Title: Numerical Modelling of Small Groups of Stone Columns

Provided by the author(s) and NUI Galway in accordance with publisher policies.
Please cite the published

version when available.
Title Numerical modelling of small groups of stone columns
Author(s) Killeen, Micheál
Publication 2012-09-21
Date
Item record http://hdl.handle.net/10379/3464
Downloaded 2019-09-25T01:34:53Z
Some rights reserved. For more information, please see the item record link above.
NUMERICAL MODELLING OF SMALL
GROUPS OF STONE COLUMNS
by
Micheál Killeen B.E.
A thesis submitted to the College of Engineering and Informatics,

National University of Ireland, Galway, in partial fulfilment of the
requirements for the Degree of Doctorate in Philosophy
June 2012
Academic Supervisor: Dr. Bryan McCabe

Professor of Civil Engineering: Prof. Padraic O’Donoghue
Declaration
I hereby declare that this thesis has not been submitted in whole or in part to any other
University as an exercise for a degree. I further declare that, except where reference is
given in the text, it is entirely my own work.
This thesis may be lent or copied by the library.
Signed: ___________________________ Date: __________________

Micheál Killeen
To my parents Michael and Kitty,
and sisters Maria and Áine
Acknowledgements
I would like to express my sincere appreciation and deep gratitude to my supervisor, Dr.
Bryan McCabe, who devoted so much time and effort to this research project. I am extremely
grateful for all his support, encouragement and advice throughout my studies. Most of all, I
appreciate his friendship.
I would also like to extend a special thanks to my academic mentors, Dr. Annette Harte and
Prof. Padraic O' Donoghue, for their contributions throughout the duration of this research.
I would like to acknowledge Keller Foundations (UK) for their generous financial support
provided to this research project. I am especially grateful to Dr. Alan Bell and Dr. Derek
Egan for their highly practical advice and valuable contributions to many technical problems.
This thesis on the behaviour of small groups of stone columns was conducted in collaboration
with research partners. I would like to thank Dr. V. Sivakumar (Queen's University, Belfast),
Dr. Jonathan Black (University of Sheffield, UK) and also Dr. Minna Karstunen and Dr.
Daniela Kamrat-Pietraszewska (Strathclyde University, Glasgow) for their assistance to this
research project. I would also like to thank Dr. Jorge Castro and Dr. Bostjan Pulko for their
help in patiently answering my many questions. I would also like to acknowledge the advice
from the PLAXIS support team.
I greatly appreciate the financial scholarship provided by the Irish Research Council for
Science, Engineering and Technology (IRCSET) to support this research project. I would also
like to express my sincere gratitude to the Geotechnical Society of Ireland for funding my trip
to Egypt to participate in the International Young Geotechnical Engineers conference in
2009. I thoroughly enjoyed this experience and it has encouraged me to pursue future studies
in geotechnical research.
I am indebted to the staff in the Department of Civil Engineering at NUI Galway who
through their dedication and enthusiasm helped to instil a passion for engineering in me
during my undergraduate studies.
I am extremely grateful to my fellow postgraduate students, all my friends and my relatives
(including those who have passed away) for their support, advice and warm friendship they
have provided to me.
Finally, my last and most important acknowledgement goes to my parents and sisters. I am
forever indebted to you for the never-ending support provided to me and all the sacrifices
made.
Míle buíochas
Abstract
Vibro stone columns, installed using the vibro replacement technique, are a cost-
effective form of ground improvement for enhancing the bearing capacity and settlement
characteristics of various soil types. Large groups of stone columns (such as used to
support embankments) are conventionally modelled using the unit cell concept, which is
based on an infinite grid of columns supporting an infinitely wide load area. Therefore
all columns are equally confined on all sides and are subject to a constant increment of
vertical stress with depth. The behaviour of small groups of stone columns supporting
small area footings is quite complex as peripheral columns are subject to a loss of lateral
confinement and the increment of vertical stress decays sharply with depth. This research
is the first comprehensive three-dimensional numerical study of the factors affecting
both the mechanisms of load transfer from columns to soil and the settlement
performance of small column groups at working loads.
PLAXIS 3D Foundation is used for this research in conjunction with the advanced
elasto-plastic Hardening Soil model, which is adopted for the parent soil and stone
backfill. The soft soil modelled is that at Bothkennar, Scotland, the former UK
geotechnical test bed. The influence of key design parameters such as area ratio, column
length, stiffness, strength and installation effects upon the settlement performance and
deformational behaviour of small groups of stone columns was investigated through a
total of 45 numerical analyses.
The modelling has shown that the area ratio and column length have a significant
influence on the settlement performance of stone columns. Moreover, they appear to be
inter-dependent as the effect of column length becomes more pronounced at low area
ratios. The influence of column confinement (i.e. increasing number of columns) was
also found to have a positive influence on the settlement performance of small groups of
stone columns. It was also shown that the influence of key design parameters upon the
settlement performance of stone columns is dependent upon the mode of deformation.
New parameters called compression and punching ratios were defined to help identify
three distinct mechanisms referred to as "punching", "block failure" and "bulging". The
occurrence of these mechanisms was verified by analysing the distribution of total shear
strain within columns and the surrounding soil and also examining the variation of stress
and strain along the length of columns. It was found that area ratio and column length,
rather than the number of columns, dictates the load transfer mechanism for small groups
of stone columns.
A more in-depth analysis of the deformational behaviour reveals that some combination
of punching and bulging occurs simultaneously, with one particular mode of deformation
more influential for a given area ratio and column length. This is consistent with the
finding that settlement improvement factors increase with column length for all
configurations of columns and suggests that a unique critical length, as proposed by
previous laboratory studies, does not exist for small groups of stone columns. The
presence of a stiff crust, an important feature of soft soil stratigraphy not captured in the
laboratory tests, was shown to have a significant influence upon the deformational
behaviour of columns. The observation of a critical length from laboratory studies is
shown in part to be due to the absence of a stiff crust (i.e. homogeneous soil samples) as
columns are more susceptible to bulging in the upper layers and thus cannot transfer the
applied load to their base.
The stress concentration ratios at the ground surface were also examined and it was
found that they are related to the mode of deformation. Moreover, it was shown that
stress concentration ratios vary considerably with column position and as such, do not
uniquely reflect the settlement performance of stone columns. Instead, the stress
concentration ratios with depth were noted and it was observed that they are constant
with depth in the yielded sections of the column and decrease towards unity at the base
of floating columns thereafter.
The numerical output in this thesis has been developed into a simplified design method
which allows the settlement of a column group to be related to that of a unit cell with
knowledge of the footing to column length ratio and the column length to layer thickness
ratio (and thereby caters for floating column groups).
Table of Contents
List of Figures i
List of Tables viii
Nomenclature ix
1 Introduction
1.1 Background 1
1.2 Aims and objectives 2
1.3 Methodology 2
1.4 Outline of thesis 3
2 Literature review
2.1 Background 5
2.1.1 Introduction 5
2.1.2 Construction processes 6
2.1.3 Applications of vibro replacement 8
2.1.4 Limitations of vibro replacement 10
2.2 General deformational behaviour of stone columns 11
2.2.1 Behaviour of single columns 11
2.2.2 Behaviour of column groups 13
2.2.3 Summary of general deformational behaviour of stone columns 18
2.3 Settlement performance of stone columns 18
2.3.1 Case studies 19
2.3.2 Laboratory studies 20
2.3.3 Numerical studies 22
2.3.4 Summary of settlement performance of stone columns 28
2.4 Column installation effects 29
2.4.1 Laboratory investigations 29
2.4.2 Observed field measurements 30
2.4.3 Simulation of installation effects in numerical models 32
2.4.4 Summary of column installation effects 35
2.5 Design methods 36
2.5.1 Unit cell concept 36
2.5.2 Cylindrical Cavity Expansion Theory (CCET) 36
2.5.3 Ultimate bearing capacity 38
2.5.4 Magnitude of settlement 40
2.5.5 Rate of settlement 46
2.5.6 Summary of design methods 48
2.6 Summary of literature review 49
3 Background and preliminary checks for Finite Element Analysis

3.1 Introduction 51
3.2 Description of material models 51
3.3 Preliminary checks to ensure accurate numerical analyses 55
3.3.1 Mesh sensitivity analyses 55
3.3.2 Influence of distance to boundary 57
3.4 Other modelling issues 59
3.4.1 Modelling of soft soil behaviour 59
3.4.2 Modelling of column-soil interface 60
3.4.3 Modelling of column installation effects 62
4 Development and validation of soft soil profile

4.1 Introduction 66
4.2 Description of load tests in a layered estuarine deposit 67
4.2.1 Background 67
4.2.2 Description of site conditions 67
4.2.3 Stone column design & construction 68
4.2.4 Material parameters 68
4.2.5 Field load tests 69
4.2.6 Axisymmetric FEA 70
4.3 Simulation of load tests in a layered estuarine deposit 71
4.3.1 Comparison of PLAXIS 3D Foundation with field records and axisymmetric
FEA 72
4.4 Development of Bothkennar soil profile 74
4.4.1 Description of Bothkennar test site 74
4.4.2 Soil classification and initial stress state 75
4.4.3 Strength characteristics 77
4.4.4 One-dimensional stiffness properties 78
4.4.5 Permeability and consolidation coefficients 79
4.5 Validation of Bothkennar soil profile 80
4.5.1 Description of field load test 80
4.5.2 Simulation of field load test with PLAXIS 3D Foundation 80
4.5.3 Comparison of PLAXIS 3D Foundation with field measurements 81
4.6 Simulation of an infinite grid of stone columns in Bothkennar 82
4.6.1 Development of material parameters for stone backfill 83
4.6.2 Comparison of PLAXIS 3D Foundation with one-dimensional compression
theory 84
4.6.3 Comparison of PLAXIS 3D Foundation with settlement database 85
4.6.4 Comparison of PLAXIS 3D Foundation with analytical design methods 87
4.7 Summary of development and validation of soft soil profile 92
5 Results of FEA: Settlement performance, deformational behaviour

and stress concentration ratios
5.1 Background 93
5.2 Details of FEA 94
5.3 Settlement analysis 98
5.3.1 Settlement performance of infinite grids of stone columns 98
5.3.2 Settlement performance of small groups of stone columns 99
5.3.3 Comparison of infinite grids with small groups of stone columns 100
5.3.4 Influence of column arrangement 103
5.3.5 Influence of column position 104
5.3.6 Influence of column compressibility 104
5.3.7 Influence of column strength 106
5.3.8 Influence of column installation effects 107
5.3.9 Influence of crust 108
5.4 Deformational behaviour of stone columns 110
5.4.1 Definitions of punching and compression ratios 110
5.4.2 Deformational behaviour of an infinite grid of stone columns 112
5.4.3 Deformational behaviour of small groups of stone columns 114
5.4.4 Comparison of infinite grid with small groups of columns 120
5.4.5 Influence of column arrangement upon deformational behaviour 122
5.4.7 Influence of column compressibility 124
5.4.10 Influence of stiff crust 127
5.5 Stress concentration ratio 131
5.5.1 Stress concentration ratios for infinite grids of stone columns 133
5.5.2 Stress concentration ratios for small groups of stone columns 133
5.5.3 Comparison of infinite grid with small groups of columns 139
5.5.4 Influence of column arrangement 142
5.5.6 Influence of column stiffness 145
5.5.9 Influence of stiff crust 151
5.5 Summary of results of FEA: Settlement performance, deformational behaviour
and stress concentration ratios 153
5.6.1 Settlement performance 153
5.6.2 Deformational behaviour 154
5.6.3 Stress concentration ratios 155
6 Results of FEA: Distribution of total shear strains and

characteristic column behaviour
6.1 Background 157
6.2 Distribution of total shear strains 158
6.2.1 Total shear strains for an infinite grid of stone columns 158
6.2.2 Total shear strains for small groups of stone columns 160
6.3 Characteristic column behaviour 168
6.3.1 Infinite grid of columns 168
6.3.2 Small groups of columns 172
6.4 Summary of results of FEA: Distribution of total shear strains and characteristic
column behaviour 188
6.4.1 Distribution of total shear strain 188
6.4.2 Distribution of vertical strain, horizontal strain and stress concentration ratios 189
7 Development of simplified design method and comparison of

findings with previous research
7.1 Introduction 190
7.2 Settlement analysis 191
7.2.1 Justification of the use of s/suc 191
7.2.2 Review of Priebe (1995) 191
7.2.3 Review of Elshazly et al. (2008a) 192
7.2.4 Comparison of s/suc for end-bearing columns from PLAXIS 3D Foundation &
Priebe (1995) 194
7.2.5 Settlement ratios for floating stone columns 195
7.3 Deformational behaviour of stone columns 199
7.3.1 Summary of findings from PLAXIS 3D Foundation 199
7.3.2 Review of Black (2006) 200
7.3.3 Comparison of PLAXIS 3D Foundation with Black (2006) 201
7.3.4 Review of Muir Wood et al. (2000) 204
7.3.5 Comparison of PLAXIS 3D Foundation with Muir Wood et al. (2000) 204
7.4 Influence of deformational behaviour upon settlement performance 207
7.5 Stress concentration ratio 210
7.5.1 Stress concentration ratios for infinite grids of stone columns 210
8 Conclusions
8.1 Introduction 216
8.1.1 Numerical modelling preliminaries 216
8.1.2 Settlement performance 217
8.1.3 Mode of deformation 217
8.1.4 Stress concentration ratios 220
8.1.5 Simplified design method 222
8.2 Recommendations for future work 222
References 224
List of Figures i
List of Figures
2 Literature review
Figure 2.1 Range of soil types treatable by vibro compaction and vibro replacement
techniques (courtesy of Keller Ground Engineering) 6
Figure 2.2 Depth vibrator and principle of vibro compaction (Sondermann & Wehr, 2004) 7
Figure 2.3 Dry bottom feed system using ‘vibrocat’ 8
Figure 2.4 Area ratios for (i) square grids, (ii) triangular girds and (iii) pad footings 9
Figure 2.5 Development of stress concentration ratios in stiffer stone columns 9
Figure 2.6 (a) Vertical displacement within the column against depth and (b) Radial
displacement at the edge of the column / initial column radius against depth
(Hughes & Withers, 1974) 12
Figure 2.7 Typical test setup examined by Narasimha Rao et al. (1992) 12
Figure 2.8 Photographs of deformed sand columns exhumed at the end of footing
penetration (Muir Wood et al., 2000) 14
Figure 2.9 Photographs of sand columns beneath strip footings at beginning, middle and end
of foundation loading process: (a) L/d = 6; (b) L/d = 10 (McKelvey et al., 2004) 16
Figure 2.10 Deformed group of (a) short columns and (b) long slender columns with
additional column outside the footing, upper part, u = 10 mm, clay (white) and
column (grey) (Wehr, 2004) 17
Figure 2.11 Settlement improvement factor against area replacement ratio for sites with
widespread loading (McCabe et al., 2009) 19
Figure 2.12 Predicted versus measured settlement improvement factors for all widespread
loadings and footings (McCabe et al., 2009) 20
Figure 2.13 Comparison of Ks for isolated and group formation (Black, 2006) 21
Figure 2.14 Illustration of block failure in group columns (Black, 2006) 21
Figure 2.15 Settlement variation with (a) EP/ES ratio and (b) degree of penetration of piles
(Balaam et al., 1977) 23
Figure 2.16 Measurement of stress concentration below an embankment (Kirsch &
Sondermann, 2003) 25
Figure 2.17 Settlement correction factor versus size ratio (a) case of layered soil, (b) case of
10.8 m thick soft clay layer and (c) case of 30 m thick soft clay layer (Elshazly et
al., 2008a) 27
Figure 2.18 Soil displacement due to pile installation (Yu, 2000) 29
Figure 2.19 Factor of restraint measured during the installation of stone columns (Kirsch,
2006) 31
Figure 2.20 Development of ground stiffness during column installation (Kirsch, 2006) 31
Figure 2.21 Zones of influence for square and triangular arrangements of stone columns 36
Figure 2.22 Equilibrium of soil element (Powrie, 2004) 37
List of Figures ii
Figure 2.23 Relationship between allowable vertical stress on stone column and undrained
shear strength (Thorburn & MacVicar, 1968) 39
Figure 2.24 Settlement diagram for stone columns in uniform soft clay (Greenwood, 1970) 40
Figure 2.25 Design curves for basic settlement improvement factor n0 with Poisson’s ratio υ =
⅓ (Priebe, 1995) 42
Figure 2.26 Settlement of single footings on groups of columns (Priebe, 1995) 43
Figure 2.27 Boundary conditions for solutions A & B proposed by Balaam & Booker (1981) 44
Figure 2.28 Dissipation of excess pore water pressure (Han & Ye, 2001) 47
Figure 2.29 Stress concentration factor. Influence of radial deformation and plastic strain
(Castro & Sagaseta, 2009) 48

Figure 3.1 Mohr-Coulomb yield surface in principal stress space (c = 0 kPa) (Brinkgreve &
Broere, 2006) 52
Figure 3.2 Hyperbolic stress-strain relation in primary loading for a standard drained triaxial
test (Schanz et al., 1999) 53
Figure 3.3 Hardening Soil yield surface in principal stress space (c = 0 kPa) (Brinkgreve &
Broere, 2006) 54
Figure 3.4 Distribution of (i) nodes and (ii) stress points within 15 node wedge elements 55
Figure 3.5 Points A, B and C beneath pad footing 56
Figure 3.6 Continuous displacement contours along inter-element boundaries beneath a 3 m
square footing 56
Figure 3.7 Flowchart outlining preliminary analysis checks 57
Figure 3.8 Influence of analysis type upon the settlement improvement factors for various
groups of columns 60
Figure 3.9 Interface elements for a group of 5 columns beneath a 3 m square pad footing 61
Figure 3.10 Influence of interface elements upon settlement improvement factors for various
Figure 3.11 Variation of cavity pressure with radius (undrained CCE) (McCabe et al., 2008) 63
Figure 3.12 Variation of (a) Excess pore pressure (pwp) and (b) total radial stress with
normalised radial distance for lateral expansions (a/a0) of (i) 1.03, (ii) 1.1 and (iii)
1.33 (McCabe et al., 2008) 64

Figure 4.1 Soil profile for Santa Barbara waste water treatment plant (columns spaced at
1.75 m × 1.75 m) 70
Figure 4.2 Conversion of surrounding stone columns into a cylindrical ring for axisymmetric
FEA 71
Figure 4.3 FE mesh for 1.75 m × 1.75 m grid of columns 72
List of Figures iii
Figure 4.4 Comparison of field load test data with numerical simulations for columns spaced
on (i) 1.2 m × 1.5 m, (ii) 1.75 m × 1.75 m and (iii) 2.1 m × 2.1 m 73
Figure 4.5 Bothkennar test site location (Nash et al., 1992a) 75
Figure 4.6 Geotechnical profile at Bothkennar test site (Nash et al., 1992a) 75
Figure 4.7 (a) Yield stress and (b) yield stress ratio from incremental load consolidation tests
(Nash et al., 1992a) 76
Figure 4.8 Variation of (a) lateral total stress (b) K0 with depth (Nash et al., 1992a) 77
Figure 4.9 Variation of (a) compression index Cc and (b) initial voids ratio e0 with depth
(Nash et al., 1992b) 78
Figure 4.10 Variation of load with time (Jardine et al., 1995) 80
Figure 4.11 Load-displacement behaviour for pad footings (Jardine et al., 1995) 81
Figure 4.12 Characteristic stiffness-strain behaviour of soil with typical strain ranges for
laboratory tests and structures (after Atkinson & Sallfors, 1991) 82
Figure 4.13 Settlement profile of Bothkennar test site for wide area loading, calculated using
one-dimensional compression theory and PLAXIS 3D Foundation 85
Figure 4.14 Comparison of settlement improvement factors for an infinite grid of end-bearing
columns with field data collated by McCabe et al. (2009) 86
Figure 4.15 Variation of normalised oedometric moduli with vertical effective stress for
Bothkennar test site 89
Figure 4.16 Distribution of minor principal stress σ3 in end-bearing stone column and
surrounding soil (A/Ac = 3.5; L = 13.9 m; Infinite grid of columns) 90
Figure 4.17 Comparison of settlement improvement factors for an infinite grid of end-bearing
columns with analytical design methods 91

Figure 5.1 Layout of columns in Bothkennar soil profile 95
Figure 5.2 Column configurations to examine influence of column confinement 95
Figure 5.3 Column configurations to examine influence of column arrangement 96
Figure 5.4 Column configurations to examine influence of column position relative to the
footing edge 96
Figure 5.5 Column configurations for parametric study of column compressibility and
strength 97
Figure 5.6 Various profiles adopted to investigate the effect of the lower Carse clay (Profile
2) and the stiff crust (Profile 3) 97
Figure 5.7 Variation of (i) footing settlement and (ii) settlement improvement factor with
column length for an infinite grid of columns 98
Figure 5.8 Variation of settlement improvement factor with (a) column length (L) and (b)
normalised column length (L/B) for small groups of columns, spaced at area
ratios of (i) 3.5, (ii) 8.0 and (iii) 14.1 101
Figure 5.9 Settlement improvement factors for small groups of end-bearing columns 102
List of Figures iv
Figure 5.10 Normalised settlement improvement factors for small groups of end-bearing
columns 102
Figure 5.11 Normalised settlement for small groups of end-bearing columns 102
Figure 5.12 Variation of settlement improvement factors with (i) column length and (ii)
volume of stone for various arrangements of stone columns 103
Figure 5.13 Influence of column position upon settlement improvement factors 104
Figure 5.14 Influence of column compressibility upon settlement improvement factors 105
Figure 5.15 Distribution of plastic points (Mohr-Coulomb) within a cross-section of (i) 4, (ii)
5 and (iii) 9 columns beneath a 3 m square footing (L = 6 m) 105
Figure 5.16 Influence of column strength upon settlement improvement factors 106
Figure 5.17 Influence of coefficient of lateral earth pressure upon settlement improvement
factors 107
Figure 5.18 Influence of the lower Carse clay and the stiff crust upon (i) settlement and (ii)
settlement improvement factors for (a) 4, (b) 5 and (c) 9 columns beneath a 3 m
square footing 109
Figure 5.19 Displacement at surface of footing, base of columns and soil surrounding base of
columns 110
Figure 5.20 Variation of punching ratios with column length for an infinite grid of stone
columns, with usoil measured in (i) a 1 m square zone and (ii) a square zone,
whose width is equal to the column spacing 111
Figure 5.21 Variation of (i) punching and (ii) compression ratios with column length for an
infinite grid of columns 112
Figure 5.22 Deformation of an infinite grid of floating stone columns at an area ratio A/AC =
14.1 for column lengths of (i) L = 1 m, (ii) L = 2 m and (iii) L = 3 m 113
Figure 5.23 Deformation of an infinite grid of 3 m long columns at area ratios of (i) 3.5, (ii)
8.0 and (iii) 14.1 113
Figure 5.24 Variation of (i) punching and (ii) compression ratios with column length for
single columns 115
Figure 5.25 Variation of (i) punching and (ii) compression ratios with column length for 2×2
Figure 5.26 Variation of (i) punching and (ii) compression ratios with column length for a 3×3
group of columns spaced at area ratios of (a) 3.5, (b) 8.0 and (c) 14.1 117
Figure 5.27 Variation of (i) punching and (ii) compression ratios with column length for a 4×4
group of columns spaced at area ratios of (a) 3.5, (b) 8.0 and (c) 14.1 118
Figure 5.28 Deformed shapes from PLAXIS 3D Foundation for a 3×3 group of columns at
area ratios of (i) 3.5, (ii) 8.0 and (iii) 14.1 for column lengths (a) L = 3 m and (b)
L=6m 119
Figure 5.29 Influence of column confinement upon (i) punching and (ii) compression ratios
for various lengths and configurations of columns spaced at area ratios of (a) 3.5,
(b) 8.0 and (c) 14.1 121
various configurations of stone columns beneath a 3 m square footing 123
List of Figures v
various column spacings beneath a 3 m square footing 124
Figure 5.32 Influence of column compressibility upon the variation of (i) punching and (ii)
compression ratios with column length for a group of (a) 4, (b) 5 and (c) 9
columns beneath a 3 m square footing 125
Figure 5.33 Influence of column strength upon the variation of (i) punching and (ii)
Figure 5.34 Influence of column installation effects upon the variation of (i) punching and (ii)
Figure 5.35 Influence of stiff crust upon the variation of (i) punching and (ii) compression
ratios with column length for a group of (a) 4, (b) 5 and (c) 9 columns beneath a 3
m square footing 130
Figure 5.36 Determination of stress concentration ratios for (i) field/laboratory studies and (ii)
numerical/analytical studies 131
Figure 5.37 Stress concentration ratio determined by (i) averaging stress over entire soil area
and (ii) averaging stress within the zone of influence for each column 132
Figure 5.38 Distribution of vertical effective stress (σ'vert) immediately beneath a 4.5 m square
footing, which is (i) unreinforced (L = 0 m) and (ii) reinforced with end-bearing
stone columns (L = 13.9 m) 132
Figure 5.39 Variation of stress concentration ratio with column length for an infinite grid of
columns 133
Figure 5.40 Variation of stress concentration ratios with column length for (i) single columns
and (ii) 2×2 groups of stone columns 134
Figure 5.41 Variation of stress concentration ratio with column length for a 3×3 group of
columns, spaced at area ratios of (a) 3.5, (b) 8.0 and (c) 14.1 136
Figure 5.42 Variation of stress concentration ratio with column length for a 4×4 group of
columns, spaced at area ratios of (a) 3.5, (b) 8.0 and (c) 14.1 137
Figure 5.43 Variation of stress concentration ratio for small groups and infinite grids of
columns spaced at area ratios of (a) 3.5, (b) 8.0 and (c) 14.1 140
Figure 5.44 Vertical stress beneath an unreinforced footing for a (i) single column, (ii) 2×2,
(iii) 3×3 and (iv) 4×4 groups of columns at an area ratio of 3.5 141
Figure 5.45 Variation of stress concentration ratio with column length for a group of 4, 5, 7
and 9 columns beneath a 3 m square footing 143
Figure 5.46 Variation of stress concentration ratio with column length for various columns
positions beneath a 3 m square footing 145
Figure 5.47 Vertical effective stress beneath a 3 m square footing for columns spaced at (i) 1.0
m, (ii) 1.5 m and (iii) 2.0 m 145
Figure 5.48 Influence of column compressibility upon the variation of stress concentration
List of Figures vi
Figure 5.49 Influence of column strength upon the variation of stress concentration ratios with
column length for a group of (a) 4, (b) 5 and (c) 9 columns beneath a 3 m square
footing 149
Figure 5.50 Influence of column installation effects upon the variation of stress concentration
Figure 5.51 Influence of stiff crust upon the variation of stress concentration ratios with
column length for a group of (a) 4, (b) 5 and (c) 9 columns beneath a 3 m square
footing 152
6 Results of FEA: Distribution of total shear strains and

Figure 6.1 Total shear strains for an infinite grid of columns for column lengths L of (a) 3 m,
(b) 8 m and (c) 13.9 m and at area ratios A/AC of (i) 3.5, (ii) 8.0 and (iii) 14.1 159
Figure 6.2 Total shear strains for single columns for column lengths L of (a) 3 m, (b) 8 m
and (c) 13.9 m and at area ratios A/AC of (i) 3.5, (ii) 8.0 and (iii) 14.1 161
Figure 6.3 Total shear strains for 2×2 groups of columns for column lengths L of (a) 3 m, (b)
8 m and (c) 13.9 m and at area ratios A/AC of (i) 3.5, (ii) 8.0 and (iii) 14.1 163
Figure 6.4 Total shear strains for 3×3 group of columns for column lengths L of (a) 3 m, (b)
Figure 6.5 Total shear strains for 4×4 group of columns for column lengths L of (a) 3 m, (b)
Figure 6.6 Distribution of (i) vertical strain, (ii) horizontal strain and (iii) stress concentration
ratio with depth for an infinite grid of columns at lengths L of (a) 3 m, (b) 8 m and
(c) 13.9 m 170
Figure 6.7 Distribution of modular ratio Ecol/Esoil with depth for an infinite grid of end-
bearing stone columns at various area ratios 171
ratio with depth for a single column at lengths L of (a) 3 m, (b) 8 m and (c) 13.9
m 173
ratio with depth for a 2×2 group of columns at lengths L of (a) 3 m, (b) 8 m and
(c) 13.9 m 176
ratio with depth for 3 m long columns, within a 3×3 group for A/AC of (a) 3.5, (b)
8.0 and (c) 14.1 179
8.0 and (c) 14.1 180
ratio with depth for 13.9 m long columns, within a 3×3 group for A/AC of (a) 3.5,
(b) 8.0 and (c) 14.1 181
8.0 and (c) 14.1 184
8.0 and (c) 14.1 185
List of Figures vii
ratio with depth for 13.9 m long columns, within a 4×4 group for A/AC of (a) 3.5,
(b) 8.0 and (c) 14.1 186
7 Development of simplified design method and comparison of

findings with previous research
Figure 7.1 Variation of settlement ratio with normalised footing width for a layered soil and
a soft clay, as determined by Elshazly et al. (2008a) 193
Figure 7.2 Comparison of settlement ratios determined by Elshazly et al. (2008a) with
Priebe (1995) 193
Figure 7.3 Deformed mesh of a foundation with B/L = 3 (case of 10.8 m-thick soft clay
layer with a 19×19 column arrangement) (Elshazly et al., 2008a) 194
Figure 7.4 Comparison of settlement ratios for groups of end-bearing stone columns from
PLAXIS 3D Foundation with Priebe (1995) 195
Figure 7.5 Variation of settlement ratios with normalised footing width (B/L) and column
length (L/H) 196
Figure 7.6 Influence of footing width (B/L) and column length (L/H) upon settlement ratios
for groups of columns 197
Figure 7.7 Relationship between α (which relates s/suc and B/L) with L/H 198
Figure 7.8 Comparison of compression ratios from Black (2006) and PLAXIS 3D
Foundation for single columns 1201
Figure 7.9 Comparison of compression ratios from Black (2006) with PLAXIS 3D
Foundation for soil profiles with and without a stiff crust 202
Figure 7.10 Influence of column location upon compression ratios measured by Muir Wood
et al. (2000) and determined PLAXIS 3D Foundation 205
Figure 7.11 Observations of punching and bulging for groups of columns, reported by Muir
Wood et al. (2000), plotted against (i) L/D and (ii) L/d 207
Figure 7.12 Variation of (i) compression ratio and (ii) punching ratio with settlement
improvement factor for (a) 1 column, (b) 2×2, (c) 3×3 and (d) 4×4 groups 208
Figure 7.13 Comparison of stress concentration ratios for an infinite grid of end-bearing
columns with field data 211
Figure 7.14 Comparison of stress concentration ratios at the surface for an infinite grid of
end-bearing columns with analytical design methods 212
Figure 7.15 Comparison of stress concentration ratios with depth from PLAXIS 3D
Foundation with (i) Priebe (1995) and (ii) Pulko & Majes (2005) for an infinite
grid of end-bearing columns 213
List of Tables viii
List of Tables
2 Literature review
Table 2.1 Increase in K0 to account for stone column installation effects by various authors 33
Table 2.2 Final solution for stresses and strains in column and soil (Balaam & Booker,
1981) 44

Table 3.1 Vertical displacement and mean effective stress at points A, B and C 58
Table 3.2 Influence of the distance to boundary upon the settlement of a 3 m square footing 58

Table 4.1 Material parameters for Santa Barbara wastewater treatment plant (Elshazly et al.,
2008b) 69
Table 4.2 Summary of adopted material parameters for the Bothkennar test site 79
Table 4.3 Summary of soil profile to 7 m depth (Jardine et al., 1995) 82
Table 4.4 Comparison of one-dimensional compression theory with PLAXIS 3D
Foundation 85
Table 4.5 Summary of material parameters for Bothkennar soil profile 87
Table 4.6 Average minor principal stresses (σ'3) and Young’s moduli (E) for Bothkennar
soil profile 90

Table 5.1 Influence of column compressibility upon settlement improvement factors for
end-bearing columns 105
Table 5.2 Influence of column strength upon end-bearing settlement improvement factors 106
Table 5.3 Influence of coefficient of lateral earth pressure upon end-bearing settlement
improvement factors 107
Table 5.4 Influence of column compressibility upon stress concentration ratios for end-
bearing columns 1146
Nomenclature ix
Nomenclature
Unless otherwise stated, the following abbreviations and symbols are used in this thesis.
However, when referring to specific publications, the original notation has been used.
Abbreviation:
CCET Cylindrical Cavity Expansion Theory
FEA Finite Element Analysis
FEM Finite Element Method
Symbols:
A Tributary area of soil per column in a large grid
AC Area of the stone column
A/AC Area ratio
B Width of square footing
CC, CS Compression and swelling indices, respectively
Ck Hydraulic change index (= ∆e/∆log(k))
D Diameter of circular footing
E Young's modulus
E50 Secant Young's modulus at 50% deviatoric stress
Eur Young's modulus for unload-reload
Ecol/Esoil Modular ratio
G Shear modulus
H Thickness of soil deposit
IL Liquidity index
IP Plasticity index
K0 Coefficient of lateral earth pressure at rest
KA, KP Coefficient of active and passive earth pressure, respectively
L Column length
N Diameter ratio (= de/dc)
Rf Failure ratio (Hardening Soil model)
Rinter Strength reduction factor (Hardening Soil model)
a Column radius
b Unit cell radius
c Cohesion
cu Undrained shear strength
Nomenclature x
cr, cv Coefficients of consolidation in radial and vertical direction, respectively

d Column diameter (= 2a)
de Unit cell diameter (= 2b)
e0 Initial voids ratio
fd Depth factor (cf. Priebe, 1995)
kvert, khorz Coefficients of vertical and horizontal permeability, respectively
mv Modulus of volume compressibility
m Power for stress dependency (Hardening Soil model)
n Settlement improvement factor (= suntreated/streated)
p, p' Mean principal total and effective stress, respectively
pA Applied pressure
ref
p Reference pressure (Hardening Soil model)
plim 'Limit pressure' at which indefinite expansion of a cavity occurs (cf. CCET)
q Deviatoric stress
s Column spacing
s Settlement of a finite group of stone columns
suc Settlement of an infinite grid of stone columns
s/suc Settlement ratio
u Vertical displacement
z Depth
w Water content
α Slope of s/suc versus B/L line

β Settlement reduction factor (= streated/suntreated)
γ Bulk unit weight
Shear strain
εy, εh Vertical and horizontal strain, respectively
κ Slope of unload/reload line on plot of υ versus ln(p')
λ Slope of isotropic normal compression line on plot of υ versus ln(p')
µ, υ Poisson's ratio
σ, σ' Total and effective normal stress, respectively
σ1, σ2, σ3 Major, intermediate and minor principal stress, respectively
σ'col/σ'soil Stress concentration ratio
σ'y, 0, σ'y, max In situ and maximum vertical effective stress, respectively
τ (= τ') Shear stress
φ Angle of internal friction
ψ Angle of dilatancy
1
Chapter 1
Introduction
1.1 Background
Vibro replacement is a popular form of ground improvement which is used to enhance
the settlement and bearing capacity characteristics of soft soils. Vertical columns of
compact stone are formed in the ground using either the top or bottom feed systems. The
vibro replacement process typically involves replacing 10–35% of the in situ soil with
crushed gravel. The high stiffness properties of the crushed gravel reduce the overall and
differential settlements of treated soil. The consolidation time is also reduced due to the
high permeability of the crushed gravel. Stone columns can be used to provide support
for a variety of loading scenarios ranging from small footings (i.e. pad/strip footings) to
wide area loadings (i.e. embankments and large floor slabs).
The design of foundations on soft soils is usually governed by settlement criteria rather
than bearing capacity due to their high compressibility. The majority of analytical design
methods developed to date contain many simplifying assumptions, such as the unit cell
concept, which assumes an infinite grid of columns supporting an infinitely wide load
area. Therefore, they do not account for the loss of lateral confinement associated with
groups of columns supporting small footings. In addition, the reduction in vertical stress
with depth beneath small footings is much sharper than that beneath wide area loadings
and, therefore, offers the possibility of partial depth treatment. While some correction
factors exist to account for the loss of lateral confinement, current design methods do not
consider the loss of lateral confinement for small groups of floating stone columns.
While the bearing capacity of small groups of stone columns has been well researched, a
dearth of information exists regarding the settlement performance of small groups of
columns. This is highlighted by McCabe et al. (2009) whose settlement database of over
20 case studies comprised only three case studies related to small groups of columns.
Similarly, the majority of numerical studies conducted are axisymmetric analyses on
Introduction 2
large groups of columns. The lack of information regarding the settlement performance
of small groups of stone columns was identified by Black (2006) who conducted some
high quality laboratory research; however, it is difficult to extrapolate the findings due to
scale effects associated with laboratory tests and also as some of the area ratios
considered are at the high end of typical values used in practice.
1.2 Aims and objectives

This research aims to investigate the loss of lateral confinement and reduction of vertical
stress with depth for small groups of floating stone columns. The following milestones are
outlined below to achieve this objective:
(a) Determine the influence of area ratio, column length and the number of columns (i.e.
column confinement) upon the settlement performance and deformational behaviour of
small groups of stone columns.
(b) For a select number of column configurations, examine the influence of column
arrangement, the position of columns relative to the footing edge, column
compressibility, column strength, column installation effects and the presence of a stiff
crust upon the settlement performance and deformational behaviour of stone columns.
(c) Develop a simplified design method which accounts for the loss of lateral confinement
and reduction of vertical stress with depth associated with small groups of floating stone
columns.
1.3 Methodology
A series of three-dimensional finite element analyses are proposed to investigate the
settlement performance and deformational behaviour of small groups of stone columns in soft
soil. PLAXIS 3D Foundation is a FE program which is specially developed for geotechnical
engineering and is adopted for this thesis. The stress-strain behaviour of the soil and stone
columns is simulated using advanced constitutive models. The soil profile adopted is that
of the well characterised Bothkennar test site, which consists of soft uniform clay overlain by
a stiff crust. The adopted soil profile and material parameters are validated and the analysis is
extended to study the influence of key design parameters such as area ratio, column length,
stiffness, strength, installation effects and the presence of a stiff crust.
Introduction 3
Two parameters are introduced to study the deformational behaviour of stone columns,
column compression and column punching ratios. These ratios are determined from the
displacement at the top and bottom of columns and are a simple method to investigate the
influence of various design parameters upon the mode of deformation of columns.
A more detailed analysis of load-transfer mechanisms for small groups of stone columns is
conducted by examining the distribution of total shear strains within stone columns and the
surrounding soil. This analysis is conducted for a select number of column lengths, which are
specifically chosen to display each mode of deformation. The effect of column spacing and
confinement upon column-column interaction and column-soil interaction for each mode of
deformation is identified.
The distribution of stress and strain along the column is also examined. These parameters
play a key role in the settlement performance of stone columns and allow a link between the
settlement performance and the deformational behaviour of columns to be established.
1.4 Outline of thesis

Chapter 2 provides a detailed description of the vibro replacement technique. A
comprehensive review into the deformational behaviour, settlement performance and
installation effects of vibro stone columns is presented. The main analytical design methods
used in practice to calculate the settlement of stone columns are also reviewed.
A background to the finite element analysis is presented in Chapter 3. The material models
used in PLAXIS 3D Foundation are described and the results of preliminary studies on mesh
sensitivity and boundary effects are presented. Other issues such as simulating the long term
behaviour of soft soil, modelling the column-soil interface and incorporating column
installation effects into the finite element model are discussed.
The accuracy of PLAXIS 3D Foundation to capture the load-deformational behaviour of

stone columns is assessed in Chapter 4. A series of load tests on stone columns in a layered
estuarine deposit are simulated and the predicted load-settlement curves are compared with
field measurements and two axisymmetric FEA. A soil profile and set of material parameters
are developed for the Bothkennar test site. The adopted soil profile and set of material
Introduction 4
parameters are validated by simulating a load test reported by Jardine et al. (1995) at the test
site. The settlement performance of an infinite grid of columns at the Bothkennar test site is
also simulated with PLAXIS 3D Foundation and compared with one-dimensional
compression theory, a settlement improvement database and analytical design methods.
The results from the primary analysis of this thesis regarding the settlement performance,
deformational behaviour and stress concentration ratios for small groups of columns are
presented and interpreted in Chapter 5. The interaction between column confinement, length
and area ratio and their influence upon the settlement performance of columns is examined.
Three modes of modes deformation are identified, which are referred to as "punching",
"block failure" and "bulging", at typical working loads levels. Finally, the relationship
between the stress concentration ratio at the surface and the mode of deformation is
examined.
The modes of deformation identified in Chapter 5 are verified by examining the distribution
of total shear strain, normal strain and stress concentration ratios within columns for specific
column lengths in Chapter 6. The distribution of shear strains within columns and the
surrounding soil allows the column-column and column-soil interactions to be investigated.
The characteristic behaviour of individual columns (i.e. corner, edge and centre columns) is
examined though the distribution of vertical and horizontal strains along the length of
columns. The variation of stress concentration ratio with depth is also examined and its
relationship with the mode of deformation is highlighted.
A simplified design method which accounts for the loss of lateral confinement and reduction
of vertical stress with depth for groups of columns supporting small loaded areas is developed
in Chapter 7. The findings from PLAXIS 3D Foundation are discussed in relation to previous
research and the existence of a critical length and the effect of the stiff crust are highlighted.
The influence of the mode of deformation upon the settlement performance of stone columns
is also discussed. Finally, the stress concentration ratios at the surface and at depth are
compared with a field and laboratory measurements and also analytical design methods.
Finally, the conclusions of the research are presented in Chapter 8. A better understanding of
small groups of stone columns is gained and the findings can be used by practitioners to
relate the settlement of small groups and infinite grids of columns.
5
Chapter 2
Literature review
2.1 Background
2.1.1 Introduction
Deep vibratory techniques such as vibro compaction and vibro replacement are forms of
ground improvement which are used to enhance the bearing capacity and settlement
characteristics of weak soils. The vibro compaction technique was developed in 1936 by the
Keller group and one of its first applications was to densify non-cohesive river-borne
granular soils (McCabe et al., 2007). The technique consists of lowering a vibrating poker
into the ground, which imparts horizontal forces to the surrounding soil and causes the soil
particles to re-arrange into a denser state. However, it was found that vibro compaction
reaches its technical and economic limits in saturated sands with high silt contents, as fine
particles attenuate the horizontal forces imparted by the vibrating poker (Sondermann &
Wehr, 2004).
The limitation of vibro compaction in cohesive soils (i.e. soils composed of a high proportion
of fine particles) was overcome in 1956 by the development of the vibro replacement
technique. As with the vibro compaction technique, a vibrating poker penetrates the soil to
form a borehole. The resulting borehole is then backfilled in stages with coarse aggregate,
which is compacted by re-lowering the poker. This process results in stone columns which
are tightly inter-locked with the surrounding soil. Stone columns can readily be formed up to
15 m and typically replace 10–35% of the in situ soil (McKelvey et al., 2004). Figure 2.1
illustrates how the vibro replacement technique extends the range of soils types treatable by
deep vibratory techniques. Vibro stone columns offer a cost-effective alternative to traditional
piled solutions when supporting moderately loaded residential, commercial and industrial
structures.
Other variations of vibro stone columns include geotextile-encased columns and vibro
concrete columns. Geotextile-encased stone columns are used in very weak soils where
Literature Review 6
insufficient lateral support is provided by the surrounding soil and excessive bulging of stone
columns occurs. The geotextile material develops tensile forces which constrain the column
and reduce bulging.
Figure 2.1 - Range of soil types treatable by vibro compaction and vibro replacement techniques (courtesy of
Keller Ground Engineering)
2.1.2 Construction processes

Poker
The vibrating pokers used for the dry and wet top feed systems are depth vibrators suspended
from a crane. The vibrating pokers are 300–500 mm in diameter and range in length from 2–5
m. This length can be increased with extension tubes to reach depths of up to 26 m
(Sondermann & Wehr, 2004). The poker consists of a cylindrical steel shell and the
horizontal vibratory forces are generated from an eccentric weight located in the lower
section. An elastic coupling is used at the top of the vibrator to prevent any vibratory forces
being transferred to the extension tubes. Supply tubes for compressed air and water, which
are used to aid the poker penetration, are also incorporated into the vibrator shell (Figure 2.2).
Dry top feed system

The poker is inserted into the ground which displaces the soil laterally. Compressed air is
delivered to the cavity as the poker is withdrawn from the ground in order to prevent the
cavity collapsing. A controlled volume of crushed stone is then tipped into the cavity from
the ground surface and subsequently compacted by re-lowering the poker. This process is
repeated until a stone column, which is tightly interlocked with the surrounding soil, is
formed up to the ground surface. This construction method is limited to competent soils that
can sustain an open borehole.
Literature Review 7
Figure 2.2 - Depth vibrator and principle of vibro compaction (Sondermann & Wehr, 2004)
Wet top feed system

The poker penetration for the wet top feed system is aided by water jets at the poker tip. The
jets support the cavity wall and dislodge any loose material during penetration. The flushed
out material is carried upwards through the annulus surrounding the poker and treated in
settlement lagoons. Field experience indicates that penetration is more effective when a larger
volume of water is used, rather than a higher pressure (Sondermann & Wehr, 2004). Once the
design depth is reached, the poker is raised slightly and crushed stone is added to the cavity
through the annulus surrounding the poker. The poker is then re-lowered to compact the stone
and this process is repeated until a stone column is formed up to the ground surface. This
system is not commonly used due to the onerous task of treating the flushed out material and
also as the water jets create a smear zone surrounding the cavity which reduces the drainage
capacity of stone columns.
Dry bottom feed system

A specially developed machine called a 'vibrocat' is used for the dry bottom feed system
(Figure 2.3). The vibrocat is different from traditional crane suspended vibrators as the
vibrator is mounted on leaders, thus ensuring vertical columns. The vibrator can also generate
pull down forces using a hydraulic winch to aid poker penetration. The supply tube for the
Literature Review 8
crushed stone is attached to the vibrator and bends inwards at the tip. Therefore columns are
formed without removing the poker, which allows the poker to case the borehole thereby
preventing collapse. Once at the design depth the poker is raised by 100–500 mm, depending
on soil conditions, and crushed stone flows into the cavity through the supply tube. The poker
is then raised and lowered a number of times until satisfactory compaction of the stone
backfill is achieved. This cycle is repeated until a column of compact stone is formed up to
the ground surface.
Figure 2.3 - Dry bottom feed system using ‘vibrocat’
2.1.3 Applications of vibro replacement

Settlement Control
Vibro replacement can be used to support a variety of structures in soft cohesive soils.
Columns can be spaced on large grids to support wide area loadings or in small groups
beneath pad/strip footings. The area ratio (A/AC) is a measure of the proportion of soil
replaced by stone, where A is the tributary area of soil per column in a large grid and AC is
the area of the stone column (Figure 2.4). The area ratio for footings is determined as the total
area of the footing divided by the total area of the columns.
Literature Review 9
1.5
= = =
# ×
(i) (ii) (iii)
Figure 2.4 - Area ratios for (i) square grids, (ii) triangular girds and (iii) pad footings
Vibro replacement primarily enhances the settlement performance of treated soils due to the
high stiffness of granular columns. If equal strain (i.e. uniform deformation) is assumed
beneath the base of footings, it can be shown from equilibrium that a larger proportion of the
applied load is taken by the stiffer stone columns (Figure 2.5). Therefore, stress concentration
ratios develop in stone columns. Field experience indicates that stress concentration ratios
typically range from 2.5–5.0 (Barksdale & Bachus, 1983). The stress concentrations which
occur in columns reduce the vertical stress on the surrounding soil which, consequently,
reduces the settlement.
Figure 2.5 - Development of stress concentration ratios in stiffer stone columns
A comprehensive review of over twenty case studies by McCabe et al. (2009) highlights the
effectiveness of vibro stone columns in enhancing the settlement performance of soft soils. In
addition to reducing the magnitude of settlement, vibro stone columns also homogenise
treated soil deposits and thus reduce differential settlement. Therefore, it is not necessary to
span ground beams between compaction points, as is required with piled solutions, and
shallow footings can be used to support the foundations.
Literature Review 10
Vibro stone columns also act as vertical drains due to the high permeability of the stone
backfill. This allows excess pore pressure to dissipate rapidly which decreases consolidation
time. However, the drainage capacity of stone columns may be reduced slightly by the
creation of a smear zone around the vibrating poker upon penetration.
Slope stability
The vibro replacement technique can be used to increase the stability of embankment side
slopes if stone columns are installed deeper than the rotational failure surface. Stone columns
have a higher shear strength than the surrounding soil and therefore increase the shearing
resistance along the failure surface. Moreover, stress concentrations ensure that a larger
proportion of the applied load is carried by the stronger column material. The effectiveness of
stone columns at increasing the stability of two full scale test embankments, founded in deep
deposits of very soft cohesive soil, was investigated by Munfakh et al. (1984). The authors
report that in addition to reducing the settlement and consolidation time, stone columns
increased the load-carrying capacity of in situ soil by approximately 50%.
Mitigation of liquefaction
Liquefaction is initiated when excess pore pressures generated from seismic motions increase
to equal the inter-granular stresses. As a result, soil loses all its shear strength and large
deformations follow. The vibro replacement technique mitigates against liquefaction by
densifying in situ soils and acting as vertical drains. The initial poker penetration densifies
granular soils which increases inter-granular forces between soil particles. Stone columns
also act as vertical drains which allows for a rapid dissipation of excess pore pressure. Vibro
stone columns were employed to support a wastewater treatment plant in Santa Barbara,
USA, which was located in a seismically active zone affected by liquefaction (Mitchell &
Huber, 1985).
2.1.4 Limitations of vibro replacement

Barksdale & Bachus (1983) state that the vibro replacement technique is most effective in
very soft to soft compressible clays/silts and in loose silty sands. However, vibro stone
columns are highly dependent on the lateral support provided by the surrounding soil and
consequently are not suitable for use in very weak soils. If the strength of the surrounding soil
is insufficient, excessive column deformation will occur. Experience indicates that stone
columns should not be installed in soils which contain layers of peat or decomposable
organics greater than 1d–2d (d = column diameter) in thickness (Barksdale & Bachus, 1983).
McCabe et al. (2009) state that considerable evidence exists to suggest that stone columns
can be formed successfully in soils with undrained shear strengths (cu) well below 15–20 kPa,
which is often quoted as the lower practical limit of the system. A small scale laboratory
study conducted by Wehr (2006a) indicates that columns can be installed in soil with cu ≥ 4
kPa.
The vibro replacement technique is also not appropriate in high sensitivity soils. The shear
strength of high sensitivity soils dramatically reduces once the grain structure is disturbed.
Therefore, the vibratory forces imparted from the poker during column installation may
remould the soil and, thus, significantly reduce the soil strength.
2.2 General deformational behaviour of stone columns

