University of Sheffield

Department of Civil and Structural Engineering

Finite Element Analysis of Embankments

on Soft Ground
Incorporating Reinforcement and Drains

by

Darien Russell
MSc, BSc

Thesis Submitted to the University of Sheffield

for the Degree of Doctor of Philosophy

June, 1992

Summary
The objectives of the research were threefold. Firstly, to improve .the
numerical modelling capability for reinforced embankments constructed
over soft compressible soils containing vertical drains. Secondly, to
demonstrate the ability to model accurately such embankments. Finally, to
develop simplified procedures to be used in the design of embankments
over soft soils.
The modifications to the finite element program, CRISP, included the
incorporation of three additional elements: modelling the reinforcement, the
soil/reinforcement interface and the vertical drains. The facility to vary
permeability with stress level was also implemented. A technique for
modelling the consolidation of soil containing vertical drains in plane strain
finite element analyses was developed and validated.
The modified program was validated in three ways. Firstly, each element
was used to analyse simple problems so that the correct formulation was
ensured. Secondly, a series of analyses was carried out of problems for
which analytical solutions were available; these problems involved collapse
of undrained subsoils and consolidation around a single vertical drain.
Thirdly, an analysis of a case history of an embankment constructed over a
normally consolidated clay, improved with vertical drains, was performed.
Based on the results of the previous finite element analyses, and an
additional analysis of an idealized two-stage constructed embankment,
simple design procedures have been proposed. Firstly, a method for the
design of single stage embankments and, secondly, a method for the
calculation of subsoil strength increases in multi-stage construction, which
can be used in conjuction with limit equilibrium analyses.
It is concluded that the finite element method is a useful technique for the
analysis of reinforced embankments over soft soils containing vertical
drains.
II

Acknowledgement
The research presented in this Thesis was carried out under the
supervision of Dr. C.C. Hird and Dr. l.C. Pyrah, as part of a SERC funded
project.
The Author would very much like to thank his supervisors for their
invaluable help and encouragement with all aspects of the research project.
Their numerous comments during the preparation of this Thesis are
gratefully acknowledged.
Dr. R.A. Jewell was involved with the SERC contract and participated in
several discussions during the project, his comments were both timely and
useful.
The Author would like to express his gratitude to all members of the
geotechnics group at the University of Sheffield who always found time to
discuss the work. The staff at the University's Computer Services
department provided advice and facilities whenever necessary.

III

Table of Contents
1. Introduction

1.1. Embankments Over Soft Soils

1.2. Methods of Analysis

1.2.1. Stability analysis

1.2.2. deformation analysis

I
2
4
5

1.2.3. summary of traditional analysis methods

1.3. Numerical Analysis

1.3.1. Constitutive modelling

1.3.2. material parameters

1.4. Aims of the Research

6
7
8
9
9

2. Literature Review: Embankment Finite Element

Analysis
2.1. Undrained Analyses

II
11

2.1.1. Unreinforced Embankments

2.1.2. Reinforced Embankments

2.2. Multi-Stage Construction

12
14
17

2.2.1. Consolidation analyses without Drains

2.2.2. subsoils containing vertical Drains

2.3. Discussion

18
21

2.3.1. Summary of embankment Finite Element

Analyses

2.3.2. Validation and Accuracy of Modelling

3. The Finite Element Program CRISP and

Modifications

3.1.1. The Historyof CRISP

3.1.2. Types of Analysis

Iv

23
25
27
29
29

3.1.3. Additional Features of CRISP

23

2.4. The Need for Further Finite Element Research

3.1. Crisp

29
30
31


3.2. Reinforcement Element

3.2.1. Constitutive Relationship

3.2.2. Reinforcement Element Stiffness Matrix
Formulation

3.2.3. Equivalent Nodal Forces

3.2.4. Analyses Using the Reinforcement Element

3.3. Interlace Element

3.3.1. Interface Element Stiffness Matrix Formulation

3.3.2. Transformation to Gobat Coordinate System

3.3.3. Interface Element Constitutive Relationship and
Modes of Behaviour

3.3.4. Equivalent Nodal Forces

3.3.5. Analyses Using the Interface Element

3.4. Drainage Element

3.4.1. Element Formulation

3.4.2. Transformation to Two-Dimensions

3.4.3. Incorporation of the Drainage Element in a
Two-Dimensional Mesh

3.4.4. Analyses Using the Drainage Element

3.5. Variation of Permeability with Stress Level

3.6. Summary

4. Collapse of Undrained Subsoils

4.1. Introduction

4.1.1. Plasticity Theory

4.1.2. reinforced embankments

4.1.3. Application of Plasticity Theory to Reinforced
Embankment Problems

4.2. Finite Element Analysis of Plasticity Problems

4.2.1. Meshes for Equivalent Loading Problems

4.2.2. Material Properties

4.2.3. Results

4.2.4. Effect of the interface element

4.2.5. predicted displacements for a rough footing on a
uniform strength/limited depth subsoil

4.2.6. further ana'ysis using modified cam-clay

33
34
35
38
39
39
39
41
42
44
45
46
46
47
48
49
49
51
52
52
52
54
55
58
58
60
61
64
69
73


4.3. Embankments with Constant Side Slope

4.3.1. Finite Element Analyses of Constant Side Slope
Loading

4.3.2. Results

4.3.3. A comparison of limit equilibrium and finite
element analysis for embankments with constant
side slope

4.4. Summary

5. Consolidation Around a Single Vertical Drain

5.1. Introduction
5.2. Comparison of Finite Element Analysis with a
Closed Form Solution
5.3. Matching Procedure for Plane Strain Analyses
5.3.1. Validation of the matching procedure
5.3.2. pore pressure variation across unit cells
5.3.3. Effect of vertical permeability on matching
procedure
5.4. Summary

6. Case History: Porto Tolle

6.1. Introduction
6.2. Porto Tolle
6.2.1. subsurface conditions
6.2.2. Construction of the trial embankment
6.2.3. Vertical drains
6.2.4. Instrumentation
6.3. Unit Cell Analyses
6.3.1. Finite element mesh and material parameters
6.3.2. in situ stresses
6.3.3. undrained analyses to assess shear strength
6.3.4. Axisymmetric unit cell consolidation analyses
6.3.5. comparison of finite element and observed
behaviour
6.3.6. plane strain unit cell consolidation analyses
6.4. Full Plane Strain Analysis

vi

77
80
81
83

86
90
90
93
104
107
114
117
120
121
121
122
122
122
124
126
126
126
129
129
129
131
135
138

6.4.1. finite element mesh and boundary conditions

6.4.2. material parameters

6.4.3. comparison of observed and predicted
behaviour

6.5. Summary

7. Multi-Stage Embankment Construction

138
139
140
145

147

7.1. Introduction

147

7.2. Idealized Two-Stage Construction

148

7.2.1. Finite element Analysis of Idealized Two-Stage

Construction

149

7.2.2. Finite Element Results

154

7.3. Subsoil Strength Increase

172

7.3.1. Simple Calculation Methods

172

7.3.2. Factors Influencing the Strength Increase of

Soft Clays

179

7.3.3. comparison of the strength increases predicted

by finite element and simple methods

183

7.3.4. Strength Increases Calculated from Simple

Methods

184

7.4. Porto Tolle

189

7.5. Summary

198

8. Conclusions and Further Work

200

8.1. Objectives and Completed Work

200

8.2. Conclusions

201

8.2.1. undrained collapse of cohesive soils

201

8.2.2. consolidation of soil around single vertical

drains

202

8.2.3. Porto tolle case history analysis

203

8.2.4. idealized two-stage construction

204

8.3. Suggestions for Further Work

205

8.3.1. further modifications to CRISP

205

8.3.2. matching procedure

206

8.3.3. simplified design procedures

206

VII

207

Appendix A - Strength of Modified Cam-clay

218

Appendix B- Consolidation of a Plane Strain Unit

Cell

Appendix C - Development of Pore Pressures Due
to Ramp Loading

224

226

References

VIII

List of Symbols

List of Symbols
a

Vector of displacement at any point in a finite element

a0

b

Vector of nodal displacement for a finite element

Bishop stress parameter

b*

B

B

B

Modified Bishop stress parameter

Footing half width
Width of a plane strain unit cell
Strain matrix

Cohesion intercept

C,
CI'

Coefficient of secondary compression

cc
ch

Coefficient of consolidation

CII

Compression index

Permeability change index

Depth of a constant depth soil

D

e

Finite element constitutive matrix

Void ratio

e0

E

E

Initial void ratio

Youngs modulus
First derivative of drainage element flow shape function matrix

F0

Nodal force matrix

FS

G

Factor of safety
Elastic shear modulus
Initial shear modulus

Hd

Embankment height
Embankment design height
Hydraulic gradient

Reinforcement stiffness per unit strain

k,1

Permeability
Horizontal permeability
x

List of Symbols
Interface element normal stiffness

k 1 Interface element normal stiffness after separation

Ic

Interface element shear stiffness

Ic

Soil permeability in smeared zone

lc

Vertical permeability

Drain permeability

k0 Initial permeability

'e

Finite element stiffness matrix in local coordinate system

K, Finite element stiffness matrix in global coordinate system

Bulk modulus of water

K0 Coefficient of earth pressure at rest

KONC Coefficient of earth pressure at rest for normally consolidated soil
KS Drainage element force stiffness matrix
length of drain
L

Length of a one-dimensional finite element

Well resistanbe dimensionless parameter

Link matrix

LL

Liquid limit

Coefficient of volume change

RIr

Displacement shape function matrix

Drainage element pore pressure shape function matrix

Nr

Relative displacement shape function matrix

OCR Over-consolidation ratio in terms of vertical effective stress

p

Mean stress invariant

P1

Plasticity index

Deviatoric stress invariant

Load applied during ramp loading

'i

Axisymmetric vertical drain discharge capacity

Plane strain vertical drain discharge capacity

r3 Radius of smeared zone

xi

List of Symbols
r

Radius of vertical drain

Radius of axisym metric unit cell

Over-consolidation ratio in terms of mean effective stress

rJr

Undrained shear strength

s 0 Undrained shear strength at surface

t

Thickness of finite element

Time

Transformation matrix

Tb Time factor for horizontal drainage

1h90

Time factor for horizontal drainage causing 90% consolidation

Pore pressure

Uh Average degree of consolidation based on pore pressure at a given

depth
Overall average degree of consolidation

U
V

Specific volume

Rate of pore water flow

Volume

wr

Relative displacement matrix for interface element

wn

Natural water content

we

Relative nodal displacement matrix for interface element

(x,y) Global coordinate system

Xd

xl

z

Width of embankment side slope

Cartesian local coordinate system for one-dimensional finite element
Depth below surface

Orientation of one-dimensional element in global coordinate system

a

13

o

Characteristic line associated with plasticity theory

L.

Rate of loading for a ramp loading embankment

Characteristic line associated with plasticity theory
Interface element cohesion
Large increment
xl'

List of Symbols
Strain
qlp'

.11
K

Gradient of swelling line in critical state frame work

Angle of shearing resistance

Drainage element flow stiffness matrix

Bulk unit weight

Unit weight of water
Specific volume of soil at critical state with p=1 kN/m2

Gradient of compression line in critical state frame work

A

(?.-ic)I?

Gradient of critical state line

Coefficient for unit cell consolidation

Poissons ratio

Rate of undrained shear strength increase with depth

Stress

a1,

Yield stress for reinforcement element

Normal stress

Vertical stress

Out of plane stress in plane strain analysis

a1 Maximum principal stress

a3 Minimum principal stress
Shear stress
(, ri) Local coordinate system

Superscripts
Effective stress
*
T

Virtual displacement
Transformed matrix

XIII

List of Symbols

Subscripts
ax

Axisymmetric conditions

p1

Plane strain conditions

xiv

List of Figures

List of Figures
Chapter 3- The Finite Element Program CRISP and Modifications
3.1

(a) Reinforcement element; (b) bilinear constitutive model.

3.2 Calculation of one-dimensional element length and orientation.

3.3

(a) Interface element; (b) constitutive relationship.

3.4 Interface element shear stress sign convention and stress regimes.
3.5 Drainage element.
Chapter 4- Collapse of Undrained Subsoils
4.1

Definition of stress characteristics for undrained loading.

4.2 Reinforced embankment configuration.

4.3

Idealized soil profiles for plasticity analyses: (a) uniform

strength/limited depth; (b) strength increasing linearly with depth.

4.4

Ideal load distributions on an idealized profile of: (a) uniform

strength/limited depth; (b) strength increasing linearly with depth.

4.5 Finite element mesh used for uniform strength/limited depth

analyses (mesh shown for smooth case).
4.6 Finite element mesh used for strength increasing with depth
analyses (mesh shown for smooth case).
4.7

Finite element approximation for strength increasing linearly with

depth.

4.8 Load displacement curves for displacement controlled analyses: (a)

smooth footing on a uniform strength/limited depth subsoil; (b)
rough footing on a uniform strength/limited depth subsoil: (c)
smooth footing on a strength increasing with depth subsoil; (d)
rough footing on a strength increasing with depth subsoil.
4.9

Comparison of vertical stress distributions at failure predicted by

plasticity and finite element analysis for a uniform strength/limited
depth idealized subsoil.

4.10 Comparison of vertical stress distributions at failure predicted by

plasticity and finite element analysis for a strength increasing linearly
with depth idealized subsoil.

xv

List of Figures
4.11 Load displacement curves for load controlled analyses: (a) subsoil of
uniform strength/limited depth; (b) subsoil with strength increasing
linearly with depth.
4.12 Surface settlement profiles for load controlled analyses: (a) subsoil
of uniform strength/limited depth ; (b) subsoil with strength increasing
linearly with depth.
4.13 Vertical stress distribution at failure predicted using three different
finite element meshes compared with plasticity theory.
4.14 Plasticity analysis of a rough rigid footing on a subsoil of uniform
strength/limited depth: (a) slip line mesh; (b) hodograph.
4.15 Comparison of the lateral displacements predicted by finite element
and plasticity analyses: (a) 2D; (b) 40; (c) 6D; (d) 80 from toe.
4.16 Stress and strength profiles used for the modified Cam-clay analysis.
4.17 Load settlement curves for a displacement controlled analysis using
modified Cam-clay to model a subsoil with strength increasing
linearly with depth.
4.18 Comparison of finite element and plasticity predicted vertical stress
distributions at failure beneath a rough footing on a subsoil with
strength increasing linearly with depth modelled using modified
Cam-clay.
4.19 Error in the undrained shear strength profile used for the modified
Cam-clay analysis (after Potts and Ganendra, 1991).
4.20 Proposed practical embankment shape to approximate ideal loading
predicted by plasticity theory (after Jewell, 1988) and a proposed
simplified constant side slope design profile.
4.21 Typical load displacement curves for a constant side slope analysis.
Curves shown for a uniform strength/limited depth analysis with
xID=5.
4.22 Comparison of the finite element predicted collapse load of a
constant side slope embankment with plasticity theory for a subsoil
with uniform strength/limited depth.
4.23 Comparison of the finite element predicted collapse load of a
constant side slope embankment with plasticity theory for a subsoil
with strength increasing linearly with depth.
4.24 Displacement vectors at failure for embankments with constant side
slopes on a subsoil of uniform strength/limited depth: (a) x/D=1; (b)
XdID3; (c) Xd/D5.

xvi

List of FIGures

4.25 Displacement vectors at failure for embankments with constant side

slopes on a subsoil with strength increasing linearly with depth: (a)
PXdIS= l .6; (b) pxJs0=4.8; (c) pxs8.0.
4.26 Maximum shear strain contours at failure for a constant side slope
embankment on a subsoil of uniform strength and limited depth
(x/D=5).

4.27 Maximum shear strain contours at failure for a constant side slope
embankment on a subsoil with strength increasing linearly with depth
(pxIs4.8).
Chapter 5- Consolidation Around a Single Vertical Drain
5.1

Unit cells and direction of flow: (a) axisym metric conditions; (b) plane
strain conditions.

5.2

Unit cell adopted for analytical solution (after: Hansbo, 1981).

5.3 Finite element mesh used for unit cell analysis comparison with
Hansbo (1981).
5.4 Comparison of finite element and analytical results for consolidation

of a unit cell: (a) L=0.0; (b) L=0.5; (c) L=3.0: (d) L5.0.
5.5 Finite element mesh used for unit cell analyses: comparison with
Jamiolkowski et al (1983).
5.6

Effect of well resistance on rate of consolidation.

5.7

Effect of drain length on rate of consolidation.

5.8 Effect of smear on rate of consolidation.

5.9 Development of surface settlement profiles in finite element
analyses: (a) without smear or well resistance; (b) without smear but
with well resistance; (c) with smear but without well resistance.
5.10 Variation of vertical strain with depth at unit cell periphery in finite
element analysis with well resistance (qjk=150m2).
5.11 Comparison of average surface settlement for axisymmetric and
plane strain analyses: (a) without smear or well resistance; (b) with
smear but without well resistance; (c) without smear but with well
resistance; (d) with smear and well resistance.
5.12 Comparison 3f average excess pore pressure at mid-depth for
axisymmetric and plane strain analyses: (a) without smear or well
resistance; (b) with smear but without well resistance; (c) without
smear but with well resistance; (d) with smear and well resistance.

xvii

List of Figures
5.13 Errors in geometry matched plane strain analyses: (a) based on
settlements; (b) based on pore pressure.
5.14 Development of surface settlement profile in geometry matched
plane strain analysis: (a) without smear or well resistance; (b)
without smear but with well resistance.
5.15 Comparison of pore pressure distribution from axisymmetric and
geometry matched plane strain analyses without smear or well
resistance (degree of consolidation=65%).
5.16 Excess pore pressure distribution after the first consolidation
increment from the axisymmetric finite element analysis without
smear or well resistance.
5.17 Comparative results for axisymmetric and matched plane strain
analyses with vertical flow (k.,=k,j, degree of consolidation based on
excess pore pressure: (a) without smear or well resistance; (b)
without smear but with well resistance.
5.18 Errors in geometry matched plane strain analyses for unit cells with
vertical permeability (k,=k,), based on excess pore pressure.
Chapter 6 - Case History: Porto Tolle
6.1

Soil profile (after Jamiolkowski and Lancellotta, 1984).

6.2 Stress history and maximum overburden from one-dimensional

consolidation tests according to Casagrande's procedure (after
Jamiolkowski and Lancellotta, 1984).
6.3 Porto Tolle Trial embankment: (a) Plan view, (b) cross section; (c)
construction schedule.
6.4 Mesh used for axisymmetric finite element analyses.
6.5

Undrained shear strength in: (a) plane strain; (b) axisymmetry.

6.6 Axisymmetric unit cell consolidation based on average surface

settlement
6.7 Excess pore pressure on centreline of the embankment: (a) 19.7m
below ground level; (b) 12.6m below ground level.
6.8 Rate of surface settlement at centreline of the embankment.
6.9 Unit cell rate of consolidation based on the average excess pore
pressure at mid-depth.
6.10 Unit cell rate of consolidation based on the average surface
settlement.
6.11 Unit cell rate of consolidation matching errors.
xviii

List of Figures
6.12 Mesh used for plane strain analysis.
6.13 Surface settlement at centreline.
6.14 Maximum horizontal displacement.
6.15 Lateral movement profiles at the inclinometer position: (a) at the end
of the construction stage; (b) at the end of the consolidation stage.
6.16 Excess pore pressure on centreline of the embankment (full plane
strain): (a) 19.7m below ground level; (b) 12.6m below ground level.
Chapter 7- Multi-Stage Embankment Construction
7.1 Idealized two-stage embankment geometry.
7.2

Idealized soil: (a) stress; (b) over-consolidation ratio; (c) initial

undrained shear strength.

7.3 Lower interface shear strength after the consolidation stage.

7.4 Finite element mesh and boundary conditions.
7.5 Settlement at of points at the subsoil surface against embankment
height
7.6 Surface settlement profiles.
7.7 Lateral movement of a section through the toe.
7.8 Maximum surface settlement against maximum lateral movement
beneath the toe.
7.9 Ratio of maximum settlement to maximum lateral movement plotted
against embankment height.
7.10 Displacement vectors at failure.
7.11 Stress paths: (a) beneath the slope; (b) beneath the crest.
7.12 Principal stress rotations: (a) beneath the slope; (b) beneath the
crest.
7.13 Modified Bishop stress parameter: (a) beneath the slope; (b)
beneath the crest.
7.14 Variation of s)ci': (a) beneath the slope; (b) beneath the crest.
7.15 Variation of s)c1: (a) beneath the slope; (b) beneath the crest.
7.16 Reinforcement strain.
7.17 Shear stress on: (a) reinforcement/fill interface; (b)
reinforcement/clay interface.
7.18 Stress state on a potential failure surface in a soil near failure
subjected to embankment loading.
xix

List of Figures
7.19 Undrained shear strength anisotropy of normally consolidated clays.
7.20 Estimation of vertical stress increase in a soil due to embankment
loading: (a) one-dimentional; (b) Gray (1936).
7.21 Predicted strength increase near centreline.
7.22 Predicted absolute strength near centreline.
7.23 Contours at the end of the consolidation stage predicted by the finite
element analysis: (a) strength increase; (b) absolute strength.
7.24 Contours of strength increase at the end of the consolidation stage:
(a) predicted by Method G; (b) predicted by Method H.
7.25 Contours of absolute strength at the end of the consolidation stage:
(a) predicted by Method G; (b) predicted by Method H.
7.26 Contours of strength difference at the end of the consolidation stage:
(a) between Method G and finite element; (b) between Method H
and finite element
7.27 Porto Tolle predicted: (a) degree of consolidation; (b) total vertical
stress.
7.28 Porto Tolle predicted: (a) undrained shear strength increase; (b)
absolute undrained shear strength..
7.29 Contours of strength increase at the end of the consolidation stage
at Porto Tolle predicted by: (a) the finite element analysis; (b) the
simplified method.
7.30 Contours of the strength difference predicted by finite element
analysis and the simplified method at the end of the consolidation
stage at Porto Tolle.
Appendix A - Undrained Shear Strength of Modified Cam-clay
Al

Definition of Mohr-Coulomb yield surface for a cohesive material.

A2

Undrained stress path for a normally consolidated modified

Cam-clay sample.

A3

Undrained stress path for a normally consolidated and

over-consolidated modified Cam-clay sample.

Appendix B - Consolidation of a Plane Strain Unit Cell

Bl

Plane strain unit cell.

xx

List of Figures
Appendix C - Development of Pore Pressures Due to Ramp Loading
Cl Ramp loading scheme.

xxi

List of Tables

List of Tables
Chapter 6 - Case History: Porto Tolle
6.1 Summary of Porto Tolle soft clay material properties.
6.2 Summary of parameters used to model Porto Tolle clay.
Chapter 7- Multi-Stage Embankment Construction
7.1 Summary of assumptions in the simple methods used to predict
strength increases.

xxi

Introduction

1. Introduction
1.1. Embankments Over Soft Soils
Recent social and economic development around the world has brought
about an increase in the construction of embankments used in highway and
railway systems, flood and irrigation projects and harbour and airport
installations. This coupled with increasing urbanisation has required the
geotechnical engineer to design and construct embankments over
increasingly weak and compressible soils.
Often the soil is sufficiently weak that the embankment cannot be
constructed in a single lift, in the area available; in such situations it is
necessary to improve the soil on which the embankment is to be placed.
This improvement can take several forms depending on the soil conditions
encountered.
Where settlement is the major consideration the rate at which settlement
occurs can be increased with the installation of vertical drains (Barron,
1948; Hansbo, 1981; Holtz et al, 1987; Holtz et al, 1991) and the
post-construction settlements therefore reduced. Preloading the
compressible soils with a surcharge is a technique used with both
embankments and settlement sensitive structures (Johnson, 1970a) and is
often used in conjunction with vertical drains (Johnson, 197Db).
Other techniques may have to be considered to ensure stability of the
embankment during construction. Stage construction can be used to
construct embankments with relatively steep side slopes (Jardine and
Hight, 1987; Ladd, 1991; Leroueil et al, 1991). This method relies on the
increase in the undrained shear strength of the subsoil during consolidation
and is therefore most beneficial when used with vertical drains. An

Introduction
increasingly popular technique is the use of tensile reinforcing material,
such as geogrids, geotextiles or steel reinforcement, placed at the base of
the embankment fill (Bonaparte and Christopher, 1987). The reinforcement
provides lateral restraint and effectively increases the bearing capacity of
the subsoil.
Other less common expedients include the use of light-weight fills, drainage
trenches, stone columns, lime columns, piles, replacement of the subsoil,
electra-osmosis, electro-injection, dynamic compaction or a combination of
the above (brief descriptions and further references for these techniques
can be obtained from: Pilot et al, 1987; Delmas et at, 1987, Lerouiel et al,
1991).
In this research three of the more popular embankment construction
techniques have been examined: 1) reinforcement at the base of the fill
material, 2) vertical drains in the subsoil and 3) multi-stage construction with
vertical drains.

1.2. Methods of Analysis

Traditionally the analysis of embankments has been divided into two parts:
firstly, stability, in which a factor of safety is calculated and, secondly,
deformation, in which displacements are calculated. However, before either
stability or deformation analysis are carried out it is necessary to decide on
the drainage conditions in the subsoil as these will influence the type of
analysis performed.
Undrained, Drained or Consolidating Behaviour
For an embankment constructed in a single lift, failure would be most likely
to occur during construction, before the excess pore pressures have had
time to dissipate and the subsoil to gain strength. If the loading period is
relatively short it is often realistically conservative to assume that the subsoil

Introduction

is undrained and a stability analysis based on total stresses could be

performed. However, care must be taken that this approach is not overly
conservative as measurement of pore pressures beneath embankments on
soft cohesive deposits (Tavenas and Leroueil, 1980) has shown that partial
drainage may still occur even during short construction periods.
For multi-stage construction, total stress analyses have the obvious
limitation that the increase in the subsoil strength due to consolidation is not
taken into account. An alternative analysis procedure proposed by Bishop
and Bjerrum (1960) is an effective stress analysis. The effective stress
analysis allows for strength increases in the subsoil but requires knowledge
of the excess pore pressures. The excess pore pressures are difficult to
calculate reliably and for major works Bishop and Bjerrum advocated that
insitu pore pressures should be measured.
Effective stress analyses also implicitly assume the complete dissipation of
shear induced excess pore pressures during failure. This implies that the
effective stress at failure in an effective stress analysis will be higher than in
a total stress analysis. This results in a higher calculated factor of safety
from an effective stress analysis than from a total stress analysis (Ladd,
1991).
A third analysis procedure for the design of multi-stage embankments (Hight
and Jardine, 1987; Leroueil et al, 1990; Ladd, 1991) has been proposed
and termed an undrained strength analysis (Ladd, 1991). In such an
analysis the failure mechanism is assumed to be undrained and the
analysis is based on total stresses, but the increase in subsoil undrained
shear strength due to consolidation is taken into account. The undrained
strength analysis of each loading stage is carried out in two parts; firstly, the
distribution of undrained shear strength in the subsoil, immediately before
loading commences, is calculated and, secondly, these strengths are used
in a conventional total stress analysis to assess the factor of safety. Ladd
3

Introduction

(1991) examined three case histories and demonstrated that the factor of
safety calculated using an effective stress analysis was approximately twice
as high as that calculated using an undrained strength analysis.
In this Thesis, where appropriate, the philosophy of an undrained strength
analysis has been used for multi-stage embankment construction (Chapter
7). Undrained analysis based on the initial undrained shear strength has
been used as a conservative method of analysis for quickly constructed
single stage em bankments (Chapter 4).

1.2.1. Stability Analysis

After consideration of the subsoil drainage conditions the analysis of stability
can be performed. Two methods that have been frequently used in design
are based on limit equilibrium and plasticity theory.
Limit Equilibrium
The most commonly used method of analysis in current design practice is
that of limit equilibrium. In this analysis a failure mechanism is postulated
and a factor of safety is defined in terms of the disturbing and resisting
forces. This method is often corn puterised with the failing mass divided into
slices; several assumptions have been postulated for the distribution of the
inter slice forces (e.g: Fellenius, 1927; Bishop, 1955; Morgenstern and
Price, 1965).
The limit equilibrium method has several significant disadvantages. The
method requires the postulation of a failure mechanism; if an incorrect
mechanism is postulated then the predicted factor of safety is in error. The
method also does not take account of the complex stress distribution which
will develop in the subsoil beneath an embankment nor for the possibility of
a progressive failure.

Introduction
Limit equilibrium analysis has been extended to the analysis of
embankments with reinforcement at the base of the fill (Fowler, 1982;
Jewell, 1982). The reinforcement provides an additional resisting force, the
magnitude of which is dependent on both the fill material and the surface
undrained shear strength of the subsoil.
Plasticity Theory
Often an accurate estimation of the collapse load of a geotechnical structure
can be made using plasticity theory. Jewell (1988) has suggested that the
plasticity solutions for two idealized soil profiles (Mandel and Saleon, 1972;
Davis and Booker, 1973) have direct relevance to the stability analysis of
embankments constructed on soft soils. These plasticity solutions predict
the collapse loads for rigid strip footings on idealized subsoils. The solutions
can also model the shear stress distribution on the underside of the footing.
This shear stress can be interpreted as the action of an infinitely stiff
reinforcement; the varying roughness factors define the bond between the
soil and the reinforcement (Houlsby and Jewell, 1988).
The plasticity solutions have been shown to match well limit equilibrium
analysis for the idealized conditions (Jewell, 1988) but further research is
required before the method can be directly applied to embankment design.

1.2.2. Deformation Analysis

Traditionally the deformation of the embankment and the subsoil has been
calculated without consideration of stability. The immediate or undrained
displacement and the long term consolidation settlement are analysed
separately.
Analytical solutions for the undrained displacements based on elastic
conditions are available (Poulos and Davis, 1974). These solutions have the
obvious disadvantage that the subsoil is often not elastic. This is especially
true with the soft, often normally consolidated, soils encountered.
5

Introduction

The consolidation settlements are calculated assuming one dimensional

consolidation theory and using parameters obtained from oedometer and/or
field tests. The stress in the subsoil can be estimated from elastic theory
(e.g: Newmark, 1942; Gray, 1936). The rate at which the settlement occurs
can be found by calculating the rate of consolidation using Terzaghi's one
dimensional consolidation theory (Terzaghi and Peck, 1967) or in the case
of subsoils containing vertical drains, using one of the available analytical
solutions (e.g. Barron, 1948; Hansbo, 1981).
Secondary compression can also be a significant factor. The estimation of
it's magnitude may be made based on laboratory test data. However,
several aspects of the secondary compression phenomenon are still the
subject of debate (Leroueil et al, 1990).
The prediction of the lateral displacement is also important, especially in
reinforced embankments where it's magnitude can significantly affect the
reinforcement forces developed. There are no analytical solutions available
for the calculation of lateral displacement. Tavenas and Leroueil (1980)
proposed empirical relationships. However, the application of these
relationships is uncertain for clays not included in their study.

1.2.3. Summary of Traditional Analysis Methods

The analysis of embankments has traditionally been separated into stability
and deformation calculations. However, displacement and stability are
interrelated and an accurate analysis requires the consideration of both
aspects. Analytical solutions are not available for such an analysis and in
order to obtain accurate predictions it is often necessary to use a numerical
analysis technique.

Introduction

1.3. Numerical Analysis

With the increased availability of computers the application of numerical
techniques for the solution of geotechnical problems has expanded. These
techniques rely on the solution of the partial differential equations which
describe the behaviour of the soil for the particular boundary conditions
considered. Although an exact solution cannot be achieved, an approximate
solution can be found by using an appropriate numerical technique. The
degree of approximation is a function of the complexity of the analysis and
the material model, the amount of computational effort used and the
experience of the analyst.
The numerical technique used in this research is the finite element method
(e.g. Zienkiewicz, 1977). The finite element analysis of geotechnical
problems relies on the discretization of a continuum into a number of
elements which are connected at nodal points. The displacement response
of each element, when subjected to loading, is defined by the element
shape, the variation of displacement within the element and the stress-strain
relationship (constitutive model) used to represent the material. The result
of the discretization process is a set of simultaneous equations which must
be solved to yield nodal displacements. The nodal displacements can then
be used to calculate the strains and stresses within each element. For
analysis of consolidation problems the excess pore pressure can be
coupled with displacements and additional nodal information is obtained.
The finite element method has the advantage over traditional analysis
techniques that the displacements and stresses within the soil are coupled
and, depending on the constitutive model, more realistic soil behaviour can
be represented. However, with this increased potential for accurate analysis
there is a corresponding increased complexity in the constitutive modelling
of the soil and therefore the input parameters required. The approximate
nature of the solution procedure results in the possibility of different
7

Introduction
solutions to the same problem if incorrect discretization or solution
techniques are used.

1.3.1. Constitutive Modelling

Increasingly complex constitutive models have been developed to represent
geotechnical materials. It is essential that the appropriate model be used for
each problem. The adopted solution technique should also be appropriate
to that model.
Linear Elastic Model
The majority of traditional deformation analyses assume a linear elastic
material. This is a gross approximation of soil behaviour if the major
displacements occur when the soil yields and becomes plastic. This is
particularly true of soft clays which are usually normally or lightly
over-consolidated. However, if the purpose of the finite element analysis is
to perform numerical comparisons with an analytical solution it may be
appropriate to use an elastic model. Also, if the information available
describing the soil is insufficient then a more complex analysis may not be
justifiable.
Elastic-Perfectly Plastic Model
Elastic-perfectly plastic models are used widely and various yield criteria
can be implemented to define when the material response changes from
elastic to plastic. Collapse occurs when a mechanism is produced. For
undrained behaviour, yield criteria which predict a constant undrained shear
strength produce theoretically correct collapse loads. However, for drained
or consolidation conditions both displacement and collapse may be
incorrectly predicted as the model does not allow for volumetric strain
hardening or softening.

Introduction
In drained conditions the Mohr-Colomb yield criterion is often used and for
mathematical convenience the flow rule is usually associated (Smith, 1982).
Such models produce excessive dilatancy and may predict an incorrect
response.
Critical State Models
Critical state theory (Schofield and Wroth, 1968; Roscoe and Burland, 1968;
Atkinson and Bransby, 1978) provides a realistic framework for the
prediction of the behaviour of normally and lightly over-consolidated clays.
These models can predict strain hardening for such soils and have been
used widely in numerical analyses to provide accurate predictions of
displacements and stresses. The increased complexity of these models
produces increasing difficulty in the analysis in which the equilibrium
condition is more difficult to enforce. This requires a larger number of
loading iterations or a more complex solution procedure.

1.3.2. Material Parameters

With more complex constitutive models more input parameters are required.
Often the site investigation data is not sufficiently accurate to justify the use
of these models for prediction. However, the more complex models may still
be useful in a parametric study.

