue to limit its use to major projects
undertaken by those who are familiar and comfortable with the methodology. This is supported
by the fact that this technique has been around for at least 30 years yet the number of engineers
using it in practice appears to be very small. Liao's method for cut-and-cover box tunnel base
slabs does appear to offer promise as a reasonable compromise between ease of use and accuracy
of results for that specific application. This offers promise that similar solutions could be
developed for other types of applications, e.g. water/wastewater treatment basins or tanks on
grade, that are very similar structurally from one project to the next.

As an interim general-purpose method that inherently incorporates spring coupling, the Beam-
Column Analogy should be used as it incorporates the lowest level multiple-parameter subgrade
model possible. As a result, it is fundamentally more accurate than Winkler's Hypothesis. The
degree of improvement offered by the Beam-Column Analogy appears to increase with
decreasing stiffness of the foundation element. However, a boundary condition involving w', the
first derivative of the settlement, w(x), at each edge of the foundation element must be dealt with
when using the Beam-Column Analogy. This issue is discussed in Horvath (1993e). Based on
study of this model to date, it is recommended that continuity of w' be assumed. This can be
achieved by specifying a zero-column-tension boundary condition at each edge of the foundation
element. It is also recommended that zero horizontal deformation boundary conditions be

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
93
imposed at each edge of the foundation element. This is to prevent calculation of fictitious
horizontal deformations of very large magnitude.

The preceding suggestion should be considered only interim until such time that higher-order
multiple-parameter subgrade models that are consistently more-accurate, such as the Reissner
Simplified Continuum, are implemented in structural analysis software available to practicing
engineers. The basic concept on how to achieve this was presented in Section 4.4.3.5.3.

Other details regarding the structural analysis aspects of mats and related foundation elements that
should also be given careful consideration are:

Attention should be given to considering in design well-known behavioral aspects of PCC in the
foundation element such as modulus reduction with time and cracked section behavior. These
issues are now considered routinely for PCC in a superstructure concrete so there is no reason
why the mat should not receive similar attention.

Although not evaluated as part of the study reported originally in Horvath (1993c, 1993d),
others have evaluated the use of "thick" elements for the foundation element in which shear
effects are considered versus the usual "thin" elements in which only simple-beam effects are
modeled (Horvilleur and Patel 1995). As discussed in Section 3.4.2.3, the inclusion of shear in
the flexural behavior of a beam effectively makes the beam more flexible. This tends to increase
differential settlement and reduce bending moments. The conclusion reached by Horvilleur and
Patel (1995) was that shear effects may be important in some cases. Therefore, it would appear to
be prudent to routinely model a mat using "thick" elements if the computer software package
used has this capability.

Finally, superstructure interaction effects are, in general, important and should be included even
for relatively modest structures such as the MIT buildings discussed here. As discussed in detail
in Section 2, the superstructure, mat and subgrade are a single, interactive unit that should be
analyzed together to the greatest extent practicable. Given the computational capabilities
available to engineers, there is no reason why this cannot be a reality on every project.
94
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
This page left blank.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
95
Section 8
References



ACI Committee 336 (1988). "Suggested analysis and design procedures for combined footings
and mats", ACI Structural Journal, American Concrete Institute, Detroit, Mich. U.S.A., Vol. 85,
No. 3, pp. 304-324.

ACI Committee 336 (1989). Closure to "Suggested analysis and design procedures for combined
footings and mats", ACI Structural Journal, American Concrete Institute, Detroit, Mich. U.S.A.,
Vol. 86, No. 1, pp. 113-116.

Bowles, J. E. (1988). "Foundation analysis and design", McGraw-Hill Book Company, New
York, N.Y., U.S.A., 4
th
edition, 1004 pp.

Burland, J. B., B. B. Broms and V. F. B. de Mello (1977). "Behaviour of foundations and
structures", Proceedings of the Ninth International Conference on Soil Mechanics and
Foundation Engineering, Japanese Society of Soil Mechanics and Foundation Engineering,
Tokyo, Japan, Vol. 2, pp. 495-546.

Chambers, R. E. (1984). "Structural plastics design manual - Volume I", ASCE Manuals and
Reports on Engineering Practice No. 63, American Society of Civil Engineers, New York, N.Y.,
U.S.A., 692 pp.

Christian, J. T. and W. D. C. Carrier III (1989). Discussion of "Elastic foundation settlements on sand
deposits" by J. E. Bowles, Journal of Geotechnical Engineering, American Society of Civil
Engineers, New York, N.Y., U.S.A., Vol. 115, No. 3, pp. 425-426.

Dawkins, W. P. (1982). "User's guide: computer program for analysis of beam-column structures
with nonlinear supports (CBEAMC)", Instruction Report K-82-6, U.S. Army Engineer
Waterways Experiment Station, Vicksburg, Miss., U.S.A., 90 pp.

DeSimone, S. V. and J. P. Gould (1972). "Performance of two mat foundations on Boston blue
clay", Proceedings of the Specialty Conference on Performance of Earth and Earth-Supported
Structures, American Society of Civil Engineers, New York, N.Y., U.S.A., Vol. I/Part 2, pp.
953-980.