2.2.1 Behaviour of single columns
Hughes & Withers (1974)
The behaviour of single stone columns was analysed by Hughes & Withers (1974) in small
scale laboratory tests. The deformation of columns was observed by taking radiographs of
lead shot markers which were located in the clay. The behaviour of various column lengths
was investigated, with length to diameter ratios (L/d) ranging from 4 to 12. Columns were
loaded to failure in a stress-controlled manner which ensured drained conditions in the clay.
While significant column bulging was observed in upper sections, negligible strain occurred
beyond a depth of 4d (Figure 2.6). Therefore the ultimate strength of single columns is
dependent upon the lateral resistance provided by the surrounding soil in the bulging zone
and installing stone columns beyond 4d will not enhance the ultimate strength. The authors
postulate that uniform bulging occurs in the upper section of the column and use cylindrical
cavity expansion theory to determine the ultimate column capacity. It is also suggested that
columns shorter than 4d would fail by punching, as the shear stress and end-bearing pressure
are lower than the bulging capacity for columns shorter than L < 4d. Another interesting
finding from the laboratory study is that only the clay within a cylinder of diameter of 2.5d is
significantly strained, which suggests that 2.5d may be the upper limit for the spacing of
stone columns.
(a) (b)
Figure 2.6 - (a) Vertical displacement within the column against depth and (b) Radial displacement at the edge
of the column / initial column radius against depth (Hughes & Withers, 1974)
Narasimha Rao et al. (1992)

Small scale laboratory tests were conducted by Narasimha Rao et al. (1992) to examine the
ultimate bearing capacity of single stone columns. The influence of the column diameter,
length and footing diameter was investigated by varying the L/d ratio from 3 to 9 and
examining a range of footing diameters, D = 1d–2d (Figure 2.7). The tests indicate that the
ultimate capacity of stone columns increases with the L/d ratio, which the authors suggest is
the governing factor of column capacity. A critical L/d ratio appears to exist between 5 and 7,
beyond which no further increase in ultimate capacity is gained. The ultimate capacity of
stone columns was also observed to increase with footing size; however, this increase was
only noticeable for columns shorter than 7d.
Figure 2.7 - Typical test setup examined by Narasimha Rao et al. (1992)
2.2.2 Behaviour of column groups

Muir Wood et al. (2000)
Muir Wood et al. (2000) conducted a series of laboratory tests in reconstituted kaolin clay to
examine the deformational behaviour of a group of floating stone columns. The spacing,
length and diameter of columns were all varied, with L/d ranging from 6 to 15. The loading
was applied through a rigid circular footing at a constant rate of displacement, which was
slow enough (0.061 mm/min) to ensure drained conditions in the clay.
The findings from the laboratory tests were verified by a plane strain Finite Element Analysis
(FEA). Four modes of deformation were observed; bulging, shearing, punching and bending.
The occurrence of column bulging was found to be dependent on the lateral restraint provided
by neighbouring columns (A in Figure 2.8); the magnitude of bulging increases and the depth
of bulging becomes shallower with increasing area ratio (A/AC). Shearing of columns
occurred when low lateral restraint is provided to columns and is most noticeable beneath the
edge of footings near the ground surface (B in Figure 2.8). The zone of influence for the
footing may be defined by a cone of undeforming soil, located directly beneath the footing
and bound by the shear planes occurring in the soil. The depth of the zone of influence is
dependent on the mobilised angle of friction of the composite soil mass, which increases at
low A/AC. Punching of columns is evident in Figure 2.8(a) and occurs when columns are not
long enough to transfer the load to depth; this mode of deformation becomes more
pronounced for closely-spaced columns, i.e. low A/AC. Bending of stone columns occurs
when columns are subject to lateral loads and can be seen in Figure 2.8(c).
The findings from the laboratory tests also indicate that columns located at mid-radius (i.e.
between central and edge columns) are the most heavily loaded, followed by central columns
and then edge columns. The authors attribute the reduced load carrying ability of edge
columns to their decreased lateral restraint. However, the reason for the reduced performance
of central columns is less clear, as the stiffness of the soil increases towards the footing
centre. The authors also suggest that a critical length may exist; however, the dominant strain
in the column is dependent on the footing diameter (D) and not the column diameter. Wehr
(2004) simulated the laboratory tests with a plane strain FEA and suggests that a critical
column length exists at 1.5D, where a marked change in column behaviour occurs. Central
and edge columns shorter than this length exhibit punching and buckling failure, respectively.
(a) (b) (c)

Figure 2.8 - Photographs of deformed sand columns exhumed at the end of footing penetration (Muir Wood et
al., 2000)
The findings from the laboratory tests also indicate that columns located at mid-radius (i.e.
between central and edge columns) are the most heavily loaded, followed by central columns
and then edge columns. The authors attribute the reduced load carrying ability of edge
columns to their decreased lateral restraint. However, the reason for the reduced performance
of central columns is less clear, as the stiffness of the soil increases towards the footing
centre. The authors also suggest that a critical length may exist; however, the dominant strain
in the column is dependent on the footing diameter (D) and not the column diameter. Wehr
(2004) simulated the laboratory tests with a plane strain FEA and suggests that a critical
column length exists at 1.5D, where a marked change in column behaviour occurs. Central
and edge columns shorter than this length exhibit punching and buckling failure, respectively.
Woo-Seok Bae et al. (2002)

Woo-Seok Bae et al. (2002) examine the horizontal resistance at the stone-soil interface for
various configurations of stone columns by conducting small scale laboratory tests, the
findings of which were verified by FEA. It was observed that failure of single end-bearing
stone columns occurs by bulging between depths 1.6d–2.8d, and also that failure is dependent
on the column diameter as opposed to the length. The failure zone for a group of stone
columns is a conical zone of undeforming soil, which corresponds to Muir Wood et al.
(2000). The depth of the conical zone of undeforming soil was found to increase for longer
columns and at lower area ratios (A/AC). It was also observed that column bulging occurred
deeper at lower A/AC, which is attributed to the increased lateral confinement. The influence
of a granular mat, which acts as a stress distribution layer, also tends to push column bulging
deeper. The authors conclude that the ultimate capacity of stone columns is influenced by
column spacing and diameter rather than column length and the granular mat.
McKelvey et al. (2004)

McKelvey et al. (2004) investigated the deformed shape, bearing capacity and stress
concentration of small groups of stone columns beneath circular, strip and pad footings. The
tests were conducted on floating stone columns, with lengths L = 6d and 10d, in both a
transparent material and kaolin clay. The transparent material, with clay-like properties,
allows for the deformed shape of the columns to be observed throughout loading. Columns
tended to deform away from neighbouring columns and bulging in central columns beneath
strip footings was significantly reduced, possibly due to the restraining effect of the two
adjacent columns (Figure 2.9). It was also observed that short columns bulged over their
entire length and tended to punch into the underlying soil. This is in contrast to long columns
which exhibit no punching and only bulged in their upper regions. The authors report that a
negligible increase in the bearing capacity of stone columns is gained by increasing column
length beyond 6d. However, the undrained stiffness of footings is observed to increase for
columns longer than 6d, which suggests that there is some merit in installing columns longer
than this critical length for settlement control. For the initial stages of loading, it was
observed that long stone columns carried a significant portion of the load (σcol/σsoil > 4) in
comparison to short stone columns (σcol/σsoil < 2). The stress concentration ratio for both long
and short stone columns tends to a constant value (σcol/σsoil = 3) at higher loads.
Ambily & Gandhi (2007)

Ambily & Gandhi (2007) investigated the behaviour of stone columns using small scale
laboratory experiments and an axisymmetric FEA. The tests are performed on kaolin clay in a
cylindrical tank and examine the influence of various design parameters. The findings
indicate that the ultimate capacity of stone columns reduces as column spacing increases and
that the loss of ultimate capacity is negligible for columns spaced further than 3d apart. It was
also observed that the ultimate column capacity increases with undrained shear strength,
surcharge load and the angle of internal friction of the stone backfill. Bulging was observed at
a depth of 0.5d when the column area alone was loaded, whereas no bulging was observed
when the entire area was loaded.
(a) (b)
Figure 2.9 - Photographs of sand columns beneath strip footings at beginning, middle and end of foundation
loading process: (a) L/d = 6; (b) L/d = 10 (McKelvey et al., 2004)
Wehr (2004)
Wehr (2004) examined the behaviour of a single stone column and a group of stone columns
using a plane strain FEA, which employed an elasto-plastic constitutive law in a Cosserat
continuum. The results from the FEA for the single column and the group of columns are
compared with experimental studies by Brauns/Witt (1978) and Hu (1995), respectively. A
wedge of undeforming stone, which induces radial and vertical displacement, is observed
from the experimental tests on a single stone column. This failure mechanism, along with a
secondary failure mechanism in the form of a shear plane along the column-soil interface, is
identified from the FEA. A wedge shaped zone of undeforming soil, which is bounded by
shear planes, is also observed from the laboratory tests on a group of stone columns (Figure
2.10). The shear zones intersect beneath the centre of the footing and cause excessive bulging
to occur in central columns. In addition to inclined shear planes, the FEA identifies extra
shear zones along the interface of internal and external stone columns. The author suggests
that the shear zones between the internal columns may unite if the spacing of the stone
columns is reduced and the lateral deformation is limited. The unification of shear zones is
also dependent upon the relative movement of the column and soil. Another interesting
observation from the FEA is that the thickness of the inclined shear planes increases with
column length and that the ‘bulging’ observed by Hu (1995) in outer columns, is actually a
large shear zone. The influence of an edge row of columns, as investigated by Hu (1995), is
quantified using the FEM and its influence on bearing capacity was found to be negligible
(peak force increased by 7%).
(a) (b)
Figure 2.10 - Deformed group of (a) short columns and (b) long slender columns with additional column
outside the footing, upper part, u = 10 mm, clay (white) and column (grey) (Wehr, 2004)
Wehr (2006b)
Wehr (2006b) extended the previous analysis (Wehr, 2004) to study the effects of footing
flexibility. A wedge shaped zone of undeforming soil was not observed for a flexible footing
and columns tended to bulge rather than buckle. However failure mechanisms in the form of
a broad vertical shear zone at the edge of the footing and several approximately wedge
shaped and parallel shear zones were observed. The number of shear zones increased with the
flexibility of the footing. The shear zones also extend to a limited depth and are dependent on
the movement of the columns relative to the soil. Wehr (2006b) concluded that a flexible
footing has a better load-carrying capacity than a rigid footing due to the formation of wider
shear zones near the footing edge and the additional shear zones between columns.
2.2.3 Summary of general deformational behaviour of stone columns

The behaviour of single columns and groups of columns was reviewed in the previous
sections. Single columns tended to bulge or punch, depending on column length. A critical
length, beyond which no increase in column capacity is gained was observed for single stone
columns within a range L = 4d–7d.
It appears that column interaction influences the deformational behaviour of stone columns as
groups of stone columns exhibit bulging, punching, shearing and bending. The occurrence of
each mode of deformation was found to be dependent upon the geometrical properties of
columns and the footing, i.e. column spacing, length, position beneath footing and footing
diameter. The critical length for groups of columns was found to be dependent on the column
diameter and the footing diameter, with L = 1.5D and L = 6d suggested by Muir Wood et al.
(2000) and McKelvey et al. (2004), respectively.
The stress concentration ratio which develops in stone columns appears to be related to the
mode of deformation (McKelvey et al. 2004). Stress concentration ratios are low for columns
which are punching into the underlying soil and increase with column length to a maximum
when column bulging is occurring. It was also observed that the position of columns beneath
footings influences the magnitude of stress concentration ratio as edge columns carry a lower
proportion of the applied load at failure due to the loss of lateral confinement.
2.3 Settlement performance of stone columns

Vibro stone columns can be used to treat soft clayey soils, which are typically characterised
by poor strength, stiffness and drainage properties. Consequently foundations on these soils
undergo large displacements at relatively low loads and settlement is usually the governing
criterion in foundation design. The rate of settlement is also another important feature of
stone column design. Consolidation time can significantly affect the progress of construction
projects, such as embankments, and therefore increase the cost of projects. By adopting vibro
stone columns, consolidation time is reduced and the indigenous soil consolidates more
quickly. As a result, the shear strength of the soil develops at a faster rate which allows
embankments to be built in a shorter space of time.
2.3.1 Case studies

McCabe et al. (2009) conducted a comprehensive review of case studies involving the
treatment of soft soils by vibro stone columns. The authors collated a series of data points
from over twenty case studies in order to develop a ‘settlement improvement database’.
While the majority of the data points are specific to wide area loadings, three case studies
relate to pad and strip footings. Figure 2.11 compares the measured data points (for wide area
loadings) with the basic improvement curve, as derived by Priebe (1995), and assuming 40°
for the angle of internal friction of the stone backfill. It was necessary to assume this value as
insufficient data was provided in many of the case studies examined; the value is typical of
what would be used in design. While some scatter exists, it is clear that the measured data
follows a similar trend to that predicted by Priebe (1995).
Figure 2.11 - Settlement improvement factor against area replacement ratio for sites with widespread loading
(McCabe et al., 2009)
The authors also examine the effectiveness of different construction techniques by comparing
measured and predicted settlement improvement factors in Figure 2.12. It appears that the
construction technique has a significant influence on settlement performance of stone
columns and also that the bottom feed system is the preferred choice for installing vibro stone
columns in soft soils. Some of the data points in Figure 2.12 suggest that vibro stone columns
are behaving worst than predicted, which may be due to a number of factors such as
workmanship and uncertainties in the measured data. The authors stress the importance of
workmanship in installing stone columns and state that the disturbance caused by imparting
excessive energy in the installation sequence may offset the benefit of a stone column.
Uncertainties also exist with the measured data, such as the methods used to predict
settlement and the ‘as-built’ dimensions of columns. On consideration of these facts and the
excellent performance of the bottom feed system, Figure 2.12 indicates that the assumed
angle of internal friction of 40° for the stone backfill is somewhat conservative.
Figure 2.12 - Predicted versus measured settlement improvement factors for all widespread loadings and
footings (McCabe et al., 2009)
2.3.2 Laboratory studies

Black (2006) examined the behaviour of large diameter single stone columns and small
groups of columns beneath a circular footing (diameter = 60 mm), using small scale
laboratory tests. Three area ratios (A/AC = 2.5, 3.6 and 5.8) were adopted for the analysis and
the effect of column length was also investigated by varying the length L = 125–400 mm.
Based on the initial slope of the 'load versus settlement' curve for both area ratios, it appears
that the group of stone columns are under-performing as the stiffness for groups is half of that
for single columns (Figure 2.13). This is attributed to a ‘block failure’ that manifests itself in
column groups, where all the columns act together and punch into the underlying soil, and
also to the reduced shoulder length between the edge of columns and the edge of the footing
(Figure 2.14).
The deformation patterns of columns were found to be dependent on both the column length
and arrangement. It was observed that both single columns and column groups that were 125
mm long, with a length to diameter ratio (L/d) ranging from 3 to 5, all punched into the
underlying soil. As the length was increased to 250 mm, it was observed that the single
column (L/d = 7–10) bulged whereas the group of columns (L/d = 11–14) continued to punch
into the underlying soil. However, column groups act as a 'block' and re-defining the L/d
ratios on the basis that the diameter for the group is more appropriate (i.e. d = 60 mm) leads
to revised L/d ratios of 4 and 6. Punching is precluded for end-bearing columns and it was
Figure 2.13 - Comparison of Ks for isolated and group formation (Black, 2006)
Figure 2.14 - Illustration of block failure in group columns (Black, 2006)
observed that all 400 mm long columns bulged. Black (2006) suggests that the critical L/d
ratio which defines the cross-over from punching to bulging failure is approximately 8 for
these drained analyses.
Another interesting finding from the tests is the relationship between the stress concentration
ratio (σcol/σsoil) and column length. No increase of vertical stress was observed in short
columns, as columns failing by end-bearing cannot sustain any extra vertical stress. However,
stress concentration ratios are observed to increase with column length, which reflects the
higher ultimate capacity of columns which are bulging. It was also observed for column
groups that the pressure recorded beneath the centre of the footings was higher than in the
columns. The increased pressure in the centre of the footing imparts a lateral force on the
surrounding columns, forcing them to bulge outwards and away from neighbouring columns.
This may explain the reduced settlement performance of groups of columns compared to
single columns, which are positioned beneath the centre of the footing.
2.3.3 Numerical studies

Balaam et al. (1977)
Balaam et al. (1977) conducted a numerical analysis to investigate the influence of column
length, spacing and stiffness upon the settlement performance of stone columns. An infinite
array of stone columns supporting an infinite load area is modelled using the unit cell concept
(see Section 2.5.1) to simplify the analysis. A uniform pressure is applied to the column-soil
unit which simulates a flexible loading (i.e. large raft, embankment).
The long term settlement performance of stone columns is estimated using the FEM and
accordingly both the column and soil are modelled as drained materials. The authors
examined the settlement performance of stone columns using an elastic and elasto-plastic
analysis. While the elastic analysis was found to slightly under-predict the settlement of the
column-soil unit at working loads (maximum discrepancy = 6%), it was deemed sufficiently
accurate and was adopted for the subsequent analyses. It is shown in Figure 2.15(a) that the
ratio of the column and soil moduli has a minor influence on the settlement performance of
stone columns. It is also noted by the authors that columns shorter than L/h = ¼ (h =
thickness of soil deposit) or columns spaced wider than de/d = 5 (A/AC = 25) yield negligible
settlement improvement factors (n). It can be seen in Figure 2.15(b) that the influence of
column length is more pronounced for closely-spaced columns. It is also interesting to note
the identical settlement performance of floating columns spaced at de/d = 2 and end-bearing
columns spaced at de/d = 5, which occupy 12.5 % and 4 % of the soil volume, respectively.
This emphasises the importance of column length upon the settlement performance of large
arrays of columns.
The rate of settlement is determined from a finite difference solution, which uses diffusion
theory to approximate consolidation theory developed by Biot (1941). The diffusion theory is
(a) (b)
Figure 2.15 - Settlement variation with (a) EP/ES ratio and (b) degree of penetration of piles (Balaam et al.,
1977)
an extension of the Terzaghi theory of one-dimensional consolidation (Terzaghi, 1925). The

influence of increasing L/h significantly increases the rate of consolidation and this effect
becomes more pronounced at close column spacings.
Domingues et al. (2007)

Domingues et al. (2007a) and Domingues et al. (2007b) conducted axisymmetric FEA to
examine the influence of column spacing and compressibility upon an infinite grid of
columns supporting a 2 m high embankment. As the embankment is a wide area loading, it is
possible to model the behaviour of the column using the unit cell concept and a flexible
loading. The p-q-θ critical state model, which is an extension of the Modified Cam Clay
model into a three dimensional stress space using the Mohr-Coulomb failure criteria, is
adopted to simulate the behaviour of both the stone and the soft soil. The end-bearing
columns are 1.0 m in diameter and formed in 5.5 m deep normally consolidated clay. The
coefficient of lateral earth pressure of the clay is increased (K0 = 0.7) to account for the
effects of column installation; this was chosen as an intermediate value between K0 = 1 –
sin(φ') (Jaky, 1944) and K0 = 1.0 (Priebe, 1995).
The influence of the column spacing was examined by conducting analyses over a range of
area ratios A/AC = 3–10. As expected, settlement improvement factors increase (n =
1.2→2.0) at lower A/AC. However, no change in differential settlement was observed. Stone
columns act as vertical drains due to their high permeability and it was found that the rate of
consolidation increased at lower A/AC. The horizontal displacement at the column-soil
interface was also found to decrease with lower A/AC, which may be attributed to the
confining effects of neighbouring columns.
The influence of column compressibility was also investigated by varying the modular ratio
(λcol/λsoil) from 10 to 100, for columns spaced at A/AC = 5.3. It was observed that both
settlement improvement factors (n = 1.3→3.8) and differential settlement increase with
increasing column stiffness. The rate of consolidation is also influenced by the column
stiffness as the time taken to reach consolidation reduces for higher column stiffness, i.e. time
taken to reach δ = 0.9δmax (δ and δmax = average settlement and final settlement on the ground
surface, respectively) was 16 weeks and 10 weeks for λcol/λsoil = 20 and 100, respectively.
The stress concentration ratio was observed by Domingues et al. (2007b) to increase linearly
from 3.9 to 14.0 with increasing modular ratio.
Andreou & Papadopoulos (2006)

Andreou & Papadopoulos (2006) investigated the influence of the applied load, area ratio,
angle of internal friction and undrained shear strength upon the horizontal displacement of a
stone column. The column is subject to a wide area loading via a rigid footing and is
modelled using the unit cell concept. The horizontal displacement is estimated using an
axisymmetric FEA, where the stone and the soil are modelled using the simple Mohr-
Coulomb model. The 20 m long end-bearing stone columns are 0.8 m in diameter. The depth
of plasticity increases in the column with increasing load (10–120 kPa) and settlement
improvement factors decrease (n = 3.1→1.9). It also appears that decreasing the area ratio
(A/AC = 14.2→5.1) and increasing the angle of internal friction (φ = 38→44°) reduces the
extent of the plastic zone in the column, which improves the settlement performance. Finally,
the authors observed that increasing the undrained shear strength of the surrounding soil (cu =
10→60 kPa) did not influence the horizontal displacements of the column.
Kirsch & Sondermann (2003)

The stress concentration occurring in stone columns is a critical parameter in numerous
settlement and slope stability design methods. Kirsch & Sondermann (2003) investigated this
parameter and compared the estimates from a FEA with two analytical solutions and a case
study. The authors present the results of the development of stress concentration as a function
of applied loading for an embankment site in Kuala Lumpur. The curve presented in Figure
2.16 indicates that the stress concentration increases with applied load; this curve is specific
to columns installed on a square grid (A/AC = 4) to 14 m, in a 16 m deposit.
Figure 2.16 - Measurement of stress concentration below an embankment (Kirsch & Sondermann, 2003)
Wehr & Herle (2006)

Wehr & Herle (2006) determined the magnitude and rate of settlement for an embankment
supported by vibro stone columns using a FEA and analytical design methods. The 9 m high
embankment was supported by 6 m long floating stone columns, which were 0.9 m in
diameter and spaced on a triangular grid at A/AC = 6.3. The embankment was constructed in
three distinct phases: (i) the initial 6 m was constructed in 40 days; (ii) rest period for
consolidation of 120 days; (iii) the final 3 m was constructed in 40 days. The magnitude of
the settlement was estimated using an analytical software program which is based upon
Priebe (1995). The rate of settlement was determined using Balaam & Booker (1981), which
is based upon Biot’s consolidation theory. A two-dimensional plane strain FEA was
undertaken where the stone and the soil are both modelled using the Mohr Coulomb model.
The stone columns were converted to two-dimensional stone walls for the FEA, by keeping
the volume of the improved soil constant. The results indicate that the stone columns
significantly reduce the settlement of the embankment (up to 40%) and also that
consolidation occurs very quickly (within 16 days). The predictions from both the FEA and
the analytical design method agree very well. The authors attribute this to the fact that the
same material model is used for both the analytical design methods and the FEA.
Consolidation also occurs very quickly which allows for compatibility between the analytical
method and the FEA.
Elshazly et al. (2008a)

Elshazly et al. (2008a) conducted an axisymmetric FEA to investigate the applicability of
settlement design methods based upon the unit cell concept (see Section 2.5.1) to groups of
stone columns. The majority of design methods are based on the unit cell concept which
assumes an infinite grid of columns supporting an infinitely wide load. However, this concept
is clearly not applicable to groups of columns as exterior columns along the edge of the
footing experience a loss of lateral confinement and also the distribution of vertical stress
decays with depth beneath small footings.
Elshazly et al. (2008a) develop modification factors to relate the settlement of finite groups
of columns (s) with the settlement of an infinite grid of columns (suc). The column length (L)
and area ratio (A/AC = 3.4) were both constant, while the number of columns beneath
footings was varied to yield normalised footing widths (B/L) from 0.5 to 4.7. Two soil
profiles were adopted for the investigation; the first soil profile is a layered estuarine deposit,
which is described in detail by Mitchell & Huber (1985), and the second soil profile is
developed from typical parameters for soft soil. The ratio of the Young’s modulus of the
stone column to the soil ranges from 1.3 to 2.6 for the first soil profile and is 8.5 for the
second soil profile. The installation effects associated with stone column installation are
incorporated by increasing the coefficient of lateral earth pressure (K0) of the soil
immediately surrounding the stone columns to 1.5 and 1.2 for the first and second soil
profiles, respectively. The loads are applied through a stone distribution blanket and the long
term settlements are determined using an axisymmetric FEA.
The authors identify two counteracting effects which are associated with the unit cell model.
The first effect concerns the distribution of vertical stress beneath pad footings and wide area
loadings. Vertical stress reduces sharply with depth beneath small footings and is negligible
beyond a depth of z = 2B. This is in contrast to the unit cell which stresses the soil profile to
the full depth and therefore increases settlement. The second effect is related to the loss of
lateral confinement that is experienced by exterior columns beneath small footings, which
increases the settlement. However, full lateral confinement is provided on all sides of
columns within an infinite grid, which results in enhanced lateral support and less settlement.
It can be seen in Figure 2.17(a) for a layered deposit that the settlement of small footings
(B/L < 2) is less than the settlement of a unit cell. Therefore it appears that the beneficial
effect of stress reduction with depth overrides the loss of lateral confinement for small
footings. However, as footing width increases (B/L > 2) it can be seen that the settlement of
small footings exceeds the settlement of a unit cell. This may be attributed to an increase in
the vertical stress in the soil with increasing footing width.
It can be seen in Figures 2.17(b) and 2.17(c) for the soft soil deposit that the settlement of
finite groups is always less than that of a unit cell. The increased footing settlement may be
attributed to the low lateral support provided by the soft soil deposit.
(a)
(b)
(c)
Figure 2.17 - Settlement correction factor versus size ratio (a) case of layered soil, (b) case of 10.8 m thick soft
clay layer and (c) case of 30 m thick soft clay layer (Elshazly et al., 2008a)
Kirsch (2008)
Kirsch (2008) conducted a FEA on two footings, which are supported by 25 columns.
Installation effects, which are based on a previous study by Kirsch (2006), are incorporated
into the FE model. The stress concentration ratio (σcol/σsoil) was measured in a sand layer
above the stone columns and was found to increase with load from 1.4 to 1.6 for central
columns and from 1.4 to 2.8 for corner columns. The measured stress-concentration ratios
increased upon reloading. The FEA is compared with field data and despite over-predicting
the ultimate capacity of columns it predicts the settlement behaviour quite well at working
loads. The influence of increasing the column length and the angle of internal friction is
found to be more pronounced at low A/AC. Another interesting finding from the study is that
column stiffness has a negligible influence on the settlement behaviour of the footing. Kirsch
(2008) suggests that columns undergo plastic deformations at relatively low loads due to
stress concentrations and are therefore insensitive to elastic stiffness parameters.
2.3.4 Summary of settlement performance of stone columns

A review of the settlement performance of stone columns from field, laboratory and
numerical studies was presented in the previous sections. The settlement performance of
large groups of columns is well understood. Field experience indicates that Priebe (1995)
predicts the settlement performance quite well and also that an angle of internal friction for
the stone column φ' = 40° is somewhat conservative for the bottom feed system. The
influence of column length and area ratio upon the settlement performance of stone columns
is very important and the arrangement of columns can be tailored to achieve a specific
settlement performance. It was found that the influence of column length becomes more
pronounced for large groups of closely-spaced columns and that long, widely-spaced columns
require less stone than short, closely-spaced columns to achieve the same settlement
performance. It was also found that columns shorter than L/h ≤ ¼ (h = thickness of soil
deposit) and wider than A/AC > 25 yield negligible settlement improvement factors.
The settlement performance of small groups of stone columns was also reviewed. It appears
that the mode of deformation changes with the number of columns, as closely-spaced
columns act together and punch as a 'block' into the underlying soil. A comparison of small
groups of columns and infinite grids of columns highlights the importance of the lateral
support to columns and the distribution of vertical stress beneath pad footings in determining
the settlement of small groups of columns. The influence of column length and the angle of
internal frictionn for the stone columns w
were found to be more pronounced at low A/AC.
Stress concentration ratios were found to vary with applied load and also with position
beneath pad footings, as corner columns carry a higher proportion of the vertical load than
centre columns.
2.4 Column installation effects

As the vibrating poker penetrates the ground, it displaces the soil and imparts horizontal
vibratory forces to the surrounding soil. However, horizontal forces are attenuated by fine
2004) and any alteration to the in situ stress state of fine
particles (Sondermann & Wehr, 2004
grained soils may, therefore, be attributed to the displacing effects of the poker. Yu (2000)
states that soil displacement due to the installation of a driven pile can be predicted using
cylindrical and spherical cavity expansion theories along the shaft and at the tip of the pile
respectively, and that the soil near the ground surface is locally affected by surface heave
(Figure 2.18).
Figure 2.18 - Soil displacement due to pile installation (Yu, 2000)
2.4.1 Laboratory investigations

Randolph et al. (1979a) conducted small scale laboratory tests to investigate soil
displacement patterns associated with the installation of a closed ended pile. As the pile tip
advances, the soil particles are displa
displaced
ced downwards and outwards, similar to that predicted
by spherical cavity expansion theory. However, once the pile tip passes the soil particles,
significant radial displacement occurs. The displacement of the soil reaches a steady state
relative to the advancing
vancing pile tip once the pile tip is more than 4d
4d–5d
5d (d = pile diameter)
beyond the soil particles. The ultimate displacement pattern of the soil is similar to that of
cylindrical cavity expansion.
The use of a video camera to observe soil displacement surrounding the installation of a flat
bottomed penetrometer was adopted by Gill & Lehane (2001). This non-intrusive method
determines the soil displacement by tracking the movement of dark beads against a light
background, which consisted of a transparent artificial material. This artificial material has
properties similar to that of a lightly over-consolidated natural clay and consists of silica
particles and paraffin. The findings show that the soil is displaced both outwards and
downwards as the penetrometer tip advances towards the beads. Once the tip passes the
beads, the soil displacement becomes radial and upwards. It was also observed that negligible
displacement occurred once the penetrometer tip was more than five penetrometer diameters
deeper than the level of the beads. The results compare well with experimental data from
other authors, which reinforces the assumption by Baligh (1985) that the kinematics of deep
penetration are such that the displacement paths are relatively insensitive to the properties of
the soil.
2.4.2 Observed field measurements

Kirsch (2006) analysed the changes to the in situ stress regime in sandy silt due to the
installation of two groups of twenty five stone columns. The variation of pore water pressure,
effective horizontal stress and soil stiffness was analysed in order to determine the post-
installation stress state of the soil. A rise in pore water pressure was recorded at various
locations immediately after installation. The data was compared with theoretical values,
predicted using an analytical formula developed by Cunze (1985) based upon cylindrical
cavity expansion theory. A good relationship was found to exist between the field and
theoretical data, although there was considerable scatter in the results. The horizontal stress
and soil stiffness, shown in Figures 2.19 and 2.20, respectively, both appear to be increased
within a zone which ranges from 4d to 8d from the centreline of columns. The horizontal
stress and soil stiffness increased as the vibrating poker advanced towards the measuring
locations. However, increases in horizontal stress and soil stiffness were offset by the effects
of remoulding and dynamic excitation within a distance 4d.
Figure 2.19 - Factor of restraint measured during the installation of stone columns (Kirsch, 2006)
Figure 2.20 - Development of ground stiffness during column installation (Kirsch, 2006)
Castro (2007) recorded the development and dissipation of excess pore water pressure which
resulted from the installation of seven stone columns in a normally consolidated clay. The
pore water pressure rose upon poker penetration and peaked as the poker tip passes the level
of the piezometers. Despite significant heave occurring at the ground surface, it was deemed
reasonable to assume that plane strain conditions exist and cylindrical cavity expansion
theory may be used to simulate the poker installation. The increase in excess pore water
pressure recorded during the initial poker penetration agreed favourably with theoretical
values, which were predicted using an analytical formula derived by Randolph et al. (1979b).
An additional assumption was required that the undrained shear strength of the surrounding
soil iss reduced, due to the increase in excess pore water pressure, which is a common
phenomenon associated with driven pile installation. However, the agreement between field
and theoretical values breaks down with the installation of subsequent columns, as the
assumed boundary conditions of plane strain are violated. The dissipation of excess pore
water pressure occurred very quickly (within 15 minutes), which is 100 times less than the
theoretical dissipation time, predicted using a finite difference technique based upon theory
developed by Soderburg (1962). This is attributed to fractures in the clays, which result from
the high pressures due to column installation and act as drainage channels.
Gäb et al. (2007) measured the pore water pressure and settlement due to the construction of
a large embankment. The embankment was founded on 11 m of loose-medium dense
lacustrine sand which was underlain by 50 m of clayey silt. The stone columns were installed
to 14.5 m and on a triangular grid at A/AC = 7.7. In total 37 columns were installed in 4 rings,
with construction starting at the outer ring and progressively moving towards the centre. The
excess pore water pressure increased as the construction sequence moved closer to the
piezometers. This increase in excess pore water pressure was a maximum at 12 m, near the
column base and was even noticeable at 20 m depth. The excess pore water pressure
dissipated very quickly in the sand (< 1 day) and very slowly in the clay. Slight heave was
also measured at the site.
Egan et al. (2008) collate data relating to the installation of stone columns from various sites
including that of Castro (2007). The authors observe that heave may occur during the
installation of stone columns and tentatively suggest a relationship between column density
and the amount of heave. The authors suggest that heave is also a function of column size,
spacing and construction method. The footing arrangement also influences the amount of
heave as small groups and strips of stone columns do not generate as much heave as large
grids.
2.4.3 Simulation of installation effects in numerical models

Increased coefficient of lateral earth pressure
The installation effects associated with stone columns may be accounted for by increasing the
coefficient of lateral earth pressure (K0) in the surrounding soil. This relatively simple
technique has been adopted by many authors, as shown in Table 2.1. Elshazly et al. (2006)
conduct an axisymmetric FEA to determine the increase in K0 by back-calculating field load
tests on stone columns, which are described by Mitchell & Huber (1985). A central column
located within a large array of columns is loaded by a circular footing. The stone columns
were installed using the ‘wet technique’ to a depth of 10.8 m, in a layered estuarine deposit.
The soil properties adopted are determined from post-installation soil samples and therefore
incorporate any change to soil properties due to column installation. The authors then
attribute any discrepancy in the load-settlement curve to a change in the stress state of the
surrounding soil and adjust K0 until the load-settlement curves match. Assuming full
confidence in the adopted soil parameters, an axisymmetric FEA indicates that K0 = 1.5
yields the best match between the actual and predicted load-settlement curves.
Elshazly et al. (2008b) extended to the previous analysis to investigate the effect of column
spacing upon the post-installation stress state of the soil. A range of column spacings were
analysed, which corresponded to A/AC ranging from 2.5 to 4.8. The back-calculated values
for the K0 ranged from 0.7 to 2.0, depending on the level of confidence in the soil parameters.
The best conservative estimates of K0 are 1.7, 1.2 and 0.85 for A/AC = 2.5, 3.7 and 4.8,
respectively. This indicates that installation stresses reduce at high A/AC (i.e. widely-spaced
columns).
Table 2.1 - Increase in K0 to account for stone column installation effects by various authors
Reference A/AC Coefficient of lateral earth pressure, K
Balaam & Booker (1977) 4 – 100 1.00
Barksdale & Bachus (1983) 4 – 10 0.75
Mitchell & Huber (1985) 2.0 – 4.9 1.00
Elshazly et al. (2006) 3.4 1.50
Domingues et al. (2007a) 3.3 – 10.0 0.70
Elshazly et al. (2008b) 2.0 – 4.9 0.85-1.70
Cylindrical cavity expansion

Debats et al. (2003) and Guetif et al. (2007) both simulate the installation effects associated
with stone columns using an axisymmetric FEA. The authors apply an undrained cylindrical
expansion to a ‘dummy material’ and then convert this material into a stone column. In order
to replicate the actual installation process, it would be necessary to expand the cavity from a
zero initial diameter to the final column diameter. However, this is theoretically impossible to
implement in a numerical model as infinite strain would be generated. A numerical analysis
by Randolph et al. (1979a) in the context of pile installation reveals that expanding a cavity
by doubling the radius is sufficient to simulate expanding a cavity from a zero initial
diameter. The ‘dummy material’ is expanded from an initial diameter of 500 mm, which is
close to a typical poker diameter, to a final column diameter of 1100 mm. A ‘dummy
material’, which is assigned a nominal stiffness, is adopted for this procedure in order to
reduce high stresses that would occur in the column during the expansion.
Both authors conducted numerical investigations into the influence of this technique upon the
stiffness development of the surrounding soil and zone of influence of the expanding column.
The increase in soil modulus is estimated from the change in mean effective stress, using an
experimental relationship developed by Biarez et al. (1998). It is postulated by Brinkgreve &
Broere (2006) that soil stiffness is directly proportional to the mean effective stress for soft
soils. The undrained cavity expansion induces large excess pore water pressures in the
surrounding soil, which are magnified at low A/AC. Following consolidation of the soil, the
mean effective stress and hence the soil stiffness are increased. Debats et al. (2003) report a
30% and 40% increase in soil stiffness within a cylindrical zone of radius 2d in the
surrounding soil for column spacings of 6 m and 10 m, respectively. However, these findings
are specific to the Mohr-Coulomb model and even higher increases in soil stiffness in a
greater zone of influence are observed when the Hardening Soil model is adopted. Guetif et
al. (2007) also observe that a stress concentration occurs in the stone column and that the
lateral coefficient of earth pressure is increased above unity in the surrounding soil.
Cylindrical cavity expansion & increased soil stiffness

Based on measurements from field tests on two groups of twenty five columns, Kirsch (2006)
simulates the installation effects of stone columns by applying both an individual cylindrical
expansion to each column and also by increasing the stiffness in an enhancement zone
surrounding the footing. Kirsch (2006) found that the best match between the predicted and
actual 'load-settlement' curves was achieved by applying a moderate cylindrical expansion to
the stone column (lateral strain, εr = 4%) and doubling the stiffness in the enhancement zone,
which exists between 2d and 5d from the centre-line of outer row columns. The load-
settlement performance of both footings was back-calculated using the FEM (incorporating
the installation effects), which was found to agree favourably with field data and analytical
design methods (Priebe, 1995 and Goughnour & Bayuk, 1979).
Kirsch (2006) investigated the influence of both the individual and the global column
installation effects upon the settlement improvement factor for a 7.2 m square footing,
supported by 25 columns. The results indicate that increasing the radial expansion of the
individual column to 8% results in a 45% increase in the settlement improvement factor. An
examination of the global installation effects suggests that increasing the stiffness of this zone
(2d–5d from the centre-line of the outer row columns) by three times its initial value yields a
25% increase in the settlement improvement factor. However, it is observed that the
additional improvement is highly dependent upon the loading stage, and that at high loads the
additional improvement gained drops once stone columns yield and plastic deformation
becomes dominant. Therefore, the installation effects do not influence the ultimate behaviour
of stone columns but they do play a positive role in enhancing the settlement behaviour.
2.4.4 Summary of column installation effects

Stone columns are formed with the aid of a vibrating poker which displaces the soil and
imparts horizontal vibrations. The forces imparted from the horizontal vibrations are
attenuated in fine grained soils and any change to in situ stress state may be attributed to the
displacing effects of the poker. It was shown in a series of small scale laboratory experiments
that the soil displacement due to the installation of a driven pile can be predicted using
cylindrical cavity expansion theory and authors have used this approach in two-dimensional
axisymmetric FEA to model stone column installation.
Field experience indicates that horizontal stress and soil stiffness increase within a zone
which ranges from 4d–8d from the centre-line of columns. However, increases in horizontal
stress and soil stiffness are offset by the effects of remoulding and dynamic excitation within
4d. It was also found that an increase in excess pore water pressure due to initial poker
penetration agrees favourably with the analytical formula developed by Randolph et al.
(1979b). Column installation effects tend to improve settlement behaviour of stone columns.
However, the additional improvement drops once stone columns yield and plastic
deformation becomes dominant. Therefore, the installation effects do not influence the
ultimate behaviour of stone columns.
A number of methods to simulate column installation effects are summarised below:

1. Increase the coefficient of lateral earth pressure (K0) in soil
- Increases in K0 ranges from 0.75 to 1.50 (K0, average = 1.0)
2. Apply cylindrical cavity expansion to stone column
- Expand a dummy material (with a nominal stiffness)
- Convert properties to stone backfill after expansion
3. Apply cylindrical cavity expansion to stone column and increase in soil stiffness
- Apply cavity expansion to stone columns
- Increase the stiffness in an enhancement zone (2d–5d from centreline of outer columns)
2.5 Design methods

2.5.1 Unit cell concept
Large arrays of columns are typically spaced on square or triangular grids to support wide
loaded areas. Figure 2.21 illustrates that the shape of the zone of influence depends upon the
column arrangement adopted. It follows from symmetry that the behaviour of each stone
column and its surrounding zone of influence within a large array will be identical. It also
follows from symmetry that the sides of the zones of influence are free of radial
displacements and shear stresses. Therefore the analysis may be simplified to one column and
its surrounding zone of influence. The unit cell concept approximates the zone of influence
by a circle of equivalent area. This concept applies to interior columns beneath wide loaded
areas, the proportion of which increases with group size (e.g. embankments, large floor
slabs). The unit cell concept is not valid for edge columns beneath wide loaded areas or small
groups of columns beneath pad/strip footings due to a loss of lateral confinement.
Figure 2.21 - Zones of influence for square and triangular arrangements of stone columns
2.5.2 Cylindrical Cavity Expansion Theory (CCET)

If an internal pressure is applied to a cavity wall, the wall will expand as shown in Figure
2.22. Suppose a general cylindrical shell at radial distance r expands by an amount y and also
that another cylindrical shell, at a radial distance r + dr expands by an amount y + dy.
The radial (∆εr) and tangential (∆εθ) strain are defined as:

Δ = −
=−

(2.1)

Δ = − =−

(2.2)
(a) (b) (c)

Figure 2.22 - Equilibrium of soil element (Powrie, 2004)
Assuming the soil can be idealised as a uniform, isotropic and linear elastic material which
obeys Hooke’s laws, then the principal total stress and strain increments in the radial (r),
tangential (θ) and vertical (z) directions are related as follows:
Δ = Δ + Δ + Δ" + 2$Δ (2.3)
Δ = Δ + Δ + Δ" + 2$Δ (2.4)
Δ" = Δ + Δ + Δ" + 2$Δ" (2.5)
%& &
where = '%'% ; $ =
'%
Considering equilibrium of the element in Figure 2.20(b) and 2.20(c), the equation for radial
equilibrium is as follows:
) + * + **+, = ) *+, + )2 *-.*+ ⁄2, (2.6)
*+ - 01 ⟹ -.*+⁄2 ≈ *+ ⁄2
) + * + **+, = ) *+, + )2 **+⁄2, (2.7)
45.-.5 01 671.8-8-9 9. 5. * × *
* + * = * (2.8)
:; <:; :=
+ = (2.9)
<
>-0-8 1 * 1.? * → 0:

:; := :;

=
(2.10)
Assuming plane strain conditions (i.e. ∆εz = 0) and combining the stress-strain equations (2.3,
2.4 and 2.5) and equation of equilibrium (2.10):
C:; C:= C:; '

=
= ) Δ + Δ + Δ" + 2$Δ , = )2$Δ − 2$Δ , (2.11)
'
) Δ + Δ + 2$Δ , = )2$Δ − 2$Δ ,

(2.12)
'
) + 2$Δ + Δ , = )2$Δ − 2$Δ ,

(2.13)
'
D + 2$ E− F+ Δ E− FG = D2$ E− F − 2$ E− FG

(2.14)
H ' '

− + 2$ H −
+ H
= −2$ H + 2$ (2.15)
H '
− + 2$ H − + 2$ + + 2$ H = 0 (2.16)
H '
H
+ − H = 0 (2.17)
Powrie (2004) states that equation 2.17 is a differential equation in terms of a single variable
y (y = radial movement). Its general solution, which may be verified by substitution, is:
J
I = + (2.18)
2.5.3 Ultimate bearing capacity

Thorburn & MacVicar (1968)
Thorburn & MacVicar (1968) present empirical and semi-empirical design curves to
calculate allowable working loads on stone columns formed in granular and cohesive soils,
respectively. The design curves were developed as the authors sought an alternative design
solution to piled foundations for low-rise developments in Glasgow, where soil types range
from fresh water alluvium to beach deposits and glacial till. A load test on a strip of three
stone columns indicated that this technique was a preferred design alternative to piles as it
reduced overall and differential settlements. The authors developed the design chart for
cohesive soils from load tests carried out at various sites and undrained axial compression
tests (Figure 2.23). The authors state that stone columns cannot be satisfactorily installed in
soils with a cu < 400 lb/ft2 (19.2 kPa), as these soils have very low passive resistances and
present difficulties in forming the stone columns.
Figure 2.23 - Relationship between allowable vertical stress on stone column and undrained shear strength
(Thorburn & MacVicar, 1968)
Hughes & Withers (1974)

As outlined in Section 2.2.1, Hughes & Withers (1974) conducted a series of laboratory tests
on single stone columns. Significant column deformation occurred in the upper sections of
columns, while negligible strain was observed below 4d (Figure 2.6). Column deformation is
idealised as uniform bulging and CCET developed by Gibson & Anderson (1961) can be
used to predict the limiting radial stress. However, on the basis of field records of quick
expansion tests the authors approximate the limiting radial stress (σrL) by:
L = M + 4 + 7 (2.19)
where σr0 = total in situ lateral stress

c = undrained cohesion
u = pore water pressure
Assuming columns are in a critical state, the ultimate vertical stress (σv') is related to the
limiting radial stress by:
'QRSTU
O ′ = E'QRSTUF ′ (2.20)
where σr' = lateral effective stress

φ' = angle of internal friction of column material
Therefore, the ultimate vertical stress for column bulging is:

'QRSTU
O ′ = E'QRSTUF M + 4 + 7 (2.21)
If the vertical forces exceed the shear resistance along the sides of the column and the
ultimate bearing capacity at the base, then column punching will occur.
2.5.4 Magnitude of settlement

Greenwood (1970)
Greenwood (1970) presents design curves for estimating consolidation settlement of soft clay
strengthened by stone columns under wide spread loadings (Figure 2.24). The curves are
developed empirically for the wet and dry methods of construction. It appears that columns
installed using the wet method provide a better settlement performance. However, this may
be attributed to the larger column diameters which result from this construction technique.
The curves neglect immediate settlement and shear displacements. In addition, columns are
assumed to be resting on firm clay, sand or harder ground.
Figure 2.24 - Settlement diagram for stone columns in uniform soft clay (Greenwood, 1970)
Aboshi et al. (1979)

Aboshi et al. (1979) propose a simple design method to determine the settlement of large
diameter sand columns. The distribution of vertical stress within sand columns (σS) and the
surrounding clay (σC) is determined by the stress concentration ratio (n) and the area ratio
(aS), which are defined as σS/σC and AC/(AS+AC), respectively. The following expressions are
obtained from the equilibrium of stresses:
= V 1V + 1 − 1V (2.22)
= /)1 + . − 11V , = X (2.23)
V = ./)1 + . − 11V , = XV (2.24)
The stress concentration ratio is estimated based on field experience. Aboshi et al. (1979)
present stress concentration ratios measured at several construction sites which range from
1.6–11.5. However, Barksdale & Bachus (1983) suggest that stress concentration ratios
typically range from 2.5–5.0. The proposed design method is based on an infinite grid of
columns and the settlement of an untreated S and treated S' soil deposit is:
Y = 0Z [\ (2.25)
Y′ = 0Z X [\ (2.26)
where mv = modulus of volume compressibility

P = applied pressure
H = thickness of soil deposit
Therefore the settlement reduction factor is:
] = X = 1/)1 + . − 11V , (2.27)
Priebe (1995)
Priebe (1995) developed a semi-empirical design method to estimate the settlement of an
infinite grid of end-bearing stone columns. The columns are assumed to be in an active state
and bulge uniformly along their length. The surrounding soil is idealised as an isotropic
elastic material, with an increased coefficient of lateral earth pressure (K0 = 1) to account for
the effects of column installation. The relationship between the applied vertical stress and the
expansion of the column is determined using CCET, the fundamental equations of which are
outlined in 2.5.2. The columns are assumed to be incompressible and a basic settlement
improvement factor (n0), which is defined as the ratio of settlements for an untreated footing
to a treated footing, is developed:
1⁄2 + X_ , ⁄
.M = 1 + ^ − 1b
aJ, X_ , ⁄
(2.28)
'c H 'c'Jd⁄J

where X_ , ⁄ = 'ccH 'cJd ⁄J
KA,c = coefficient of active earth pressure of column = tan2(45 – φc/2)

φc = angle of internal friction of column
µ = Poisson’s ratio of soil
It is clear from Figure 2.25 that the settlement improvement factors are significantly
influenced by both column spacing (A/AC) and the strength of the stone backfill (φc).
Figure 2.25 - Design curves for basic settlement improvement factor n0 with Poisson’s ratio υ = ⅓ (Priebe,
1995)
Priebe (1995) modifies the basic settlement improvement factor (n0) to account for column
compressibility and the effect of overburden stress. The assumption of column
incompressibility yields infinite settlement improvement factors for an area ratio A/AC = 1.
This is clearly unrealistic as settlement improvement factors should at best be equal to the
ratio of the compression moduli of the column and soil. The effect of column compressibility
is accounted for by increasing the area ratio A/AC by an amount ∆A/AC, which is dependent
on the ratio of the compression moduli of the column and soil. The modification for column
compressibility results in lower settlement improvement factors (n1).
The basic settlement improvement factor also neglects the effect of overburden stress. The
pressure difference at the column-soil interface is assumed to be constant with depth and does
not account for the unit weight of the column and the surrounding soil. Overburden stress
increases with depth and reduces the pressure difference at the column-soil interface.
Therefore, consideration of the overburden stress reduces column bulging and yields higher
settlement improvement factors. A depth factor (fd) is introduced to account for the effect of
overburden stress. The depth factor is defined as the ratio of the original pressure difference
to the ‘new’ pressure difference (which incorporates overburden stress). The settlement of
stone columns is directly related to the depth factor and modified settlement improvement
factors are calculated as n2 = fd×n1.
Priebe (1995) develops design charts relating the settlement of a group of columns beneath
pad footings (s) to the settlement of an infinite grid of columns (s∞). The design curves
account for the stress distribution beneath pad footings and assume a reduced bearing
capacity for the outer columns. It appears from Figure 2.26 that the settlement ratio (s/s∞) is
independent of footing area (A). However Priebe states that the footing area must be
calculated from the area ratio A/AC, which compensates for footing size as larger footing
areas have higher A/AC ratios and hence yield lower settlement improvement factors. Priebe
also states that this compensation provides acceptable results for area ratios A/AC up to 10.
Figure 2.26 - Settlement of single footings on groups of columns (Priebe, 1995)
The curves in Figure 2.26 indicate that the settlement ratio (s/s∞) reduces rapidly with depth,
especially for smaller groups of columns. This results from a decay of vertical stress with
depth beneath pad footings and hence the influence of the depth factor (fd) is reduced for pad
footings. Therefore, Priebe suggests dividing the subsoil into layers and calculating the
settlement of each layer individually to avoid over-estimating settlements of pad footings.
The settlement is calculated using the following formula:
f
∆ = )⁄h L ?L − ⁄h i ?i ,
gV .
(2.29)
where dL and dU are the lower and upper bound depths of the layer considered.
Balaam & Booker (1981)