1.4. Aims of the Research

The use of reinforcement, vertical drains and multi-stage construction is
increasing in the construction of embankments overlying soft soils. The
analysis of these problems is far from straightforward and traditional
techniques are not always sufficient!y accurate to be a reliable method for
design. Numerical analysis using the finite element method overcomes
several of the disadvantages of the traditional methods and produces
coupled predictions for displacements, stresses and pore pressures. The
9

Introduction

finite element method is therefore an excellent tool with which to study the
behaviour of embankments constructed over soft ground.
However, accurate numerical predictions require careful analysis and it is
often necessary to conduct a study in which material parameters, boundary
conditions and solution techniques are varied. Therefore, It may be
unrealistic to perform a finite element analysis at an initial stage of the
design process, instead simple design procedures may be used for
preliminary analysis.
The aims of the present research were:
1. To modify an existing finite element program, CRISP (Gunn and
Britto, 1984), to better model reinforced embankments over soft soils
containing vertical drains (Chapter 3).
2. To benchmark the modified program against analytical solutions for
both the undrained collapse of idealized soils and the consolidation
behaviour of soil surrounding a single vertical drain (Chapters 4 and
5).
3. To develop procedures so that soil systems containing vertical drains
can be analysed accurately using the finite element method (Chapter
5).
4. To compare finite element predictions with observed behaviour for an
embankment constructed over soft soil incorporating vertical drains
(Chapter 6).
5. To develop simple procedures which aid routine design of
embankments on soft soils containing vertical drains, reinforcement
and stage constructed embankments (Chapters 4 and 7).

10

Literature Review

2. Literature Review: Embankment

Finite Element Analysis
Published finite element analyses of embankments constructed over soft
ground can be divided into two categories. Firstly, undrained analyses of
embankments which are constructed sufficiently quickly that consolidation
of the subsoil is negligible. Secondly, analyses of embankments which are
constructed in more than one lift or slowly enough that consolidation effects
are significant. This literature review is divided into these two sections with
particular attention being given to analyses in which reinforcement or
vertical drains have been modelled. All analyses reviewed have been
carried out in plane strain.
The review is a selective account of the significant contributions to the area
of finite element analysis of embankments, mainly constructed over soft
ground. Each author's work is described and, where relevant to the present
research, the main findings stated. There then follows a discussion of
difficulties which have been encountered and areas in which the analysis
may be improved.

2.1. Undrained Analyses

Undrained analyses were first conducted using elastic constitutive models
considering unreinforced embankments. Subsequent studies considered
increasingly sophisticated constitutive models and, as the use of
reinforcement became more widespread, analysts included idealizations of
reinforcement and soil/reinforcement interface behaviour.

11

Literature Review

2.1.1. Unreinforced Embankments

Brown and King
Brown and King (1966) performed one of the first finite element studies of
an embankment problem. They analysed built up and cut down
embankments using an elastic model to represent the soil. The authors
speculated on possible failure mechanisms and compared the factors of
safety calculated from the finite element predicted stress distributions with
Taylor's (1948) stability parameters.
dough and Woodward
Clough and Woodward (1967) compared elastic embankments, over rigid
subsoils, constructed in a single increment with those constructed in several
increments. The authors concluded that it is essential to model the
construction sequence to produce accurate embankment displacement
predictions.
Further research included a parametric study of the stiffness of a subsoil of
finite depth. The authors concluded that the vertical stress was relatively
insensitive but horizontal and shear stresses were strongly dependent on
subsoil stiffness.
dough and Woodward also performed a non-linear e'astic analysis of the
Otter Brook Dam. The elastic properties of the soil were modified at the end
of each of 14 lifts using data obtained from a series of triaxial tests. The
finite element predicted and observed displacements, for two positions,
were in good agreement.
Kulhawy and Duncan
Kulhawy and Duncan (1972) analysed the Oroville Dam. The embankment
fill was represented using a hyperbolic model (Duncan and Chang, 1970)
and the subsoil was assumed to be rigid. The predicted displacements were
12

Literature Review

in good agreement with the observed values. Embankment stresses did not
compare favourably with those measured but the authors expressed
concern over the measured values. The finite element predicted tension
zones agreed closely with visible tension cracks.
Smith and Hobbs
Smith and Hobbs (1974) compared finite element analyses with centrifuge
model tests. The subsoil was represented with an undrained
elastic-perfectly plastic soil model. The effect of the elastic parameters and
the distance of the mesh boundaries were studied. The authors showed
that, for their analyses, the finite element method predicted collapse loads
agreed well with limit equilibrium calculations.
Wroth and Simpson
Wroth and Simpson (1972) analysed a test embankment constructed at
King's Lynn, UK. The subsoil was modelled using Cam-clay (Schofield and
Wroth, 1968) and the embankment loading by equivalent vertical loads
applied over a number of increments. Undrained analyses were carried out
representing the short term conditions. The predicted vertical and horizontal
displacements were in good agreement with those observed. However, the
predicted pore pressures were not in such good agreement. The authors
suggested that the disagreement may have been caused by consolidation
near the drained boundaries during construction.
The authors also carried out drained analyses to simulate long term
conditions. The predicted displacements were again in good agreement with
those observed.

13

Literature Review

2.1.2. Reinforced Embankments

Boutrop and Holtz
Boutrop and Holtz (1983) performed finite element analyses of reinforced
and unreinforced embankmerits constructed over a soft clay. The
embankment fill was modelled as an elasto-plastic material with a
Drucker-Prager failure criterion. The authors compared reinforced and
unreinforced analyses and observed that the reinforcement reduced the
shear stresses in the subsoil and the differential settlement of the top of the
embankment.
Rowe
Rowe and his co-researchers performed several finite element studies of
the behaviour of reinforced embankments. Their research included both
back analyses and analyses of idealized problems. The subsoil was
modelled as an elasto-plastic material with a Mohr-Coulomb failure criterion
and a non-associated flow rule. In order to better represent granular
materials the embankment was idealized using the same model but with the
stiffness varying with stress level. The geotextile was modelled as a
membrane with axial stiffness but no flexural rigidity. Slip could occur
between the reinforcement and soil. All studies reviewed consider single
stage construction in which the subsoil was modelled as undrained.
The first of these analyses, Rowe (1982), consisted of a parametric study of
a reinforced embankment constructed at Pinto Pass, USA. The analysis
used a large deformation finite element formulation. Rowe et al (1984) and
Rowe (1984) analysed a second reinforced embankment, at Bloomington
Road, Ontario, USA, using the same finite element program. These two
studies confirmed the beneficial effects of reinforcement on stability and
deformations.

14

Literature Review

Rowe and Soderman (1985) proposed an extended limit equilibrium

procedure in which reinforcement strain level was considered. The
procedure was developed using results from a finite element parametric
study of an idealized embankment. The method can be used to assess the
likely increase in the factor of safety when using reinforcement.
Rowe and Soderman (1987) reviewed their finite element procedure and
used plasticity solutions to predict failure for an idealized embankmeni The
authors found that the finite element results were within 7% of the predicted
plasticity failure heights for highly reinforced embankments.
Rowe and Mylleville (1988) analysed steel reinforced embankments and
showed that the failure mechanism was dependent on the amount of
reinforcement used. Yielding of the reinforcement occurred well before a
subsoil failure mechanism developed.
Rowe and Mylleville (1989) presented results from several finite element
analyses in which they showed that the subsoil shear strains exceeded the
reinforcement strain significantly at working loads. An -increase in the
reinforcement modulus was shown to increase the failure height and also
the subsoil shear strain. The authors pointed out that these increased shear
strains would have important consequences for strain softening soils.
Mylleville and Rowe (1991) presented further evidence of the subsoil shear
strains exceeding the reinforcement strain and discussed the implications
for practical design.
Rowe and Mylleville (1990) showed that the subsoil shear strains
corresponding to only modest reinforcement strains were large enough to
cause concern, particularly when considering strain softening soils.
Analyses were also carried out with a higher strength surface crust. The
authors showed that the crust, a) increased the failure height, b) reduced
reinforcement strains and c) reduced the maximum shear strains in the
subsoil. The authors compared finite element analyses with one and two
15

Literature Review

layers of reinforcement and concluded that an embankment constructed

using two layers of reinforcement of a given modulus behaved in essentially
the same way as one constructed with a single layer of twice the modulus.
Basset and Guest
Basset and Guest (1990) compared results from model tests and finite
element analyses, using a version of the finite element program CRISP,
with plasticity predicted failure loads and displacement fields for a clay
subsoil of limited depth. The finite element analysis seriously over-predicted
failure loads and under-predicted displacements. This discrepancy was a
result of discretization errors at the rough boundaries where, as shown by
the plasticity solution, discontinuities develop. The use of interface elements
to model such discontinuities has been discussed by Hird et al (1990) and
Van Langen and Vermeer (1991). Basset and Guest showed that the slip
line field (e.g: Atkinson, 1981) inferred from finite element and model tests
was similar to that predicted by plasticity theory.
Kwok, Hird and Pyrah
Kwok (1987) implemented a reinforcement element and an interface
element in the finite element program CRISP81. Analyses of both real and
idealized embankments were carried out. The modified version of CRISP8I
was used in a parametric study to investigate the effect of the reinforcement
stiffness, the subsoil conditions, the geometry and the embankment
constitutive model. An analysis of a trial embankment constructed at
Stanstead Abbotts, UK, (Basset, 1987) was also performed. Kwok
concluded that in the short term considerable reductions of deformations
may be achieved by the insertion of reinforcement, provided that the
reinforcement is sufficiently stiff and strong. Examination of the
embankment model showed that an elastic-perfectly plastic model produced
realistic behaviour. However, care must be taken that unrealistic tensile
stresses do not develop.
16

Literature Review

Hird and Kwok (1990a) presented results from two embankment analyses
using the modified version of CRISP. Interface elements were used to
model the soil/reinforcement interfaces and it was shown that useful
information regarding the transmission of shear stresses from the soil to the
reinforcement could be extracted.
Hird and Kwok (1990b) carried out a parametric study of an idealized
embankment considering the effect of the reinforcement stiffness and the
subsoil depth. The authors concluded that provided the reinforcement was
sufficiently stiff and strong: 1) subsoil deformations would be significantly
reduced, 2) the development of tension in the reinforcement would be in
accordance with the principles suggested by Jewell (1988) and 3) very stiff
reinforcement may cause arching in the embankment fill. Study of the effect
of subsoil depth showed that for a subsoil of constant strength the effect of
reinforcement reduced with increasing depth.
Hird and Pyrah (1990) performed a parametric study of a second trial
embankment constructed at Stanstead Abbotts, UK The embankment was
modelled with both elements and equivalent vertical loads. The finite
element predicted vertical displacements compared well with those
observed. The horizontal displacements agreed less well and according to
the authors may have been attributed to the relatively coarse mesh. The
analyses modelling the embankment using equivalent vertical loads
produced unrealistic reinforcement stress and strain profiles, this may have
been due to the large deformations which occurred in a highly compressible
peat layer.

2.2. Multi-Stage Construction

Analyses of stage constructed embankments must allow for the dissipation
of the excess pore pressures which develop during construction.
Numerically this can be achieved in several ways. Firstly, a partially drained
17

Literature Review

situation can be simulated by specifying a reduced value for the bulk

modulus of water. Secondly, the excess pore pressure profile within the
subsoil can be fixed to some assumed value. Thirdly, a fully coupled
consolidation approach can be used.
The analysis is further complicated by the frequent inclusion of vertical
drains which make the assumption of plane strain conditions invalid. In
order to analyse accurately an embankment constructed on a subsoil in
which vertical drains have been installed a rational approach to the
modelling of the drains must be developed. Finite element analyses in
which modelling of vertical drains has been attempted are reviewed in
Section 2.2.2.

2.2.1. Consolidation Analyses Without Drains

Smith and Hobbs
Smith and Hobbs (1976) presented one of the first coupled consolidation
finite element analyses of embankments constructed over a soft clay. The
stage construction of two embankments was analysed. In both cases the
rate of centreline settlement was predicted accurately. The excess pore
pressures were in close agreement with observed values for one
embankment but less favourable agreement was achieved for the second.
The authors stated that the poor agreement in the second analysis may
have been due to anomalies in the field records.
Wroth
Wroth (1977) analysed the MIT test embankment. The embankment was
built quickly and then allowed to consolidate for five years after which it was
loaded as quickly as possible to failure. Predictions of the amount of fill
required to produce failure were invited. Wroth represented the subsoil
using a critical state soil model. The five year wait period was simulated by
using a reduced bulk modulus for water to estimate the in-situ conditions
18

Literature Review

before the final loading stage which was then modelled as completely
undrained. The finite element analysis could not correctly predict failure as
the clay subsoil exhibited strain softening behaviour. However, the predicted
displacements and excess pore pressures due to the additional fill agreed
favourably with the observed values.
Imperial College
Several finite element studies have been carried out at Imperial College.
Bond (1984) compared different drainage conditions and illustrated the
strengthening effects of a surface crust for a 'typical' soft clay using an
extended modified Cam-clay constitutive model to represent the subsoil.
The same set of soil parameters were use by Smith (1984) to investigate
some effects of stage constructed embankments. The embankment was
constructed in Im lifts with partial dissipation of excess pore pressure
specified after each increment. Smith observed large principal stress
rotations during the first undrained loading stage and that these rotations
were greatest below the toe of the embankment. Interestingly no rotations
greater that 5 were observed during subsequent loading or consolidation
stages. Smith also plotted the variation of the undrained shear strength ratio
with the maximum principal effective stress and the vertical effective stress
(s)a 1' and sJa ') against embankment height and observed that neither of
these ratios fell below the normally consolidated value.
Further details of these analyses can be found in Jardine and Hight (1987).
Schafer
Schafer (1987) carried out fully coupled consolidation analyses of reinforced
embankments using a finite element program which included an extended
version of modified Cam-clay, used to model the soil, and bar elements,
used to model the reinforcement.

19

Literature Review

The program was used to analyse the Mohicanville Dike (Schafer, 1987;
Duncan et al, 1988). The subsoil consisted of peat and soft clay on which a
previous embankment had failed. Steel reinforcement was used to allow the
construction of the dike to the design height. The finite element analysis of
these highly complex subsoil conditions showed good agreement for
reinforcement stresses and subsoil displacements. However, the predicted
excess pore pressures were less good. The discrepancies were explained
as being due to creep occurring in the subsoil and ageing of the fill
material.
Schafer (1987) also analysed the test embankments at St Alban, Canada,
constructed on Champlain Clay. The author found that the initial results
were not in good agreement with the observed behaviour. A reinterpretation
of preconsolidation pressures improved the quality of the analysis with
respect to predicted failure, displacements and excess pore pressures. The
author suggested that the lack of agreement, when using the measured soil
properties, occurred because of the inability of the constitutive model to
reproduce the brittle behaviour of Champlain Clay.
Almeida
Almeida (1984) used a version of CRISP to compare finite element
analyses with centrifuge tests (Almeida et al 1985; Almeida et al, 1986). The
finite element analyses did not accurately predict failure, but this may have
been due to the embankment being modelled as elastic. Subsoil stress
paths predicted by Almeida were similar to those predicted by Smith (1984).
Almeida (1984) implemented a variation of permeability with stress level
relationship and presented results indicating improved pore pressure
prediction when using this algorithm. The predictions of excess pore
pressure using fully coupled consolidation were, generally, encouraging.

20

Literature Review

Dluzewski and Term aat

Dluzewski and Termaat (1990) used a novel method to analyse stage
constructed embankments. The procedure consisted of an undrained
elasto-plastic analysis of loading stages followed by an elastic consolidation
analysis for the wait stages. This method of analysis simplified the
calculations and reduced the computer time required for solution.

2.2.2. Subsoils Containing Vertical Drains

Zeng, Xie and Shi
Zeng et al (1987) conducted fuHy coupled consolidation analyses to assess
the effectiveness of two methods for representing vertical drains. The first
method, developed by Shinsha et al (1982), calculated an equivalent
horizontal permeability for a plane strain analysis in which the vertical drains
were represented as sheets. The second method analysed the same
cross-section but was converted into a three dimensional problem by
modelling half a unit cell (Chapter 5) width into the plane using
three-dimensional elements.
The authors showed an improvement when using the three dimensional
analysis but acknowledged the large increase in computation cost.
Hird and Kwok
Hird and Kwok (1986) performed a parametric study of a test embankment
constructed at Stanstead Abbotts, UK The subsoil consisted of clay and
peat into which vertical drains were installed.
The effect of the vertical drains was allowed for by globally increasing the
subsoil permeability by an estimated factor. The observed displacements
and reinforcement strains were within the predicted range, although the
predictions of pore pressure were not good. The error in the excess pore
pressure prediction were most likely caused by an under-estimation of the
21

Literature Review

global permeability. The peat showed marked variation of permeability with

stress level and the authors indicated that this should be allowed for in
future analysis.
Poran
Poran (1986) analysed the Stanstead Abbotts trial embankment. The soil
was represented using a 'Bounding Surface Plasticity' model and the
reinforcement as a linear viscoelastic material.
An attempt was made to represent the vertical drains installed in the
subsoil. An axisymmetric analysis of a cylindrical unit cell, with drainage
occurring both horizontally, inwards, and vertically, upwards, was compared
with a plane strain unit cell with drainage towards the upper boundary only.
An 'equivalent' vertical permeability for the plane strain analysis was
interpolated to produce a similar average degree of consolidation at two
points in time. This equivalent vertical permeability was then used in a full
plane strain analysis.
The authors stated that it was difficult to compare the predicted pore
pressures with the observed values as several major assumptions had to be
made due to a lack of input information. This led to a wide range of
predictions and the accuracy of the procedure for modelling the vertical
drains could not be assessed.
Sanchez and Sagaseta
Sanchez and Sagaseta (1990) used a version of the finite element program
CRISP to carry out a back analysis of an embankment built over a soft clay
subsoil in which vertical drains were installed to reduce the required
construction time. The authors developed a method of equating the drain
spacing in a plane strain finite element analysis with the field drain spacing
by manipulating the horizontal subsoil permeability. However, using this

22

Literature Review

method the derived plane strain vertical drain spacing is only applicable for
one degree of consolidation.
A consolidation analysis of the embankment was compared with a partially
drained analysis in which the bulk compressibility of the pore water was
manipulated. The agreement of the finite element analyses with the
observed behaviour was not good but the consolidation and partially
drained analyses produced similar results.

2.3. Discussion
2.3.1. Summary of Embankment Finite Element Analyses
The finite element analyses reviewed above have been divided into two
categories: firstly, undrained analyses (which have been used to assess the
short term behaviour of a single stage constructed embankment) and,
secondly, analyses which have considered the dissipation of the excess
pore pressures due to the consolidation of the subsoil. Particular attention
has been placed on analyses which have modelled reinforcement andlor
vertical drains.
The development of the constitutive models used to represent the subsoil
has been an important aspect of the analysis of embankments constructed
over soft soils. The earliest analyses used an elastic subsoil model (Brown
and King, 1966) but quickly developed with a non-linear elastic model
(dough and Woodward, 1967), an elastic-perfectly plastic model (Smith
and Hobbs, 1974) and a critical state model (Wroth and Simpson, 1972).
Improvements of the constitutive modelling for the embankment material
have also been made and several authors have indicated that accurate
modelling of the embankment is essential to ensure realistic predictions.
Kwok (1987) performed a parametric study of embankment models. Hird
and Pyrah (1990) compared the use of finite elements, using an
23

Literature Review

elastic-perfectly plastic constitutive model with a Mohr-Coulomb yield

criterion, with equivalent vertical loads and found the equivalent vertical
loading produced unrealistic results. Almeida et al (1985) indicated that the
use of an elastic embankment model produced unrealistic failure
predictions.
Over the last decade with the availability of faster computers with larger
data storage, it has become possible to perform finite element analyses of
increasingly complex problems involving larger numbers of elements. This
has allowed the analysis of reinforced embankments in which the
reinforcement has been modelled using bar elements and the interface with
interlace or slip elements (Rowe, 1982; Boutrup and Holtz, 1983; Kwok,
1987; Schafer, 1987, Hird and Kwok, 1990a). The use of the finite element
method for reinforced embankment analysis has provided evidence which
has allowed a better understanding of the action of the reinforcing material
and facilitated the development of methods for their design (Rowe and
Soderman, 1982). The use of parametric studies of back analyses and
idealized reinforced embankments has also provided insight into the
possible improvements provided by reinforcement
In order to model accurately embankments in which strengthening of the
subsoil occurs due to consolidation, it is necessary to model the dissipation
of excess pore water pressures with time. Coupled consolidation finite
element analyses have been shown to be capable of predicting the pore
pressure response in the field and in centrifuge tests (Smith and Hobbs,
1976; Wroth, 1977; Schafer, 1987; Almeida, 1984).
An area which has been less widely researched is the modelling of vertical
drains in a subsoil subjected to embankment loading. Several authors have
proposed approximate procedures for the modelling of vertical drains. Hird
and Kwok (1986) and Poran (1986) allowed for the effect of the vertical
drains by increasing the vertical permeability of the subsoil in plane strain
24

Literature Review

analyses. This approach increases the rate of consolidation but fails to

model the localized consolidation around individual drains. The permeability
increase factor is also difficult to define. Shinsha et al (1982) and Sanchez
and Sagaseta (1990) developed methods in which an equivalent drain
spacing in a plain strain finite element analysis could be calculated. These
methods produce an equivalent average degree of consolidation at only one
stage of the analysis and the methods cannot account for the effects of well
resistance or smear which have been shown to be significant in certain
situations (Jamiolkowski et al, 1983).

2.3.2. Validation and Accuracy of Modelling

It is necessary to validate a finite element program before sufficient
confidence in the ability to model real situations is achieved. The validation
can be approached in two ways. Firstly, by analysing relevant idealized
problems for which analytical solutions are available. If satisfactory
agreement is obtained then finite element analyses can be performed to
examine more complex problems for which analytical solutions are not
available. Secondly, the back analysis of case histories enables the finite
element predictions to be compared against actual behaviour. Differences
between observed and predicted behaviour must be assessed carefully in
order to highlight deficiencies in the analysis as opposed to inaccuracies in
input parameters or observed results.
Several authors have compared undrained plane strain finite element
analyses with limit equilibrium solutions (Brown and King, 1966; Smith and
Hobbs, 1974; Rowe and Soderman, 1985) and with plasticity solutions for
strip footings (Rowe and Soderman, 1987; Basset and Guest, 1990; Hird et
al, 1990). Comparisons of analytical and finite element analyses for
consolidation of soil using vertical drains are rare, and the Author is not
aware of any comparison made between strength increases beneath

25

Literature Review

embankments predicted by finite element analyses and those predicted by

simple design methods.
The improvement in constitutive modelling and the refinement of analysis
techniques have resulted in the potential to model embankment problems
more accurately. However, these improvements must be shown to
reproduce observed behaviour. This can only be achieved by the back
analysis of embankments or model tests.
Back analysis requires high quality case histories which provide sufficient
soil data, detailed construction information, accurate instrumentation and
the extended recording of observed behaviour. The definition of the subsoil
conditions is very difficult and parameters such as mass permeability can be
unreliable. During and after construction it is necessary to monitor pore
water pressure and settlement at several positions and to record lateral
movements, particularly near the toe.
Several authors have attempted back analyses and with the refinement of
input parameters it is possible to produce good agreement between
observed and finite element predicted pore pressures and displacements.
However, it is not possible to assess the accuracy of the finite element
analysis until sufficient confidence in the input material parameters and
observed results is achieved.
Centrifuge testing can provide data which can be used as an alternative to
full scale embankments. Several authors have used centrifuge tests for
comparison with finite element analyses (Almeida, 1984; Basset and Guest,
1990). Centrifuge tests have the advantage, over full scale embankments,
that they are relatively cheap and provide reproducible results under
controlled laboratory conditions.

26

Literature Review

2.4. The Need for Further Finite Element

Research
A large amount of research into the finite element analysis of embankments
constructed over soft clays has been carried out over the last few decades.
Advances in the constitutive modelling of the subsoil and fill and the
representation of reinforcing materials and their interface with the soils have
been made. However, there are several areas that have been less well
researched, and further analyses are presented in this Thesis, as outlined
below:
1. Plasticity solutions are available for the collapse loads of rigid strip
footing over various subsoils. Jewell (1988) has suggested that
these solutions are relevant to the design of single stage reinforced
embankments. In order to assess the validity of these solutions finite
element analyses have been compared with plasticity theory for
undrained loading of two idealized subsoils. Further analyses have
been performed and a simplified design method which is based on
plasticity theory has been proposed (Chapter 4).
2. Vertical drains have become increasing popular in embankment
construction. Few attempts have been made to develop a rational
approach to their inclusion in plane strain finite element analyses
and these methods have significant limitations. In Chapter 5 the
consolidation behaviour of single drains is investigated by
comparison of finite element analyses with an analytical solution. A
new procedure for modelling vertical drains in plane strain finite
element analyses is developed and validated.
3. Few authors have carried out analyses of stage constructed
embankments and the Author is not aware of any comparisons of
27

Literature Review

simple design methods with finite element analyses. In Chapter 7 an

idealized two stage constructed embankment is analysed and the
strength increases predicted in a fully coupled analysis are
compared with a simple design procedure. The procedure is then
used to analyse the Porto Tolle trial embankment (analysed in
Chapter 6) and again comparisons of strength increases predicted
by finite element analysis and the simple design procedure are
corn pared.

28

Crisp and Modifications

3. The Finite Element Program

CRISP and Modifications
CRISP is a finite element program for the analysis of geotechnical
problems; the program has been modified and used in all analyses
presented. Details of the use of the modified program are given in Russell
(1992). In this Chapter the 1984 version of CRISP (Gunn and Britto, 1984)
is described and then the modifications to the program, to enable the
efficient analysis of reinforced embankments on soft ground incorporating
vertical drains, are presented.

3.1. Crisp
3.1.1. The History of Crisp
The computer programs known as CRISP were developed at Cambridge
University. Work started in 1975 when the program was originally called
MZOL but in 1976, after additional work, it was renamed CRISTINA. The
program was developed further and in 1982 given the name CRISP
(itical state Erogram). The 1982 version of the program (Gunn and
Britto, 1982), referred to as CRISP82, has been used in previous research
of reinforced embankments at the University of Sheffield (Kwok, 1987).
In 1984 a new version of the program was produced (CRISP84). The
double precision version of the 1984 program has been modified, by the
Author, and used in this research. In the Thesis this version of the program
will be referred to as CRISP.
Other versions of the program are available. In 1987 a reduced program
was made available to coincide with the publication of Britto and Gunn

29

CnSD

and Modifications

(1987). In 1990 the program was extensively rewritten for use on personal
computers (CRISP9O).

3.1.2. Types of Analysis

Analyses can be carried out in plane strain, axisymmetry (with axisymmetric
loading) or three dimensions. The soil can be modelled as drained,
undrained or as undergoing fully-coupled consolidation.
The method by which the program allows for drained or undrained
conditions is by the specification of an appropriate value for the bulk
modulus of water (I(). The bulk modulus of water is then included in the
formulation of the element stiffness matrices.
Types of Constitutive Model Available
The five constitutive models available are: anisotropic elastic,
inhomogenous elastic (Young's modulus varying with depth), Cam-clay,
modified Cam-clay and elastic-perfectly plastic.
The elastic-perfectly plastic model can be used with four yield criteria: Von
Mises, Tresca, Drucker-Prager and Mohr-Coulomb.
Types of Finite Element Available
Eight elements are available. The basic four elements are the linear strain
triangle, the linear strain quadrilateral, the cubic strain triangle and the linear
strain brick. Each of these elements have nodal displacements as
unknowns and each element is also available with pore pressure nodes for
consolidation analysis. All elements are formulated using full integration.
Theoretical limitations on the use of linear strain elements have been
identified. For example, when using full integration Sloan and Randolph
(1982) has shown that failure loads for undrained analyses using
elastic-perfectly plastic constitutive models are overestimated in certain
circumstances. The linear strain triangle and linear strain quadrilateral
30

Cnso and Modifications

over-estimate axisymmetric failure loads. This over-prediction can be

avoided by using the higher order cubic strain elements or reduced
integration.
When analysing an embankment in plane strain, due to the complex
geometry and soil conditions, it is often necessary to use a large number of
elements. This is particularly the case when considering reinforcement
and/or vertical drains. Cubic strain elements cannot be used economically
for such problems, and so for the present research linear strain elements
have been used to model the soil.
A subsoil containing vertical drains forms a three-dimensional problem.
However, the three-dimensional analysis of an embankment would be
impractical, even when using modern super computers, due to the large
number of elements required for accurate modelling. The present
embankment analyses have been carried out in plane strain and a
procedure has been developed for modelling accurately the
three-dimensional rate of consolidation around a single vertical drain in a
two-dimensional plane strain analysis (Chapter 5).

3.1.3. Additional Features of Crisp

Material Non-Linearity
An incremental tangent stiffness solution procedure is used by CRISP for
analyses including non-linear constitutive models. The result is that for any
increment in which an element's stiffness changes an error in equilibrium is
generated as the change in stiffness of the element is ignored.
For the elastic-perfectly plastic constitutive models it is possible to correct
the stress state back to the yield surface by calculating 'out of balance'
loads which are applied in the next increment. CRISP corrects the stresses
back along a line normal to the yield surface (Owen and Hinton, 1980). This

31

CrisD and Modifications

is an arbitrary assumption so that even though the yield criterion is not

violated a large number of increments must still be used to ensure that the
stress state never passes beyond the yield surface by a large amount
Tangent stiffness parameters are also used for the critical state constitutive
models; 'out of balance loads' are not calculated as the correct position on
the yield surface is unknown. In order to reduce these errors it is essential
that a large number of increments are used. CRISP produces information
on the rate of change in the size of the yield locus and recommended
values are discussed by Britto and Gunn (1987).
An alternative solution technique is the use of iterations to ensure that
displacements and forces are consistent at the end of each increment.
CRISP has been compared with a finite element program using an iterative
solution technique (Potts et al, 1990) and was shown to produce a correct
solution provided sufficient increments were used.
To ensure that the CRISP analyses are accurate it is necessary to perform
several analyses with increasing numbers of increments until a further
increase in the number of increments produces no change in the final
increment displacements. Where practical this procedure has been adopted
in the analyses presented.
Geometric Non-Linearity
In analyses using non-linear constitutive models the specified changes in
load or displacement conditions are applied over a number of increments.
The result of each increment is a set of nodal displacements and as these
displacements become large the equilibrium equations based on the
undeformed geometry may no longer be accurate.
This geometric non-linearity can be approximated in CRISP by updating the
nodal coordinates at the end of each increment. The updated nodal
coordinates are then used to calculate the new element stiffness matrices.
This coordinate updating is only a first approximation to a large strain theory
32

Cris p and Modifications

(Cook, 1981). If excessively large strains occur the solution may no longer
be reliable.
Embankment Construction
CRISP allows for the simulation of the construction of an embankment by
specifying an initial mesh containing elements which represent the fill
material. CRISP then allows these elements to be added as the analysis
proceeds. To avoid using a large number of thin elements the self weight of
the added elements is applied over a number of increments. Excavation can
be simulated by removing elements.
Restarted Analyses
The original CRISP program outputted information at the end of each
increment so that an analysis could be restarted from some earlier point.
The file produced became extremely large for any reasonable number of
increments. In order to reduce its size the program has been modified so
that restart information is only saved for specified increments. The analysis
can then be restarted from any one of these increments.

3.2. Reinforcement Element

To analyse accurately reinforced embankments it is necessary to model the
reinforcing material. This could be achieved using thin quadrilateral
elements. Whilst this method has the advantage that no new elements need
be added to CRISP, the aspect ratio of the elements modelling the
reinforcement may become large enough to cause numerical problems. A
second approach, which overcomes this problem, is to model the
reinforcement using one-dimensional bar elements, Figure 3.la. This
method has the additional advantage that no extra degrees of freedom are
added to the mesh.

33

Crisp and Modifications

1
y
I=-i

S node

(ci)

strain
(b)

Figure 3.1 a) Reinforcement element; b) bi-linear constitutive model.

Following a preliminary study (Russell, 1988) it was decided that a bar

element of the type developed by Kwok (1987) should be added to CRISP.
The element was implemented for plane strain analyses only.

3.2.1. Constitutive Relationship

The reinforcement is usually a polymer material with high tensile strength
and little bending resistance. Therefore, in the element formulation the
flexural stiffness has been ignored. To maintain a simple constitutive
relationship the reinforcement is modelled as bi-linear elastic, Figure 3.lb.
Four parameters are required, the elastic Young's modulus (E 1 ), the
Young's modulus after yielding (E2), the yield stress

(ar)

and the

reinforcement thickness (t). The element thickness is required as an input

parameter, even though the element is formulated as one dimensional, and
is used to calculate the reinforcement element stiffness matrix.
It should be noted that the axial stiffness of a geotextile or geogrid is usually
expressed as force per unit width per unit strain (kNIm). This is commonly

34


Crisp and Modifications

referred to as the 'reinforcement modulus' (J) and is equal to the Young's

modulus (E) multiplied by the thickness (t) (for plane strain conditions).
When a bi-linear relationship is assumed the material behaves elastically
with a Young's modulus E 1 until the yield stress is reached after which the
modulus is changed to E 2. If unloading occurs the original modulus is
restored.

3.2.2. Reinforcement Element Stiffness Matrix Formulation

The reinforcement element, Figure 3.la, has three displacement nodes so
as to be compatible with the linear strain elements which are used to model
the soil.
The displacement at any point along the element, a, can be related to the
nodal displacements, a, by the element shape functions, N, as
a = Na0

.............. ................ .......... ........................(3.1)

For the element shown in Figure 3.la the shape functions, N, in the local
coordinate system, (), are
N=[ O.5(_1) o.5(^1)

(1_2) ]

........................(3.2)

The element strains, a, are expressed in terms of the nodal displacements

as
g =Bae

.................................................................(3.3)

where
ON

(34

........................................S UUUUUUUUUU ................U

In which x' is the coordinate system defining the element length, i.e. x'=O at
node I and x'=L at node 2, where L is the length of the element.
However, the shape functions are in terms of the element local coordinate
system, (p,). Using the chain rule for differentiation it can be shown that
35

CrisD and Modifications

(35)

N_ Nd,

ox'

, dx'

By consideration of the element geometry, Figure 3.la

'36

---

dx' - L

Differentiating the shape functions, Equation 3.2 and using the chain rule
B

=[

(0.5 - ) (0.5

-F

2 ]

.................................(

3.7)

The element stiffness matrix for a displacement finite element (Eg.

Zienkiewicz, 1977) is
Ke=j" BTDBd(VO1)

....

...............................................(3.8)

where D is the constitutive matrix relating stresses to strains as a=Da.

The cross sectional area for a unit width of the element in a plane strain
analysis is constant and equal to the element thickness, t, therefore
Equation 3.8 becomes
....... ...................... ................

(K) = tf11 B TDBd

To transform Equation 3.9 into the x' coordinate system it must be multiplied
by , which from Equation 3.6 is L12. Thus
(Ke)xirrJ'l B TDBIt

. .............................................( 3.10)

Numerical integration is used to calculate the element stiffness matrix for

each reinforcement element. Two point Gaussian integration would produce
an exact solution, but three Gauss points have been used to provide extra
data.

