Copyright Iste 2022 / File for personal use of Robert Charlier only

Deterministic Numerical Modeling

of Soil–Structure Interaction

Copyright Iste 2022 / File for personal use of Robert Charlier only
Copyright Iste 2022 / File for personal use of Robert Charlier only
 Series Editor
Marc Boulon

Deterministic
Numerical Modeling of
Soil–Structure Interaction

Edited by

Stéphane Grange
Diana Salciarini

Copyright Iste 2022 / File for personal use of Robert Charlier only
First published 2021 in Great Britain and the United States by ISTE Ltd and John Wiley & Sons, Inc.

Apart from any fair dealing for the purposes of research or private study, or criticism or review, as
permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced,
stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers,
or in the case of reprographic reproduction in accordance with the terms and licenses issued by the
CLA. Enquiries concerning reproduction outside these terms should be sent to the publishers at the
undermentioned address:

ISTE Ltd John Wiley & Sons, Inc.

27-37 St George’s Road 111 River Street
London SW19 4EU Hoboken, NJ 07030
UK USA

www.iste.co.uk www.wiley.com

© ISTE Ltd 2021

The rights of Stéphane Grange and Diana Salciarini to be identified as the authors of this work have been
asserted by them in accordance with the Copyright, Designs and Patents Act 1988.

Library of Congress Control Number: 2021947467

British Library Cataloguing-in-Publication Data

A CIP record for this book is available from the British Library
ISBN 978-1-78630-798-9

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Contents

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

Chapter 1. Hydro-mechanically Coupled Interface Finite

Element for the Modeling of Soil–Structure Interactions:
Application to Offshore Constructions . . . . . . . . . . . . . . . . . . . . . . . 1
Benjamin C ERFONTAINE and Robert C HARLIER
1.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1. The finite element method (FEM) . . . . . . . . . . . . . . . . . . . . . . 2
1.1.2. Review of existing contact formulations . . . . . . . . . . . . . . . . . . . 3
1.1.3. Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2. Governing equations of the interface problem . . . . . . . . . . . . . . . . . . . 7
1.2.1 .Mechanical problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.2. Flow problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.2.3. Couplings between mechanical and flow problems . . . . . . . . . . . . . 15
1.3. Numerical formulation of the element . . . . . . . . . . . . . . . . . . . . . . . 16
1.3.1. Space and fluid pressure discretization . . . . . . . . . . . . . . . . . . . . 16
1.3.2. Mechanical problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.3.3. Flow problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.3.4. Time discretization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.3.5. Stiffness matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.4. Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.4.1. Suction caissons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.4.2. Problem description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.4.3. Tension loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.4.4. Lateral loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.5. Conclusion and perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
1.5.1. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
1.5.2. Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
1.6. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

Copyright Iste 2022 / File for personal use of Robert Charlier only
vi Deterministic Numerical Modeling of Soil–Structure Interaction

Chapter 2. DEM Approach for the Modeling of Geotechnical Structures in

Interaction with Reinforcements. . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Pascal VILLARD
2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.2. Discrete modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.2.1. General concepts of the discrete modeling approach . . . . . . . . . . . . 56
2.2.2. Specific interaction between discrete particles and reinforcement
elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.2.3. Numerical strategy for geotechnical structure modeling
using DEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.3. Application of the DEM to geotechnical structures in interaction
with rigid piles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.3.1. Load transfer mechanisms within granular embankments
over a network of piles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.3.2. Load transfer mechanisms within granular embankments
over a network of piles under cyclic loadings . . . . . . . . . . . . . . . . . . . . 71
2.4. Application of the DEM to geotechnical structures in interaction
with flexible and deformable reinforcement – comparison with
experiment results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.4.1. Numerical and experimental behavior of geosynthetic tubes
filled with granular material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
2.4.2. Numerical and experimental behavior of granular embankments
reinforced with geosynthetic in areas prone to subsidence . . . . . . . . . . . . . 87
2.5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
2.6. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

Chapter 3. SSI Analysis in Geotechnical Engineering Problems

Using a Finite Difference Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
Daniel DIAS and Orianne JENCK
3.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
3.2. The finite difference method using an explicit scheme . . . . . . . . . . . . . . 102
3.3. Application of the finite difference method to soil–structure
interaction problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.3.1. Structural elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.3.2. Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
3.3.3. Constitutive models for soil . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.3.4. Dimension of the problem . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.3.5. Monotonic, quasi-static cyclic and dynamic loadings . . . . . . . . . . . . 112
3.4. Some application examples in the geotechnical engineering field . . . . . . . . 112
3.4.1. Reinforced retaining walls . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
3.4.2. Tunneling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Contents vii