Balaam & Booker (1981) propose an analytical solution to determine the settlement of an
infinite array of end-bearing stone columns. The stone backfill and the surrounding soil are
both idealised as linear elastic materials, which are defined by Young’s modulus (E) and
Poisson’s ratio (ν); the selection of appropriate values must take into consideration the stress
level encountered beneath the foundation. The analysis is essentially the compression of a
cylindrical body between smooth (raft) and rough (substratum) plates, with a laterally
restrained smooth wall. An axisymmetric FEA undertaken by the authors indicates that a
triaxial state of stress exists in the column and also that field quantities remote from the
substratum are insensitive to a smooth or rough boundary condition being assumed at the
substratum. Therefore, it is possible to assume a smooth substratum and hence to find an
exact analytical solution. This solution is based on CCET, the fundamental equations of
which are outlined in 2.5.2.
As a first approximation (solution A), the column is assumed to be laterally restrained. The
authors report that this assumption yields reasonable results, however a stress discontinuity
(∆σr) occurs at the column-soil interface. As the column is much stiffer than the surrounding
soil, it will attract more load and hence will develop a higher radial stress than the
surrounding soil. This stress discontinuity displays itself as column bulging in reality and it is
necessary to account for this. Therefore a second solution (B) is developed for zero vertical
movement of the raft and a laterally expanding column, which imparts a radial stress equal
and opposite to ∆σr at the column-soil interface (Figure 2.27).
Figure 2.27 - Boundary conditions for solutions A & B proposed by Balaam & Booker (1981)
The final solution is found by super-imposing solutions A & B (Table 2.2). The relationship
between strain and the average applied stress qA is determined by integrating the vertical
stresses across the soil surface.
Table 2.2 - Final solution for stresses and strains in column and soil (Balaam & Booker, 1981)
Region 1 Region 2
Stone column Clay
"
zH {| H H }
I y Dy G
|H zH
zH ~ |H
) ' − 2 ' + $' y, D + |H zH E + $ + $ H FG
zH ~ |H
) ' − 2 ' + $' y, D + |H zH E + $ − $ H FG
~zH
" ) ' + 2$' − 2 ' y, D + 2$ + 2 |H zH G
& &
Lame’s parameters: = '%'% ; $ =
'%
H {|H zH }

and where y = )zH
H H |
H
H ,
a = radius of stone column

b = radius of unit cell
Relationship between strain and average applied stress, qA:
6J = ) ' + 2$' 1 + + 2$ − 1 − 21 ' − y, (2.30)
6J = ) ' + 2$' ⁄ + + 2$ 1 − ⁄ − 2 ⁄ ' − y, (2.31)
= 6J /) ' + 2$' ⁄ + + 2$ 1 − ⁄ − 2 ⁄ ' − y, (2.32)
The settlement reduction factor (β) is defined as:
] = /6J 0Z (2.33)
] = /) ' + 2$' ⁄ + + 2$ 1 − ⁄ − 2 ⁄ ' − y, (2.34)
Balaam & Booker (1985)

Several authors have found that under certain circumstances an elastic analysis can grossly
overestimate the effectiveness of stone columns in reducing the settlement of a foundation.
Balaam & Booker (1985) propose an interaction analysis which contains some simplifying
assumptions to account for column yielding. The analytical solution for the elastic response
developed by Balaam & Booker (1981) indicates that the major principal stresses will be
close to vertical. Furthermore, there may also be significant yielding of the column, but little
yield in the surrounding clay. As a result, Balaam & Booker (1985) make the following
assumptions:
(i) stone columns are in a triaxial stress state
(ii) yielding may occur in columns and no yielding occurs in the surrounding soil
(iii) no shear stress develops along the stone-soil interface
(iv) the behaviour of stone columns is idealised as an elasto-plastic material satisfying
the Mohr-Coulomb yield criterion
The proposed solution which incorporates these assumptions was validated by a comparison
with a FEA. The clay and the column material were both treated as dilatant materials, which
satisfy the Mohr-Coulomb yield criterion and have a non-associated flow rule. Appropriate
material parameters and geometrical dimensions were selected to comprehensively test the
validity of the assumptions. A good agreement exists between both methods and, therefore,
the solution proposed by Balaam & Booker (1985) is an efficient and accurate way to
calculate the reduction in settlement due to stone columns.
Pulko & Majes (2005)

Pulko & Majes (2005) extend the elastic analysis proposed by Balaam & Booker (1981) to
account for confined column yielding. Therefore the soil is assumed to remain in an elastic
state and the column material is assumed to behave as a perfectly elastic-plastic material
satisfying the Mohr-Coulomb yield criterion. The design method accounts for both applied
stress levels and the in situ overburden stress. Column yielding, which is accounted for by
dilatancy theory developed by Rowe (1962), occurs once active conditions are reached in the
column. The elastic strains, and if applicable plastic strains, are integrated along the length of
the column to determine settlement improvement factors. The settlement results of the design
method are in excellent agreement with a FEA for non-dilatant soil and stone materials.
2.5.5 Rate of settlement

Han & Ye (2001)
It is well established from field studies that stone columns increase the rate of consolidation.
The high permeability of stone backfill ensures that columns act as vertical drains. The low
compressibility of stone columns further increases the rate of consolidation as stress
concentrations develop in columns which reduces the vertical stress on the soil. Han & Ye
(2001) develop a simplified method to compute the rate of consolidation of stone column
reinforced foundations. The formats of vertical and radial flows are similar to those of the
Terzaghi 1D solution and the Barron solution for drain wells in fine grained soils,
respectively. Modified coefficients of consolidation in the radial (cr') and vertical (cv')
directions are introduced to account the stone column-soil modular ratio:
' '
U = E1 + .Q H 'F OU = O E1 + .Q H 'F
where cr, cv = coefficient of consolidation in radial and vertical direction, respectively

ns = steady state stress concentration ratio
N = diameter ratio (= de/dc)
The effectiveness of stone columns in increasing the rate of consolidation is shown Figure
2.28, where the proportion of excess pore pressure dissipation due to the stiffness of the stone
column is shown. It can also be seen that vertical stress transfers from the soil to the stone
columns as consolidation proceeds. The proposed solution was compared with a numerical
study by Balaam & Booker (1981) and a reasonable agreement was observed. However, the
proposed solution tends to under-estimate the rate of consolidation initially, while a reverse in
this trend is observed for rates of consolidation greater than 40 %. The may be explained as
the proposed solution does not account for lateral expansion of columns. Initially, the pore
water pressure in the soil takes the entire load which forces the column to compress. This
mechanism acts as a relief for the soil which increases the rate of consolidation. However, as
consolidation occurs, stress concentration increases in columns which cause columns to
expand into surrounding soil. This increases the excess pore water pressure in the soil and
reduces the rate of consolidation.
Figure 2.28 - Dissipation of excess pore water pressure (Han & Ye, 2001)
Castro & Sagaseta (2009)

Castro & Sagaseta (2009) develop equivalent coefficients of consolidation which account for
the lateral movement of the stone column during the consolidation process. The behaviour of
the column is modelled as an elastic material or as an elasto-plastic dilatant material with a
Mohr-Coulomb failure criterion. Vertical stress increases in columns with consolidation and
column yielding may occur. The stiffness of columns reduces once yielding occurs, which
results in increased radial deformability of the column. Therefore columns in a plastic state
expand and increase the excess pore water pressure in the soil. The solution developed by
Castro & Sagaseta (2009) allows the depth and time of column yielding to be determined.
Therefore it is possible to accurately determine the stresses and strains occurring in columns
at any stage in the loading history. A comparison with Han & Ye (2001) of the development
of stress concentration ratio with time is shown in Figure 2.29.
Figure 2.29 - Stress concentration factor. Influence of radial deformation and plastic strains (Castro & Sagaseta,
2009)
2.5.6 Summary of design methods

The majority of design methods for the ultimate bearing capacity and settlement performance
of stone columns are based on the unit cell concept and Cylindrical Cavity Expansion Theory
(CCET). The unit cell concept assumes that an infinite grid of columns support an infinitely
wide load area. It follows from symmetry that the behaviour of each column within an
infinite grid is identical and the analysis may be simplified to one column and its surrounding
zone of influence. Consequently, the unit cell concept is only applicable to interior columns
within large groups. The stresses and strains which develop in the soil surrounding an
expanding cylindrical shell may be determined using CCET.
Empirical and analytical design methods to determine the ultimate bearing capacity of
columns were presented in the previous section. These design methods indicate that the
ultimate bearing capacity of stone columns is highly dependent upon the passive resistance of
the surrounding soil, especially in the upper sections of columns (i.e. near the ground
surface).
A series of design methods, ranging from empirical to analytical, are presented in the
previous section to determine the magnitude of settlement for stone columns. Priebe (1995)
develops a semi-empirical design method which contains some simplifying assumptions.
Columns are assumed to be in an active state and uniform bulging occurs along the length of
columns. Modification factors are introduced to account for column compressibility and
overburden stresses. Balaam & Booker (1981) propose a more rigorous theoretical solution
than Priebe (1995), but model the stone columns and surrounding soil as linear elastic
materials and, therefore, fail to account for column yielding. This can significantly over-
estimate the effectiveness of stone columns at reducing the settlement of foundation. Balaam
& Booker (1985) and Pulko & Majes (2005) extend this solution to account for column
yielding through an interaction analysis and analytical design method, respectively. The
behaviour of stone columns is idealised as elastic-rigid plastic and column yielding is
captured using dilation theory developed by Rowe (1962).
Stone columns are an effective method to increase the rate of consolidation as their high
permeability allows them to act as vertical drains and also as stress concentrations, which
develop in stone columns, reduce the vertical stress on the surrounding soil. Han & Ye (2001)
develop a simplified solution to compute the rate of consolidation for stone column
reinforced foundations. This solution is extended by Castro & Sagaseta (2009) who model the
stone column as an elasto-plastic dilatant material and account for lateral expansion. This
design method can be used to determine the depth and time of yielding, which allows for an
accurate determination of stresses and strains occurring in columns at any stage in the loading
history.
2.6 Summary of literature review

Vibro stone columns are a popular form of ground improvement, which enhance the
settlement performance and bearing capacity of treated soils. Vertical columns of compact
stone are formed in the ground using the top or bottom feed systems. Stone columns can be
used to support a wide variety of loading scenarios ranging from small footings to wide area
loadings.
The deformational behaviour of stone columns is well understood and it is established that
single columns or large groups of columns exhibit either a punching or bulging mode of
deformation. However, the behaviour of small groups of stone columns is more complex and
Muir Wood et al. (2000) demonstrate that small groups of columns can fail by punching,
bulging, bending and shearing. This may be attributed to the loss of lateral confinement for
peripheral columns and also the reduction in vertical stress with depth beneath small loaded
areas.
Stone columns can be adopted to treat soft clayey soils, which are typically characterised by
poor strength, stiffness and drainage properties. Consequently, foundations on these soils
undergo large displacements at relatively low loads, and settlement is usually the governing
criterion in foundation design. While the deformational behaviour of small groups of columns
is well understood, there exists a dearth of information regarding the settlement performance
of small groups of columns. This is highlighted by McCabe et al. (2009), who collated a
settlement database from field records, and observed that only three out of 20 case studies
relate to small groups of stone columns. Analytical design methods contain many simplifying
assumptions, such as the unit cell concept which is only applicable to large groups of stone
columns. Elshazly et al. (2008) conduct an axisymmetric FEA to examine the settlement
performance of finite groups of columns; however, the majority of numerical studies are
based on large groups of columns supporting wide area loadings. The lack of information
regarding the settlement performance of small groups of stone columns was identified by
Black (2006), who conducted a series of high quality laboratory tests. However, it is difficult
to extrapolate the findings due to scale effects and also as some of the area ratios considered
are at the high end of typical values used in practice.
It was shown in the laboratory tests that column interaction plays an important role in the
deformational behaviour of small groups of stone columns. Therefore, it is necessary to
model columns at discrete locations, rather than approximating their presence by cylindrical
rings, as is the case with axisymmetric analyses. Furthermore, the decay of vertical stress
with depth will have an important role on the settlement performance of stone columns, and
must be captured when analysing the settlement performance of small groups of stone
columns. It is proposed in this thesis to conduct a three-dimensional FEA using PLAXIS 3D
Foundation to examine the influence of various design parameters, such as area ratio, column
length and the number of columns, upon the settlement performance and deformational
behaviour of small groups of stone columns.
51
Chapter 3
Background and preliminary checks for
Finite Element Analysis
3.1 Introduction
The Finite Element Method (FEM) is a powerful tool used for analysing complex engineering
problems. The FEM is an approximate technique which uses the principle of virtual work to
estimate the distribution of stresses and strains throughout a continuum. PLAXIS 3D
Foundation is a three-dimensional FE program which is specifically tailored for geotechnical
applications. This program is ideal for capturing the complex behaviour of small groups of
stone columns and was adopted for the subsequent FEA. The behaviour of the soil and stone
is simulated with advanced constitutive models, which are described in the following chapter.
In addition, as the FEM is an approximate technique it is necessary to carry out a number of
preliminary checks, such as mesh sensitivity and distance to the boundary, to ensure accurate
numerical analyses.
3.2 Description of material models

The behaviour of real soil is highly non-linear, with both strength and stiffness depending on
the stress and strain level (Potts & Zdravkovic, 1999). Furthermore, real soil often exhibits
time-dependent behaviour and anisotropic tendencies. The behaviour of soil may be
approximated to varying degrees of accuracy using material models. An overview of the
material models used in the subsequent FEA is given below:
Linear Elastic model

This is based on Hooke’s law of elasticity and idealises soil as a linear elastic material, thus
precluding the development of irreversible strains. The material behaviour is defined by two
parameters, Young’s modulus (E) and Poisson’s ratio (ν). This model is too crude to
accurately capture the complicated stress-strain behaviour of soil and is only adopted to
represent structural elements e.g. concrete, steel.
Background and preliminary checks for FEA 52
Mohr-Coulomb model
This is a first order model which idealises soil as an elastic-perfectly plastic material. The
behaviour of soil before failure is approximated by Hooke’s law of elasticity. The failure of
soil is based upon the Mohr-Coulomb failure criterion which is defined by two parameters,
angle of internal friction (φ) and cohesion (c). This failure criterion is an extension of
Coulomb’s friction theory and its yield surfaces in principal stress space are shown in Figure
3.1. The model does not generate irreversible strains below the yield surfaces (Mar, 2002).
However, irreversible plastic strains resulting from shearing are captured using a non-
associated flow rule, which is defined by an angle of dilation (ψ).
Figure 3.1 - Mohr-Coulomb yield surface in principal stress space (c = 0 kPa) (Brinkgreve & Broere, 2006)
A constant Young’s modulus defines the behaviour of soil before failure and it is important to
choose an appropriate value which reflects the stress path and stress level experienced by the
soil. A drawback to the Mohr-Coulomb model is that it fails to accurately capture the
stiffness response of soils due to the simplistic assumption of linear elasticity before failure
and is only used as a first approximation of soil behaviour (Brinkgreve & Broere, 2006).
While the Mohr-Coulomb model captures the failure behaviour of soil quite well in drained
conditions, it over-estimates the undrained shear strength, as continuous dilation is assumed
once yielding occurs (Potts & Zdravkovic, 2001).
Hardening Soil model

The Hardening Soil model is an advanced elasto-plastic constitutive model which can be used
to simulate the behaviour of both soft and stiff soils (Schanz, 1998). The model is an
extension of the hyperbolic model developed by Duncan & Chang (1970). It supersedes the
hyperbolic model as it is based on the theory of plasticity rather than elasticity, includes soil
dilatancy and introduces a yield cap (Schanz et al., 1999). It accounts for both shear and
volumetric hardening, thus capturing irreversible strains caused by deviatoric and
compression loadings, respectively.
The basis for the Hardening Soil model is that the relationship between the deviatoric stress
(q) and the vertical strain (ε1) for a primary triaxial loading may be approximated by a
hyperbola. Therefore, the yield curve for a standard drained triaxial test, which is shown in
Figure 3.2, can be defined by:
6z 6
' = 6 < 6
2M 6z − 6
(3.1)
where qa = asymptotic value of shear strength

qf = ultimate deviatoric stress
E50 = secant Young’s modulus at 50% deviatoric stress
The secant Young’s modulus is used instead of the initial modulus (Ei) as this is easier to
define from triaxial tests. The Hardening Soil model also captures the stress path dependency
of soil stiffness using an unload-reload modulus (Eur). The ultimate deviatoric stress (qf) is
derived from the Mohr-Coulomb failure criterion and related to the asymptotic value of shear
strength (qa) by a failure ratio (Rf):
6-. 6z
6 = f + . 8 =
3 − -. 6 (3.2)
where φ = angle of internal friction

c = cohesion
p = mean total stress
Figure 3.2 - Hyperbolic stress-strain relation in primary loading for a standard drained triaxial test (Schanz et al.,
1999)
Soils tend to exhibit a stress level dependency which results in an increasing stiffness with
increasing confining pressure. The stress dependency of soil stiffness is captured by the
Hardening Soil model using a power law:

=

(3.3)
where Eref is the reference stiffness corresponding to the reference confining pressure (σref).
The confining pressure is taken as the minor or major principal stress for triaxial or
oedometer tests, respectively. The stress dependency is determined by the parameter, m,
which ranges from 0.5 for Norwegian sands and silts (Janbu, 1963) to 1.0 for soft clays
(Brinkgreve & Broere, 2006).
Once the ultimate deviatoric stress is reached, the Mohr-Coulomb failure criterion is satisfied
and perfectly plastic yielding occurs. In addition to capturing plastic shear strains, which
result from decreasing stiffness, the Hardening Soil model also accounts for the volumetric
strains due to dilatancy. Plastic volumetric strains are determined from the stress-dilatancy
theory developed by Rowe (1962). A cap yield surface is introduced to account for
volumetric strains resulting from isotropic loadings. The yield surfaces for the Hardening Soil
model in principal stress space are shown in Figure 3.3.
Figure 3.3 - Hardening Soil yield surface in principal stress space (c = 0 kPa) (Brinkgreve & Broere, 2006)
The Hardening Soil model does not include anisotropic strength and stiffness or time
dependent behaviour of soil. Also, it does not account for the strain dependent nature of soil
stiffness and therefore its application to dynamic loadings is limited.
3.3 Preliminary checks to ensure accurate numerical analyses

3.3.1 Mesh sensitivity analyses
The accuracy of the FEM depends not only on the level of sophistication of the material
model, but also on the number and type of the elements into which the domain is discretised.
PLAXIS 3D Foundation uses 15 node wedge elements which are composed of 6 node
triangles in the horizontal direction and 8 node quadrilaterals in the vertical direction (Figure
3.4).
(i) (ii)
Figure 3.4 - Distribution of (i) nodes and (ii) stress points within 15 node wedge elements
The influence of the number of elements upon the accuracy of the FEM was investigated by
conducting a mesh sensitivity analysis. In the subsequent parametric studies, both the number
of columns and column spacing are varied beneath pad footings which results in various
footing sizes. Mesh sensitivity analyses are conducted for six different footing sizes where
the accuracy of medium and fine meshes are compared against very fine (vf) meshes.
The vertical displacement (uy) and mean effective stress (p') were measured at three points A,
B and C which are 0, 1 and 2 m below the centre of footings, respectively (Figure 3.5). When
conducting mesh sensitivity analyses it is prudent to examine stresses in the zone of interest,
as the distribution of stress within elements is derived from lower order equations than the
displacement. Therefore, stress converges slower than displacement with increasing mesh
density and, consequently, the distribution of stress within an element will not be as accurate
as the displacement. Vertical displacements were also examined for the mesh sensitivity
analysis as the settlement performance of stone columns is the primary focus of this thesis.
The soil profile adopted for the subsequent FEA and the mesh sensitivity analysis is that of
the well characterised Bothkennar test site, which is described in detail in Section 4.4.
Figure 3.5 - Points A, B and C beneath pad footing
The accuracy of medium and fine meshes is determined by comparing the normalised error
for vertical displacement (uy) and mean effective stress (p') against very fine meshes:
,
Normalised error for vertical displacement, uy = ,
× 100

Normalised error for mean effective stress, p' =
× 100
In addition to checking convergence of stress and strain with an increasing number of

elements, the mesh is also checked for discontinuities at inter-element boundaries.
Discontinuities typically occur in regions of rapid changes of stress and strain, i.e. beneath
the edge of footings, and can be overcome by refining the mesh in these regions. An example
of continuous displacement contours at inter-element boundaries is shown in Figure 3.6.
10.5 m
14.5 m
Figure 3.6 - Continuous displacement contours along inter-element boundaries beneath a 3 m square footing
In the subsequent FEA, the settlement performance and deformational behaviour of various
configurations of columns is examined at 50 kPa, which is a typical working load for the
Bothkennar test site. However, most of the mesh sensitivity analyses were conducted at 25
kPa in order to save computational time.
A flowchart outlining the steps taken to ensure accurate numerical analyses is shown in
Figure 3.7.
Figure 3.7 - Flowchart outlining preliminary analysis checks
The number of elements used for each mesh and the normalised error for the vertical
displacement and mean effective stress is shown in Table 3.1. It is clear that the values for
medium and fine meshes are converging towards the very fine mesh and also that the
normalised error between the fine and very fine meshes for the vertical displacement and
mean effective stress is quite low (maximum error is no greater than 2%).
3.3.2 Influence of distance to boundary

The pad footings modelled in the subsequent parametric studies are surrounded by a zone of
soil, which undergoes no lateral displacement along its outer boundary. Therefore, it is
necessary to position the boundary at a sufficient distance from the footing so that boundary
conditions do not influence the results. A sensitivity analysis was conducted on a 3 m square
footing, where the distance to the boundary from the footing centre ranged from 2B–12B.
The footings were loaded to 25 kPa and the vertical displacement and mean effective stress at
points A, B and C for the different boundary distances are listed in Table 3.2. It appears that
Background and preliminary checks for Finite Element Analysis 58
Table 3.1 - Vertical displacement and mean effective stress at points A, B and C
Ftg width, B No. of elements No. of elements Vert displ, uy (mm) Normalised Error (%) Mean eff stress, p' (kPa) Normalised Error (%)
Mesh
(m) in top work plane in 3D mesh A B C A B C A B C A B C
Medium 302 7852 11.6 10.8 9.2 0.9 2.4 8.1 21.5 24.6 25.5 3.1 0.3 0.9
Fine 302 12684 11.7 10.5 8.5 0.5 0.4 0.3 20.8 24.5 25.3 0.4 0.1 0.0
2
Very fine 700 29400 11.7 10.6 8.5 20.9 24.5 25.3
Medium 356 11036 48.3 45.4 38.4 1.5 1.5 0.9 26.5 29.7 29.2 2.2 0.5 0.1
3 Fine 356 16376 48.7 45.7 38.6 0.8 0.8 0.5 25.8 29.7 29.2 0.5 0.3 0.1
Very fine 678 31188 49.1 46.1 38.7 25.9 29.8 29.3
Medium 536 16616 59.4 55.7 43.5 1.1 1.1 0.9 25.7 30.4 24.4 0.4 0.4 0.0
4 Fine 536 23048 59.9 56.2 43.8 0.3 0.3 0.3 25.8 30.5 24.4 0.1 0.1 0.0
Very fine 822 35346 60.1 56.3 43.9 25.9 30.5 24.4
Medium 480 20640 85.4 82.1 71.3 1.8 1.8 1.6 25.9 30.7 32.6 0.8 0.6 0.5
4.5 Fine 1124 48332 86.8 83.5 72.4 0.2 0.2 0.2 26.1 30.9 32.8 0.1 0.1 0.1
Very fine 1370 58910 87.0 83.6 72.5 26.1 30.9 32.8
Medium 392 10976 122.8 118.8 105.6 2.7 2.6 2.5 273.2 32.3 36.2 25.0 1.0 2.0
6 Fine 738 31734 125.5 121.3 107.8 0.6 0.6 0.5 370.9 32.6 35.4 1.7 0.2 0.2
Very fine 1476 63468 126.2 122.0 108.4 364.5 32.6 35.5
Medium 738 20664 170.2 165.1 149.2 1.3 1.2 1.2 30.7 34.5 37.6 6.0 0.4 0.4
8 Fine 738 31734 171.5 166.2 150.2 0.5 0.5 0.5 28.8 34.6 37.7 0.2 0.2 0.2
Very fine 1532 58216 172.4 167.1 151.0 28.9 34.6 37.8
Table 3.2 - Influence of the distance to boundary upon the settlement of a 3 m square footing
Distance to boundary Vertical displacement, uy (mm) Normalised Error (%) Mean effective stress, p' (kPa) Normalised Error (%)
(B = footing width) Point A Point B Point C Point A Point B Point C Point A Point B Point C Point A Point B Point C
2B 61.9 58.4 50.2 26.2 26.8 29.3 26.8 30.6 29.9 0.5 2.8 2.0
4B 53.2 50.0 42.1 8.4 8.5 8.5 26.2 30.1 29.5 1.7 1.0 0.6
6B 49.4 46.3 38.9 0.6 0.6 0.3 25.9 29.8 29.3 2.9 0.1 0.1
8B 49.1 46.1 38.7 0.0 0.0 0.2 25.9 29.8 29.3 2.9 0.1 0.2
10B 48.9 45.9 38.7 0.3 0.3 0.2 26.6 29.8 29.3 0.1 0.1 0.1
12B 49.1 46.1 38.8 0.0 0.0 0.0 26.7 29.8 29.3 0.0 0.0 0.0
positioning the boundary closer than 6B to the centre of the footing induces more settlement
in the pad footing. The distance to the boundary from the centre of the footing is
conservatively chosen at 8B for all the subsequent FEA.
3.4 Other modelling issues

3.4.1 Modelling of soft soil behaviour
The long term behaviour of cohesive soils, which develop excess pore pressure during
loading, can be modelled by two methods with PLAXIS 3D Foundation:
(i) Undrained loading followed by consolidation analysis
(ii) Drained analysis
Method (i) is a closer simulation of reality than method (ii), however it is far more time
consuming. In method (i), the soil is specified to behave in an undrained manner during the
application of footing loads. Initial undrained settlements are computed and long term
settlements are then determined by conducting a consolidation analysis. The soil can be
defined using either total or effective strength parameters for this method. However, it is not
possible to capture the increase in soil shear strength with consolidation or the stress
dependency of soil stiffness when using total strength parameters and, consequently, effective
strength parameters are used in all the subsequent FEA. The initial undrained response of the
soil is simulated by adding a very large bulk modulus to the pore water stiffness matrix,
which ensures that all of the stresses generated from the footing loading will be taken by the
pore water. The consolidation analysis is then conducted by removing the extra bulk
modulus. The stresses in the pore water pressure are then transferred to the soil matrix and
the resulting settlements for primary consolidation are computed. The behaviour of the soil
for method (ii) is again defined using effective strength and stiffness parameters. However, it
is not possible to separate the settlements from the initial undrained response and primary
consolidation for the second method.
The consistency of methods (i) and (ii) is compared in Figure 3.8, which shows the variation
of settlement improvement factors with column length for three arrangements of stone
columns beneath a 3 m square footing. For the purposes of comparing the consistency of both
methods it suffices to examine their relative performance; a detailed analysis of the results is
provided in chapter 5. It appears that method (ii) predicts slightly higher settlement
improvement factors than method (i) for footings supported by 9 columns, while the opposite
is observed for the footing supported by 5 columns. However, both methods predict a similar
variation of settlement improvement factor with column length for all configurations of
columns (maximum normalised error is no greater than 7%). To save computational time,
method (ii) is therefore adopted for all the subsequent FEA, unless otherwise stated.
Settlement improvement factor, n

1.0 2.0 3.0 4.0
0
4 cols - s=2.0 m
5 cols - s=1.0 m
2
9 cols - s=1.0 m
Method (i)
Column length, L (m)
4 Method (ii)
10
12
14
Figure 3.8 - Influence of analysis type upon the settlement improvement factors for various groups of columns
3.4.2 Modelling of column-soil interface

Interface elements are available in PLAXIS 3D Foundation to model the interaction between
smooth and rough surfaces i.e. between piles/basement walls and soil. Interface elements can
simulate gap and slip displacements which are normal and parallel to the interface,
respectively. The 16 node elements consist of 8 pairs of nodes (2 nodes at the same point; 1
for the soil and 1 for the wall) and are shown in Figure 3.9.
The element behaviour is modelled as elastic-plastic, with the Coulomb criterion adopted to
distinguish between elastic and plastic behaviour. The loss of strength at the interface is
modelled with a strength reduction factor (Rinter), which relates the interface strength to the
soil strength through friction angle (φ) and cohesion (c):
R = RS QR (3.4)
81.R = RS 81.QR ≤ 81.QR (3.5)

Figure 3.9 - Interface elements for a group of 5 columns beneath a 3 m square pad footing
In reality elements have zero thickness, but are assigned a virtual thickness in order to
determine element stiffness. Gap and slip displacements are calculated from the oedometric
(Eoed,i) and shear (Gi) moduli, respectively. The moduli are related by the expression:
1 − R
,R = 2$R
1 − 2R
(3.6)
where υi = 0.45
Guetif et al. (2007) adopt rigid interface elements (i.e. Rinter = 1) on the basis that stone
columns are tightly interlocked with the surrounding soil and a perfect bond exists along the
column-soil interface. However Gäb et al. (2008), Elshazly et al. (2008a), Domingues et al.
(2007a) and many other authors model a perfect bond along the column-soil interface by
omitting interface elements.
The influence of interface elements upon the settlement performance of various

configurations of columns beneath a 3 m square footing is presented in Figure 3.10. It can be
seen that column arrangements with interface elements yield lower settlement improvement
factors i.e. they over-predict the settlement of stone columns. This may be attributed to the
elastic-plastic material model and the increased Poisson’s ratio (υi = 0.2→0.45) assigned to
interface elements in PLAXIS 3D Foundation. It can be shown in equation 3.6 that an
increased Poisson’s ratio yields lower shear moduli, which increases slip displacement.
The influence of interface elements also appears to become more pronounced with an
increasing number of columns. It will be shown in chapter 5 that closely-spaced columns
punch into the underlying soil, whereas columns at higher A/AC tend to bend and bulge.
Therefore, as A/AC decreases, a larger proportion of the applied load is transferred along the
side of columns and along the interface elements, which induces more displacement.
It is clear from Figure 3.10 that a similar variation of settlement improvement factors with
column length was observed for columns modelled with and without interface elements. As
columns are tightly interlocked with the surrounding soil it was deemed appropriate to model
the column-soil interface in the subsequent FEA by omitting interface elements, in keeping
with the prior work of several other authors.

1.0 2.0 3.0 4.0
0
4 cols - s=2.0 m
2 5 cols - s=1.0 m
9 cols - s=1.0 m
Interface
4
No interface
10
12
14
Figure 3.10 - Influence of interface elements upon settlement improvement factors for various groups of
columns
3.4.3 Modelling of column installation effects

It was shown in the literature review (see Section 2.4) that column installation effects
increase the horizontal stress and stiffness in the surrounding soil. It was also postulated that
any changes to the in situ stress levels in fine grained soils are solely due to the displacing
effects of the poker, as the forces generated from horizontal vibrations are attenuated.
Measurements from field and laboratory studies suggest that the effects of column installation
can be accurately predicted by Cylindrical Cavity Expansion Theory (CCET). McCabe et al.
(2008) conducted a three-dimensional FEA to assess the ability of PLAXIS 3D Foundation to
simulate the column installation process using cavity expansion.
The change in the stress state of the soil due to column installation may be predicted by an
undrained CCE formulation developed by Gibson & Anderson (1961). As the poker expands
a hole from an initial diameter of zero to the final diameter of a constructed stone column, the
lateral strain is effectively infinite. Consequently, Egan et al. (2008) state that it is necessary
to assume cavity expansion pressure reaches its limit value (plim) in order to accurately
simulate column installation effects.
The relationship between cavity pressure and lateral expansion is derived from CCET and is
shown in Figure 3.11, where the cavity pressure (p) is normalised by the limit pressure and
the current borehole radius (a) is normalised by the initial radius (a0). It can be seen that
considerable lateral expansion is required to reach the limit pressure.
1.0
Normalised Cavity Pressure (p/plim )
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
a = 1.15a0 a = 1.4a0
0.1
0.0
1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0
Normalised Expansion (a/a0 )
Figure 3.11 - Variation of cavity pressure with radius (undrained CCE) (McCabe et al., 2008)
McCabe et al. (2008) simulated the lateral expansion of a 600 mm diameter column in the
Bothkennar soil profile using PLAXIS 3D Foundation. The Mohr Coulomb and Hardening
Soil models were adopted to simulate the behaviour of the stone column and soil,
respectively. Various degrees of lateral expansion were applied (a/a0 = 1.03, 1.06, 1.1, 1.33,
1.67) and the excess pore pressure and radial stress in the soil were measured at mid-depth
along a 5 m long column.
The variation of excess pore pressure and radial stress with normalised radial distance from
the centre of the stone column is shown in Figures 3.12(a) and 3.12(b), respectively. The
excess pore pressure and radial stress is also compared with theoretical curves, developed by
Randolph et al. (1979b) and Gibson & Anderson (1961), respectively. Two curves are
presented for radial stress in Figure 3.12(b); the first curve corresponds to limit pressure in
the expanding cavity (i.e. field conditions) and the second curve corresponds to the actual
lateral expansion applied in the numerical analysis.
200 200
Change in radial total stress (kPa)

PLAXIS 3-D
PLAXIS 3-D
150 150 CCE
Randolph et al (1979)
CCE (plim)
Excess pwp (kPa)
100 100
50 50
0 0
-50 -50
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Normalised radial distance (r/a) Normalised radial distance (r/a)
(a-i) (b-i)
200

200
PLAXIS 3-D
PLAXIS 3-D 150 CCE
150
Excess pwp (kPa)
Randolph et al (1979) CCE (plim)
100 100
50 50
0 0
-50 -50
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
(a-ii) (b-ii)
200 200
PLAXIS 3-D
150 PLAXIS 3-D 150 CCE
Randolph et al (1979) CCE (plim)
Excess pwp (kPa)
100 100
50 50
0 0
-50 -50
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
(a-iii) (b-iii)
Figure 3.12 - Variation of (a) Excess pore pressure (pwp) and (b) total radial stress with normalised radial
distance for lateral expansions (a/a0) of (i) 1.03, (ii) 1.1 and (iii) 1.33 (McCabe et al., 2008)
It can be seen in Figure 3.12 that PLAXIS 3D Foundation data tends towards the plim curve as
lateral expansion increases. It can also be seen that PLAXIS 3D Foundation captures the
variation of radial stress with radial distance from the stone column quite well. However, a
large scatter for excess pore pressure and radial stress is evident, especially for a < r < 2a.
The level of scatter increases with lateral expansion and would not be acceptable at the level
required to reach plim. Furthermore, the lateral extent of the scatter precludes this technique
for simulating the installation effects of groups of stone columns. The results suggest that
PLAXIS 3D Foundation is currently incapable of simulating the column installation process
by applying large strain cavity expansions. An approximate technique to increase the stress
state of the soil by increasing the coefficient of horizontal stress (K0) is recommended by the
authors and this is investigated in chapter 5.
66
Chapter 4
Development and validation of soft soil
profile
4.1 Introduction
The FEM is a powerful investigative tool which has many advantages over laboratory
experiments and large scale field tests. However, the FEM is still an approximate technique
which idealises real-life situations into a set of continuum components and adopts
constitutive models to simulate soil behaviour. As a result, it is necessary to validate the
output of the FEM to ensure that the real-life situation is accurately modelled.
The ability of PLAXIS 3D Foundation to accurately capture the settlement performance of

stone columns is validated by simulating three field load tests, described by Mitchell &
Huber (1985). The load tests were conducted on single columns within a large group, which
are founded in a layered estuarine deposit. The predicted load-settlement curves are
compared with observed field data and also with two independent axisymmetric FE
simulations of the field tests, which also allows the merits of a three-dimensional FEA to be
evaluated.
The main aim of this thesis is to examine the behaviour of small groups of stone columns in
soft clay. A model of the soft soil profile at the well-characterised Bothkennar test site is
developed (using the Hardening Soil model in PLAXIS) in this chapter. The Bothkennar test
site consists of a soft uniform clay, which is overlain by a stiff crust and is representative of
many sites where the applicability of stone columns is of growing interest. The adopted soil
profile and material parameters are validated by back-analysing a load test on an unreinforced
rigid pad footing at the Bothkennar test site, described by Jardine et al. (1995).
Finally, the accuracy of PLAXIS 3D Foundation to predict the settlement performance of

stone columns in the Bothkennar soil profile is examined. The settlement performance of an
infinite grid of columns is simulated and compared against one-dimensional compression
Development and validation of soft soil profile 67
theory, a settlement improvement database (collated by McCabe et al., 2009) and a selection
of analytical design methods.
4.2 Description of load tests in a layered estuarine deposit

4.2.1 Background
In 1976 the vibro replacement technique was used to support a waste water treatment plant
located in Santa Barbara, California. Over 6500 stone columns were installed at the site in
what was the first major application of the vibro replacement technique along the west coast
of America (Mitchell & Huber, 1985). The vibro replacement technique was preferred to
traditional foundation solutions such as piling and ‘remove and replace’ as it generated less
noise and required less construction time. The technique also offered the extra benefit of
reducing the liquefaction potential of the site, which is located in a seismically active area.
The settlement performance of the stone column foundations was assessed by conducting 28
load tests on individual columns within large groups of columns as construction proceeded.
The field tests are described in detail by Mitchell & Huber (1985) who also conduct an
axisymmetric FEA to simulate the field load tests and to predict the long term settlements of
the wastewater treatment plant.
4.2.2 Description of site conditions

The site is adjacent to the Pacific Ocean and is located in a historic tidal estuary. The upper
1–3 m is a recent fill of clayey sand containing wood, masonry, rubble, asphalt, glass and
metal in variable concentrations. The recent fill is underlain by soft estuarine deposits which
increase in thickness from 5–16 m across the site. The estuarine deposits consist of inter-
layered silty and sandy clay to clayey and silty sand, with occasional layers of sandy silt.
Some layers and lenses of sand, with minor amounts of gravel are locally present, usually in
the lower portions of the deposits. The estuarine deposits are underlain by older marine
deposits, which are thought to extend to 600 m. These deposits consist of clayey sand and
silty sand, with lesser amounts of sandy clay and sandy silt. The ground water level is located
at 1.5 m below the ground surface.
4.2.3 Stone column design & construction

The stone columns were formed using the ‘wet’ system, whereby a poker penetrated the soil
with the aid of water jets. A controlled quantity of stone was tipped into the resulting
borehole and the poker was re-lowered to compact the stone and force it laterally into the
surrounding soil. This process of adding and compacting stone was repeated until a stone
column, which was tightly inter-locked with the surrounding soil, was formed up to the
ground surface. The final diameter of stone columns ranged from 0.81–1.22 m, with an
average of 1.07 m. The stone columns were installed on square or rectangular patterns and
spacing ranged from 1.2 m × 1.5 m in the most heavily loaded areas to 2.1 m × 2.1 m in areas
between buildings. The length of columns varied from 9–15 m across the site and all columns
penetrated 0.3 m into the underlying older marine deposits.
4.2.4 Material parameters

Undisturbed soil samples were obtained from various depths using thin walled Shelby tubes
after column installation. Despite a large degree of variation within estuarine and marine
deposits, Mitchell & Huber (1985) classify the samples as either cohesive or cohesionless,
depending on whether the predominant soil type was clay or sand, respectively. The
cohesionless soils were assumed to be free draining in the vicinity of the stone columns and
were only subject to drained triaxial compression tests. The cohesive soils were not
considered to be completely free draining and were subject to both consolidated undrained
and drained triaxial tests; these tests were used to establish the short and long term behaviour
of the cohesive soils, respectively. The gravel used for the stone columns was reconstituted
into specimens and subjected to consolidated drained triaxial tests. Mitchell & Huber (1985)
average the results of several triaxial tests and convert the results into parameters for the
Duncan & Chang (1970) material model.
While the majority of the parameters for the Hardening Soil model can be readily converted
from the Duncan & Chang (1970) material model, the latter model does not account for stress
path dependency or formulate a cap yield surface. Therefore, it is necessary to define unload-
reload (Eur) and oedometric (Eoed) moduli for the soil layers. However, the settlement of the
field load tests are insensitive to Eur and Eoed, as the field load tests are monotonically loaded
and do not develop large isotropic stresses (Elshazly et al., 2008a). Therefore, practicable
values of Eur = 5E50 and Eoed = E50 are selected for the Hardening Soil model. A Poisson’s
ratio υ = 0.2 is adopted for both the estuarine and marine deposits, which yields reasonable
lateral deformations and does not under-estimate final settlements.
Elshazly et al. (2008b) state that it is not appropriate to use high permeability values for the
stone column and cohesionless layers as the column is infiltrated with silt and clay particles
resulting from column installation and also as a large proportion of fine particles exist in the
cohesionless layers. Accordingly, a vertical permeability kz = 10-5 m/s and 10-6 m/s is adopted
for the stone columns and cohesionless soil layers, respectively. The vertical permeability for
cohesive layers is within the range 10-8 < kz < 10-7 m/s, as suggested by Lambe & Whitman
(1979) for low liquid limit silts and clays. Elshazly et al. (2008b) adopt a ratio of horizontal
to vertical permeability kx/kz = 2 and state that this ratio does not affect the settlement of the
field load tests, as stone columns act as vertical drains and therefore reduce the influence of
horizontal permeability.
A summary of the material parameters for the Santa Barbara wastewater treatment plant
quoted by Elshazly et al. (2008b) are shown in Table 4.1. It should be noted that the failure
ratio (Rf) for the cohesive marine deposit is different to the original value of Rf = 0.84 quoted
by Mitchell & Huber (1985) and is possibly an error.
Table 4.1 - Material parameters for Santa Barbara wastewater treatment plant (Elshazly et al., 2008b)
Parameter Estuarine Estuarine Marine Marine Gravel
cohesive cohesionless cohesive cohesionless
Dry unit weight, γd (kN/m3) 15 15 17 17 18.6
Saturated unit weight, γsat (kN/m3) 19 19 20 20 21.6
Cohesion, c' (kPa) 0 0 0 0 0
Angle of internal friction, φ' (°) 34 38 34 37 41
Poisson’s ratio, υ 0.2 0.2 0.2 0.2 0.2
Secant Young’s modulus, E50ref (kPa) 8500 17000 8700 12600 29200
Power for stress dependency, m 0.65 0.69 0.65 0.90 0.59
Reference pressure, pref (kPa) 100 100 100 100 100
Failure ratio, Rf 0.87 0.69 0.67* 0.67 0.86
Coefficient of vert. permeability, kz (m/s) 1 × 10-8 1 × 10-6 1 × 10-8 1 × 10-6 1 × 10-5
Coefficient of horiz. permeability, kx (m/s) 2 × 10-8 2 × 10-6 2 × 10-8 2 × 10-6 2 × 10-5
* Mitchell & Huber (1985) adopt Rf = 0.84
4.2.5 Field load tests

The settlement performance of single columns within large groups was assessed for various
column spacings. The tests were conducted in accordance with ASTM D1194-72, with a 1.2
m deep circular concrete slab used as a loading plate. The area of the concrete slab was equal
to the tested area for each column and, therefore, increased with column spacing. The load
was applied in 45 kN increments once the settlement rate dropped to 0.25 mm/hr and was
increased to a maximum of 350–400 kN, with the final load maintained for 6 hrs after the
settlement rate slowed to 0.25 mm/hr. The criterion for increasing the vertical load at 0.25
mm/hr was arbitrarily chosen to reduce the length of the load tests and additional settlement
would have occurred if a longer time increment was adopted (Mitchell & Huber, 1985).
Therefore, the recorded load-settlement curves are not specific to fully drained conditions.
4.2.6 Axisymmetric FEA

Mitchell & Huber (1985)
Mitchell & Huber (1985) conduct an axisymmetric FEA to simulate field load tests for three
configurations of stone columns: (i) 1.2 m × 1.5 m; (ii) 1.75 m × 1.75 m and (iii) 2.1 m × 2.1
m. The behaviour of the soil layers is captured using the hyperbolic model developed by
Duncan & Chang (1970). While the thickness of the cohesive and cohesionless layers varied
across the site, Mitchell & Huber (1985) assume a thickness of 1–2 m for the estuarine
deposits and 1–2.4 m for the marine deposits, with approximately 60% more cohesionless
soil in the older marine deposits (Figure 4.1). To account for the vibro-compaction effect due
to column installation, K0 was chosen as 1.0 and 0.5 for the estuarine deposits and older
marine deposits, respectively. The presence of the large group of columns surrounding the
centrally loaded column is accounted for by modelling cylindrical rings of stone columns in
the FEA. The thickness of the rings is chosen to maintain a constant area replacement ratio.
The rings of stone columns are concentrically spaced at column spacings for each
arrangement.
Figure 4.1 - Soil profile for Santa Barbara waste water treatment plant (columns spaced at 1.75 m x 1.75 m)
Elshazly et al. (2008b)

Elshazly et al. (2008b) also conduct an axisymmetric FEA similar to Mitchell & Huber
(1985), but adopt the more advanced Hardening Soil model and account for the presence of
the surrounding group of columns differently. The presence of the columns is again
accounted for by modelling cylindrical rings and varying the thickness of the rings to
maintain a constant area replacement ratio. However, the spacing between the concentric
rings is an average of the distance to the orthogonal and diagonal columns (Figure 4.2).
Figure 4.2 - Conversion of surrounding stone columns into a cylindrical ring for axisymmetric FEA
4.3 Simulation of load tests in a layered estuarine deposit

As part of this thesis, the field load tests described by Mitchell & Huber (1985) were
simulated for the first time using a three-dimensional FEA in PLAXIS 3D Foundation.
Similar to Elshazly et al. (2008b), the Hardening Soil model is adopted in conjunction with
the material parameters outlined in Table 4.1 - with the exception of the failure ratio (Rf) for
the cohesive marine deposit, which was selected as Rf = 0.84, in keeping with the original
reference Mitchell & Huber (1985). The three-dimensional analysis accurately captures the
confinement effect of the surrounding columns, rather than approximating their presence by
cylindrical rings. A constant length L = 11.1 m and diameter d = 1.07 m was adopted for the
field load tests. The FE mesh for columns spaced on a 1.75 m × 1.75 m grid is shown in
Figure 4.3.
The behaviour of the soft clay is modelled using methods (i) and (ii), as outlined in Section
3.4.1. The loading process is accounted for in method (i) by applying the load in 45 kN
increments and allowing the soil to consolidate until the settlement rate dropped below 0.25
mm/hr. The extra settlement due to the pause periods can be seen as vertical drops in the
load-settlement curves shown in Figure 4.4. In contrast all the layers are assumed to behave
in a drained manner for method (ii) and do not develop excess pore pressures.
19.25 m Circular loading cap (D = 1.97 m)

CL
Stone columns:
- d = 1.07 m
- L = 11.1 m
CL
19.2 m
LEGEND
Fill
Concrete footing
Stone column
Estuarine cohesionless
Estuarine cohesive
Marine cohesionless
Marine cohesive
Figure 4.3 - FE mesh for 1.75 m × 1.75 m grid of columns
4.3.1 Comparison of PLAXIS 3D Foundation with field records and axisymmetric FEA
The observed load-settlement curves from the 28 field load tests and numerical simulations
are shown in Figure 4.4. A large degree of variation is observed in the field load tests, which
Mitchell & Huber (1985) attribute to the non-uniformity of soil properties in the estuarine
deposits and the variable depth to older marine deposits.
It can be seen in Figure 4.4 that all the numerical analyses over-predict the load-settlement
response from the field tests and this becomes more pronounced as column spacing
decreases. The distribution of fine particles throughout the soil profile is quite varied and
Mitchell & Huber (1985) state that soil layers defined as cohesionless may have contained
enough fine particles to prevent free drainage during the short duration of the load tests. This
would explain the stiffer response of the observed field measurements.
A comparison of the two axisymmetric analyses reveals that Mitchell & Huber (1985) tend to
predict less settlement than Elshazly et al. (2008b), especially for closely spaced columns.
This may be explained as Mitchell & Huber (1985) position the cylindrical rings of stone
columns at radii equal to the orthogonal distance between columns, while Elshazly et al.
(2008b) position the cylindrical rings of stone columns at radii equal to average distance
between orthogonal and diagonal columns (Figure 4.2). In addition, both analyses are based
on different material models, as Mitchell & Huber (1985) adopt the Duncan & Chang (1970)
Applied Pressure (kPa) Applied Pressure (kPa) Applied Pressure (kPa)

0 40 80 120 160 0 30 60 90 120 150 0 30 60 90
0 0 0
2 2
2
4 4
4
Settlement (mm)
Settlement (mm)
Settlement (mm)
6 6
6
8 8
8
10 10
Field data Field data Field data
Mitchell & Huber (1985) Mitchell & Huber (1985) Mitchell & Huber (1985)
10
12 Elshazly et al. (2008) 12 Elshazly et al. (2008) Elshazly et al. (2008)
PLAXIS 3D: Method (i) PLAXIS 3D: Method (i) PLAXIS 3D: Method (i)
PLAXIS 3D: Method (ii) PLAXIS 3D: Method (ii) PLAXIS 3D: Method (ii)
14 14 12
(i) (ii) (iii)
Figure 4.4 - Comparison of field load test data with numerical simulations for columns spaced on (i) 1.2 m × 1.5 m, (ii) 1.75 m × 1.75 m and (iii) 2.1 m × 2.1 m
74
hyperbolic model and Elshazly et al. (2008b) adopt the more advanced Hardening-Soil
model. It can be seen that Mitchell & Huber (1985) tend to predict less settlement as the
applied load increases. This may be attributed to the limitations of the hyperbolic model,
which is based on the theory of elasticity.
The presence of the surrounding columns and the material model are two important factors to
consider when modelling the load-deformation behaviour of stone columns. PLAXIS 3D
Foundation appears to capture the load-settlement response of the columns closet to the field
data (Figure 4.4). This is most noticeable for closely spaced columns and reflects the accurate
modelling of the surrounding columns at discrete locations, rather than approximating their
presence by cylindrical rings. This highlights the advantage of PLAXIS 3D Foundation over
the axisymmetric analyses.
It can also be seen in Figure 4.4 that methods (i) and (ii) predict very similar load-
displacement curves. This indicates that the excess pore pressures generated in the cohesive
layers from the undrained loading in method (i) are almost fully dissipated by the time the
next load increment is applied. This may be due to the short drainage length to the
cohesionless layers and to the presence of the stone columns which reduce the radial drainage
length. The close agreement between both methods gives confidence to the use of method (ii)
for the subsequent FEA.
4.4 Development of Bothkennar soil profile

4.4.1 Description of Bothkennar test site
The Bothkennar test site is located on the south side of the Firth of Forth, near Grangemouth
in Scotland (Figure 4.5). The site was purchased in 1987 by the Science, Engineering and
Research Council as the national test site for soft soil engineering in the UK. The site
primarily consists of soft uniform clay, commonly referred to as Carse clay, which is
extensively characterised in Geotechnique (1992). The Carse clay was deposited in a stable
marine environment and its thickness ranges from 13–19 m across the site. The Carse clay
rests upon Bothkennar gravel and is overlain by a 1.5 m stiff crust. Piezometers indicate that
the ground water level is hydrostatic and varies from 0.5–1.0 m below ground level (Hight et
al., 1992).
Figure 4.5 - Bothkennar test site location (Nash et al., 1992a)
4.4.2 Soil classification and initial stress state