Incorporation of the Reinforcement Element in a Two-Dimensional Mesh

In general the reinforcement element will be at an angle a to the global (x)
axis. A transformation matrix T is used to convert the element stiffness

36

Criso and Modifications

matrix, (Kj 1,, to the global element stiffness matrix K in global coordinate
system (x,y).
ICT T(K)_r

................ (3.1 1)

where
000
0 0
0 0 ............................................ (3.12)
0000 Cs
cs0

r =

in which c=cosa and s=sina.

Calculation of the Reinforcement Length and Angle of Orientation
In analyses in which the option to update the nodal coordinates has been
used it is desirable for the length and angle of orientation of the
reinforcement element to be calculated at the start of each increment
These coordinates are then used in the calculation of the new element
stiffness matrices.
The continuum elements have an isoparametric formulation, therefore, the
element sides may be curved. The reinforcement element is approximated
by two straight lines connecting the three nodes, Figure 3.2, and the length
is calculated as
L= L1 -i L2 ............................................................

(3.13)

and the angle of orientation as

ct=O.5(cci +cL2)

...................................................... (3.14)

The error introduced by this approximation, for relatively coarse meshes, is

less than 1% (Kwok, 1987) and becomes smaller as the mesh is refined.

37

Crisp and Modifications

cx

Q.5(a1+c)

LL1+L2

modified cj eo me try

deformed geometry

L1

Figure 3.2 - Calculation of one-dimensional element length and

orientation.

3.2.3. Equivalent Nodal Forces

It is necessary to convert the element stresses into equivalent nodal forces,
F. These forces are found by integrating the stresses over the volume of
the element
F e r JB Tad(VO1)

.................................................... (3.15)

The element is used in plane strain analyses; therefore, consider a unit

width of constant thickness, t. Using the local coordinate system (a).
Equation 3.15 becomes
(Fe)

tJ'1 B Tad()

................................ ................(3.1 6)

Converting to the coordinate system (x')

(Fe) rJ' BTcyd()
Again a three point Gaussian integration scheme is used to evaluate
Equation 3.17 and the nodal forces are resolved into the global coordinate
system (x,y).
38

CrlsD and Modifications

3.2.4. Analyses Using the Reinforcement Element

To check the correct implementation of the reinforcement element two
problems were analysed. These test problems were analysed previously by
Kwok (1987).
The first analysis involved a single reinforcement element with a bi-linear
constitutive relationship. The element was restrained at one end and a load
applied at the other. The load displacement response predicted by the finite
element analysis was as defined by the constitutive model.
The second analysis involved a three bar structure for which the
displacements could be calculated analytically. The finite element analysis
produced the expected displacements.

3.3. Interlace Element

An interface element has been implemented for plane strain analyses and
can be used to model slip between the reinforcement and soil or at a rough
boundary. The element also allows interface shear and normal stress to be
retrieved easily. The element is similar to the Goodman (1968) relative
displacement element and was implemented previously in CRISP82 by
Kwok (1987).

3.3.1. Interface Element Stiffness Matrix Formulation

The interface element, Figure 3.3a, has a quadratic displacement variation
so as to be compatible with the linear strain elements used to model the
soil.
If a set of nodal forces, F0, are acting on the interface element such that
nodal displacements, a0, occur then it is possible to define a set of relative
local displacements at any position along the element, w, such that

39


CIISD and Modifications

U)
U)
a)
I-

In

T=6+qtan o

.1-I

U)

L
a
a)
-c

14

U)

nodal point

relative displacement
(b)

(a)

Figure 3.3 - a) Interface element; b) constitutive relationship.

f atop a bottom
L

top
bottom
8q a

.. ....... ......................... ....... ( 3.18)

The shape functions for the interface element are defined in Equation 3.2
and can be rewritten
N=[

Ni (112 tS3 ]

................................. ...............( 3.19)

The relative nodal displacements in terms of the actual nodal

displacements, q, and the shape functions are

We = Nrae

............. ................................................ (3.20)

where
-N, 0 -N2 0 N2 0 N1 0 -N3 0
0 -N1 0 -N2 0 N2 0 N., 0 -N3
L

.. ( 3.21)

Nr-iF

N ]

The stresses can then be related to the relative displacements by a

constitutive matrix, D, as

ci=Dw 9 ..............................................................(3.22)
The vector of stresses, a, has components of normal stress,

a,,

and shear

stress, t. The constitutive matrix has components of shear stiffness, k5, and
normal stiffness, k,.
40

CIISD and Modifications

D-1 k

.............................. (3.23)

0 k

Equation 3.23 implies that there is no coupling of the shear and normal
stiffnessses of the interface.
The principal of virtual work can now be used to define a set of nodal forces
which are in equilibrium with the state of internal stress. It is assumed that a
set of virtual nodal displacements, a0*, cause a set of virtual nodal relative
displacements,

where

w=Nra............................................................. (3.211.)
and from the principal of virtual work
,jlJ*Tad()

aTFe .._. J1

(3.25)

Substituting for a and w, in Equation 3.25 and dividing both sides by a.*T
Fe= 1

N r3N rd(

) a e

.............................................. (3.26)

The element stiffness matrix for the interface element is therefore

(Ke)=lhi F'J
TDN rd()

............................................. (3.27)

Transforming to the x' coordinate system

(Ke)=J1iNDNrd(,)

(3.28)

3.3.2. Transformation to Global Coordinate System

The element stiffness matrix from Equation 3.28 has been defined in the x'
coordinate system and must be transformed into the global coordinate
system (x,y). This is achieved using a transformation matrix, such that,
KegTT(Ke),1T, where

41

Crisp and Modifications

t
0
0
0
0
0

T=

0
t
0
0
0
0

0
0
t
0
0
0

0
0
0
t
0
0

0
0
0
0
t
0

0
0
0
0
0
t

...... . (3.29)

in which
coscc sina
-Sifla

................. (3.30)

To calculate the orientation of the element in the two-dimensional global

coordinate system, a, and the element length, L, a reference plane through
the centre of the element is defined by averaging the positions of the pairs
of nodes. The orientation angle and element length are calculated
according to Equations 3.13 and 3.14.

3.3.3. Interface Element Constitutive Relationship and Modes

of Behaviour
The interlace element has been implemented as elastic-perfectly plastic
with a Mohr-Coulomb yield criterion, Figure 3.3b. The elastic shear stiffness,
k5, is set to zero when the soil yields. The normal stiffness, k 1, is not
changed when yielding occurs but can be set to a modified value, kM. if the
normal stress becomes tensile. Kwok (1987) showed that the interface
element is insensitive to the value of shear and normal stiffness except
when very low values are used.
Five modes of behaviour can be represented:
1 Non-slip - the element is behaving elastically with a normal stiffness,
k, and shear stffness, k.
2 Slip - when the shear stress reaches the bond strength, governed by
the Mohr-Colomb relationship, slip occurs; the shear stiffness, lc, is
then set to zero.
42

Crisp and Modifications

3 Unloading - if unloading occurs then the original shear stiffness is

restored.
4 Separation - if the normal stress becomes tensile then the shear
stiffness is set to zero and the normal stiffness is allowed to change
to a specified value, k,4.
5 Rebonding - if separation has occurred and during some later
increment the interface normal stress becomes compressive then
the original shear and normal stiffnesses are restored.
The mode of behaviour is identified at each of the three integration points
and the appropriate shear and normal stiffnesses are used to calculate the
element stiffness matrix.
Oscillation of stresses has often been observed when using interface
elements. To try to minimise this problem Kwok (1987) carried out a study in
which two methods of smoothing the value of the normal stress were
evaluated. The first method was a local stress smoothing technique (Hinton
and Campbell, 1974) and the second was a simple averaging of the normal
stress calculated at integration points along the element. Kwok showed that
the stress averaging technique was almost as efficient as the local
smoothing method. The interface element has been implemented using the
stress averaging technique so as to minimize oscillations of shear stresses
efficiently.
The stress averaging technique involves calculating an average normal
stress and using this value to calculate the Mohr-Coulomb shear strength.
Improved performance may be achieved by using a more complicated
integration rule (Hohberg, 1990).
The interface shear stress sign convention is defined in Figure 3.4.
Compressive normal stresses are positive. This Figure also identifies four

43

CrisD and Modifications

POSITION a

UPPER___________
VIIRFAcE\+
REIN
FOREI(NT

LOVER
IUERFAcE

,0

POSITION b

E7_
/0

ck ci

\+\

Figure 3.4 - Interface element shear stress sign convention and shear
stress regimes.
shear stress regimes which may develop, by differential movement of the fill
and subsoil, in a reinforced embankment.
3.3.4. Equivalent Nodal Forces

The equivalent nodal forces, which are in equilibrium with the internal
stresses, are calculated for the interface element at the end of each
increment. Stresses (c, an), which act on a plane through the centre of the
interface element, are used to calculate the nodal forces F.
Consider an infinitesimally small section of the element. The forces which
are in equilibrium with the stresses are

dFx = (tcoscx-.. sin a)
cm

ciiy=(tsina-i-cmncosa)c

(3.31a)
(3.31b)
44

Cris p and Modifications

where a defines the orientation of the interface element central reference

plane in the global (x,y) coordinate system, Section 3.3.2.
If a set of virtual displacements, a*, are applied to the element then the
virtual displacements at any position along the element can be related to the
nodal virtual displacements by the element shape functions
a* = a

......................................................... (3.32)

The work done by the nodal forces due to the virtual displacements is
equivalent to the work done by the internal stresses
(dFe)=f1 N(rcosaansina)d,

(dFe) = J 11 N(tsina+ancosa)d

(3. 33a)
(3.33b)

The equivalent nodal forces are then evaluated by numerical integration.

These forces are acting on both faces of the interface element but in
opposite senses.

3.3.5. Analyses Using the Interlace Element

A finite element analysis was carried out to ensure that the Mohr-Coulomb
yield criterion had been implemented correctly. The test involved a single
element to which a shear displacement was applied. The shear stress
response was compared with that defined by the constitutive relationship
and found to be acceptable. Further details of this analysis can be found in
Kwok (1987).
Results from an analytical solution developed by Hird and Russell (1990)
were used to compare with a second finite element analysis. The analytical
solution allows the calculation of the shear stress distribution along a rough
boundary of a block which is loaded in compression. Results from the finite
element analysis were in excellent agreement with those predicted using

45

Crisp and Modifications

the analytical solution; comparisons are presented by Hird and Russell

(1990).

3.4. Drainage Element

In consolidation finite element analyses of subsoil containing vertical drains
it is often adequate to represent the drain as a boundary with zero excess
pore pressure. However, this approximation is in error if well resistance, due
to the drain's finite permeability, restricts the flow along the drain.
To model the well resistance accurately an element, termed a drainage
element, has been developed for use in plane strain and axisymmetric
consolidation analyses (Russell, 1990). The drainage element can be used
to represent a thin layer of relatively high permeability in consolidation
analyses, for example a sand lamination in a clay soil or, as for the present
research, to model a vertical drain in a clay soil.

3.4.1. Element Formulation

The drainage element is compatible with linear strain elements having three
displacements nodes and two pore pressure nodes, Figure 3.5.
Discretising the continuity and equilibrium equations (Biot, 1941) and using
a fully implicit approximation over time, the element stiffness matrix is
IKS L
Ke
I M

................... (3.31)

where
KS=J' BTDBd(vol)

stiffness matrix

L=f" BmNPd(vot)

link matrix

,tf'

permeability matrix

ETkEd(VOt)

in which At is an increment of time, N are the displacement shape functions,

B are the first derivative of the displacement shape functions, N'1 are the
46


Crlso and Modifications

In

displacement node
+ pore pressure node

Figure 3.5 - Drainage element.

pore pressure shape functions, E are the first derivative of the pore
pressure shape functions, k is the permeability matrix, D is the constitutive
matrix, m=(

o j and is the unit weight of water.

3.4.2. Transformation to Two-Dimensions

The element stiffness matrix, Equation 3.34, is in the x' coordinate system
and must be transformed to the global coordinate system (x,y) using a
transformation matrix
= TT(Ke) T

...................... (3.35)

where
cosa sina

1= 0

o
o

o
0

0
o
o 00
0
0 00
cosa sincx
0 cosa Sina 0 0
o
0
o -0 1 0
o
0
o
a 01
o

(3. 36)

The element length, L, and angle of orientation, a, are defined as for the
reinforcement element, Figure 3.2.
47

CrisD and Modifications

3.4.3. Incorporation of the Drainage Element in a

Two-Dimensional Mesh
The drainage element has been transformed so that it can be used as a
two-dimensional element, but before it is included in a two-dimensional
mesh the significance of each of the terms in the element stiffness matrix,
Equation 3.34, must be considered.
A fully coupled Biot consolidation analysis with a fully implicit approximation
over the time increment (Booker and Small, 1975) is used in CRISP. For an
increment of time, At, the discretized form of the continuity and equilibrium
equations are:
equilibrium

KSAd+LAu=AF1

continuity

LTAd-AtcDAu=AF2

where Ad, Au, AF1 and AF2 are increments of nodal displacement, pore
pressure, nodal force and nodal flow respectively.
Considering each term separately:
1 KSAd - is the change of nodal displacement, Ad, caused by an
increment of nodal force AF1.
2 LAu - is the change of nodal pore pressure, Au, caused by an
increment of nodal force AF1.
3 LTAd - is the change of nodal displacement, Ad, caused by an
increment of flow, AF2.
4 AtcDAu - is the change of nodal pore pressure, Au, caused by.an
ncrement of flow, AF2.
Terms 1 and 4 are relatively easy to understand, in that they have the usual
finite element meaning. KS is the force stiffness ascribed to the element
multiplied by the defined area; similarly, Q can be considered as the flow
48

Cris and Modifications

stiffness of the element which is related to the drain's permeability and

defined area. However, the element itself is one-dimensional and whilst it
does have force and flow stiffnesses, it does not have a real cross sectional
area when included in a two-dimensional mesh. The linkage terms 2 and 3,
are therefore zero.
Expressed another way, the element is not a simple one-dimensional
consolidation element of unit cross sectional area (where terms L and

LT

would be non zero) but a bar element with force and flow stiffness but no
real thickness. The element stiffness matrix, Equation 3.34, is reduced to

IKS 0
Ke L 0 Atb

...................(3.37')

3.4.4. Analyses Using the Drainage Element

The correct implementation of the drainage element has been checked in
two series of analyses. Firstly, comparing plane strain analyses using
drainage elements, to represent relatively permeable laminations within a
less permeable soil, with previous research using a different finite element
program (Abid, 1986; Abid and Pyrah, 1990). Secondly, axisymmetric
analyses of inward drainage to a vertical drain were compared with an
analytical solution (Hansbo, 1981). The latter analyses are fully documented
in Chapter 5.

3.5. Variation of Permeability with Stress Level

The modified Cam-clay constitutive model provides a realistic framework for
the analysis of soft clay soils. However, an observed behaviour of soft clay,
not implemented in the critical state models in CRISP, is the reductior of
permeability with increasing stress level (Tavenas et al, 1983). In order to
assess the influence of this effect modifications to the program have been
made.

49

CrisD and Modifications

Tavenas et al (1983) reviewed several permeability/void ratio relationships.

They compared each relationship with carefully derived experimental results
from intact soil samples taken from 14 different sites in Canada, USA and
Sweden. They concluded that Taylor's (1948) relationship, Equation 3.38,
was the most representative for soils of initial void ratio less than 2.5 at
volumetric strain levels normally encountered.
Taylor (1948) suggested an empirical linear relationship between the
logarithm of the permeability, k, and the void ratio, e
log

k= log ko - (e0__e). ......................................( 3.38)

Where k0 and e0 are the initial values of the permeability and void ratio
respectively and Ck is the permeability change index. Tavenas et al (1983)
showed that the permeability change index can be reasonably
approximated as Ck=O.5e0.
The permeability stress level relationship, Equation 3.38 (with Ck=O.5e0),
has been implemented in CRISP. The option is available when using
modified Cam-clay and if the option is selected the program calculates a
new horizontal and vertical permeability at the end of each increment.
Analyses, of an idealized triaxial test, were carried out in which the
permeability and void ratio were monitored and found to agree with
calculated values using Taylor's relationship. The variation of permeability
with stress level was used in an analysis of the Porto Tolle case history
(Chapter 6). A comparison of analyses of this case history showed
negligible difference with and without varying the permeability with stress
level.
Two aspects of the selection of permeability parameters which have been
found to be more significant are the initial values of both vertical and
horizontal permeabilities and ensuring that the coefficient of consolidation
throughout the clay depth has been modelled correctly (Section 7.3.2).
50

Crisp and Modifications

Results of analyses using the permeability variation algorithm are not

presented, but this aspect of soft clay behaviour will become more important
as the soil becomes more compressible and this algorithm may be of use in
future analyses.

3.6. Summary
The finite element program CRISP is a versatile program which can be used
to model geotechnical problems. However, in order to model accurately
reinforced embankments over soft soils incorporating vertical drains several
modifications were necessary:
I The implementation of a bi-linear elastic one-dimensional element to
model the reinforcing material in plane strain analyses.
2 The implementation of an elastic-perfectly plastic relative
displacement interface element to model the soillreinforcement
interface in plane strain analyses.
3 The implementation of a one-dimensional bar element which can be
used to model vertical drains in axisymmetric or plane strain
analyses.
4 The implementation of an algorithm which allows the permeability to
be varied with stress level when using the modified Cam-clay
constitutive model.
The modified version of the program has been used in all analyses
presented in the remaining Chapters.

51

Collaose of Undrained Subsoils

4. Collapse of Undrained Subsoils

4.1. Introduction
Plasticity theory (Eg. Atkinson, 1981) is a method frequently used for the
prediction of the collapse of geotechnical structures. It has been suggested
(Jewell, 1988) that plasticity solutions for the collapse of strip footings on
undrained soils can be used to predict the short term stability of reinforced
embankments built over soft cohesive deposits.
Plasticity theory is of limited use in that it can only predict collapse loads
and displacement patterns for specific problems. The finite element method
is a more flexible analysis technique which can reproduce plasticity
solutions as well as providing additional information for more complex
boundary conditions.
In this Chapter relevant plasticity theory is briefly discussed with reference
to two idealized soil profiles. Finite element solutions for rigid strip footings
are then presented which approximate closely to the plasticity theory.
Finally, finite element analyses of embankment loading are carried out to
verify a simple design procedure.

4.1.1. Plasticity Theory

The application of the theory of plasticity allows upper and lower bounds of
the collapse loads for a geotechnical structure to be calculated. If yield and
compatibility conditions are satisfied, but equilibrium conditions are ignored,
the solution is an upper bound of the collapse load. If equilibrium and yield
conditions are satisfied, but compatibility conditions are ignored, the solution
is a lower bound. If the upper and lower bounds coincide then the true
collapse load has been obtained.

52

Collaose of Undrained Subsoils

fI

cr1'

if3

a
x

Figure 4.1 - Definition of stress characteristics for undrained loading.

Coincident upper and lower bound solutions can be found using methods
which depend on the solution of hyperbolic partial differential equations.
One such solution technique is the method of characteristics (Eg. Atkinson,
1981). This method involves the extension of two sets of slip lines from
known boundaries, the directions of the slip lines being defined by the
boundary conditions. At the intersection of two slip lines the state of stress
can be calculated. The slip lines are referred to as a and 13, or stress,
characteristics and are associated with positive and negative shear stresses
respectively, Figure 4.1. For undrained soil the a and

13

characteristics

intersect at 900.
From the slip line field it is possible to consti uct a vector displacement
diagram which defines the relative displacement of areas bounded by slip
lines. The displacement diagram is referred to as a hodograph.

53

CollaDse of Undrained Subsoils

Reinforcement

Subsoil

Figure 4.2 - Reinforced embankment configuration.

Plasticity theory assumes that the soil is perfectly plastic satisfying the
normality condition and that the flow rule is associated. It is also assumed
that the coaxiality condition, which states that the directions of principal
stress and principal strain increment coincide, is valid.
A set of strain increment characteristics, also referred to as velocity
characteristics or zero extension lines, can be defined in a similar way to the
stress characteristics shown in Figure 4.1. For an undrained soil which is
undergoing purely plastic deformations and has an associated flow rule,
characteristics of stress and strain increment coincide.

4.1.2. Reinforced Embankments

(Embanlanents overlying soft clay soils often make use of geotextile
reinforcement, placed at the surface of the subsoil (Figure 4.2), to allow
construction to the design height within the area available (Bonaparte and
Christopher, 1987). The stability of the embankment is most critical at the
end of a loading stage, after which the subsoil consolidates and gains
strength. The reinforcement is only required to maintain the factor of safety

54

CollaDse of Undrained Subsoils

above unity until the subsoil has consolidated sufficiently to maintain

stability without the assistance of the reinforcement.
The reinforcement can improve stability in two ways. Firstly by resisting the
shear stresses developed due to the lateral thrust in the embankment.
Secondly, if sufficiently strong, the reinforcement may provide restraint to
the surface of the subsoil.
The force in the reinforcement due to the lateraJ thrust of the embankment

has been termed P and the force in the reinforcement due to the spreading
of the subsoil as ftidn (Jewell, 1988).
The stability of an embankment reinforced at its base has been discussed
by Hird and Jewell (1990). Three classes of failure can be identified: internal
stability of the embankment, instability of the subsoil and overall instability
involving the embankment and the subsoil. The second of these has been
selected for investigation; comparisons of finite element analyses and
plasticity theory are made for relevant subsoil conditions.

4.1.3. Application of Plasticity Theory to Reinforced

Embankment Problems
The loading due to an embankment is similar to that of a rigid strip footing.
A heavily reinforced embankment is analogous to a rough footing in which
the soil surface shear strength, is fully mobilised. A smooth footing can
be thought of as an embankment in which the reinforcement is sufficiently
strong to resist the shear stress generated by the embankment but unable
to provide any restraint at the subsoil surface.
Plasticity solutions have been developed for two idealized soil profiles.
Firstly, a subsoil of uniform strength over a limited depth with a rough rigid
layer beneath, Figure 4.3a (Mandel and Salencon,

1972)

and secondly, a

55

CoilaDse of Undrained Subsoils

pX

subsoil of constant strength (s)

rough rigid base
(a)

su
subsoil with strength increasing
linearly and indefinitely with depth

p
z
(b)
Figure 4.3 - Idealized soil profiles for plasticity analyses: (a) uniform
strength/limited depth; (b) strength increasing linearly with depth.

subsoil with strength increasing linearly and indefinitely with depth, Figure
4.3b (Davis and Booker, 1973).
Plasticity theory gives the distribution of vertical stress on the underside of
the rigid footing, shown in a non-dimensional form in Figure 4.4. These are
ideal distributions which make the best possible use of the available subsoil
strength and provide a meaningful comparison for finite element analyses.

56


CollaDse of Undrained Subsoils

25

20

15
Jo
b
10

5
x/D

10

Figure 4.4a - Ideal load distribution on an idealized soil profile of

uniform strength!imited depth (Mandel and Salenon, 1972).
30
25
rough

5
01
0

5
1ox/Su.

10

Figure 4.4b - Ideal load distribution on an idealized soil profile of

strength increasing linearly with depth (Davis and Booker, 1973).
57

ColIase of Undrained Subsoits

4.2. Finite Element Analysis of Plasticity

Problems
CRISP has been used to perform plane strain undrained analyses of the
plasticity problems defined in Figure 4.3. Both rough and smooth footings
were modelled so that four situations were examined. Linear strain triangles
using an elastic-perfectly plastic constitutive model, with a Tresca yield
criterion, were used to model the soil; this is equivalent to the yield criteria
used for the plasticity solutions. Interface elements have been used to
model all rough boundaries, the interface strength is taken as equal to that
of the soil adjacent to the element and is assumed to be purely cohesive.

4.2.1. Meshes for Equivalent Loading Problems

In all cases advantage was taken of symmetry and only half of the loaded
width was modelled.
For the soil of uniform strength and limited depth (0), the mesh for the
smooth loading case is shown in Figure 4.5. The width of the loading was
relatively large (20D). Sufficient distance was allowed beyond the edge of
the loading that the conditions at the left-hand boundary of the mesh had no
significant effect on the solution. It was assumed that the soil rests on a
perfectly rough, perfectly rigid surface, modelled using a row of interface
elements. In the rough case, the same mesh was used except that a row of
interface elements was also included along the loaded surface, which was
restrained horizontally.
For the soil with strength increasing linearly with depth, the mesh for the
smooth loading case is shown in Figure 4.6. As the soil is of unl!mited
depth, the mesh had to be sufficiently extensive to ensure that none of the
boundaries affected the solution significantly. The extent of the failure zone
was initially estimated from charts produced by Houlsby and Jewell (1988).
58

CollaDse of Undrained Subsoils

287 linear strain triangles

D=4.Om

39 interface elements
1488 degrees of freedom

3D

1OD

DI
restrained
horizontally

restrained
horizontally

restrained vertically
and horizontally

Figure 4.5 - Finite element mesh used for uniform strength/limited depth
analyses (mesh shown for smooth case).

243 linear strain triangles

B=1 2.Om

1096 degrees of freedom

1.5B

S
C

S.

C
S
I)

BI
horizontally and
vertically restrained

Figure 4.6 - Finite element mesh used for strength increasing linearly
with depth analyses (mesh shown for smooth case).

59

CollaDse of Undrained Subsoils

For the rough loading analysis a row of interface elements was included
along the surface and no horizontal movement was allowed.

4.2.2. Material Properties

In working with an elastic-perfectly plastic material to obtain comparisons
with the plasticity solutions, it was important to restrict elastic strains as
much as possible. A high value of shear modulus was therefore used
(G=6000kPa). Undrained conditions can be modelled by specifying a large
bulk modulus for water, but if this parameter is too large, numerical
problems may occur (Griffiths, 1985). A bulk modulus for water
K.5x1 O5kPa (equal to a total stress Poissons ratio v=O.4925) was found to
be satisfactory.
The soil yielded according to a Tresca yield criterion when the shear stress
equalled either a constant value, s=3OkPa (limited depth case), or a value
which increased with depth from the subsoil surface, s 0=2.5kPa, at a
constant rate of p=2kPa/m (strength increasing with depth case). For the
case of strength increasing linearly with depth the strength profile could not
be modelled exactly, but it was possible to divide the mesh into several
horizontal layers and to specify a different strength for each layer, Figure
4.7. The resulting stepped distribution was found to give sufficiently
accurate results.
When interface elements were used their strength was set equal to the
strength of the adjacent idealized soil. To limit the relative displacements
prior to slip the shear and normal stiffness parameters were set to high
values, k=6x1 O3kNIm2 and k5x1 O5kN/m2.

60

Collapse of Undrained Subsoils

4
E

1:
-10
-12

10

15
20
Shear strength S (lcPa)

25

30

Figure 4.7 - Finite element approximation for strength increasing

linearly with depth.

4.2.3. Results
Displacement Controlled Analyses
In order to model the rigid strip footing assumed by the plasticity solutions,
plane strain finite element analyses were carried out by applying increments
of uniform vertical displacement along the loaded surface. Failure was
deemed to have occurred when a further increase in vertical displacement
caused no significant increase in vertical stress at any point on the
displaced boundary. Failure was reached in approximately 100 increments.
The normal stresses in the upper interface elements (equivalent to the
vertical stress distributions at the subsoil surface) were compared with the
vertical stress distribution predicted by plasticity theory. The load-settlement
curves for three points on the subsoil surface beneath the footing, for each
of the four cases, are shown in Figure 4.8.

61

Colla se of Undrained Subsoils

p

700

x0
Ax=5D

600
500

xlOD

400

I1-

300
U
1-

200
100
C

10
15
settlement (mm)

20

25

Figure 4.8a - Load displacement curves from displacement controlled

analyses for a smooth footing on a limited uniform strength/depth
subsoil.
700

x0
Ax5D

600
'500
a.

x=1 OD

400
.1-

m
300
U
.1-

200
100

10
15
settlement (mm)

20

25

Figure 4.8b - Load displacement curves from displacement controlled

analyses for a rough footing on a uniform strength/limited depth subsoil.
62


Collause of Undrained Subsoils

70

-x0
-Ax5D

60
,50

x1 OD

B-

SI..

40

1S
.
0

30

.4I-

S
)

20
10
0*
0

10
15
settlement (mm)

20

25

Figure 4.8c - Load displacement curves from displacement controlled

analyses for a smooth footing on a strength increasing with depth
subsoil.
70

-.x0
-Ax5D
-xlOD

60
50
ag 40
S
I.4ii

30
U
.4I-

> 20
10

10
15
settlement (mm)

20

25

Figure 4.8d - Load displacement curves from displacement controlled

analyses for a rough footing on a strength increasing with depth subsoil.
63

CollaDse of Undrained Subsoils

For soil of uniform strength and limited depth the distribution of total vertical
stress for both the rough and the smooth footings are shown in
non-dimensional form, together with the distribution predicted by plasticity
theory, Figure 4.9. The corresponding distributions for the case of strength
increasing linearly with depth is shown in Figure 4.10.
Load Controlled Analyses
In addition to the displacement controlled analyses, load controlled analyses
were performed in which the ideal plasticity distributions were applied in
increments at the ground surface. The load was applied in 100 equal
increments, i.e. 1 % of the failure load predicted by plasticity theory was
applied in each increment. Only rough cases were examined.
As the applied load reached the plasticity load, the incremental
displacements increased rapidly, Figure 4.11. An attempt was made to
increase the load above that predicted by the plasticity solution by 5%
applied over 50 increments, but the analyses failed due to numerical
problems. The settlements profiles under full plasticity load are shown in
Figure 4.12 for soils of uniform strength and limited depth and soil of
increasing strength with depth.

4.2.4. Effect of the Interface Element

Plasticity theory predicts a displacement discontinuity at the rough
boundaries. At a discontinuity an infinite strain gradient may develop when
the soil yields. In the analyses presented interface elements have been
used at all rough boundaries to model the infinite gradient which develops;
this is achieved when the, zero thickness, interface element reaches its
yield stress and slip of the interface occurs. In order to assess the effect of
the interface element a series of analyses was carried out in which different
meshes were used to analyse the rough footing resting on the idealized soil
of uniform strength over a limited depth, described previously.
64


CollaDse of Undrained Subsoils

25
20
15
In

10
5
01
0

5
x/D

I
10

Figure 4.9 - Comparison of the vertical stress distribution at failure

predicted by plasticity theory and finite element analysis for the uniform
strengtMimited depth Idealized subsoil (values interpolated to nodes
from adjacent elements).
30
25
20
U)

10
5
0

10

/x/SLb

Figure 4.10- Comparison of the vertical stress distribution at failure

predicted by plasticity theory and finite element analysis for the strength
increasing linearly with depth idealized subsoil (values interpolated to
nodes from adjacent elements).
65

Collapse of Undrained Subsoils

100
90
,- 80
,70
260
>.
4C)

4m

a 40
a.

0
U

220
10
0
10

10

15

20

settlement (mm)

Figure 4.1 Ia - Load displacement curve for a load controlled analysis

on a subsoil of uniform strengtMimited depth.

100
90
- 80
.a70
U
260

o 40
a.
U

220
10
0
20

10

10
20
settlemeni (mm)

30

40

Figure 4.11 b - Load displacement curve for a load controlled analysis

on a subsoil with strength increasing linearly with depth.
66

Coltaose of Undrained Subsoils

40
30

?20
E
.10
Ew
.11-

110

10

20
20

10

10

20

30

40

distance from toe (m)

Figure 4.12a - Surface settlement profile at failure for load controlled

analysis of a rough footing on a subsoil with uniform strengtMimited
depth.
40
30
20
110
ES

20
30
I
- 6

I
0

10

12.

distance from toe (m)

Figure 4.12b - Surface settlement profile at failure for load controlled

analysis of a rough footing on a subsoil with strength increasing linearly
with depth.
67

Collapse of Undrained Subsoils

30

25

20
:10
U)

15
10
5
C

10

x/D

Figure 4.13 - Vertical stress distributions at failure predicted using three

different finite element meshes compared with plasticity theory.
Two analyses, in addition to the analysis described previously, were
performed. Firstly, an analysis using the mesh shown in Figure 4.5 but with
no interface elements. Secondly, an analysis in which an extra row of
triangular elements was introduced to the original mesh but, as for the first
additional analysis, no interface elements were used. Each analysis was
performed as described previously and the predicted vertical stress
distributions at failure are presented along with the plasticity theory and the
finite element analysis using interface elements (Figure 4.9) in Figure 4.13.
From Figure 4.13 it can be seen that the interface elements used to model
the upper and lower rough boundaries produce a superior solution. The
error in the predicted stress, at x/D=5, is 4% as compared to an error of
37% at the same point for the mesh without interface elements. Even when
68

CollaDse of Undrained Subsoils

the number of rows of linear strain triangles was increased, from three to
four, the vertical stress distribution at failure still over-predicts the plasticity
solution by a large amount, 24% at xID=5, even though the number of
degrees of freedom is similar to the original mesh including interface
elements.
This result shows that the interface element can be used to analyse,
accurately and economically, the behaviour at rough soil boundaries. The
interface element also allows interface shear and normal stress information
to be extracted easily from the analysis.

4.2.5. Predicted Displacements for a Rough Footing on a

Uniform Strength/Limited Depth Subsoil
Using the method of characteristics it is possible to construct the slip line
mesh for the rigid strip footing on soil of uniform strength and limited depth
(Mandel and Salencon, 1973), Figure 4.14a, from which a hodograph can
be established, Figure 4.14b. In the hodograph heave beyond the toe has
not been shown. As noted earlier the hodograph provides information on
the incremental relative displacement of blocks of soil for an increment of
vertical displacement of the footing.
The incremental lateral displacements, predicted by both plasticity theory
and the finite element analysis, at 2D, 4D, 60 and 80 from the footing toe
are shown in Figure 4.15. The finite element lateral incremental
displacements represent the lateral displacement from increment 90 to 100
predicted using the mesh shown in Figure 4.5 (with additional interface
elements to model the upper rough boundary). At this stage of the analysis
most of the soil beneath the footing is behaving plastically and should
therefore provide an accurate comparison with the plasticity theory. The
predicted incremental lateral displacements from the finite element analysis
and plasticity theory are generally in good agreement.
69

Collapse of Undrained Subsoils

Towards the centreline the finite element displacements are larger than the
plasticity displacements (Figures 4.15b, c and d). This may have been
caused by the rigid area near the centreline in the hodograph, Figure 4.14b,
being larger than that predicted by plasticity theory. This over-prediction of
the extent of the rigid zone is a result of the relatively coarse slip line field
which was used for the graphical solution, Figure 4.14a. If a finer slip field
had been drawn, then the vertex of block 1, Figure 4.14a, would have
coincided with the centreline of the footing, as predicted by plasticity theory.
This would have reduced the extent of the rigid block and increased the
lateral displacements predicted from Figure 4.14b.
At a distance 20 from the toe, Figure 4.14a, the trend reverses and the
plasticity displacements are larger than the finite element displacements.
Two factors may have contributed to this. Firstly, the undrained behaviour is
modelled by using a high value for the bulk modulus of water, but any
non-infinite bulk modulus will result in some volumetric strain which will tend
to produce smaller lateral displacements. This error is cumulative, becoming
more significant with distance from the centreline. Secondly, if some of the
soil was still elastic at increment 90 then a reduced lateral displacement
would result.