Douglas, R. A. (1995). Discussion of "Modelling of geosynthetic-reinforced engineered granular
fill on soft soil" by S. K. Shukla and S. Chandra, Geosynthetics International, Industrial Fabrics
Association International, St. Paul, Minn., U.S.A., Vol. 2, No. 4, pp. 771-775.

Fraser, R. A. and L. S. Wardle (1976). "Numerical analysis of rectangular rafts on layered
foundations", Gotechnique, The Institution of Civil Engineers, London, U.K., Vol. 26, No. 4, pp.
613-630.

Ghosh, C. and M. R. Madhav (1994a). "Settlement response of a reinforced shallow earth bed",
Geotextiles and Geomembranes, Elsevier Applied Science, Oxford, U.K., Vol. 13, No. 10, pp.
643-656.
96
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
Ghosh, C. and M. R. Madhav (1994b). "Reinforced granular fill-soft soil system: confinement
effect", Geotextiles and Geomembranes, Elsevier Applied Science, Oxford, U.K., Vol. 13, No.
11, pp. 727-741.

Ghosh, C. and M. R. Madhav (1994c). "Reinforced granular fill-soft soil system: membrane
effect", Geotextiles and Geomembranes, Elsevier Applied Science, Oxford, U.K., Vol. 13, No.
11, pp. 743-759.

Haliburton, T. A. (1971). "Soil structure interaction: numerical analysis of beams and beam-
columns", Technical Publication No. 14, Oklahoma State University, School of Civil
Engineering, Stillwater, Okla., U.S.A., 179 pp.

Handy, R. L. (1985). "The arch in soil arching", Journal of Geotechnical Engineering, American
Society of Civil Engineers, New York, N.Y., U.S.A., Vol. 111, No. 3, pp. 302-318.

Harrop-Williams, K. (1989b). "Arch in soil arching", Journal of Geotechnical Engineering,
American Society of Civil Engineers, New York, N.Y., U.S.A., Vol. 115, No. 3, pp. 415-419.

Harrop-Williams, K. (1989c). "Geostatic wall pressures", Journal of Geotechnical Engineering,
American Society of Civil Engineers, New York, N.Y., U.S.A., Vol. 115, No. 9, pp. 1321-1325.

Hemsley, J. A. (ed.) (2000). "Design applications of raft foundations", Thomas Telford Ltd.,
London, U.K., 626 pp.

Hetnyi, M. (1946). "Beams on elastic foundation", The University of Michigan Press, Ann
Arbor, Mich., U.S.A., 255 pp.

Horvath, J. S. (1979). "A study of analytical methods for determining the response of mat
foundations to static loads", thesis presented to the Polytechnic Institute of New York in
Brooklyn, N.Y., U.S.A. as partial fulfillment of the requirements for the degree of Doctor of
Philosophy, 241 pp.

Horvath, J. S. (1983a). "New subgrade model applied to mat foundations", Journal of
Geotechnical Engineering, American Society of Civil Engineers, New York, N.Y., U.S.A., Vol.
109, No. 12, pp. 1567-1587; with errata in Vol. 110, No. 8 (August 1984), p. 1171.

Horvath, J. S. (1983b). "Modulus of subgrade reaction: new perspective", Journal of
Geotechnical Engineering, American Society of Civil Engineers, New York, N.Y., U.S.A., Vol.
109, No. 12, pp. 1591-1596; with errata in Vol. 110, No. 8 (August 1984), p. 1171.

Horvath, J. S. (1984). "Simplified elastic continuum applied to the laterally loaded pile problem -
part I: theory", Laterally Loaded Deep Foundations: Analysis and Performance, Special
Technical Publication No. 835, American Society for Testing and Materials, Philadelphia, Pa.,
U.S.A., pp. 112-121.

Horvath, J. S. (1988a). "Numerical analysis of beams and beam-columns with linear and non-
linear spring supports using finite differences", Research Report No. CE/GE-88-1, Manhattan
College, Civil Engineering Department, Bronx, N.Y., U.S.A.



Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
97
Horvath, J. S. (1988b). "Further evaluation of the coefficient, or modulus, of subgrade reaction,
k, using an extension of Reissner's simplified elastic continuum concept", Research Report No.
CE/GE-88-2, Manhattan College, Civil Engineering Department, Bronx, N.Y., U.S.A.

Horvath, J. S. (1988c). "Determination of the coefficient of subgrade reaction, k, for beams",
Research Report No. CE/GE-88-3, Manhattan College, Civil Engineering Department, Bronx,
N.Y., U.S.A.

Horvath, J. S. (1988d). "Historical review and critique of mathematical models for plate- and
beam-type foundation element subgrades", Research Report No. CE/GE-88-4, Manhattan
College, Civil Engineering Department, Bronx, N.Y., U.S.A.

Horvath, J. S. (1989a). Discussion of "Suggested analysis and design procedures for combined
footings and mats" by ACI Committee 336, ACI Structural Journal, American Concrete Institute,
Detroit, Mich. U.S.A., Vol. 86, No. 1, pp. 112-113.

Horvath, J. S. (1989b). "Base pressure and earth pressure measurements; Nrnberg subway",
Research Report No. CE/GE-89-1, Manhattan College, Civil Engineering Department, Bronx,
N.Y., U.S.A.