The soil at the Bothkennar test site was classified by Nash et al. (1992a) following a range of
index tests, which are presented in Figure 4.6. The moisture content increases from 30% in
the crust to 80% at 8 m and then decreases to 40% above the gravel layer. The Atterberg
limits are also shown in Figure 4.5 and when plotted on the plasticity chart indicate that the
soil is an inorganic clay of high plasticity. However, in accordance with BS 5930, and
following findings by Paul et al. (1992) that the clay fraction of the soil varies from 35–50%,
the soil is classified as a silty clay. The organic content was found by loss on ignition at
425°C and ranges from 3–8%. The bulk density varies from 1600–1800 kg/m3 and the
adopted profile for all subsequent FEA is shown in Figure 4.6.
Adopted unit weight (γ)

profile for FEA
Figure 4.6 - Geotechnical profile at Bothkennar test site (Nash et al., 1992a)
Hight et al. (1992) suggest that post depositional processes such as erosion, changes in
groundwater levels and bonding occurred at the Bothkennar test site. According to the
geological history a maximum drop of 15 kPa in effective overburden pressure could have
occurred at the Bothkennar test site. This implies that over-consolidation ratios should be
high in the upper layers of the Carse clay and reduce with depth. The over-consolidation ratio
determined on the basis of this stress history is compared with data from incremental load
tests by Nash et al. (1992a) in Figure 4.7, where yield stress ratio is equivalent to over-
consolidation ratio. While the profile agrees favourably in the upper layers, it under predicts
the over-consolidation ratio below 4 m where a constant value of 1.55 better represents the
actual stress state. It is suggested that the soil below 4 m may be influenced by ageing, which
accounts for the observed over-consolidation ratio being larger than that due to the stress
history. The profile adopted for all subsequent FEA is shown in Figure 4.7 where the over-
consolidation ratio based on a 15 kPa drop in effective overburden pressure defines the upper
2.5 m of the soil profile and a constant over-consolidation ratio of 1.5 defines the soil below
this level.
Adopted OCR
profile for FEA
(a) (b)
Figure 4.7 - (a) Yield stress and (b) yield stress ratio from incremental load consolidation tests (Nash et al.,
1992a)
The profiles of the lateral total stress measured in situ and the coefficient of lateral earth
pressure (K0) with depth are presented by Nash et al. (1992a) in Figures 4.8(a) and 4.8(b),
respectively. The coefficient of lateral earth pressure was determined from a self-boring
pressuremeter, spade cells and dilatometer tests. The coefficient of lateral earth pressure is
high in the upper layers, which Hight et al. (1992) state is consistent with recent groundwater
fluctuations, and decreases to 0.6–0.9 in the lower Carse layers. The distribution of horizontal
stress with depth adopted for the FEA is shown in Figure 4.8.
Adopted K0
profile for FEA
(a) (b)
Figure 4.8 - Variation of (a) lateral total stress (b) K0 with depth (Nash et al., 1992a)
4.4.3 Strength characteristics

The strength characteristics of reconstituted Carse clay were investigated by Allman &
Atkinson (1992). Samples were removed from depths between 3.5–6.5 m, reconstituted and
formed into a slurry 1.25 times the liquid limit. The slurry was then one-dimensionally
consolidated and transferred to the triaxial cell for testing, where the samples were one-
dimensionally compressed or swelled to normally-consolidated or slightly over-consolidated
states before shearing. Most samples reached well defined constant stress ratios (M = q/p') at
shear strains above 15%. The critical stress ratios, which correspond to zero rate of dilation,
are Mc = 1.38 and Me = 1.00 for compression and extension, respectively. These critical
stress ratios are equivalent to critical state friction angles of φc = 34° and φe = 37° for
compression and extension, respectively. These friction angles are larger than would be
expected for a high plasticity clay but this is attributed to the large proportion of angular silt
particles present in the Carse clay. While no effective cohesion was observed in the tests, a
nominal value of 1 kPa and 3 kPa was adopted for the Carse clay and crust, respectively, to
ensure numerical stability in all the subsequent FEA. A higher value of cohesion is adopted
for the crust as overburden stresses are lowest in this layer and the stress state is therefore
closer to the Mohr Coulomb yield surface.
4.4.4 One-dimensional stiffness properties

Nash et al. (1992b) undertook a comprehensive study to examine the one-dimensional
stiffness behaviour of Carse clay. Samples were recovered from the ground at various depths
and were subject to a series of incremental load, constant rate of strain and restricted flow
tests. The majority of the tests were incremental load tests, conducted in a fixed ring
oedometer cell, which were used to establish the voids ratio (e) and coefficient of
compressibility (CC) with depth. On the basis of the oedometer results from standard
incremental load tests, the soil profile at the Bothkennar test site is divided into a crust, upper
and lower Carse clay layers, as shown in Figure 4.9.
Compression index, Cc Initial voids ratio, e0

0.0 0.5 1.0 1.5 0.0 1.0 2.0 3.0
0 0
Adopted CC Adopted e0
2 2
profile for FEA profile for FEA
4 4
6 6
Depth (m)
Depth (m)
8 8
10 CC / (1+e0) CC 10
12 12
14 14
16 16
18 18
20 20
(a) (b)
Figure 4.9 - Variation of (a) compression index Cc and (b) initial voids ratio e0 with depth (Nash et al., 1992b)
The behaviour of the samples prior to yielding was not routinely measured by Nash et al.
(1992b). However, Allman & Atkinson (1992) also investigated the stiffness behaviour of
reconstituted Carse clay and found the slopes of the normal compression (λ) and swelling (κ)
lines to be 0.181 and 0.025, respectively. The ratio between these indices (λ/κ = 7.2) was
used to determine the coefficients of swelling (CS) from the one-dimensional data presented
by Nash et al. (1992b), i.e. λ/κ = CS/CC = 7.2. The one-dimensional stiffness parameters are
converted into three-dimensional parameters for the Hardening Soil model using the
following expressions (Brinkgreve & Broere, 2006):
2.31 ! 9M f

= (4.1)

2.31 ! 9M 1 ! 1 2f

(4.2)
V 1
It is expected that the influence of the stiff crust will, to a certain extent, mask the behaviour
of stone columns in the upper sections, i.e. prevent bulging. In addition, the stiff crust will
absorb a higher proportion of the applied load and thus minimise the influence of elevated
stress levels beneath the edge of rigid footings. However, the stiff crust is an important
feature of soft soil stratigraphy; for example, the soil profile at Kinnegar, Belfast consists of
soft Belfast ‘sleech’ overlain by relatively stiff layers of fill and silty sand (McCabe, 2002).
Therefore, the presence of the stiff crust will yield a more realistic analysis of the settlement
performance and deformational behaviour of small groups of stone columns.
4.4.5 Permeability and consolidation coefficients

The hydraulic conductivity characteristics were examined by Leroueil et al. (1992) using
various laboratory and in situ tests. The self boring permeameter gives the most realistic
profiles and indicates that the horizontal hydraulic conductivity (kh0) increases from 1.2×10-9
m/s at 3 m to 2.1×10-9 m/s at 6 m and then decreases to 7.5×10-10 m/s at 15 m. The vertical
hydraulic conductivities (kv0) observed from laboratory tests are lower than the equivalent
horizontal values and an anisotropy ratio (kh0/kv0) of 1.5–2.0 was found. The variation of
hydraulic conductivity (k) with voids ratio (e) was examined in order to define a hydraulic
change index (Ck = ∆e/∆logk). This was found to be adequately defined by Ck = 0.5e0.
A summary of the parameters developed for the Bothkennar test site is outlined in Table 4.2.
Table 4.2 - Summary of adopted material parameters for the Bothkennar test site
Soil Parameter Crust Upper Carse Lower Carse
clay clay
Depth (m) 0.0 – 1.5 1.5 – 2.5 2.5 – 14.5
Bulk unit weight, γ (kN/m3) 18.0 16.5 16.5
Over-consolidation ratio - - 1.5
Pre-overburden stress (kPa) 15 15 -
Coefficient of lateral earth pressure, K0 1.5 1.0 0.75
Effective cohesion, c' (kPa) 3 1 1
Angle of internal friction, φ (°) 34 34 34
Initial voids ratio, e0 1.0 1.2 2.0
Compression index, CC 0.07 0.25 1.12
Swelling index, CS 0.01 0.03 0.16
Reference pressure, pref (kPa) 13 20 30
Vert. coefficient of permeability, kvert (m/day) 6.9 × 10-5 6.9 × 10-5 6.9 × 10-5
Horz. coefficient of permeability, khorz (m/day) 1.0 × 10-4 1.0 × 10-4 1.0 × 10-4
4.5 Validation of Bothkennar soil profile

4.5.1 Description of field load test
The bearing capacity and load-displacement behaviour of two rigid pad footings was
investigated at the Bothkennar test site by Jardine et al. (1995). The short and long term
behaviour of the footings was examined as the first footing (Pad A) was loaded to failure and
the second footing (Pad B) was loaded to 67% of the ultimate bearing capacity of Pad A. The
loading rate is shown in Figure 4.10 and it can be seen that Pad A was loaded to failure in a
short space of time while Pad B continued as a maintained load test for more than 2 years.
The loading was applied using kentledge blocks; pauses in the loading rate occurred
overnight and whenever the settlement rate exceeded 8 mm/hr. The footings were founded at
0.8 m below ground level and Pads A and B were 2.2 m and 2.4 m in width, respectively.
Applied Pressure (kPa)
120 Test B (continues

for 2 years)
80
Test A
40
0
0 20 40 60 80 100 120 140
Time (hrs)
Figure 4.10 - Variation of load with time (Jardine et al., 1995)
The footings were instrumented with pneumatic piezometers, spade cells, inclinometers and
magnetic extensometers. The surface settlement was measured with a precise level, giving a
nominal resolution of 0.1 mm. In total ten targets were set into the concrete pads and 16
targets were augured into the surrounding soil.
4.5.2 Simulation of field load test with PLAXIS 3D Foundation

The load test on Pad A was simulated with PLAXIS 3D Foundation as an undrained loading
due to the short duration of the load test. Consequently, the Carse clay layers were specified
to behave in an undrained manner. However, the upper layers of the Bothkennar test site
consist of a weathered crust underlain by a shelly layer (Nash et al., 1992a). These layers are
combined together to form the crust in the adopted soil profile. As the ground water level
varies from 0.5–1.0 m below ground level and the lower portion of the crust consists of a
shelly layer, it was deemed appropriate to model the crust as a drained material.
The load test on Pad B was not simulated with PLAXIS 3D Foundation as this load test
contained an unload-reload loop and the recorded footing displacement includes both primary
and secondary settlement (Jardine et al., 1995). The aim of this thesis is to examine the long
term primary settlement response of small groups of stone columns supporting pad footings.
In addition, the Hardening Soil model does not account for secondary settlement and is
therefore unable to accurately simulate the load test on Pad B.
4.5.3 Comparison of PLAXIS 3D Foundation with field measurements

The load-displacement behaviour of Pad A recorded by Jardine et al. (1995) is shown in
Figure 4.11. Pad A reaches an ultimate bearing capacity of 138 kPa and the settlement at the
end of the loading was approximately 190 mm. The simulated load-displacement curve from
PLAXIS 3D Foundation is also shown in Figure 4.11. It can be seen that PLAXIS 3D
Foundation predicts the settlement behaviour of Pad A quite well.
140
120
Jardine et al. (1995)
Applied pressure (kPa)
Plaxis 3D Foundation
100
80
60
40
20
0
0 20 40 60 80 100 120 140 160 180 200
Vertical displacement (mm)
Figure 4.11 - Load-displacement behaviour for pad footings (Jardine et al., 1995)
However, it appears that PLAXIS 3D Foundation slightly under-estimates the stiffness

response of the field load test, especially at low applied pressures, which are of interest in this
thesis. This may be attributed to increased soil stiffness at low strain levels, which is not
accounted for in the Hardening Soil model. The stiffness parameters for the Hardening Soil
model are determined from laboratory tests, where the specimens are subject to relatively
large strain (Figure 4.12). PLAXIS 3D Foundation under-estimates the stiffness response of
Pad A at low applied pressure, as shear strain is lowest at this load level.
Figure 4.12 - Characteristic stiffness-strain behaviour of soil with typical strain ranges for laboratory tests and
structures (after Atkinson & Sallfors, 1991)
Another contributing to the under-estimation of the stiffness response by PLAXIS 3D

Foundation may be that the soil profile encountered by Jardine et al. (1995) at the test
location is slightly different to the soil profile adopted for the FEA. It can be seen in Table
4.3 that Jardine et al. (1995) encountered a 0.3 m shelly layer and also that the Carse clay
occurs at a shallower depth. The Carse clay would yield a relatively stiff response under
undrained loading, which would result in a stiffer footing response.
Table 4.3 - Summary of soil profile to 7 m depth (Jardine et al., 1995)

Stratum Depth Soil type Typical index parameters
(m) < 2 µm (%) w (%) IP (%) IL γ (kN/m3)
I 0.0 – 1.0 Weathered clayey silt crust 15 40 20 0.4 18.0
II 1.0 – 1.3 Shelly layer Not applicable
III 1.3 – 2.2 Soft clayey silt with some 15 50 30 0.6 17.0
shell fragments
IV 2.2 – 7.0 Soft black silty clay with 20 – 40 60 – 75 30 – 50 0.6 – 15.5 –
fine mottling and 1.0 16.5
occasional silt laminae
In conclusion, it can be seen that PLAXIS 3D Foundation captures the settlement

performance of the field load test in soft clay quite well. This is encouraging and validates the
adopted soil profile and the choice of material parameters.
4.6 Simulation of an infinite grid of stone columns in Bothkennar

The settlement performance of an infinite grid of stone columns in the Bothkennar soil profile
is simulated using PLAXIS 3D Foundation. A set of material parameters is developed for the
stone backfill from previous numerical studies and field records. However, before these
material parameters are validated, the settlement performance of an untreated widespread
loading is compared with one-dimensional compression theory. This ensures that the
settlements, and not just the settlement improvement factors, are in the correct order of
magnitude.
The settlement performance of an infinite grid of stone columns is then analysed over a
typical range of A/AC values presented in the settlement database by McCabe et al. (2009).
The unit cell concept was adopted to simplify the analysis of an infinite grid of columns to
one column and its surrounding zone of influence (see Section 2.5.1). Settlement
improvement factors determined from PLAXIS 3D Foundation are compared with a
settlement database of field tests (collated by McCabe et al., 2009) and a selection of current
analytical design methods.
The settlement performance of stone columns at the Bothkennar test site was determined at
50 kPa. The ultimate bearing capacity recorded by Jardine et al. (1995) for an unreinforced
footing at the Bothkennar test site was 138 kPa (Figure 4.11). Applying typical factors of
safety (2.5–3.0) yields allowable bearing pressures in the range of 46–55 kPa. Therefore, 50
kPa is deemed a typical working load for the Bothkennar test site.
4.6.1 Development of material parameters for stone backfill

The bulk unit weight for the stone backfill γ = 1900 kg/m3 is representative of stone columns
in soft soils and is similar to values adopted by Mitchell & Huber (1985), Domingues et al.
(2007a) and Gäb et al. (2008). Coefficients of permeability in the vertical and horizontal
direction of ky = kh = 1.7 m/day are adopted for the stone backfill, which are similar to the
values adopted by Elshazly et al. (2008b).
The angle of internal friction adopted for stone columns φ = 45° is chosen on the basis of
findings by McCabe et al. (2009), who conducted a review of field tests on stone columns
and found that φ = 40° is conservative for columns formed using the bottom feed system. A
nominal value for cohesion c = 1 kPa is adopted to ensure numerical stability. The angle of
dilatancy is determined from the empirical relationship ψ = φ – 30°, developed by Bolton
(1986). A positive angle of dilatancy ensures that columns dilate when subject to high stress
ratios.
Barksdale & Bachus (1983) report that Young's moduli (E) for stone columns, which are
back-calculated from measured settlements and recommended by other authors, typically
range from 30–58 MPa. However, these values are specific to columns formed using the top
feed system and may under-estimate Young’s moduli for stone columns following the
assertion by McCabe et al. (2009) that columns formed using the dry bottom feed system
yield a better settlement performance. Consequently, a higher Young’s modulus E50 = 70
MPa is adopted for the stone columns. A similar value was also adopted by Gäb et al. (2008),
who conducted a numerical investigation of stone columns supporting an embankment in
loose-medium compacted sand and weak clayey silt. The unload-reload Young’s modulus Eur
= 210 MPa is also similar to Gäb et al. (2008) and its selection is based on the relationship
(Eur = 3E50) proposed by Brinkgreve & Broere (2006).
4.6.2 Comparison of PLAXIS 3D Foundation with one-dimensional compression theory

The accuracy of PLAXIS 3D Foundation to predict the settlement of an untreated infinitely
wide footing at the Bothkennar test site is compared with one-dimensional compression
theory. The concrete footing is 0.6 m in thickness and is located 0.6 m below ground level.
The initial stress state and one-dimensional stiffness parameters, as outlined in Sections 4.3.2
and 4.3.4, respectively, are used in conjunction with equations 4.3 and 4.4 to estimate the
settlement of an infinitely wide footing (suc) at the Bothkennar test site.
If σ'y,0 + ∆σ'y < σ'y, max:
V ′,M + ∆′
_ = \ ^ 5 b
1 + 9M ′,M
(4.3)
If σ'y,0 + ∆σ'y > σ'y, max:
∆′,z V ′,M + ∆′

_ = \ ^ 5 + 5 b
1 + 9M ′,M 1 + 9M ′,z
(4.4)
where σ'y, 0 and σ'y, max = in situ and maximum vertical effective stress, respectively
The Bothkennar soil profile is divided into seven layers and the settlement is calculated on
the basis of the vertical effective stress at the centre of each layer (Table 4.4). The settlement
for each layer and the variation of settlement with depth is shown in Figures 4.13(i) and
4.13(ii), respectively. It can be seen in Figure 4.13(i) that a slight variation exists between
PLAXIS 3D Foundation and one-dimensional compression theory for the settlement of each
layer with depth, which may be attributed to a difference in the definition of the cap yield
surface. However, it is clear from Figure 4.13(ii) that PLAXIS 3D Foundation predicts the
settlement response of an untreated infinitely wide footing at the Bothkennar test site quite
well (normalised error for the total settlement is less than 3%).
Table 4.4 - Comparison of one-dimensional compression theory with PLAXIS 3D Foundation

One-dimensional compression theory PLAXIS 3D Foundation
Layer σ'y,0
Thickness σ'y, max σ'y,0 + ∆σ'y * suc σ'y,0 suc
(mm)(kPa) (kPa) (kPa) (mm) (kPa) (mm)
1 0.0 - 0.918 33 72 12 18 9
2 0.9 - 1.925 40 79 36 25 29
3 1.9 - 3.935 53 89 185 35 169
4 3.9 - 5.949 72 102 127 47 120
5 5.9 - 7.962 92 116 88 60 86
6 7.9 - 9.976 111 129 61 73 63
7 9.9 - 11.9
89 131 143 40 86 46
8 11.9 - 13.9
102 150 156 23 100 36
Total settlement 572 Total settlement 558
* ∆σ'y includes additional pressure due to weight of concrete, i.e. ∆σ'y = 50 + 0.6×(24-18) ≈ 54 kPa
Vertical displacement per layer (mm) Vertical displacement (mm)

0 50 100 150 200 0 200 400 600
0 0
2 2
4 4
Depth (m)
Depth (m)
6 6
8 8
10 10
Plaxis 3D Foundation Plaxis 3D Foundation
12 12
1D compression theory 1D compression theory
14 14
(i) (ii)
Figure 4.13 - Settlement profile of Bothkennar test site for wide area loading, calculated using one-dimensional
compression theory and PLAXIS 3D Foundation
4.6.3 Comparison of PLAXIS 3D Foundation with settlement database

The results from the numerical analysis of an infinite grid of end-bearing columns are
compared with a settlement database of field data collated by McCabe et al. (2009) in Figure
4.14. The field data agrees quite well with Priebe's (1995) basic design curve (n0), assuming
an angle of friction φ = 40°. It can also be seen in Figure 4.14 that while PLAXIS 3D
Foundation captures the variation of settlement improvement factors (n) with area ratio
(A/AC) quite well, it tends to over-estimate the field data.
However, McCabe et al. (2009) suggest that settlement improvement factors for some of the
case studies may have been under
under-estimated due to the following reasons:
(i) The time at which settlement was measured may not be consistent for all tests and
untreated settlement may be under
under-estimated,
estimated, as primary consolidation was not complete
when the measurements were taken.
(ii) Not all of the field data are specific to end
end-bearing
bearing columns and, therefore, the measured
settlement includes the settlement of the soil deposit underneath the base of columns.
The high n values predicted by PLAXIS 3D Foundation may also be related to the high
compressibility of the lower Carse clay layer relative to the stone columns, i.e. high modular
ratio (Ecol/Esoil). Field experience suggests that Ecol/Esoil typically ranges from 10–50
10 (Castro
& Sagaseta, 2009). However, columns at the Bothkennar test site are formed in a soft clay
with an average Ecol/Esoil = 100 (see Figure 6.7) for the lower Carse clay layer. This is
significantly higher than the upper limit suggested by Castro & Sagaseta (2009), but is not
unrealistic as a finite element study by Barksdale & Bachus (1983) indica
indicates
tes Ecol/Esoil up to
100 for columns formed in soft cohesive soil.
11
Venmans (1998) Cooper and Rose (1999)
10 Munfakh et al (1983) Greenwood (1991)
Settlement Improvement Factor, n
9 Raju et al (2004) De Cock and d'Hoore (1994)

Greenwood (1970) Baumann and Bauer (1974)
8 Watt et al (1967) Greenwood (1974)
Goughnour and Bayuk (1979) Raju (1997)
7
Bell (1993) Kirsch (1979)
6
Increase to Priebe's (1995) n0 to account for Ecol/Esoil
5 (based on Domingues et al., 2007a)
4
Plaxis 3D Foundation (φ'=45°)
3
2
26.2
1 18.2
Priebe's (1995) n0 with φ'=40°, υ= ⅓
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Area Replacement Ratio, 1/r = A/Ac
Figure 4.14 - Comparison of settlement improvement factors for an infinite grid of end-bearing
end bearing columns with
field data collated by McCabe et al. (2009)
A finite element study by Domingues et al. (2007a) shows that high Ecol/Esoil leads to an
increase in n values. The authors observed that increasing Ecol/Esoil from 10 to 100 and from
40 to 100 yielded an increase of n values in the range 80
80–200%. These increases
creases are applied
to Priebe’s (1995) basic design (n0) and are shown in Figure 4.14.. It can be seen that PLAXIS
3D Foundation agrees quite well with field data when adjusted for a high Ecol/Esoil. This gives
confidence to the selection of the material parameters adopted for the stone backfill (see
Section 4.6.1) and also the ability of PLAXIS 3D Foundation to predict the settlement
performance of large groups of stone columns in soft clay.
4.6.4 Comparison of PLAXIS 3D Foundation with analytical design methods

The settlement performance of an infinite grid of stone columns from PLAXIS 3D
Foundation is compared with various analytical design methods in this section. The analytical
design methods chosen for comparison are Balaam & Booker (1981), Pulko & Majes (2005)
and Priebe (1995). All of these design methods are based on the unit cell concept, which is
consistent with the boundary conditions adopted for the numerical analysis. Analytical design
methods are based on more simplified material models than those adopted by PLAXIS 3D
Foundation and, consequently, it is necessary to adopt an initial stress state and a set of
material parameters which are consistent with the numerical analysis.
Determination of initial stress state

The influence of the initial stress state upon the settlement performance of stone columns is
taken into account by both Priebe (1995) and Pulko & Majes (2005):
• Priebe (1995) sets a minimum coefficient of lateral earth pressure K0,soil = 1.0 to account
for the effects of column installation. However, K0 > 1.0 for the crust and upper Carse
clay at the Bothkennar test site (see Section 4.3.2). Priebe (1995) is modified to account
for higher K0 values in the upper layers by changing the pressure difference causing
bulging (= KA,col.σcol – K0,soil.σsoil) at the column-soil interface.
• Pulko & Majes (2005) do not specify a minimum K0 and the K0 profile developed in
Section 4.3.2 for the numerical analysis is adopted for this design method.
The coefficients of lateral earth pressure and material parameters adopted for the stone
backfill in the analytical design methods are outlined in Table 4.5.
Table 4.5 - Summary of material parameters for Bothkennar soil profile

Soil Stone backfill
Depth Unit weight, γ K0 Unit weight, γ φ ψ
(m) (kN/m3) (kN/m3) (°) (°)
0.6 – 1.5 18 1.50 19 45 15
1.5 – 2.5 16 1.00 19 45 15
2.5 – 14.5 15 0.75* 19 45 15
*K0 = 1.0 for Priebe (1995)
Determination of Young's moduli

The Hardening Soil model is adopted by PLAXIS 3D Foundation to simulate the behaviour
of stone columns and the surrounding soil. This is an advanced constitutive model that
accounts for the stress path and stress level dependency of soil stiffness. In contrast the more
simplistic models adopted for the analytical design methods (i.e. Mohr-Coulomb) define the
stiffness response of each soil layer with a constant Young's modulus. Therefore, the
selection of appropriate stiffness parameters must take into consideration the stress path and
stress level experienced by the stone column and the surrounding soil. These factors are
discussed below:
(i) Consideration of the stress path

The Hardening Soil model accounts for the stress path dependency of soil by defining two
stiffness moduli: (i) E50, a secant modulus at 50% strength and (ii) Eur, an unload-reload
modulus. E50 is appropriate for primary loading of soils, while Eur is used for excavation or
tunnelling problems. The loading of an infinite grid of stone columns at the Bothkennar test
site is considered as a primary loading and E50 is chosen to represent the stiffness response of
the stone and surrounding soil in the analytical design methods.
(ii) Consideration of the stress level

Soils exhibit stress level dependency which results in an increasing stiffness with confining
pressure. The relationship between soil stiffness and confining pressure is defined in equation
3.3 using a power law. The confining pressure is taken as the minor or major principal stress
for triaxial or oedometer tests, respectively.
The parameter 'm' for the Bothkennar soil profile can be determined from one-dimensional
stiffness data presented by Nash et al. (1992b). The values presented by Nash et al. (1992b)
are converted into oedometric moduli (Eoed) using equation 4.1. The oedometric moduli are
then normalised by the effective overburden stress (σ'y,0), at the level which the samples were
taken from the soil, and plotted against σ'y,0 in Figure 4.15. In this case the major principal
stress is adopted for the confining pressure as the samples are subject to oedometric tests. It
can be seen for the lower Carse clay that Eoed/σ'y,0 is constant with an increasing σ'y,0, which
suggests that Eoed is directly proportional to σ'y,0. This indicates that m = 1 is appropriate for
the lower Carse clay layer, which is consistent with Brinkgreve & Broere (2006) for soft
soils. Insufficient data is available to determine the parameter 'm' for the crust and the upper
Carse clay and similar values to the lower Carse clay were adopted. A lower value of 'm' is
more appropriate for the granular material and m = 0.3 is used for stone columns, which is
similar to Gäb et al. (2008).
Normalised stiffness, Eoed / σ'y,0

0 20 40 60 80
0
Crust
Vertical effective stress, σ'y,0 (kPa) 20
Carse Clay
40
60 Lower
Carse Clay
80
100
120
140
Figure 4.15 - Variation of normalised oedometric moduli with vertical effective stress for Bothkennar test site
While the parameter 'm' is now determined for the Bothkennar soil profile and stone columns,
it is still necessary to define the confining pressure for stone columns in the field condition. A
FEA conducted by Balaam & Booker (1985) indicates that columns are in a triaxial state of
stress during loading. Therefore, the minor principal stress (σ3) defines the confining pressure
for the stone column. The surrounding soil also carries a proportion of the applied vertical
load and, consequently, is compressed in vertical direction. As with stone columns, the minor
principal stress is used to define the confining pressure for the surrounding soil.
The variation of minor principal stress with depth for stone columns and the surrounding soil
is presented in Figure 4.16 for an infinite grid of end-bearing columns (A/AC = 3.5). It can be
seen that the average minor stress is slightly higher in the stone column, which may be
explained as this material is stiffer than the surrounding soil and therefore takes more load.
The average minor stress for each layer is tabulated in Table 4.6 and the resulting Young’s
moduli are determined from equation 3.3.
Principal effective stress, σ3 (kPa)

0 20 40 60 80 100
0
σ'_3 - soil
2 σ'_3 - column
Depth (m)
6
10
12
14
Figure 4.16 - Distribution of minor principal stress σ3 in end-bearing stone column and surrounding soil
(A/Ac = 3.5; L = 13.9 m; Infinite grid of columns)
Table 4.6 - Average minor principal stresses (σ'3) and Young’s moduli (E) for Bothkennar soil profile
Soil Stone backfill
Depth σ'ref Eref (= E50) σ'3 E σ'ref Eref (= E50) σ'3 E
(m) (kPa) (kPa) (kPa) (kPa) (kPa) (kPa) (kPa) (kPa)
0.6 – 1.5 13 1068 28 2304 100 70000 35 51215
1.5 – 2.5 20 506 28 703 100 70000 32 49848
2.5 – 14.5 30 231 53 410 100 70000 57 59120
Comparison of n values from PLAXIS 3D Foundation with analytical design methods

The settlement performance of an infinite grid of columns predicted by PLAXIS 3D
Foundation is compared with the analytical design methods in Figure 4.17. It appears that
PLAXIS 3D Foundation captures the variation of settlement improvement factors with area
ratio quite well.
The design method proposed by Balaam & Booker (1981) significantly over-estimates the
settlement performance of stone columns, which is due to the simplified assumption of linear
elasticity for the stone backfill (Castro & Sagaseta, 2009). This design method is shown as
it forms the basis of Pulko & Majes (2005) and, also, as it demonstrates that linear
elasticity is not a valid assumption for stone columns.
19
Plaxis
PLAXIS3D 3D
Foundation
Foundation
17 Priebe:
Priebe:n0
n0 (1995)

Priebe:
Priebe:n2
n2 (1995)
(1995)
15 Pulko
Pulko&&Majes
Majes(2005)
(2005)
Balaam
Balaam&&Booker
Booker(1981)
(1981)
13
11
7
Priebe: n2
5 K0, lower Carse clay = 1.00
K0, lower Carse clay = 0.75
3
1
2 4 6 8 10 12 14 16
Area ratio, A/Ac
Figure 4.17 - Comparison of settlement improvement factors for an infinite grid of end-bearing columns with
analytical design methods
The design method proposed by Pulko & Majes (2005) is an extension of Balaam & Booker
(1981) which accounts for column yielding and, hence, predicts more realistic n values.
While Pulko & Majes (2005) predicts a similar settlement performance to PLAXIS 3D
Foundation for A/AC > 8.0, it tends to over-predict n values at low A/AC. This may be
attributed to the assumption of rigid-plastic behaviour for stone columns by Pulko & Majes
(2005), in contrast to elasto-plastic behaviour (i.e. Hardening Soil model) adopted in PLAXIS
3D Foundation. The main difference between these models is that the rigid-plastic model
does not account for elastic strains developed when columns are in a plastic state. Therefore,
the rigid-plastic model predicts less settlement, and thus higher n values, than the elasto-
plastic model. The extent of plasticity becomes more pronounced for an infinite grid of
columns at low A/AC, as columns are better confined and can carry a larger proportion of the
applied load. Consequently, the discrepancy between the rigid-plastic model and the elasto-
plastic model is most pronounced at low A/AC.
The basic (n0) and modified (n2) settlement improvement factors determined from Priebe
(1995) are also shown in Figure 4.17. It appears that the basic design curve (n0) under-
estimates the settlement performance of stone columns. Priebe (1995) modifies the basic
design curve to account for column compressibility and the effect of overburden stress, which
leads to a significant increase in n values. The modified n2 values agree favourably with the
numerical analysis, although the settlement performance is slightly over-estimated. This over-
estimation may be attributed to the assumption of K0 = 1 for the lower Carse clay layer,
which is higher than that adopted in Pulko & Majes (2005) and PLAXIS 3D Foundation.
Settlement improvement factors were re-calculated for Priebe (1995) n2 with K0 = 0.75 for
the lower Carse clay layer. It can be seen that the re-calculated n values are very close to
Pulko & Majes (2005) and PLAXIS 3D Foundation.
4.7 Summary of development and validation of soft soil profile

Field load tests on single columns, located within large groups and founded in a layered
estuarine deposit were simulated using PLAXIS 3D Foundation. The predicted load-
settlement curves were compared against field measurements and two independent
axisymmetric FEA. It was found that PLAXIS 3D Foundation predicts the load-
settlement curves closest to the field measurements, which reflects the ability of a three-
dimensional analysis to model the surrounding columns at discrete locations, rather than
approximating their presence by a cylindrical ring, as is the case with axisymmetric
analyses. This highlights the importance of accurately capturing the presence of
surrounding columns and gives confidence to the use of PLAXIS 3D Foundation for the
subsequent FEA.
A soil profile and set of material parameters were developed for the well characterised
Bothkennar test site, which consists of a soft uniform clay overlain by a stuff crust. This
soil profile is representative of many sites where the applicability of stone columns is of
growing interest and, therefore, forms the basis for the subsequent FEA. The adopted
soil profile and choice of material parameters were validated by simulating a field load
test, which was conducted on an unreinforced (i.e. no stone columns) rigid pad footing at
the Bothkennar test site.
The ability of PLAXIS 3D Foundation to capture the load deformation behaviour of

stone columns in the Bothkennar soil profile was also assessed. The settlement
performance of an infinite grid of columns was simulated and the settlement
performance was compared with a settlement database of field measurements and
various analytical design methods. The importance of modular ratio upon the settlement
performance of stone columns was highlighted and it was also shown that linear
elasticity is not valid assumption for the behaviour of stone columns.
93
Chapter 5
Results of FEA: Settlement performance,
deformational behaviour and stress
concentration ratios
5.1 Background
The main focus of this research is to develop a better understanding of the behaviour of small
groups of stone columns in soft soils. A series of FEA were conducted using PLAXIS 3D
Foundation to examine the influence of key design parameters upon the settlement
performance and deformational behaviour of small groups of stone columns. The key design
parameters and considerations investigated are as follows:
• Column length (L)
• Area ratio (A/AC)
• Column confinement
• Column arrangement
• Column position relative to footing edge
• Column compressibility
• Column strength
• Column installation effects
• The presence of a stiff crust
The design of foundations on soft soils is usually governed by settlement rather than bearing
capacity criteria, due to their high compressibility (Priebe, 1976). Therefore, the settlement
performance of stone columns at working load levels is of the upmost importance. The
degree of confinement provided to columns is dependent upon the number of columns
and the area ratio. The influence of column confinement is investigated by analysing the
settlement performance of different configurations of columns (i.e. different group sizes)
over a range of area ratios. The settlement performance of each configuration of columns
is examined for various column lengths, which allows the relationship between column
Results of FEA: Settlement performance, deformational behaviour and stress concentration ratios 94
length, area ratio and column confinement to be determined. Finally, the influence of the
key design parameters and practical considerations outlined above is investigated for a
select number of column configurations, which were specifically chosen to cover a wide
range of area ratios.
The modes of deformation for small groups of stone columns are examined with the aid of
new parameters called compression and punching ratios. In previous studies the modes of
deformation have usually been determined at ultimate conditions; however, this study
examines the modes of deformation at typical working loads. This allows a direct link to be
made between the deformational behaviour and settlement performance of stone columns.
The stress concentration ratio is an important parameter which compares the vertical stress in
columns to the vertical stress in the surrounding soil. For this research the vertical stress used
to determine this parameter is measured beneath the base of footings (i.e. at the top of stone
columns). Stress concentration ratios are used in analytical design methods to determine the
settlement performance and the potential of stone columns to stabilise slopes; the variation of
this ratio with the modes of deformation and the key design parameters is investigated.
5.2 Details of FEA

The various configurations of columns analysed are outlined in the following section. The
Bothkennar soil profile, which was validated in the previous chapter, was adopted for all of
the subsequent FEA. The footings are 0.6 m thick and founded 0.6 m below ground level.
The column diameter is normally not a variable in design as the poker is of fixed diameter
(430 mm is the most commonly used size) and the constructed column diameter depends on
the properties of the surrounding soil. A diameter of 0.6 m is typical of stone columns
constructed in soft soils using the bottom feed system. The column length was increased in 1
m increments (from 0 m, i.e. no columns, to 13.9 m, i.e. the base of the Carse clay profile
adopted) to examine the effect of floating and end-bearing columns (Figure 5.1).
Area ratio and column confinement

The degree of confinement provided to individual stone columns depends on A/AC and the
number of columns within a group. The influence of column confinement was determined by
analysing various configurations of columns such as single columns, 2×2, 3×3, 4×4 groups
Figure 5.1 - Layout of columns in Bothkennar soil profile
and infinite grids of columns (see Figure 5.2). Single columns and infinite grids provide a
useful frame of reference for the other configurations of columns as these two cases
correspond to zero and full column confinement, respectively. Columns are positioned on a
square grid and spaced at 1.0, 1.5 and 2.0 m, which corresponds to A/AC of 3.5, 8.0 and 14.1,
respectively. This is a typical range of A/AC for small loaded areas, as can be seen in the
settlement database developed by McCabe et al. (2009); columns at A/AC < 3.5 would not be
practicable or economic to construct and columns at A/AC > 14.1 would typically be used to
support wide area loadings such as embankments, which can tolerate larger settlements. The
edge of the footing is located at a distance of half the column spacing (0.5×s) from the
centreline of the outer row of columns; the influence of the footing overhang is examined in a
subsequent parametric study.
Figure 5.2 - Column configurations to examine influence of column confinement
Column arrangement
The column arrangements adopted to support a 3 m square pad footing are shown in Figure
5.3. The arrangements of 4, 5 and 9 columns are carefully chosen to allow the benefit of
individual columns (i.e. centre, edge and corner) to be assessed. A comparison of the 4 and 5
column groups allows the benefit of an extra central column to be assessed, while a
comparison of the 5 and 9 column groups allows the benefit of four edge columns to be
assessed. A group of 7 columns may also be adopted to support small pad footings in practice
and this configuration was compared with a group of 9 columns.
Figure 5.3 - Column configurations to examine influence of column arrangement
Column position
The effect of footing overhang and column position was examined for a 3 m square pad
footing by positioning a 2×2 group of columns progressively closer to the footing edge
(Figure 5.4). It is well known from elastic theory that the distribution of vertical stress
beneath rigid footings is non-uniform and high vertical stress develops beneath the edge of
footings. The column spacings investigated were 1.0, 1.5 and 2.0 m.
Figure 5.4 - Column configurations to examine influence of column position relative to the footing edge
Column compressibility and strength

The influence of column stiffness and strength upon the settlement performance and
deformational behaviour of groups of 4, 5 and 9 columns beneath a 3 m square footing (see
Figure 5.5) is investigated. The benchmark parameters adopted for the column stiffness (E50
= 70 MPa; Eur = 210 MPa) and strength (φ = 45°; ψ = 15°) are developed in Section 4.6.1.
• The stiffness parameters are varied from E50 = 30–70 MPa on the basis of a range
proposed by Barksdale & Bachus (1983). The unload-reload Young’s modulus is defined
by the relationship Eur = 3E50 (Brinkgreve & Broere, 2006).
• The angle of friction for the stone backfill was varied from φ = 40–50° and the angle of
dilatancy was defined as ψ = φ – 15° (Bolton, 1986).
Figure 5.5 - Column configurations for parametric study of column compressibility and strength
Column installation effects

It was shown in Section 3.4.3 that the most appropriate method to simulate column
installation effects in PLAXIS 3D Foundation is to increase the coefficient of lateral pressure
(K0) in the surrounding soil. A review of numerical analyses which adopted this methodology
(see Section 2.4.3) suggests that K0 is increased in the range 0.75–1.50, with an average of
1.00. As the in situ stress state in the upper layers at the Bothkennar test site is quite high (K0
> 1.0), only the stress state of the lower Carse clay layer is modified to investigate the effects
of column installation. A range of K0 values (0.75, 1.00 and 1.25) were examined for the
lower Carse clay.
Stiff crust layer

The soil profile adopted for the Bothkennar test site is modified to investigate the effect of the
lower Carse clay and the stiff crust (Figure 5.6). The settlement performance, deformational
behaviour and stress concentration ratio for various column arrangements outlined in Figure
5.5 are investigated for the various profiles shown in Figure 5.6.
Figure 5.6 - Various profiles adopted to investigate the effect of the lower Carse clay (Profile 2) and the stiff
crust (Profile 3)
5.3 Settlement analysis

As described earlier, the ultimate bearing capacity (qult) recorded by Jardine et al. (1995) for
an unreinforced rigid footing at the Bothkennar test site was 138 kPa (see Section 4.5.3).
Applying typical factors of safety (2.5–3.0) yields allowable bearing pressures in the range of
46–55 kPa. As a result, the settlement performance of stone columns was measured at 50 kPa,
which was deemed a typical working load. The extent of settlement improvement is
quantified by means of a settlement improvement factor (n), defined as the ratio of the
settlement of an untreated footing to the settlement of a treated footing:
988909.8 7.89189? 8-.5
.=
988909.8 89189? 8-.5
5.3.1 Settlement performance of infinite grids of stone columns

The influence of column length upon the settlement performance of an infinite grid of stone
columns is shown in Figure 5.7. A negligible reduction in footing settlement is observed in
Figure 5.7(i) for columns shorter than L = 2 m. This may be attributable to the stiff upper
layers at the Bothkennar test site which extend to 1.9 m beneath the footing base, as installing
stone columns in an already competent layer will not reduce footing settlement. However, a
significant reduction in footing settlement is observed with increasing column length
thereafter which yields high settlement improvement factors (Figure 5.7(ii)). It can also be
seen that the increase in settlement improvement factors with column length becomes more
pronounced at low area ratios. This suggests that columns are at their most effective when
closely-spaced.
Footing settlement (mm) Settlement improvement factor, n
0 150 300 450 600 1 3 5 7 9 11
0 0
A/Ac = 3.5
2 2 A/Ac = 8.0
A/Ac = 14.1
4 4
6 6
8 8
10 10
A/Ac = 3.5
12 A/Ac = 8.0 12
A/Ac = 14.1
14 14
(i) (ii)
Figure 5.7 - Variation of (i) footing settlement and (ii) settlement improvement factor with column length for an
infinite grid of columns
5.3.2 Settlement performance of small groups of stone columns

The influence of column length and column confinement upon the settlement performance of
small groups of stone columns is shown in Figure 5.8(a). As with an infinite grid of columns,
a negligible increase in settlement improvement factors is observed for all columns shorter
than L = 2 m.
It can be seen in Figure 5.8(a-i) that settlement improvement factors for columns at low area
ratios (i.e. A/AC = 3.5) increase continuously with column length. A jump in settlement
improvement factors is observed when moving from floating to end-bearing columns (i.e. L =
13→13.9 m) at low area ratios. The improved settlement performance may be explained as
the proportion of applied load transferred to the base of columns is supported by a rigid
stratum in the case of end-bearing columns, rather than a compressible stratum in the case of
floating columns. The jump in settlement improvement factors suggests that closely-spaced
columns transfer a significant proportion of the applied load to their base. The magnitude of
the jump increases for larger groups of columns, which indicates that the proportion of
applied load transferred to the base of columns increases for larger groups of columns. This
reflects the enhanced lateral confinement associated with larger groups of columns.
The variation of settlement improvement factors with column length for various groups of
columns at A/AC of 8.0 and 14.1 is shown in Figures 5.8(a-ii) and 5.8(a-iii), respectively.
Similar to columns at low area ratios, a continuous increase in settlement improvement
factors is observed with increasing column length for all columns. However, the rate of
increase in settlement improvement factors reduces for column lengths longer than L = 5 m.
This suggests that long widely-spaced columns transfer a smaller proportion of the load to the
base of columns and may be deforming closer to the ground surface.
It is clearly shown in Figure 5.8(a) that the enhanced levels of lateral confinement associated
with larger groups of columns yields higher settlement improvement factors. However, it
appears that the variation of settlement improvement factors with column length is dictated
by area ratio rather than the number of columns, as a continuous increase in settlement
improvement factors is observed with column length at low area ratios (i.e. A/AC = 3.5),
while the increase in settlement improvement factors with column length tends to reduce
beyond L = 5 m at higher area ratios (i.e. A/AC = 8.0 and 14.1).
The influence of the footing width (B) upon the settlement performance of small groups of
columns is shown in Figure 5.8(b). For the column configurations analysed, it appears that
A/AC governs the variation of settlement improvement factors with column length. The
classical Boussinesq (1885) solution for vertical stress distribution beneath an unreinforced
square footing suggests that the stress increment applied to the footing is no longer perceived
at L/B = 2 and so improving the settlement characteristics of the soil below this level, by
installing stone columns, should not enhance the settlement performance. However, it appears
that this significant depth (L/B = 2) does not influence settlement improvement factors of
columns at low area ratios, as a continuous increase in settlement improvement factors is
observed with increasing column length beyond this level.
5.3.3 Comparison of infinite grids with small groups of stone columns

The influence of A/AC and column confinement upon the settlement performance of infinite
grids and small groups of end-bearing columns is shown in Figure 5.9. It is clear that A/AC
has a significant influence on the settlement performance of stone columns, as a decrease in
area ratio yields very high settlement improvement factors. The effect of column confinement
is also evident as a larger group of stone columns yields higher settlement improvement
factors, especially at low area ratios.
Settlement improvement factors for small groups of end-bearing stone columns (n) are
normalised by settlement improvement factors for an infinite grid of end-bearing stone
columns (nuc) and presented in Figure 5.10. The n/nuc ratio allows the influence of column
confinement upon small groups of columns to be assessed, as an infinite grid of columns has
full confinement and is an ideal benchmark. An increase in n/nuc is observed with both an
increasing number of columns and an increasing area ratio. It can be seen in Figure 5.10 that
n/nuc increases with area ratio, from 0.19 to 0.72 for single columns and from 0.45 to 0.82 for
a 4×4 group of columns. The increase in n/nuc for larger groups of columns reflects the
increased level of lateral confinement. However, the increase in n/nuc with increasing area
ratio may be explained as the influence of lateral confinement upon settlement improvement
factors reduces at high area ratios. Consequently, the difference between settlement
improvement factors for small groups (n) and infinite grids (nuc) of stone columns reduces at
high area ratios, which yields an increase in n/nuc.
Settlement improvement factor, n Settlement improvement factor, n Settlement improvement factor, n

1 2 3 4 5 6 1.0 1.3 1.6 1.9 2.2 2.5 1.0 1.1 1.2 1.3 1.4 1.5
0 0 0
A/AC = 3.5 A/AC = 8.0 A/AC = 14.1
2 1 col 2 1 col 2 1 col
2×2 2×2 2×2

3×3 3×3 3×3
4 4×4 4 4×4 4 4×4
6 6 6
8 8 8
10 10 10
12 12 12
14 14 14
(a-i) (a-ii) (a-iii)
Settlement improvement factor, n Settlement improvement factor, n Settlement improvement factor, n

1 2 3 4 5 6 1.0 1.3 1.6 1.9 2.2 2.5 1.0 1.1 1.2 1.3 1.4 1.5
0 0 0
Normalised column length, L/B

2 2 2
4 4 4
6 6 6
8 8 8
10 A/AC = 3.5 10 A/AC = 8.0 10 A/AC = 14.1

1 col - B = 1 m 1 col - B = 1.5 m 1 col - B = 2 m
2×2 - B = 2 m 2×2 - B = 3.0 m 2×2 - B = 4 m
12 3×3 - B = 3 m 12 12 3×3 - B = 6 m
3×3 - B = 4.5 m
4×4 - B = 4 m 4×4 - B = 6.0 m 4×4 - B = 8 m
14 14 14
(b-i) (b-ii) (b-iii)
Figure 5.8 - Variation of settlement improvement factor with (a) column length (L) and (b) normalised column length (L/B) for small groups of columns, spaced at area ratios
of (i) 3.5, (ii) 8.0 and (iii) 14.1
The relationship between the settlement of small groups of columns and infinite grids (s/suc)
is shown in Figure 5.11. Priebe (1995) also developed a design chart (Figure 2.24) which
relates the settlement of small groups of columns to infinite grids. However, Priebe (1995)
does not directly account for the influence of footing area and relies on A/AC to compensate
for a decrease in s/suc with an increasing footing area. It can be seen in Figure 5.11 that both
the area ratio and the number of columns have a significant influence on the settlement
performance of small groups of columns. For similar area ratios, it can be seen that s/suc
increases for a larger group of columns. This may be attributed to an increase in footing size
for larger groups of columns. It can also be seen that the increase in s/suc for larger groups of
columns is more pronounced at high area ratios (A/AC = 14.1). This highlights the positive
confining effects of closely-spaced columns.
11
1 col
Settlement improvement
2×2
9
3×3
4×4
factor, n
7 Unit Cell
Increasing
confinement
5
1
3 5 7 9 11 13 15
Area ratio, A/Ac
Figure 5.9 - Settlement improvement factors for small groups of end-bearing columns
1.0
Normalised n value, n/nuc
0.8
0.6
0.4
1 col
2×2
0.2
3×3
4×4
0.0
3 5 7 9 11 13 15
Area ratio, A/AC
Figure 5.10 - Normalised settlement improvement factors for small groups of end-bearing columns
0.5
Settlement ratio, s/suc
0.4
0.3
0.2
A/Ac=3.5
0.1 A/Ac=8.0
A/Ac=14.1
0.0
0 5 10 15 20
No. of columns
Figure 5.11 - Normalised settlement for small groups of end-bearing columns
5.3.4 Influence of column arrangement

The settlement performance of the various configurations of columns supporting a 3 m square
footing (see Figure 5.3) is shown in Figure 5.12(i). As expected, it can be seen that settlement
improvement factors increase with both column length and the number of columns. It appears
that columns at low area ratios continue to yield higher increases in settlement improvement
factors with increasing column length, which is consistent with earlier findings. A continuous
increase in settlement improvement factors is observed for the 7 and 9 column groups and a
jump in settlement improvement factors is again observed when moving from floating to end-
bearing columns (i.e. L = 13→13.9 m). This suggests that the 7 and 9 column groups transfer
a significant proportion of the applied load to the base of columns.
The arrangement of columns can be tailored to achieve a specific settlement performance, i.e.
long columns at high area ratios or short columns at low area ratios. The volume of stone
required for each configuration of columns is shown in Figure 5.12(ii). It appears that
installing an extra centre column in a 2×2 group is only beneficial for long columns. It also
appears that short columns at low area ratios are the most efficient configuration to achieve
high settlement improvement factors. However, it must be noted that the volume of stone
does not entirely reflect the overall cost of construction. Other factors such as the time taken
to form stone columns must also be taken in account, i.e. a large number of short columns
will take longer to construct than a small number of long columns due to the time taken to
remove the poker. The longer length of time taken to construct a larger number of columns
translates itself as an additional cost and must also be taken into consideration.
Settlement improvement factor, n Settlement improvement factor, n

1.0 2.0 3.0 4.0 1.0 2.0 3.0 4.0
0 0
B=3m B=3m
4 cols 5 4 cols
2
5 cols 5 cols
10
Volume of stone (m3)
4 7 cols 7 cols
9 cols 9 cols
15
6
20
8
25
10
30
12 35 Each data point represents

lengthening the column by 1 m
14 40
(i) (ii)
Figure 5.12 - Variation of settlement improvement factors with (i) column length and (ii) volume of stone for
various arrangements of stone columns
5.3.5 Influence of column position

The influence of column position relative to the edge of footings (see Figure 5.4) upon
settlement improvement factors is shown in Figure 5.13. It appears that small benefits can be
gained by keeping columns closer to the footing edge. This may be attributed to high stress
concentrations which occur beneath the edge of rigid footings. Therefore, columns placed in
this zone have the potential to absorb more load and develop higher settlement improvement
factors. A FEA by Wehr (2004) to examine the group behaviour of stone columns
demonstrates that shear zones develop at the edges of a pad footing and extend to a depth
beneath the centre of the footing. Positioning columns closer to the edge of footings, and thus
closer to these shear zones, may also explain the enhanced settlement performance.