4.2.6. Further Analysis using Modified Cam-clay

Finite element solutions using an elastic-perfectly plastic constitutive model
for the soil can be compared directly with plasticity theory. However, this soil
model may not necessarily the best for representing soft clays and in the
analyses in Chapters 6 and 7, modified Cam-clay (Roscoe and Burland,
1968) has been used to model the subsoil. For this reason a comparison
was also made between the load distnbutioi' predicted by plasticity theory
and from a finite element analysis using the modified Cam-clay model to
represent the soil.

73

CollaDse of Undrained Subsoils

0
0
2

I\

4]
S

40

80

120

calculated shear strength (kpa)

over consolidation ratio

stress (kPa)
150

ID

20

30

40

effective
vertical
stress

'\
-- pore water
\ \

pressure

D 81

101
121

\'

Figure 4.16 - Stress and strength profiles used for the modified
Cam-clay analysis.

The analysis was conducted for a rough footing resting on a soil with
strength increasing linearly with depth, the material parameters, using
standard notation, were X=0.25, ic=0.07, f=3.0, M=1.0 and v'0.3. The
predicted stress and strength profiles, Figure 4.16, were defined according
to the equations developed in Appendix A. An error in the strength profile of
Figure 4.16, pointed out by Potts and Ganendra (1991), is discussed below.
The analysis was performed using the mesh shown in Figure 4.6. Additional
interface elements were included at the surface to model the rough
boundary and the interface strength was s 0 =4.25kPa. The undrained shear
strength was calculated to increase at a rate p=2.7OkPaIm. The load
settlement curves for three points at the subsoil surface are shown in Figure
4.17. Failure was defined to have occurred after 600 increments with a
displacement of 0.06m.
At failure the vertical stress distribution beneath the footing can be
compared with the plasticity solution, Figure 4.18. The prediction made
using the modified Cam-clay model is in close agreement with the plasticity
74

Collapse of Undrained Subsoils

90
80
,70
U
0

.60
j50
m 40
U
C,

30

> 20
10
0

0.1

0.2

0.3
0.4
settlement (mm)

0.5

0.6

Figure 4.17 - Load settlement curves for a displacement controlled

analysis using modified Cam-clay to model a subsoil with strength
increasing linearly with depth.

25
20
,15
10
5
0
0

4
ox/Su

Figure 4.18 - Comparison of finite element and plasticity solutions for

the stress distribution at failure beneath a rough footing on a subsoil
with strength increasing linearly with depth (modified Cam-clay).
75

Collapse of Undrained Subsoils

solution. Therefore, if modified Cam-clay constitutive model is selected in

order to improve the modelling of the pre-failure behaviour, the prediction of
collapse need not be adversely affected.
The number of increments used has increased by six times when compared
with the equivalent elastic-perfectly plastic analysis, Section 4.2.2. Only a
limited study was carried out to investigate the effect of the number of
increments. However, the number of increments used is similar to that used
to predict failure of a footing problem using the modified Cam-clay
constitutive model with an incremental tangent stiffness solution (Potts et al,
1990).
Error in the Undrained Shear Strength Profile
As pointed out by Potts and Ganendra (1991), in their discussion of Hird, et
al (1990), the assumption of a linear undrained shear strength distribution,
shown in Figure 4.16 is not strictly correct. The correct profile is defined by
Equation 4.1 which is non-linear
s=(2.3oz+5.5o)(^)

.....................................(4.1)

The two predictions of the undrained shear strength profile, Figure 4.16 and
Equation 4.1, are shown in Figure 4.19. Also shown is the percentage error
between the two predictions. The maximum error is 22% at the surface and
rapidly decreases to less than 5% beyond 2m below ground level.
A best fit line through the profile predicted by Potts and Ganendra produces
values of s=4.13kPa and p2.6lkPa/m (compared with the values
s=4.25kPa and p=2.7OkPaIm used in Figure 4.18). If these values are
used to replot the data in Figure 4.18 a negligible difference occurs in the
position of the data points.
The effect of the over-large interface strength (4.25kPa rather than
3.49kPa) appears to have had little effect on the final result The analysis
76

Collapse of Undrained Subsoils

4
E
1a.

0
8

10

12
0

10

15

20

25

30

35

40

Undrained eheor etrenath (kPa) or Error (%)

Figure 4.19 - Error in the undrained shear strength profile used for the
modified Cam-clay analysis (after Potts and Ganendra, 1991).

showed that slip occurred on the interface beneath the loaded area even
though the shear strength was lower at some integration points in
neighbouring soil elements. This implies that yielding was constrained in
those elements to some degree due to the discretization error at the
surface.

4.3. Embankments with Constant Side Slope

Jewell (1988) has suggested that the design of an embankment built on
either of the subsoils shown in Figure 4.3 can make use of the plasticity
solutions to calculate the embankment profile

for

a single lift construction.

Using the strength parameters defined in Figure 4.3, if an embankment is to

be constructed to a design height Hd with a fill of unit weight y and a factor of
safety FS, then the vertical axis in Figure 4.4, cYJs, becomes FSyHIs0,
77

CollaDse of Undrained Subsoils

uniform strength/limited depth

strength increasing with depth

FSyH
suo

FS-yH
SUD

design curve

px/Suo

x/D
design curve

(a)
H fill

(b)

Hfill

Hd

Hd

7-

x
practical crosssection

(c)

Xd I

x
practical crosssection

(d)
constant side, slope
practical crosssection

Figure 4.20 - Proposed practical embankment shape to approximate

ideal loading predicted by plasticity theory (after: Jewell, 1988) and a
proposed simplified constant side slope design profile.

78

CollaDse of Undrained Subsoils

Figure 4.20a and b. The plasticity curve can then be used to define the
minimum safe single stage side slope length, Xd, Figures 4.20a and b, and
cross section, Figures 4.20c and d.
The embankment shape which corresponds to plasticity theory involves a
uniform slope within the plasticity region (x>O) and a stable, steeply sloping
wedge beyond (x<O). Jewell (1988) compared limit equilibrium analyses of
the plasticity cross sections (Figures 4.20c and d) with the p'asticity
solutions. Limit equilibrium analysis of subsoil with strength increasing
linearly with depth were in good agreement with the plasticity curve but limit
equilibrium analysis of the uniform strength and limited depth case
exceeded the plasticity distribution by a large amount, this difference is
discussed in Section 4.3.2. The embankment shape suggested by the
plasticity cross section would be difficult to construct and would take
additional land due to the triangle of soil required to avoid a vertical soil face
at the toe.
Also shown on Figures 4.20c and cj is a proposed embankment profile with
a constant side slope which starts at the toe (x=O) of the plasticity solution
and at a distance Xd intersects the plasticity profile at the required design
height, Hd. This profile would be easier to construct and would require less
land than the plasticity cross section suggested by Jewell (1988). In order to
investigate the proposed constant side slope profile a series of finite
element analyses was carried out
Finite element analyses were performed for fully reinforced embankments
with constant side slopes on both types of idealized subsoil. The results of
these analyses were compared with the plasticity solution to assess the
validity of a design method using the plasticity distribution to predict the
slope length of a constant side slope embankment.

79

CollaDse of Undrained Subsoils

4.3.1. Finite Element Analyses of Constant Side Slope

Loading
A series of load controlled analyses was carried out in order to investigate
the possibility of using the plasticity solutions in the design of constant side
slope embankments. The meshes shown in Figures 4.5 and 4.6, with
additional interface elements at the surface, were used with the
embankment fill represented by vertical loads. The subsoil was modelled as
undrained using an elastic-perfectly plastic model with the same parameters
as described in Section 4.2.2. The upper nodes of the surface interface
elements were restracned n the t'orizonta diTection Thus moOeng the hilly
reinforced case.
Analyses were carried out using five values of the design length of the side
slope, x., for each of the two idealized subsoils. The procedure for
calculating the applied loading was as follows. The side slope length, Xi,
was chosen. The position on the abscissa of Figures 4.20a and b was
therefore known. The embankment loading at Xd was found from the
plasticity solutions, using a factor of safety of unity, as aV='yHd. The constant
side slope loading was defined as increasing uniformly from the toe (c=O at
x=O) to the crest (;=yH at x=xd ) using vertical loads applied to the top of
the interface elements.
The uniform strength/limited depth mesh, Figure 4.5, has a footing half
width of I OD. For the constant side slope analyses the ratios of xID were I,
3, 5, 7 and 9. The strength increasing linearly with depth mesh, Figure 4.6,
has a footing half width of 9.6p/s 0. For the constant side slope analyses the
ratios of Xdp/S were 1.6, 3.2, 4.8, 6.4 and 8.0.
In each analysis the crest position remained constant and the vertical loads
were increased proportionally across the loaded area.

80

Collanse of Undrained Subsoils

450
400
350
u 300
0
0
U

250
200
150
100
50
0
350

50
250 150 50
settlement (mm)

150

250

350

Figure 4.21 - Typical load displacement curve for a constant side slope
analysis. Curve shown for a uniform strength limited depth subsoil
analysis with XdID=5.

4.3.2. Results
The failure of the analysis was assessed from load-displacement curves for
three points on the subsoil surface; these were located beneath the toe, the
crest and the centreline. Typical curves are presented in Figure 4.21.
Clearly defined failure loads could be assessed for all analyses as the load
acting at the increment after which large displacements occurred at one or
more of the observed points.
For each of these analyses the vertical stress applied beneath the crest was
normalised with s and plotted against the plasticity solution, Figure 4.22
and Figure 4.23 for the uniform strength and limited depth case and the
strength increasing linearly with depth case respectively.
81


CoIIaDse of Undrained Subsoils

25
20
.15
(I)

10
5
0

6
5
4
x/D or xd/D

10

Figure 4.22 - Comparison of the predicted collapse load of a constant

side slope embankment with plasticity theory for a subsoil with uniform
strength and limited depth.

30
25
20
15
10
5
0

- Plasticity theory
0

Constant side slope

6

10

px/S or pxJS

Figure 4.23 - Comparison of the predicted collapse load of a constant

side slope embankment with plasticity theory for a subsoil with strength
increasing linearly with depth.
82

CollaDse of Undrained Subsoils

The finite element predictions of the undrained failure of embankments with

constant side slopes compare well with the profiles derived from plasticity
theory. The vertical stress distribution from plasticity theory may be used to
obtain a preliminary assessment of the length of slope required to build an
embankment to the design height in a single lift.

4.3.3. A Comparison of Limit Equilibrium and Finite Element

Analysis for Embankments with Constant Side Slope
Jewell (1988) compared limit equilibrium analyses with plasticity solutions
for the two idealized subsoils. The slip circle analyses gave good agreement
with the plasticity solution for the case of strength increasing linearly with
depth, but the analyses of the uniform strength and limited depth subsoil
over-predicted the stability by a large amount.
Using the finite element predicted displacements, the displacement vectors
at failure can be produced, and an indication as to the mode of failure of the
subsoil may be obtained. Displacement vectors at failure are plotted for
three of the constant side slope analyses in Figures 4.24 and 4.25. The
embankments shown indicate their relative size and position for each
analysis.
The displacement vectors for the soil with strength increasing linearly with
depth, Figure 4.25, show a rotational mechanism, thus confirming that a slip
circle analysis is applicable in this case. The displacement vectors at failure
for the uniform strength and limited depth case, Figure 4.24, show different
mechanisms. When the side slope is very short, relative to the subsoil
depth, a rotational mechanism can still develop, Figure 4.24a. However, as
the side slope length becomes greater the subsoil is too shallow for a
rotational mechanism to develop and the subsoil then fails with a
translational mechanism with small rotational zones at either end, Figures
4.24b and c. For such a mechanism a slip circle analysis is no longer
83

Collapse of Undrained Subsoils

-....

Figure 4.24a - Displacement vectors at failure for a constant side slope

embankment on a subsoil of uniform strength and limited depth;
xdIDl.

- -

Figure 4.24b - Displacement vectors at failure for a constant side slope

embankment on a subsoil of uniform strength and limited depth;
Xd/D5.

Figure 4.24c - Displacement vectors at failure for a constant side slope

embankment on a subsoil of uniform strength and limited depth;
xdID=9.

84

Collapse of Undrained Subsoits

Figure 4.25a - Displacement vectors at failure for a constant side slope

embankment on a subsoil with strength increasing linearly with depth;
pxjs 0 1 .6.

Figure 4.25b - Displacement vectors at failure for a constant side slope

embankment on a subsoil with strength increasing linearly with depth;
PXdI5uo4.8.

85

CollaDse of Undrained Subsoils

Figure 4.25c - Displacement vectors at failure for a constant side slope

embankment on a subsoil with strength increasing linearly with depth;
pxJs0=8.O
appropriate, as observed by Jewell (1988).
In Figure 4.26 and Figure 4.27 the maximum shear strain contours at failure
for two of the constant side slope analyses are shown. The regions of high
shear strain indicate zones in which a slip surface is likely to develop. The
intense shear gradients which are modelled accurately using the interface
elements cannot be shown on the contour plot as the interface element has
no thickness. These maximum shear strain contours further illustrate the
failure mechanisms indicated in Figures 4.24 and 4.25.

4.4. Summary
A modified version of the finite element program CRISP has successfully
predicted the undrained collapse of a strip footing on two idealized soil
profiles using an elastic-perfectly plastic constitutive model. Both rough and
86

Collap se of Undrained Subsoils

Figure 4.26 - Maximum shear strain (%) contours at failure for an

embankment with constant side slope on a subsoil of uniform strength
and limited depth (x/D=5)

Figure 4.27 - Maximum shear strain (%) contours at failure for an

embankment with constant side slope on a subsoil with strength
increasing linearly with depth (pxjs04.8)

87

Collaose of Undrained Subsoits

smooth footings were analysed representing different degrees of base

restraint in a reinforced embankment. Predictions of failure loads agreed
closely with plasticity theory for displacement controlled analyses, Figures
4.9 and 4.10.
Load controlled failures, which are more relevant to the construction of
reinforced embankments, have been shown to be predictable using
plasticity theory although the surface settlement profiles from finite element
analyses and plasticity theory are different, Figure 4.12.
A displacement controlled analysis was performed modelling the subsoil
using the modified Cam-clay constitutive model and having a profile of
strength increasing, almost linearly, with depth. Comparison of finite
element analysis and plasticity theory shows that, provided a sufficient
number of increments are used, undrained failure, using the modified
Cam-clay constitutive model, can be predicted accurately, Figure 4.18.
The lateral displacements predicted using finite element analysis for a
subsoil of uniform strength and limited depth are in good agreement with
the displacements predicted by plasticity theory, Figure 4.15. The
comparisons were made for incremental lateral displacements, close to
collapse, in a displacement controlled analysis.
A series of load controlled analyses of fully reinforced embankments with
constant side slopes overlying the two idealized soil profiles was performed.
From these analyses a simple design method, based on the plasticity
solutions, which predicts the length of the embankment side slope for a
required design height has been validated, Figures 4.22 and 4.23.
Both translational and rotational failure mechanisms have been observed in
the finite element analyses, Figures 4.24 and 4.25. The type of mechanism
which develops is dependent on the depth of the soft subsoil compared to
the length of the embankment side slope. When performing a limit
88

Collapse of Undrained Subsoils

equilibrium analysis it is essential that the correct type of mechanism is

analysed. Failure to do so may lead to a serious over-estimate in the
stability of the embankment (Jewell, 1988).
In all finite element analyses presented interface elements have been used
to model the soil interlace with rough boundaries. The interface was
specified as having an undrained shear strength equal to that of the
adjacent soil. The interface element has the capability of modelling
accurately the discontinuity which develops at the rough boundary and
produces superior predictions of failure loads compared to meshes in which
interface elements are not used, Figure 4.13. The interface element also
allows shear an normal stress information to be extracted easily from the
finite element analysis.

89

Consolidation Around a Sin gle Vertical Drain

5. Consolidation Around a Single

Vertical Drain
5.1. Introduction
The installation of vertical drains in soft ground prior to embankment
construction has become a well used construction technique since the
invention of prefabricated band drains (Kjelleman, 1948) and their
subsequent development. The drains increase stability by speeding up
consolidation and therefore allow embankment construction to progress
more quickly.
Vertical drains are most likely to be needed in clay soils with little or no
fabric of the kind that facilitates natural drainage (Rowe, 1972), but they can
also be effective in laminated or layered soils, where the permeability of the
soil is markedly anisotropic (Home, 1964). The quantification of mass
permeability is widely recognised as one of the most difficult aspects of
geotechnical design (Jamiolkowski et al, 1983). Perhaps because of the
uncertainty associated with the permeability parameters, designers have
apparently been content, until now, to rely on simplified analytical methods
to quantify the effects of the drains. The most widely used methods (Barron,
1948; Hansbo, 1981) assume that consolidation takes place in a uniform
soil column with linear compressibility characteristics in the absence of
lateral movement. Such restrictive conditions are not likely to be realized in
normally or lightly over-consolidated soils under embankment loading. The
finite element method offers a less restrictive analysis which is also capable
of incorporating the effects of reinforcement and stage construction.
Furthermore, finite element analysis leads directly to the prediction of
deformations and eliminates the need to extract information from a
90

Consolidation Around a Sin gle Vertical Drain

3D view

Plan View
arrows indicate
direction of flow

(c>\

(a

-elk-'II--II-(b

Figure 5.1 - Unit cell and direction of flow: (a) axisymmetric conditions
(b) plane strain conditions.
consolidation analysis for subsequent use in a stability calculation. Despite
its merits, however, the finite element analysis of vertical drains has only
occasionally been reported (Zeng et al, 1987; Kumamoto et al, 1988;
Sanchez and Sagaseta, 1990).
In this Chapter a methodology for the finite element analysis of ground
incorporating vertical drains is developed. An obvious difficulty is that most
numerical analyses of embankment problems are conducted for plane strain
conditions, whereas the soil around an individual vertical drain is more
appropriately modelled as axisymmetric, Figure 5.1. It is necessary,
91

Consolidation Around a Sin gle Vertical Drain

Applied stress

NL

Drain
Smear zone

Figure 5.2 - Unit cell adopted for analytical solution (After: Hansbo,
1981).
therefore, to find a logical basis for the plane strain analysis. As previous
approaches to this problem have left scope for improvement, Section 2.2.2,
the equivalence of plane strain and axisymmetric consolidation has been
re-examined and a new matching procedure developed.
92

Consolidation Around a Sin gle Vertical Drain

The numerical analyses have been performed using the modified version of
CRISP, Chapter 3, using the drainage element, Section 3.4, to model the
vertical drains. Validation of the analysis was undertaken in two stages.
Firstly, the results of axisymmetric analyses were compared with Hansbo's
(1981) closed form solutions for consolidation around a single vertical drain.
Secondly, comparative plane strain and axisymmetric analyses were
conducted to test the matching procedure under the simplified conditions
assumed in its derivation. In Chapter 6 the matching procedure is tested
under more realistic ground conditions.

5.2. Comparison of Finite Element Analysis with

a Closed Form Solution
Analytical solutions for the consolidation of soil by vertical drains have
invariably involved the study of an axisymmetric unit cell, i.e. a cylinder of
soil around a single vertical drain (Figure 5.la), under simplified boundary
conditions. Figure 5.2 shows a unit cell of fixed external radius R and initial
length I containing a drain of radius r. The impervious bottom boundary is
not free to move vertically, while the downward movement of the upper
boundary may be permitted to develop freely under a constant applied
stress (free strain) or may be constrained to be uniform (equal strain); the
assumption of equal strain is usually adopted for mathematical
convenience. As it is compressed, the soil exhibits uniform stress-strain
behaviour. Drainage within the soil may take place both vertically and
radially, although it is often reasonable to assume that radial flow
dominates. The permeability of the soil, k, may reduce to a lower value, lc,
within a smeared zone of radius r 5 caused by drain installation. Flow, in the
drain takes place vertically towards the top and is governed by the
permeability, k, and radius of the drain.

93

Consolidation Around a Sin gle Vertical Drain

The available theories vary in their degrees of rigour and complexity. The
most rigourous, but mathematically most complex, solutions are based on
Blot's (1941) consolidation theory and were derived for equal strain
boundary conditions by Yoshikuni and Nakando (1974) and Onoue (1988).
Both solutions allow for combined vertical and radial drainage in the soil and
finite drain permeability (well resistance), but only Onoue's solution
considers smear. Although the upper boundary is displaced uniformly,
variation of vertical strain is permitted with both radius and depth. Less
rigourous solutions, allowing for both well resistance and smear but
neglecting vertical flow in the soil, have been obtained by Barron (1948),
Hansbo (1981) and Zeng and Xie (1989). These are genuine equal strain
analyses where the vertical strain is assumed to be uniform with radius and
depth; the latter is clearly an approximation if well resistance is significant.
The solutions of Hansbo and Zeng and Xie are relatively easy to compute
compared with either Barron's solution or the more rigourous equal strain
solution. Solutions for free strain boundary conditions have been obtained in
the absence of well resistance (Barron, 1948) and even then tend to be
complex.
It has been shown (Hansbo, 1981; Zeng and Xie, 1989; Onoue, 1988) that
Hansbo's relatively simple theory compares well, not only with the
philosophically similar solutions of Barron and Zeng and Xie, but also with
more rigourous solutions. Being also simple to apply, the theory has gained
wide acceptance and has been used in parametric studies (Jamiolkowski et
al, 1983). It has therefore been used to validate the finite element analysis.
Under an instantaneous step loading the average degree of consolidation
based on pore pressure, tih , on a horizontal plane at depth z and at time t is
predicted by Hansbo to be
Vh=1_eTI41(5.1)

94

Consolidation Around a Sin g le Vertical Drain

where Th is the time factor for horizontal drainage (c,,tI4R2) and

i.t=In(nls) +(wk3) 3I4+z(2I_z)Wkr ...........................(5.2)
in which n = RIr and s=rJr.
Several groups of axisymmetric finite element analyses were carried out to
obtain the results for comparison with Hansbo's theory. The mesh and
boundary conditions for the first group of analyses, involving well resistance
but no smear, are given in Figure 5.3. Loads were applied to the upper
boundary of the mesh under a free strain rather than an equal strain
condition, although, as will be shown later, the displacement of the upper
boundary was practically uniform. Hansbo incorporated linear stress-strain
behaviour of the soil by assuming a constant constrained modulus in the
calculation of ch; in the finite element analyses linear elastic properties were
specified, E'=l O4kN/m2, and u'=O. Hansbo's neglect of vertical flow in the
soil was matched by setting its vertical permeability to zero. The horizontal
permeability was taken to be I OmIs. The drain was assumed to possess
negligible stiffness, but it's finite permeability was varied to give different
values of discharge capacity, q=kitr2 , and hence different values of the
dimensionless parameter introduced by Yoshikuni and Nakando (1974),
L=8kI2/mq. At any particular stage in the analysis, the average degree of
consolidation at the base of the drain could be evaluated by performing a
simple numerical integration (trapezoidal rule) of the nodal pore pressures
along the base of the mesh. The results are compared with those calculated
by Hansbo for Th<2 in Figure 5.4. It can be seen that the agreement is
generally excellent. Also shown are almost identical results obtained by
Yoshikuni and Nakando (1974).
Three more groups of analyses were conducted, with the aim of duplicating
the parametric study reported by Jamiolkowski et al (1983). The effect of
well resistance in the absence of smear was again investigated but a
different mesh, Figure 5.5a, was used so that the geometry of the unit cell
95

Consolidation Around a Sin gle Vertical Drain

orizontally
estrained,
rn pervious

hon
resi
dral
eler
dral
bou

i.Om

I .JFE1

tica Ily
restrained,
imoervious

Figure 5.3 - Finite element mesh used for unit cell analyses comparison
with Hansbo (1981).

would be consistent with that adopted in the other studies. Assumptions

regarding the boundary conditions and material parameters remained
unchanged. In the next group of analyses, the effect of the drain length was
investigated for fixed values of the drain permeability. The variation of the
drain length was achieved by shrinking or stretching the mesh in Figure
5.5a in the vertical direction so that R/l varied between 0.2 and 1.0. Finally,
the effect of the smear zone was studied in the absence of well resistance
96

Consolidation Around a Sin g le Vertical Drain

100
90
c8
c 70

0
1U

-a

0
VI

c 50
0
(3

40
30
20
10
0
0.01

0.1

10

time factor

Figure 5.4a - Comparison of finite element and analytical results for

consolidation of a unit cell : L=O.O
100
90
#-' 80
c 70
0

a
-o
0
11

c 50
0
0

40
30
20
10
0
0.01

0.1

10

time factor

Figure 5.4b - Comparison of finite element and analytical results for

consolidation of a unit cell : L=O.5
97

Consolidation Around a Sin gle Vertical Drain

100
finite element
--Honsbo
-*Yoshilcuni

90

c 50
0
U

.40
30
20
10
0
0.01

0.1

'U

time factor

Figure 5.4c - Comparison of finite element and analytical results for

consolidation of a unit cell : L=3.O

100
-R-

B0

finite element
-Hansbo

70

Yoshikuril

90

60
C
0
U

1-

50

40
L

0)
.

30
20
10+-
0.01

0.1

'U

time factor

Figure 5.4d - Comparison of finite element and analytical results for

consolidation of a unit cell : L=5.O
98

Consolidation Around a Sin gle Vertical Drain

(a)

fixed in x
fixed
drainc
eleme
or dn
bound

=20.Om

___] in y
R=1O.Om

(b)

WI/V 1,1L'LL-L--1 $ l=O.5m

I' r=1.Om
I

R=1O.Om

Figure 5.5 - Finite element meshes used for unit cell analyses:
comparisons with Jamiolkowski et al (1983).

99

Consolidation Around a SinQie Vertical Drain

for a fixed drain spacing. In this case the analysis could be conducted more
economically by using a single row of elements, Figure 5.5b, whose upper
and lower boundaries were subject to the same conditions as those applied
to the corresponding boundaries of the unit cell. To model the effect of
smear, soil elements near the drain were reduced in size and given a
permeability 10 times lower than the main body of soil (i.e. k/k5=10). An
increase in the drain radius was also used to provide a value of s=rJr in the
desired range.
Results for these three sets of predictions are presented alongside the
theoretical predictions in Figures 5.6, 5.7 and 5.8. In each figure the time
factor for 90% consolidation at mid-depth of the drain, Th , is expressed as
a ratio of the corresponding theoretical factor for an ideal case in which
there is no smear and no well resistance. Once again the agreement
between the finite element analyses and the theoretical results is generally
excellent. The trends shown in Figures 5.6, 5.7 and 5.8 are in close
agreement with those shown by Jamiolkowski et al (1983). This is true
despite a subtle difference of approach in the latter case, namely that in
order to establish ThgQ the average degree of consolidation was not just
evaluated at the mid-depth but over its entire length.
As noted above, the finite element results were obtained under free strain
conditions yet have been compared with an equal strain theory. It is of
interest to plot the displacements of the upper boundary of the mesh to
examine how far the boundary conditions for the finite element analyses
deviate from the theoretical ones. The development of the surface
displacement with time for three representative analyses is given in Figure
5.9; the settlements are expressed as proportions of the final settlement,
which itself is uniform across the unit cell. The only significant
non-uniformity occurs temporarily above the unusually thick smear zone in
Figure 5.9c. The results explain why very similar solutions were obtained for
free and equal strain conditions by Barron (1948).
100

Consolidation Around a Sinale Vertical Drain

4
3.5
3
U
0
01
I-

25

0
0i

0.5

1000 2000 3000 4000 5000 6000 7000 8000 900010000

discharge capacIty/soil permeability

Figure 5.6 - Effect of well resistance on rate of consolidation.

B
7
6
a
-05
0
014

03
01
I-

2
1
0

10

15

20
25 30 35
drain length (m)

40 45

50

Figure 5.7 - Effect of drain length on rate of consolidation.

101

Consolidation Around a Sin gle Vertical Drain

B
7

2
1

1.2

1.4

1.6
1.8
2
2.2 2.4
smear zone radius/drain radius s

2.6

2.8

Figure 5.8 - Effect of smear on rate of consolidation.

The spatial variation of the vertical strain within the unit cell, neglected by
Hansbo, is also readily obtainable from the finite element analyses. Figure
5.10 shows the variation obtained on the periphery of the unit cell at various
stages of an analysis with high well resistance, when the largest variation of
strain might be anticipated. The strain is expressed as a proportion of the
final strain, which was uniform with depth. It is demonstrated that the
variation of strain can be significant, although its effect on the overall rate of
settlement, as given by Hansbo's solution, appears small.

102

Consolidation Around a Sin gle Vertical Drain

0

-10

Th=0.0025
-Th=0.025

g-2o
C
1-

m-

Th=0.1 25

-9-

Th=0.5
-Th=2.5

-50
C

-60

1:;:
-90
-100

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

r/R

Figure 5.9a - Development of surface settlement profile in finite element

analyses without smear or well resistance.
0
Th=0.0025

-10
g-20

Th=0.025

4-

C
S-

Th=0.1 25

-9-

-40

Th=0.5

-50
Th=2.5
-ATh=1 2.5

-70
-80
-90
-100
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
r/R

Figure 5.9b - Development of surface settlement profile in finite element

analyses without smear but with well resistance (qjk=150m2).
103

Consolidation Around a Sinile Vertical Drain

0
Th=0.0025
-iTh0.025

-10

TIi=0.1 25
-aTh=0.5

-40
I,
.
C

-50

-N-

Th=2.5

-60

Th=12.5

-90
-100
U U.1 U.Z U..

0.4 0.5 0.6 0.7 0.8 0.9

r/R

Figure 5.9c - Development of surface settlement profile in finite element

analysis with smear (s=3.0) but without well resistance.

5.3. Matching Procedure for Plane Strain

Analyses
Previous approaches for modelling the effect of vertical drains in plane
strain analyses (Zeng et al, 1987; Sanchez and Sagaseta, 1990) have
involved matching the time taken for a given degree of consolidation to be
achieved by horizontal drainage under plane strain and axisymmetric
conditions. Thus neglecting the effect of well resistance,
Th,,, B2
pi

- ThaxR2
ax

where 2B is the drain spacing in plain strain and the subscripts (or subscript
extensions) p1 and ax identify the plane strain or axisymmetric conditions
respectively. Either or both the drain spacing and the soil permeability may
be manipulated to satisfy Equation 5.3. Unfortunately, the resulting values
do not then apply for other than the chosen degree of consolidation, since
104

Consolidation Around a Sinale Vertical Drain

a
Th=O.0025
-Th=0.025

-0.1
-0.2

Th=O.1 25
-aTh=0.5

-0.3
-0.4

Th=2.5
-ATh12.5

-0.6
-0.7
-0.8
-0.9

10 20 30 40 50 60 70 80 90 100
vertical strain/final vertical strain (%)

Figure 5.10 - Variation of vertical strain with depth at unit cell periphery
in finite element analysis with well resistance (q/k=150m2).

the ratio of Th,JT,,, is not constant. Furthermore, in the presence of well

resistance, the degree of consolidation varies with depth. In view of the
generally good agreement between Hansbo's (1981) theory and finite
element analyses seen in the previous section, a better basis for matching
plane strain and axisymmetric unit cells was developed by adapting
Hansbo's theory for plane strain.
Consider a plane strain unit cell of half width B containing a drain with
discharge capacity per unit length. For matching purposes the drain is
assumed to possess negligible thickness and no smear zone. As shown in
Appendix B, it is possible to apply Equation 5.1 with Thcht/4B2 and
IL=

(2I3)+2z(2l-z)(kJBQ). For the rate of consolidation to be matched in the

plane strain and axisymmetric unit cells, equality of the average degree of
consolidation at every time and at every level in the cell is required. Hence
105

Consolidation Around a Sin gle Vertical Drain

(5.4)
from Equation 5. I

IELI
II hpI

(55

J.Lax

or
= C h t

(5 6

B2 .Lpj R2.Lax

Two types of matching are defined, firstly, geometry matching in which the
permeability of the soil is the same in the plane strain and axisymmetric
analyses and the width of the plane strain unit cell is adjusted. If the soil
parameters are identical in the axisymmetric and plane strain unit cell, then
Equation 5.6 becomes
B2 gg=R2 .t

........................................................................(5.7)

The relevant expressions for

and

may be substituted into Equation

5.7 and the terms rearranged to give

B2_R2[ln(.)