Horvath, J. S. (1989c). "Subgrade models for soil-structure interaction analysis", Foundation
Engineering: Current Principles and Practices, American Society of Civil Engineers, New York,
N.Y., U.S.A., pp. 599-612.

Horvath, J. S. (1989d). Discussion of "Pile horizontal soil modulus values" by T. D. Smith,
Journal of Geotechnical Engineering, American Society of Civil Engineers, New York, N.Y.,
U.S.A., Vol. 115, No. 7, pp. 1037-1038.

Horvath, J. S. (1990). "A theoretical study of the effect of anchors on the performance of shallow
foundations under vertical loads", Research Report No. CE/GE-90-3, Manhattan College, Civil
Engineering Department, Bronx, N.Y., U.S.A.

Horvath, J. S. (1993a). "Analysis of anchored foundations", preprint paper No. 93-0078, 72
nd

Annual Meeting, Transportation Research Board, Washington, D.C., U.S.A.

Horvath, J. S. (1993b). "Cut-and-cover tunnel subgrade modeling", preprint paper No. 93-0087,
72
nd
Annual Meeting, Transportation Research Board, Washington, D.C., U.S.A.

Horvath, J. S. (1993c). "Subgrade modeling for soil-structure interaction analysis of horizontal
foundation elements", paper presented at a special joint meeting of the Structural and
Geotechnical groups of the American Society of Civil Engineers Metropolitan Section, New
York, N.Y., U.S.A.

Horvath, J. S. (1993d). "Subgrade modeling for soil-structure interaction analysis of horizontal
foundation elements", Research Report No. CE/GE-93-1, Manhattan College, Civil Engineering
Department, Bronx, N.Y., U.S.A.

Horvath, J. S. (1993e). "Beam-column-analogy model for soil-structure interaction analysis",
Journal of Geotechnical Engineering, American Society of Civil Engineers, New York, N.Y.,
U.S.A., Vol. 119, No. 2, pp. 358-364; with errata in Vol. 119, No. 7 (July 1993), p. 1183.
98
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
Horvath, J. S. (1993f). "Cut-and-cover tunnel subgrade modeling", Transportation Research
Record No. 1415, Transportation Research Board, Washington, D.C., U.S.A.

Horvath, J. S. (1994). Discussion of "The effect of prestressing on the settlement characteristics
of geosynthetic-reinforced soil" by S. K. Shukla and S. Chandra, Geotextiles and Geomembranes,
Elsevier Applied Science, Oxford, U.K., Vol. 13, No. 11, pp. 761.

Horvath, J. S. (1998a). "The compressible inclusion function of EPS geofoam: an overview of
concepts, applications, and products", Research Report No. CE/GE-98-1, Manhattan College,
Civil Engineering Department, Bronx, N.Y., U.S.A.

Horvath, J. S. (1998b). "The compressible-inclusion function of EPS geofoam: analysis and
design methodologies", Research Report No. CE/GE-98-2, Manhattan College, Civil Engineering
Department, Bronx, N.Y., U.S.A.

Horvath, J. S. (2000a). "Coupled site characterization and foundation analysis research project -
rational selection of for drained-strength bearing capacity analysis", Research Report No.
CE/GE-00-1, Manhattan College, Civil Engineering Department, Bronx, N.Y., U.S.A.

Horvath, J. S. (2000b). "Integral-abutment bridges: problems and innovative solutions using EPS
geofoam and other geosynthetics", Research Report No. CE/GE-00-2, Manhattan College, Civil
Engineering Department, Bronx, N.Y., U.S.A.

Horvath, J. S. (2000c). "Coupled site characterization and foundation analysis research project -
further research into the rational selection of for bearing capacity analysis under drained-
strength conditions", Research Report No. CE/GE-00-3, Manhattan College, Civil Engineering
Department, Bronx, N.Y., U.S.A.

Horvath, J. S. (2001a). "Geomaterials research project - geofoam and geocomb geosynthetics: a
bibliography through the second millennium A.D.", Research Report No. CGT-2001-1,
Manhattan College, School of Engineering, Center for Geotechnology, Bronx, N.Y., U.S.A.

Horvath, J. S. (2001b). "Soil-structure interaction research project - analysis of vertically
anchored foundation elements", Research Report No. CGT-2001-3, Manhattan College, School
of Engineering, Center for Geotechnology, Bronx, N.Y., U.S.A.

Horvath, J. S. (2002). "Integrated site characterization and foundation analysis research project -
static analysis of axial capacity of driven piles in coarse-grain soil", Research Report No. CGT-
2002-1, Manhattan College, School of Engineering, Center for Geotechnology, Bronx, N.Y.,
U.S.A.

Horvilleur, J. F. and V. Patel (1995). "Mat foundation design - a soil-structure interaction
problem", Design and Performance of Mat Foundations; State-of-the-Art Review, E. J. Ulrich
(ed.), American Concrete Institute, Detroit, Mich., U.S.A., pp. 51-94.

Kerr, A. D. and W. J. Rhines (1967). "A further study of elastic foundation models", Report No.
S-67-1, New York University, School of Engineering and Science, Bronx, N.Y., U.S.A., 23 pp.




Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
99
Kerr, W. C. and S. K. Saxena (1977). "Role of cracking in deformations of shallow foundations",
ASCE 1977 Fall Convention preprint paper No. 3000, American Society of Civil Engineers, New
York, N.Y., U.S.A.

Lambe, T. W. and R. V. Whitman (1969). "Soil mechanics", John Wiley & Sons, Inc., New York,
N.Y., U.S.A., 553 pp.

Landva, A. O., A. J. Valsangkar, J. C. Alkins and P. D. Charalambous (1988). "Performance of a
raft foundation supporting a multistorey structure", Canadian Geotechnical Journal, National
Research Council of Canada, Ottawa, Ont., Canada, Vol. 25, No. 1, pp. 138-149.

Liao, S. S. C. (1991). "Estimating the coefficient of subgrade reaction for tunnel design", final
draft report of an internal research project sponsored by Parsons Brinckerhoff, Inc., New York,
N.Y., U.S.A.

Liao, S. S. C. (1995). "Estimating the coefficient of subgrade reaction for plane strain
conditions", Geotechnical Engineering, published for The Institution of Civil Engineers by
Thomas Telford Services Ltd., London, U.K., Vol. 113, No. 3, pp. 166-181.

Nair, K. and C.-Y. Chang (1973). "Flexible pavement design and management - materials
characterization", NCHRP Report No. 140, Highway Research Board, Washington, D.C., U.S.A.

Poulos, H. G. and E. H. Davis (1974). "Elastic solutions for soil and rock mechanics", John
Wiley & Sons, Inc., New York, N.Y., U.S.A., 411 pp.

Poulos, H. G. and E. H. Davis (1980). "Pile foundation analysis and design", John Wiley & Sons,
Inc., New York, N.Y., U.S.A., 397 pp.

Reissner, E. (1958). "A note on deflections of plates on a viscoelastic foundation", Journal of
Applied Mechanics, Vol. 25/Transactions of the American Society of Mechanical Engineers, Vol.
80, pp. 144-155.

Reissner, E. (1967). "Note on the formulation of the problem of the plate on an elastic
foundation", Acta Mechanica, Vol. 4, pp. 88-91.

Rhines, W. J. (1965). "Foundation models for continuously supported structures", thesis
presented to New York University, Bronx, N.Y., U.S.A. in partial fulfillment of the requirements
for the degree of Doctor of Philosophy, 236 pp.

Roark, R. J. and W. C. Young (1975). "Formulas for stress and strain", McGraw-Hill Book
Company, New York, N.Y., U.S.A., 5
th
edition, 624 pp.

Schiffman, R. L., R. V. Whitman and J. C. Jordan (1970). "Settlement problem oriented computer
language", Journal of the Soil Mechanics and Foundation Engineering Division, American
Society of Civil Engineers, New York, N.Y., U.S.A., Vol. 96, No. SM2, pp. 649-669.

Scott, R. F. (1981). "Foundation analysis", Prentice-Hall, Inc., Englewood Cliffs, N.J., U.S.A.,
545 pp.


100
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
Shukla, S. K. and S. Chandra (1994a). "The effect of prestressing on the settlement characteristics
of geosynthetic-reinforced soil", Geotextiles and Geomembranes, Elsevier Applied Science,
Oxford, U.K., Vol. 13, No. 8, pp. 531-543.

Shukla, S. K. and S. Chandra (1994b). "A study of settlement response of a geosynthetic-
reinforced compressible fill-soft soil system", Geotextiles and Geomembranes, Elsevier Applied
Science, Oxford, U.K., Vol. 13, No. 9, pp. 627-639.

Shukla, S. K. and S. Chandra (1994c). "A generalized mechanical model for geosynthetic-
reinforced foundation soil", Geotextiles and Geomembranes, Elsevier Applied Science, Oxford,
U.K., Vol. 13, No. 12, pp. 813-825.

Shukla, S. K. and S. Chandra (1995). "Modelling of geosynthetic-reinforced engineered granular
fill on soft soil", Geosynthetics International, Industrial Fabrics Association International, St.
Paul, Minn., U.S.A., Vol. 2, No. 3, pp. 603-618.

Soil-structure interaction - the real behaviour of structures (1989). The Institution of Structural
Engineers, London, U.K., 120 pp.

Spangler, M. G. and R. L. Handy (1982). "Soil engineering", Harper & Row, New York, N.Y.,
U.S.A., 4
th
edition, 819 pp.

Stark, T. D. and Vettel, J. J. (1991). "Effective stress hyperbolic stress-strain parameters for clay",
Geotechnical Testing Journal, American Society for Testing and Materials, Philadelphia, Pa., U.S.A.,
Vol. 14, No. 2, pp. 146-156.

Terzaghi, K. (1943). "Theoretical soil mechanics", John Wiley and Sons, Inc., New York, N.Y.,
U.S.A., 510 pp.

Terzaghi, K. (1955). "Evaluation of coefficients of subgrade reaction", Gotechnique, The
Institution of Civil Engineers, London, U.K., Vol. 5, No. 4, pp. 297-326.

Timoshenko, S. P. and J. M. Gere (1961). "Theory of elastic stability", McGraw-Hill Book
Company, New York, N.Y., U.S.A., 2
nd
edition, 541 pp.