1.0 1.2 1.4 1.6 1.8
0
4 cols - B = 3 m
2 s = 1.0 m
s = 1.5 m
4 s = 2.0 m
10
12
14
Figure 5.13 - Influence of column position upon settlement improvement factors
5.3.6 Influence of column compressibility

The influence of column compressibility upon the settlement performance of various column
arrangements supporting a 3 m square footing (see Figure 5.5) is shown in Figure 5.14. As
expected, settlement improvement factors increase with column stiffness (E50, col). This is
consistent with Domingues et al. (2007a) who conducted a FEA on large groups of stone
columns (A/AC = 5.3) supporting an embankment. The increase in settlement improvement
factors with increasing column stiffness for end-bearing columns is shown in Table 5.1 and it
can be seen that the effect of column stiffness is more significant for the group of 9 columns.
The distribution of plastic points along a cross-section of 4, 5 and 9 columns is shown in
Figure 5.15. The Mohr-Coulomb plastic points represent soil elements in a plastic stress state,
i.e. which lie on the Mohr-Coulomb failure line. It can be seen that the extent of plasticity
reduces for larger groups of columns. This may be explained as the average applied load
carried per column reduces for larger groups. This suggests that larger groups of columns are
in more of an elastic state and hence are more sensitive to changes in elastic stiffness moduli.
1.0 2.0 3.0 4.0
0
4 cols - s = 2.0 m
5 cols - s = 1.0 m
2 9 cols - s = 1.0 m
E50, col = 30 MPa
E50, col = 50 MPa

4
E50, col = 70 MPa
10
12
14
Figure 5.14 - Influence of column compressibility upon settlement improvement factors
Table 5.1 - Influence of column compressibility upon settlement improvement factors for end-bearing columns
Settlement improvement factors % reduction
No. of cols Column compressibility, E50 (MPa) Column compressibility, E50 (MPa)
30 50 70 30 50 70
4 1.57 1.64 1.69 -7 -3 -
5 1.76 1.89 1.94 -9 -2 -
9 2.60 3.23 3.68 -29 -12 -
Crust
Upper Carse clay
Lower Carse clay
(i) (ii) (iii)

Figure 5.15 - Distribution of plastic points (Mohr-Coulomb) within a cross-section of (i) 4, (ii) 5 and (iii) 9
columns beneath a 3 m square footing (L = 6 m)
5.3.7 Influence of column strength

The influence of column strength upon the settlement performance of various column
configurations supporting a 3 m square footing (see Figure 5.5) is shown in Figure 5.16 for
friction angles in the range φ = 40–50°. It can be seen that settlement improvement factors
increase with column strength. The increase in settlement improvement factors with
increasing column strength from the benchmark strength of φ = 45° to 50° for end-bearing
columns is shown in Table 5.2. It can be seen that an increase in column strength has the
smallest effect on the group of 9 columns. This may be explained as the angle of internal
friction only becomes mobilised once columns are in a plastic state. Therefore, increasing the
angle of internal friction (i.e. increasing column strength) will have a larger influence on
column configurations which exhibit the largest degree of plasticity. Similarly reducing the
angle of internal friction from the benchmark strength of φ = 45° to 40° increases the extent
of plasticity within columns and, consequently, has the smallest effect upon columns which
exhibit the largest degree of plasticity.
1.0 2.0 3.0 4.0 5.0
0
4 cols - s = 2.0 m
5 cols - s = 1.0 m
2 9 cols - s = 1.0 m
φcol = 40°
φcol = 45°
4
φcol = 50°
10
12
14
Figure 5.16 - Influence of column strength upon settlement improvement factors
Table 5.2 - Influence of column strength upon end-bearing settlement improvement factors
Settlement improvement factors % reduction / increase
No. of cols Angle of internal friction, φ (°) Angle of internal friction, φ (°)
40 45 50 40 45 50
4 1.43 1.69 2.09 -15 - 24
5 1.57 1.94 2.49 -19 - 29
9 2.46 3.68 4.32 -33 - 17
5.3.8 Influence of column installation effects

The effects of column installation are investigated by increasing the coefficient of lateral
earth pressure (K0) in the lower Carse clay layer. The influence of K0 upon the settlement
performance of various configurations of columns (see Figure 5.5) is shown in Figure 5.17. It
can be seen that increasing K0 enhances the settlement performance of all column
configurations, especially for groups of 4 and 5 columns. The smallest increase in settlement
improvement factors is observed for the group of 9 columns (Table 5.3). This suggests that
the mode of deformation for the group of 9 columns is less dependent on the lateral support
provided by the surrounding soil than the groups of 4 and 5 columns. In fact, a slight decrease
in settlement improvement factors is observed when increasing K0 from 1.0 to 1.25 for a
group of 9 columns. This may be explained as the untreated footing benefits more from an
increase in K0 than the settlement of a group of columns and, therefore, settlement
improvement factors decrease slightly. It will be shown later on in Section 5.4.3 that columns
at low area ratios tend to punch, while columns at high area ratios tend to bulge and are
therefore more dependent on the lateral support provided by the surrounding soil.

1.0 2.0 3.0 4.0 5.0
0
4 cols - s=2.0 m
5 cols - s=1.0 m
2 9 cols - s=1.0 m
K0 = 0.75
4 K0 = 1.00
K0 = 1.25
6
10
12
14
Figure 5.17 - Influence of coefficient of lateral earth pressure upon settlement improvement factors
Table 5.3 - Influence of coefficient of lateral earth pressure upon end-bearing settlement improvement factors
Settlement improvement factors % reduction / increase
No. of cols Coefficient of lateral earth pressure, K0 Coefficient of lateral earth pressure, K0
0.75 1.00 1.25 0.75 1.00 1.25
4 1.69 1.88 2.10 - 11 24
5 1.94 2.24 2.49 - 16 29
9 3.68 4.00 3.91 - 9 6
5.3.9 Influence of crust

The influence of the lower Carse clay and the stiff crust at the Bothkennar test site upon the
settlement performance of various arrangements of columns (see Figure 5.5) supporting a 3 m
square footing is shown in Figure 5.18. Profile 1 is the standard Bothkennar soil profile and
settlement improvement factors for this profile in Figure 5.18 were presented earlier in Figure
5.12. The differences between Profiles 1, 2 and 3 are shown in Figure 5.6.
Settlement performance
Profile 2 consists of a stiff crust which is underlain by upper Carse clay. This soil profile is
stiffer at depth than Profile 1, as it does not contain the lower Carse clay layer. This is
reflected in Figure 5.18(i) as the untreated (L = 0 m) and treated settlements are lower than
Profile 1. Profile 3 does not contain the stiff crust and consists entirely of upper Carse clay.
While it is softer than Profile 1 near the surface, it is stiffer at depth and this is again reflected
in Figure 5.18(i) as the treated settlements are lower than Profile 1. It is interesting to note
that the untreated settlements (L = 0 m) for Profiles 1 and 3 are similar, which suggests that
the increase in settlement due to the absence of the stiff crust is offset by the increased
stiffness at depth, as the lower Carse clay is replaced by upper Carse clay.
Settlement improvement factors

The importance of the soil stiffness at depth is evident in the variation of settlement
improvement factors with column length (Figure 5.18(ii)). It can be seen that settlement
improvement factors are lowest for Profile 1. This may be related to the weak lower Carse
clay which is not present in Profiles 2 and 3.
It can also be seen in Figure 5.18(ii) that settlement improvement factors increase
significantly with column length up to L = 3 m in Profile 3. However, the rate of increase in
settlement improvement factors with column length slows down thereafter. This is most
pronounced for smaller groups of columns (i.e. higher A/AC) and suggests that a critical
length may exist for columns formed in Profile 3. This profile is similar to that used in
laboratory studies described in the literature review (see Sections 2.2 and 2.3) and may
explain why authors have postulated critical lengths in the past. However, it can be seen that
critical lengths are less well defined for more realistic soil profiles. Therefore, this highlights
the benefits of the FEM which accurately captures the influence of the stiff crust.

0 20 40 60 1.0 1.5 2.0 2.5
0 0
2 2

4 4
6 6
8 8
10 10
4 cols - 2.00 m 4 cols - 2.00 m
Profile 1 Profile 1
12 12
Profile 2 Profile 2
Profile 3 Profile 3
14 14
(a-i) (a-ii)

0 20 40 60 1.0 1.5 2.0 2.5 3.0
0 0
2 2
4 4
6 6
8 8
10 10
5 cols - 1.00 m 5 cols - 1.00 m
Profile 1 Profile 1
12 12
Profile 2 Profile 2
Profile 3 Profile 3
14 14
(b-i) (b-ii)

0 20 40 60 1.0 2.0 3.0 4.0 5.0
0 0
2 2
4 4
6 6
8 8
10 10
9 cols - 1.00 m 9 cols - 1.00 m
Profile 1 Profile 1
12 12
Profile 2 Profile 2
Profile 3 Profile 3
14 14
(c-i) (c-ii)
Figure 5.18 - Influence of the lower Carse clay and the stiff crust upon (i) settlement and (ii) settlement
improvement factors for (a) 4, (b) 5 and (c) 9 columns beneath a 3 m square footing
5.4 Deformational behaviour of stone columns

5.4.1 Definitions of punching and compression ratios
It was shown by Muir Wood et al. (2000) that groups of stone columns can exhibit many
modes of deformation such as punching, bulging, shearing and bending. New parameters,
referred to as punching and compression ratios, were defined in this thesis to help identify the
different modes of deformation.
Punching ratio
Depending on the adopted configuration of columns, a proportion of the applied load may be
transferred to the base of the columns which results in columns punching vertically into the
underlying soil. In order to compare the punching of columns beneath different footing sizes,
the displacement at the base of columns is normalised by the displacement of the footing:
7_ − 7QR
[7.ℎ-.5 18- =
7RS¡ (5.1)
where ucol = displacement at base of columns

usoil = displacement of soil surrounding the base of columns
ufooting = displacement at surface of footings
The displacement of the soil surrounding the base of columns (usoil) is determined by
averaging soil displacements within a zone surrounding each column. The size of the zone
adopted to determine usoil will influence the magnitude of the punching ratio, as soil
displacement decreases with distance from columns. Two zones, shown in Figure 5.19, were
compared in order to find a consistent method of calculating the punching ratio. The first
zone (i) is a 1 m square zone surrounding the column and the second zone (ii) is also a square
zone, whose width is equal to the column spacing. The size of the second zone will always be
greater than or equal to the first zone as column spacing ranges from 1.0–2.0 m in this study.
Figure 5.19 - Displacement at surface of footing, base of columns and soil surrounding base of columns
The variation of punching ratio - determined by methods (i) and (ii) - with length for an
infinite grid of columns is shown in Figure 5.20. The variation of punching ratio with length
for A/AC = 3.5 is not shown as this corresponds to a column spacing of 1.0 m; the zones for
methods (i) and (ii) coincide at this spacing and no difference will be observed. It can be seen
in Figure 5.20 that methods (i) and (ii) predict a similar variation of punching ratio with
length. It is also clear that the punching ratio is always higher for method (ii). This may be
explained as the displacement of the surrounding soil is averaged over a larger area for the
method (ii). This leads to a lower average displacement of the soil and, consequently, higher
punching ratios, as soil displacement decreases with distance from the column. The punching
ratio is determined using the method (i) for the subsequent analyses, as the zone in which the
soil displacement is averaged is the same size for all column spacings and thus allows for a
consistent analysis.
Punching ratio, (ucol-usoil)/ufooting

0.00 0.05 0.10 0.15 0.20 0.25
0
2
Unit cell
10
(i) - A/Ac = 8.0
(i) - A/Ac = 14.1
12 (ii) - A/Ac = 8.0
(ii) - A/Ac = 14.1
14
Figure 5.20 - Variation of punching ratios with length for an infinite grid of columns, with usoil measured in (i)
a 1 m square zone and (ii) a square zone, whose width is equal to the column spacing
Compression ratio
The compression ratio is an important indicator of the mode of deformation and is defined as
the proportion of surface settlement transferred to the column base. Low compression ratios
suggest that columns are transferring the load to depth, rather than deforming along their
length. One limitation of the compression ratio is that is does not differentiate between axial
and radial deformation. However, axial deformation due to column compressibility results in
a steady increase in compression ratios, while radial deformation due to column bulging can
be identified by a rapid increase in compression ratios.
7RS¡ − 7_
0f9-. 18- =
7RS¡ (5.2)
5.4.2 Deformational behaviour of an infinite grid of stone columns

The variation of punching ratio with area ratio and column length for an infinite grid of
columns is shown in Figure 5.21(i). It can be seen that the punching ratio increases with
column length for all columns shorter than L = 3 m. It can also be seen in Figure 5.21(ii) that
the increase in punching ratio is coupled with low compression ratios. This suggests that
columns do not deform along their length and punching is therefore the dominant mode of
deformation for short columns.
Punching ratio, (ucol-usoil)/ufooting Compression ratio, (ufooting-ucol)/ufooting

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
Unit cell
2 2 A/Ac = 3.5
A/Ac = 8.0

4 4
A/Ac = 14.1
6 6
8 8
Unit cell
10 10
A/Ac = 3.5
12 A/Ac = 8.0 12
A/Ac = 14.1
14 14
(i) (ii)
Figure 5.21 - Variation of (i) punching and (ii) compression ratios with column length for an infinite grid of
columns
It can be seen in Figure 5.21(i) that the punching ratio increases with column length up to a
maximum at L = 3 m. This increase may be attributed to a reduction in the stiffness of soil
layers in the upper 3 m of the Bothkennar test site, i.e. the crust is the stiffest layer, followed
by the upper Carse clay and then the lower Carse clay (see CC values in Table 4.2).
Therefore, as column length increases, the columns bases are founded in more compressible
soil strata which results in higher punching ratios (Figure 5.22). It is interesting to note that
the maximum punching ratio occurs at L = 3 m, despite the fact that the base of 2 m long
columns coincides with the weakest part of the soil profile, i.e. the top of the lower Carse clay
layer. It seems intuitive that the maximum punching ratio would occur when the column
bases are formed in the weakest part of the soil profile. However, it can be seen in Figure
5.22(ii) that the base of 2 m long columns extends just beyond the bottom of the upper Carse
clay layer, which reduces the displacement of the soil at the top of the lower Carse clay layer,
thus, reducing the punching ratio.
(i) (ii) (iii)

Figure 5.22 - Deformation of an infinite grid of floating stone columns at an area ratio A/AC = 14.1 for column
lengths of (i) L = 1 m, (ii) L = 2 m and (iii) L = 3 m
It is also evident in Figure 5.21(i) that the punching ratio increases with increasing area ratio
and this is most pronounced at L = 3 m. The deformation of an infinite grid of 3 m long
columns at various area ratios is shown in Figure 5.23. It can be seen in Figure 5.23(i) that
columns at low area ratios act with the surrounding soil and punch as a single entity into the
underlying soil. This mode of deformation is referred to as 'block' failure and is a form of
column punching. It occurs for large groups of closely-spaced columns and is characterised
by low punching and low compression ratios. High punching ratios usually denote punching
failure; however, punching ratios are low for 'block' failure as columns deform with the
surrounding soil. A similar mode of deformation is observed by Black (2006) for a group of
three columns at low area ratios (A/AC = 2.5 and 3.6). It can be seen in Figures 5.23(ii) and
5.23(iii) that column interaction decreases at higher area ratios, as columns tend to act more
individually and punch into the underlying soil.
(i) (ii) (iii)

Figure 5.23 - Deformation of an infinite grid of 3 m long columns at area ratios of (i) 3.5, (ii) 8.0 and (iii) 14.1
While it can be seen in Figure 5.21 that columns at all area ratios exhibit similar variations of
punching and compression ratios with column length up to L = 3 m, a marked difference is
observed with area ratio thereafter:
• Columns at low area ratios continue to exhibit relatively low and constant punching
ratios for column lengths L > 3 m. Compression ratios also remain relatively low, which
indicates that columns do not deform significantly along their length. This suggests that
'block' failure remains the dominant mode of deformation with increasing length for an
infinite grid of columns at low area ratios. A slight increase in the punching ratio is
observed for column lengths L > 11 m.
• Columns at high area ratios exhibit a decrease in punching ratio and increase in
compression ratio for column lengths L > 3 m. The increase in compression ratios
indicates that columns are deforming along their length and may be bulging closer to the
ground surface. Columns which are bulging cannot transfer the applied load to the base
of columns which consequently results in low punching ratios.
5.4.3 Deformational behaviour of small groups of stone columns

Single columns and 2×2 groups
The influence of area ratio and column length upon punching ratios for single columns and
2×2 groups is shown in Figures 5.24(i) and 5.25(i), respectively. It can be seen that punching
ratios increase with column length up to a maximum at L = 3 m, which is similar to an
infinite grid of columns. The increase in punching ratios is again coupled with relatively low
compression ratios as shown in Figures 5.24(ii) and 5.25(ii) for single columns and 2×2
groups, respectively. Therefore, it appears that punching is the dominant mode of
deformation for short columns.
It can be seen in Figures 5.24(i) and 5.25(i) that an increase in area ratio leads to a decrease in
punching ratio for 1 m long single columns and 2×2 groups of columns, respectively. This
may be explained as the soil surrounding at the base of 1 m long columns is significantly
influenced by the footing overhead. An increase in area ratio for groups of columns
corresponds to an increase in footing width, and wider footings induce more displacement in
the soil at z = 1 m (z = depth beneath footing). As a consequence, this reduces the differential
displacement between the base of columns and the surrounding soil, which results in lower
punching ratios.

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
1 column
2 2 A/Ac = 3.5
A/Ac = 8.0

4 4
A/Ac = 14.1
6 6
8 8
1 column
10 10
A/Ac = 3.5
12 A/Ac = 8.0 12
A/Ac = 14.1
14 14
(i) (ii)
Figure 5.24 - Variation of (i) punching and (ii) compression ratios with column length for single columns

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
2×2 cols
2 2 A/Ac = 3.5
A/Ac = 8.0
4 4
A/Ac = 14.1
6 6
8 8
2×2 cols
10 10
A/Ac = 3.5
12 A/Ac = 8.0 12
A/Ac = 14.1
14 14
(i) (ii)
Figure 5.25 - Variation of (i) punching and (ii) compression ratios with column length for 2×2 groups of
columns
However, it can also be seen that an increase in area ratio leads to an increase in punching
ratios for L = 2, 3 and 4 m. It is established that punching is the dominant mode of
deformation for short stone columns, which implies that columns develop shear stress and
end-bearing pressure along the side and at the base of columns, respectively. An increase in
area ratio corresponds to an increased footing width for groups of columns. This implies that
the total load taken per column also increases with area ratio. Columns of similar length
which are 'punching' have the capacity to carry similar shear force along their sides.
Therefore, an increase in the total load taken per column increases the portion of the load
transferred to the base of columns, which consequently increases the punching ratio.
A reduction in the punching ratio is coupled with an increase in compression ratios for single
columns and 2×2 groups which are longer than L = 3 m. This suggests that columns are
deforming along their length and that bulging is the dominant mode of deformation. It is
interesting to note that the reduction in punching ratios and increase in compression ratios
becomes more pronounced with increasing area ratio. This is consistent with the findings
from an infinite grid of columns and indicates that column bulging becomes more
pronounced with increasing area ratio.
3×3 and 4×4 groups

The variation of punching ratio with column length for 3×3 and 4×4 groups of columns is
shown in Figures 5.26(i) and 5.27(i), respectively. Punching ratios are observed to increase
with length up to a maximum at L = 3 m, which is again similar to findings from single
columns, 2×2 groups and infinite grids of columns. The increase in punching ratios is
coupled with low compression ratios, as shown in Figures 5.26(ii) and 5.27(ii) for 3×3 and
4×4 groups, respectively. This indicates that columns do not deform along their length and
are transferring the applied load to the base of columns. Therefore, it appears that punching is
again the dominant mode of deformation for short columns. It can be also seen that punching
ratios increase at higher area ratios, which reflects the loss in lateral confinement and increase
in the total force carried per column with increasing with area ratio.
The deformed shape of a 3×3 group of columns at L = 3 m and 6m is shown in Figure 5.28. It
can be seen in Figure 5.28(i) that closely-spaced columns are acting as a 'block' with the
surrounding soil. This is consistent with relatively low punching and compression ratios,
evident in Figures 5.26 and 5.27 for 3×3 and 4×4 groups, respectively. It appears that 'block'
failure remains the dominant mode of deformation with increasing column length for closely-
spaced columns. In contrast, it can be seen that columns at higher area ratios which are longer
than L = 3 m deform significantly with increasing column length. A reduction in punching
ratios and increase in compression ratios suggests that columns are deforming along their
length and do not transfer the applied load to the base of columns (see Figure 5.28). It can
also be seen that external columns tend to bend outwards and away from central columns.
This mechanism was also observed by McKelvey et al. (2004) in small scale laboratory tests.

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
3×3 cols
A/Ac = 3.5
2 2
Centre

4 4 Edge
Corner
6 6
8 8
3×3 cols
10 A/Ac = 3.5 10
Centre
12 Edge 12
Corner
14 14
(a-i) (a-ii)

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
3×3 cols
A/Ac = 8.0
2 2
Centre
4 4 Edge
Corner
6 6
8 8
3×3 cols
10 A/Ac = 8.0 10
Centre
12 Edge 12
Corner
14 14
(b-i) (b-ii)

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
3×3 cols
A/Ac = 14.1
2 2
Centre
4 4 Edge
Corner
6 6
8 8
3×3 cols
10 A/Ac = 14.1 10
Centre
12 Edge 12
Corner
14 14
(c-i) (c-ii)
Figure 5.26 - Variation of (i) punching and (ii) compression ratios with column length for a 3×3 group of
columns spaced at area ratios of (a) 3.5, (b) 8.0 and (c) 14.1

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
4×4 cols
A/Ac = 3.5
2 2
Centre

4 4 Edge
Corner
6 6
8 8
4×4 cols
10 A/Ac = 3.5 10
Centre
12 Edge 12
Corner
14 14
(a-i) (a-ii)

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
4×4 cols
A/Ac = 8.0
2 2
Centre
4 4 Edge
Corner
6 6
8 8
4×4 cols
10 A/Ac = 8.0 10
Centre
12 Edge 12
Corner
14 14
(b-i) (b-ii)

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
4×4 cols
A/Ac = 14.1
2 2
Centre
4 4 Edge
Corner
6 6
8 8
4×4 cols
10 A/Ac = 14.1 10
Centre
12 Edge 12
Corner
14 14
(c-i) (c-ii)
Figure 5.27 - Variation of (i) column and (ii) column ratios with column length for a 4×4 group of columns
spaced at area ratios of (a) 3.5, (b) 8.0 and (c) 14.1
Figure 5.28 - Deformed shapes from PLAXIS 3D Foundation for 3×3 group of columns at area ratios of (i) 3.5, (ii) 8.0 and (iii) 14.1 for column lengths (a) L = 3 m and (b) L
=6m
The variation of punching ratios for individual columns (i.e. centre, edge and corner columns)
within a 3×3 and 4×4 group can also be seen in Figures 5.26 and 5.27, respectively. For
columns shorter than L = 3 m, it appears that punching ratios are lowest for centre columns,
followed by edge and corner columns. This trend is also evident for all lengths of closely-
spaced columns (i.e. A/AC = 3.5), which suggests that this variation of punching ratios with
column length is associated with a punching mode of deformation. Columns which are
punching impart shear stresses to the surrounding soil which tends to drag the surrounding
soil downwards. Punching ratios are lowest for centre columns as the soil surrounding the
base of the column is dragged downwards on all sides by edge and corner columns. However,
it can be seen for widely-spaced columns (i.e. A/AC = 8.0 and 14.1) which are longer than L
= 3 m that punching ratios for centre columns exceeds punching ratios for corner and edge
columns. It is evident in Figure 5.28 that long widely-spaced edge columns tend to bulge and
bend outwards, which reduces punching at the base of columns. Centre columns are better
confined and can transfer more of the applied load to the base.
5.4.4 Comparison of infinite grid with small groups of columns

The influence of column confinement upon punching and compression ratios is shown in
Figures 5.29(i) and 5.29(ii), respectively. It can be seen in Figure 5.29(a-i) for columns at low
area ratios that punching ratios reduce with an increasing number of columns. It was
established previously that punching failure is the dominant mode of deformation for closely-
spaced floating columns. Furthermore, punching ratios are lowest for interior columns within
closely-spaced groups, as shear stresses from neighbouring columns drag the soil surrounding
the base of columns downwards. Therefore, as the number of interior columns increases, the
average punching ratio of closely-spaced groups of columns reduces.
A reduction in the punching ratio with an increasing number of columns is also observed in
Figures 5.29(b-i) and 5.29(c-i) for columns at higher area ratios. It is shown in Figure
5.28(iii) that column bending occurs in the outer row of widely-spaced columns. Bending
reduces the ability of columns to transfer the applied load to depth and hence results in lower
punching ratios. For a constant area ratio, footing width increases with an increasing number
of columns. Columns become increasingly slender as footing width increases and,
consequently, are more susceptible to bending. Therefore, punching ratios reduce with an
increasing number of columns at high area ratios.

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
A/AC = 3.5
2 2 1 col
2×2

4 4 3×3 - average
4×4 - average
Unit Cell
6 6
8 8
A/AC = 3.5
10 1 col 10
2×2
3×3 - average
12 12
4×4 - average
Unit Cell
14 14
(a-i) (a-ii)

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
A/AC = 8.0
2 2 1 col
2×2
3×3 - average
4 4
4×4 - average
Unit Cell
6 6
8 8
A/AC = 8.0
10 1 col 10
2×2
3×3 - average
12 12
4×4 - average
Unit Cell
14 14
(b-i) (b-ii)

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
A/AC = 14.1
2 2 1 col
2×2
4 4 3×3 - average
4×4 - average
Unit Cell
6 6
8 8
A/AC = 14.1
10 1 col 10
2×2
3×3 - average
12 12
4×4 - average
Unit Cell
14 14
(c-i) (c-ii)
Figure 5.29 - Influence of column confinement upon (i) punching and (ii) compression ratios for various
lengths and configurations of columns spaced at area ratios of (a) 3.5, (b) 8.0 and (c) 14.1
It can be seen in Figure 5.29(i) that unit cell conditions yield the lowest punching ratios for
columns at low area ratios. However, this is not the case for high area ratios and it can be
seen that punching ratios for the unit cell are larger than those for groups of columns longer
than 7 m and 5 m in Figures 5.29(b-i) and 5.29(c-i), respectively. The external columns in
groups experience a loss of lateral confinement and bend outwards, which reduces the
punching ratio. An infinite grid of columns cannot bend due to symmetry and punching ratios
are higher, despite the increased level of column confinement.
The relationship between column confinement and compression ratios is shown in Figure
5.29(ii). It can be seen for closely-spaced columns that the average compression ratios of a
group reduces as the number of columns increases. This is consistent with findings from the
settlement performance, where the jump in settlement improvement factors observed for end-
bearing columns with an increasing number of columns. This indicates that larger groups of
columns transfer higher proportions of the applied load to the base, which reduces the
deformation along the length of columns.
5.4.5 Influence of column arrangement upon deformational behaviour

The variation of punching and compression ratios with length for various column
arrangements (see Figure 5.3) supporting a 3 m square footing is shown in Figure 5.30. An
increase in the punching ratio with length is observed for all short columns. This is coupled
with relatively low compression ratios, which indicates that punching is the dominant mode
of deformation. An increase in punching ratios is observed for fewer supporting stone
columns. This may be related to an increase in the total load taken per column as the number
of supporting columns reduces. Columns of similar length which are punching have the
capacity to develop similar shear force along the sides. Therefore, any extra load taken by
columns (i.e. for fewer supporting columns) is transferred to the base of columns, which
results in higher punching ratios.
While all configurations of columns exhibit a similar variation of punching and compression
ratios with increasing length for short columns, a marked change is observed with increasing
length thereafter. The groups of 4 and 5 columns exhibit a significant reduction in punching
ratio and a simultaneous increase in compression ratio. This indicates that these column
groups are not transferring the applied load to the base of columns and are deforming closer

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
B=3m
2 2 4 cols
5 cols - average

4 4 7 cols - average
9 cols - average
6 6
8 8
B=3m
10 4 cols 10
5 cols - average
12 7 cols - average 12
9 cols - average
14 14
(i) (ii)
Figure 5.30 - Variation of (i) punching and (ii) compression ratios with column length for various
configurations of stone columns beneath a 3 m square footing
to the ground surface. This is consistent with previous findings for long widely-spaced
columns. In contrast to this behaviour, relatively low compression ratios are observed for the
groups of 7 and 9 columns. This suggests that punching failure remains the dominant mode of
deformation with increasing column length.

The influence of column position relative to the edge of footings (see Figure 5.4) upon
punching and compression ratios is shown in Figure 5.31. For columns shorter than L = 3 m,
it can be seen that the punching ratio increases for columns positioned closer to the edge of
footings. It is established that punching is the dominant mode of deformation for all short
columns. Columns which are punching develop shear stress along the side of columns, which
tends to drag the surrounding soil downwards. The influence of shear stress becomes more
pronounced as columns are positioned closer together and the surrounding soil is displaced
more uniformly, which results in lower punching ratios. However, no change in punching
ratios is observed for column lengths longer than L = 3 m when bulging becomes the
dominant mode of deformation for longer columns. While no change in the punching ratio
occurs for columns which are bulging, the compression ratio increases for columns which are
positioned closer to the edge of footings. This may be explained as closely-spaced columns
are less susceptible to bending and, therefore, exhibit less compression.

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
4 cols - B = 3 m
2 2 s = 1.0 m
s = 1.5 m

4 4 s = 2.0 m
s = 1.0 m (B = 2 m)
6 6
8 8
4 cols - B = 3 m
10 s = 1.0 m 10
s = 1.5 m
12 s = 2.0 m 12
s = 1.0 m (B = 2 m)
14 14
(i) (ii)
Figure 5.31 - Variation of (i) punching and (ii) compression ratios with column length for various column
spacings beneath a 3 m square footing
The deformational behaviour of a 2×2 group spaced at 1.0 m and supporting a 2 m and a 3 m
square pad footing is also compared in Figure 5.31. It appears that the mode of deformation is
different for the group of columns beneath the 2 m square footing (i.e. A/AC = 3.5). Increased
punching ratios and lower compression ratios indicate that punching remains the dominant
mode of deformation for all column lengths. Therefore, it is postulated that area ratio is more
critical than column spacing in governing the deformational behaviour of columns.
5.4.7 Influence of column compressibility

The influence of column compressibility upon the deformational behaviour of various
column arrangements supporting a 3 m square footing (see Figure 5.5) is shown in Figure
5.32. It can be seen that an increase in column stiffness results in an increase in punching
ratios and decrease in compression ratios. While the increase in punching ratios is relatively
uniform for all column configurations, a larger decrease in compression ratios is observed for
the group of 9 columns. This is consistent with findings from the settlement performance of
stone columns, as the group of 9 columns are in more of an elastic state and are therefore
more influenced by changes in elastic stiffness moduli.

The influence of column strength upon the deformational behaviour of various column
arrangements supporting a 3 m square footing (see Figure 5.5) is shown in Figures 5.33. It

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
4 cols - s = 2.0 m
2 2 E_50 = 30 MPa

E_50 = 50 MPa
4 4
E_50 = 70 MPa
6 6
8 8
4 cols - s = 2.0 m
10 10
E_50 = 30 MPa
12 E_50 = 50 MPa 12
E_50 = 70 MPa
14 14
(a-i) (a-ii)

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
5 cols - s = 1.0 m
2 2 E_50 = 30 MPa
E_50 = 50 MPa
4 4
E_50 = 70 MPa
6 6
8 8
5 cols - s = 1.0 m
10 10
E_50 = 30 MPa
12 E_50 = 50 MPa 12
E_50 = 70 MPa
14 14
(b-i) (b-ii)

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
9 cols - s = 1.0 m
2 2 E_50 = 30 MPa
E_50 = 50 MPa
4 4
E_50 = 70 MPa
6 6
8 8
9 cols - s = 1.0 m
10 10
E_50 = 30 MPa
12 E_50 = 50 MPa 12
E_50 = 70 MPa
14 14
(c-i) (c-ii)
Figure 5.32 - Influence of column compressibility upon the variation of (i) punching and (ii) compression ratios
with column length for a group of (a) 4, (b) 5 and (c) 9 columns beneath a 3 m square footing
Punching ratios, (ucol-usoil)/ufooting Compression ratios, (ufooting-ucol)/ufooting

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
4 cols - s = 2.0 m
2 2 φ = 40°
φ = 45°

4 4
φ = 50°
6 6
8 8
4 cols - s = 2.0 m
10 10
φ = 40°
12 φ = 45° 12
φ = 50°
14 14
(a-i) (a-ii)

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
5 cols - s = 1.0 m
2 2 φ = 40°
φ = 45°
4 4
φ = 50°
6 6
8 8
5 cols - s = 1.0 m
10 10
φ = 40°
12 φ = 45° 12
φ = 50°
14 14
(b-i) (b-ii)

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
9 cols - s = 1.0 m
2 2 φ = 40°
φ = 45°
4 4
φ = 50°
6 6
8 8
9 cols - s = 1.0 m
10 10
φ = 40°
12 φ = 45° 12
φ = 50°
14 14
(c-i) (c-ii)
Figure 5.33 - Influence of column strength upon the variation of (i) punching and (ii) compression ratios with
column length for a group of (a) 4, (b) 5 and (c) 9 columns beneath a 3 m square footing
can be seen that compression ratios decrease with an increasing angle of friction for the stone
column. This results in a corresponding increase in punching ratios. The changes in punching
and compression ratios are more pronounced for fewer columns and only negligible changes
are observed when increasing the angle of internal friction from 45° to 50° for the group of 9
columns.

The influence of the coefficient of lateral earth pressure (K0) for the lower Carse clay layer
upon the deformational behaviour of various column arrangements supporting a 3 m square
footing (see Figure 5.5) is shown in Figures 5.34. It appears that increasing K0 has a minor
influence upon punching and compression ratios. This is consistent with Kirsch (2006) who
conducted a FEA on a group of stone columns and observed that column installation effects
do not influence the deformational behaviour of stone columns but do enhance the settlement
performance of stone columns.
5.4.10 Influence of stiff crust

The influence of the lower Carse clay and the stiff crust upon the deformational behaviour of
various column arrangements (see Figure 5.5) supporting a 3 m square footing is shown in
Figure 5.35. Profile 1 is the standard Bothkennar soil profile and the punching and
compression ratios for this profile in Figure 5.35(i) were presented earlier in Figure 5.30.
Punching ratio
It can be seen in Figure 5.35(i) that punching ratios increase with column length up to a
maximum at L = 3 m for Profile 1. A similar variation is observed for Profile 2, however
punching ratios reach a maximum at L = 2 m and are lower than Profile 1 for column lengths
L > 3 m. This reflects the increased stiffness at depth for Profile 2, as the lower Carse clay in
the standard Bothkennar soil profile (Profile 1) is replaced with the stiffer upper Carse clay.
Profile 3 consists entirely of upper Carse clay and, therefore, allows for the influence of the
stiff crust to be assessed. It can be seen that punching ratios are largest near the surface for
this soil profile and decrease with increasing column length thereafter. Punching ratios for
Profile 3 are higher than Profiles 1 and 2 for short columns, but this trend reverses for
columns longer than L = 3 m.
Compression ratio
The variation of compression ratios with column length is shown in Figure 5.35(ii). It can be
seen that compression ratios are lowest for Profile 1, which indicates that columns formed in
Profiles 2 and 3 are more susceptible to a bulging mode of deformation. This may be
explained by a number of reasons:
(i) The lower Carse clay in Profile 1 is replaced by the stiffer upper Carse clay in Profiles 2
and 3. Therefore, the base of floating columns in Profile 2 and 3 is formed in a more
competent soil layer. The enhanced support at the base of columns reduces punching and
forces columns to compress more along their length.
(ii) The absence of the stiff crust (i.e. Profile 3) leads to the highest compression ratios. The
stiff crust prevents bulging in the upper sections of columns, where overburden stresses
are lowest.
It appears that the modes of deformation for columns formed in a homogenous soil profile, as
used in the laboratory tests described in the literature review (see Sections 2.2 and 2.3), are
very different to columns formed in a more realistic soil profile, as modelled in PLAXIS 3D
Foundation. This is discussed in more detail in Section 7.3.3.

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
4 cols - s = 2.0 m
2 2 K0_soil = 0.75

K0_soil = 1.00
4 4
K0_soil = 1.25
6 6
8 8
4 cols - s = 2.0 m
10 10
K0_soil = 0.75
12 K0_soil = 1.00 12
K0_soil = 1.25
14 14
(a-i) (a-ii)
Punching ratios, (ucol-usoil)/ufooting Compression ratio, (ufooting-ucol)/ufooting

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
5 cols - s = 1.0 m
2 2 K0_soil = 0.75
K0_soil = 1.00
4 4
K0_soil = 1.25
6 6
8 8
5 cols - s = 1.0 m
10 10
K0_soil = 0.75
12 K0_soil = 1.00 12
K0_soil = 1.25
14 14
(b-i) (b-ii)
Punching ratios, (ucol-usoil)/ufooting Compression rattio (ufooting-ucol)/ufooting

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.2 0.4 0.6 0.8 1.0
0 0
9 cols - s = 1.0 m
2 2 K0_soil = 0.75
K0_soil = 1.00
4 4
K0_soil = 1.25
6 6
8 8
9 cols - s = 1.0 m
10 10
K0_soil = 0.75
12 K0_soil = 1.00 12
K0_soil = 1.25
14 14
(c-i) (c-ii)
Figure 5.34 - Influence of column installation effects upon the variation of (i) punching and (ii) compression
ratios with column length for a group of (a) 4, (b) 5 and (c) 9 columns beneath a 3 m square
footing

0.00 0.10 0.20 0.30 0.40 0.0 0.2 0.4 0.6 0.8 1.0
0 0
2 2

4 4
6 6
8 8
10 4 cols - s = 2.0 m 10 4 cols - s = 2.0 m

Profile 1 Profile 1
12 Profile 2 12 Profile 2
Profile 3 Profile 3
14 14
(a-i) (a-ii)
5 columns
Punching ratio, (ucol-usoil)/ufooting Column compression, (ufooting-ucol)/ufooting
0.00 0.10 0.20 0.30 0.40 0.0 0.2 0.4 0.6 0.8 1.0
0 0
2 2
4 4
6 6
8 8
10 5 cols - s = 1.0 m 10 5 cols - s = 1.0 m

Profile 1 Profile 1
Profile 3 Profile 3
14 14
(b-i) (b-ii)

0.00 0.10 0.20 0.30 0.40 0.0 0.2 0.4 0.6 0.8 1.0
0 0
2 2
4 4
6 6
8 8
10 9 cols - s = 1.0 m 10 9 cols - s = 1.0 m

Profile 1 Profile 1
Profile 3 Profile 3
14 14
(c-i) (c-ii)
Figure 5.35 - Influence of stiff crust upon the variation of (i) punching and (ii) compression ratios with column
length for a group of (a) 4, (b) 5 and (c) 9 columns beneath a 3 m square footing
5.5 Stress concentration ratio

The ratio of vertical effective stress in the column and the surrounding soil is known as the
stress concentration ratio. It is an important parameter which indicates the efficiency of stone
columns at enhancing the settlement performance of treated soils. The stress concentration
ratio is typically measured at the top of stone columns. Stress concentration ratios from field
tests (Munfakh et al., 1984) and laboratory studies (Black 2006, McKelvey et al., 2004 and
Muir Wood et al., 2000) are determined from pressure transducers which are located at
discrete locations in columns and the surrounding soil, as shown in Figure 5.36(i). Stress
concentration ratios are therefore relative to a specific point and hence a specific stress level.
However, stress concentration ratios from numerical and analytical studies (Pulko & Majes,
2005) are measured more comprehensively, as vertical stress is averaged over the entire
column and soil areas, as illustrated in Figure 5.36 (ii).
(i) (ii)
Figure 5.36 - Determination of stress concentration ratios for (i) field/laboratory studies and (ii)
numerical/analytical studies
Stress concentration ratios from PLAXIS 3D Foundation are defined by two methods, which
are illustrated in Figure 5.37:
(i) The first method compares the average vertical stress in each column (σ'col) to the
average vertical stress in the entire soil area beneath the footing (σ'soil, average).
(ii) The second method compares the average vertical stress in each column (σ'col) to the
average vertical stress in the soil within a square zone of influence surrounding each
column (σ'soil); the width of the zone of influence is equal to the column spacing.
The first method allows the stress levels in corner, edge and centre columns to be directly
compared with each other, as the average stress over the entire area of soil is used. The
second method is a measure of the work each column is doing relative to the stress in the
surrounding zone of influence and, therefore, indicates the efficiency of each column.
(i) (ii)
Figure 5.37 - Stress concentration ratio determined by (i) averaging stress over entire soil area and (ii)
averaging stress within the zone of influence for each column
Stress concentration ratios determined from both methods will be identical for single
columns, 2×2 groups and infinite grids as due to symmetry, the area of soil beneath footings
corresponds to the zone of influence for each column. However, this is not the case for
groups of columns larger than 2×2 configurations as the stress in the zone of influence varies
for different columns. By way of example, the distribution of vertical effective stress
immediately beneath a 4.5 m square concrete footing, which is (i) unreinforced (L = 0 m) and
(ii) reinforced with a 3×3 group of end-bearing (L = 13.9 m) stone columns, is shown in
Figure 5.38. It can be seen that high vertical stress develops beneath the edges of the rigid
unreinforced footing. The high stress levels beneath the edges of the footing reduce for the
reinforced case as columns absorb a larger proportion of the load and, therefore, reduce the
stress on the surrounding soil.
(i)
(ii)
Figure 5.38 - Distribution of vertical effective stress (σ'vert) immediately beneath a 4.5 m square footing, which
is (i) unreinforced (L = 0 m) and (ii) reinforced with end-bearing stone columns (L = 13.9 m)
5.5.1 Stress concentration ratios for infinite grids of stone columns

The variation of stress concentration ratios with column length for an infinite grid of columns
is shown in Figure 5.39. It can be seen that the stress concentration ratio remains relatively
constant with increasing column length. It was shown previously (see Section 5.4.1) that an
infinite grid of columns exhibit punching and bulging modes of deformation. It is also well
known that all columns within an infinite grid deform equally due to symmetry.
Consequently, no shear stress develops along the column-soil interface near the surface and
the stress concentration ratio measured at the surface depends on the stiffness properties of
the column and the surrounding soil. Therefore, the stress concentration ratio does not change
with increasing column length.
A slight reduction in the stress concentration ratio is observed at high area ratios. Column
confinement reduces at high area ratios, which consequently reduces the stiffness of stone
columns and hence yields lower stress concentration ratios.
Stress concentration, σ'col / σ'soil

1 3 5 7 9
0
2
10 Unit cell
A/Ac = 3.5
12 A/Ac = 8.0
A/Ac = 14.1
14
Figure 5.39 - Variation of stress concentration ratio with column length for an infinite grid of columns
5.5.2 Stress concentration ratios for small groups of stone columns

Single columns and 2×2 groups
The variation of stress concentration ratios with column length for single columns and 2×2
groups is shown in Figures 5.40(i) and 5.40(ii), respectively. An increase in stress
concentration ratios with column length is observed for all columns shorter than L = 3 m. It
was shown previously (see Section 5.4.2) that punching is the dominant mode of deformation
for short single columns and 2×2 groups. Groups of columns which are punching develop
shear stress along the sides and end-bearing pressure at the base of columns. The load-
carrying capacity of columns increases with increasing column length as shear stress can
develop over a larger area. As a result, stress concentration ratios increase in concert.
Stress concentration, σ'col / σ'soil Stress concentration, σ'col / σ'soil

1 3 5 7 9 1 3 5 7 9
0 0
2 2

4 4
6 6
8 8
10 1 column 10 2×2 cols

A/Ac = 3.5 A/Ac = 3.5
12 A/Ac = 8.0 12 A/Ac = 8.0
A/Ac = 14.1 A/Ac = 14.1
14 14
(i) (ii)
Figure 5.40 - Variation of stress concentration ratio with column length for (i) single and (ii) 2×2 groups of
stone columns
Stress concentration ratios continue to increase with column length for closely-spaced
columns (i.e. A/AC = 3.5), which is consistent with previous findings that closely-spaced
columns punch into the underlying soil. While large increases in stress concentration ratios
are initially observed with increasing column length, the rate of increase in stress
concentration ratios with column length reduces significantly for column lengths L > 3 m.
This may attributed to large percentage increases in column length for short columns (100%
for L = 1 → 2 m) compared to longer columns (10% for L = 10 → 11 m).
It can also be seen in Figures 5.40(i) and 5.40(ii) that stress concentration ratios for columns
at A/AC = 8.0 and 14.1 reach maximum values at L ≈ 4 m. Stress concentration ratios remain
relatively constant with increasing column length thereafter for columns at A/AC = 8.0, which
indicates that the mode of deformation has changed from punching to bulging. Column
bulging occurs at the point of lowest lateral support, which occurs near the top of the lower
Carse clay (z ≈ 3 m). Increasing the column length for this mode of deformation does not
enhance the load-carrying capacity of columns and, therefore, does not increase stress
concentration ratios. A slight reduction in stress concentration ratios for columns at A/AC =
14.1 is observed with increasing column length in the range 4–7 m and 4–5 m for single
columns and 2×2 groups, respectively. A combination of punching and bulging modes of
deformation may be occurring over this range of lengths. However, bulging becomes more
prevalent as column length increases and stress concentration ratios decrease slightly towards
constant values.
A reduction in stress concentration ratios is observed for higher area ratios at L = 1 m. This is
consistent with findings from the deformational behaviour of single columns and 2×2 groups,
as shown in Figures 5.24(i) and 5.25(i), respectively, where a reduction in column punching
is observed for higher area ratios at L = 1 m. The footing size increases at higher area ratios,
which stresses the soil deeper and induces more displacement in the soil surrounding the base
of columns at z = 1 m. Therefore, the end-bearing resistance at the base of columns, and
consequently the load-carrying capacity of stone columns, reduces with increasing area ratio.
However, it is interesting to note that stress concentration ratios increase at a faster rate at
higher area ratios for L = 2, 3 and 4 m. The average load carried per column increases with
footing area. Therefore, a larger force is carried along the sides and at the base of columns.
This results in high punching ratios, as shown in Section 5.4.3, and also forces columns to
carry a larger proportion of the applied load which results in higher stress concentration
ratios.
3×3 and 4×4 groups

The variation of stress concentration ratio with column length for 3×3 and 4×4 groups of
columns is shown in Figures 5.41 and 5.42, respectively. The stress concentration ratio is
measured using two methods, as outlined at the start of this section. The first method, which
is shown in Figures 5.41(i) and 5.42(i), compares the average stress in columns with the
average stress in the entire soil area beneath the footing. The second method, which is shown
in Figures 5.41(ii) and 5.42(ii), compares the average stress in columns with the average
stress in the soil surrounding each column.
Variation of stress concentration ratio with column position

It can be seen in Figures 5.41(i) and 5.42(i) that stress levels are highest in corner columns,
followed by edge and centre columns. This is due to the position of columns relative to the
edge of footings. It is well known from elastic theory that elevated stress levels develop
beneath the edge of rigid footings and, therefore, positioning columns in this location gives
them the potential to absorb more load. It can also be seen that the variation in stress
Stress concentration ratio, σ'col / σ'soil,average Stress concentration, σ'col / σ'soil

1 3 5 7 9 1 3 5 7 9
0 0
2 2

4 4
6 6
8 8
3×3 cols 3×3 cols
A/Ac = 3.5 A/Ac = 3.5
10 10
Centre Centre
12 Edge 12 Edge
Corner Corner
14 14
(a-i) (a-ii)
Stress concentration, σ'col / σ'soil, average Stress concentration, σ'col / σ'soil

1 3 5 7 9 1 3 5 7 9
0 0
2 2
4 4
6 6
8 8
3×3 cols 3×3 cols
A/Ac = 8.0 A/Ac = 8.0
10 10
Centre Centre
12 Edge 12 Edge
Corner Corner
14 14
(b-i) (b-ii)

1 3 5 7 9 1 3 5 7 9
0 0
2 2
4 4
6 6
8 8
3×3 cols 3×3 cols
A/Ac = 14.1 A/Ac = 14.1
10 10
Centre Centre
12 Edge 12 Edge
Corner Corner
14 14
(c-i) (c-ii)
Figure 5.41 - Variation of stress concentration ratio with column length for a 3x3 group of columns, spaced at
area ratios of (a) 3.5, (b) 8.0 and (c) 14.1

1 3 5 7 9 1 3 5 7 9
0 0
2 2

4 4
6 6
8 8
4×4 cols 4×4 cols
A/Ac = 3.5 A/Ac = 3.5
10 10
Centre Centre
12 Edge 12 Edge
Corner Corner
14 14
(a-i) (a-ii)

1 3 5 7 9 1 3 5 7 9
0 0
2 2
4 4
6 6
8 8
4×4 cols 4×4 cols
10 A/Ac = 8.0 10 A/Ac = 8.0
Centre Centre
12 Edge 12 Edge
Corner Corner
14 14
(b-i) (b-ii)