^*ln(s)_]

= (R2_-_2!(2!z_z2)

q., Q)

(5.8)

The condition for geometric matching may be obtained by considering the

case of negligible well resistance (qw and Q , cc). Therefore, geometric
matching, including the effect of smear, is achieved if
_R2[ln()+tln(s)_]=0

........................................(5.9)

or

1In()+*ln(s)_

............................................... (5:10)

The effect of well resistance is matched independently if

.............................(5.11)

106

Consolidation Around a Sing le Vertical Drain

or
Qww2j .......................................................................(5.12)
A second procedure, referred to as permeability matching, is to maintain the
same drain spacing in the plane strain and axisymmetric analyses and to
adjust the plane strain horizontal soil permeability to achieve a matched
analysis. From Equation 5.6 with B=R
.................................................(5.13)
substituting for the factors

and i.La(

2k
3[ln()+.ln(s)_..]
and
Qw = ..........................................................................( 5.15)

Although geometry and permeability matching procedures have been

described separately, they can be used in combination by setting B to a
desired value and re-deriving the requirements for k, and This is useful
in avoiding the necessity of modelling small equivalent drain spacings with
excessively large numbers of finite elements in a full plane strain analysis.
5.3.1. Validation of the Matching Procedure
In order to validate the above theory, finite element analyses were
undertaken for both an axisymmetric unit cell and the equivalent plane
strain case, determined with the aid of Equations 5.10 and 5.12 or 5.14 and
5.15. The effects of both well resistance and smear were studied. As the
matching theory considers only horizontal drainage in the soil, the vertical
permeability of the soil was initially neglected. The mesh of Figure 5.5a was
again used, stretched if necessary in the horizontal plane to achieve
geometric matching. The stretching was done non-uniformly in such a way
107

Consolidation Around a Sinule Vertical Drain

that soil elements near the drain remained sufficiently small. The previously
adopted values of the elastic parameters (E' and u') and horizontal
permeability of the soil were retained, except when permeability was
changed in accordance with the matching rules. As before the drain was
assumed to possess no stiffness. In analyses with smear, the size of the
smear zone and its permeability were fixed so that s=5 and kII=2. In the
analyses with well resistance, the discharge capacity was chosen to give
L=2.
The equivalence of these plane strain and axisymmetric analyses can be
assessed on the basis of both settlement and pore pressure. Results for
the overall average degree of consolidation, Ti, calculated from the
settlements of the surface nodes are given in Figure 5.11. The style of
calculation was the same as that used to compute Ti,, at a given depth from
nodal pore pressures as described previously. The corresponding results for
tih at the mid-depth of the drain are shown in Figure 5.12. For both Ti and
Uh it

can be seen that the geometry and permeability matching procedures

produce almost identical results. However, in the absence of well resistance

(Figures 5.11a, 5.11b, 5.12a and 5.12b) there are noticeable differences
between the plane strain and axisymmetric results, especially at small time
factors. When well resistance is significant (Figures 5.1 ic, 5.1 lb, 5.12c and
5.12d) better overall matching is achieved.
The matching errors displayed in Figures. 5.11 and 5.12 may be assessed,
for geometric matching, by referring to Figures. 5.13a and 5.13b. The errors
which are presented as differences in the degree of consolidation achieved
at a given time factor, are probably attributable to the displacements at the
upper boundary. For example, Figure 5.14a shows the displacement
profiles at various stages of a matched analysis without smear or well
resistance. In this plane strain analysis the settlement profile exhibits a
much larger variation across the unit cell than in the corresponding
108

Consolidation Around a Sin gle Vertical Drain

0
10
Axisymmetric

Permeability matched

60
S
!

Geometry matched

-70

o 80
90
100

0.01

0.1

IU

Time factor Th

Figure 5.1 Ia - Comparison of the average surface settlement for

axisymmetric and matched plane strain analyses : without smear or well
resistance.
0
-10

i :
Permeability matched

50
-60
S

-70

Geometry matched

0 80
90
100
o.01

10

0.1
Time factor Th

Figure 5.1 lb - Comparison of the average surface settlement for

axisymmetric and matched plane strain analyses : with smear but
without well resistance.
109

Consolidation Around a Sin gle Vertical Drain

0
-10

Axisymmetric

Permeability matched

-60
0

f -70

Geometry matched

-80
-90
-100
0.01

0.1

I',

Time factor Th

Figure 5.1 Ic - Comparison of the average surface settlement for

axisymmetric and matched plane strain analyses : without smear but
with well resistance.
0
-10
c,-20
::

Axisymmetric

Permeability matched

-50
-60
Geometry matched

1:;:
-90
-100
0.01

0.1

'U

Time factor Th

Figure 5.11d - Comparison of the average surface settlement for

axisymmetric and matched plane strain analyses : with smear and well
resistance.
110

Consolidation Around a Sin gle Vertical Drain

100
90

Axisymmetric

70
60

matched

50

40
U

Geometry matched

30

0)
U

20
10

a
b.0 1

0.1

'U

Time factor Th

Figure 5.12a - Comparison of the average excess pore pressure at

mid-depth for axisymmetric and matched plane strain analyses : without
smear or well resistance.
100
90
g8
Axisymmetric

70
-a 60

Permeability matched

50

40
30

Geometry matched

DI

20
10
C
O.01

0.1

IU

Time factor Th

Figure 5.12b - Comparison of the average excess pore pressure at

mid-depth for axisymmetric and matched plane strain analyses : with
smear but without well resistance.
111

Consolidation Around a SinQie Vertical Drain

100
90
80
Axisymmefric
70
matched
a 60
50

40

Geometry matched

! 30
Dl

o 20
10
0
0.01

0.1

'U

lime factor Tb

Figure 5.12c - Comparison of the average excess pore pressure at

mid-depth for axisymmetric and matched plane strain analyses : without
smear but with well resistance.
100
90
g8

Axisymrnetric
matched

50

40

Geometry matched

30
o 20
10
0
0.01

0.1

10
Time factor Th

Figure 5.12d - Comparison of the average excess pore pressure at

mid-depth for axisymmetric and matched plane strain analyses : with
smear and well resistance.
112

Consolidation Around a Single Vertical Drain

10

F-'

'-F

x
a

..

10

15
0.01

0.1

10

TImB factor Th

Figure 5.13a - Errors in geometry matched plane strain analyses:

based on settlements; results of Figure 5.11.

10

F'
'-F

x
a

10

15
0.01

0.1

10

Time factor Th

Figure 5.13b - Errors in geometry matched plane strain analyses:

based on pore pressure; results of Figure 5.12.
113

Consolidation Around a Single Vertical Drain

axisymmetric analysis (Figure 5.9a) and clearly violates the equal strain
assumption on which the matching theory is based. The improvement of the
matching procedure obtained with well resistance, most marked at small
time factors, corresponds to the attainment of slightly more uniform
settlement profiles, Figure 5.14b. Fortunately the differential settlements
seen in these unit cell analyses are less likely to occur in full embankment
analyses, because of the stiffness contribution of the embankment
elements, and superior matching can be anticipated.

5.3.2. Pore Pressure Variation Across Unit Cells

It is important to realize that, even if perfect matching is achieved, the
excess pore pressures at corresponding points in the axisymmetric and
plane strain unit cells will not be the same. This is illustrated in Figure 5.15
for a pair of analyses in which the difference in 1 h was negligible. In
particular the maximum values of the excess pore pressures on the
periphery of the unit cell (midway between drains) differ significantly.
The ratio of the pore pressures across the axisymmetric and matched unit
cells (uJu) has been derived in Appendix B, and at the periphery of the
unit cells is
u

21n(n)-1

.....................(5.15)

- 3(ln(n)_)
From Equation 5.15 the ratio of the pore pressures at the periphery of the
test unit cell. Figure 5.5, is 0.72. This compares well with the finite element
predicted value of 0.68, from Figure 5.15. Equation 5.15 can therefore be
used to predict the pore pressure which would develop in an axisymmetric
unit cell from plane strain analysis results.
A notable difference in the theoretical solution (Barron, 1948; Hansbo,
1981) and the finite element solution to the consolidation of the unit cell is
the excess pore pressure distribution at low time factors. As pointed out by
114

Consolidation Around a Sin g le Vertical Drain

0
Th=O.025
-Th=0.25

-10
gzo
4-

C
S-

Th=0.75
-aTh=1.75

-40
-50

-N-

Th=3.75

C
11-

-60

1:;:

-90
-100

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

x/9

Figure 5.14a - Development of surface settlement profile in geometry

matched plane strain : without smear or well resistance.
0
Th=0.025

-10
g-20

Th=O.25

4-

C
0

Th=0.75

-B-

Th=1 .75
. -50

-N-

Th=3.75

C
1-

1:;:

-90
-100

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

x/B

Figure 5.14b - Development of surface settlement profile in geometry

matched plane strain : without smear but with well resistance.

115

Consolidation Around a Single Vertical Drain

60
o 50
0.

40

0.30
m
0

-20
1
S

a
x

Lii

0.1

0.2 0.3

0.4 0.5 0.6

r/R or x/B

0.7

0.8

0.9

Figure 5.15 - Comparison of pore pressure distributions from

axisymmetric and geometry matched plane strain analyses without
smear or well resistance (degree of consolidation = 65%).
Barron, under equal strain conditions, the excess pore pressure distribution
across an axisymmetric unit cell, at all times, must be parabolic. This results
in an excess pore pressure greater than the applied load at early time
factors near the periphery of the unit cell. In the finite element solution the
initial excess pore pressure distribution is developed by applying the loading
in an undrained increment; this results in the initial excess pore pressure
being equal to the applied loading at every pore pressure node iii the mesh.
In order to limit the effect of pore pressure oscillations, the initial time step
was chosen such that at no node was the excess pore pressure greater
than the applied load after the first consolidation increment. The variation of
pore pressure, at mid-depth, across the axisymmetric unit cells is shown in
Figure 5.16 at the end of the first consolidation increment of an analysis with
no smear zone and no well resistance. It is interesting to note that, even at
this low time factor, a parabolic variation of excess pore pressure can be
seen to have developed. However, the fact that the initial finite element
116

Consolidation Around a Single Vertical Drain

100
90
a
-

80
70

m
I.
III
m
L

a.

L40
30
20
10
0

0.1

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

r/R

Figure 5.16 - Excess pore pressure distribution after the first

consolidation time increment from axisymmetric finite element analysis
without smear or well resistance.
excess pore pressure distribution cannot conform to the parabolic
distribution discussed by Barron may contribute to the small differences
between the theoretical and finite element results at low time factors, Figure
5.13.

5.3.3. Effect of Vertical Permeability on Matching Procedure

So far the effect of vertical flow in the soil has not been considered. In order
to check its influence on the matching procedure, two further pairs of
analyses were conducted with equal horizontal and vertical permeabilities in

the soil. Specimen results for tih at the mid-depth of the drain are shown in
Figure 5.17 and indicate that the introduction of vertical flow does not

invalidate the matching. In fact the matching errors in 1h in Figure 5.18, are
significantly reduced. This can be explained by the action of the vertical flow
117

Consolidation Around a Sin gle Vertical Drain

100

go

Axisymmetrl

c 50
0
U

40
30

I Matched plane strain

20
10
0-I-
0.01

0.1

'U

Time factor Th

Figure 5.17a - Comparative results for axisymmetric and matched plane

strain analyses with vertical flow (k=k,), degree of consolidation based
on excess pore pressure without smear or well resistance.

100
90
c8O
70
0

60
50
40

20
10
0.01

0.1

10

Time factor Th

Figure 5.17b - Comparative results for axisymmetric and matched plane

strain analyses with vertical flow (kk,), degree of consolidation based
on excess pore pressure without smear but with well resistance.
118

Consolidation Around a Single Vertical Drain

10

a
6
4

9 -2
a

-4
-6

-a
-10
-12
0.01

0.1

10

Time factor Th

Figure 5.18 - Errors in geometry matched plane strain analyses for unit
cells with vertical permeability (k,=kh) based on excess pore pressure.

in promoting more uniform settlement profiles in the plain strain analysis. As

in previous cases (Figure 5.13), the matching errors based on settlement
were little different from those based on pore pressure and therefore they
have not been presented. The influence of the vertical flow on the rate of
consolidation is evident from the separation of the finite element and
theoretical curves, Figure 5.17, but this is likely to be reduced in practical
cases by two factors: firstly, the width of the unit cell (drain spacing) is likely
to be less than that adopted for this study (PJI=BIl=O.5) and, secondly, the
vertical permeability is likely to be less than the horizontal permeability
(Jamiolkowski et al, 1983).

119

Consolidation Around a Sin gle Vertical Drain

5.4. Summary
A methodology has been developed for representing vertical drains in plane
strain finite element analyses of embankments on soft ground, and has
been validated by analysing the consolidation of soil around a single drain.
The methodology includes the use of a drainage element to represent the
vertical drains, Section 3.4. The performance of CRISP has been checked
under axisymmetric conditions against the theoretical solutions of Hansbo
(1981) and found to be satisfactory. The trends obtained by Jamiolkowski et
al (1983) in a parametric study of welJ resistance and smear have been
reproduced successfully.
A procedure has been derived which permits equivalent plane strain and
axisymmetric analyses to be conducted. Throughout the analysis, the
average degrees of consolidation are matched at every level in the soil,
although the pore pressures at corresponding points are not the same. The
equivalent plane strain analysis can be achieved by manipulating the drain
spacing andlor the horizontal permeability of the soil. The validity of the
proposed matching procedure has been examined for drains with and
without well resistance and smear, installed in a uniform soil with linear
compressibility characteristics. The procedure is generally successful; at no
time does the error in the average degree of consolidation, either at the
mid-depth of the drain or over its whole length exceed 11%.
In practice the soil behaviour is likely to be non-linear and to vary with
depth. Also in a full embankment analysis significant lateral strains may
occur in the soil. Chapter 6 describes analyses of a case history in which a
preloading embankment was constructed over a soil into which vertical
drains had been installed. The matching procedure is used to model the
vertical drains in these more realistic soil conditions and comparisons of
observed and finite element predicted behaviour are made.
120

Case History: Porto Tolie

6. Case History: Porto Tolle

6.1. Introduction
A thermal power plant constructed on the Po river delta at Porto Tolle, Italy,
required the construction of several large steel tanks. In order to limit
settlement after the placement of these tanks pre-loading embankments
were constructed and vertical drains installed in the subsoil. The amount of
vertical drain used at the site (approximately I ,700,000m) justified a
comprehensive site investigation followed by the construction of a trial
embankment beneath which several types of vertical drain were installed.
The trial was extensively monitored and used to assess each drain type.
The quality of the field and laboratory data (Hansbo et al, 1981) allowed the
input parameters for a finite element analysis to be chosen with unusual
confidence. Also, the large amount of instrumentation and sustained
monitoring of the trial embankment enabled several comparisons between
finite element analysis and observed behaviour to be made. The Porto Tolle
case history therefore provides an excellent opportunity for comparing finite
element analysis with reliable observed behaviour.
This Chapter is divided into three main Sections. Firstly, the Porto Tolle site
and construction of the trial embankment are discussed. Secondly, analyses
of axisymmetric and plane strain unit cells are presented; these analyses
allow an assessment to be made of the matching procedure, develQped in
thapter 5, for realistic soil conditIons. Finally, full plane strain analysis is
presented and compared with the published observed behaviour.

121

Case History: Porto Tolle

6.2. Porto Tolle

6.2.1. Subsurface Conditions
The Porto Tolle site was subjected to intensive investigation (Croce et al,
1973; Garassino et al, 1979; Jamiolkowski et al, 1980; Hansbo et al, 1981;
Jamiolkowski and Lancellotta, 1984). These investigations showed the
subsoil, Figure 6.1, to consist of 6-9m of medium dense sand and silty sand,
a 20-23m deep layer of silty-clay beneath which is a deep deposit of dense
silty sand. The silty-clay deposit is normally consolidated, Figure 6.2, and
contains very thin lenses and seams of silty sand from a few millimetres to a
few centimetres thick. Jamiolkowski and Lancellotta (1984) observed that
the permeable inclusions were discontinuous and that the clay layer
behaved as a single consolidating stratum with drainage occurring towards
the top and bottom. In the upper and lower part of the clay layer organic
materials and gas pockets were observed. A summary of the published soft
clay soil parameters is presented in Table 6.1. For the upper and lower silty
sands only descriptive information was available.
Reclamation plants maintain the ground water level at 2-2.5m below mean
sea level.

6.2.2. Construction of the Trial Embankment

The original elevation of the site was between 1 and 2m below mean sea
level; before construction of the embankment the ground level was raised,
using a sand fill, to a level of 1-0.5m below mean sea level. All finite element
analyses assume that the original ground level is at 0.5m below mean sea
level and that pore pressures due to the sand layer completely dissipate
before commencement of construction. This assumption was adopted as
the vertical drains were installed in November 1976, over 4 months before
construction started.

122

Case History: Porto Tolle

yellow to brown
medium dense
sand and silty
sand

110
o 20
a,

light bown
soft silty day

.0

0.
U,
-D

30
brown dense
silty sand

Figure 6.1 - Soil profile (after. Jamiolkowski and Lancellotta, 1984).

0
-4
-8
-12
E
16
0

20
-24
-28
-32

100

200

300
Stress (kPa)

400

500

600

Figure 6.2 - Stress history and maximum overburden from

one-dimensional consolidation tests according to Casagrande's
procedure (after: Jamiolkowski and Lancellotta, 1984)
123

Case History: Porto Tolle

VALUE

Compression Index

C, = 0.30 - 0.45

Triaxial compression

4'

= 29

Plane strain compression

4',

= 32

4)'
4)'

PARAMETER

Horizontal permeability
Vertical permeability

k,1 = 7.7x1 0 - 9.6x1 0.10 m/s

K

= 1.5x1 0 0 - 5.5x10 rn/s

Total unitweight

y= 18.6kN/m3

Natural water content

w, = 35-40%

Liquid limit

LL=50 - 60%

Plasticity index

P1=25 - 40%

Table 6.1 - Summary of Porto Tolle soft clay material properties.

The cross section and plan view of the embankment are shown in Figure
6.3a and b and the rate of construction in Figure 6.3c. Four types of vertical
drain were installed in the clay stratum, Figure 6.3b. The finite element
analyses concentrate on the Geodrain area, although little variation in
performance of the four zones was observed (Hansbo et al, 1981). The
embankment was constructed at a constant rate to a height of 5.5m in a
period of 3.5 months; this implies a load due to the fill of 99kN1m 2 beneath
the crest (unit weight of fill l8kN/m3 , Jamiolkowski and Lancellotta, 1984).

consolidation period of 10 months was then allowed before removal of the

embankment.
6.2.3. Vertical Drains

The vertical drains were installed in a triangular grid as reported by

Mogilarardi and Torstensson (1977). The Geodrain spacing was 3.80m and
the equivalent diameter of the drain was 62mm (Jamiolkowski et al, 1983).
This drain spacing results in an area of influence of each drain of 3.9 g m and
124

Case History: Porto Toile

330m
Jetted

Soildrain

Geodrain

Sandwick

(a)

30m

5.5m

66m

(b)
___ 120
E
80

-D
D 40
0

a,
>
00
b

12

1978

1977

time: number of months (Jan=1)

(c)
Figure 6.3 - Porto Tolle trial embankment: (a) Plan view: (b)
Cross-section; (c) Construction schedule.
125

Case History: Porto Tolle

therefore the unit cell radius is 1.995m.

No information regarding the discharge capacity of the drains was available
and for the initial finite element analyses conservative values were
assumed.

6.2.4. Instrumentation
The trial embankment was heavily instrumented with settlement plates.
several types of piezometer and vertical and horizontal inclinometers
(Garassino et al, 1979). Field measurements were taken during the
construction and consolidation stages.

6.3. Unit Cell Analyses

A series of fully coupled consolidation analyses has been performed. The
aims of these analyses were, firstly, to assess the likely affect of the well
resistance of the vertical drain and, secondly, to assess the quality of the
matching procedure developed in Chapter 5 under more realistic soil
conditions. In all of these analyses only the clay layer, Figure 6.1, has been
modelled.

6.3.1. Finite Element Mesh and Material Parameters

The finite element mesh and boundary conditions used for the axisymmetric
unit cell analyses are shown in Figure 6.4.
The clay layer was modelled using modified Cam-clay (Roscoe and
Burland, 1968). The material parameters were obtained from the published
data, summarized in Table 6.1, as follows. The compression index varies
between 0.30 and 0.45, corresponding to a ? of 0.13-0.195. A single value,
?.=0.16, was used in all analyses; this approximates to the mean of the
range and also agrees well with an empirical relationship between ? and
plasticity index proposed by Yudhbir and Wood (1989). For many clays K
126

Case History: Porto Toile

drained

ho rizoni
restrain
d ra i nag
element
or drair
boundar

rntally
med

drain rad
O.031m
tically restrained
drained

R=2.Om

Figure 6.4 - Mesh used for axisymmetric finite element analyses.

127

Case History: Patio ToIIe

approximately 20% of X (Wroth, 1984), therefore, ic was estimated as 0.032.

To ensure that the initial finite element voids ratio was approximately correct
a value for F of 2.58 was used. To ensure that the predicted undrained
shear strengths in plane strain and axisymmetry were correct, Section 6.3.3,
different values of M were used (see Appendix A). The values were derived
from the angle of shearing resistance, 4)', in triaxial and plane strain
compression tests as M=i.16 and M=0.92, which correspond to 4'29
and $'=32 (where the subscripts ax and p1 refer to axisymmetry and piane
strain respectively). A typical value for the Poissons ratio, v'=0.3, was used
(Tomlinson, 1986). The horizontal and vertical permeability parameters,
k,=4. lxi 0m/s and k=3.5x1 0 10m1s, were the averages of the respective
ranges which were based on a combination of field and laboratory tests.
The above parameters are summarised in Table 6.2.
Parameter

Value

A.

0.16

0.032

2.58

1.16

M1

0.92

v'

0.3

4.lxlO9mIs
Ic

3.5xi 010m/s

Table 6.2 - Summary of parameters used to model Porto Tolle clay.

128

Case History: Porto Tolle

6.3.2. In Situ Stresses

The vertical effective and total stresses were derived using a unit weight for
the soil of 18.6kN/m3, both above and below the water table which was
assumed to be I .4m below ground level. The coefficient of earth pressure at
rest was calculated according to modified Cam-clay theory, Appendix A, as
(K)=0.64 and (K0)=0.74. These values are considerably higher than
published, I(=O.48-O.54 (Garassino et al, 1979), but must be used if the
correct response to one-dimensional loading is to be obtained. To calculate
the size of the initial yield locus the layer was modelled as be normally
consolidated, Figure 6.2.

6.3.3. Undrained Analyses to Assess Shear Strength

Two analyses were carried out to check that the undrained shear strength of
the soil was modelled correctly in both plane strain and axisymmetry. The
mesh shown in Figure 6.4 was used with all boundaries undrained, the right
hand boundary was incrementally displaced and the deviator and shear
stress of the elements monitored. When no increase of stress occurred for
an increment of displacement the unit cell was said to have failed. The finite
element predicted undrained shear strength profiles are plotted against the
theoretical values (using equations from Appendix A) in Figure 6.5.
The results presented were obtained using 50 increments with total
displacements of 0.08 and 0.25m for the axisymmetric and plane strain
analyses respectively. Additional analyses with 200 increments and double
the number of elements in both the vertical and horizontal direction
produced negligible differences in the predicted undrained shear strengths

6.3.4. Axisymmetric Unit Cell Consolidation Analyses

Two axisymmetric analyses were conducted using the mesh shown in
Figure 6.4. The embankment loading was represented by a uniform stress
of 99 kN/m 2 applied to the top boundary of the mesh. The load was applied
129

Case History: Porto Tolle

0
Theory
A
Finite element

b0
LI,

E -15
0

.o-20
.r
.1-

0.

ci-25
30

35
0

10

20

30

40

50

60

70 80

undrained shear strength (kpa)

Figure 6.5a - Undrained shear strength in plane strain.

0

Theory
A
Finite element

b0

E 15
0

-C
-I-

a
0-25

30

35
0

10 20 30

40 50 60

70 80

undrained shear strength (kPa)

Figure 6.5b - Undrained shear strength in axisymmetry.

130

Case History: Porto]pflQ

at a constant rate over 50 equal time increments totalling 3.5 months. The
consolidation was modelled using 50 equal time increments totalling 10
months.
The first analysis used drainage elements, Section 3.4, to itode1 the vertica
drain, with a discharge capacity of l4Om3Iyear corresponding to the
minimum likely value (Holtz et at, 1991). The second modelled the dra n as
infinitely permeable by setting the excess pore pressure to zero at the left
hand mesh boundary, Figure 6.4. The average surface settlement for each
analysis is plotted against time in Figure 6.6; as cai be seen even The
lowest likely discharge capacity has a negligible effect on the rate of
consolidation. The maximum hydraulic gradient in the drain, used for the
first analysis, occurred transiently near the surface and was equal to 0.11.
As the actual discharge capacity is likely to be significantly higher than that
assumed, the effect of well resistance can be ignored and the vertical drain
modelled as a boundary with zero excess pore pressure.
In the remainder of this Chapter the analysis without vertical drainage
elements will be referred to as the axisymmetric analysis.

6.3.5. Comparison of Finite Element and Observed Behaviour

The rate of consolidation of the axisymmetric unit cell can be compared with
the observed pore pressure dissipation and rate of surface settlement
during the trial. The unit cell analyses neglect the sand layers and the lateral
deformation of the soil; however, as the sand is relatively incompressible
and the relevant observations were made beneath the embankment
centreline these effects were considered negligible.
The excess pore pressure at the periphery of the unit cell is compared with
observed values in Figures 6.7a and b for depths of 19.7 and 12.6m below
ground level respectively; the peripheral value was used for comparison as
the piezometers were placed at the centre of the triangularly spaced drains.

131

Case History: Porto Tolle

?ioa
E
'-200
E 300
41-

.0

U -

a
1L
a

800
900

6
8
time (months)

10

12

14

Figure 6.6 - Axisymmetric unit cell consolidation based on average

surface settlement.

The settlement of the top right hand corner of the unit cell is compared with
observed values in Figure 6.8.
During construction the agreement between predicted and observed pore
pressures and settlements is very good, therefore validating the choice of
material parameters. However, during the ten month consolidation period,
although the predicted settlement is still very close to that observed Figure
6.8, the predicted pore pressures show a marked deviation from the
monitored values, Figure 6.7. The finite element predictions of both pore
pressures and settlements are consistent, with the settlement reaching a
final constant value as the excess pore pressures approach zero. This trend
is not seen in the observed values. Although the observed settlement
appears to be levelling off, implying the end of primary compression, there
is little dissipation of the excess pore pressure after the end of construction.
Jamiolkowski and Lancellotta (1984) suggested that the observed pore
132

Case History: Porto Totle

70
60

U
0

50
I-

[40
30
0

aI,

ci
x
10
0
0

- observed

6
8
time (months)

10

12

14

A finite element

Figure 6..7a - Excess pore pressure 19.7m below ground level on the
centreline of the embankment.
80
70
U

a.60
50
a-40
-30
U

20

x
10
0
0

observed

6
8
time (months)

10

12

14

A finite element

Figure 6.7b - Excess pore pressure 12.6m below ground level on the
centreline of the embankment.
133

Case History: Porto Toile

0
- 100

observed

A-- finite element

20D

o --500
60O
" 700
0

8O0
F-900
1000
0

6
B
time (months)

10

12

14

Figure 6.8 - Rate of surface settlement at centreline of the embankment.

pressures during the consolidation stage may have been unreliable, and
stated that the unsatisfactory long term performance of the piezometers
may have been due to the presence of organic gas. However, the
malfunction of the piezometers may have masked the effect of phenomena,
such as destructuring of the soil (Burland, 1990), which would also have
resulted in higher observed excess pore pressures than predicted by the
finite element analysis. As the monitored excess pore pressures during the
consolidation stage may be in error, the material parameters used for the
axisymmetric analysis were retained in further analyses.
From Figure 6.7 it is interesting to note that the pore pressure at 19.7m
below ground level is lower than at 12.6m below ground level. This
behaviour appears incorrect as with an infinitely permeable drain the rate of
consolidation of an isotropic material should be the same at every depth,
and the effect of the small vertical permeability would imply lower pore
134

Case History: Porto Tolle

pressures closer to the top or bottom boundaries. The cause of this

apparent anomaly is the way in which the modified Cam-clay model has
been implemented in CRISP. The coefficient of consolidation depends on
both the permeability and the compressibility of the material; in CRISP the
permeability is constant for a given material but the compressibility, which is
dependent on both stress level and voids ratio, varies with depth. This
results in the coefficient of consolidation reducing with depth in a modified
Cam-clay analysis and is the cause of the above variation in the dissipation
of excess pore pressure with depth. At Porto Tolle this variation was not
large but as will be seen in Chapter 7 this effect may cause a much larger
variation in other circumstances.

6.3.6. Plane Strain Unit Cell Consolidation Analyses

Two plane strain analyses have been carried out to assess the quality of the
matching procedure developed in Chapter 5. In the first analysis the width of
the unit cell has been changed and the soil permeability left the same as for
the axisymmetric analysis; this is referred to as geometry matching. In the
second analysis the horizontal permeability has been changed whilst
maintaining the same unit cell dimensions; this is referred to as permeability
matching.
Details of the procedure are given in Chapter 5. Applying Equations 5.10
and 5.14, in which the effect of smear and well resistance has been ignored,
the plane strain unit cell width for the geometry matched case is B=4.5m
and the horizontal permeability for the permeability matched analysis is
kh 8.OxI 010m/s.
The two analyses in plane strain were conducted in the same way as those
in axisymmetry. The vertical drain was again modelled as infinitely
permeable. Comparisons of the average surface settlement and average
excess pore pressure at mid depth are shown in Figures 6.9 and 6.10. The
average values were calculated using the trapezium rule. Only the geometry
135

Case History : Porto Tolle

60

a-

o 30
a1
0
U
fn

x
0
0

110

00
0

6
8
time (months)

10

12

14

Figure 6.9 - Unit cell rate of consolidation based on the average excess
pore pressure at mid-depth.

0
_i00
..- 200
C
0

E 300
0

-I-I-

400

500

1-

L
11

600

o-700

U
I-

-BOO
900

6
a
time (months)

io

12

14

Figure 6.10 - Unit cell rate of consolidation based on the average

surface settlement.
136

Case History: Porto Tone

1
0

I0
'--2
IDI

c 3
CI
4U

E 4
5
6
0

6
8
time (months)

10

12

14

-- pore pressure A settlement

Figure 6.11 - Unit cell rate of consolidation matching errors.

matched analysis has been plotted as the difference between the two plane
strain analyses was too small to be noticeable.
For both settlement and pore pressure the agreement between
axisymmetric and plane strain analyses is very good. It is possible to define
matching errors as
AVERAGE EXCESS PORE PRESSURE
(ERROR) pwp - DIFFERENCE IN PREDICTED
100 ........(6.1)
APPLIED LOAD

PREDICTED AVERAGE SETTLEMENT x

(ERROR)sj= DIFFERENCE IN FINAL
100
SETTLEMENT

....................(6.2)

These errors are plotted against time in Figure 6.11 from which it can be
seen that the matching error is never greater than 6%.

137

Case History: Porto Tolle

nbankment
modelled using
ent

drair
sand

I loads

clay

drair
sanc

restrainea

horizontally and
verticall v restrained

restrainea

Figure 6.12 - Mesh used for full plane strain analysis.

6.4. Full Plane Strain Analysis

Having established that well resistance was not a significant factor and that
the proposed matching procedure was reliable a full plane strain analysis
was carried out.

6.4.1. Finite Element Mesh and Boundary Conditions

The finite element mesh used for the full plane strain analysis is shown in
Figure 6.12. Both the upper and lower sands have been modelled as
drained layers. The unit cell mesh, Figure 6.4, has been stretched to a width
of 4.5m, as used in the geometry matched unit cell analysis, and repeated
to model the area of soil improved with vertical drains. The boundary
conditions used in the analysis are also shown. The embankment was

138

Case History: Porto Tolle

modelled using vertical loads applied in the same number of increments and
time as for the unit cell analyses.
6.4.2. Material Parameters
Clay Layer and Drains
The material parameters for the clay layer were as defined in Table 6.2. The
drains were modelled as lines of nodes at which no excess pore pressure
could develop, thus representing an infinitely permeable drain (Section
6.3.4).
Sand Layers
Only descriptive information was available for the upper and lower sand
layers and it was therefore necessary to adopt typical parameter values to
model these materials. The layers were modelled as drained
elastic-perfectly plastic materials for which five parameters are required: the
cohesion (c'), the angle of shearing resistance (4'), the Young's modulus
(E), the Poissons ratio (v') and the unit weight (y). The cohesion was taken
to be zero, c'=O. The angle of shearing resistance was defined as 4'=44.4
following Jewell (1990). A typical value for the Poissons ratio under drained
conditions was used, v'=0.2, (Lade, 1977). The unit weight of the sand was
assumed to be the same as for the clay, i.e. y=18.6kNIm3. Again following
Jewell (1990), the Youngs modulus for the soil was derived from an
empirical relationship proposed by Hardin and Black (1966) for the initial
elastic shear modulus of soil, G1,
Gj=7OO(27() ...(6.3)
where s r=lOOkN/m2 is included for dimensional consistency,s' is assumed
to be equivalent to the mean effective stress (p') for plane strain conditions
and e is the voids ratio, assumed equal to 0.5. Equation 6.3 was used to
evaluate an initial shear modulus, G 1. In order to use a shear modulus more
139

Case History: Porto Tolle

0
E
E
-I-

a,

Ea,
-I-I-

a,
a,

100
-300
-400
-500

'II-

-600

"3

-700

a,

C
a,

-k- finite element

-200

a,

U
U

observed

-800

I-

4C
a,
U

-900
-10004
0

6
8
time (months)

10

12

14

Figure 6.13 - Surface settlement at centreline.

appropriate to the strain levels that may develop in the embankment before
yielding, a value for the shear modulus of G=0.5G1 was used. The Young's
modulus was then related to the shear modulus as E=2(1+v')G.
Three sand layers were defined, namely: z=0-7.5m, 29-42m and 42-62m
below ground level and the Young's modulus values in these layers were
5.9x104, 14.5x104 and 17.4xlO4kNIm2 respectively.

6.4.3. Comparison of Observed and Predicted Behaviour

Settlement
The surface settlements on the embankment centreline are plotted in Figure
6.13. The observed and predicted values are in good agreement. However,
while the predicted settlement has reached a steady value at the end of the
analysis, indicating complete consolidation, the observed settlement is
140

Case History: Porto Tolle

0
observed

k-- finite element

60
80
100
12a
E
x
a
E
16C
0

6
8
time (months)

10

12

14

Figure 6.14 - Maximum horizontal displacement.

continuing at a reduced rate. This may be due to secondary compression.

Garassino et al (1979) quoted a value for the coefficient of secondary
compression of c=0.460.22% measured in oedometer tests. If it is
assumed that 90% consolidation occurs in 6 months and that secondary
compression then starts, the estimated settlement over the remaining 7.5
months is approximately 15mm. This suggests that the inability of CRISP to
model secondary compression is not a significant limitation in this case.
Lateral Movement
An inclinometer was installed beneath the embankment approximately 9m
from the toe. The development of the maximum lateral movement, which
was predicted by the finite element analysis to occur at around lOm below
ground level, is plotted in Figure 6.14. Again the agreement between
analysis and observation is encouraging. The observed lateral displacement
141

Case History: Porto Tolle

has not reached a constant value at the end of the period considered; this
may again be explained by the continued creep of the soil which is not
modelled in the finite element analysis.
The lateral movement profiles at the end of construction and at the end of
consolidation are plotted in Figure 6.15. The maximum values of lateral
movement are in good agreement, although the observed small inward
movement of the surface has been exaggerated in the finite element
analysis. This may be a result of modelling the embankment simply using
vertical loads which does not allow the development of outward shear
stresses at the ground surface, as may occur in practice.
The inward movement cannot be explained as being caused simply by
settlement of the embankment. If the embankment was assumed to remain
the same length (33m) and the settlement increased linearly from the toe to
the centreline value of 900mm, then the toe would move inwards by only
2mm. Most of the predicted inward movement must be caused by the initial
elastic behaviour of the sand. It may be noted that for a footing moving
rigidly into an elastic layer an inward movement at either end of the footing
is predicted (Poulos and Davis, 1974).
Pore Pressure
Piezometers were placed on the centreline of the embankment at depths of
19.7 and 12.6m below the surface. The piezometers were positioned at the
centre of a triangle of drains so as to record the highest pore pressures. In
Figure 6.16 comparisons of the observed and finite element predicted pore
pressures are shown.
The excess pore pressure calculated from the finite element program is for
a plane strain analysis and as discussed in Chapter 5 the matching
procedure ensures that the average pore pressure across the axisymmetric
and plane strain unit cell are equal at every depth and at every time.
However the distribution of pore pressure across each unit cell is not the
142

Case History: Porto Tolle

0
observed
5
finite element

10
15

25
30
35
40
45
50
200 150 100 50
0
lateral displacement (mm)

100

150

Figure 6.15a - Lateral movement profile at the inclinometer position at

the end Qf the construction stage.
0
observed

A
finite element

10
15

-a
30
35
40
45
50
0
200 150 100 50
lateral displacement (mm)

100

150-

Figure 6.1 5b - Lateral movement profile at the inclinometer position at

the end of the consolidation stage.
143

Case History: Porto Tolle

80
70
a
a-

. 60

20
U

10
0

6
8
time (months)

observed

10

12

14

A finite element

Figure 6.16a - Excess pore pressure 19.7m below ground surface on

centreline.
80
70
a
a, 60
50
a40
S

.30

observed

6
8
time (months)

10

12

14

A finite element

Figure 6.16b Excess pore pressure 12.6m below ground surface on

centreline.
144

Case History: Porto Tolle

same and in the present case, at the periphery of the unit cell, the
axisymmetric pore pressure is only 72% of that predicted in plane strain, as
predicted using Equation B33, Appendix B. It is therefore necessary to
correct the plane strain pore pressure at the periphery of the unit cell by
multiplying by a factor of 0.72 before comparison with the observed data
can be made.
From Figure 6.16 it can be seen that the predicted excess pore pressures
during construction are in good agreement with those observed. However,
as seen in the axisymmetric unit cell analyses, Section 6.3.5, the agreement
after the end of construction is not good. As previously discussed the likely
cause of this discrepancy is the unsatisfactory long term performance of the
piezometers due to the presence of organic gas, resulting in incorrect
recorded values. However, other factors may result in an increased
observed excess pore pressure (Eg. destructuring of the subsoil) and
should ideally be considered in future analyses.