Timoshenko, S. P. and J. M. Gere (1972). "Mechanics of materials", D. Van Nostrand Company,
New York, N.Y., U.S.A., 552 pp.

Timoshenko, S. P. and S. Woinowsky-Krieger (1959). "Theory of plates and shells", McGraw-
Hill Book Company, New York, N.Y., U.S.A., 2
nd
edition, 580 pp.

Tomlinson, M. J. (1986). "Foundation design and construction", John Wiley & Sons, Inc., New
York, N.Y., U.S.A., 5
th
ed., 842 pp.

Ulrich, E. J. (1991). "Subgrade reaction in mat foundation design", Concrete International,
American Concrete Institute, Detroit, Mich., U.S.A., April, pp. 41-50.

Vesic, A. B. and W. H. Johnson (1963). "Model studies of beams resting on a silt subgrade",
Journal of the Soil Mechanics and Foundations Division, American Society of Civil Engineers,
New York, N.Y., U.S.A., Vol. 89, No. SM1, pp. 1-31.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
101
Vesic, A. S. and S. K. Saxena (1970). "Analysis of structural behavior of AASHO road test rigid
pavements", National Cooperative Highway Research Program Report 97, Highway Research
Board, Washington, D.C., U.S.A.

Vlasov, V. Z. and N. N. Leont'ev (1960). "Beams, plates and shells on elastic foundations (Balki,
plity i obolochki na upragom osnovanii)", Gosudartstvennoe Izdatel'stvo, Fiziko-
Matematicheskoi Literatury, Moscow, U.S.S.R.; translated from Russian and published in 1966
by the Israel Program for Scientific Translations, Jerusalem, Israel, in cooperation with The
National Aeronautics and Space Administration and The National Science Foundation,
Washington, D.C., U.S.A., 357 pp.

Wolf, J. P. (1994). "Simple physical models for foundation vibration - a guided tour",
Proceedings of the 13
th
International Conference on Soil Mechanics and Foundation
Engineering, Volume 2, A. A. Balkema, Rotterdam, The Netherlands, pp. 663-668.

Yin, J. H. (1997a). "Modelling geosynthetic-reinforced granular fills over soft soils",
Geosynthetics International, Industrial Fabrics Association International, St. Paul, Minn., U.S.A.,
Vol. 4, No. 2, pp. 165-185.

Yin, J. H. (1997b). "A nonlinear model of geosynthetic-reinforced granular fill over soft soil",
Geosynthetics International, Industrial Fabrics Association International, St. Paul, Minn., U.S.A.,
Vol. 4, No. 5, pp. 523-537.
102
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
This page left blank.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
103
Section 9
Notation



a
v
= coefficient of compressibility of soil
A = cross-sectional area of a beam
A
v
= shear area of a beam

b = half the width of the base slab of a box tunnel constructed using the cut-and-cover method

i
p
c = constant coefficient (i = 1, 2, )
i
w
c = constant coefficient (i = 1, 2, )
C
r
= recompression index of soil

{d} = flexural displacement vector for a beam or beam-column
D = flexural stiffness of an elastic plate; also the drained constrained modulus of soil
D(x,y) = D (plate stiffness) as an explicit function of x and y

E = Young's modulus
EI = flexural stiffness of a beam
EI(x) = EI as an explicit function of x

g = shear stiffness of a shear layer in mechanical multiple-parameter models
G = elastic shear modulus

H = thickness (depth) of elastic continuum

I = moment of inertia of a beam
I(x) = I as an explicit function of x

k = spring stiffness; also coefficient of subgrade reaction
k(x,y) = k (coefficient of subgrade reaction) as an explicit function of x and y
k
eq
= equivalent spring stiffness for two or more springs in series
k
W
= Winkler's coefficient of subgrade reaction
k
W
(x) = k
W
as an explicit function of x
k
W
(x,y) = k
W
as an explicit function of x and y
i
W
k = k
W
at point i
o
W
k = constant value of k
W
(x,y) or
i
W
k

l = beam length
L = beam length

p = vertical normal stress (subgrade reaction stress or subgrade reaction) at the contact between
the bottom of a horizontal foundation element and underlying subgrade; also the vertical
normal stress applied to a subgrade; also the horizontal stress between a laterally loaded deep
foundation and the subgrade (limited to colloquial use only as in "p-y" method)
{p} = subgrade reaction vector for a beam or beam-column under uniaxial flexural loading
104
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
p(x) = p as an explicit function of x
p(x,y) = p as an explicit function of x and y
p
i
= p(x,y) at point i
p' = first derivative of p with respect to x or both x and y as appropriate
P = axially applied beam force (defined as positive in compression)
P
cr
= column buckling force

q = reactions at the bottom of a superstructure supported on a horizontal foundation element; also
loads applied at the top of a horizontal foundation element
{q} = flexural load (force) vector for a beam or beam-column
q(x) = q as an explicit function of x
q(x,y) = q as an explicit function of x and y

[S] = flexural stiffness matrix for a beam or beam-column
[S'] = [S] modified to account for concentrated springs at each node