1 3 5 7 9 1 3 5 7 9
0 0
2 2
4 4
6 6
8 8
4×4 cols 4×4 cols
A/Ac = 14.1 A/Ac = 14.1
10 10
Centre Centre
12 Edge 12 Edge
Corner Corner
14 14
(c-i) (c-ii)
Figure 5.42 - Variation of stress concentration ratio with column length for a 4x4 group of columns, spaced at
concentration ratios between individual columns becomes more pronounced for lower area
ratios. The external row of columns are positioned a distance 0.5×s (s = column spacing)
from the edge of footings. As column spacing decreases for lower area ratios, columns are
positioned closer to the edge of footings and hence closer to the zone of elevated stress levels.
Therefore, the variation in stress concentration ratios is most significant at low area ratios.
Variation of stress concentration ratio with column length and area ratio
The variation of stress concentration ratios with column length for 3×3 and 4×4 groups of
columns at A/AC = 3.5 is shown in Figures 5.41(a-i) and 5.42(a-i), respectively. An increase
in stress concentration ratios with increasing column length is observed for all columns up to
L = 4 m and L = 3 m for 3×3 and 4×4 groups of columns, respectively. A slight reduction in
stress concentration ratios with increasing column length is subsequently observed for corner
columns up to L ≈ 7 m, while stress concentration ratios in edge and centre columns continue
to increase or remain constant. As outlined previously, corner columns carry a greater
proportion of the applied load due to the position of columns relative to the footing edge.
However, corner columns are more susceptible to bulging and bending as column length
increases. Therefore, corner columns cannot sustain the high levels of vertical stress and
transfer the applied load to edge and centre columns. Consequently, stress concentration
ratios continue to increase with column length for edge and centre columns.
The transfer of vertical stress from corner columns to edge and centre columns is also evident
in the variation of stress concentration ratios, measured using the second method (σ'col/ σ'soil).
It can be seen in Figures 5.41(a-ii) and 5.42(a-ii) for 3×3 and 4×4 groups of columns,
respectively, that corner and edge columns initially carry a larger proportion of the applied
load relative to the surrounding soil. However, as column length increases a larger proportion
of the load is carried by centre columns relative to the surrounding soil. This reflects the
higher levels of lateral confinement associated with central columns.
The variation of stress concentration ratios with column length for columns spaced at high
area ratios (i.e. A/AC = 8.0 and 14.1) is shown in Figures 5.41(b-i) and 5.41(c-i) for 3×3
groups of columns and 5.42(b-i) and 5.42(c-i) for 4×4 groups of columns. As with closely-
spaced columns, it can be seen that stress concentration ratios increase with column length up
to L = 3 m. This is again consistent with a punching mode of deformation and is in keeping
with previous findings for short widely-spaced columns. A slight reduction in stress
concentration ratios with increasing column length is subsequently observed for corner
columns up to L ≈ 5 m, while stress concentration ratios continue to increase for centre
columns at A/AC = 8.0. This indicates that the heavily loaded corner columns transfer the
applied vertical load to centre columns. No change in stress concentration ratio with
increasing column length is observed thereafter, which suggests that bulging is the dominant
mode of deformation. While no transfer of stress is evident for columns at an area ratio A/AC
= 14.1, it can be seen that stress concentration ratios for all columns (i.e. corner, edge and
centre) reduce slightly with increasing column length up to L ≈ 6 m and are constant
thereafter. The absence of stress transfer from corner to centre columns may be attributed to
the relatively uniform stress levels in all columns. The slight reduction in stress concentration
ratios with is also evident from single and 2×2 groups of columns and suggests that both
punching and bulging modes of deformation are occurring. Bulging becomes more prevalent
as column length increases and no change in stress concentration ratios is observed with
increasing column length.
The positive influence of column confinement can be seen in Figures 5.41(b-ii) and 5.41(c-ii)
for 3×3 groups of columns and 5.42(b-ii) and 5.42(c-ii) for 4×4 groups of columns. Stress
concentration ratios - measured using the second method (i.e. the vertical stress in columns
relative to the local stress level) - are highest in the centre columns. This reflects the
increased levels of lateral confinement associated with centre columns. This is in contrast to
edge and corner columns which bulge and bend outwards away from central columns.
5.5.3 Comparison of infinite grid with small groups of columns

The influence of the number of columns upon the variation of stress concentration ratios is
shown in Figures 5.43. It can be seen in Figure 5.43(a) for closely-spaced columns (i.e. A/AC
= 3.5) that stress concentration ratios increase for larger groups of columns. In addition to the
enhanced lateral confinement associated with larger groups of columns, the increase in stress
concentration ratios is also due to the position of columns relative to the footing edge. The
distribution of vertical stress beneath unreinforced footings for a single column, 2×2 group,
3×3 group and 4×4 group of columns is shown in Figure 5.44. As the number of columns
increases, the external rows of columns are positioned progressively closer to the edge of
footings. Therefore, these columns are located in the zone of elevated stress levels and have

1 3 5 7 9
0

4
8 A/AC = 3.5
1 col
10 2×2
3×3 - average
12 4×4 - average
Unit Cell
14
(a)

1 3 5 7 9
0
2
8 A/AC = 8.0
1 col
10 2×2
3×3 - average
12 4×4 - average
Unit Cell
14
(b)

1 3 5 7 9
0
2
8 A/AC = 14.1
1 col
10 2×2
3×3 - average
12 4×4 - average
Unit Cell
14
(c)
Figure 5.43 - Variation of stress concentration ratio for small groups and infinite grids of columns spaced at
(i) (ii) (iii) (iv)

Figure 5.44 - Vertical stress beneath an unreinforced footing for (i) single column, (ii) 2×2, (iii) 3×3 and (iv) 4×4 groups of columns at an area ratio of 3.5
the potential to absorb a larger proportion of the applied load. Consequently, the average
stress concentration ratios for the group of columns increases.
A similar trend is observed for short columns at higher area ratios (i.e. A/AC = 8.0 and 14.1)
in Figures 4.43(b) and 4.43(c). It was established previously that punching is the dominant
mode of deformation for this configuration of columns. An increase in stress concentration
ratios with an increasing number of columns is again due to the position of columns relative
to the edge of the footing. However, once column length exceeds L = 3 m, the mode of
deformation changes to bulging and no difference in stress concentration ratios is observed.
Columns which are bulging fail at the same location (i.e. the weakest part of the soil profile).
Therefore, the ultimate load is the same for all columns and is not dependent upon the
position of columns relative to the edge of footings.
The stress concentration ratios for a unit cell are relatively uniform as columns are located
within an infinite grid and therefore are not subject to the elevated stress levels near the edge
of rigid footings.
5.5.4 Influence of column arrangement

The variation of the stress concentration ratio with column length for various arrangements of
stone columns (see Figure 5.3) beneath a 3 m square footing is shown in Figure 5.45. Stress
concentration ratios for groups of 4 and 5 columns (A/AC = 8.0 and 6.4, respectively)
increase with column length up to a maximum at L = 3 m. This suggests that punching is the
dominant mode of deformation which corresponds with previous findings for short columns
at high area ratios. A slight decrease in stress concentration ratios is observed with increasing
column length up to L = 6 m, which is followed by constant stress concentration ratios
thereafter. The slight reduction in stress concentration ratios with column length indicates
that the mode of deformation is changing from punching to bulging. Column bulging
becomes more prominent as column length increases, which results in lower stress
concentration ratios. It appears that bulging is the only mode of deformation for column
lengths longer than L = 6 m as stress concentration ratios are constant. Increasing the length
of columns which are bulging or bending does not enhance the load-carrying capacity and,
therefore, does not result in increased stress concentration ratios. It can also be seen in Figure
5.45 that stress concentration ratios are higher for the group of 4 columns than the group of 5
columns. The lower stress concentrations ratios for the group of 5 columns are observed due
to the location of the central column in a region of lower stress levels (see Figure 5.44).
Therefore, this column takes less of the applied load and the average stress concentration
ratio for the group is reduced.
Stress concentration, σ'col / σ 'soil, average

1 3 5 7 9
0
Column length, L (m) 4
B=3m
10
4 cols
5 cols - average
12
7 cols - average
9 cols - average
14
Figure 5.45 - Variation of stress concentration ratio with column length for a group of 4, 5, 7 and 9 columns
beneath a 3 m square footing
The variation of stress concentration ratios with length for the groups of 7 and 9 columns is
quite different to groups of 4 and 5 columns. Stress concentration ratios for groups of 7
columns and 9 columns increase at a slower rate with column length up to L = 3 m than the
smaller groups of columns. Furthermore, stress concentration ratios for the groups of 7 and 9
columns continue to increase with length longer than L = 3 m, which suggests that punching
remains the dominant mode of deformation with increasing column length. However, the rate
of increase in stress concentration ratio reduces with increasing column length and stress
concentration ratios are eventually constant for the group of 7 columns longer than L = 8 m.
The magnitude of stress concentration ratios is higher for the group of 9 columns than the
group of 7 columns. This is due to the position of columns relative to the footing edge. The
outer ring of columns for a 9 column group are positioned closer to the footing edge and
therefore closer to the elevated stress levels which develop in this region.

The influence of column position relative to the footing edge (see Figure 5.4) upon stress
concentration ratios is shown in Figure 5.46. The variation of stress concentration ratios with
column length is quite similar to previous findings for columns at high area ratios. Stress
concentration ratios increase with column length up to L = 3 m, which indicates that all
column configurations are punching into the underlying soil. However, no increase in the
stress concentration ratio is observed with increasing column length thereafter, which
indicates that the mode of deformation has changed to bulging. A slight decrease in stress
concentration ratios with column length is observed in the range 3–6 m for columns spaced at
2.0 m, which indicates that a transition from punching to bulging occurs for this range of
column lengths.
It can be seen in Figure 5.46 that the position of columns relative to the footing edge has a
significant influence upon stress concentration ratios. Stress concentration ratios increase as
columns are positioned closer to the footing edge; an increase of c.50% is observed at L = 3
m. This is due to elevated stress levels which occur beneath the edge of rigid footings, as
shown in Figure 5.47. Positioning columns closer to this zone allows columns to absorb more
load and hence develop higher stress concentration ratios. However, it was shown in Section
5.3.5 that the position of columns relative to the footing edge has a relatively minor influence
upon the settlement performance of stone columns. Therefore, stress concentration ratios
measured at the surface do not define the settlement behaviour of a group of columns. This is
an interesting finding and suggests that the variation of stress concentration ratios with depth,
which is not routinely measured in laboratory or field tests, may play an important role in
determining the settlement performance of stone columns. The variation of stress
concentration ratio with depth is investigated as part of this thesis in Section 6.3.
The influence of footing overhang upon stress concentration ratios is also investigated by
comparing a 2×2 group of columns spaced at 1.0 m beneath a 2 m and a 3 m square footing.
It can be seen in Figure 5.46 that stress concentration ratios for the 3 m square footing
increase with column length to a maximum at L = 3 m and are constant thereafter. In contrast,
a continuous increase in stress concentration ratios with increasing column length is observed
for the 2 m square footing. Furthermore, it can be seen that stress concentration ratios are
higher for the 2×2 group of columns beneath a 2 m footing. This is again due to the position
of the columns close to the footing edge.
It appears that the position of columns and the mode of deformation both play important roles
in the magnitude of stress concentration ratios.

1 3 5 7 9
0

4
8
4 cols - B = 3 m
10 s = 1.0 m
s = 1.5 m
12 s = 2.0 m
s = 1.0 m (B = 2 m)
14
Figure 5.46 - Variation of stress concentration ratio with column length for various columns positions beneath a
3 m square footing
(i)
(ii)
(iii)
Figure 5.47 - Vertical effective stress beneath a 3 m square footing for columns spaced at (i) 1.0 m, (ii) 1.5 m
and (iii) 2.0 m
5.5.6 Influence of column stiffness

The influence of column stiffness upon stress concentration ratios for various arrangements
of columns (see Figure 5.5) beneath a 3 m square footing is shown in Figure 5.48. Column
stiffness is controlled by the secant Young’s modulus (E50) and it can be seen that an increase
in E50 leads to higher stress concentration ratios.
It can be seen for columns shorter than L = 3 m that the influence of columns stiffness upon
stress concentration ratios becomes more pronounced as column length increases up to L = 3
m. However, the variation in stress concentration ratios with E50 does not change with
increasing column length thereafter for groups of 4 and 5 columns, as shown in Figure
5.48(a) and 5.48(b), respectively. This may be related to the bulging mode of deformation
which becomes more prevalent for groups of 4 and 5 columns which are longer than L = 3 m.
It was shown previously that the load-carrying capacity, and hence the stress concentration
ratio, of columns which are bulging does not increase with column length.
The variation of stress concentration ratios with column length for a group of 9 columns is
shown in Figure 5.48(c). As with the groups of 4 and 5 columns, the influence of column
stiffness becomes more pronounced with increasing column length up to L = 3 m. Stress
concentration ratios continue to increase at a slower rate with increasing column length
thereafter for E50 = 70 MPa. However, no increase in stress concentration ratios is observed
with column length for E50 = 30 MPa, which suggests that the mode of deformation has
changed from punching to bulging.
A linear relationship appears to exist between column stiffness and stress concentration ratios
as shown by stress concentration ratios for end-bearing columns in Table 5.4. Reductions in
Young’s moduli of 29% (E50 = 70 MPa→50 MPa) and 57% (E50 = 70 MPa→30 MPa) yield
average reductions in stress concentration ratios of 12% and 31%, respectively.
Table 5.4 - Influence of column compressibility upon stress concentration ratios for end-bearing columns
Stress concentration ratio, σ'col / σ'soil % reduction
No. of cols Column compressibility, E50 (MPa) Column compressibility, E50 (MPa)
30 50 70 30 50 70
4 4.6 5.7 6.4 -28 -11 -
5 4.3 5.4 6.1 -30 -11 -
9 4.1 5.3 6.3 -35 -15 -

The influence of column strength upon the stress concentration ratio for various arrangements
of columns (see Figure 5.5) beneath a 3 m square footing is shown in Figure 5.49. The
Stress concentration, σ'col / σ'soil, average

1 3 5 7 9
0

4
4 cols - s = 2.0 m
10
E_50 = 30 MPa
12 E_50 = 50 MPa
E_50 = 70 MPa
14
(a)

1 3 5 7 9
0
2
5 cols - s = 1.0 m
10
E_50 = 30 MPa
12 E_50 = 50 MPa
E_50 = 70 MPa
14
(b)

1 3 5 7 9
0
2
9 cols - s = 1.0 m
10
E_50 = 30 MPa
12 E_50 = 50 MPa
E_50 = 70 MPa
14
(c)
Figure 5.48 - Influence of column compressibility upon the variation of stress concentration ratios with column
length for a group of (a) 4, (b) 5 and (c) 9 columns beneath a 3 m square footing
variation of stress concentration ratio with column length for groups of 4 and 5 columns is
shown in Figures 5.49(a) and 5.49(b), respectively. Negligible changes in stress concentration
ratios are observed with increasing column strength for short columns. However, it can be
seen that increasing the strength of columns which are longer than L = 3 m leads to
significant increase in stress concentrations ratios. This suggests that the influence of column
strength upon stress concentration ratios is dependent on the mode of deformation. Column
strength is controlled by the angle of internal friction and this parameter is only mobilised
when columns are in a plastic state. Columns which are punching do not mobilise the column
strength as stresses are concentrated along the column-soil interface and at the base of
columns. Therefore, stress concentration ratios do not increase with column strength for
columns which exhibit punching.
The variation of stress concentration ratio with column length for a group of 9 columns is
shown in Figure 5.49(c). As with the groups of 4 and 5 columns, no variation of stress
concentration ratios is observed with column strength for short columns. However, a decrease
in column strength, from φ = 45° to 40°, leads to a reduction in stress concentration ratios.
Stress concentration ratios are relatively constant with increasing column length for columns
longer than L = 3 m and defined by φ = 40°. This suggests that this configuration of columns
is bulging. In contrast, a slight increase in stress concentration ratios with column length is
observed for columns longer than L = 3 m and defined by φ = 45°. This is consistent with a
punching mode of deformation. Furthermore, no variation of stress concentration ratios is
observed when increasing column strength from φ = 45° to 50°. Therefore, the stronger group
of columns also exhibits punching and, consequently, stress concentration ratios are not
influenced by a change in the column strength.

The influence of the coefficient of lateral earth pressure for the lower Carse clay layer upon
stress concentration ratios for various arrangements of columns (see Figure 5.5) beneath a 3
m square footing is shown in Figure 5.49. It appears that this parameter has a negligible
influence upon stress concentration ratios. However, slight increases in stress concentration
ratios are observed for the groups of 4 and 5 columns at the transition from punching to
bulging, i.e. for column lengths in the range of 3–5 m and 3–7 m for groups of 4 and 5
columns, respectively.

1.0 3.0 5.0 7.0 9.0
0

4
8
4 cols - s = 2.0 m
10
φ = 40°
12 φ = 45°
φ = 50°
14
(a)

1.0 3.0 5.0 7.0 9.0
0
2
8
5 cols - s = 1.0 m
10
φ = 40°
12 φ = 45°
φ = 50°
14
(b)

1.0 3.0 5.0 7.0 9.0
0
2
8
9 cols - s = 1.0 m
10
φ = 40°
12 φ = 45°
φ = 50°
14
(c)
Figure 5.49 - Influence of column strength upon the variation of stress concentration ratios with column length
for a group of (a) 4, (b) 5 and (c) 9 columns beneath a 3 m square footing

1.0 3.0 5.0 7.0 9.0
0

4
8
4 cols - s = 2.0 m
10
K0_soil = 0.75
12 K0_soil = 1.00
K0_soil = 1.25
14
(a)

1.0 3.0 5.0 7.0 9.0
0
2
8
5 cols - s = 1.0 m
10
K0_soil = 0.75
12 K0_soil = 1.00
K0_soil = 1.25
14
(b)

1.0 3.0 5.0 7.0 9.0
0
2
8
9 cols - s = 1.0 m
10
K0_soil = 0.75
12 K0_soil = 1.00
K0_soil = 1.25
14
(c)
Figure 5.50 - Influence of column installation effects upon the variation of stress concentration ratios with
column length for a group of (a) 4, (b) 5 and (c) 9 columns beneath a 3 m square footing
5.5.9 Influence of stiff crust

The influence of the lower Carse clay and stiff crust upon stress concentration ratios for
various arrangements of columns beneath a 3 m square footing is shown in Figure 4.49. It can
be seen that stress concentration ratios increase with column length up to L = 3 m for all soil
profiles. This suggests that punching is the dominant mode of deformation for all short
columns. It can also be seen that stress concentration ratios for columns formed in Profile 2
are higher than Profile 1. This may be related to the increased stiffness of the soil layers at
depth in Profile 2 compared to Profile 1. It was shown earlier that columns formed in Profile
2 yield higher compression ratios as the base of columns are formed in stiffer soil strata.
Therefore, columns formed in Profile 2 carry a larger proportion of the applied load and
hence yield higher stress concentration ratios.
A further increase in stress concentration ratios is observed for columns formed in Profile 3.
This is due to the absence of the stiff crust at the ground surface. The soil at the ground
surface in Profile 3 consists of upper Carse clay, which yields a higher modular ratio
(Ecol/Esoil) than Profile 1. This results in higher levels of vertical stress in the stone columns,
which consequently leads to higher stress concentration ratios.
Stress concentration ratio, σ'col/σ'soil, average

1 3 5 7 9
0

4
10 4 cols - s = 2.0 m
Profile 1
12 Profile 2
Profile 3
14
(a)

1 3 5 7 9
0
2
10 5 cols - s = 1.0 m
Profile 1
12 Profile 2
Profile 3
14
(b)

1 3 5 7 9
0
2
10 9 cols - s = 1.0 m
Profile 1
12 Profile 2
Profile 3
14
(c)
Figure 5.51 - Influence of stiff crust upon the variation of stress concentration ratios with column length for a
group of (a) 4, (b) 5 and (c) 9 columns beneath a 3 m square footing
5.6 Summary of results of FEA: Settlement performance, deformational

behaviour and stress concentration ratios
Various arrangements of columns were analysed over a typical range of area ratios to
examine the relationship between area ratio, column length and column confinement (i.e.
increasing the number of columns). The influence of key design parameters was examined for
a select number of column configurations, which were specifically chosen to cover a wide
range of area ratios. The settlement performance and deformational behaviour of stone
columns was measured at 50 kPa, which was deemed a typical working load for the
Bothkennar test site.
5.6.1 Settlement performance

The settlement performance was measured using a settlement improvement factor, n, which is
defined as the ratio of the settlement for an untreated and treated footing. It was found that
the area ratio, column length and column confinement all have a positive influence upon the
settlement performance of stone columns. Moreover, the influence of area ratio and column
length appears to be inter-related as the influence of column length becomes more
pronounced at low area ratios. A continuous increase in settlement improvement factors with
increasing column length was observed for columns at low area ratios, while the increase
settlement improvement factors with column length tails off for long columns at higher ratios.
Influence of key design parameters:

• A trade off exists between area ratio and column length, as short closely-spaced columns
and long widely-spaced columns can be adopted to achieve a specific settlement
performance. This raises the possibility of optimising the design of stone columns by
minimising the volume of stone; however, other factors such as construction time must
also be taken into account.
• Small benefits can be gained by positioning columns closer to the edge of rigid footings.
This may be attributed to the elevated stress levels in this region. However, the presence
of the stiff crust may be masking the beneficial effects of column position, to a certain
extent.
• The stiffness and strength of stone columns were found to a have significant influence on
the settlement performance of stone columns at low and high area ratios, respectively.
The influence of these parameters is related to the degree of plasticity within columns,
and hence the mode of deformation.
• The coefficient of lateral earth pressure has a positive influence on the settlement
performance of stone columns and this is most pronounced at high area ratios.
• The presence of the stiff crust has a significant influence upon the settlement
performance of stone columns. It was found that the improvement in settlement
improvement factors tails of significantly for homogenous soils samples, i.e. without a
stiff crust. This profile is similar to that adopted in laboratory studies (see Sections 2.2
and 2.3) and may explain why authors have postulated critical lengths in the past.
However, critical lengths are less well defined for more realistic soil profiles.
5.6.2 Deformational behaviour

The deformational behaviour of stone columns was examined in this thesis using newly
defined punching and compression ratios, which essentially are a measure of the normalised
punching at the base of columns and axial strain along the length of columns, respectively.
Columns which are punching transfer a significant proportion of the applied load to the base
of columns and, consequently, this mode of deformation is characterised by high punching
and low compression ratios. In contrast, columns which are bulging deform extensively along
their length and cannot transfer the applied load to the base of columns. Therefore, the
bulging mode of bulging is characterised by low punching and high compression ratios.
Using these ratios, it was possible to determine which mode of deformation (i.e. punching or
bulging) was most influential for each configuration of columns.
Three modes of deformation were identified from the numerical analysis: (i) punching; (ii)
‘block’ failure and (iii) bulging. The punching mode of deformation is observed for all short
columns (i.e. L < 3 m), as an increase in punching ratios is coupled with relatively low
compression ratios. Punching remains the primary mode of deformation for long single
columns at low area ratios, as relatively high punching and low compression ratios are
observed with increasing column length. ‘Block’ failure is characterised by low punching
and compression ratios and this mode of deformation is observed for large groups of columns
at low area ratios. ‘Block’ failure is effectively an extension of punching, as columns act as a
single entity with the surrounding soil and punch uniformly, thus yielding low punching
ratios. The bulging mode of deformation is observed for long columns at high area ratios, as a
significant decrease in punching ratios, coupled with an increase in compression ratios is

observed with increasing column length. In culmination, it appears that the number of
columns does not fundamentally change the mode of deformation, and that the deformational
behaviour of columns is governed by the area ratio and column length.

• The influence of column stiffness and strength is most pronounced for columns at low
and high area ratios, respectively. Columns at low area ratios in more of an elastic state
and, therefore, are more susceptible to changes in the elastic stiffness moduli. In contrast,
the extent of plasticity within groups of columns is more pronounced at high area ratios
and these configurations of columns are more influenced by changes in the strength
characteristics of stone columns.
• The coefficient of lateral earth pressure does not influence the deformational behaviour
of columns, which is consistent with a numerical study conducted by Kirsch (2006).
• The presence of the stiff crust has a significant influence on the deformational behaviour
of stone columns. It was observed that column configurations formed in a homogenous
soil sample exhibit lower punching and higher compression ratios, which suggests a
bulging mode deformation is more pronounced. Column bulging occurs at the point of
lowest lateral resistant, which is generally near the ground surface. Therefore, removing
the stiff crust allows columns to bulge closer to the ground surface.
5.6.3 Stress concentration ratios

The stress concentration ratio is an important parameter which compares the vertical effective
stress in columns (σ’col) to the surrounding soil (σ’soil). It is generally measured at the ground
surface and is defined by two methods in this thesis:
(i) The first method compares the average vertical effective stress in columns (σ’col) to
the average stress in the entire soil area beneath the footing (σ’soil, average).
(ii) The second method compares the average vertical stress in columns (σ’col) to the
column (σ’soil); the width of the zone of influence is equal to the column spacing.
Stress concentration ratios do not increase with column length for an infinite grid of columns.
This is due to symmetry which ensures that the behaviour of an infinite grid of columns is
identical and, consequently, no shear stress develops along the side of the columns near the
ground surface. Therefore, stress concentration ratios are only dependent upon the modular
ratio (Ecol/Esoil). In contrast, stress concentration ratios for a group of stone columns are
dependent upon the mode of deformation. Columns which are punching develop low stress
concentration ratios, which increase with column length. However, no increase in stress
concentration ratios is observed with increasing column length for columns which exhibit a
bulging mode of deformation. Bulging occurs at the point of lowest lateral resistance and
increasing the column length does not enhance the load-carrying capacity of stone columns.

• Stress concentration ratios measured using the method (i) indicate that corner corners
carry a higher proportion of the applied load, followed by edge columns and centre
columns. This may be attributed to the elevated levels of vertical stress beneath the edge
of rigid footings.
• Stress concentration ratios measured using method (ii) suggests that centre columns carry
the highest proportion of the applied load relative to the stress in the surrounding zone of
influence. This reflects the increased levels of lateral confinement associated with centre
columns.
• Increasing the stiffness and strength of columns yields higher stress concentration ratios.
This is most pronounced for columns which exhibit a bulging mode of deformation, as
columns which are punching transfer the applied load along the column-soil interface.
• Increasing the coefficient of lateral earth pressure has a minor influence on the stress
concentration ratios of columns, which is consistent with previous findings from the
mode of deformation.
157
Chapter 6
Results of FEA: Distribution of total shear
strains and characteristic column
behaviour
6.1 Background
The characteristic behaviour of individual columns is examined in the following chapter. The
load-transfer mechanisms are identified for various column configurations at three specific
column lengths: 3, 8 and 13.9 m. The column lengths are chosen on the basis of previous
findings from the deformational behaviour of columns (see Section 5.4) to examine specific
modes of deformation; 3 m long columns are chosen to investigate punching; 8 m long
columns are chosen to investigate a combination of punching-bulging failure and 13.9 m end-
bearing columns are chosen to investigate bulging failure, as punching is precluded.
It is established that the deformational behaviour of columns is highly dependent upon the
support provided by the surrounding soil. The type of stress imparted from columns to the
surrounding soil depends upon the dominant mode of deformation, i.e. columns which are
bulging primarily impart radial stress, while columns which are punching impart a
combination of shear and end-bearing stress. The response of the surrounding soil to each of
these stresses plays a key role in determining the settlement performance of stone columns.
The distribution of shear strains within columns and in the surrounding soil is therefore very
important. This is investigated in the following sections and allows the modes of deformation
identified in Section 5.4, which were determined by compression and punching ratios, to be
verified.
The deformational behaviour of columns is also examined by investigating the distribution of

vertical and horizontal strain within columns. The magnitude and location of maximum strain
within columns defines the deformational behaviour of columns. The influence of column
confinement, column length and area ratio upon these strains is examined in the following
Results of FEA: Distribution of total shear strains and characteristic column behaviour 158
chapter. The stress concentration ratio is an important parameter in determining the

settlement performance of stone columns. It was shown in Section 5.5 that its magnitude at
the surface does not uniquely define the settlement performance of columns. Therefore, the
variation of this parameter with depth is examined.
6.2 Distribution of total shear strains

6.2.1 Total shear strains for an infinite grid of stone columns
The distribution of total shear strain (γ) for an infinite grid of columns at lengths of 3, 8 and
13.9 m is shown in Figures 6.1(a), 6.1(b) and 6.1(c), respectively. It can be seen in Figure
6.1(a) that the majority of shear strain develops beneath the base of 3 m long columns. This
suggests that punching is the dominant mode of deformation. It appears that the distribution
of shear strain beneath the base of columns becomes more uniform at low area ratios, which
is indicative of ‘block’ failure (Figure 6.1(a-i)). In contrast, it can be seen in Figures 6.1(a-ii)
and 6.1(a-iii) that column punching becomes more localised at higher area ratios.
The mode of deformation for 8 m long floating columns is shown in Figure 6.1(b). It can be
seen in Figure 6.1(b-i) that the majority of shear strain continues to develop beneath the base
of columns at low area ratios. It appears that an infinite grid of columns at low area ratios do
not develop shear stress along the sides of columns and transfer the applied load to the base
of columns, which is again consistent with ‘block’ failure. A change in the mode of
deformation is observed for 8 m long columns at higher area ratios in Figures 6.1(b-ii) and
6.1(b-iii). While a fluctuation in shear strain is observed within columns, it is clear that
bulging develops at the top of the lower Carse clay layer (i.e. weakest part of the soil profile).
Columns at higher area ratios experience a loss of lateral confinement which results in
column bulging. The loss of lateral confinement becomes more pronounced with increasing
area ratio and the magnitude of shear strain increases in concert.
No punching is observed in Figure 6.1(c) as the end-bearing columns (L = 13.9 m) are resting
on a rigid stratum. It can be seen that bulging always occurs at the top of the lower Carse clay
layer for all area ratios. The magnitude of column bulging increases at high area ratios, which
reflects the loss of lateral confinement and, also, the higher force taken per column with
increasing area ratio.
(×10-3) (×10-3) (×10-3)
(c-i) (c-ii) (c-iii)

Figure 6.1 - Total shear strains for an infinite grid of columns for column lengths L of (a) 3 m, (b) 8 m and (c)
13.9 m and at area ratios A/AC of (i) 3.5, (ii) 8.0 and (iii) 14.1
6.2.2 Total shear strains for small groups of stone columns

Single columns
The development of total shear strain for a single column at lengths of 3, 8 and 13.9 m is
shown in Figures 6.2(a), 6.2(b) and 6.2(c), respectively. Similar to an infinite grid of
columns, it can be seen in Figure 6.2(a) that the majority of shear strain develops beneath the
base of 3 m long columns. This indicates that column punching is the dominant mode of
deformation for short columns. However, it can also be seen that additional shear strain
develops along the side of single columns. This is not observed for an infinite grid of
columns, as columns within an infinite grid deform equally due to symmetry and
consequently no shear stress develops along the side of columns.
The distribution of total shear strain for 8 m long floating columns is shown in Figure 6.2(b).
It can be seen in Figure 6.2(b-i) that shear strain develops along the side and at the base of
columns at low area ratios, which indicates that punching is the dominant mode of
deformation. It can also be seen that the magnitude of shear strain at the base of columns is
considerably reduced compared to 3 m long columns. This may be explained as longer
columns can generate a larger shear force along the side of columns and also a higher base
resistance is available as columns are formed in a deeper, more competent soil stratum. It can
be seen in Figures 6.2(b-ii) and 6.2(b-iii) that the mode of deformation changes from
punching to bulging with increasing area ratio for 8 m long columns. While some shear strain
develops along the side and at the base of columns, the majority of shear strain develops
within columns near the top of the lower Carse clay layer.
The distribution of shear strains for single end-bearing columns is shown in Figure 6.2(c) and
indicates that no bulging occurs in columns at low area ratios. It appears that columns at low
area ratios develop shear strains along the sides of columns in upper sections. In contrast, the
majority of shear strain develops within columns at higher area ratios, which suggests that
bulging is the primary mode of deformation.
Localised shear zones also develop beneath the edges of pad footings for columns at higher
area ratios. These shear zones develop as a results of the large displacement discontinuity
between the pad footing and the surrounding soil; however, they do not appear to extend
beyond the stiff crust.
(×10-3) (×10-3) (×10-3)
Bulging at top of
Shear strain develops ‘lower Carse clay’
along sides and at layer
base of column

Figure 6.2 - Total shear strains for single columns for column lengths L of (a) 3 m, (b) 8 m and (c) 13.9 m and
at area ratios A/AC of (i) 3.5, (ii) 8.0 and (iii) 14.1
2×2 group of columns

The distribution of shear strain within a 2×2 group of columns for lengths of 3, 8 and 13.9 m
is shown in Figures 6.3(a), 6.3(b) and 6.3(c), respectively. It can be seen in Figure 6.3(a) that
the majority of shear strain develops beneath the base of 3 m long columns, which is similar
to findings from single columns and infinite grids of columns. It can also be seen in Figure
6.3(a-i) that relatively uniform strain develops beneath the base of columns and no shear
strain develops in the central zone of soil bounded by the columns. This suggests that
columns are punching as a ‘block’ with the surrounding soil. The development of shear strain
along the outer sides of columns is also evident in Figure 6.3(a-i). It is difficult to determine
whether this strain is generated from friction along the side of columns or whether it is a
shear zone extending from the corner of the pad footing. It can be seen in Figures 6.3(a-ii)
and 6.3(a-iii) that column punching becomes more localised and columns tend to act more
individually with increasing area ratio. The development of shear zones from the corner of
pad footings to the base of columns is also observed for 2×2 groups at higher area ratios.
As column length increases, Figure 6.3(b-i) illustrates how the magnitude of shear strain
beneath the base of columns reduces and a more uniform distribution of shear strain develops
along the side and at the base of columns. It can be also seen that no shear strain develops in
the central zone of soil bounded by the columns which again suggests that 'block' failure
remains the dominant mode of deformation with increasing column length for 2×2 groups at
low area ratios. In contrast, the mode of deformation appears to change from punching to
bulging with increasing column length for 2×2 groups of columns at higher area ratios.
Column bulging is evident in Figures 6.3(b-ii) and 6.3(b-iii) as the majority of the shear strain
develops within columns, near the top of the lower Carse clay layer. Shear zones also develop
from the corners of pad footings to the point of maximum bulging in columns at high area
ratios. It also appears that shear zones bridge between the points of maximum bulging in
columns and create a zone of undeforming soil between the base of the footing and the points
of column bulging. This zone of undeforming soil was also observed by Muir Wood et al.
(2000) and Wehr (2004) in small scale laboratory tests and numerical studies, respectively.
This zone of soil moves downwards and displaces the surrounding soil, causing the columns
to bulge and bend outwards.
(×10-3) (×10-3) (×10-3)
As A/AC ↑
Uniform strain at Localised failure
base of columns
⇒ ‘block’ failure
at column bases
Development of
shear zone
Low γ in soil
bounded by outer
columns
Development of
Shear strain shear zone
develops in soil

Figure 6.3 - Total shear strains for 2×2 groups of columns for column lengths L of (a) 3 m, (b) 8 m and (c)
13.9 m and at area ratios A/AC of (i) 3.5, (ii) 8.0 and (iii) 14.1
The distribution of shear strains for end-bearing columns is shown in Figure 6.3(c). While no
bulging occurs in columns at low area ratios, it appears from Figure 6.3(c-i) that shear strain
develops in the soil bounded by the columns. This causes columns to bend outwards and
away from the centre of the footing. Shear strains are also observed along the side of
columns, which indicates that the columns may still be transferring the applied load through
side friction. Figures 6.3(c-ii) and 6.3(c-iii) indicate that bulging failure continues to occur for
end-bearing columns at high area ratios. The development of a shear zone, which extends
from the corners of pad footings to the points of maximum bulging, and bridges the soil
between columns is again evident.

The distribution of shear strain for a 3×3 group of columns at lengths of 3, 8 and 13.9 m is
shown in Figures 6.4(a), 6.4(b) and 6.4(c), respectively. It can be seen in Figure 6.4(a-i) that
the majority of shear strain develops beneath the base of columns, which indicates that
punching is the dominant mode of deformation for 3 m long columns at low area ratios.
Similar to previous findings, Figures 6.4(a-ii) and 6.4(a-iii) illustrate that punching becomes
more localised and columns tend to behave individually at higher area ratios. It appears that
columns at low area ratios are behaving as a ‘block’ with the surrounding soil, as shear strain
at the base of columns is relatively uniform and shear strains do not develop in the soil
between columns. The development of a shear zone which extends from the corner of pad
footings to the base of columns is also evident, especially at high area ratios.
The punching mode of failure continues to occur with increasing length for columns at low
area ratios, as shown in Figure 6.4(b-i) for 8 m long columns. While some shear strain
develops within the upper sections of external column, which may result in column bending,
the majority of shear strain develops along the outer side of external columns and at the base
of columns. Therefore, it appears that ‘block’ failure remains the dominant mode of
deformation with increasing column length for a group of columns at low area ratios. In
contrast, it can be seen in Figures 6.4(b-ii) and 6.4(b-iii) that columns at higher area ratios
tend to bulge with increasing column length. Bulging occurs at the top of the lower Carse
clay layer. The shear zone which extends from the corners of pad footings to the points of
maximum bulging in external columns becomes more defined with increasing column length.
The shear zone bridges between the points of bulging in the central and outer columns, which
creates a zone of undeforming soil.
(×10-3) (×10-3) (×10-3)
Uniform strain at Localised failure

base of columns at column bases
⇒ ‘block’ failure
Development of
shear zone
Low γ in soil
bounded by outer
columns

Figure 6.4 - Total shear strains for 3×3 group of columns for column lengths L of (a) 3 m, (b) 8 m and (c) 13.9
m and at area ratios A/AC of (i) 3.5, (ii) 8.0 and (iii) 14.1
The mode of deformation changes from punching to bulging once columns are end-bearing
on a rigid stratum. In contrast to single columns and groups of 2×2 columns, bulging occurs
in 3×3 groups of end-bearing columns at low area ratios (Figure 6.4(c-i)). It can be seen in
Figures 6.4(c-ii) and 6.4(c-iii) that the bulging mode of deformation continues to occur with
increasing column length for 3×3 groups of columns at high area ratios. Furthermore, shear
strain in external columns is more acute, which highlights the positive confining effect
provided to the central column. It also appears that larger deformation occurs at a slightly
shallower depth in external columns compared to central columns. Initially all columns tend
to bulge at the same depth, i.e. at the top of the lower Carse clay. However, the bulging of
external and central columns mobilises the passive resistance of the soil which is bounded by
the central and exterior columns. This increases the lateral resistance at this depth, which
enhances the confinement provided to the central column, and hence forces bulging deeper.
The magnitude of bulging is also reduced, as bulging occurs in a deeper and stiffer soil
stratum. Bulging in external columns continues to occur at the top of the lower Carse clay
layer, as columns bulge and bend outwards towards the unconfined sides.

The distribution of shear strain for a 4×4 group of columns of lengths 3, 8 and 13.9 m is
shown in Figures 6.5(a), 6.5(b) and 6.5(c), respectively. It can be seen in Figure 6.5(a) that all
3 m long columns punch into the underlying soil. Figure 6.5(a-i) shows that shear strain
develops along the outer sides of external columns at low area ratios. It can also be seen in
Figure 6.5(a-i) that relatively uniform shear strain develops beneath the base of columns,
which again suggests that columns at low area ratios are acting as a ‘block’ with the
surrounding soil. It is evident in Figures 6.5(a-ii) and 6.5(a-iii) that column punching
becomes more pronounced with increasing area ratio, which is consistent with findings from
an infinite grid of columns, single columns, 2×2 groups and 3×3 groups of columns.
It can be seen in Figure 6.5(b-i) that punching remains the dominant mode of deformation
with increasing column length for columns at low area ratios. Shear strains develop beneath
the base of columns and along the outer side of external columns. It is also clear that low
levels of shear strain develops in central columns and in the soil bounded by external
columns, which provides further evidence to suggest that columns at low area ratios are
deforming as a ‘block’. However, column bulging becomes more pronounced with increasing
Development of
shear zone

Figure 6.5 - Total shear strains for 4×4 group of columns for column lengths L of (a) 3 m, (b) 8 m and (c) 13.9
m and at area ratios A/AC of (i) 3.5, (ii) 8.0 and (iii) 14.1
area ratio, as shown in Figures 6.5(b-ii) and 6.5(b-iii). Bulging occurs at the top of the lower
Carse clay layer in external columns and occurs slightly deeper in central columns. The
development of a shear zone, which extends from the corner of pad footings to the point of
maximum bulging, is again observed.
The distribution of shear strain for end-bearing columns is shown in Figure 6.5(c). It can be
seen that bulging occurs for all area ratios as column punching is precluded. It appears that
the magnitude and location of bulging is similar for central and external columns at low area
ratios. However, as the area ratio increases it can be seen that the magnitude of shear strain
increases and bulging occurs at shallower depths in external columns compared to central
columns. The development of shear zones from the corner of the pad footing is again evident
and they join together beneath the centre of the footing to form a zone of undeforming soil.
6.3 Characteristic column behaviour

The characteristic behaviour of each column is presented in this section which shows the
relationship between vertical strain (εy), horizontal strain (εh) and stress concentration ratio
within the columns. The difference in behaviour for corner, edge and centre columns is also
highlighted. The strains are reported in the x, y and z directions in PLAXIS 3D Foundation.
However, as the columns are circular it is useful to investigate the vertical and horizontal
strains. The vertical strains (εy) are determined directly from PLAXIS 3D Foundation and the
horizontal strains (εh) are determined as the average of the two orthogonal strains, i.e. εh =
½(εx+ εz).
6.3.1 Infinite grid of columns

Distribution of vertical and horizontal strain
The distribution of vertical strain, horizontal strain and stress concentration ratio with depth
for an infinite grid of columns is shown in Figure 6.6. It can be seen in Figure 6.6(a-i) that
vertical strain is low within 3 m long columns and primarily develops beneath the base of
columns. This coincides with findings from the distribution of total shear strains (see Section
6.2.1) that an infinite grid of 3 m long columns punch as a ‘block’ into the underlying soil. It
can be seen in Figure 6.6(a-ii) that horizontal strain within 3 m long columns is also quite low
and appears to be related to the distribution of vertical strain. The magnitude of vertical and
horizontal strain increases with area ratio, which reflects the loss of lateral confinement and
increase in the total load taken per column with increasing area ratio. In addition, it is
interesting to note that no horizontal strain develops in the soil below the base of columns.
This indicates that the unreinforced soil is subject to one-dimensional boundary conditions,
imposed by the wide area loading.
The distribution of vertical strain with depth for an infinite grid of 8 m long columns is
shown in Figure 6.6(b-i). It can be seen for columns at low area ratios (i.e. A/AC = 3.5) that
the majority of vertical strain develops beneath the base of columns. In contrast, significant
vertical strain develops within columns at higher area ratios. It is interesting to note that
vertical strain within columns increases with area ratio, while the vertical strain beneath the
base of columns remains unchanged. This may be explained as the wide area loading stresses
the soil profile to the full depth and, therefore, induces similar vertical strain in the soil
beneath the base of columns. As with 3 m long columns, it appears that the distribution of
vertical strain within 8 m long columns is related to the horizontal strain (Figure 6.6(b-ii)). It
appears that both vertical and horizontal strain within columns increases with area ratio and
that the maximum horizontal strain occurs at the same depth (z ≈ 2.5 m), close to the top of
the lower Carse clay layer (i.e. the weakest point in the soil profile). A significant amount of
horizontal strain occurs within columns at high area ratios (i.e. A/AC = 8.0 and 14.1) which is
consistent with column bulging.
The distribution of vertical strain for an infinite grid of end-bearing stone columns is shown
in Figure 6.6(c-i). Vertical strain increases with depth to a maximum at z ≈ 2.5 m and
decreases thereafter, eventually becoming negligible at z = 8 m. The horizontal strain is
shown in Figure 6.6(c-ii) and is similar to the distribution of vertical strain with depth. This
indicates that vertical strains are primarily developed from horizontal strains.
Distribution of stress concentration ratio

The distribution of stress concentration ratios in the upper layers of the Bothkennar test site,
for columns at low area ratios (i.e. A/AC = 3.5) is similar for all lengths and is best illustrated
in the inset in Figure 6.6(b-iii). It appears that stress concentration ratios decrease with depth
within the crust. This is related to a decrease in modular ratio (Ecol/Esoil) with increasing
overburden pressure, as shown in Figure 6.7. The stiffness of the soil and column increases
Vertical strain, εy (× 10-3) Horizontal strain, εh (× 10-3) Stress concentration, σ'col/σ'soil

0 -60 -120 -180 -240 0 15 30 45 60 1 3 5 7 9
0
4
Depth, z (m)
10 Unit Cell Unit Cell Unit Cell

A/Ac = 3.5 A/Ac = 3.5 A/Ac = 3.5
12 A/Ac = 8.0 A/Ac = 8.0 A/Ac = 8.0
A/Ac = 14.1 A/Ac = 14.1 A/Ac = 14.1
14

0 -60 -120 -180 -240 0 15 30 45 60 1 3 5 7 9
0
2 σ'col/σ'soil ↑ at
transition to
softer layers
4
Depth, z (m)

A/Ac = 3.5 A/Ac = 3.5 A/Ac = 3.5
12 A/Ac = 8.0 A/Ac = 8.0 A/Ac = 8.0
A/Ac = 14.1 A/Ac = 14.1 A/Ac = 14.1
14

0 -60 -120 -180 -240 0 15 30 45 60 1 3 5 7 9
0
4
Depth, z (m)

A/Ac = 3.5 A/Ac = 3.5 A/Ac = 3.5
12 A/Ac = 8.0 A/Ac = 8.0 A/Ac = 8.0
A/Ac = 14.1 A/Ac = 14.1 A/Ac = 14.1
14
Figure 6.6 - Distribution of (i) vertical strain, (ii) horizontal strain and (iii) stress concentration ratio with depth
for an infinite grid of columns at lengths L of (a) 3 m, (b) 8 m and (c) 13.9 m
Modular ratio, Ecol/Esoil

0 50 100 150 200
0
Depth, z (m)
6
10
A/Ac = 3.5
12 A/Ac = 8.0
A/Ac = 14.1
14
Figure 6.7 - Distribution of modular ratio Ecol/Esoil with depth for an infinite grid of end-bearing stone columns
at various area ratios
with depth according to the power law outlined in equation 3.3. The rate of increase in
stiffness with overburden pressure is controlled by the parameter, ‘m’, which is defined as 0.3
and 1.0 for the column and soil, respectively. The higher ‘m’ value results in a faster increase
in Esoil compared to Ecol with depth. Therefore, Ecol/Esoil reduces with depth in the crust,
which leads to lower stress concentration ratios.
An increase in stress concentration ratios is observed at the transition between the crust and
the softer upper Carse clay layer. This may be explained as Ecol/Esoil increases sharply for this
layer and columns, therefore, carry a larger proportion of the applied load. A slight reduction
in stress concentration ratios with depth is again observed throughout the upper Carse clay
layer, followed by a sharp increase at the transition to the softer lower Carse clay layer. Stress
concentration ratios continue to reduce with depth in the lower Carse clay layer and revert to
unity (i.e. σcol/σsoil = 1) at the base of floating columns, as no extra vertical load can be
sustained in the columns at this level.
It can be seen in Figure 6.6(c-iii) that stress concentration ratios are related to the modular
ratio for an infinite grid of columns at A/AC = 3.5, as stress concentration ratios reduce by
c.55% in the lower Carse clay layer (σcol/σsoil = 5.0 to 2.3) which coincides with a similar
drop in modular ratios for the same layer (Ecol/Esoil = 160 to 68), as shown in Figure 6.7.
While a similar distribution of stress concentration ratio with depth is observed for all short
columns (Figure 6.6(a-iii)), a marked change is observed with increasing area ratio for 8 m
and 13.9 m long columns, as shown in Figures 6.6(b-iii) and 6.6(c-iii), respectively. Despite a
large degree of scatter, it appears that stress concentration ratios reduce with increasing
horizontal strain. It can be seen in Figure 6.6(ii) that horizontal strain increases with
increasing area ratio. This reflects the loss of lateral confinement and, also, the higher load
carried per column with increasing area ratio. It can be seen that sections of columns which
develop high horizontal strain cannot sustain the vertical stress and, consequently, yield low
stress concentration ratios. However, it is interesting to note that in sections of columns
which develop low horizontal strain, stress concentration ratios increase for higher area ratios
(Figure 6.6(c-iii)).
6.3.2 Small groups of columns

Single columns
It can be seen in Figure 6.8(a-i) that vertical strain is very small in 3 m long columns at low
area ratios (i.e. A/AC = 3.5) and that the majority of vertical strain develops in the soil
beneath the base of columns. It also appears that the distribution of vertical strain is closely
related to horizontal strain; it can be seen in Figure 6.8(a-ii) that the majority of horizontal
strain also develops beneath the base of columns. The development of horizontal strain
beneath the base of single columns is in contrast to an infinite grid of columns. This may be
explained as the soil beneath the base of single columns is unconfined and spreads outwards
when subjected to vertical stress, thus generating horizontal strain. A slight increase in
vertical and horizontal strain is also observed with increasing area ratio within 3 m long
columns. This reflects the loss of lateral confinement and increase in the total force taken per
column with increasing area ratio. Finally, it appears that punching is the dominant mode of
deformation for all 3 m long columns as the majority of vertical strain develops beneath the
base of columns.
The distribution of vertical and horizontal strain along 8 m long columns is shown in Figures
6.8(b-i) and 6.8(b-ii), respectively. A marked changed in the distribution of vertical and
horizontal strain within columns is observed with increasing area ratio. It can be seen in
Figure 6.8(b-i) for low area ratios that the majority of vertical strain develops beneath the
base of columns, which suggests that punching remains the dominant mode of deformation.

0 -10 -20 -30 0 5 10 15 1 3 5 7 9
0
4
Depth, z (m)
10 1 column 1 column 1 column

A/Ac = 3.5 A/Ac = 3.5 A/Ac = 3.5
12 A/Ac = 8.0 A/Ac = 8.0 A/Ac = 8.0
A/Ac = 14.1 A/Ac = 14.1 A/Ac = 14.1
14

0 -10 -20 -30 0 5 10 15 1 3 5 7 9
0
2 σ'col/σ'soil ↓ in
section which
4 is bulging
Depth, z (m)

A/Ac = 3.5 A/Ac = 3.5 A/Ac = 3.5
12 A/Ac = 8.0 A/Ac = 8.0 A/Ac = 8.0
A/Ac = 14.1 A/Ac = 14.1 A/Ac = 14.1
14

0 -10 -20 -30 0 5 10 15 1 3 5 7 9
0
section which
4 is bulging
Depth, z (m)

A/Ac = 3.5 A/Ac = 3.5 A/Ac = 3.5
12 A/Ac = 8.0 A/Ac = 8.0 A/Ac = 8.0
A/Ac = 14.1 A/Ac = 14.1 A/Ac = 14.1
14
for a single column at lengths L of (a) 3 m, (b) 8 m and (c) 13.9 m
It is interesting to note that the magnitude of vertical strain at the base of columns is reduced
compared to 3 m long columns. This may be explained as the base of 8 m long columns is
founded in a deeper, more competent soil stratum. It can also be seen in Figure 6.8(b-i) for
columns at high area ratios that vertical strain develops within columns and beneath the base
of columns. This suggests that punching and bulging modes of deformation may be occurring
simultaneously. However, it can be seen that the majority of the vertical strain develops
within columns rather than at the base, which suggests that bulging is the dominant mode of
deformation. The development of high horizontal strains within columns is also evident in
Figure 6.8(b-ii), which is consistent with a bulging mode of deformation. It can be seen in
Figure 6.8(b-i) that the magnitude of vertical strain and the extent of the column which is
strained both increase at high area ratios. This is due to the loss of lateral confinement and
increase in the total load taken per column. Maximum bulging occurs at a depth between 2.0–
2.5 m, which coincides with the top of the lower Carse clay layer and is consistent with the
depth of maximum bulging for an infinite grid of columns.
The distributions of vertical and horizontal strain within end-bearing columns are shown in
Figures 6.8(c-i) and 6.8(c-ii), respectively. It can be seen that no horizontal strain develops in
single columns at low area ratios which indicates that columns are not bulging. However, a
small amount of vertical strain is observed near the top of columns at low area ratios. This
indicates that columns are transferring the applied load through shear stress along the sides of
columns, which is similar to a punching mode of deformation. The bulging mode of
deformation remains prevalent for columns at higher area ratios, as a significant degree of
vertical and horizontal strain is observed in the upper sections of columns. Similar to 8 m
long floating columns, the magnitude of strain and the extent of the column which is strained
both increase with increasing area ratio.