6.5. Summary
The matching procedure developed in Chapter 5 has been applied to a
normally consolidated clay deposit modelled using modified Cam-clay. The
comparisons of the rate of consolidation based on both average surface
settlement and average pore pressure at the mid depth of axisymmetric and
plane strain unit cells are good. The matching procedure has therefore been
shown to be capable of providing an accurate representation of soil
improved using vertical drains analysed in plane strain.
A full plane strain analysis of the Porto Tolle trial embankment has been
carried out in which the matching procedure was used to calculate the
spacing of the vertical drains in plane strain. The finite element predicted
vertical and lateral displacements and pore pressure on the centreline are,
with the exception of excess pore pressures after construction, in good
agreement with the observed values. The study shows that with accurate
145

Case History: Porto Tolle

soil parameters obtained from high quality field and laboratory tests and
with careful finite element modelling accurate predictions of the behaviour of
soils improved with vertical drains can be made.

146

Multi-Stage Embankment Construction

7. Multi-Stage Embankment
Construction
7.1. Introduction
The maximum single stage height of an embankment, with a given side
slope length, built on soft clay can be estimated using plasticity theory (as
described in Chapter 4). If the required embankment height is greater than
this then a multi-stage construction approach could be used in which the
subsoil is allowed to consolidate and gain strength between each, of
several, loading stages. Vertical drains are often installed in the subsoil to
increase the rate of consolidation and therefore shorten the required
consolidation stages.
The design of a multi-stage embankment requires the accurate prediction of
the strength increase of the subsoil and the application of the correct
analysis method to each loading stage. Three approaches have been
summarized by Ladd (1991).
I Total stress analysis - Construction and failure are assumed to occur
in a sufficiently short time so that the strength increase of the subsoil
is negligible. The shear strength of the subsoil is based on
unconsolidated undrained laboratory tests or field vane shear tests.
This method cannot take account of the increase of shear strength
due to consolidation and is only applicable to a single lift
embankment analysis.
2 Effective stress analysis - Effective strength parameters are
measured in consolidated drained laboratory tests and insitu
measured pore pressures are used to calculate the available shear
strength along a potential slip surface. The implicit assumption is
147

Multi-Stage Embankment Construction

that no excess pore pressures are generated during failure.

However, observations indicate that embankment failures generally
occur in a matter of minutes. This would imply the generation of
significant pore water pressure during failure, thus reducing the
effective stress and therefore the factor of safety from that which
would be predicted from an effective stress analysis.
3 Undrained strength analysis - The effective stresses are sed to
predict the undrained shear strength of the subsoil. These strengths
are then used in a total stress analysis. This procedure, proposed by
Ladd (1991), Jardine and Hight (1987) and Leroueil et al (1990),
allows for increases of strength in the subsoil and appreciates that
failure is likely to occur sufficiently quickly so as to be undrained in
nature.
The undrained strength analysis provides a logical method for the analysis
of stage constructed embankments. However, the designer still has to
decide how to estimate the effective stresses in the subsoil at any time and
how to use these stresses to predict accurately the undrained shear
strength before analysis of a loading stage. In this Chapter a finite element
analysis of an idealized two stage construction is performed and the
undrained shear strength increases of the subsoil are calculated. Simple
methods of estimating the subsoil undrained shear strength increase are
then proposed and compared with the finite element predictions. These
simple methods are also used to predict undrained shear strength increases
for the Porto Tolle case history presented in Chapter 6 and compared with
the finite element results.

7.2. Idealized Two-Stage Construction

An idealized reinforced embankment constructed in two lifts on a soft clay
subsoil containing vertical drains has been modelled using CRISP. The
finite element analysis has been used to predict the strength increase of the
148

MuIti-Staae Embankment Construction

subsoil after a period of consolidation. The finite element analyses were

carried out in conjunction with limit equilibrium analyses, of the same
idealized embankment, performed by Jewell (1991)1.

7.2.1. Finite Element Analysis of Idealized Two-Stage

Construction
The material parameters and embankment geometry were chosen to be
typical of embankments built over soft clay subsoils. Reinforcement at the
base of the embankment and vertical drains in the subsoil were modelled.
Fully coupled consolidation analyses have been performed in which the
subsoil was allowed to consolidate with the embankment modelled as fully
drained. The embankment was constructed in two lifts applied relatively
quickly with a period in between in which significant consolidation occurred.
Geometry
The embankment and subsoil geometry are defined in Figure 7.1. The soft
clay was assumed to overly a rough rigid layer. The heights of the first and
second lift were calculated by Jewell (1990), using a limit equilibrium
method, to provide a factor of safety of 1.1 at the end of each lift.
Soft Clay Subsoil
The soft clay was modelled using modified Cam-clay, with parameters:
M=0.7, =0.25, c=0.05, 1=3.0, v'=0.3 and y=18kNIm 3. Following
Appendix A these parameters imply 4'=24, K0.83 and (sJa')=0.21 8.
An over-consolidated profile was modelled by assuming that the present
water table at I m below ground level had at some time previously been
lowered to 2m below ground level. The above material parameters and
stress history defines the subsoil strength and initial stress conditions
shown in Figure 7.2.
1

Private communication.
149

Multi-Stage Embankment Construction

6.5m

1O.Om
1.2m
2.6m

1O.Om

Figure 7.1 - Idealized two-stage embankment geometry.

The horizontal permeability was k h =3x10m/s with a negligible vertical

permeability so that flow was predominantly towards the vertical drains,
which were the only draining boundaries. Vertical flow was minimised so
that the subsoil surface did not consolidate and gain strength quicker than
the soil mass. Hansbo's (1981) equation for consolidation around a vertical
drain could therefore be used to generate comparisons between simpler
design methods and the finite element analyses.
Embankment Fill
The fill was modelled as a fully drained elastic-perfectly plastic material with
a Mohr-Coulomb yield criterion representing a granular fill. The fill was
defined as having a bulk density of y bU =20kN/m3 and elastic parameters:
E=lO5kN/m2 and v'=0.2. The strength parameters were 4'=30 and
c'=5kNIm2.

150

Multi-Stage Embankment Construction

30

60

90

0.8

1.2

1.6

2.0

16

24

0
2
93kPa
l9kPa/m
4-.,
0
a)
-o
8
10
(a) stress

(b) OCR

(c) strength

Figure 7.2 - Idealized soil (a) stress (b) Over-consolidation ratio (C)
Initial undrained shear strength.

The small cohesion was necessary in order to avoid numerical problems in

the finite element solution and is justifiable as some suction would be
present in the fill material. This embankment model may allow tensile
stresses to develop, so the analysis was monitored to ensure that any
tensile stresses did that develop did not become unrealistically large.
Reinforcement
The reinforcement was modelled as a linear elastic material with a stiffness
J=2000kN/m placed at the base of the embankment.
Vertical Drains
A design consolidation period of 6 months was allowed in which a system of
vertical drains would produce 60-70% consolidation. Using Hansbo's (1981)
equation for the consolidation of soil around a vertical drain and using a
typical effective drain diameter of 65mm (Jamiolkowski et al, 1983) a drain
spacing of 2.5m was estimated to cause approximately 62% dissipation of
151

Multi-Stage Embankment Construction

excess pore water pressure beneath the centre of the embankment. This
calculation assumes that the drains have no well resistance and cause no
smearing of the soil during installation and that the subsoil undergos
one-dimensional loading.
In the finite element analysis the drains were modelled as infinitely
permeable by defining boundaries at which the excess pore pressures were
set to zero. The matching procedure, Chapter 5, was used to provide a
finite element drain spacing of 4m by modifying the horizontal subsoil
permeability to k,1=1.77x1OmIs. The positions of the drains in the finite
element analysis are shown in Figure 7.1.
Soil/Reinforcement Interfaces
In order to represent accurately the behaviour of the reinforcement it was
necessary to model the (lower) subsoil/reinforcement and (upper)
reinforcement/fill interfaces. The stiffness parameters of both interfaces
were: normal stiffness,

k=1O4kNIm2

and, shear stiffness, k=1O6kN!m2.

Stiffness values in this range have been shown by Kwok (1987) to produce
satisfactory results.
The upper interface yield criterion was modelled with 4=300 and =5kNIm2
so that the full shear strength of the fill material was assumed to be
mobilised along the upper interface at failure.
The lower interface yield criterion was modelled as purely cohesive with a
value of =3.93kN/m2, equal to the initial undrained shear strength of the
subsoil surface, during the first lift and the consolidation stage. The interface
strength was increased for the final lift to represent the enhanced subsoil
surface strength. The increased interface shear strength was the average of
the undrained shear strengths in the surface elements, of each unit cell, as
shown in Figure 7.3.

152

Multi-Stage Embankment Construction

20

16
a

.c12
.4-

a'
C
I-

-I-

I-

0
UI

6
8
10
12
Distance from foe Cm)

14

16

iB

Figure 7.3 - Lower interface shear strength after the consolidation

stage.
Construction Sequence
The first lift of 2.6m was applied in 20 days, during which some
consolidation of the elements adjacent to the drains occurred. A further 180
days was allowed in which the subsoil consolidated and gained strength.
The embankment was then constructed at a steeper side slope, so that the
crest position remained constant, to a height of 3.8m in 20 days, Figure 7.1.
Each of the loading stages was modelled using 200 increments and the
consolidation period occurred over 50 time increments of increasing size.
Undrained Loading to Failure
At the end of both the first and second lifts an undrained analysis was
carried out in order to assess the factor of safety. A uniform surcharge load
was applied to the crest of the embankment, and increased until failure of
153

Multi-StaQe Embankment Construction

reinforcement and
interface level

a)5

C .i-.
C
LO
4- N

C.
C
i0
..-. N

-c

position of
vertical drains

Figure 7.4 - Finite element mesh and boundary conditions.

the subsoil occurred. Failure was defined as the increment during which the
increase in settlement was greater than the increase in the height of fill,
represented by the surcharge loading. This definition of failure has been
previously used by Rowe and Soderman (1987) and is discussed further in
Section 7.2.2.
Finite Element Mesh and Boundary Conditions
The mesh and boundary conditions used for the finite element analysis are
shown in Figure 7.4. Four vertical drains have been modelled at the
positions shown.

7.2.2. Finite Element Results

Displacements
The vertical displacements of points directly below the toe, the crest and the
centreline on the surface of the subsoil are plotted against embankment
154

Multi-Stag e Embankment Construction

6.00

TOE

undrained
loading

5.00

CREST

E
CENTRELINE

'4.00
.

0)
w

second
lift

3.00
.

*NON

E
2.00

consolidation stage

Id

first lifI-'1

1 .00

0.00

'I

1.2-1.0 0.8 0.6 0.4 0.2 0.0 0.2 0.4 0.6 0.8

Settlement (m)

Figure 7.5 - Settlement of points at the subsoil surface against

embankment height.
height in Figure 7.5. At the end of the first lift the settlement response is
non-linear indicating significant yielding of the subsoil. During the
consolidation stage significant settlement occurs with no increase of
embankment height as the subsoil consolidates and gains strength. The
final lift, to 3.Bm, again shows a non-linear response. Beyond a height of
3.8m the undrained surcharge loading causes large displacements, with the
curves for the three positions flattening out and indicating the approach of
failure.
A change of gradient of the crest and centreline settlement results can be
seen when the analysis changes from consolidated to undrained, i.e. at the
end of the second lift and the start of surcharging. A steeper settlement load
curve would be expected in an undrained analysis as the large bulk
modulus of water, KN=5xlO5kNIm2, allows negligible volumetric strain and
155

Multi-Stag e Embankment Construction

the settlement due to the loss of water from the system can no longer occur.
Further research into the change from consolidated to undrained analyses
may be required to ensure that the undrained failure predicted is accurate.
The maximum centreline surface settlement at the end of the consolidation
was 691mm, this value is comparable to an estimate of the one dimensional
consolidation settlement beneath the embankment crest, performed before
the finite element analysis, of 0.71m. This calculation was based on an
average coefficient of volume change, m=2.2x10 3m2/kN, and assuming
that 62% of the excess pore pressures dissipated during the consolidation
stage, Section 7.2.1.
The centreline settlement increased to 799mm by the end of the second lift
as a result of mainly distortional deformations, although some small volume
change did occur. At failure the final centreline settlement was 1147mm,
indicating large distortional deformations with the development of a
rotational mechanism, Figure 7.10.
The surface settlement profiles at the end of each of the three stages and at
failure are shown in Figure 7.6. Heave occurred near the toe and the
greatest settlement occurred beneath the crest. The ratio of the maximum
heave to the maximum settlement increased with the application of load.
The ratio was 0.48 after the first lift, 0.31 after consolidation, 0.38 after the
second lift and 0.68 at failure.
The lateral displacements of a section through the toe are shown in Figure
7.7. The maximum lateral movement occurs at a depth of 1-2.5m below
ground level. The lateral movement profile is typical of that observed in finite
element analyses but differs from those often observed which are
significantly more concave below the peak movement.
The maximum surface displacement is plotted against the maximum lateral
movement in Figure 7.8. The form of this curve is similar to those produced
by Jardine and Hight (1987) and Leroueil et al (1991) with the ratio of lateral
156

Multi-Stage Embankment Construction

1000

500
E
0
1C
0

500

1000

1500
15

10

10
0
5
Distance from toe (m)

-- f si 1ff I -I- co gO.

-- 2nd

ili -9-

15

20

1dhra

Figure 7.6 - Surface settlement profiles.

2
3
E
4-

a0

a
g
1200

1000

400
600
800
Lateral movement (mm)

200

-.- 1 st lift -+-- consoi "c-- 2nd lift -9- failure

Figure 7.7 - Lateral movement of a section through the toe.

157

Multi-Stage Embankment Construction

1400
1200
E
E 1000
end of consolidation

4-

800
end of 2nd lift
' 60C
E

20(

end of first lift

0

200

400
600
800
1000
Maximum lateral movement (mm)

1200

Figure 7.8 - Maximum surface settlement against maximum lateral

movement beneath the toe.
deformation to settlement being greater during loading than during
consolidation. It is interesting to note that at failure the ratio is approaching
unity, similar behaviour being observed physically for the Queensborough
bypass and a centrifuge model test (see Jardine and Hight, 1987).
Factor of Safety
In order to define the failure height of the embankment the ratio of the
change in the maximum surface settlement, over ten increments, to the
depth of fill placed has been plotted in Figure 7.9. When the ratio is equal to
unity the settlement of the subsoil is greater than the depth of fill placed and
there is no net gain in height, therefore failure is deemed to have occurred.
From this graph the failure of both undrained loading stages can be defined
to give an indication of the factors of safety of the first and second lifts. The
first lift fails with an additional surcharge equivalent to 1.33m of fill and the

158

Multi-Stage Embankment Construction

1.60
1.40
1.20
1.00
0.80
0.60
0.40
0.20
0.00
0.00

1.00 2.00 3.00 4.00 5.00 6.00

Total fill height (m)

Figure 7.9 - Ratio of maximum settlement to maximum lateral

movement against embankment height (ds=change in maximum
settlement, dh=height of fill added).
second lift with 1.40m. This additional loading can be used to define a factor
of safety as
FS=

.......................................................................(7.1)

where H is the height before surchargeing and AH is the height of surcharge

required to cause failure.
From this definition the factor of safety of the first lift is 1.5 and the second
lift is 1.4. The corresponding factors of safety calculated by Jewell (1991)
using a limit equilibrium approach was 1.1 for both lifts. However, it is
difficult to directly compare the finite element and limit equilibrium factors of
safety for several reasons
159

Multi-Sta g e Embankment Construction

I The finite element analysis allowed 20 days for the placing of the
first and second lifts in which time the soil adjacent to the drains
would have partially consolidated and gained strength. This would
result in a higher factor of safety as the limit equilibrium assumed
undrained loading.
2 The strength increases during the consolidation stage adopted in the
limit equilibrium calculations were significantly different to those
observed in the finite element analysis. This aspect is discussed
further in Section 7.3.
3 The limit equilibrium calculation used a factor of safety based on the
available shear strength and the applied loading. The calculated
factor of safety of 1.1, therefore, implied that the embankment height
could be increased by a factor of 1.1 across the whole section. The
finite element calculation applied a surcharge loading along the crest
and not on the side slope. The mechanism which develops in the
finite element analysis would have been different to that produced if
the embankment loading was proportionally increased at all points.
4 When predicting undrained collapse using modified Cam-clay a
large number of increments need to be used to ensure that the
collapse load is not over-predicted. The only certain method of
prediction is to perform a sensitivity study with increasing numbers
of increments until the collapse is unaffected by an increase in the
number of increments. Such a sensitivity study was not carried out
for in this case but a large number of increments were used. The
first lift failed in the 660th increment the first 200 increments were
consolidated, the second lift failed in the 690th increment with the
first 450 increments consolidated.
To compare the limit equilibrium and finite element predicted failure directly
it would be necessary to carry out a limit equilibrium analysis using the finite
160

Multi-Stage Embankment Construction

'-

JI

I I

II

______
-

..

-------

iimii ii I

--_'-/..'MY_'//I 111111111

ISIHUSI P

Figure 7.10- Displacement vectors at failure.

element predicted strength at the end of the consolidation stage and to

perform a finite element analysis in which the second lift loading is
undrained and the additional surcharge loading is applied proportionally
across the whole embankment.
The displacement vectors at failure are shown in Figure 7.10. For a subsoil
of this type and geometry a rotational failure mechanism would be expected
to develop, Chapter 4. Such a mechanism can be seen to have developed
at failure.
Subsoil Stresses
The finite element analysis can be used to investigate the subsoil stresses
at key positions. Investigations beneath the slope and the beneath the crest
have been carried out, considering elements at increasing distances from
vertical drains. The positions of the elements (A, B, C, D E and F) are
shown in Figure 7.1.
In Figure 7.11 the stress paths for the six positions in the subsoil are
plotted. Figure 7.11a shows the stress paths for points beneath the slope.
161

Multi-Stag e Embankment Construction

45
40
35
30
25
0

20
15
IC

20

10

40

30

60

50

60

70

Figure 7.1 Ia - Stress paths of the subsoil beneath slope (* indicate

change of loading).

45
40
35
30
25
0

20
15
'C

10

20

30

40
p.

50

60

70

60

Figure 7.IIb - Stress paths of the subsoil beneath crest (* indicate

change of loading).
162

Multi-StaQe Embankment Construction

None of these elements reach the critical state but it can be seen that the
response during loading is increasingly undrained as the distance from the
drain increases. Position C, furthest from the drain, shows an almost
undrained response whilst position A, closest to the drain, shows a virtually
drained response, and is almost parallel to the K0 line during the
consolidation stage, until the undrained loading is applied. Figure 7.11b
shows the stress paths for positions close to the centreline and the general
trend of behaviour is the same as beneath the slope. Some of the soil is
approaching the critical state at failure of the embankment. These stress
paths follow a similar pattern to those observed by Almeida et al (1985,
1986) and Smith(1984).
The rotation of the principal stress direction within the subsoil is of
significance as the subsoil undrained shear strength is known to be
dependent on the orientation of principal stresses (Eg: Ladd, 1991; Hight et
al, 1987), this aspect will be discussed in more detail in Section 7.3. The
rotations beneath the crest are shown in Figure 7.12b. At position 0
principal stress directions are exchanged during the first increment with the
horizontal stress becoming the maximum principal stress. However, apart
from this, the principal stress rotations are less than 100 and conditions
beneath the crest approximate to those in a one dimensional compression
test The rotations beneath the slope, Figure 7.12a, are much larger
throughout the analysis although the rotations reduce sharply as failure is
approached. This variation of rotation across the subsoil is complicated by
the inclusion of the vertical drains but compares well with that observed by
Smith (1984).
The parameter introduced by Bishop (1966)
b= 05 ............................................................................(7.2)
1

can be modified so that

c' and a' are the in plane maximum and minimum

principal stresses and a2' is the out of plane stress, ar'. Now b is redefined
163

Multi-StaQe Embankment Construction

90.0
60.0
(3

o 300

(3
.4-

a
'- 00
UI
d3
L
-U
C
-I-

U
-I-

-90.0
0.00

1.00

2.00

3.00

4.00

5.00

6.00

Embankment height (m)

Figure 7.12a - Principal stress rotations beneath the slope.

90.0
60.0
U

o 300
(3
-I-

a
-' 00
In
0
L
0)
0
-U

-30.0

o_
a

-4-

-4-

0
E

-90.0
0.00

1.00

2.00

3.00

4.00

500

6.Oa

Embankment height (m)

Figure 7.12b- Principal stress rotations beneath the crest.

164

Multi-Stage Embankment Construction

1.0
0.5
0.0
.0

1.0
1.5
2.0
0.00

1.00

2.00
3.00
4.00
Embankment height (m)

5.00

6.00

Figure 7.13a - Modified Bishop stress ratio beneath the slope.

1.0
0.5
0.0
.0

1.0
1.5
2.0
0.00

1.00

3.00
4.00
2.00
Embankment height (m)

5.00

6.00

Figure 7.13b - Modified Bishop stress ratio beneath the crest.

165

Multi-Sta g e Embankment Construction

as

b.=a.............(7.3)
Oi

This implies that, if a'>a1' then b>1 if

'<c'

then b*<0 and if a3 >a> 1 then

btb.
The variation of b* with embankment height has been plotted in Figure 7.13.
Beneath both the slope and crest during the first lift the element nearest the
drain has an out of plane principal stress that is less than the minimum in
plane principal stress. This is consistent with the soil nearest the drain
consolidating quickly, causing little out of plane stress, and the soil further
from the drain becoming increasingly undrained and undergoing stress
increases in all directions due to the increased vertical stress. During the
consolidation stage for all positions b* reaches a value between 0.0 and 0.5.
In the second lift the elements away from the drain are approaching the
failure value of b*=0.5 with the soil adjacent to the drain having a relatively
constant value. Finally the undrained loading causes all points to approach
a value of b*=O.5. This is consistent with Figure 7.11 in which the stress
paths are approaching the critical state.
The strength at the end of each increment can be calculated from the void
ratio, Appendix A, Equation A30. The ratio of the strength to the effective
vertical stress (sJc') is plotted in Figure 7.14 and the ratio of the strength to
the maximum in plane principal effective stress (sJci 1 ) is plotted in Figure
7.15. Two points are noted from these graphs; firstly, at no stage of the
analysis does either ratio fall below the normally consolidated ratio
sJcy' 1=0.218 and, secondly, the ratio sJ; ' is always equal to or greater than
the ratio s/a1' as the effective vertical stress can never be greater the
maximum in plane principal stress. These points are significant when
developing strategies for calculating strength increases.

166

Multi-Stage Embankment Construction

0.30

0.28

- 0.26
>
b
' 0.24

0.22

0.20
0.00

1.00

2.00
3.00
4.00
Embankment heIght Cm)

5.00

6.00

Figure 7.14a - Variation of sJ& beneath slope.

0.30

0.28

- 0.26
>
b
0.24

0.22

0.20
0.00

1.00

3.00
4.00
2.00
Embankment height (m)

5.00

6.00

Figure 7.14b - Variation of sJc? beneath the crest.

167

Multi-Stage Embankment Construction

0.30

0.28

- 0.26
b
' 0.24

0.22

o.2a
0.00

1.00

3.00
4.00
2.00
Embankment h&ght (m)

5.00

6.00

Figure 7.15a - Variation of sJa'1 beneath the slope.

0.30

0.28

- 0.26
b
(1)

0.24

0.22

0.20
0.00

1.00

2.00
3.00
4.00
Embankment height (m)

5.00

6.00

Figure 7.15b - Variation of s/a' 1 beneath the crest.

168

Multi-Stage Embankment Construction

During the first lift the strength ratios for all points reduce until yielding
occurs. Once the soil has yie'ded the value of sJa1' is dependent on the
ratio qip'. By comparing Figure 7.11 with 7.15 it can be seen that an
increasing qlp' ratio causes an increase in s,/a' and vice versa. Points 0, E
and F are at the critical state, Figure 7.11, with a constant sJc 1' of
approximately 0.28. During the loading stages the undrained shear strength
remains virtually constant. The increase in the ratio is, therefore, a result of
a reduction in the effective stress due to increasing pore pressures.
Reinforcement and Interface Behaviour
The reinforcement strain is plotted at the end of each of the four stages in
Figure 7.16. From the profiles it can be seen that the reinforcement strain is
slightly reduced near vertical drains (at 2, 6, 10, and 14m from toe). This is
due to the soil adjacent to the drains consolidating and gaining strength
more quickly. The reinforcement strain increases during the consolidation
stage. This increase is not caused purely by the differential settlement of the
embankment but also by to an increasing lateral movement, Figure 7.7. The
reinforcement strain profile at failure shows a final reinforcement strain of
over 8%.
The interface elements can be used to plot the shear stress distributions at
the interfaces above and below the reinforcement, Figure 7.17. The sign
convention used to define the direction of the interface shear stresses is
shown in Figure 3.4.
The (upper) reinforcement/fill shear stress profile, Figure 7.17a, shows
significant oscillation of stress, particularly at failure. However, trends can
be discerned from the profiles and in all cases the shear stress on the
reinforcement is outwards for the first 8-lOm beyond the toe,indicating that
the embankment is spreading laterally and inducing a tensile strain in the
reinforcement. Beyond this (nearer the centreline) the embankment

169

Multi-Stage Embankment Construction

0.00
1.00
2.00
3.00
4.00

! :::

7.00
8.00
9.00 +0.00

2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00

Distance from toe (m)

-M- 1 st lift -I-- consol. -3K- 2nd lift -8- faflure

Figure 7.16 - Reinforcement strain.

provides an inward shear stress indicating that the embankment is providing

some restraint to the subsoil which is spreading outwards.
The (lower) reinforcement/subsoil interface shear stress profile is shown in
Figure 7.17b. For the majority of the profile the shear stress on the
reinforcement is in an outward direction, indicating that the soil is moving
outwards relative to the reinforcement At the end of the first lift some of the
interface has reached failure at 3.93kPa. Again variation of the shear stress
across a unit cell can be observed as the shear stress tends to be reduced
near a drain and increased away from it (drains are positioned at 2, 6, 10
and 14m from the toe). After the consolidation stage the interface shear
stress, generally, reduces with none of the soil at failure. For the second lift
and the undrained loading to failure the interface shear strength was
170

Multi-Stage Embankment Construction

20.0
15.0
a

10.0

0
-

I.1UI
I-

a
10.1
15.
20.

0.0

2.0

4.0

6.0
10.0
8.0
Distance from toe (m)

12.0

14.0 16.0

-- 1st lIft -4-- consol. -*- 2nd lift -8- failure

Figure 717a - Shear stress on the reinforcement/fill interlace.

4.0
2.0
a
UI
UI
UI
4UI
L
U
-

4.1

U)

6.
8.
0.0

2.0

6.0
8.0
10.0
Distance from toe (m)

4.0

-a I st lift

- -

12.0

14.0

16.0

consol -*-- 2nd lift s-- failure

Figure 7.17b - Shear stress on the reinforcement/clay interface.

171

Multi-Stage Embankment Construction

increased, according to Figure 7.3, to take account of the increase in the

surface strength of the subsoil. After the second lift the shear stress has, in
general, increased with some of the elements reaching their limiting
strength. The profile plotted at failure was taken from a slightly earlier
increment to avoid the oscillations which occurred during the failure
increment. The shear strength has been reached over almost all of the
interface indicating slip between the reinforcement and the soil. At failure
the shear stress is in an outward direction along the whole of the interface.

7.3. Subsoil Strength Increase

In order to carry out an accurate stability analysis of a multi-stage
embankment it is necessary to make a safe assessment of the increase in
undrained shear strength during the consolidation stage of construction and
then to use these in a total stress limit equilibrium analysis. In this Section
methods for estimating the increase in undrained shear strength and the
assumptions which account for the observed behaviour of soft clay are
discussed. The strength increases predicted by several simple methods are
compared with the increases observed from the idealised two stage
construction finite element analysis. The simple methods which best fit with
the finite element results have then been used to predict strength increases
for the Porto Tolle case history (Chapter 6) and these are also compared
with finite element results.

7.3.1. Simple Calculation Methods

Calculation of Undrained Strength from Vertical Effective Stress
Before construction of an embankment the subsoil will almost always be in
a K0 condition with the principal stresses vertical and . horizontal. As
construction proceeds the stress conditions become more complex with
rotation of the principal stress directions at all positions in the subsoil except
directly beneath the embankment centreline. Figure 7.18 shows a subsoil
172

Multi-Stage Embankment Construction

III

passive
compression

- active
compression
simple
shear

Figure 7.18 - Stress state on a potential failure surface in a soil near

failure subject to embankment loading.
near failure. A potential rotational failure surface is shown, but depending on
the subsoil conditions, the mechanism could equally well be translational
(see Section 4.3.3). From the orientation of the slip surface the principal
stress directions at I, II and III can be inferred, with I undergoing active
compression, II simple shear and III passive extension. This complex stress
distribution is further complicated by the inclusion of vertical drains. From
Figure 7.12 it can be seen that for positions beneath the mid-slope and the
centreline there is a variation of the principle stress rotation as the distance
from a vertical drain increases.
In the laboratory it is possible to perform a number of tests with the principal
stress directions in fixed positions. Figure 7.19 shows results of a series of
such tests on normally consolidated clays and silts (as summarized by
Ladd, 1991), with each line representing the best fit to a number of tests
results. It can be argued that an undrained triaxial compression test
173

Multi-Stag e Embankment Construction

.2 0.40
0

C
a)

0.30

______

U,
I-

0
a)

triaxia] compression
direct smpte shear :

0.20

U)

a)
C

0.10

0
I-

0
C

0.00 L
0.0

20.0

40.0

60.0

80.0

100.0

plasticity index (%)

Figure 7.19 - Undrained strength anisotropy of normally consolidated

clays.
represents the stress state at I, and correspondingly a direct simple shear
test at II and a triaxial extension test at III. To obtain a simple analysis
method an average of these three results would give a measure of the
average strength along the failure surface shown in Figure 7.18. The
average value from the three tests corresponds approximately with the
direct simple shear test result, although this is not the case with varved
deposits which often produce low undrained strengths in simple shear. The
Cam-clay constitutive model implemented in CRISP does not reproduce the
observed strength anisotropy but calculates a constant undrained shear
strength independent of the direction of the principal stresses.
The strength ratio (sJal')NC can be used to provide a safe estimate of the
undrained shear strength of the subsoil at any time providing the value of

a1'

is known. The variation of the strength ratio with embankment height has
174

Multi-Stage Embankment Construction

been plotted for the idealised study in Figures 7.14 and 7.15. From these
plots it can been seen that the initial strength ratio provides a lower bound
to the strength ratio during construction. However, It is difficult to simply yet
accurately estimate the value of ' and an easier quantity to evaluate is the
effective vertical stress ;'. Using this value with the initial strength ratio
must provide a conservative strength as ;'<a1 . The strength at any time
can therefore be calculated from
S()

c,

.......................(7.4.)

NC

Vertical Stress Increase due to Embankment Loading

In order to apply Equation 7.4 it is necessary to calculate the increase in the
effective vertical stress (A;') due to the embankment loading. This involves
firstly, the calculation of the total vertical stress increase (A;) and secondly,
the excess pore water pressure (u) due to incomplete consolidation. In this
Section two methods for the estimation of the total vertical stress increase
are proposed.
Method I (Figure 7.20a) - In this method the load is assumed to act
one-dimensionally, i.e. the loading of the subsoil is due only to the weight of
fill material above. This results in no stress increase, and therefore no
strength increase, beyond the toe.
Method 2 (Figure 7.20b) - This method relies on the use of an elastic
solution to the embankment loading problem (Gray, 1936). The application
of the solution is only strictly valid for elastic conditions. However, the
solution does provide a convenient method of calculation of the vertical
stress increase and has been shown to provide realistic values (Watson et
al, 1984).
These methods provide an upper and lower bound on the sophistication
which is appropriate in this type of calculation. An intermediate calculation
method assumes that a constant load spreading angle can be applied at the
175

Multi-Stage Embankment Construction

(a) method 1

LqO

O<&Tv<ri
I

I
I

I
I

(b) method 2 - Cray (1936)

b

-4

-4

Xct
]

Figure 7.20 - Estimation of vertical stress increase in a soil due to

embankment loading: a) one-dimensional; b) according to Gray (1936).