T = membrane tension in the Filonenko-Borodich mechanical model

u = x-axis displacement

v = y-axis displacement

w = z-axis displacement
w(x) = w as an explicit function of x for a beam under uniaxial bending
w(x,y) = total settlement of a horizontal foundation element at foundation level; also total
settlement at the subgrade surface
w
i
= w(x,y) at point i
w' = first derivative of w with respect to x or both x and y as appropriate

x = horizontal axis; also the longitudinal axis of a beam

y = horizontal axis; also the horizontal displacement of a laterally/moment loaded deep
foundation element (limited to colloquial use only as in "p-y" method)

z = vertical (depth) axis defined as positive downward

v
= parameter for shear effects on beam flexure

= displacement (general)

= viscoelastic coefficient

= Poisson's ratio

= node rotation.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
105
Appendix A
Background Information re SSIH



A.1 PROGRAM CAPABILITIES

A.1.1 Introduction

SSIH is intended to be a general program for research-related soil-structure interaction analysis
of horizontal foundation elements resting on a subgrade. At the present time, it is limited to two-
dimensional problems only and allows for the following three model components:

a superstructure consisting of a one-story rigid frame (horizontal beams supported on vertical
columns) that is supported on a

foundation element (beam or plate under plane-strain conditions) resting on a

subgrade.

Some of the types of problems that can be analyzed using the current versions of this program
are:

beams of any width,

plates of unit width under plane-strain conditions to simulate one-dimensional flexure of a
mat or slab-on-grade, and

railway track systems.

The differential equations defining the behavior of both the structural elements and subgrade
are solved using the finite element method. Models containing up to 51 nodes/50 elements total can
be accommodated currently.
Because SSIH is an academic research program, its capabilities are undergoing continual
refinement and upgrading. Consequently, it is important to check the "most recent revision" date
that is given in the FORTRAN source code and printed at the heading of the output file against
that given in the SSIH user manual. The most recent operational version of this program is dated
April 16, 1999.

A.1.2 Summary of Current Capabilities

A.1.2.1 General

Three explicit unknown primary variables are assumed at each element node:

vertical displacement, w;

rotation, ; and

106
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
subgrade reaction, p, at the foundation element-subgrade interface.

A.1.2.2 Structural

The types of structural elements used in the current version of SSIH are:

a beam or plane-strain plate element suspended above the foundation on one or more columns to
simulate a superstructure with the following characteristics:

o linear-elastic behavior, and

o uniform flexural stiffness (may be applied incrementally to simulate multi-story
construction);

vertical column elements to connect the foundation element and supported superstructure with the
following characteristics:

o each column can have a different EI/L for flexural behavior and AE/L for axial behavior (all
behavior linear elastic),

o no horizontal loading allowed although concentrated moments can be applied at the nodes at
each end of the column,

o no horizontal displacements allowed although node rotation can occur, and

o top of column has full moment fixity to the superstructure element but the connection at the
bottom of the column to foundation element can be either fully fixed or hinged;

a horizontal "true" beam-column to simulate the foundation element with the following
characteristics:

o linear- or non-linear elastic flexural behavior under transverse loading. Non-linear behavior
accounted for by allowing for cracked-section behavior according to Branson's equation for
reinforced portland-cement concrete;

o rigid under axial loading;

o width (if beam) and moment of inertia constant within each element for each load-step
iteration;

o shear effects on flexural stiffness (which effectively reduces stiffness) can be included in
addition to the usual flexural effects;

o concentrated transverse and rotational springs with linear-elastic behavior can be placed at
any node to simulate support from, or continuity with, other structural elements that are not
modeled explicitly. These transverse springs should not be confused with the soil subgrade or
vertical ground anchors that are discussed subsequently;

o vertical ground anchors, either prestressed or passive, with linear-elastic behavior can be
placed at any foundation node. Note that anchor prestressing, if any, is always performed

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
107
before any superstructure placement/loading and/or direct foundation loading takes place;
and

o up to one plane of symmetry can be invoked (if used it must be placed at right side of
foundation model).

The following boundary conditions can be imposed on the structural components:

any node on the foundation element can be fully fixed against vertical displacement, rotation or
both;

any node on the superstructure can be fixed against rotation; and

in neither case can non-zero nodal displacements and/or rotations be specified.

With regard to the types of loading that can be accommodated:

applied to the superstructure:

o concentrated moment at any node, and

o uniformly distributed load across entire superstructure. This load may be divided into dead
and live load components, with each component applied in a different number of increments;

applied directly to the foundation element (in addition to loads transmitted from the
superstructure):

o constant axial (horizontal) load. This load is always applied in one increment and as part of
the ground-anchor prestressing or first load increment (if no anchors are present);

o concentrated vertical load, moment or both at any node. These may be applied in an arbitrary
number of increments of equal magnitude; and

o uniformly distributed load on any element. These may be applied in an arbitrary number of
increments of equal magnitude.

A.1.2.3 Subgrade

Three types of subgrade models can be used:

Winkler's Hypothesis. The Winkler coefficient of subgrade reaction can be different for each
finite-element of the foundation element.

Pasternak's Hypothesis (Beam-Column Analogy). The edge boundary condition involving the
first derivative of transverse displacement (w') can be continuous or discontinuous (see Section
A.2.3 for a discussion concerning selection).