The distribution of stress concentration ratios along the length of single columns is shown in
Figure 6.8(iii). It can be seen in Figure 6.8(a-iii) that stress concentration ratios for 3 m long
columns are high at the top and reduce with depth to unity (i.e. σ'col/σ'soil = 1) at the base of
columns. It is well established that punching is the dominant mode of deformation for short
columns, which implies that columns carry the applied load through a combination of shear
stress and end-bearing resistance. Consequently, the applied vertical stress, and hence the
stress concentration ratio, reduces with depth. Stress concentration ratios approach unity at
the base of columns as no extra vertical stress in columns can be sustained at this point.
A slightly different distribution of stress concentration ratios is observed for 3 m long

columns at high area ratios in Figure 6.8(a-iii). It can be seen in Figure 6.8(a-ii) that
horizontal strain increases with area ratio, which is consistent with the fact that columns at
higher area ratios carry a larger proportion of the applied load. This explains the high stress
concentration ratios at high area ratios in the lower sections of columns. However, the
increase in stress concentration ratio with increasing area ratio is offset by the effects
horizontal strain in the upper sections of columns.
The distribution of stress concentration ratios with depth for 8 m and 13.9 m long columns is
shown Figures 6.8(b-iii) and 6.8(c-iii), respectively. It can be seen that stress concentration
ratios increase with area ratio for sections of columns which are subject to low horizontal
strains. This may be attributable to an increased load that is taken per column with increasing
area ratio and is consistent with the findings from infinite grids of columns. However, it
appears that the magnitude of stress concentration ratios is significantly reduced in the upper
sections of columns, where column bulging occurs. Columns in a state of plasticity cannot
sustain high vertical stress and, consequently, a significant reduction in vertical stress, and
hence stress concentration ratios, occurs. The reduction of stress concentration ratios appears
to be most pronounced at the top of the bulged section of columns.

The distributions of vertical and horizontal strain with depth for 2×2 groups of 3 m long
columns are shown in Figures 6.9(i) and 6.9(ii), respectively. It can be seen in Figure 6.9(a-i)
that the majority of vertical strain develops beneath the base of 3 m long columns, which
suggests that punching is the dominant mode of deformation. It also appears that the vertical
and horizontal strains are closely related; the majority of horizontal strain also develops
beneath the base of columns (Figure 6.9(a-ii)). The magnitude of vertical and horizontal
strains is also observed to increase with increasing area ratio.

0 -10 -20 -30 -40 0 5 10 15 20 1 3 5 7 9
0
4
Depth, z (m)
10 2×2 cols 2×2 cols 2×2 cols

A/Ac = 3.5 A/Ac = 3.5 A/Ac = 3.5
12 A/Ac = 8.0 A/Ac = 8.0 A/Ac = 8.0
A/Ac = 14.1 A/Ac = 14.1 A/Ac = 14.1
14

0 -10 -20 -30 -40 0 5 10 15 20 1 3 5 7 9
0
section which
4 is bulging
Depth, z (m)

A/Ac = 3.5 A/Ac = 3.5 A/Ac = 3.5
12 A/Ac = 8.0 A/Ac = 8.0 A/Ac = 8.0
A/Ac = 14.1 A/Ac = 14.1 A/Ac = 14.1
14

0 -10 -20 -30 -40 0 5 10 15 20 1 3 5 7 9
0
section which
4 is bulging
Depth, z (m)

A/Ac = 3.5 A/Ac = 3.5 A/Ac = 3.5
12 A/Ac = 8.0 A/Ac = 8.0 A/Ac = 8.0
A/Ac = 14.1 A/Ac = 14.1 A/Ac = 14.1
14
for a 2×2 group of columns at lengths L of (a) 3 m, (b) 8 m and (c) 13.9 m
The distribution of vertical and horizontal strain with depth for 2×2 groups of 8 m long
columns is shown in Figures 6.9(b-i) and 6.9(b-ii), respectively. It can be seen in Figure
6.9(b-i) for low area ratios that the majority of vertical strain develops beneath the base of
columns. This is again consistent with a punching mode of deformation. In contrast, it can be
seen that the majority of vertical strain develops within columns at high area ratios, which
suggests that bulging is the dominant mode of deformation. It can be seen in Figure 6.9(b-ii)
that columns at high area ratios develop large horizontal strains, which is consistent with a
bulging mode of deformation.
The distribution of vertical and horizontal strain with depth for 2×2 groups of end-bearing
columns is shown in Figures 6.9(c-i) and 6.9(c-ii), respectively. It can be seen that the
vertical and horizontal strain in columns at low area ratios is quite low, which suggests that
columns do not fail by bulging. This is quite interesting as punching failure is precluded for
end-bearing columns. It appears that closely-spaced columns develop shear stress along the
side of the columns to resist the applied load. A significant increase in horizontal strain is
observed with an increasing area ratio, which indicates that columns at high area ratios are
bulging.

The distribution of stress concentration ratio with depth is shown in Figure 6.9(iii). It can be
seen for all short columns in Figure 6.9(a-iii) that stress concentration ratios are high at the
surface and decrease steadily towards unity at the base of columns. This may be attributed to
the punching mode of deformation as shear stress develops along the side of columns which
results in a reduction of vertical stress with depth. A similar distribution of stress
concentration ratio is also evident for 8 m and 13.9 m long closely-spaced columns (i.e. A/AC
= 3.5) in Figures 6.9(b-iii) and 6.9(c-iii), respectively.
A change in the distribution of stress concentration ratio at higher area ratios (A/AC = 8.0 and
14.1) is observed as column length increases. It can be seen in Figures 6.9(b-iii) and 6.9(c-iii)
that stress concentration ratios reduce with depth in the stiff crust (i.e. z < 0.9). This again
may be explained as columns carry the applied load through shear stress along the sides of
the columns in this section. However, the magnitude of stress concentration ratios at this
depth is lower than stress concentration ratios for closely-spaced columns (i.e. A/AC = 3.5),
which is a direct result of the bulged column in the Carse clay. Columns which are bulging
are in a contained state of plasticity which reduces their load-carrying capacity and, therefore,
the vertical stress in bulged sections of columns. The vertical stress increases with depth in
bulged sections of columns due to overburden stress, which increases column confinement
and hence allows columns to carry a larger proportion of the load. Stress concentration ratios
decrease thereafter as vertical stress is carried through shear stress and also as the modular
ratio reduces with increasing overburden stress.

columns is shown in Figures 6.10(i) and 6.10(ii), respectively. It can be seen in Figure 6.10(i)
that the majority of vertical strain develops beneath the base of columns, which suggests that
punching is the dominant mode of deformation. The distribution of vertical and horizontal
strain for individual columns (i.e. centre, edge and corner columns) with depth is also shown
in Figure 6.10(i) and 6.10(ii). The variation in vertical and horizontal strain between
individual columns is negligible for columns at low area ratios, but becomes more
pronounced as area ratio increases. It can be seen in Figure 6.10(c-i) and 6.10(c-ii) that
vertical and horizontal strains are lowest in centre columns, followed by edge and corner
columns. This highlights the increased levels of confinement associated with central columns.
columns is shown in Figures 6.11(i) and 6.11(ii), respectively. It can be seen in Figure 6.11(a-
i) for 8 m long columns at low area ratios that the majority of vertical strain develops beneath
the base of columns, which suggests that punching remains the dominant mode of
deformation. A change in the mode of deformation is observed with increasing area ratio for
8 m long columns. A significant proportion of horizontal strain is observed within columns at
area ratios of 8.0 and 14.1 in Figures 6.11(b-ii) and 6.11(b-iii), respectively. While a certain
amount of vertical strain develops at the base of columns, it is clear that column bulging is
the dominant mode of deformation. The variation in strain levels with column position is only
observed for columns at low area ratios, where the positive effects of column confinement are
again evident as vertical and horizontal strains are lowest in central columns.
Stress concentation,
Vertical strain, εy (× 10-3) Horizontal strain, εh (× 10-3) σ'col / σ'soil, average
0 -5 -10 -15 -20 0 3 5 8 10 1 3 5 7 9
0
4
Depth, z (m)
8
3×3 cols 3×3 cols 3×3 cols
10 A/Ac = 3.5 A/Ac = 3.5 A/Ac = 3.5
Centre Centre centre
12 Edge Edge edge
Corner Corner corner
14
Stress concentration,
0 -10 -20 -30 -40 0 5 10 15 20 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 8.0 A/Ac = 8.0 A/Ac = 8.0
12 Edge Edge edge
14
Vertical strain, εy (× 10-3) Radial strain, εr (× 10-3) σ'col / σ'soil, average
0 -10 -20 -30 -40 0 5 10 15 20 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 14.1 A/Ac = 14.1 A/Ac = 14.1
12 Edge Edge edge
14
for a 3 m long column, within a 3×3 group for A/Ac of (a) 3.5, (b) 8.0 and (c) 14.1
0 -2 -4 -6 -8 -10 0 1 2 3 4 5 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 3.5 A/Ac = 3.5 A/Ac = 3.5
12 Edge Edge edge
14
0 -5 -10 -15 -20 0 5 10 15 20 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 8.0 A/Ac = 8.0 A/Ac = 8.0
12 Edge Edge edge
14
0 -10 -20 -30 -40 0 10 20 30 40 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 14.1 A/Ac = 14.1 A/Ac = 14.1
12 Edge Edge edge
14
for a 8 m long column, within a 3×3 group for A/AC of (a) 3.5, (b) 8.0 and (c) 14.1
0 -1 -2 -3 -4 -5 0 1 2 3 4 5 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 3.5 A/Ac = 3.5 A/Ac = 3.5
12 Edge Edge edge
14
0 -5 -10 -15 -20 0 5 10 15 20 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 8.0 A/Ac = 8.0 A/Ac = 8.0
12 Edge Edge edge
14
0 -10 -20 -30 -40 0 10 20 30 40 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 14.1 A/Ac = 14.1 A/Ac = 14.1
12 Edge Edge edge
14
for a 13.9 m long column, within a 3×3 group for A/AC of (a) 3.5, (b) 8.0 and (c) 14.1
The distribution of vertical and horizontal strain for 13.9 m long columns is shown in Figures
6.12(i) and 6.12(ii), respectively. Punching failure is precluded for end-bearing columns and
bulging appears to be the dominant mode of deformation for all columns. A significant
increase in the magnitude of vertical and horizontal strain is observed with increasing area
ratio, which may be attributed to the loss of lateral confinement and increase in the total load
taken by columns with increasing area ratio. No variation in the levels of vertical and
horizontal strain is observed with column position for end-bearing columns.

The distribution of stress concentration ratios for 3 m long columns is shown in Figure
6.10(iii). It can be seen that stress concentration ratios reduce steadily from high values, at the
surface, to unity, at the base of floating columns. This is related to the punching mode of
deformation and is consistent with previous findings for short single columns and 2×2
groups. A significant variation in stress concentration ratios with column position is observed
in Figure 6.10(iii). It can be seen that stress concentration ratios at the surface are highest for
corner columns, followed by edge and centre columns. This is due to the position of columns
relative to the footing edge and is analysed in more detail in Section 5.5.2. The variation of
stress concentration ratios is more pronounced at lower area ratios.
6.11(iii). It can be seen for closely-spaced columns in Figure 6.11(a-iii) that stress
concentration ratios reduce steadily with depth, which is consistent with a punching mode of
deformation. However, a change in the variation of stress concentration ratios with depth is
observed for higher area ratios (Figures 6.11(b-iii) and 6.11(c-iii)). It appears that stress
concentration ratios for columns at high area ratios are related to horizontal strain. A
significant proportion of horizontal strain occurs in long columns at high area ratios which
reduces the load-carrying capacity of columns and hence results in lower stress concentration
ratios. This is similar to findings for single columns and 2×2 groups. The largest drop in
stress concentration ratios is evident at the top of the bulged section of columns. It can be
seen that stress concentration ratios increase with depth to the bottom of the bulged section,
due to overburden stress, and then decrease to unity at the base of floating columns.
The variation of stress concentration ratios with column position is also evident in Figure
6.11(iii). Vertical stress at the surface is highest for corner columns, followed by edge and
centre columns. This is again attributable to the position of columns relative to the footing
edge. No change in stress concentration ratio with column position is observed with depth
thereafter for closely-spaced columns. However, it can be seen for columns at higher area
ratios that stress concentration ratios are highest for centre columns, followed by edge and
corner columns over the bulged section of columns. This highlights the positive effects of
column confinement provided to centre columns.
A similar variation of stress concentration ratios for 8 m long columns is observed for end-
bearing columns in Figure 6.12(iii). It can be seen for closely-spaced columns in Figure
6.12(a-iii) that stress concentration ratios reduce steadily with depth, which is consistent with
a punching mode of deformation. The distribution of stress concentration ratios for columns
at higher area ratios is shown in Figures 6.12(b-iii) and 6.12(c-iii). Similar to single columns
and 2×2 groups, it can be seen that the magnitude of stress concentration ratios reduces with
increasing levels of horizontal strain.

columns is shown in Figures 6.13(i) and 6.13(ii), respectively. It appears that the distributions
of vertical and horizontal strain are very similar to 3×3 groups of columns. The majority of
vertical strain develops beneath the base of columns which indicates that punching is the
dominant mode of deformation. A variation in strain levels with column position is also
observed and becomes more pronounced with increasing area ratio. It can be seen that
vertical and horizontal strains are lowest in centre columns, followed by edge and corner
columns. This may be attributed to the loss of lateral confinement associated with outer
columns.
The distribution of vertical and horizontal strain for 8 m long columns is shown in Figures
6.14(i) and 6.14(ii), respectively. It can be seen in Figure 6.14(a-i) that punching remains the
dominant mode of deformation for closely-spaced columns, as the majority of vertical strain
occurs beneath the base of columns. However, a change in the mode of deformation is
observed with increasing area ratio. It can be seen in Figure 6.14(ii) that considerable
horizontal strain occurs in columns, which indicates that column bulging is occurring.
0 -10 -20 -30 -40 0 5 10 15 20 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 3.5 A/Ac = 3.5 A/Ac = 3.5
12 Edge Edge edge
14
0 -20 -40 -60 -80 0 10 20 30 40 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 8.0 A/Ac = 8.0 A/Ac = 8.0
12 Edge Edge edge
14
0 -30 -60 -90 -120 0 10 20 30 40 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 14.1 A/Ac = 14.1 A/Ac = 14.1
12 Edge Edge edge
14
0 -2 -4 -6 -8 0 1 2 3 4 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 3.5 A/Ac = 3.5 A/Ac = 3.5
12 Edge Edge edge
14
0 -10 -20 -30 0 5 10 15 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 8.0 A/Ac = 8.0 A/Ac = 8.0
12 Edge Edge edge
14
0 -15 -30 -45 -60 0 10 20 30 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 14.1 A/Ac = 14.1 A/Ac = 14.1
12 Edge Edge edge
14
0 -2 -4 -6 -8 0 1 2 3 4 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 3.5 A/Ac = 3.5 A/Ac = 3.5
12 Edge Edge edge
14
0 -10 -20 -30 0 5 10 15 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 8.0 A/Ac = 8.0 A/Ac = 8.0
12 Edge Edge edge
14
0 -20 -40 -60 -80 0 10 20 30 40 1 3 5 7 9
0
4
Depth, z (m)
8
10 A/Ac = 14.1 A/Ac = 14.1 A/Ac = 14.1
12 Edge Edge edge
14
for a 13.9 m long column, within a 4×4 group for A/AC of (a) 3.5, (b) 8.0 and (c) 14.1
While it appears that column bulging is the dominant mode of deformation, it can still be
seen that some of the applied load is transferred to the base of columns as vertical strains are
developed in this region. The variation of strain level with column position is clearly visible
for closely-spaced columns. The positive influence of column confinement is evident as
vertical and horizontal strains are lowest for centre columns. The variation of strain level with
column position is less well defined as area ratio increases. It appears that centre columns
exhibit slightly lower vertical and horizontal strain.
The distribution of vertical and horizontal strain for end-bearing columns is shown in Figures
6.15(i) and 6.15(ii), respectively. It can be seen that bulging is the dominant mode of
deformation for all area ratios. A significant increase in vertical and horizontal strain is also
evident with increasing area ratio. The variation of strain levels with column position can also
be seen for columns at higher area ratios. It can be seen in Figure 6.15(b-i) that the magnitude
of vertical strain is highest and maximum strain occurs at the deepest level for centre columns
spaced at A/AC = 8.0. The maximum vertical strain also occurs deeper for centre columns
which are spaced at A/AC = 14.1. However, the maximum vertical and horizontal strain is
highest for corner columns.
Distribution of stress concentration ratios

The distribution of stress concentration for 3 m long columns is shown in Figure 6.13(iii).
The distribution of stress concentration ratios appears to be similar to a 3×3 group. Stress
concentration ratios reduce steadily with depth from high values at the surface, which is
consistent with a punching mode of deformation. The variation of stress concentration ratios
with column position is also evident and this becomes more pronounced with decreasing area
ratio.
6.14(iii). The distribution of stress concentration ratio with depth for closely-spaced columns
again reduces steadily with depth. This indicates that columns are punching into the
underlying soil and is consistent with previous findings from single columns, 2×2 groups and
3×3 groups. However, a marked change in the distribution of stress concentration ratios is
observed with increasing area ratio. It appears that stress concentration ratios are related to
the horizontal strains. The development of horizontal strain in columns reduces the load-
carrying capacity of columns, which results in lower stress concentration ratios. It is also
interesting to note that while stress concentration ratios are lowest for centre columns at the
surface, a reverse in this trend is observed in lower sections of columns which are bulging.
The distribution of stress concentration ratios for end-bearing columns is shown in Figure
6.15(iii). It can be seen that stress concentration ratios follow a similar trend to 3×3 groups of
columns and reduce steadily with depth for closely-spaced columns. However, it can be seen
that the magnitude of stress concentration ratios reduces for higher area ratios, as vertical and
horizontal strain increase. It can be seen that stress concentration ratios increase for centre
columns in sections of columns which are bulging.
6.4 Summary of results of FEA: Distribution of total shear strains and

Three modes of deformation were identified in Chapter 5 using newly defined punching and
compression ratios. However, these ratios only indicate the general mode of deformation, and
it is necessary verify these modes of deformation by conducting a more in-depth analysis of
the load-transfer mechanisms. The distribution of shear strain within columns and in the
surrounding soil is examined for three column lengths, namely 3, 8 and 13.9 m. These
column lengths were specifically chosen to allow the load-transfer mechanism for each mode
of deformation to be examined; 3 m long columns were chosen to investigate punching
failure, 8 m long columns were chosen to investigate a combination of punching and bulging
and 13.9 m long columns were chosen to investigate bulging failure, as punching is
precluded. The distribution of vertical strain, horizontal strain and stress concentration ratio
along the length of columns was also examined.
6.4.1 Distribution of total shear strain

The punching mode of deformation was evident for short columns, and also long single
columns at low area ratios. This is consistent with the distribution of total shear strains, as
significant shear strains develop along the side and beneath the base of columns. ‘Block’
failure becomes more prominent as the number of columns increases for columns at low area
ratios. Shear strain develops along the outer side of exterior columns and beneath the base of
columns, while negligible shear strain develops in the soil bounded by central and exterior
columns. This indicates that the columns and surrounding soil deform as a single entity,
which is consistent ‘block’ failure. Bulging failure is evident as large shear strains develop
within columns. This mode of deformation occurs for long columns at high area ratios. In
addition to the three modes of deformation, the development of shear planes was also
observed. The shear planes extend from the corner of pad footings to the point of maximum
bulging and create a zone of undeforming soil. This is consistent with previous findings by
Muir Wood et al. (2000); however, it is important to note that the shear planes are a
secondary mode of deformation, but may become more pronounced at higher load levels.
6.4.2 Distribution of vertical strain, horizontal strain and stress concentration ratios
The distribution of vertical and horizontal strain is consistent with the shear strains. It was
also observed that punching or bulging modes of deformation are not clearly defined and
some combination of both modes of deformation occurs for all configurations of columns.
The distribution of stress concentration ratios is significantly influenced by the vertical and
horizontal strain within columns. Sections of columns which exhibit large vertical strains are
in a contained state of plasticity and cannot sustain high vertical loads. Consequently, stress
concentration ratios reduce in these regions.
190
Chapter 7
Development of simplified design method
and comparison of findings with previous
research
7.1 Introduction
The results from the numerical analysis regarding the settlement performance, deformational
behaviour and stress concentration ratios for small groups of stone columns are placed in
context to previous research in this chapter.
The loss of lateral confinement for small groups of end-bearing stone columns has been
studied previously, and the findings of these studies are compared with those obtained using
PLAXIS 3D Foundation. The previous research is extended through investigation of the
influence of column confinement upon the settlement performance of small groups of floating
stone columns. A simplified design method is proposed which allows the settlement of small
groups and infinite grids of columns to be related with knowledge of the footing-width to
column-length ratio and column-length to soil-thickness ratio.
Three modes of deformation referred to as "punching", "block failure" and "bulging", were
identified in the previous chapters for small groups of stone columns. The development of
these modes of deformation is compared with small scale laboratory studies, conducted by
Muir Wood et al. (2000) and Black (2006). The importance of capturing the influence of the
stiff crust, a salient feature of soft soil profiles, is highlighted and the existence of a critical
length, proposed by previous laboratory research, is investigated.
The influence of the mode of deformation upon the settlement performance of stone columns
is also investigated in the following chapter. The modes of deformations were identified
using compression and punching ratios and the variation of these parameters with settlement
improvement factors for different configurations of columns is presented and analysed.
Development of simplified design method and comparison of findings with previous research 191
Stress concentration ratios at the surface of an infinite grid of columns are compared with a
database, which consists of field and laboratory measurements, and also analytical design
methods. The distribution of stress concentration ratios with depth in not well understood and
this is examined and compared with analytical design methods.
7.2 Settlement analysis

The behaviour of small groups of stone columns beneath pad footings is quite complex as the
external columns are subject to a loss of lateral confinement and also the vertical stress
decays sharply with depth. Therefore, it is helpful to conduct a three-dimensional numerical
analysis to accurately capture the settlement performance of small groups of stone columns.
The influence of column confinement upon the settlement performance of small groups of
stone columns was examined in Chapter 5. It was found that the degree of column
confinement depends on many parameters such as the number of columns, area ratio and
column length. The influence of the number of columns upon the settlement performance of
stone columns is expressed as a settlement ratio (s/suc), which is defined as the ratio of the
settlement of small groups of columns to that of infinite grids of columns.
7.2.1 Justification of the use of s/suc

It was established in Chapters 5 and 6 that the area ratio and column length play key roles in
determining the mode of deformation of stone columns. While the number of columns has a
positive influence upon the settlement performance of stone columns, it does not
fundamentally change the load transfer mechanisms. Therefore, it is deemed acceptable to
compare the settlement of small groups with infinite grids of columns (s/suc) at similar area
ratios and lengths.
7.2.2 Review of Priebe (1995)

Priebe (1995) is the only analytical method which uses the s/suc term to estimate group
settlements. Design curves are presented in Figure 2.24 which relate s/suc to the number of
columns and the normalised column length (L/d). The main points regarding these design
curves are summarised below:
• The design curves are based on numerous calculations which account for the reduction in
vertical stress with depth beneath small loaded areas and also consider a lower bearing
capacity for the outer ring of columns in the column group.
• No direct reference is made to the footing area (A) in Figure 2.24 and it appears that s/suc
is independent of the area ratio (A/AC).
• However, for a given number of columns, an increase in footing area leads to larger
settlements (s) due to the greater depth of stress. The corresponding increase in A/AC
leads to an increase in suc. Priebe (1995) claims that the increase in s and suc
"compensate" such that s/suc ratios are acceptable for A/AC < 10.
7.2.3 Review of Elshazly et al. (2008a)

Elshazly et al. (2008a) examined the influence of column confinement upon the settlement
performance of stone columns and present graphs of s/suc as a function of the normalised
footing width (B/L). A series of axisymmetric FEA are conducted on two soil profiles:
(i) Layered estuarine deposit, as described by Mitchell & Huber (1985)
- L = 10.8 m; end-bearing on an older marine deposit
- Ecol/Esoil = 1.3–2.6
(ii) Soft clay deposit
- L = 10.8 m and 30 m; end-bearing on dense silty sand
- Ecol/Esoil = 8.5
The loading is applied through a gravel distribution layer to various load levels, ranging from
working loads to ultimate loads. A more detailed description of the FEA by Elshazly et al.
(2008a) is provided in Section 2.3.3.
The variation of s/suc with B/L for the two soil profiles is shown in Figure 7.1:
• It can be seen that s/suc values increase somewhat asymptotically with B/L for profile (i).
For profile (ii) a rapid increase in s/suc values is observed at low B/L followed by a
gradual decrease with increasing B/L.
• It can also be seen that s/suc are much higher for profile (ii). Elshazly et al. (2008a)
attribute this to the high modular ratio (Ecol/Esoil). The Young's modulus of the stone
column (Ecol) is similar for both profiles and a higher Ecol/Esoil indicates that the soil
stiffness is lower for profile (ii). This suggests that the lateral support provided by the
surrounding soil has a significant influence on s/suc.
The design curve developed by Priebe (1995) is adapted to match the numerical analysis by
Elshazly et al. (2008a) (L/d = 10.8/1.06 = 10.2) and is shown in Figure 7.2. As mentioned
previously, Priebe (1995) does not directly relate s/suc to the footing area. However, the
footing width (B) can be determined by assuming the columns are positioned on a square grid
and the distance from the centre-line of the outer row of columns to the edge of the footing is
half the column spacing.
3.0
- Profile (ii); L = 30 m
- Ecol/Esoil = 8.5
2.5
2.0 - Profile (ii); L = 10.8 m

- Ecol/Esoil = 8.5
1.5 - Proile (i); L = 10.8 m

- Ecol/Esoil = 1.3–2.6
1.0
0.5
0.0
0 1 2 3 4 5
Normalised footing width, B/L
Figure 7.1- Variation of settlement ratio with normalised footing width for a layered soil and a soft clay, as
determined by Elshazly et al. (2008a)
3.0
- Profile (ii); L = 30 m * K0 refers to soil surrounding the footing
- K0* = 0.55 (not immediately beside the columns)
2.5
2.0 - Profile (ii); L = 10.8 m

- K0* = 0.55
1.5 - Profile (i); L = 10.8 m

- K0* = 0.38–0.44
1.0
Priebe (1995)
0.5 - K0 = 1.0
0.0
0 1 2 3 4 5
Figure 7.2 - Comparison of settlement ratios determined by Elshazly et al. (2008a) with Priebe (1995)
While the variation of s/suc with B/L for profile (i) is quite similar to Priebe (1995), it can be
seen in Figure 7.2 that Priebe (1995) predicts lower s/suc. This may be related to the lateral
support provided by the surrounding soil:
• Elshazly et al. (2008a) account for column installation effects by increasing K0 in the soil
immediately surrounding stone columns. However, the soil surrounding the footing (i.e.
outside the outer row of columns) is modelled as a normally consolidated soil and the
lateral stress is defined by K0 = 1 – sin(φ') (Jaky, 1944). This results in K0 = 0.38–0.44
and 0.55 for profiles (i) and (ii), respectively.
• Priebe (1995) accounts for column installation effects by adopting K0 = 1.0 for the
surrounding soil. A higher lateral stress state in the surrounding soil provides enhanced
support to the columns and, therefore, results in lower s/suc.
Another possible explanation of the high s/suc values determined by Elshazly et al. (2008a)
relative to Priebe (1995) is the flexible loading conditions. Elshazly et al. (2008a) determine
s/suc from the maximum settlement, which occurs near the edge of footings (Figure 7.3). This
is higher than the settlement of rigid footings, assumed by Priebe (1995), and is another
contributing factor to the higher s/suc values.
Figure 7.3 - Deformed mesh of a foundation with B/L = 3 (case of 10.8 m-thick soft clay layer with a 19×19
column arrangement) (Elshazly et al., 2008a)
7.2.4 Comparison of s/suc for end-bearing columns from PLAXIS 3D Foundation &
Priebe (1995)
The design curve developed by Priebe (1995) in Figure 2.24 is adapted to match the
numerical analyses conducted using PLAXIS 3D Foundation (i.e. L/d = 13.9/0.6 = 23.2) and
the variation of s/suc with the number of columns is shown in Figure 7.4. In order to remain
consistent with the assumptions of Priebe (1995), only s/suc values for end-bearing columns
from PLAXIS 3D Foundation are shown in Figure 7.4.
As expected, it can be seen in Figure 7.4 that an increase in the number of columns leads to
higher s/suc values. An increase in the number of columns results in higher levels of column
confinement and also higher levels of vertical stress with depth (due to larger footing areas).
Therefore, the boundary conditions for groups of columns approach unit cell conditions with
an increasing number of columns and s/suc values tend towards unity.
It can also be seen in Figure 7.4 that s/suc values generally increase with A/AC and this
becomes more significant for larger groups of columns. The design curve developed by
Priebe (1995) agrees favourably with PLAXIS 3D Foundation for A/AC = 3.5, but diverges at
higher area ratios. This confirms that Priebe’s (1995) method is only acceptable for column
groups having low area ratios and is un-conservative for more widely-spaced columns.
0.5
0.4
0.3
Priebe (1995)
A/Ac = 3.5
0.2
PLAXIS 3D Foundation
0.1 A/Ac=3.5
A/Ac=8.0
A/Ac=14.1
0.0
0 5 10 15 20
Number of columns
Figure 7.4 - Comparison of settlement ratios for groups of end-bearing stone columns from PLAXIS 3D
Foundation with Priebe (1995)
7.2.5 Settlement ratios for floating stone columns

Currently, no study has investigated the variation of s/suc for floating stone columns. The
effect of floating columns is introduced into Figure 7.5 using the L/H term, where H is the
thickness of the soil deposit (= 13.9 m for the Bothkennar deposit in this research). The
influence of A/AC upon s/suc values for floating columns is also shown in Figure 7.5. For
clarity, only data specific to three column lengths are presented, namely: L/H = 0.22, 0.58
and 1.00, which correspond to L = 3, 8 and 13.9 m, respectively.
Influence of L/H
For a given B/L, Figure 7.5 shows that s/suc reduces with decreasing L/H. A decrease in L/H
corresponds to a thicker layer of soil beneath the base of columns. The settlement of this
layer is much higher for the infinite grid of columns relative to small groups, due to the
constant stress with depth. Therefore, a decrease in L/H leads to much higher suc and
consequently lower s/suc.
1.0
A/Ac = 3.5
A/Ac = 8.0
0.8 A/Ac = 14.1
Priebe (1995)
A/Ac = 3.5
0.6
L/H = 1.00
0.4
L/H = 0.58
L/H = 0.22
0.2
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Figure 7.5 - Variation of settlement ratios with normalised footing width (B/L) and column length (L/H)
Influence of A/AC
It appears that A/AC only influences s/suc for end-bearing stone columns (L/H = 1.0). It was
shown in Section 5.3.1 that the settlement performance of an infinite grid of stone columns
improves significantly at low A/AC. This is attributed to the increased levels of lateral
confinement and it can be seen in Figure 5.7(ii) that the influence of lateral confinement
becomes most pronounced for long columns (L > 10 m). Therefore, a significant reduction in
suc occurs for long, end-bearing columns at low A/AC, which leads to higher s/suc.
In contrast to end-bearing columns, it appears that s/suc does not vary with A/AC for floating
stone columns (L/H = 0.22 and 0.58). For a given L/H and B/L, it can be seen in Figure 7.5
that s/suc values are similar for different A/AC. The influence of A/AC upon the settlement
performance of stone columns may be hidden in s/suc. The settlement of floating stone
columns is comprised of settlement from two sections: (i) over the length and (ii) beneath the
base of columns. The settlement of the soil beneath the base of columns is quite high for an
infinite grid of columns, due to the constant stress with depth. This results in high suc which
yields low s/suc and hence hides the influence of A/AC upon s/suc.
Influence of B/L
A somewhat linear relationship appears to exist between s/suc and B/L for floating stone
columns. However, care must be exercised in assuming a linear relationship for all B/L
values as it can be seen from the distribution of s/suc with B/L for Priebe (1995) that s/suc
approaches unity somewhat asymptotically at high B/L ratios.
The variation of s/suc with B/L for a wide range of column lengths is shown in Figure 7.6. It
can be seen that the influence of A/AC upon s/suc becomes more pronounced as L/H
increases. This may be explained as the settlement of soil beneath the base of columns
reduces at high L/H values. However, it must be noted that for the majority of column
lengths, A/AC does not influence s/suc and a near-linear relationship exists between s/suc and
B/L for a given L/H.
1.0
L/H = 0.22
Normalised settlement, s/suc
L/H = 0.36
0.8
L/H = 0.58
Priebe (1995) L/H = 0.79
A/Ac = 3.5 L/H = 1.0
0.6
s/suc = 0.76(B/L)
R2 = 0.78 s/suc = 0.51(B/L)
R2 = 0.97
0.4 s/suc = 0.31(B/L)
R2 = 0.99 s/suc = 0.17(B/L)
R2 = 0.97 s/suc = 0.1(B/L)
R2 = 0.95
0.2
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Figure 7.6 - Influence of footing width (B/L) and column length (L/H) upon settlement ratios for groups of
columns
Development of a simple design equation

A simple design equation relating s/suc, B/L and L/H is developed as follows:
(i) Establish a relationship between s/suc and B/L
If s/suc and B/L are assumed to be directly proportional, then their relationship can be
defined by:
/_ = £/> (7.1)
The lines of best fit between s/suc and B/L are shown in Figure 7.6, along with their
corresponding R2 values. R2 is the coefficient of determination and is sometimes referred
to as the proportion of explained variance. It is a measure of the relative predictive power

of a model and varies between 0 and 1; the closer R2 is to 1, the better the predictive
power of the model. It can be seen that all R2 values are close to unity, especially for
floating columns, which justifies the assumption that s/suc is directly proportional to B/L.
(ii) Establish a relationship linking L/H to s/suc and B/L

The slopes of s/suc versus B/L for each L/H (defined as α) are plotted against L/H in
Figure 7.7. A line of best-fit, which assumes a quadratic relationship between α and L/H,
is defined by:
£ = 0.61>/\ + 0.1>/\ + 0.06 (7.2)
It can be seen that R2 = 1 for this line which indicates that the quadratic equation predicts
the relationship between α and L/H perfectly.
1.0
0.8
α = 0.61(L/H)2 + 0.1(L/H) + 0.06
α = (s/suc) / (B/L)
R2 = 1
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Normalised column length, L/H
Figure 7.7 - Relationship between α (which relates the s/suc and B/L) with L/H
(iii) Combine equations 7.1 and 7.2

Equation 7.1 and 7.2 are combined to yield a simple design equation relating s/suc to B/L
and L/H:
/_ = £. />
/_ = 0.61>/\ + 0.1>/\ + 0.06. /> (7.3)
The above equation (7.3) is an interesting and useful finding which allows designers to relate
the settlement of small groups and infinite grids of columns by knowing B/L and L/H.
Therefore, instead of conducting rigorous and time consuming three-dimensional numerical

analyses, the settlement of small groups of columns (s) can be easily determined from
equation 7.3. The settlement of an infinite group of columns (suc) can be quickly determined
from the analytical design methods described in Section 2.5.4. It was shown in Figure 4.16
that Priebe’s (1995) n2 design curve and Pulko & Majes (2005) predict the settlement
performance of stone columns closest to PLAXIS 3D Foundation. However, McCabe et al.
(2009) shows that Priebe's basic design curve (n0) predicts the field measurements of
settlement improvement factors quite well. It may be the case that Priebe's basic design curve
is limited to typical field conditions and cannot capture the settlement performance of stone
columns at high modular ratios (e.g. the Bothkennar soil profile).
It should be noted that equation 7.3 is specific to the benchmark parameters, developed in
Sections 4.4 and 4.6.1. It was shown previously that design parameters such as modular ratios
(Ecol/Esoil) and the column strength (φ) have a significant influence upon the settlement
performance of stone columns. However, it is difficult to predict their influence upon s/suc
values, as the relative influence of Ecol/Esoil and φ upon s and suc may cancel each other.
7.3 Deformational behaviour of stone columns

The deformational behaviour of stone columns was investigated in Chapters 5 and 6. The
following section summarises the main findings from PLAXIS 3D Foundation and identifies
the main parameters governing the deformational behaviour of stone columns. The findings
from other studies regarding the deformational behaviour of stone columns are examined and
placed in context of the new numerical analysis.
7.3.1 Summary of findings from PLAXIS 3D Foundation

The influence of key design parameters upon the deformational behaviour of stone columns
was investigated in Section 5.3. Punching and compression ratios were defined in order to
help identify three distinct modes of deformation: (i) punching, (ii) 'block' failure and (iii)
bulging. The load transfer mechanisms for each of these modes of deformation were verified
in Chapter 6 by analysing the distribution of total shear strain in the column and surrounding
soil and, also, the distribution of stress and strain along the length of columns. It was found
that A/AC and L had a significant influence on the mode of deformation.
A short summary of the main characteristics of each mode of deformation is given below:
(i) Punching
- Typically occurs for short columns and small groups of long columns at low A/AC.
- Characterised by low compression ratios and high punching ratios.
- Shear stress and end-bearing pressure develop along the sides and at the base of columns.
(ii) 'Block' failure
- Typically occurs for large groups of columns at low A/AC.
- Characterised by low compression ratios and low punching ratios.
- Columns act as a single entity and displace as a 'block' into the underlying soil, which
ensures low differential settlement at the base of columns and consequently low
punching ratios.
- Shear stress develops along the outer sides of external columns and end-bearing pressure
at the base of columns.
(iii) Bulging
- Typically occurs for long columns at high A/AC.
- Characterised by high compression ratios and low punching ratios.
- Columns bulge at the point of lowest lateral resistance, which generally occurs near the
ground surface where overburden stresses are lowest.
- Bulging precludes the transfer of the applied load to depth which yields low punching
ratios.
7.3.2 Review of Black (2006)

Black (2006) conducted a series of small scale laboratory tests to investigate the settlement
performance and deformational behaviour of single columns and small groups of columns. A
detailed description of the study is provided in Section 2.3.2 and the main findings are
summarised below:
• The deformed shape of columns was determined upon completion of loading by
removing the aggregate from columns and backfilling the subsequent voids with either a
hot wax or a cement-water mixture. The loading tests were terminated at high strain
levels (footing settlement/diameter = 0.18–0.27), in comparison to PLAXIS 3D
Foundation (footing settlement/diameter = 0.02), which implies that the deformed shape
of columns was specific to ultimate conditions.
• Black (2006) observed that the displacement at the base of single columns becomes
negligible for L/d > 8 and postulates that the mode of deformation changes from
punching to bulging at this point.
• 'Block' failure was observed for small groups of columns whereby the columns acted as a
single entity. The mode of deformation also changes from punching to bulging at L/d =
8, however d is defined as the outer diameter of the group in this instance.
• The findings by Black (2006) are consistent with McKelvey et al. (2004) who conducted
a small scale laboratory study on small groups of stone columns and suggested that a
critical column length exists and lies in the range L/d = 6–10.
7.3.3 Comparison of PLAXIS 3D Foundation with Black (2006)

The penetration at the base of single columns and footing settlement is presented by Black
(2006) and these have been combined to determine compression ratios (defined in equation
5.2). The variation of compression ratios with normalised column length (L/d) from Black
(2006) is compared with PLAXIS 3D Foundation in Figure 7.8.
Compression ratio, (ufooting - ucol)/ufooting

0.0 0.2 0.4 0.6 0.8 1.0
0.0
Normalised column length, L/d
5.0
10.0
15.0
PLAXIS 3D Foundation Black (2006)

20.0 A/Ac=2.5
A/Ac = 3.5
A/Ac = 8.0 A/Ac=3.5
A/Ac = 14.1 A/Ac=5.8
25.0
Figure 7.8 - Comparison of compression ratios from Black (2006) and PLAXIS 3D Foundation for single
columns
Influence of the stiff crust

It can be seen in Figure 7.8 that compression ratios from PLAXIS 3D Foundation are much
lower than Black (2006). This is due in part to the stiff crust at the Bothkennar test site, which
is modelled by PLAXIS 3D Foundation, but not present in the homogeneous soil samples
tested by Black (2006). The influence of the stiff crust upon the deformational behaviour of
stone columns, and more specifically upon compression ratios, is shown in Section 5.4.9.
While the influence of the stiff crust upon the deformational behaviour of single columns was
not investigated exclusively, it can be seen in Figure 7.9 that compression ratios increase
dramatically with the absence of the stiff crust for a group of 4 columns beneath a 3 m square
footing (i.e. A/AC = 8.0) and agree more closely with Black (2006).
The stiff crust tends to confine columns near the surface, where columns are subject to the
lowest overburden stress and hence are most susceptible to bulging. Therefore, the presence
of the stiff crust enhances the load-carrying capacity of columns and allows a greater
proportion of the load to be transferred to the base of columns. This highlights the advantages
of the FEM, which is capable of capturing the presence of the stiff crust and, therefore,
capable of modelling more realistic soil profiles.

0.0 0.2 0.4 0.6 0.8 1.0
0.0
Normalised column length, L/d
5.0
10.0
Absense of crust
increases column
compression
15.0
PLAXIS 3D Foundation Black (2006)

20.0 (4 cols - B = 3 m) A/Ac=2.5
A/Ac = 8.0 A/Ac=3.5
A/Ac = 8.0 (no crust) A/Ac=5.8
25.0
Figure 7.9 - Comparison of compression ratios from Black (2006) with PLAXIS 3D Foundation for soil profiles
with and without a stiff crust
Influence of applied load level

It can be seen in Figure 7.9 that that an increase in the normalised column length (L/d) leads
to an increase in compression ratios. This indicates that a larger proportion of the applied load
is absorbed within columns rather than transferred to the base. Black (2006) defines column
bulging when negligible displacement occurs at the base of columns and observes that the
transition between punching and bulging modes of deformation is clearly defined at L/d = 8
for ultimate conditions.
However, it appears that the transition between punching and bulging is more gradual at
typical working load levels, as determined by PLAXIS 3D Foundation. It can be seen in
Figures 7.8 and 7.9 that compression ratios are less than 1.0 for all floating columns. This
indicates that displacement occurs at the base of all floating columns - even in the case of
long columns at high area ratios (A/AC = 8.0 and 14.1), where bulging is established as the
dominant mode of deformation. This finding is quite interesting and suggests that the mode
of deformation at working load levels is a combination of punching and bulging, with one
particular mode more influential for a given A/AC and L.
It was shown previously in Figure 5.15 that the extent of plasticity within columns increases
for fewer supporting columns beneath a 3 m square footing, i.e. as the average load taken per
column increases. Column plasticity reduces the ability of columns to transfer the applied
load to the base of columns and, consequently, results in higher compression ratios. This may
explain the high compression ratios observed by Black (2006) as the deformed shape of
columns is determined at ultimate conditions, which ensured that columns were in an
advanced state of plasticity. In contrast, columns at low working load levels are in more of an
elastic state and can transfer a larger proportion of the load to the base of columns.
Influence of column length

The main aim of this thesis is to examine the key design parameters influencing the
settlement performance of small groups of stone columns. Therefore, it is necessary to
examine the deformational behaviour at working load levels rather than at ultimate
conditions. The findings from PLAXIS 3D Foundation indicate that some degree of punching
occurs at the base of all floating columns and this is most significant for columns at low
A/AC. It was shown in Section 5.3.2 that increasing the length of columns exhibiting
punching enhances the settlement performance of these stone columns. Therefore, as all
columns exhibit some degree of punching, the settlement performance of columns always
increases with column length and in the general case no critical length exists for the
settlement performance of columns.
Influence of the material parameters

Black (2006) suggests that a modular ratio Ecol/Esoil = 10 is appropriate for the laboratory
study. The modular ratio for an infinite grid of columns at the Bothkennar test site, modelled
in PLAXIS 3D Foundation, is shown in Figure 6.7 and is much higher than Black (2006).
However, it was shown in Section 5.4.6 that the effect of Ecol/Esoil on compression ratios is of
minor significance in the range Ecol = 30–70 MPa.
Black (2006) used a basalt aggregate to form the stone columns. This was subject to direct
shear and triaxial compression tests which indicated angles of internal friction (φ') of 46° and
43°, respectively. These values are very similar to those adopted in PLAXIS 3D Foundation
(φ' = 45°) and therefore would not contribute towards the discrepancy in the results.
7.3.4 Review of Muir Wood et al. (2000)

Muir Wood et al. (2000) conducted a series of small scale laboratory tests to investigate the
influence of column arrangement and length upon the deformational behaviour of groups of
floating stone columns. A detailed description of the study is provided in Section 2.2.2 and
the main points are summarised below:
• The deformed shape of columns was determined at the end of loading by removing the
sand in the columns and backfilling the voids with plaster of Paris. Similar to Black
(2006), the loading tests were terminated at high strain levels (footing
settlement/diameter = 0.3) which again implies that the deformed shape of columns is
specific to ultimate conditions.
• It was observed that the deformational behaviour of columns varies with the position of
columns beneath the circular footings; bulging in central columns was pushed deeper due
to the enhanced lateral support provided by the surrounding columns.
• It was also observed that punching was most pronounced for short, closely-spaced
columns (i.e. low A/AC) and did not occur for columns longer than L/D = 1.0. On this
basis Muir Wood et al. (2000) suggest that a critical length may exist, but the dominant
strain in columns is dependent on the footing diameter rather than on the column
diameter.
7.3.5 Comparison of PLAXIS 3D Foundation with Muir Wood et al. (2000)

Muir Wood et al. (2000) present the penetration at the base of columns as a proportion of
footing settlement (ucol/ufooting) for various configurations of columns. These values are
converted into compression ratios (i.e. 1 - ucol/ufooting) as defined in this thesis and plotted
against column position in Figure 7.10; column position is defined as the normalised radial
distance from the centre of the footing to the column (r/r0), where r0 is the radius of the
footing. For clarity, only data for column lengths L/D = 1.0 is presented in Figure 7.9, as
punching was not observed by Muir Wood et al. (2000) for longer columns.
Compression ratios from PLAXIS 3D Foundation are also presented in Figure 7.10. As with
Black (2006), the laboratory tests described by Muir Wood et al. (2000) are conducted in a
homogeneous soil sample and for consistency, the data from PLAXIS 3D Foundation is
specific to soil profiles with and without a stiff crust. The column configurations from
PLAXIS 3D Foundation correspond to groups of 4, 5 and 9 columns beneath a 3 m square
footing, which correspond to A/AC = 8.0, 6.4 and 3.5, respectively. In the interest of
consistency, only compression ratios for 3 m long columns (i.e. L/B = 1.0) from PLAXIS 3D
Foundation are presented in Figure 7.10. The position of columns for PLAXIS 3D
Foundation is defined as the normalised radial distance of columns from the footing centre
(r/r0), where r0 is the equivalent footing radius (r0 = B/√π).
Normalised radial distance, r/r0

0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2 PLAXIS 3D Foundation

A/Ac = 3.5
A/Ac = 6.4
0.4 A/Ac = 8.0
A/Ac = 3.5
A/Ac = 6.4 no crust
A/Ac = 8.0
0.6
Muir Wood et al. (2000)
Column compression A/Ac = 3.3
increase towards A/Ac = 4.2
0.8 A/Ac = 10.0
footing edge
1.0
Figure 7.10 - Influence of column location upon compression ratios measured by Muir Wood et al. (2000) and
determined PLAXIS 3D Foundation
It can be seen in Figure 7.10 that the findings from PLAXIS 3D Foundation are consistent
with Muir Wood et al. (2000), as higher compression ratios are observed with increasing
A/AC and for columns positioned towards the footing edge.
While the absence of a stiff crust leads to an increase in compression ratios from PLAXIS
3D Foundation, it can be seen that Muir Wood et al. (2000) observed higher compression
ratios for all A/AC. This may be explained as follows:
• The angle of shearing resistance (φ' ≈ 30°) for the sand columns in the laboratory tests is
much lower than PLAXIS 3D Foundation (φ' = 45°). It was shown in Section 5.4.7 that
decreasing φ from 50° to 40° had a significant influence on compression ratios, with an
increase of compression ratios in the order of 0.05–0.15 observed for 3 m long columns.
• The deformed shape of columns was determined at ultimate conditions. As with Black
(2006), loading was terminated at high strain levels which ensured the columns were in
an advanced state of plasticity and hence yield high compression ratios.
Influence of footing width

Muir Wood et al. (2000) suggest that a critical length may exist for small groups of stone
columns, but that the dominant strain in columns is dependent on the footing diameter (D),
rather than the column diameter. The laboratory tests conducted by Muir Wood et al. (2000)
were simulated with a plane strain FEA by Wehr (2004), who also found that the critical
length is dependent on the footing diameter. Wehr (2004) observed a marked change in the
deformational behaviour of columns at L/D = 1.5.
However, these findings are in contrast to McKelvey et al. (2004) and Black (2006) who
suggest that a critical length may exist in the range L/d = 6–10, and is not dependent on the
footing width. In light of these more recent findings, the data presented by Muir Wood et al.
(2000) was re-examined in this thesis plotted in Figure 7.11.
Muir Wood et al. (2000) observed punching and bulging for different configurations of
columns and these are plotted against L/D in Figure 7.11(i). The mode of deformation
changes from punching to bulging in the range L/D = 1.0–1.5, which prompted the authors to
postulate that the mode of deformation is controlled by the footing diameter. However, the
observed modes of deformation are plotted against L/d in Figure 7.11(ii) and it is clear that,
with the exception of two tests out of total of 14, that the mode of deformation changes from
punching to bulging in the range L/d = 9.1–13.6. This is consistent with more recent findings.
The results from this thesis indicate the area ratio and column length, rather than the footing
width, dictate the mode of deformation. It was established in Chapter 5 that different
configurations of area ratio and column length yield either a punching-type (i.e. punching of
‘block’ failure) or a bulging mode of deformation. For example, it was shown in Figure
6.3(b-i) that a 2×2 group of 8 m long columns at A/AC = 3.5 (i.e. B = 2 m; L/B = 4) exhibit a
punching mode of deformation, while a 4×4 group of 8 m long columns at A/AC = 14.1 (i.e.
B = 8 m; L/B = 1) exhibit a bulging mode of deformation (Figure 6.5(b-iii)).
0.0 0
D=100 mm D=100 mm
0.2 2
d d
0.4 4
0.6 d = 11 mm Punching d = 11 mm
6 d = 17.5 mm
Punching d = 17.5 mm
0.8
8
1.0
10 Possible
1.2 Postulated critical length
critical length 12 at L/d = 9.1
1.4
1.6 14 Bulging
1.8 Bulging 16
L/D L/d
(i) (ii)
Figure 7.11 - Observations of punching and bulging for groups of columns, reported by Muir Wood et al.
(2000), plotted against (i) L/D and (ii) L/d
7.4 Influence of deformational behaviour upon settlement performance