176

Multi-StaQe Embankment Construction

toe and that this angle reduces linearly towards the centreline. If the load
spreading angle was zero, this method would be equivalent to method I
(above).
Generation of Excess Pore Water Pressure due to Embankment Loading
At the end of a loading stage excess pore pressures wilt have been
generated in the subsoil. For the purpose of this investigation the simplest
possible estimation of the value of these excess pore pressure increase
(Au) has been made, i.e. that the excess pore pressure increase is equal to
the increase in total vertical stress calculated using Grays (1936) elastic
theory
AU=AY V ...............................................................................(7.5)

This approach is only correct beneath the centreline of the embankment as

no account has been taken of the generation of pore pressures due to
shear stress increase. Other simple approaches to the calculation of excess
pore pressure generation have been used. Murray and Symons (1974)
used Skempton's (1954) equation for the pore pressure developed in a
saturated soil under undrained conditions
Au=Ac 3 +A(Aa i_ Ac 3) ..........................................................(7.6)
and assumed that Skemptons pore pressure parameter A=1 at failure for a
normally consolidated soil, so that
Au=Aa i .............................................................................(7.7)
with Aa1 calculated using an elastic solution.
Parry (1972) used a similar approach for the comparison of predicted and
observed pore pressures for an embankment constructed over a soft clay
soil. The excess pore pressure was linked to the stress increment by
Au=aAAai

..........................................................................(7.8)

177

Multi-Stage Embankment Construction

where A is the pore pressure parameter measured at failure for undisturbed

samples in an undrained triaxial compression test, Acy1 is the increment in
the maximum principal total stress calculated using elastic theory and a was
a constant value equal to 1.25 for the embankment studied.
Burland (1972) proposed a method for the prediction of the generation of
excess pore pressures which allowed for local failure of the subsoil. The
method still relied on elastic solutions for the calculation of stresses but
showed encouraging results when predictions were compared with
observations for three embankments constructed over soft clays.
The method used for the calculation of excess pore pressures in this Thesis
is likely to underestimate the actual pore pressures generated under
undrained conditions. Whilst the present method has the merit of simplicity.
further calculations should be carried out to assess the applicability of other
relationships.
Estimation of Dissipation of Excess Pore Water Pressure

The excess pore water pressure generated during loading will reduce
during the consolidation stages. In order to calculate the effective stress
conditions, and thus the undrained shear strength, it is necessary to
estimate the amount of dissipation which occurs. When analysing a subsoil
containing vertical drains it is likely that the flow will be predominantly
horizontal and it may be appropriate to use one of the solutions for a unit
cell reviewed in Chapter 5. For this study Hansbo's (1981) solution has
been used. For the idealised two-stage analysis the vertical permeability
was negligible so that it was possible to apply Hansbo's equation without
consideration of vertical flow. In general the vertical permeability will not be
negligible and soil close to horizontal drained boundaries *ill co'.solidate
more quickly than would be predicted using purely radial flow in a unit cell
calculation. The increased rate of consolidation would lead to surface
strengthening which is important, particularly for reinforced embankments,
178

Multi-Stage Embankment Construction

where the soil strength at the soil/reinforcement interface significantly

affects the failure load of the subsoil. Using Hansbo's equation should
therefore give a conservative estimate of the dissipation of pore water
pressure in this critical region provided the material parameters used, and
assumptions regarding well resistance and smear, are realistic.
Hansbo's equation can be used to calculate the average degree of
consolidation as
Uh =1_e T

......................................................................(7.9)

where Th is the time factor and i is a factor dependent on geometry, smear

and well resistance; these factors are discussed in detail in Chapter 5.
Equation 7.9 can be used to estimate the dissipation of excess pore water
pressure during a consolidation stage. If the time taken to place the fill is
sufficiently short, the consolidation which occurs during construction, is
negligible. However, if the rate of loading is relatively slow a calculation of
the pore pressure dissipation can be made using the equations developed
in Appendix C, which modify Hansbo's equation to allow for ramp loading
- - at(1 Uhl

eT1)

Th

(7.10)

where a is the rate of loading.

7.3.2. Factors Influencing the Strength Increase of Soft Clays

In the previous Sections a method for calculating the strength increase of a
clay soil has been proposed and the method can be summarized as
1 Calculate the increment of total vertical stress in the subsoil using
either a one-dimensional approach or an elastic solution, Figure
7.20.
2 Assume that the increase in vertical stress causes an equal
increment of excess pore pressure at the end of the loading stage.
179

Multi-Stage Embankment Construction

3 Calculate the dissipation of the excess pore water pressure, before

the start of the next loading stage, using Equation 7.9 and/or
Equation 7.10 and combine with the increment of total vertical
stress to calculate the increment of effective vertical stress.
4 Calculate the new undrained shear strength using the increment of
effective vertical stress with Equation 7.4.
This procedure could be used to calculate the strength increase in any clay
subsoil. However, some other aspects of the behaviour of soft clays need
be taken into account to achieve an accurate estimation of the subsoil
strength increase.
Lightly Over-Consolidated Clays
It is unlikely that a soft clay subso(( will be normally consolidated;
over-consolidation could occur for one of several reasons, such as the
removal of overburden, fluctuations in ground water level, dessication of the
clay surface or ageing. The over-consolidation has several important
consequences in relation to the strength increase due to loading of the soil.
In order to evaluate these effects the soil behaviour will be discussed within
a critical state frame work.
The plane strain undrained strength ratio of an over-consolidated (OC) clay
is shown in Appendix A to be
(
ISI \

( \
_lSjl

1+2Ko) OCR1+2Ko,)

ocIvc(1+2KoNc)

(710

(1+2k0)

As the over-consolidation ratio (OCR) is, by definition, greater than I and,

from Appendix A, K>KONc , the over-consolidated strength ratio must be
greater than the normally consolidated ratio, which is often observed to be
0.220.02 (Hight et al, 1987; Ladd, 1991).
When embankment construction commences the soil is reloaded from its
over-consolidated state and modest strength increases will be expected
180

Multi-Sta g e Embankment Construction

until the preconsolidation pressure is reached. For a typical ratio of ic/A.=0.2

the strength increase in the over-consolidated range is only 20% of that in
the normally consolidated range for the same change in vertical stress. It is
therefore a realistically conservative assumption to ignore any strength
increase due to effective stress increases whilst the soil is
over-consolidated. Any soil which remains in the over-consolidated range is
assumed to fail at its insitu undrained shear strength.
A second effect is the increased rate of pore water pressure dissipation in
the over-consolidated range (Tavenas and Leroueil, 1980; Hight et al, 1987;
Leroueil et al, 1991). The change in volume of an over-consolidated clay
compared with that of a normally consolidated clay for the same variation in
vertical effective stress is small. Therefore, for soil loaded in the
over-consolidated range the excess pore pressures generated may be
small.
Variation of the Coefficient of Consolidation
When using Hansbo's (1981) equation to assess the rate of consolidation of
a soil it is implicitly assumed that the coefficient of consolidation is constant.
In reality the coefficient initially varies with depth and also varies with time
as the loading is applied and consolidation occurs. The coefficient of
consolidation is defined as

Chm'y..........................................................................(7.11)
where the coefficient of volume change for the initial one-dimensional
conditions can be defined as m=AJcy,'(1+e), k is the permeability and 're, is
the unit weight of water.
The coefficient of volume change is dependent on both the void ratio and
the vertical effective stress and will therefore vary with depth. For the
idealized case described in Section 7.2, the values of m at the top and
bottom of the soft clay were 9.4x1 0 and 1 .4x1 0 3m2/kN respectively. This

181

Multi-Stacie Embankment Construction

implies an initial coefficient of consolidation almost seven times larger at the

bottom than the top. This effect is offset, to some extent, by the observed
decrease of permeability with deceasing void ratio, Tavenas et at (1983),
proposed the following relationship
logk=logko_()

........................................................(7.12)

where e0 and k0 are the initial void ratio and permeability respectively.
Applying Equation 7.12, to the idealized case, results in a permeability
which is five times greater at the surface than at the base. Combining the
effects of the variation of the coefficient of volume change and permeability
gives a slightly larger coefficient of consolidation at the surface than at the
base.
In the finite element analysis of the idealized case the variation of the
coefficient of volume change was automatically accounted for when using
the modified Cam-clay constitutive model. However, the permeability was
constant throughout the depth. Therefore, the upper layers of the finite
element mesh consolidated more quickly than those lower down.
In general, when performing a finite element analysis of an embankment
constructed on a soft soil it is necessary to model accurately the coefficient
of consolidation throughout the depth of the deposit. To achieve this, using
the modified Cam-clay model implemented in CRISP, it would be necessary
to define several horizontal layers of elements and define a different value
for the permeability in each.
As well as its initial variation with depth, the permeability will also display
further changes with loading. This can be considered using Equation 7.12.
The variation of permeability with stress level has been implemented in
CRISP, Section 3.4, but has not been used in the idealized analysis.

182

Multi-Stag e Embankment Construction

7.3.3. Comparison of the Strength Increases Predicted by

Finite Element and Simple Methods
In this Section the simple analysis methods outlined above are applied to
the idealized embankment and compared with the strength increases
derived from the finite element analysis at the end of the consolidation
stage.
Strength Increases Computed from Finite Element Results
The void ratio, e, for the centroid of each element is included in the output
from the finite element analysis and can be used to calculate the undrained
shear strength in plane strain (Appendix A, Equation A34) as
re0e
M
s=_1ZexpL

..(7.13)

where e0=r-1.
The undrained shear strength increases for each row of elements in each
unit cell are then averaged to give eight values of strength with depth (per
unit cell).
The matching technique (developed in Chapter 5) only ensures that the
average value of the excess pore pressure in the plane strain unit cell is
equivalent to that which would occur in an axisymmetric unit cell. The
variation of the excess pore pressure across the unit cells will be different
(Section 5.3.2) and therefore the undrained shear strength variation across
the unit cell will be different. In order to use the results from a plane strain
finite element analysis to predict undrained shear strengths which are likely
to occur in the subsoil, it is necessary to average the void ratio at various
levels in each of the un!t cells and to calculate the undrained shear strength
from these average values.

183

Multi-Stage Embankment Construction

7.3.4. Strength increases Calculated from Simple Methods

A total of eight combinations of the assumptions outlined in Sections 7.3.1
and 7.3.2 were used to compare simple design methods with the finite
element results. These combinations are summarized in Table 7.1 and the
main assumptions are outlined below
I Calculation of loading - the stress increase in the subsoil was
calculated either using a one-dimensional approach or Gray's (1936)
elastic solution, Figure 7.20.
2 Variation of the coefficient of volume change, m - the value of m
varies significantly with depth in the finite element analysis. Simple
calculations have been carried out with either a constant m or with
the value m varying with depth as in the finite element analysis.
3 Pore pressure dissipation in the over-consolidated range - increased
rates of the dissipation of the excess pore pressure occur in the
over-consolidated range. This observation has been accounted for
in the simple analyses by considering the two extremes of
behaviour. At one extreme, the rate of consolidation is assumed
equal in both the over-consolidated and normally consolidated soil
At the other extreme, immediate dissipation is assumed in the
over-consolidated range. The remaining excess pore water pressure
dissipates during the consolidation period according to Equation 7.5.
In all cases only the vertical effective stress increase above the
preconsolidation pressure has been used to compute the increase in
strength. Positions at which the soil has not reached the preconsolidation
pressure are assumed to have no change in strength.

184

Multi-Stage Embankment Construction

Method

Loading

m varies with Immediate pore pressure

depth

dissipation in OC range

one-dimensional

no

no

Gray

no

no

one-dimensional

yes

no

Gray

yes

no

one-dimensional

no

yes

Gray

no

yes

one-dimensional

yes

yes

Gray

yes

yes

Table 7.1 - Summary of the assumptions used in the simple methods

used to predict strength increases.

Comparison of Simple Methods and Finite Element Results for Strength

Increase
All eight simple methods were used to compute the strength increase in the
unit cell closest to the centreline. At this point the conditions are almost
one-dimensional and comparisons between the simple methods and the
finite element analysis should be most favourable. The calculated undrained
shear strength increases at the end of the consolidation stage are plotted in
Figure 7.21 and the absolute value of strength in Figure 7.22.
From Figure 7.21 it can be seen that the methods which assume a constant
coefficient of volume change with depth (A, B, E and F) do not predict
accurately the trend of strength increase observed in the- finite element
results. As discussed above this result is somewhat artificial, arising from
the variation of coefficient of volume change in the finite element analysis
without the compensating variation of permeability. However, when the
variation of the coefficient of volume change is used in the simple
185

Multi-Stage Embankment Construction

0.0
FE
-Method A

-1.0
-2.0
-3.0

Method B
-aMethod C

-4.0
4-

Method 0

-7.0
-8.0
-9.0
-10.0
0.0

2.0

1.0

6.0
3.0
5.0
4.0
Strength Increase (kPa)

7.0

8.0

Figure 7.21a - Predicted strength increase near centreline; Methods A-D.

0.0
-1.0

FE

-2.0

Method E

-3.0

Method F
-aMethod C

-4.0

-N-

.c
-5.0
4a

Method H

-7.0
-8.0
-9.0
-10.0
1.0

2.0

5.0
6.0
4.0
3.0
Strength increase (kPa)

7.0

8.0

Figure 7.21b - Predicted strength increase near centreline; Methods E-F.

186

Multi-Stage Embankment Construction

0.0
FE
-Method A

-1.0
-2.0
-3.0

Method B
-eMethod C

-4.0
E -5.0

Method 0

-7.0
-B.0
-9 .c
-b.c.
0.0

5.0

10.0

15.0
20.0
Strength (kPa)

25.0

30.0

Figure 7.22a - Predicted strength near centreline; Methods A-D.

0.0
-1.0

FE

-2.0

Method E

-3.0

Method F
-aMethod G

-4.0
-C

-5.0

Method H

a-6.0
-7.0
-8 .a
-9.0
-10.0
0.0

5.0

10.0

20.0
15.0
Strength (kPo)

25.0

30.0

Figure 7.22b - Predicted strength near centreline; Methods E-F.

187

Multi-Stage Embankment Construction

calculations, (C, 0, G and H) it is reassuring that the finite element predicted

variation of strength increase with depth at the centreline can be
qualitatively reproduced. For comparisons of the undrained shear strength
increases over the whole cross section, to be made below, only methods
which allow for the variation of coefficient of volume change with depth will
be considered.
The effect of the method of calculation of the vertical stress due to
embankment loading can be observed by comparing method C with D and
method G with H. Near the surface the predicted strength increases are
identical, but lower down Gray's elastic solution predicts a smaller vertical
stress (and hence strength) increase. Gray's elastic solution was developed
for a semi-infinite half space whereas the finite element analysis considers a
lOm depth of clay overlying a rough rigid layer. A comparison of Gray's
solution with an elastic finite element analysis showed only minor
differences in the predicted vertical stresses, with the maximum difference
occurring at the surface, but in an elasto-plastic analysis of the subsoil the
rigid bottom boundary effect on the vertical stress distribution may be more
significant. In order to provide more general comparisons, the simple
methods were used to calculate strength increases over the whole cross
section employing both the one-dimensional and Gray vertical stress
distributions.
The effect of immediate dissipation of pore water pressure in the
over-consolidated range can be observed by comparing Figures 7.21a and
b. In all cases the analyses assuming immediate dissipation (E, F, G and H)
predicted a larger increase in strength and produced a strength increase
profile which was closer to that predicted by the finite element analysis.
Therefore, only methods which assume immediate dissipation of excess
pore water pressure in the over-consolidated range will be considered
further (i.e. Methods G and H).

188

Multi-Stage Embankment Construction

Following the study of strength increases at the centreline methods G and H

have been used to predict strength increases throughout the subsoil.
Contours of the finite element predicted strength increase and absolute
strength after the consolidation stage are shown in Figure 7.23.
The corresponding strength increase contours for methods G and H are
plotted in Figure 7.24 with the absolute strength in Figure 7.25. In all of the
contour plots it can be seen that the simple methods predict accurately the
trend for the strength increase and absolute strength as defined from the
finite element results. The differences between the finite element and simple
methods for the undrained shear strength are plotted in contour form in
Figure 7.26 (a positive value indicates a lower predicted strength increase
from the simple method than from the finite element analysis).

7.4. Porto Tolle

An analysis of the strength increase predicted for the Porto Tolle case
history, reported in Chapter 6, can be carried out in the same manner as
that described for the idealized case. For the idealized case the simplified
methods which best reproduced the finite element analysis strength
increases were methods G and H, see Table 7.2. These same assumptions
were used to calculate the strength increase at Porto Tolle.
No consideration of the behaviour of the soil in the over-consolidated range
was required as the soil was modelled as normally consolidated.
Comparison of the predicted degree of dissipation of excess pore pressures
using the simplified procedure with those predicted from the finite element
analysis are shown in Figure 7.27a for the end of the consolidation period.
The degree of consolidation near the top of the clay layer is greater in the
finite element analysis. This is due to the vertical flow of pore water which is
not accounted for in the simplified (Hansbo) analysis. At greater depths the
predicted dissipation agrees well, indicating that the variation of the
189

Multi-StaQe Embankment Construction

Figure 7.23a - Contours of the strength increase at the end of the

consolidation stage predicted by the finite element analysis.

5.0

_____- 10.0
0.0
__________

25.O

Figure 7.23b - Contours of the absolute strength at the end of the

consolidation stage predicted by the finite element analysis.

190

Multi-Stage Embankment Construction

Figure 7.24a - Contours of the strength increase at the end of the

consolidation stage predicted using Method G (see Table 7.2).

Figure 7.24b - Contours of the strength increase at the end of the

consolidation stage predicted using Method H (see Table 7.2).

191

Multi-Stage Embankment Construction

5.0

______- 10.0
15.0

20.0

________
25.0

Figure 7.25a - Contours of the absolute strength at the end of the

consolidation stage predicted using Method G (see Table 7.2).

5.0
#
1 u.1u
15.0
-.-----------.--

___
-

25.0

Figure 7.25b - Contours of the absolute strength at the end of the

consolidation stage predicted using Method H (see Table 7.2).

192

Multi-Stacie Embankment Construction

maxlmum=2.96kPa
mlnlmum=-O.47kPa

Figure 7.26a - Contours of the strength difference between Method G

and finite element at the end of the consolidation stage.

madmum=2.47kPa
IP%IWI

i.fl flUfl,n

Figure 7.26b - Contours of the strength difference between Method H

and finite element at the end of the consolidation stage.

193

Multi-Stage Embankment Construction

5.0

10.0

?_15.0
-FE

20.0

Honsbo

25.0

30.0
80.0

85.0

90.0

95.0

100.0

Degree of pore pressure dissipation (%)

Figure 7.27a - Predicted degree of consolidation.

5.0
FE
10.0

15.0

-4Onedimensional
Gray

20.0

25.0

30.01

65.0 70.0 75.0 80.0

85.0 90.0

95.0 100.0

Total vertical stress increase (kPa)

Figure 7.27b - Predicted increase in total vertical stress.

194

MuItl-Stane Embankment Construction

coefficient of volume change with depth, which would have occurred in the
finite element analysis, was correctly reproduced by the simple method.
Figure 7.27b shows a comparison of the total stress increase near the
centreilne, calculated using both a one-dimensional approach and using the
Gray (1936) elastic solution with the average finite element prediction in the
unit cell closest to the centreline. The one-dimensional approach predicts a
uniform stress increase equal to the applied load, and is increasingly in
error with depth. The elastic solution represents more accurately the profile
of total stress increase with depth, but still exceeds the finite element
prediction.
After combining the predicted degree of consolidation (Figure 7.27a) and
the increase in total vertical stress (Figure 7.21 b) to predict the increase in
effective vertical stress, the undrained strength ratio, (s/a')=O.272, can
be used to predict the strength increase. These predictions are compared
with the undrained shear strengths calculated from the finite element
analysis near the centreline in Figure 7.28a. The underestimation of the
degree of consolidation causes the simplified method to underestimate the
strength at the surface. The one-dimensional calculation of total stress
increase (Method G) results in a large error in the predicted strength.
However, the superior total stress predictions using the elastic solution
(Method H) result in a better prediction of the strength increase and produce
the observed trend at the centreline.
The Gray solution has been used to calculate the stresses throughout the
soft clay layer and predictions of strength increase from both the finite
element analysis and the simplified method are presented in Figure 7.29. A
comparison of the two predictions is made in Figure 7.30 where the strength
calculated using the simplified method has been subtracted from those
predicted in the finite element analysis. The maximum error on the safe side
is 14.1 OkPa and this is due to the simplified method neglecting the effect of
vertical permeability. The maximum unsafe error is -3.37kPa and is caused
195

Multi-Stage Embankment Construction

5.
E
lethod G

ia

lethod H
?15

20

25

30 .0
16.0

18.0

24.0
22.0
20.0
sfrengfh increase kPa)

26.0

28.0

Figure 7.28a - Predicted undrained shear strength increase

5.0
FE
10.0

Method C
Method H

15.0

20.0

25 .a

30.0 -I--
40.0

50.0

60.0

70.0
80.0
Strength (kpa)

90.0

100.0

Figure 7.28b - Predicted absolute undrained shear strength.

196

Multi-Stage Embankment Construction

Figure 7.29a - Contours of the strength increase at the end of the

consolidation stage predicted by the finite element analysis of Porto Tolle.

Figure 7.29b- Contours of the strength increase at the end of the

consolidation stagepredicted by the simplified method at Porto Tolle.
197

Multi-Stag e Embankment Construction

max1mum 14.1 kPa

mInImum-3.37kPa

_)

Figure 7.30 - Contours of the strength difference predicted by the finite

element analysis and the simplified method at the end of the
consolidation stage at Porto Tolle.

by the failure of the elastic solution to accurately predict the vertical stress
increase. The unsafe error is relatively small and occurs only at a depth at
which failure mechanisms are unlikely due to the increase of strength with
depth.

7.5. Summary
The idealized study has provided information on the behaviour of a
multi-stage construction of an embankment over a soft cohesive deposit. As
expected analysis showed that multi-stage construction allows a
significantly higher embankment to be constructed than would be possible
in a single lift.
Comparison of limit equilibrium and finite element analysis has proved
inconclusive. A higher factor of safety was found from the finite element
198

Multi-Stage Embankment Construction

analysis. This was due, in part, to the consolidation which occurred during
construction, in the finite element analysis, producing strength increases
that were not allowed for in the (undrained) limit equilibrium analysis. Also
the method by which the embankment was brought to failure in the finite
element analysis was different from that used in the limit equilibrium
analysis.
The undrained shear strength increases for the idealized multi-stage
analysis have been calculated from the void ratios predicted by the finite
element analysis. Simplified calculations of strength increases have also
been made and comparisons of these methods with the finite element
analysis have shown reasonable agreement. The simplified strength
increase calculation procedures have also been applied to the Porto Tolle
case history, Chapter 6. Agreement for this case was not as good, but the
size of the error produced still indicates that the simple procedures for the
prediction of undrained shear strength increase may be used.
The good agreement achieved despite a major assumption made in the
simple methods regarding the generation of excess pore pressure during
loading. The proposed procedure used a one-dimensional approach
ignoring the pore pressures which would result from distortion. Such a
method would underestimate the excess pore pressure generated and,
therefore, would overestimate both the effective stress and the undrained
shear strength. Further investigation of this aspect is needed.

199

Conclusions and Further Work

8. Conclusions and Further Work

8.1. Objectives and Completed Work
The research reported in this thesis had several objectives;
To modify an existing finite element computer program so as to better
model rein forced embankments constructed over soft clay subsoils
involving vertical drains. These modifications involved the
incorporation of three additional elements to model: 1) the
reinforcement, 2) the reinforcementlsoil interface and 3) the vertical
drains. A facility to model the variation of permeability with stress level
has been incorporated for use with the modified Cam-clay constitutive
model.
To benchmark the program by comparing finite element analyses with
analytical solutions. These comparisons were made for both the
undrained failure of reinforced and unreinforced subsoils and for the
consolidation of soil around a single vertical drain.
To compare the results of finite element analysis of well documented
case histories with observed behaviour. The case history analysed
was a pre-loading embankment constructed over a soft normally
consolidated clay improved using vertical drains.
To develop simplified design procedures for the effect of both
reinforcement and vertical drains upon embankment stability. This was
divided into two parts. Firstly, a design method for single stage
embankments, built with a constant side slope, based on published
plasticity solutions was developed. Secondly, for the design of
multi-stage constructed embankments a design philosophy was
proposed in which undrained shear strengths were calculated at the
start of each loading stage. These strengths were then used in a total
200

Conclusions and Further Work

stress limit equilibrium analysis to assess the factor of safety of the

embankment. Methods of estimating the undrained strength increases
were investigated.

8.2. Conclusions
The following conclusions have been drawn from the results presented in
the previous chapters.

8.2.1. Undrained Collapse of Cohesive Soils

Plasticity solutions for the undrained collapse of rigid strip footings on
cohesive subsoils have been compared with a series of finite element
analyses so as to confirm the ability to numerically model such problems.
Once confidence in the finite element program was established, it was then
used to model similar embankment problems for which analytical solutions
were not available. Two relevant idealized subsoils were considered: firstly,
a subsoil with strength increasing linearly and indefinitely with depth and,
secondly, a subsoil with constant strength over a limited depth overlying a
rough rigid layer.
The displacement controlled finite element analysis has been shown to
be in good agreement with the plasticity solutions. Analyses were
carried out using both elastic-perfectly plastic (with a Tresca yield
criterion) and modified Cam-clay constitutive models.
Load controlled finite element analysis of the same idealized subsoils
has shown that the surface settlement profile does not significantly
affect the ultimate surface stress distribution. This conclusion is
relevant to the application of plasticity solutions, which assume a rigid
footing, to the design of embankments which are not rigid.
The interface element has been shown to significantly improve the
quality of the solution when analysing problems of this type containing
201

Conclusions and Further Work

rough boundaries. The improvement in the solution cannot be

matched by using a mesh with additional two dimensional elements
with a similar number of degrees of freedom.
The displacements predicted from the finite element analyses have
been shown to be in good agreement with those predicted from
plasticity solutions for a subsoil of uniform strength and limited depth.
The mechanism of failure in the two types of subsoil considered has
been shown to agree with the mechanisms predicted by plasticity
theory. For a uniform strength and limited depth subsoil the failure is
predominantly translational, and for a subsoil with strength increasing
with unlimited depth the mechanism is rotational.
A design procedure has been validated, using finite e(ement anatyses,
in which the plasticity solutions can be used to predict the safe uniform
side slope length for a single lift embankment under undrathed
conditions.

8.2.2. Consolidation of Soil Around Single Vertical Drains

The behaviour of the soil around a single vertical drain has been studied
and compared with analytical solutions.
Comparisons between analytical and finite element results show that
the drainage element implemented in CRISP correctly models a single
vertical drain in an axisymmetric analysis.
The predicted analytical trends for drains with we'l resistance and soil
smear, caused by drain installation, have been correctly reproduced.
A procedure for analysing a soil containing vertical drains in a plane
strain finite element analyses has been developed. This matching
procedure has been shown to produce good results for an elastic soil
in a variety of conditions, which include the effects of smear, well
resistance and vertical pore water flow in the soil.
202

Conclusions and Further Work

8.2.3. Porto Tolle Case History Analysis

A case history of an embankment constructed over a normally consolidated
clay in northern Italy has been used as a case history for which finite
element analyses could be carried out and compared with observed
behaviour. All material parameters chosen for the analysis were obtained
from published data at the outset and no attempt was made to revise the
parameters subsequently.
A series of matching analyses was performed in which the rates of
consolidation of axisymmetric and plane strain unit cells were
compared. In all cases the rate of consolidation of both unit cells was
in good agreement. These analyses indicated that the matching
procedure is suitable for more realistic soil conditions, in which
material properties vary with depth and plastic deformations occur.
A full plane strain analysis of the Porto Tolle subsoil was carried out
The vertical drains were installed at a spacing derived from the
matching procedure. Comparisons of finite element and observed
displacements were in good agreement The predicted and observed
excess pore pressures near the embankment centreline were in
reasonable agreement during construction of the embankment
However, during the consolidation stage differences in the observed
and predicted behaviour occurred with the finite element excess pore
pressures almost completely dissipating whilst the observed values
showed little dissipation. Jamiolkowski and Lancellotta (1984)
suggested that the observed piezometer data were in error due to the
presence of organic gas in the subsoil. This would explain the
discrepancy between observed and predicted behaviour but it is also
important to note that other phenomenon, such as destructuring of the
subsoil, may have contributed to this difference.

203

Conclusions and Further Work

8.2.4. Idealized Two-Stage Construction

In order to assess the finite element modelling of multi-stage embankment
construction an idealized two-stage reinforced embankment was analysed.
The material parameters and geometry were chosen to give a situation
which is typical of such embankments in practice.
The analysis confirmed that the height of an embankment could be
increased above its single stage height by using a multi-stage
construction approach, making use of the increase of undrained shear
strength during consolidation stages.
Comparisons of the finite element predicted factors of safety and limit
equilibrium factors of safety were inconclusive. The difference in the
assumptions for the two ana'yses made comparison difficult as
consolidation occurred during the finite element construction stages
and the method by which the embankment was made to fail was not
the same in each case.
The predicted lower interface shear stress distribution indicated that
slip occurred along the entire underside of the reinforcement at failure.
Reinforcement strains of over 8% were predicted at failure.
Stress paths at various positions in the subsoil reflected expected
behaviour and agreed with results of previous research.
A simplified procedure for the prediction of the increases in undrained
shear strength during stage construction has been proposed. The
result of applying this procedure agreed well with the finite element
predicted strength increases for the idealized subsoil. The procedure
was also used to predict the strength increases for the Porto Tolle
case history and the agreement with the finite element predictions was
reasonably good considering the simplifications made.

204

Conclusions and Further Work

8.3. Suggestions for Further Work

8.3.1. Further Modifications to CRISP
The interface element has been implemented as a total stress
element. This limits use of the element as it cannot be used to
represent an drained frictional interface correctly and cannot predict
strength increases for consolidating soils. An effective stress
modification to the present element would allow more accurate
predictions of collapse after consolidation of soft clay subsoils.
Several aspects of the behaviour of soft clays are neglected in the
current constitutive models in CRISP; their inclusion would lead to
better modelling of these materials and improved predictions of
embankment behaviour. Such aspects include strength anisotropy and
the present model could be modified to better represent the varaIEon
of strength along a failure surface in an embankment analysis. A
second problem with the modified Cam-clay model is the high value
for the coefficient of earth pressure, K0 , which must be used to
accurately model one-dimensional conditions. Also, many soft clays
exhibit brittle behaviour and strain softening; this aspect is not
modelled in CRISP but is likely to be important in the practice. No
attempt has been made to model the creep behaviour of the soft clays;
creep deformations can be significant but are not considered in the
current version of the program.
CRISP uses an incremental solution procedure. The size of the
loading increments is hard to assess before analysis. A suitable
increment size can only .be found by performing several analyses, with
increasing numbers of increments, and monitoring the displacements
predicted. Iterative techniques can provide more efficient solutions and

205

Conclusions and Further Work

eliminate the need to perform sensitivity studies on the number of

increments.

8.3.2. Matching Procedure

The matching procedure has not been fully tested for all soil
conditions. Additional analyses of over-consolidated soils and
situations were the lateral displacements of the soil are larger should
be carried out.
The matching of the undrained strength increases in plane strain and
axisymmetry have not been fully investigated. Analyses in which the
increase in shear strength and comparisons of failure loads of unit
cells could be carried out to verify the procedure.

8.3.3. Simplified Design Procedures

Comparison of finite element, limit equilibrium and plasticity analyses
for single stage construction are in good agreement and indicate that
any of the three methods is suitable for design. However, further
research is required before similar simplified design methods for
multi-stage embankments can be recommended. Comparisons of
finite element and limit equilibrium analyses of loading stages must be
performed to ensure that limit equilibrium analyses produce realistic
results. Further investigation into the simple prediction of undrained
shear strength is required including the generation of pore pressures
during loading, the applicability of analytical solutions, such as Hansbo
(1981), for elasto-plastic soils and calculations of the increase in
stresses due to embankment loading.

206

ApDendix A

Appendix A - Strength of Modified Cam-clay

In this Appendix equations for the undrained shear strength, in both plane
strain and axisymmetry, are developed within the modified Cam-clay
framework (Roscoe and Burland, 1968).

A.1. Relationship Between the Stress Invariant q and the

Undrained Shear Strength s
Definitions of the generalised stress invariant, q, used in critical state and
the maximum shear stress at failure must be found and related in order to
derive definitions for the undrained shear strength of modified Cam-clay.
The generalised deviator stress is defined in terms of the pricipal stresses
as

(Al)
where a1>a2>a3.
The undrained shear strength is defined as
Su =

1"
, a1

G3

.....................(A2)

Triaxial Conditions
In a triaxial test a2 =a3 , therefore from Equations Al and A2

SU

..............................................................................(A3)

where the subscript TX represents triaxial conditions.

207

ADoendix A

a'

Figure Al - Definition of Mohr Coulomb Yield surface for a cohesive

material.
Triaxial Conditions
In a triaxial test the stress invariants are

p'=(ci 1

+2ci3')13 and q(cT 11-a31 ), SO

that
3(a _ci)
M=..................................................................... (A9)
( +2a)
rearranging Equation A9 and combining with Equation A7
MTX =

6sin'(AID)
3sin4

Plane Strain
Using the definition of the stress invariant q from Equation Al and
p'(a 1 +a21+a3 )13 and substituting for the intermediate principal stress
defined in Equation A5 it can be shown that
/(a _a)
.......(All)

MpL =
(

+)
209

ADnendix A

N
N

ln(p')

Figure A2 - Undrained stress path for a normally consolidated modified

Cam-clay sample.
combining with Equation A7
IVIPL

if5

dPL ................................................................... (Al 2)

A.3. Undrained Shear Strength of a Normally Consolidated

Modified Cam-clay
Figure A2 shows a sample of I( normally consolidated clay at position J the
sample is then sheared undrained until failure at L. As the sample is
normally consolidated at J it must lie on the yield surface
..(A13)
rearranging and substituting i'=qIp'

210

ApDendix A

(p;' (+M

2M2

From geometry and using the critical state gradient for the normal
consolidation lines and critical state lines of ?. and the gradient of the
swelling line K
i. i\ f ,\A
I PL I - I Pic

s.

.......................(P15)

iII I
j I PJ

where A=Q.K)/
Combining Equation A14 and A15
(P16)
Point L is at the critical state, therefore q=Mp', so that
=

A/7pj(tIj2

02)

qI.

A(P17')

Equation A17 defines the deviator stress at failure in terms of the initial
stresses for any normally consolidated modified Cam-clay sample subject to
undrained loading.
Triaxial Conditions
Substituting Equation A3 into Equation A17

SurxMr4'{ 2M

(/1 8)

As the sample is K0 normally consolidated at J, the inital stress invariants

are p'=a'(l+2K)I3 and q'(l-K), where K is the coefficient of
earth pressure at rest for the normally consolidated sample and a' is the
vertical preconsolidation pressure, therefore Equation A18 becomes

211

App endix A
____

S UTX M

(1 +2KONC)((
6

Ovc

)2

I KONC
1 + 2KONC 2M 2

.(A19)

Plane Strain Conditions

Combining Equation A6 with Equation A17
P'J

SUPL=MPL-J. 2M,

(420)

L )

again substituting for initial conditions

A

SUpL

M (1+2I<'0) ( 1KONC 2 g
L(1+2KONC) 2ML)(A21)
3J

A.4. Undrained Shear Strength of an Over-Consolidated

Modified Cam-clay
Figure A3 shows a K0 normally consolidated sample at A and a lightly
over-consolidated sample at B. If the sample undergo undrained shearing
they would reach the critical state line at A* and B*.
From the defined geometry and gradients of the swelling and compression
tines

()

= ()...................................................................( A22)

The over consolidation ratio in terms of mean effective stresses is defined

as

..........................................(/\23)
and at the critical state q.=Mp'. and q.=Mp'., Equation A22 can be
rewritten as

212

Appendix A

VB
VA

In(p')

Figure A3 - Undrained stress paths for normally consolidated and lightly

over-consolidated modified Cam-clay samples.