Reissner Simplified Continuum (RSC). A single layer of finite thickness (the same for each
element) with a Young's modulus that:

108
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
o is constant with depth (Model 1),

o increases linearly with depth (Model 2), or

o increases with the square root of depth (Model 3).

A.1.2.4 Structure-Subgrade Coupling

To produce an efficient solution, the nodes for both the foundation element and subgrade-
surface are the same, which is the same as saying that they are perfectly coupled. The only
implication of significance in terms of behavior is that foundation lift-off from the subgrade
cannot be modeled explicitly and, if liftoff should occur, a tensile contact stress will be
calculated. This situation is not expected to arise in most problems. In fact, foundation lift-off is
something that probably would be avoided in most design cases. The potential for foundation lift-
off will be handled in a future version of SSIH by checking for tensile contact stresses after a load
increment is applied, and re-doing the load step with an artificially low value of subgrade
stiffness for the element(s) where it occurred. This process would be iterated until a satisfactory
solution was achieved. Note that in an earlier attempt to deal with uplift, the primary variable for
pressure was set to zero for any node exhibiting uplift. However, this was found to produce
incorrect results for the RSC subgrade because of boundary-condition considerations, so this
approach was ultimately abandoned in favor of the present situation.

A.2 COMMENTS AND SUGGESTS RE PROGRAM OPERATION

A.2.1 Structural Modeling of Foundation Element

At first glance, a beam and plane-strain plate appear to be identical. However, there are some
subtle differences between the two that need to be addressed. These differences arise out of how
to couple the flexural behavior of the foundation and displacement of the subgrade.
There are four unique combinations of foundation and subgrade that can be handled by SSIH:
combinations of a beam or plane-strain plate on a Winkler
mmm
or RSC subgrade. Consider first a
beam on a Winkler subgrade.
There is a conceptual and dimensional inconsistency between:

the Winkler coefficient of subgrade reaction, k
W
, which has dimensions of force/length cubed (=
force/length (i.e. a spring constant)/unit width of beam/unit length of beam);

beam stiffness, EI, which is based on the full width of the beam; and

applied loads, which are either distributed per unit length of the beam or concentrated.

There are two solution alternatives:

The more commonly used one is to simply multiply k
W
by the beam width to get a modulus of
subgrade reaction (as Vesic called it) that has dimensions of force/length squared (= force/length
(i.e. a spring constant) per unit length of beam). Essentially, this integrates the subgrade support
per unit width of the beam over the beam width.

mmm
Issues related to this discussion for the Beam-Column Analogy (Pasternak) subgrade are essentially the
same as a Winkler subgrade.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
109
Alternatively, one could divide the beam stiffness and loads by the beam width, so that a unit
width of beam is effectively analyzed. Essentially, this is the same as a plane-strain plate.

This latter approach was used initially in SSIH because it is theoretically more correct when one
considers a beam on an RSC subgrade. This is because in the derivation of the RSC (and the
Pasternak-Type and Winkler-Type simplified continua for that matter), it was assumed that a stress
was applied to the subgrade, not a line load as is assumed to be the subgrade reaction in a beam
problem. However, this approach runs into problems if a beam of variable width, e.g. a strap footing
or combined footing of variable width, is analyzed. Therefore, SSIH was modified so that the
appropriate stiffness factors that deal with a the subgrade are multiplied by the beam width. This
approach was verified against the previous one. Note that this issue does not arise with the plane-
strain plate foundation.

A.2.2 Subgrade Modeling

When it comes to Young's modulus, most subgrades in practice are best modeled as a layered
system where each artificial layer has a constant Young's modulus. Because SSIH cannot
accommodate layered systems directly, it is necessary to do some pre-analysis to develop the
approximate equivalent isotropic, homogeneous single-layer model (Model 1) that is the basic one
incorporated into SSIH. The exceptions to this are the two explicit solutions (Models 2 and 3) for a
variable Young's modulus that are available if the RSC subgrade option is chosen.
A weighted-average methodology based on linear-elastic theory was described in Horvath (1988b)
that is similar to that used by Fraser and Wardle (1976). The general validity of calculating equivalent
values of elastic parameters for a layered system is discussed in Burland et al. (1977) and Tomlinson
(1986). No correction (i.e. increase) in Young's modulus to account for embedment of the foundation
element below the surrounding grade is suggested. This is consistent with a recommendation by
Christian and Carrier (1989) that the embedment effect is small and can be ignored.

A.2.3 Boundary Conditions

A.2.3.1 Structural

Regardless of the subgrade model used, there are four boundary conditions (two at each end of the
foundation element) that involve the foundation element itself:

either the displacement or shear must be zero, and

either the rotation or moment must be zero.

For most foundation elements, the edge shears and moments will be zero.

A.2.3.2 Subgrade

Whether or not there are boundary conditions involving the subgrade depends on the model used.
None exist for a Winkler subgrade.
For a Beam-Column-Analogy (Pasternak) subgrade, only w', the first derivative of the surface
displacement at each end of the foundation element, needs to be considered. The available choices are
that w' is:

110
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
discontinuous. This allows a kink to occur in the subgrade surface but produces a concentrated
force acting upward at each end of the foundation element; or

continuous. This implies that vertical shears within the subgrade are continuous at the edges.