The settlement performance and deformational behaviour of small groups of stone columns
were analysed separately in Sections
Sections 5.3 and 5.4, respectively. Compression and punching
ratios were defined to help identify different modes of deformation for various
configurations of columns, as summarised in Section 7.2.1. It was shown previously that an
increase in column length leads to higher compression ratios, lower punching ratios (for L >
3 m) and higher settlement improvement factors (n). However, the variation
variat of these
parameters with column length depends on the mode of deformation. Essentially,
Essentially the aim of
this section is to form a link between the modes of deformation and the settlement
performance of columns at typical working load levels.
Column punching (L ≤ 3 m and small groups of columns at A/AC = 3.5)

Column punching typically occurs for short columns (L ≤ 3 m) and for small groups of
closely-spaced
spaced columns. The occurrence of column punching is most evident for single
columns at A/AC = 3.5, as shown in Figure 7.12(a). It can be seen in Figure 7.12(a-i)
7.12(a that the
influence of compression ratios upon n values is more significant for columns which are
punching compared to columns at higher area ratios. Furthermore, it appears that the
influence
luence of compression ratios upon n values is most pronounced for short columns (i.e. L <
3 m), as the largest increase in n values with compression ratios is observed for this range of
column lengths (compression ratios and n values are 0 and 1, respectiv
respectively,
ely, for unreinforced
footings, i.e. L = 0 m). Similarly, it can be seen in Figure 7.12(a
7.12(a-ii)
ii) that the influence of
punching ratios upon n values is most pronounced for columns which are punching.
1.0 0.20
1 col 1 col
A/Ac = 3.5 A/Ac = 3.5
0.8 A/Ac = 8.0 A/Ac = 8.0
Compression ratio
0.15 L=3m
Punching ratio
11 m A/Ac = 14.1 A/Ac = 14.1
5m
0.6
8m 0.10
8m
0.4
5m 11 m
0.05
0.2
L=3m
0.0 0.00
1.0 1.5 2.0 2.5 3.0 3.5 1.0 1.5 2.0 2.5 3.0 3.5
(a-i) (a-ii)
1.0 0.20
2×2 2×2
A/Ac = 3.5 A/Ac = 3.5
0.8 A/Ac = 8.0 A/Ac = 8.0
Compression ratio
0.15
Punching ratio
A/Ac = 14.1 A/Ac = 14.1
0.6 11 m L=3m
0.10 5m
0.4 8m
8m
11 m
0.05
0.2
5m
L=3m
0.0 0.00
1.0 1.5 2.0 2.5 3.0 3.5 1.0 1.5 2.0 2.5 3.0 3.5
(b-i) (b-ii)
1.0 0.20
3×3 3×3
A/Ac = 3.5 A/Ac = 3.5
0.8 A/Ac = 8.0 A/Ac = 8.0
Compression ratio
0.15
Punching ratio
A/Ac = 14.1 A/Ac = 14.1

0.6
11 m 0.10 L=3m
0.4 5m
8m 8m
0.05 11 m
0.2
5m
L=3m
0.0 0.00
1.0 1.5 2.0 2.5 3.0 3.5 1.0 1.5 2.0 2.5 3.0 3.5
(c-i) (c-ii)
1.0 0.20
4×4 4×4
A/Ac = 3.5 A/Ac = 3.5
0.8 A/Ac = 8.0 A/Ac = 8.0
Compression ratio
0.15
Punching ratio
A/Ac = 14.1 A/Ac = 14.1

0.6
0.10 L=3m
11 m
0.4
5m
8m 11 m
8m 0.05
0.2
5m
L=3m
0.0 0.00
1.0 1.5 2.0 2.5 3.0 3.5 1.0 1.5 2.0 2.5 3.0 3.5
(d-i) (d-ii)
Figure 7.12 - Variation of (i) compression ratio and (ii) punching ratio with settlement improvement factor for
(a) 1 column, (b) 2×2, (c) 3×3 and (d) 4×4 groups
'Block' failure (Large groups of columns at A/AC = 3.5)

'Block' failure is similar to column punching and typically occurs for large groups of closely-
spaced columns. The punching mode of deformation, evident in Figure 7.12(a) for single
columns, transforms into 'block' failure as the number of columns increases. 'Block' failure is
most clearly defined for a 4×4 group of columns at A/AC = 3.5, as shown in Figure 7.12(d).
It can be seen in Figures 7.12(a-i to d-i) that the line corresponding to A/AC = 3.5 moves
outwards and downwards, which indicates that an increase in n values and decrease in
compression ratios occurs for larger groups of columns. This suggests that n values for
columns exhibiting 'block' failure are more sensitive to changes in compression ratios. The
influence of 'block' failure upon the variation of punching ratio with n values is shown in
Figure 7.12(d-ii). It can be seen in Figures 7.12(a-ii to d-ii) that as the number of columns
increase, punching ratios reduce significantly for short columns. However, punching ratios
remain relatively unchanged with an increasing number of long columns. Therefore, it
appears that n values for columns exhibiting 'block' failure are also very sensitive to
punching ratios.
Column bulging (L > 3 m, A/AC = 8.0 and 14.1)

Column bulging typically occurs for long widely-spaced columns and its influence upon the
variation of compression ratios with n values can be seen in Figures (a-i to d-i). It appears
that the increase in n values with column length for A/AC = 14.1 is much lower than for
punching or 'block' failure (A/AC = 3.5). While the increase in n values with compression
ratios becomes slightly more pronounced at closer spacings (A/AC = 8.0) and for larger
groups, it appears that n values are not strongly dependent upon compression ratios for the
bulging mode of deformation. The variation of punching ratios with n values is shown in
Figure 7.12(a-ii to d-ii). It can be seen that the variation of n values with punching ratios is
similar for all configurations of columns at A/AC = 14.1. Moreover, the near-vertical curve
between punching ratios and n values suggests that n values are not strongly dependent upon
punching ratios for the bulging mode of deformation.
Implications for design of stone columns

It was shown that the settlement performance of stone columns depends upon many factors
such as the configuration of columns, material parameters and the soil profile. In order to
achieve an optimum column arrangement, a trade-off exists between area ratio and column
length, i.e. short closely-spaced columns or long widely-spaced columns. The relationship
between area ratio and column length is an inter-dependent one, as different configurations of
columns result in different modes of deformation. Each mode of deformation has its own
characteristics which makes its advantageous in different soil profiles.
It was shown in Figure 7.12 that the settlement performance of columns which exhibit
punching, and especially 'block' failure, are most sensitive to changes in compression ratios.
This is consistent with previous findings that the influence of column stiffness upon the
settlement performance of small groups of columns is most pronounced for these modes of
deformation. It was also observed that columns which exhibit punching and 'block' failure are
most sensitive to changes in punching ratios. This reflects the load-transfer mechanism, as a
large proportion of the applied load is transferred to the base of columns. Therefore these
modes of deformation are not highly dependent upon the properties of the surrounding soil
and are most efficient is soil profiles with weak upper layers.
In contrast, it can be seen that influence of compression and punching ratios upon n values
diminishes for the bulging mode of deformation. Column bulging is highly dependent upon
the lateral support provided by the surrounding soil and generally occurs near the ground
surface, where overburden stresses are lowest. The bulging mode of deformation makes
efficient use of the high stiffness properties of the stiff crust. Columns are confined in the
upper layers which tends to push bulging deeper and hence increase the load-carrying
capacity of columns.
7.5 Stress concentration ratio

7.5.1 Stress concentration ratios for infinite grids of stone columns
Comparison of stress concentration ratios with laboratory and field data
Stress concentration ratios (at the surface) from PLAXIS 3D Foundation are compared with
laboratory and field measurements in Figure 7.13. The data presented by Aboshi et al. (1979)
and Greenwood (1991) is from field measurements which pertain to large diameter sand
columns and stone columns in soft clay, respectively. A large range of stress concentration
ratios was observed by Aboshi et al. (1979), however, Barksdale & Bachus (1983) suggest
that stress concentration ratios typically lie in the range of 2.5–5.0. In addition to the field
data presented in Figure 7.13, stress concentration ratios from laboratory tests conducted by
Charles & Watts (1983) are also shown. It can be seen in Figure 7.13 that stress concentration
ratios predicted by PLAXIS 3D Foundation agree quite well with field and laboratory data.
15
Stress concentration ratio, σcol/σsoil

Aboshi et al. (1974)
13 Charles & Watts (1983)
Munfakh et al. (1983)
11 Greenwood (1991)
7
Barksdale & Bachus (1983)
5
1
2 4 6 8 10 12 14 16
Area ratio, A/AC
Figure 7.13 - Comparison of stress concentration ratios for an infinite grid of end
end-bearing
bearing columns with field
data
Comparison of stress concentration ratios with analytical design methods

Stress concentration ratios determined from PLAXIS 3D Foundation at the surface are
compared with analytical design methods in Figure 7.14. The data from the numerical
analyses are specific to infinite grids of end
end-bearing columns and are therefore consistent
with the assumptions of the analytical design methods.
It can be seen in Figure

gure 7.14 that the Balaam & Booker (1981) method greatly over-estimates
over
stress concentration ratios. This is most significant at high area ratios and is due to the
assumption of elastic behaviour for the columns (Castro & Sagaseta, 2009). In contrast,
PLAXIS
IS 3D Foundation and the other analytical design methods account for plastic
deformations which develop in columns and therefore predict lower, more realistic stress
concentration ratios.
It can also be seen in Figure 7.14 that Priebe (1995) generally predicts
dicts lower stress
concentration ratios than Pulko & Majes (2005),
(2005) especially at low area ratios. This is due to
the different assumptions inherent in each design method. Priebe (1995) assumes columns are
in an active state throughout loading, while Pulko & Majes (2005) account for the elastic
behaviour of columns at low load levels. Stress concentration ratios
ratio are much higher for
columns in an elastic state and consequently Pulko & Majes (2005) predict higher stress
concentration ratios than Priebe (1995). This is most pronounced at low area ratios as
columns are better confined and hence carry a larger proportion of the applied load when in
an elastic state.
Perhaps the most important finding from Figure 7.14 is that both Priebe (1995) and Pulko &
Majes (2005) predict higher surface stress concentrations ratios than PLAXIS 3D Foundation.
This is quite interesting given that both of these design methods predict similar settlement
improvement factors to PLAXIS 3D Foundation (see Section 4.6.4). Considering the equation
of equilibrium developed by Aboshi et al. (1979) in equation (2.27), it seems intuitive that
higher stress concentration ratios at the surface would yield higher settlement improvement
factors. However, this is not the case and it appears that the distribution of stress
concentration ratios with depth, rather than at the surface, may have more of an influence
upon the settlement performance of stone columns and thus merits further investigation.
15
Stress concentration ratio, σcol/σsoil
13 Priebe (1995)
Pulko & Majes (2005)
11 Balaam & Booker (1981)
1
2 4 6 8 10 12 14 16
Area ratio, A/Ac
Figure 7.14 - Comparison of stress concentration ratios at the surface for an infinite grid of end-bearing
columns with analytical design methods
Variation of stress concentration ratio with depth

The variation of stress concentration ratios with depth from PLAXIS 3D Foundation is
compared with Priebe (1995) and Pulko & Majes (2005) in Figures 7.15(i) and 7.15(ii),
respectively. The distribution of stress concentration ratios with depth from PLAXIS 3D
Foundation for an infinite grid of end-bearing columns was presented and discussed
previously in Section 6.3.1. The extent of plasticity within closely-spaced columns is quite
low and as a result, stress concentration ratios follow a similar variation with depth to
modular ratios. The maximum stress concentration ratio occurs at z = 1.9 m, which
corresponds to the top of the lower Carse clay layer (i.e. the weakest part of the soil profile).
Stress concentration ratios are lower in the overlying soil, even though these soil layers are in
an elastic state. This may be explained as an infinite grid of columns do not develop shear
stress along their sides and hence the maximum vertical load which columns can carry at the
surface is limited by the vertical stress where bulging occurs. The extent of plasticity within
columns at higher area ratios (i.e. A/AC = 8.0 and 14.1) is more significant and despite the
large scatter, a constant distribution of stress concentration ratios with depth is observed in
Figure 7.15 for yielded sections of columns. It can also be seen that stress concentration
ratios reduce with depth in the lower sections of columns which are in an elastic state.
Stress concentration ratio, σ'col/σ'soil Stress concentration ratio, σ'col/σ'soil

1 3 5 7 9 1 3 5 7 9
0 0
2 2
4 4
Depth, z (m)
Depth, z (m)
6 6
8 8
PLAXIS 3D Foundation PLAXIS 3D Foundation
A/Ac = 3.5 A/Ac = 3.5
10 A/Ac = 8.0 10 A/Ac = 8.0
A/Ac = 14.1 A/Ac = 14.1
Priebe (1995) Pulko & Majes (2005)
12 A/Ac = 3.5 12 A/Ac = 3.5
A/Ac = 8.0 A/Ac = 8.0
A/Ac = 14.1 A/Ac = 14.1
14 14
(i) (ii)
Figure 7.15 - Comparison of stress concentration ratios with depth from PLAXIS 3D Foundation with (i) Priebe
(1995) and (ii) Pulko & Majes (2005) for an infinite grid of end-bearing columns
Comparison with Priebe (1995)

It can be seen in Figure 7.15(i) that Priebe (1995) predicts higher stress concentration ratios
than PLAXIS 3D Foundation at the surface. This may be explained as Priebe (1995) assumes
that columns are in an active state throughout loading, in comparison to PLAXIS 3D
Foundation where sections of columns within the crust are in an elastic state. As mentioned
earlier, stress concentration ratios at the surface of an infinite grid of columns are limited by
the vertical stress at the location of column bulging and are therefore lower than Priebe
(1995). This highlights the advantage of the FEM which is capable of capturing all the factors
influencing the behaviour of columns, rather than treating each layer individually.
It can be seen in Figure 7.15(i) that Priebe (1995) predicts lower stress concentration ratios
than PLAXIS 3D Foundation within yielded sections of columns. This is most pronounced
for closely-spaced columns (A/AC = 3.5) at z = 1.9 m and may be explained as Priebe (1995)
fails to account for stress concentration ratios developed when columns are in an elastic state
(i.e. at low load levels). However, stress concentration ratios for closely-spaced columns
reduce at a faster rate with depth for PLAXIS 3D Foundation, as the lower sections of
columns are in an elastic state. It can also be seen in Figure 7.15(i) that stress concentration
ratios at higher area ratios (i.e. A/C = 8.0 and 14.1) are relatively constant with depth in
yielded sections of columns compared to Priebe (1995). This is again attributed to the
assumption by Priebe (1995) that columns are in an active state throughout loading.
It is interesting to note that Priebe (1995) generally under-predicts stress concentration ratios
but still yields similar settlement improvement factors to PLAXIS 3D Foundation. This may
explained as stress concentration ratios determined from Priebe (1995) do not contain the
modification factor for overburden stress, which is taken into account when determining
settlement improvement factors (n2).
Comparison with Pulko & Majes (2005)

As with Priebe (1995), it can be seen in Figure 7.15(ii) that Pulko & Majes (2005) predict
higher stress concentration ratios in sections of columns located within the crust compared to
PLAXIS 3D Foundation. This is again explained as the vertical stress at the surface of an
infinite grid of columns in PLAXIS 3D Foundation is dependent upon the vertical stress at
the location of column bulging. In contrast, Pulko & Majes (2005) treat each layer
individually and columns are in an plastic state in the crust.
It can be seen in Figure 7.15(ii) that Pulko & Majes (2005) predict higher stress concentration
ratios than Priebe (1995) and capture the variation of stress concentration ratios with depth
much closer to PLAXIS 3D Foundation. This highlights the importance of considering the
loading history and accounting for stress concentration ratios developed when columns are in
an elastic state.
It also appears that Pulko & Majes (2005) slightly over-estimate stress concentration ratios
compared to PLAXIS 3D Foundation and this may be due to the high modular ratios adopted
for the analytical design methods. The soil profile adopted for the analytical design methods
consists of three layers, each defined by a single Young's modulus. In contrast, PLAXIS 3D
Foundation accounts for the stress dependency of soil stiffness and captures the increase of
soil stiffness with depth. The modular ratios adopted for each layer in Pulko & Majes (2005)
are higher than PLAXIS 3D Foundation which increases the magnitude of stress
concentration ratios and also extent of plasticity within columns.
The influence of higher stress concentration ratios upon the settlement performance of stone
columns is evident in Figure 4.16, as Pulko & Majes (2005) predict higher settlement
improvement factors than PLAXIS 3D Foundation at low area ratios (A/AC = 3.5). This may
be explained as the majority of columns are in an elastic state at low area ratios and are
therefore more susceptible to differences in modular ratios. However, influence of the higher
stress concentration ratios is negligible at higher area ratios, as Pulko & Majes (2005) predict
similar settlement improvement factors to PLAXIS 3D Foundation (Figure 4.16). The extent
of plasticity within columns is more pronounced at higher area ratio and hence the influence
of modular ratios upon the settlement performance of columns reduces.
The distribution of stress concentration ratios with depth clearly indicates when columns are
in an elastic and plastic stress state. This is quite useful for stability analyses where a
rotational failure surface may intersect columns at different depths. However, it difficult to
form a direct relationship between stress concentration ratio and settlement improvement
factors for the following reasons:
(i) Settlement improvement factors are based on the increment of vertical stress (∆σcol and
∆σsoil) with depth rather than the total vertical stress(σcol/σsoil), which is used to define
stress concentration ratios, i.e. σcol/σsoil == (∆σcol + γ'col.z)/(∆σsoil + γ'soil.z). Furthermore,
the increment of vertical stress reduces with depth beneath small loaded area.
(ii) The distribution of stress concentration ratios only shows the stress state at the end of
loading and does not capture the proportion of the load applied to columns in an elastic
and plastic state. The majority of footing settlement is comprised from plastic strain and
it is necessary to predict the yield load.
Therefore it is difficult to apply a simple equation, such as that developed by Aboshi et

al. (1979) in equation (2.27), to estimate the settlement performance a small groups of
stone columns from the distribution of stress concentration ratios with depth.
216
Chapter 8
Conclusions
8.1 Introduction
A series of three-dimensional FEA were undertaken to examine the influence of key design
parameters upon the settlement performance and deformational behaviour of small groups of
stone columns. PLAXIS 3D Foundation is adopted for this research in conjunction with the
advanced elastic-plastic Hardening Soil model. A set of material parameters and soil profile
were developed for the well characterised Bothkennar test site, which formed the basis of this
research. The key design parameters investigated were area ratio, column length,
confinement, stiffness, strength, installation effects and the presence of a stiff crust. The main
conclusions are summarised in this chapter.
8.1.1 Numerical modelling preliminaries

In advance of performing the primary analyses of this thesis, it was necessary to arrive at a
reliable methodology to capture the behaviour of small groups of stone columns in soft clay.
The main findings from the preliminary analysis are as follows:
- The long term settlement performance of small groups of stone columns at typical working
loads can be accurately modelled using drained analyses in PLAXIS 3D Foundation, as
these were shown to give a similar output to more time-consuming coupled analyses (i.e.
undrained loading followed by a consolidation analysis).
- Simulating column installation effects by applying large strain cavity expansions cannot
be accurately implemented in a FE program which is implicitly based on small strain
theory. However, an alternative approach to simulate column installation effects is to
increase the coefficient of lateral earth pressure (K0) in the surrounding soil.
- The analytical design method developed by Balaam & Booker (1981) grossly over-
estimates settlement improvement factors, due to the assumption of linear elasticity for
stone. It was shown that Priebe's (1995) (n2) modified design curve (when accounting for
an increased K0 in the surrounding soil, which is not a standard adjustment) agrees quite
well with PLAXIS 3D Foundation. Pulko & Majes (2005) is a more rigorous theoretical
Conclusions 217
solution than Priebe (1995), and was shown to predict a similar settlement response to
PLAXIS 3D Foundation.
8.1.2 Settlement performance

The settlement performance of small groups of stone columns in the Bothkennar soil profile
was determined at 50 kPa, a typical working stress for columns in soft soil. It was found that
an increase in the area ratio, column length and the number of columns all have a positive
influence on the settlement performance of small groups of stone columns. Moreover, the
influence of column length and area ratio are somewhat coupled, as the effect of column
length becomes more pronounced at low area ratios. A continuous increase in settlement
improvement factors is observed with increasing column length at low area ratios, while the
increase in settlement improvement factors tails off with increasing column length at high
area ratios.
Influence of key design parameters and practical considerations

- For a given number of columns, their proximity to the footing edge has a relatively minor
influence on the settlement performance of stone columns.
- Increasing the column stiffness and strength enhances the settlement performance of stone
columns. The influence of these parameters is linked to the mode of deformation, as
increasing the column stiffness and strength is most significant at low and high area ratios,
respectively. This is discussed in more detail in the following section.
- Accounting for column installation effects by increasing the coefficient of lateral earth
pressure (K0) in the surrounding soil improves the settlement performance of stone
columns. The influence of K0 is most noticeable at high area ratios.
- The presence of the stiff crust also had a significant influence on the settlement
performance of stone columns and this is discussed in the following section.
8.1.3 Mode of deformation

The modes of deformation for various configurations of columns were identified with the aid
of new parameters, referred to as punching and compression ratios. Three distinct modes of
deformation were observed: (i) punching, (ii) 'block' failure and (iii) bulging. The occurrence
of each mode of deformation was verified by analysing the distribution of shear strains within
Conclusions 218
columns and the surrounding soil. The main characteristics of each mode of deformation are
outlined below:
(i) Punching:
- Characterised by high punching and low compression ratios.
- Typically occurs for short columns at all area ratios and small groups of columns at low
area ratios.
- High shear strains develop along the sides and beneath the base of columns which
exhibit punching.
(ii) 'Block' failure:
- Characterised by low punching and high compression ratios.
- Most prominent in large groups of columns at low area ratios.
- Columns act as a single entity and punch uniformly into the underlying soil, thus
resulting in low punching ratios. 'Block' failure is an extension of the punching mode of
deformation, as both modes of deformation develop similar load transfer mechanisms.
- High shear strains develop along the external sides of outer columns and beneath the
base of columns. No shear strains develop in the central zone of soil bounded by
columns, which confirms that columns punch as a 'block' with the surrounding soil.
(iii) Bulging:
- Characterised by low punching and high compression ratios.
- Prevalent in long columns at high area ratios.
- This mode of deformation is highly dependent upon the lateral support provided by the
surrounding soil and as such, bulging generally occurs near the ground surface where
overburden stresses are lowest.
- High shear strains develop within columns near the top of the lower Carse clay layer,
i.e. the weakest part of the soil profile.
In addition to the three modes of deformation, the development of shear planes which extend
from the corners of pad footings to the point of column bulging beneath the centre of footings
were also observed. It should be noted that the shear planes are a secondary mode of
deformation; however they may become more prominent at higher load levels, as observed by
Muir Wood et al. (2000).
The deformational behaviour of small groups of stone columns is governed by the area ratio
and column length. While the number of columns enhances the settlement performance of
Conclusions 219
stone columns for a given area ratio, it does not fundamentally change the mode of
deformation. This is an important findings which was used in the development of a simplified
design method.
Influence of column length

The three modes of deformation postulated from compression and punching ratios are a
simplified description of the general deformational behaviour. A more in-depth analysis of
the variation of punching and compression ratios with column length reveals that some
combination of punching and bulging occurs simultaneously, with one particular mode of
deformation more influential for a given area ratio and column length. For example, long
columns at high area ratios exhibit some degree of punching, even though bulging is
established as the general mode of deformation.
The increase in settlement improvement factors with column length is most pronounced for
columns at low area ratios, for which punching is the dominant mode of deformation. Shear
stress develops along the sides of columns and increasing the column length directly
improves the settlement performance of stone columns. A certain degree of punching is
evident for all the configurations of columns analysed, which is consistent with the finding
that settlement improvement factors increase with column length for all configurations of
columns. Therefore, the existence of a unique critical column length, proposed from previous
laboratory studies, is not supported by this numerical modelling.
Influence of stiff crust

The presence of a stiff crust, a real feature of the stratigraphy in soft soil sites that has not
been recreated in previous laboratory studies, has a significant influence on the deformational
behaviour of small groups of columns. The stiff crust tends to confine columns in the upper
layers and forces bulging to occur in deeper layers, which enhances the columns’ load-
carrying capacity.
The presence of the stiff crust also contributes towards a combination of punching and
bulging modes of deformation. Previous research from laboratory tests observed well-defined
crossovers between punching and bulging modes of deformation at column lengths in the
range L/d = 6–10. However, this has been shown to be in part due to the absence of the stiff
crust in the laboratory tests (i.e. homogeneous soil samples). Columns in a homogeneous soil
Conclusions 220
sample are more susceptible to bulging near the ground surface, which reduces the ability of
columns to transfer the applied load to the base of columns and hence reduces punching
potential.
Influence of key design parameters and practical considerations

- The position of column relative to the footing edge has a relatively minor influence upon
the deformational behaviour of small groups of columns.
- The influence of column stiffness and strength upon the settlement performance of small
groups of stone columns is linked to the mode of deformation. Increasing the column
stiffness is most significant at low area ratios, as columns are in more of an elastic state
and thus are more susceptible to changes in elastic stiffness moduli. Conversely,
increasing the column strength is most significant at high area ratios, as the extent of
plasticity within columns is most pronounced.
- Increasing K0 in the surrounding soil enhances the settlement performance of stone
columns. This is most noticeable at high area ratios as columns which exhibit a bulging
mode of deformation are highly dependent upon the lateral support provided by the
surrounding soil. While increasing K0 in the surrounding soil enhances the settlement
performance of stone columns, it does not change their deformational behaviour.
8.1.4 Stress concentration ratios

The ratio of the vertical effective stress in columns to that in the soil is referred to as the
stress concentration ratio. This parameter is generally measured at the ground surface and is
quantified using two methods in this thesis:
(i) the first method compares the average vertical effective stress in columns (σ'col) to the
average vertical stress in the entire soil area beneath the footing (σ'soil, average).
(ii) the second method compares the average vertical effective stress in columns (σ'col) to the
column (σ'soil); the width of the zone of influence is equal to the column spacing.
Stress concentration ratios at the ground surface

Stress concentration ratios do not vary with the mode of deformation for an infinite grid of
columns. This is due to the symmetry which ensures that all columns deform identically.
Conclusions 221
Stress concentration ratios from PLAXIS 3D Foundation are in the typical range (2.5–5.0)
proposed by Barksdale & Bachus (1983).
In contrast to an infinite grid of columns, stress concentration ratios for small groups of stone
columns are dependent on the mode of deformation. An increase in stress concentration ratios
is observed with increasing column length for columns which are punching. However, no
increase in stress concentration ratios is observed with increasing column length for bulging.
Influence of key design parameters and practical considerations:

- Stress concentration ratios at the surface vary with the position of columns relative to the
footing edge. Stress concentration ratios measured using method (i) are highest in corner
columns, followed by edge and centre columns. This is attributed to the elevated stress
levels which develop beneath the edge of rigid footings. However, stress concentration
ratios measured using method (ii) are highest in centre columns, followed by edge and
corner columns. This highlights the positive confining effects provided to centre columns.
- Increasing the column stiffness and strength leads to higher stress concentration ratios.
This is most significant for columns exhibiting a bulging mode of deformation, as columns
which are punching transfer the applied load along the column-soil interface.
- Increasing K0 in the surrounding soil has a minor influence on stress concentration ratios.
This is consistent with the previous findings that K0 does not affect the deformational
behaviour of columns.
Stress concentration ratios with depth

It is interesting to note that the position of columns relative to edge of the footing has a
significant influence on stress concentration ratios, but only a minor influence on the
settlement performance of stone columns. The variation of stress concentration ratios with
depth was also examined and it was found that stress concentration ratios are dependent upon
the deformational behaviour of columns. Stress concentration ratios are relatively constant
with depth in yielded sections of columns and decrease with depth in concert with modular
ratios in elastic sections of columns.
Conclusions 222
8.1.5 Simplified design method

The behaviour of small groups of stone columns is quite complex as columns are subject to a
loss of lateral confinement along the outer edge and also a reduction in vertical stress with
depth beneath small loaded areas. The influence of the number of columns upon the
settlement performance of small groups of stone columns was quantified using a settlement
ratio (s/suc), which is defined as the ratio of the settlement of a small group to that of an
infinite grid of columns.
Settlement ratios for end-bearing columns increase with area ratio and were found to agree
favourably with Priebe’s (1995) s/suc values at low area ratios. Previous research is only
focussed on end-bearing columns at low area ratios. However, this thesis examines the
influence of both area ratio and floating columns (i.e. partial-depth treatment) upon s/suc for
the first time.
It was observed that area ratio only influences s/suc values for end-bearing stone columns.
Moreover, for a given normalised column length (L/H), a near-linear relationship exists
between s/suc and the normalised footing width (B/L). Therefore, it appears that the footing
width, and hence the stress distribution in the underlying soil, governs s/suc values. A
simplified design method was developed which allows the settlement of a small groups of
columns to be quickly estimated from B/L, L/H and suc. It is recommended that Pulko &
Majes (2005) be used as an estimate of suc.
8.2 Recommendations for future research

This research has focussed on the influence of key design parameters upon the settlement
performance and deformational behaviour of small groups of stone columns in soft soils. A
series of three-dimensional FEA were undertaken in conjunction with the advanced elasto-
plastic Hardening Soil model. A set of material parameters were developed for the well
characterised Bothkennar test site, which consists of a soft uniform clay overlain by a stiff
crust. The findings from the deformational behaviour of small groups of stone columns
allowed a simplified method to be developed which relates the settlement of small groups and
infinite grids of columns. However, this simplified method needs to be validated against field
load tests on various configurations of columns. This would require measurements of
Conclusions 223
settlement at various depths, using extensometers, for unreinforced and reinforced footings.
There is also further potential to extend the design method to other soil profiles:
- It was shown that the modular ratio and stress state in the surrounding soil both play an
important role in the settlement performance of stone columns. It would be valuable to
examine their influence upon settlement ratios for floating stone columns.
- Although the range of B/L and L/H investigated in this thesis covers a practicable range,
investigating the settlement performance of small groups of stone columns in a soil
profile of varying thickness would allow an even greater range of B/L and L/H to be
examined.
This thesis has focussed on the long term settlement performance of small groups of columns.
In addition to improving the settlement performance, stone columns also act as vertical drains
and enhance the drainage capacity of soils. It is well established that the soil carries a large
proportion of the vertical stress at the start of loading due to the high undrained stiffness.
However, vertical stress is transferred from the soil to the stone columns as consolidation
proceeds. The dissipation of excess pore pressure for small groups of stone columns has not
been investigated to date and it would also be useful to examine how the position of columns
relative to the footing edge affects the stress transfer during consolidation.
The installation effects associated with vibro stone columns have been touched on in this
thesis. Installing stone columns imposes a complex stress regime on the surrounding soil; the
displacing effects of the poker increase the soil stiffness, while the horizontal vibrations from
the poker can lead to remoulding and consequently a loss of soil strength. More field
measurements are essential to examine the influence of different soil types and column
configurations upon column installation effects.
224
References
Aboshi, H., Ichimoto, I., Enoki, M. and Harada, K. (1979). The compozer - a method to improve
characteristics of soft clays by inclusion of large diameter sand columns. Proc. Int. Conf. on Soil
Reinforcement: Reinforced Earth and Other Techniques, Paris, Vol. 1, 211–216
Allman, M. A. and Atkinson, J. H. (1992). Mechanical properties of reconstituted Bothkennar soil.
Géotechnique, Vol. 42, No. 2, 289–301
Ambily, A. P. and Gandhi, S. R. (2007). Behaviour of stone columns based on experimental and FEM
analysis. Journal of Geotechnical and Geoenvironmental Engineering, Vol. 133, No. 4, 405–415
Andreou, P. and Papadopoulos, V. (2006). Modelling stone columns in soft clay. Proc. 6th European
Conf. on Numerical Methods in Geotechnical Engineering, Graz, 777–780
Atkinson, J. H. and Sallfors, G. (1991). Experimental determination of soil properties. Proc. 10th
European Conf. on Soil Mechanics and Foundation Engineering, Florence, Vol. 3, 915–956
Barksdale, R. D. and Bachus, R. C. (1983) Design and construction of stone columns, Report no.
FHWA/RD-83/026, National Information Service, Springfield, Virginia, USA
Balaam, N. P. and Booker, J. R. (1981). Analysis of rigid rafts supported by granular piles. Int.
Journal for Numerical and Analytical Methods in Geomechanics, Vol. 5, 379–403
Balaam, N. P. and Booker, J. R. (1985). Effect of stone column yield on settlement of rigid
foundations in stabilised clay. Int. Journal for Numerical and Analytical Methods in
Geomechanics, Vol. 9, 331–351
Balaam, N. P., Poulos, H. G. and Brown, P. T. (1977). Settlement analysis of soft clays reinforced
with granular piles. Proc. 5th Southeast Asian Conf. on Soil Engineering, Bangkok, Vol. 1, 81–91
Baligh, M. M. (1985). Strain path method. Journal of the Geotechnical Engineering Division, ASCE,
Vol. 111, No. 9, 1108–1136
Biarez, J., Gambin, M., Gomes-Corriea, A., Falvigny, E. and Branque, D. (1998). Using
pressuremeter to obtain parameters of elastoplastic models for sands. Proc. 1st Int. Conf. on Site
Characterization, Atlanta, Georgia, USA, 747–752
Biot, M. A. (1941). General theory of three-dimensional consolidation. Journal of Applied Sciences,
Vol. 12, 155–164
Black, J. A. (2006). The settlement performance of a footing supported on soft clay reinforced with
vibrated stone columns. PhD thesis, Queen's University of Belfast
Bolton, M. D. (1986). The strength and dilatancy of sands. Géotechnique, Vol. 36, No. 1, 65–78.
Boussinesq, J. (1885). Applications des Potentiels à L’étude de L’équilibre et du Mouvement des
Solides Élastiques, Gauthier-Villars, Paris
References 225
Brauns, J. (1978). Die anfangstraglast von schottersäulen im bindigen untergrund. Die Bautechnik,
263–270
Brinkgreve, R. B. J and Broere, W. (2006). PLAXIS 3D Foundation Manual Version 2. PLAXIS BV
Castro, J. and Sagaseta, C. (2009). Consolidation around stone columns. Influence of column
deformation. Int. Journal for Numerical and Analytical Methods in Geomechanics, Vol. 33, Issue
7, 851–877
Castro, J. (2007). Pore pressures during stone column installation. Proc. 18th European Young
Geotechnical Engineers' Conference, Ancona
Charles, J. A. and Watts, K. S. (1983). Compressibility of soft clay reinforced with granular columns.
Proc. 8th European Conf. on Soil Mechanics and Foundation Engineering, Helsinki, Vol. 1,
347–352
Cunze, G. (1985). Ein beitrag zur abschätzung des porenwasserüberdruckes beim rammen von
verdrängungspfählen in bindigen böden. Dissertation University of Hannover
Debats, J. M., Guetif, Z. and Bouassida, M. (2003). Soft soil improvement due to vibro-compacted
columns installation. Proc. Int. Workshop on Geotechnics of Soft Soils - Theory and Practice,
Noordwijkerhout, Netherlands, 551–556
Domingues, T. S., Borges, J. L. and Cardoso, A. S. (2007a). Parametric study of stone columns in
embankments on soft soils by finite element method. Applications of Computational Mechanics
in Geotechnical Engineering V: Proc. 5th Int. Workshop, Guimaraes, 281–291
Domingues, T. S., Borges, J. L. and Cardoso, A. S. (2007b). Stone columns in embankments on soft
soils. Analysis of the effect of the gravel deformability. 14th European Conf. on Soil Mechanics
and Geotechnical Engineering, Madrid, Vol. 3, 1445–1450
Duncan, J. M. and Chang, C. Y. (1970). Nonlinear analysis of stress and strain in soil. ASCE Journal
of the Soil Mechanics and Foundation Division, Vol. 96, No. 5, 1629–1653
Egan, D., Scott, W. and McCabe, B. A. (2008). Installation effects of vibro replacement stone
columns in soft clay. Proc. 2nd Int. Workshop on the Geotechnics of Soft Soils, Glasgow, 23–30
Elshazly, H. A., Hafez, D. and Mosaad, M. (2006). Back calculating vibro-installation stresses in
stone columns reinforced ground. Proc. of the ICE - Ground Improvement, Vol. 26, Issue 2, 47–
53
Elshazly, H. A., Hafez, D. H. and Mossaad, M. E. (2008a). Reliability of conventional settlement
evaluation for circular foundations on stone columns. Journal of Geotechnical and Geological
Engineering, Vol. 26, No. 3, 323–334
Elshazly, H., Elkasabgy, M. and Elleboudy, A. (2008b). Effect of inter-column spacing on soil
stresses due to vibro-installed stone columns: Interesting findings. Journal of Geotechnical and
Geological Engineering, Vol. 26, No. 2, 225–236
Gäb, M., Schweiger, H. F., Thurner, R. and Adam, D. (2007). Field trial to investigate the
performance of floating stone columns. Proc. 14th European Conf. on Soil Mechanics and
Geotechnical Engineering, Madrid, 1311–1316
Gäb, M., Schweiger, H. F., Kamrat-Pietraszewska, D. and Karstunen, M. (2008). Numerical analysis
of a floating stone column foundation using different constitutive models. Proc. 2nd Int.
Workshop on Geotechnics of Soft Soils, Glasgow, 137–142
References 226
Institution of Civil Engineers. (1992). Bothkennar soft clay test site: characterization and lessons
learned. Géotechnique, Vol. 42, No. 2, 161–378.
Gibson, R. E. and Anderson, W. F. (1961). In-situ measurements of soil properties with the
pressuremeter. Civil Engineering and Public Works Review, Vol. 56, No. 658, 6l5–6l8
Gill, D.R. and Lehane, B.M. (2001). An optical technique for investigating soil displacement patterns.
Geotechnical Testing Journal, Vol. 24, 324–329
Goughnour, R. R. and Bayuk, A. A. (1979). Analysis of stone column-soil matrix interaction under
vertical load. Proc. Int. Conf. on Soil Mechanics Reinforcement, Paris, 271–277
Greenwood, D. A. (1970). Mechanical improvement of soils below ground surface. Geotechnical
Engineering, Proc. of Institution of Civil Engineers, London, 9–20
Greenwood, D. A. (1991). Load tests on stone columns. In Deep Foundation Improvements: Design,
Construction and Testing, Esrig M. I. and Bachus R. C. (eds.), American Society for Testing and
Materials, Philadelphia, ASTM STP 1089, 141–171
Guetif, Z., Bouassida, M. and Debats, J.M. (2007). Improved soft clay characteristics due to stone
column installation. Computers and Geotechnics, Vol. 34, Issue 2, 104–111
Han, J. and Ye, S. L. (2001). Simplified method for consolidation rate of stone column reinforced
foundations. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, Vol. 127, No.
7, 597–603
Hight, D. W., Bond, A. J., and Legge, J. D. (1992). Characterization of the Bothkennar clay: an
overview. Géotechnique, Vol. 42, No. 2, 303–347
Hu, W. (1995). Physical modelling of group behaviour of stone column foundations. PhD thesis,
University of Glasgow
Hughes, J. M. and Withers, N. J. (1974). Reinforcing of soft cohesive soils with stone columns.
Ground Engineering, Vol. 7, No. 3, 42–49
Jaky, J. (1944). The coefficient of earth pressure at rest. Journal of the Union of Hungarian Engineers
and Architects, 355–358 (in Hungarian).
Janbu, J. (1963). Soil compressibility as determined by oedometer and triaxial tests. Proc. European
Conf. on Soil Mechanics and Foundation Engineering, Wiesbaden, Vol. 1, 19–25
Jardine, R. J., Lehane, B. M., Smith, P. R. and Gildea, P. A. (1995). Vertical loading experiments on
rigid pad foundations at Bothkennar. Géotechnique, Vol. 45, No. 4, 573–597
Kirsch, F. and Sondermann, W. (2003) Field measurements and numerical analysis of the stress
distribution below stone column supported embankments and their stability. Int. Workshop on
Geotechnics of Soft Soils - Theory and Practice, Noordwijkerhout, Netherlands, 595–600
Kirsch, F. (2006). Vibro stone column installation and its effect on the ground improvement. Int.
Conf. on Numerical Simulation of Construction Processes in Geotechnical Engineering for
Urban Environment, Bochum, 115–124
Kirsch, F. (2008). Evaluation of ground improvement by groups of vibro stone columns using field
measurements and numerical analysis. Proc. of the 2nd Int. Workshop on the Geotechnics of Soft
Soils, Glasgow, 241–248
Lambe, T. W. and Whitman, R. V. (1979). Soil mechanics (SI Version). John Wiley and Sons, Inc.
References 227
Leroueil, S., Lerat, P., Hight, D. W. and Powell, J. J. M. (1992). Hydraulic conductivity of a recent
estuarine silty clay at Bothkennar. Géotechnique, Vol. 42, No. 2, 275–288
Mar, A. (2002). How to undertake finite element based geotechnical analysis. NAFEMS Publication
McCabe, B. A. (2002). Experimental investigations of driven pile group behaviour in Belfast Soft
Clay. PhD Thesis, Department of Civil Engineering, University of Dublin, Trinity College
McCabe, B. A., McNeill, J. A. and Black, J. A. (2007). Ground improvement using the vibro-stone
column technique. Transactions of the Institution of Engineers of Ireland, Galway
McCabe, B. A., Killeen, M. M. and Egan, D. (2008). Challenges faced in 3-d finite element modeling
of stone column construction. Proc. Joint Symposium on Concrete Research and Bridge and
Infrastructure Research in Ireland, NUI Galway, 393–400
McCabe, B. A., Nimmons, G. J. and Egan, D. (2009). A review of field performance of stone columns
in soft soils. Proc. Institution of Civil Engineers - Geotechnical Engineering, Vol. 162, No. 6,
323–334
McKelvey, D., V. Sivakumar, Bell, A.L. and Graham, J. (2004). Modelling vibrated stone columns in
soft clay. Proc. Institution of Civil Engineers - Geotechnical Engineering, Vol. 157, No. 3, 137–
149
Mitchell, J.K. and Huber, T. R. (1985). Performance of a stone column foundation. Journal
Geotechnical Engineering, ASCE, Vol. 11, No. 2, 205–223
Muir Wood, D., Hu, W. and Nash, D. F. T. (2000). Group effects in stone column foundations: model
tests. Geotechnique 50(6): 689–698.
Munfakh, G. A., Sarkar, S. K. and Castelli, R. J. (1984). Performance of a test embankment founded
on stone columns. Proc. Int. Conf. on Advances in Piling and Ground Testing, London, 259–265
Narasimha Rao, S., Madhiyan, M. and Prasad, Y. V. S. N. (1992). Influence of bearing area on the
behaviour of stone columns. Indian Geotechnical Conf., Calcutta, 235–237
Nash, D. F. T., Powell, J. J. M. and Lloyd, I. M. (1992a). Initial investigations of the soft clay test site
at Bothkennar. Géotechnique, Vol. 42, No. 2, 163–181
Nash, D. F. T., Sills, G. C. and Davison, L. R. (1992b). One-dimensional consolidation testing of the
soft clay from Bothkennar. Géotechnique, Vol. 42, No. 2, 241–256
Paul, M. A., Peacock, J. D. and Wood, B. F. (1992). The engineering geology of the Carse clay of the
national soft clay research site, Bothkennar. Géotechnique, Vol. 42, No. 2, 183–198
Potts, D.M. and Zdravkovic, L. (1999). Finite element analysis in geotechnical engineering: theory.
Thomas Telford
Potts, D.M. and Zdravkovic, L. (2001). Finite element analysis in geotechnical engineering:
application. Thomas Telford
Powrie, W. (2004). Soil mechanics: concepts and applications. Spon Press
Priebe, H.J. (1976). Evaluation of the settlement reduction of a foundation improved by vibro-
replacement. Bautechnik, No. 5, 160–162 (in German)
Priebe, H. J. (1995). The design of vibro replacement. Ground Engineering, Vol. 28, No. 10, 31–37
References 228
Pulko, B. and Majes, B. (2005). Simple and accurate prediction of settlements of stone column
reinforced soil. 16th International Conference on Soil Mechanics and Geotechnical Engineering,
Osaka, Vol. 3, 1401–1404
Randolph, M. F., Carter, J. P. and Wroth, C. P. (1979a). Driven piles in clay - the effects of
installation and subsequent consolidation. Géotechnique, Vol. 29, No. 4, 361–393
Randolph, M. F., Steenfelt, J. S. and Wroth C. P. (1979b). The effect of pile type on design
parameters for driven piles. Proc. 7th European Conf. on Soil Mechanics and Foundation
Engineering, London, 2, 107–114
Rowe, P. W. (1962). The stress-dilatancy relation for static equilibrium of an assembly of particles in
contact. Proc. Royal Society A, Vol. 269, No. 1339, 500–527
Schanz, T., (1998). Zur modellierung des mechanischen verhaltens von reibungsmaterialen.
Habilitation, Stuttgart Universität
Schanz, T., Vermeer, P.A., Bonnier, P.G., (1999). The hardening-soil model: Formulation and
verification. Beyond 2000 in Computational Geotechnics - 10 Years of PLAXIS International, R.
B. J. Brinkgreve (ed.), A. A. Balkema, Rotterdam, Netherlands, 281–290
Soderberg, L. O. (1962). Consolidation theory applied to foundation pile time effects. Géotechnique,
Vol. 12, No. 3, 217–225
Sondermann, W. and Wehr, W. (2004). Deep vibro techniques, Ground Improvement, 2nd Edition,
edited by M.P. Moseley and K. Kirsch, 57–92, Spon Press
Terzaghi, K. (1925). Erdbaumechanik auf bodenphysikalischer grundlage, F. Deuticke, Vienna
Thornburn, S. and MacVicar, R. S. L. (1968). Soil stabilization employing surface and depth
vibrators. The Structural Engineer, Vol. 46, No. 10, 309–316
Wehr, J. and Herle, I. (2006). Exercise on calculation of stone columns - Priebe method and FEM.
Proc. 6th European Conf. on Numerical Methods in Geotechnical Engineering, Graz, 773–776.
Wehr, W. (2004). Stone columns - single columns and group behavior. Proc. 5th Int. Conf. on Ground
Improvement Techniques, Malaysia, 329–340
Wehr, J. (2006a). The undrained cohesion of the soil as a criterion for the column installation with a
depth vibrator. TRANSVIB 2006, Gonin, Holeyman and Rocher-Lacoste (ed.), Editions du LCPC.
Paris
Wehr, W. C. S. (2006b). Stone columns - group behaviour and influence of footing flexibility. Proc.
6th European Conf. on Numerical Methods in Geotechnical Engineering, Graz, 767–772
Witt, K. J. (1978). Versagensmechanismus einzeln belasteter schottersäulen im bindigin untergrund
bei plötzlicher belastung. B. Sc. thesis, University of Karlsruhe
Woo-Seok Bae, Bang-Woong Shin, Byung-Chul An and Joo-Sub Kim (2002). Behaviours of
foundation system improved with stone columns. Proc. 12th Int. Offshore and Polar Engineering
Conf., Kitakyushu, Japan, 675–678
Yu, H. (2000). Cavity expansion methods in geomechanics. Kluwer Academic Publishers BV,
Dordrecht, NL

Title: Numerical Modelling of Small Groups of Stone Columns

Uploaded by

Copyright:

Available Formats

Title: Numerical Modelling of Small Groups of Stone Columns

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Title: Numerical Modelling of Small Groups of Stone Columns

Uploaded by

Copyright:

Available Formats

Provided by the author(s) and NUI Galway in accordance with publisher policies.

Please cite the published

Title Numerical modelling of small groups of stone columns

Author(s) Killeen, Micheál

Item record http://hdl.handle.net/10379/3464

A thesis submitted to the College of Engineering and Informatics,

Academic Supervisor: Dr. Bryan McCabe

This thesis may be lent or copied by the library.

Signed: ___________________________ Date: __________________

3 Background and preliminary checks for Finite Element Analysis

4 Development and validation of soft soil profile

5 Results of FEA: Settlement performance, deformational behaviour

6 Results of FEA: Distribution of total shear strains and

7 Development of simplified design method and comparison of

3 Background and preliminary checks for Finite Element Analysis

4 Development and validation of soft soil profile

5 Results of FEA: Settlement performance, deformational behaviour

6 Results of FEA: Distribution of total shear strains and

7 Development of simplified design method and comparison of

3 Background and preliminary checks for Finite Element Analysis

4 Development and validation of soft soil profile

5 Results of FEA: Settlement performance, deformational behaviour

cr, cv Coefficients of consolidation in radial and vertical direction, respectively

α Slope of s/suc versus B/L line

1.2 Aims and objectives

1.4 Outline of thesis

The accuracy of PLAXIS 3D Foundation to capture the load-deformational behaviour of

2.1.2 Construction processes

Dry top feed system

Wet top feed system

Dry bottom feed system

Figure 2.3 - Dry bottom feed system using ‘vibrocat’

2.1.3 Applications of vibro replacement

Figure 2.5 - Development of stress concentration ratios in stiffer stone columns

2.1.4 Limitations of vibro replacement

2.2 General deformational behaviour of stone columns

Narasimha Rao et al. (1992)

2.2.2 Behaviour of column groups

(a) (b) (c)

Woo-Seok Bae et al. (2002)

McKelvey et al. (2004)

Ambily & Gandhi (2007)

2.2.3 Summary of general deformational behaviour of stone columns

2.3 Settlement performance of stone columns

2.3.1 Case studies

2.3.2 Laboratory studies

Figure 2.14 - Illustration of block failure in group columns (Black, 2006)

2.3.3 Numerical studies

an extension of the Terzaghi theory of one-dimensional consolidation (Terzaghi, 1925). The

Domingues et al. (2007)

Andreou & Papadopoulos (2006)

Kirsch & Sondermann (2003)

Wehr & Herle (2006)

Elshazly et al. (2008a)

2.3.4 Summary of settlement performance of stone columns

2.4 Column installation effects

Figure 2.18 - Soil displacement due to pile installation (Yu, 2000)

2.4.1 Laboratory investigations

Signed: _________ Date:

Δ = Δ + Δ + Δ" + 2$Δ (2.3)

Δ = Δ + Δ + Δ" + 2$Δ (2.4)

Δ" = Δ + Δ + Δ" + 2$Δ" (2.5)

) + + +, = ) +, + )2 -.+ ⁄2, (2.6)

+ - 01 ⟹ -.+⁄2 ≈ *+ ⁄2

) + + +, = ) +, + )2 +⁄2, (2.7)

45.-.5 01 671.8-8-9 9. 5. ×

>-0-8 1 1.? → 0:

H ' '

L = M + 4 + 7 (2.19)

= /)1 + . − 11V , = X (2.23)

V = ./)1 + . − 11V , = XV (2.24)

] = X = 1/)1 + . − 11V , (2.27)

'c H 'c'Jd⁄J

H {|H zH }

6J = ) ' + 2$' 1 + + 2$ − 1 − 21 ' − y, (2.30)

6J = ) ' + 2$' ⁄ + + 2$ 1 − ⁄ − 2 ⁄ ' − y, (2.31)

= 6J /) ' + 2$' ⁄ + + 2$ 1 − ⁄ − 2 ⁄ ' − y, (2.32)

] = /6J 0Z (2.33)

R = RS QR (3.4)

81.R = RS 81.QR ≤ 81.QR (3.5)