()

............(1\.24)

= ( i)

The coefficient of earth pressure for the normally consolidated and

over-consolidated samples, KONC and K. respectively, can be subsituted into
Equation A24
+ 2K0NC)

3q9.
[ (1+2Kn)ci..a ] = [ (1+2KnNr)a11A

(1+2Kn)

c4A]

(A25)

aVB

now the overconsolidation ratio, OCR, is defined as

.................................................(A26)
avB

Substituting Equation A26 into Equation A25 and rearranging

213

AnDendix A

(qB.'

_(qA.'i'

Hr
..aVB)

(i ^2K0)

1OCR(i +2KONc)1

avA4(1+2KnN()J

(1+2K0)

J ...................

(A27)

An identical equation to Equation A27 can be derived for heavily

over-consolidated samples where the initial state is on the dry side of the
critical state line.
Triaxial Conditions
Point A in Figure A3 is normally consolidated and substituting Equation A3
Into Equation A27
(S(jrx' (SUTX'

, I

I(1+O)

IULt((1+2IcoNc)'1

'avioc= 1 \aVc1NCL(1+2KflN()JL (1+21(n)

i

(A28)

........
where the subscripts OC and NC represent over-consolidated and normally
consolidated states respectively
Plane Strain Conditions
Similarly for plane strain conditions substituting Equation A6 into Equation
A27
(SPL) _.(SUPL) ((

+)

NC(l+2Kiwt)J[

(1+21(n)

.......

(A29)

A.5. Calculation of the Undrained Shear Strength of Modified

Cam-clay From Void Ratio Data
The finite element analysis produces values for the void ratio during a
modified Cam-clay analysis. From this data the undrained shear strength
can be derived as follows.
For any stress state if the specific volume, v, is known then at the critical
state
v =1'?.ln(j,)

.................................................................... (ft30)
214

A pp endix A

where the subscript cs represents critical state conditions.

Rearranging Equation A30
p1cs=exp[rv]

(A31

At the critical state q 1 =Mp',, therefore Equation A31 becomes

qcs=Mexp[1jV]( A32)
Substituting for triaxial conditions, from Equation A3, Equation A32
becomes
sul)c=fexp[l'jt]

(A33)

Substituting for plane strain conditions, from Equation A6, Equation A32
becomes
suPL =,fJexP[ r'v]

........................................................

(A34)

A.5. Calculation of the Coefficient of the Earth Pressure at

Rest
Calculation of the undrained shear strength, Equations A19, A21, A28 and
A29, requires the calculation of the coefficient of earth pressure at rest.
Calculations are outlined below for both normally and over consolidated
soils.
Coefficient of Earth Pressure at Rest for a Normally Consolidated Soil
(KoNc)

Empirical relationships for the coefficient of earth pressure at rest have

been proposed for normally consolidated clays. For example Jaky (1944)
KONC -1sin4 ' ...................................................................( A35)
215

ADgendix A
From critical state theory it is also possible to derive an equation which
ensures zero strains in the horizontal directions for an increment of vertical
load, as
KoNC

=:011c................................................................ (A36)

where
11ONC(1+v')(lIt) ^ 3iA

3(1_2v')

=1 .....................................(A36a)

M2-11ONC

In the current research a value of G=oo has been used so that Equation
A36a becomes
_3A+JA2+4M2

....................................................(

Equations A35 and A36 can be applied for the Porto Tolle case history
(Chapter 6) where 4,'32, M=0.92 and A0.8. The coefficients of earth
pressure at rest for the normally consolidated soil are 0.47 and 0.74 for the
empirical and theoretical relationships respectively.
In all analyses presented in this Thesis the theoretical relationship, Equation
A36 has been used to calculate Analyses of unit cells with insitu
stresses defined using this relationship and loaded one-dimensionally
produced correct stress paths. However, comparative analyses using the
empirical relationship produced unrealistic stress paths.
Coefficient Of Earth Pressure at Rest for an Over-Consolidated Soil (Is)
Several empirical relationships have been suggested for the coefficient of
earth pressure at rest for over-consolidated soils, for exam Ii Schmitt (1966)
Ko =KQNC(OCR) ................................................................ (A38)

where OCR is the over-consolidation ratio and a is an empirical constant.

216

Acrnendix A

Wroth (1975) proposed a theoretically correct relationship based on the

assumption of elastic unloading
KO=(OCR)KONC _ ' (OCR_I)
I uA

(139)

In which v' is the Poissons ratio.

For the lightly overconsolidated soils, considered in this Thesis, Equation
A39 has been used to calculate the coefficieint of earth pressure at rest
However, it is noted that as the over-consolidation ratio (OCR) becomes
higher the predicted coefficient, using this relationship, overestimates the
observed value which are better represented by Equation A38.

A.6. Triaxial Versus Axisymmetric Conditions

The equation derived in this Appendix for triaxial conditions have assumed
a2 =a3 . The equation have been applied to axisymmetric conditions where

in general a2' is not equal to a3 . However in the analyses presented this

difference has been negligible.

217

ADDendix B

Appendix B - Consolidation of a Plane Strain

Unit Cell
Average Degree Of Consolidation
A plane strain unit cell of width B, shown in Figure BI, has a central vertical
drain of zero thickness and a vertical discharge capacity Q., . Hansbo's
(1981) Theory can be adapted for such a cell as follows.
If v represents the rate of pore water flow towards the drain and i is the
hydraulic gradient at distance x from the drain then from Darcy's law

vx= k!

.................................................................................(BI)

where k is the horizontal soil permeability.

Equation BI may be rewritten as

V- k u
where is the unit weight of water and u the excess pore pressure.
Consider a horizontal slice of soil of thickness dz. The flow in the slice at
distance x from the drain is equal to the change in volume within a block of
soil of width (B-x), so that

V=

(B_x) dz .................................................................... (B3)

where g is the strain in the z direction

Substituting Equation B3 into Equation B2 and rearranging
(B4)
218

Aooendlx B

drain

2dz

Figure BI - Plane strain unit cell.

Consider next the corresponding slice of drain. As only vertical flow occurs
in the drain, the change of flow from the entrance to the exit of the slice,
dQ11is

dQ1 =(Q-1) dzdt

Yw 8z x=O

.......................................................... (B5)

The horizontal inflow to the slice from each side, dO2, is

dQ2 = 7w

ax

dzdt

.....(B6)

For continuity
219

ADDendix B

dQ1 +2dQ2

=0

.....................(B7)

Substituting for dQ1 and dQ2 into Equation B7

dzdt+

1w az2

1w

xo

dzdt= 0

...................................... (B8)

Rearranging
Qw(2u -o

(u

X)o2kZ2)

B9

Substituting Equation B4 into Equation B9

&i'wac Qw(o2tI

+)

=0

(BlO)

x=O

Rearranging

( 2 u

- 2By
Qw t

Bli

Integrating Equation Bil with depth and introducing the boundary

conditions that at z=0, u=0 and at z=L, =0
(u)= 2Yw(&_)

(B12)

Integrating Equation B4 and introducing the boundary conditions defined by

Equation B12.
U=i[(2BX_X2)+(2&_Z2)]

(B13)

Let be the average excess pore pressure across the section at depth z, so
that
tIB =Judx

......................................................................... (B14)

Substituting u from Equation B13 into Equation B14 and integrating

220

Aooendix B

vB2Yw[2+2k(2&z2)]

.(B15)

Now
= (TI

- -(TI v.-.j.

where m is the coefficient of volume change

Substituting Equation B16 into Equation B15

B27wm4^2k(2&_Z2)]1

(B17)

Integrating Equation B17 and applying the boundary condition that at t=O,
11= v.

(B18)

_____

Rearranging

t=_B2;;mvin(&)

........................................................ (B19)

where

[+ B2 (2Lz_z2)]

..................................................... (B20)

Introducing expressions for the time factor

Th =

............................................................................ (B21)

and the coefficient of consolidation

Chmky........................................................................... (B22)
Equation B19 can be rearranged as

221

Appendix B

v = ti o exp[_ Th ]

............ (B23)

Therefore, the average degree of consolidation at depth z is given by

(B24)
Ratio of Excess Pore Pressure in Axisymmetric and Plane
Strain Unit Cells
Consider a drain without well resistance or smear. It can be shown
(Hansbo, 1981) that the excess pore pressure at a distance r from the
centreline in an axisymmetric unit cell of radius R is

(L) _(r2;r)]

In

(B25)

Similarly, for a plane strain unit cell of half width B the excess pore pressure
at a distance x from the centreline is given by Equation B13 which for a
drain with no well resistance (Q - ) becomes
u=[8x_]

............................................................ (B26)

Expressing Equation B25 and Equation B26 as a ratio

2
4
(,(r'
k,,, R2 In)+5

U ,k2B2

X2
-

2B2

If the axisymmetric and plane strain unit cells are matched using geometry
matching (Section 5.3) then

B= RJ(In(n)_)

(B28)

and

222

App endix B
.(B29)

k,=k

Substituting Equation B28 and Equation B29 into Equation B26

- L 'r) Upl -

,2

2R2

r.1
^ 2R2]

(B 30)

[3(ln(n)_)][j.]

Substituting nR/rand introducing the geometric ratio, a, where

a= .L. =

................................................................ (B31)

Equation B30 becomes

[2ln(na)_a2+t]

(B32)

[3(ln(n)_)][2a_a2]
At the periphery of the unit cells a =1 and ignoring the small term (1/n2)
u
UpS -

21n(n)-1

........(B33)

3(ln(n)_)

223

AnDendix C

Appendix C - Development of Pore Pressures

due to Ramped Loading
Hansbo (1981) developed an analytical solution for the consolidation of a
soil cylinder by radial flow towards a central vertical drain.
LJh

= uoe(T1)(c1)

where u0 is the initial excess pore pressure throughout the soil cylinder, Th is
the time factor for radial consolidation and

is a parameter dependent on

the radius of the drain and soil cylinder, the well resistance and the amount
of smear which has taken place on installation.
A similar equation for plane strain conditions was derived in Appendix B,
Equation B24. Both equations assume that a load has been applied
instantaneously to the top of the unit cell so that no dissipation of pore
pressure has taken place.
In practice loads cannot be applied instantaneously and a certain period for
application must be allowed. In this Appendix Hansbo's analytical solution is
modified to allow for the effect of a load applied at a constant rate, i.e. ramp
loading.
Figure Cl shows a load applied at a constant rate over of a time T so that
the final load is Q. At any time, t, the applied load, q, can be defined as
q =t

................................................................................. (C2)

During appliction of the load it is necessary to calculate the excess pore

pressure developed, at say time t1.
Consider an infinitely small time increment from t to t+dt, during this time the
load has increased from q to q+dq, as this is an infinitely small time the
loading is undrained. So that from Equation C2 the increase in excess pore
pressure is
du=dq=dt

....................................................................... (C3)

The excess pore pressure can then dissipate for a time t1-t.
224

ADDendix C

dq

dt

ti

Figure Cl - Ramp loading scheme.

From Equation Cl and substituting cx=8c,/D 2j.t and considering the
increment of load we have
dtih

= (du) e

r0

................................................................. (C4)

Substituting for du from Equation C3 into Equation C4

di

h = . e(10r0 dt

...................................................................(C5)

The excess pore pressure developed at time t1 can be found by integrating

Equation C5 from t0 to t=t1

U,, =

e"')dt .................................................................(C6)

therefore
U,, = -(i - ...................................................................(C7)
Equation C7 defines the excess pore pressure developed by a ramp
loading. If a series of ramp loadings and consolidation periods are used the
Equation C7 and Equation Cl can be alternated using the calculated values
for the average excess pore pressure as input to the next stage.

225

References

References
I Abid, M.M. (1980). The consolidation behaviour of laminated clays using
the finite element method. PhD Thesis, Sheffield University.
2 Abid, M.M. and Pyrah, l.C.(1990). The consolidation behaviour of finely
laminated clays. Computers and Geotechnics, Vol. 10, PP. 307-323.
3 Almeida, M.S.S. (1984). Stage constructed embankments on soft clays.
PhD Thesis, University of Cambridge.
4 Almeida, M.S.S, Britto, A.M, and Parry, R.H.G. (1985). Centrifuged
embankments on strengthened and unstrengthened clay foundations.
Gotechnique, Vol. 35, pp. 425-441.
5 Almeida, M.S.S, Britto, A.M and Parry, R.H.G. (1986). Numerical
modelling of a centrifuged embankment on soft clay. Canadian
Geotechnical Journal, Vol. 23, pp. 103-114.
6 Atkinson, J.H. and Bransby, P.L (1981). The mechanics of soils - An
introduction to critical state soil mechanics. Pub: McGraw Hill.
7 Atkinson, J.H. (1981). Foundations and slopes. Pub: McGraw Hill.
8 Barron, R.A. (1948). Consolidation of fine-grained soils by drain wells.
Trans. ASCE, 113, Pp. 718-742.
9 Bassett, RH. (1987). Original design of the trial embankment.
Prediction Symposium on a Reinforced Embankment, King's College,
London.
10 Bassett, RH. and Guest D.R. (1990). Model and analytical
comparisons of the behaviour of reinforced embankments on soft
foundations. Proc. Int Reinforced Soil Conf., Glasgow, PP. 461-467.
11 Blot, M.A. (1941). General theory of three dimensional consolidation. J.
Appi. Phs. Vol. 12, PP. 155-1 64.
12 Bishop, A.W. (1955). The use of slip circle in the stability analysis of
slopes. Geotechnique No. 5, P p. 7-17.
13 Bishop, A.W. (1966). The strength of soils as geotechnical materials.
Gotechnique, Vol. 10, pp. 91-1 28.

226

References

14 Bishop, A.W. and Bjerrum L. (1960). The relevance of triaxial the test
to the solution of stability problems. Proc. Research Conf. on Shear
Strength of Cohesive Soils, ASCE, pp. 437-501.
15 Bonaparte, R. and Christopher, B.R. (1987). Design and Construction
of reinforced embankments over weak foundations. Transportation
Research Record 1153, pp. 26-39.
16 Bond, A. (1984). The behaviour of embankments on soft clay. MSc
Dissertation, Imperial College, London.
17 Booker, J.R. and Small, J.C. (1975). An investigation of the stability of
numerical solutions of Blot's equations of consolidation. Int. J. Solids and
Structures. Vol. 11, pp. 907-911.
18 Boultrop, E. and Holtz, R.D. (1983). Analysis of embankments on soft
ground reinforced with geotextiles. Proc. 8th ECSMFE. Vol. 1, pp. 469-472.
19 Britto, A.M and Gunn, M.J. (1987). Critical state soil mechanics via
finite elements. Pub: Ellis Horwood.
20 Brown, C.B. and King, I.P. (1966). Automatic embankment analysis:
equilibrium and instability conditions. Geotechnique, Vol. 16, No. 3, pp.
209-219.
21 Burland, J.B. (1972). A method of estimating the pore pressures and
displacements beneath embankments on soft, natural clay deposits.
Proceedings of the Roscoe Memorial lecture, pp. 505-536, Pub: G.T. Foulis
and Co.
22 Burland, J.B. (1990). On the compressibility and shear strength of
natural clays. Gotechnique Vol. 40, pp. 329-378.
23 Clough, R.W. and Woodward, R.J. (1967). Analysis of embankment
stresses and deformations Journal of Soil Mechanics and Foundations
Division ASCE , Vol. 9, SM4, pp. 529-549.
24 Cook, R.D. (1981). Concepts and applications of finite element
analysis. Pub: Wiley.
25 Croce, A., Calabresi, G. and Viglianni, C. (1973). In situ investigation
on pore pressures in soft clay. Proceedings of the 8th ICSMFE, Moscow,
Vol. 2.2, pp. 53-60.
26 Davis, E.H. and Booker, J.R. (1973). Effect of increasing strength with
depth on bearing capacity of clays. Gotechnique Vol. 23, pp. 551-563.
227

References

27 Delmas, P., Magnan, J.P. and Soyez, B. (1987). New techniques for
building embankments on soft soils. Embankments on Soft Clays, Chapter
6, Bulletin of the Public Works Research Centre, Athens, Greece.
28 Dluzewski, J.M. and Termaat, R.J. (1990). Consolidation by finite
element method in engineering problems. 2" European Speciaflity
Conference on Numerical Methods in Geotechnical Engineering, pp.

213-222.
29 Duncan, J.M. and Chang, C.Y. (1970). Non-linear analysis of stress
and strain in soils. J. Soil Mech. Fdns. Div., ASCE, Vol. 96, SM 5, pp.
1629-1 653.
30 Duncan, J.M., Schafer, V.R., Franks, L.W. and Collins, S.A. (1987).
Design and performance of a reinforced embankment for mohicanville dike
No. 2 in Ohio. Transportation Research Record 1153, pp. 15-25.
31 Fellenius, W. (1927). Erdstatische berechnungen mit reibung and
kohaesion, Ernst, Berlin.
32 Fowler, J. (1982). Theoretical design considerations for
fabric-reinforced embankments. Proc 2 mt. Con?. Geotextiles, VoJ. 3, Las
Vegas, pp. 665-670.
33 Garassino, A., Jamiolkowski, M., Lancellotta, ft and Tonghini, M.
(1979). Behaviour of pre-loading embankments on different vertical drains
with reference to soil consolidation characteristics. Proc. 7th European Conf.
Soil Mech. Brighton Vol. 3, pp. 213-21 8.
34 Goodman, R.E., Taylor, R.L. and Brekke, T.L (1968). A model for the
mechanics of jointed rock. J. Soil Mechanics and Foundation Div. ASCE,
Vol 94, pp. 637-659.
35 Gray, H. (1936). Stress distribution in elastic solids. Proc. 1 mt. Conf.
Soil Mech. and Found. Eng., Vol. 2, p. 157.
36 Griffiths, D.V. (1985). The effect of pore fluid compressibility on failure
loads in elasto plastic soils. mt. J. Numerical and Analytical Methods in
Geomechanics, Vol 9, pp. 253-259.
37 Gunn, M.J. and Britto, A.M. (1982). CRISP: users' and programmers'
guide. Engineering Department, Cambridge University.
38 Gunn, M.J. and Britto, A.M. (1984). CRISP84-users and programmers
guide. Engineering Department, Cambridge University.

228

References

39 Hight, D.W., Jardine, R.J. and Gens, A.(1987). The behaviour of soft
clays. Embankrnents on Soft Clays, Chapter 2, Bulletin of the Public Works
Research Centre, Athens, Greece.
40 Hansbo, S. (1981). Consolidation of fine-grained soils by prefabricated
drains. Proc. 10th ICSMFE, Stockholm, Vol. 3, pp. 677-682.
41 Hansbo, S., Jamiolkowski, M. and Kok, L. (1981). Consolidation by
vertical drains. Geotechnique, Vol 31, No. 1, pp. 45-66.
42 Hardin, B.O. and Black, W.L (1966). Sand stiffness under various
triaxial stresses. Journal of Soil Mechanics and Foundation Engineering,
ASCE, Vol. 92, SM2.
43 Hinton, B. and Campbell, J.S. (1974). Local global smooting of
discontinuous finite element functions using a least squares method.
International Journal of Numerical Methods in Engineering, Vol. 8, pp.
461-480.
44 Hird, C.0 and Kwok, C.M. (1986). Predictions for the Stanstead
Abbotts trial Embankment. Prediction Symposium on a Reinforced
Embankment, King's College, London.
45 Hird, C.0 and Kwok, C.M. (1 990a). Finite element studies of interface
behaviour in reinforced embankments on soft ground. Computer and
Geotechnics, Vol. 8, pp. 111-131.
46 Hird, C.0 and Kwok, C.M. (199Db). Parametric studies of the
behaviour of a reinforced embankment. Proceedings of the 4th International
Conference on Geotextiles, Geomembranes and Related Products, The
Hauge, Netherlands.
47 Hird, C.0 and Russell, D. (1990). A benchmark for soil-structure
interface elements. Computer and Geotechnics, Vol. 10, pp. 139-147.
48 Hird, C.C, Pyrah, I.C. and Russell, D. (1990). Finite element analysis
of the collapse of reinforced embankments on soft ground. Vol. 40, No. 4,
pp. 633-640.
49 Hird, C.C. and Pyrah; I.C. (1990). Predictions of the behaviour of a
reinforced embankment on soft ground, Performance cf reinforced soil
structures, British Geotechnica I Society.
50 Hird, C.0 and Jewell, R.A. (1990). Theory of reinforced embankments
Reinforced Embankments: Theory and Practice in the British Isles. pp.
115-1 39. Pub: Thomas Telford.
229

References

51 Hohberg, J.M. (1990). A note on spurious oscillations in FEM joint

elements. Earthquake Engineering and Structural Dynamics, Vol. 19, pp.
773-779.
52 Holtz, R.D., Jamiolkowski, M., Lancellotta, R. and Pedroni, S. (1987).
Performance of band-shaped drains. CI RIA Report 364.
53 Holtz, R.D., Jamiolkowski, M., Lancellotta, R. and Pedroni, S. (1991).
Prefabricated vertical drains - design and performance. CIRIA, Pub:
Butterworth Heinemann.
54 Home, M.R. (1964). The consolidation of stratified soil with vertical and
horizontal drainage. mt. J. of Mech. Sc., Vol. 6, pp. 187-197.
55 Houlsby,G.T. and Jewell, R.A. (1988). Analysis of unrein forced and
reinforced embankments on soft clays by plasticity theory. Proc. 6th mt.
Conf. on Numerical Methods in Geomechanics, lnnsbruck, Vol. 2, pp.
1443-1448.
56 Jacky J. (1944). The coefficient of earth pressure at rest. Magyar
Mernok es Epitesz Egylet Kozloyne.
57 Jamiolkowski, M., Lancellotta, R. and Tordelle, M.L (1980).
Geotechnical properties of Porto Tolle NC silty clay. 6th ECSMFE, Varna,
Bulgaria, pp. 151-179.
58 Jamiolkowski, M., Lancellotta, R. and Wolski W. (1983).
Precompression and speeding up consolidation. General report, 8th
ECSMFE, Helsinki, Vol. 3, pp. 1201-1226.
59 Jamiolkowski, M. and Lancellotta, R.(1984). Embankment on vertical
drains - pore pressures during construction. Proc. Int. Conf. Case Histories
in Geotechnical Engineering, St Louis, MO, USA, Vol. 1, pp. 275-278.
60 Jardine R.J. and Hight D.W. (1987). The behaviour and analysis of
embankments on soft clay. Embankments on Soft Clays, Chapter 3, Bulletin
of the Public Works Research Centre, Athens, Greece.
61 Jewell, R.A. (1982). A limit equilibrium design method for embankment
on soft soils. Proc. 2 lnt. Conf. on Geotextiles, Las Vegas, Vol. 3, pp.
671-676.
62 Jewell, R.A. (1988). The mechanics of a reinforced embankment on
soft soils. Geotextiles and Geomembranes, Vol. 7, pp. 237-273.

230

References

63 Jewell, R.A. (1990). Strength and deformation in reinforced soil design.

Proceedings of the 4 International Conference on Geotextiles,
Geomembranes and Related Products, The Hauge, Netherlands.
64 Johnson, S.J. (I 970a). Precompression for improving foundation soils.
J. Soil Mech. Fnds. Div., ASCE, Vol. 96, SMI, pp. 111-144.
65 Johnson, S.J. (1970b). Foundation precompression with vertical sand
drains. J. Soil Mech. Fnds. Div., ASCE, Vol. 96, SMI, pp. 145-175.
66 Kjellman, M. (1948). Accelerating consolidation of fine-grained soils by
means of cardboard wicks. Proc. 1" Conf. Soil Mech., Rotterdam, Vol. 2,
pp. 302-305.
67 Kulhawy, F.H. and Duncan, J.M. (1972). Stresses and movements in
Oroville Dam. Journal of Soil Mechanics and Foundations Division ASCE.
Vol. 98, SM7, pp. 653-665.
68 Kumamoto, N., Sumioka, N., Moriwaki, T. and Yoshikuni, H. (1988).
Settlement behaviour of improved ground with a vertical drain system. Soils
and Foundations. Vol. 28, No. 1, pp. 77-88.
69 Kwok, C.M. (1987). Reinforced embankments on soft ground. PhD
Thesis, University of Sheffield.
70 Ladd, C.C. (1991). Stability evaluation during staged construction. J.
Geotechnical Engineering, ASCE, Vol. 117, No. 4, pp. 540-615.
71 Lade, P.V. (1977). Elasto-plastic stress-strain theory for cohesionless
soils with curved yield surfaces. mt. Jrnl Solids and Structures, Vol. 13, pp.
1019-1 035.
72 Leroueil, S., Bouclin, G., Tavenas F., Bergeron, L and La Rochelle,
P. (1990). Permeability anisotropy as function of strain. Canadian
Geotechnical Journa', Vol. 27, pp. 568-579.
73 Leroueil, S., Magnan, J. and Tavenas, F. (1991). Embankments on
soft clays. Published: Ellis Horwood.
74 Mandel, J. and Salenon, J. (1972). Force portante d'un soil sur
asside rigide (etude theoretique). Gotechnique, Vol. 22, pp. 79-93.
75 Mongilardi, E. and Tortensson, B. (1977). 8000,000 metres of
Geodrain in the Porto Tolle thermoelectric power plant. Goetechnical report
from Terrafigo, Stockholm.
231

References

76 Morgenstern, N.R. and Price, V.E. (1965). The analysis of stability of

general slip surfaces. Gotechnique, Vol. 15, PP. 79-93.
77 Murray, R.T. and Symons, l.F. (1974). Embankments on soft
foundations: settlement and stability study at Tickton in Yorkshire. TRRL
Laboratory Report 643.
78 Mylleville, B.L.J. and Rowe, R.K. (1991). On the design of reinforced
embankments on soft brittle clays. Geosynthetics 91 Conference, Atlanta,
USA, pp. 395-408.
79 Newmark, N.M. (1948). Influence charts for the computation of
stresses in elastic foundation. University of Illinois Eng. Exp. Sta. Bul., 378.
80 Onoue, A. (1988). Consolidation by vertical drains taking well
resistance and smear into consideration. Soils and Foundations, Vol. 28,
No. 4, pp. 165-174.
81 Owen, D.R.J. and Hinton, E. (1980). Finite elements in plasticity:
theory and practice. Pub: pineridge press, Swansea.
82 Parry, R.H.G. (1970). Overconsolidation in soft clay deposits.
Gotechnique, Vol. 20, pp. 442-446.
83 Parry, R.H.G. (1972). Stability analysis for low embankments.
Proceedings of the Roscoe Memorial lecture, pp. 643-668, Pub: G.T. Foulis
and Co.
84 Pilot, G., Chaput, D. and Queyroi. D. (1987). Improvement of clay
properties - Usual methods. Embankments on Soft Clays, Chapter 5,
Bulletin of the Public Works Research Centre, Athens, Greece.
85 Potts, D.M., Ganendra, D. and Day, R.A. (1990). increment size and
solution accuracy in finite element analysis. 4th CRISP Users workshop,
Surrey University.
86 Potts, D.M. and Ganendra, D. (1991). Finite element analysis of the
collapse of reinforced embankments on soft ground. Discussion:
Gotechnique, Vol. 41, No. 4, pp 627-630.
87 Poran, C.J., Kaliakin, V.N., Herrmenn, LR., Romstadd, K.M., Lee,
D.F. and Shen, C.K. (1986). Prediction of trial embankment behaviour
Hertfordshire county council - Stanstead Abbotts. Prediction Symposium on
a Reinforced Embankment, King's College, London.

232

References

88 Poulos, H.G. and Davis, E.H. (1974). Elastic solutions for soil and
rocks. Wiley, New York.
89 Roscoe, K.H. and Burland, J.B. (1968). On the generalised
stress-strain behaviour of "wet" clay. In: Engineering Plasticity, Ed. J.
Heyman and F.A. Leckie, Cambridge University Press, pp. 535-609.
90 Rowe, P.W. (1972). The relevance of soil fabric to site investigation
practice. Gotechnique, Vol. 22, No. 2, PP. 1995-300.
91 Rowe, R.K. (1982). The analysis of an embankment constructed on a
geotextlle. Proc. 2 Int. conf. on Geotextiles, Las Vegas, Vol. 2, pp.
677-682.
92 Rowe, R.K. (1984). Reinforced embankments: Analysis and design.
Journal of Geotecnical Engineering Division, ASCE. Vol. 110, GT2, pp.
231-246.
93 Rowe, R.K., Maclean, M.P. and Soderman, 1(1. (1984). Analysis of a
geotextile reinforced embankment constructed on peat. Canadian
Geotecnical Journal. Vol. 21, pp. 563-576.
94 Rowe, R.K. and Soderman, K.L. (1985). An approximate method for
estimating the stability of geotextile-rein forced embankments. Canadian
Geotechnical Journal, Vol. 22, No. 3, pp. 392-398.
95 Rowe, R.K. and Soderman, K.L (1987). Stabilization of very soft soils
using high strength geosynthetics: the role of finite element analyses
Geotextiles and Geomembranes, Vol. 6, pp. 53-80.
96 Rowe, R.K. and Mylleville, B.LJ. (1988). The analysis of
steel-reinforced embankments on soft clay foundations. Proc. 6th tnt. conf.
on Numerical Methods in Geomechanics, lnnsbruck, pp. 1273-1278.
97 Rowe, R.K. and Mylleville, B.LJ. (1989). Consideration of strain in the
design of reinforced embankments. Geosynthetics '89 Conference, San
Diego, USA, pp. 124-1 35.
98 Rowe, R.K. and Mylleville, B.LJ. (1990). Implications of adopting an
4th
allowable geosynthetic strain in estimating stability. Proceedings of the
International Conference on Geotextiles, Geomembranes and Related
Products, The Hauge, Netherlands, pp. 131-136.
99 Russell, D. (1988). A comparison of CRISP84 and CR!SP81. University
of Sheffield, Internal Report.
233

References

100 Russell, D. (1990). An element to model thin, highly permeable

materials in two dimensional finite element consolidation analyses. 2"
European Speciallity Conference on Numerical Methods in Geotechnical
Engineering, pp. 303-310.
101 Russell, D. (1992). DRCRISP User's manual . Internal report,
University of Sheffield.
102 Sanchez, J.M. and Sagaseta, C. (1990). Analysis of stage
construction of embankments on soft soil. 2" European Speciallity
Conference on Numerical Methods in Geotechnical Engineering, pp.
303-310.
103 Schmidt, B. (1966). Discussion of earth pressures at rest related to
stress history. Canadian Geotechnical Journal, Vol 3, pp. 239-242.
104 Schafer, V.R. (1987). Analysis of reinforced embankments and
foundations overlying soft soils. PhD Thesis, Virginia Polytechnic Institute
and State University.
105 Schofield, A.N. and Wroth, C.P. (1968). Critical state soil mechanics.
London: Mcgraw-Hill.
106 Shinsha et al (1982). Consolidation settlement and lateral
displacements of soft soil ground improved by sand drains. Soils and
Foundations, May, pp. 7-12 (in Japanese).
107 Skempton, A.W. (1954). The pore pressure coefficients A and B.
Gotechnique, Vol. 4, No. 4, pp. 143-147.
108 Sloan, S.W. and Randolph, M.F. (1982). Numerical prediction of
collapse loads using finite element methods. International Journal of
Numerical and Analytical Methods in Geomechanics, Vol. 6, pp. 47-48.
109 Smith, l.M. and Hobbs R. (1974). Finite element analysis of
centrifuged and built-up slopes. Gotechnique, Vol. 24, pp. 531-559.
110 Smith, l.M. and Hobbs R. (1976). Blot analysis of consolidation
beneath embankments. Gotechnique, Vol. 26, pp. 149-171.
111 Smith, IM. (1982). Programming the finite element method - with
application to geomechanics. Wiley, London.
112 Smith, P.R. (1984). The constructiuon of multistage embankments on
soft clay, MSc Dissertation, Imperial College, London.
234

References

113 Tavenas, F., Jean, P., Leblond, P. and Leroueil, S. (1983). The
permeability of natural soft clays, part II: Permeability characteristics.
Canadian Geotechnical Journal, Vol. 20, pp. 645-659.
114 Tavenas, F. and Leroueil, S. (1980). The behaviour of embankments
on clay foundations. Canadian Geotechnical J., Vol. 17, No. 2, pp. 236-260.
115 Taylor, D.W. (1937). Stability of earth slopes. Journal of the Boston
Society of Civil Engineers, Vol. 24, pp. 197.
116 Taylor, D.W. (1948). Fundamentals of soil mechanics. Pub:
McGraw-Hill, New York.
117 Terzaghi, K. and Peck, R.B. (1967). Soil mechanics in engineering
practice. Pub: Wiley, London.
118 Tomlinson, M.J. (1987). Foundation design and construction. Fifth
Edition, Pub: Longman.
119 Van Langen, H. and Vermeer, P.A. (1991). Interface elements for
singular plasticity points. Int. J. Num. Meth in Geomech, Vol. 15, pp.
301-315.
120 Watson, H., Crooks, J.H.A., Williams, R.S. and Yam, C.C. (1984).
Performance of preloaded and stage loaded structures on soft soils in
Trinidad. Gotechnique, No. 34, pp. 239-259.
121 Wroth, C.P. and Simpson, B. (1972). An induced failure at a triai
embankment. Part II - finite element computations. Proc. Spec. Conf. on
Performance of Earth and Earth-Supported Structures, ASCE, Lafayette,
Vol. 1, pp. 65-79.
122 Wroth, C.P. (1975). In-situ measurement of initial stresses and
deformation characteristics. Proc. Spec. Conf. in In-situ Measurement of
Soil Properties. ASCE, Rayleigh, North Caroline, pp. 181-230.
123 Wroth, C.P. (1977). The predicted performance of soft clay under a
trial embankment loading based on the Cam-clay model. Finite Elements in
Geomechanics, Willey, Chapter 6. (ed. Gudehus).
124 Wroth, C.P. (1984). The interpretation of in-situ soil tests
Geotechnique, Vol. 34, No. 4, pp. 449-489.
125 Yoshikuni, H. and Nakanodo, H. (1974). Consolidation of soils by
vertical drain wells with finite permeability.Soils and Foundations. Vol. 14,
No. 2, pp. 35-46.
235

References

126 Yudbir, C. and Wood, D.M. (1989). Recent developments in

laboratory strength and deformation testing. General report. Proc. I 2th lnt.
Conf. Soil Mech., Rio de Janeiro.
127 Zeng G.X., Xie, K.H. and Shi,Z.Y. (1987). Consolidation analysis of
sand-drained ground by F.E.M. Proc. 8th Regional Conf Soil Mech., Kyoto.
Vol. 1, PP. 139-142.
128 Zeng G.X. and Xie, K.H. (1989). New development of vertical drain
theories. Proc. 12th lnt. Conf. Soil Mech., Rio de Janeiro. Vol. 2, pp.
1433-1438.
129 Zienkiewicz, O.C. (1977). Finite Element Method. McGraw-Hill,
London.

236