Based on limited study to date as summarized in Horvath (1993e), the latter assumption seems to
produce calculated results that are much closer to those produced by the more-advanced RSC model
as well observed in the case history presented in Section 7 of this report. However, SSIH allows a
choice between the two cases should the former boundary condition be preferred for some reason.
Finally, for the RSC model two boundary conditions must be considered at each end of the
foundation element:

w' and

p', the first derivative of the foundation-subgrade contact stress, p.

The approach taken in SSIH with regard to these boundary conditions is to assume physical
continuity of the vertical shear stress both within and at the edge of the foundation element. This
assumption involves both p' and w' and was the approach suggested by Reissner (1958, 1967). This
assumption is also consistent with that suggested above for the Beam-Column Analogy. Note that no
assumption is made with regard to the pattern of subgrade displacements.
This approach produces results that are intuitively pleasing, namely, a gradual buildup of subgrade
reaction stress occurs toward the edges of the foundation element which replicates the edge-stress
concentrations predicted both by elastic theory as well as observed in advanced numerical methods
for a continuum such as the finite-element method. In addition, a punching type of settlement can
occur at the edges of the foundation element.
On the other hand, Rhines (1965) assumed continuity of subgrade deflections at each edge of the
foundation element. With his approach, a concentrated edge reaction is always produced which is a
cruder approximation of the edge-stress concentrations that occur beneath actual foundation elements.
Rhines' boundary condition assumptions are not available in SSIH.
Note that the above results for both the Beam-Column Analogy and RSC models are for a
boundary condition at the physical edge of the foundation element. A symmetry boundary is handled
by simply noting that both p' and w' are zero on the plane of symmetry.
Note also that the reason that three explicit unknowns ("primary variables") listed in Section
A.1.2.1 are assumed at each foundation/subgrade-surface node is that this facilitates implementation
of the boundary condition on p'. If p is eliminated as an explicit unknown by expressing it in terms of
the displacement vector (which includes both the real displacements and rotations), it becomes
algebraically complex to deal with the p' boundary condition.

A.3 PROGRAM LOGIC

The main program is divided into four sections:

initial screen prompts for I/O file names,

pre-processor,

processor, and

post-processor.

Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
111
Unless noted otherwise, all subroutines are called from the main program.
The pre-processor phase begins with subroutine INPUT that reads all required input data from the
input file specified at the appropriate screen prompt and echoes it to the output file that was specified
at the appropriate screen prompt. Subroutine TITLE is called from INPUT to write the problem
heading to the output file. Any applied loads are placed in the appropriate locations within the global
force (load) vector. Subroutine MODULUS is called from INPUT to calculate the various soil moduli
used for some of the components of the stiffness matrix. The only exception to this is if the specified
subgrade is a Winkler type with a constant value of k
W
,
o
W
k , that is input directly into the stiffness
matrix during the processor phase (MODULUS is bypassed in such a case). Subroutine FORCE is
then called to determine the size of the problem to be solved and assigning the solution equation
numbers by using the nodal boundary conditions on displacement and/or rotation. These are handled
by using the ID-matrix concept in which a nodal displacement or rotation specified to be zero is
eliminated from solution to minimize the size of the solved problem and the global force vector is
thereby condensed vertically to eliminate those terms that correspond to an applied boundary
condition on displacement or slope.
The processor phase begins by calling subroutine STIFFE to calculate and assemble the basic 6x6
stiffness matrix for each foundation element finite-element in turn. Subroutine STIFFG then places
the element stiffness values in the appropriate location in the global stiffness matrix, skipping the
appropriate rows and columns for which a nodal boundary condition has been specified. If external
structural springs are included, subroutine STIFFSPR is called to place the springs in the global
stiffness matrix. If a structure is modeled explicitly, subroutine STIFFSS is called to calculate the
stiffness components of the superstructure element and place them in the global stiffness matrix. This
is followed by solution for the unknown primary variables using subroutine SOLVE. The solver used
in this subroutine is completely general in that it works on the full global stiffness matrix even though
it may be banded (but non-symmetrical if the RSC subgrade model is used) and very sparse. Note that
the banded property is true only if only a foundation is modeled, as nodes for a superstructure totally
destroy the banded property (the global stiffness matrix will still be very sparse).
The post-processor phase consists of calling subroutine RESULTS. Here, the following parameters
are calculated at each end of each finite-element:

bending moment and shear in the foundation element, and

first derivative of the contact stress, i.e. p' (for a RSC subgrade only).

Two iterations are made for each increment of load application. A predictor-corrector approach is
used. The first iteration uses the foundation stiffness at the beginning of the load increment. A new
stiffness is calculated, and the second iteration is based on the average of the original and new
stiffnesses.
The above processor/post-processor cycle is repeated for each increment of load application
specified. The program executes relatively quickly, as all storage requirements of the current version
are satisfied using RAM. The only hard-drive read/write operations are for the normal I/O.
112
Soil-Structure Interaction Research Project
Basic SSI Concepts and Applications Overview
Manhattan College School of Engineering Center for Geotechnology Report No. CGT-2002-2
This page left blank.

You might also like