3.4.3. Soft soil improvement using vertical rigid piles . . . . . . . . . . . . . . . 121

3.5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
3.6. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

Chapter 4. Macroelements for Soil–Structure Interaction . . . . . . . . . . . 143

Diana S ALCIARINI, Stéphane G RANGE, Claudio TAMAGNINI and
Panagiotis KOTRONIS

4.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

4.2. The concept of generalized forces: Eurocode 8 recommendations. . . . . . . . 145
4.3. Macroelements for shallow foundations . . . . . . . . . . . . . . . . . . . . . . 148
4.3.1. Generalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
4.3.2. Macroelements formulated in the framework of hardening
elastoplasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
4.3.3. Macroelements formulated in the framework of hypoplasticity . . . . . . . 151
4.4. The considered macroelements . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
4.4.1. The elastoplastic macroelement . . . . . . . . . . . . . . . . . . . . . . . . 154
4.4.2. The hypoplastic macroelement . . . . . . . . . . . . . . . . . . . . . . . . . 156
4.5. Case study: seismic response of a reinforced concrete viaduct . . . . . . . . . . 160
4.5.1. Features of the viaduct . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
4.5.2. The finite element model of the viaduct and its foundations . . . . . . . . . 161
4.5.3. Seismic input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
4.6. Calibration of the macroelements . . . . . . . . . . . . . . . . . . . . . . . . . 164
4.7. Results of the numerical simulations . . . . . . . . . . . . . . . . . . . . . . . . 167
4.7.1. Forces and displacements in the structural elements . . . . . . . . . . . . . 167
4.7.2. Displacements of the abutment and the foundations . . . . . . . . . . . . . 170
4.8. Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
4.9. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175

Chapter 5. Urban Seismology: Experimental Approach to Soil–Structure

Interaction Towards the Concept of Meta-city . . . . . . . . . . . . . . . . . . . 181
Philippe G UÉGUEN, Philippe ROUX and Andrea C OLOMBI
5.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
5.1.1.Observation of soil–structure interaction under weak
and strong seismic loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
5.1.2. Contamination of urban seismic motion by the vibration
of buildings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
5.1.3. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
5.2. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203

List of Authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

Copyright Iste 2022 / File for personal use of Robert Charlier only
Copyright Iste 2022 / File for personal use of Robert Charlier only
 Introduction

The deterministic numerical modeling of Soil–Structure Interaction (SSI) uses

the tools of Mechanics, and of their transposition, according to the methods of the
numerical discretization. An SSI analysis should take into account the differences
between the properties of the soil and the structure, possibly leading to local relative
displacements. In addition, the presence of water enforces to consider the interaction
on the hydro-mechanical framework, involving pore pressures and fluid flow.

Non-deterministic methods, making use of artificial intelligence and progressive

learning, are not considered here.

The geotechnical works are usually designed referring to the following three
situations:
1) The pseudo-static case: where the loadings are nearly time independent
(a small number of cycles are accepted).
2) The cyclic case: involving a large number of slow cycles (slow versus the
time scales of any physical phenomena involved).
3) The dynamic (mostly seismic) case: where inertia forces are taken into
account.

In the pseudo-static and cyclic cases, the SSI is defined from a local point of
view, whereas in dynamic conditions it is examined from a more global angle. The
soil–structure interfaces, transfer zones of the loads acting between soil and
structure, and often zones of large localized deformations have a significant
importance. Special strategies and experimental characterization tests have been
developed for describing them.

This volume gives an overview on the main modeling methods developed in

geotechnical engineering in order to describe the SSI in various situations:

Copyright Iste 2022 / File for personal use of Robert Charlier only
x Deterministic Numerical Modeling of Soil–Structure Interaction

– The classical and well-known finite element method (FEM) using interface or
contact elements available for coupled hydro-mechanical problems and in which the
local plastic energy dissipation contributes to the classical global damping.
– The distinct element method (DEM) in which the contact zones between soil
and structure are described by local adapted interaction laws.
– The finite difference method (FDM) using explicit algorithm in which interface
or contact elements are available and straightforward for coupled hydro-mechanical,
highly nonlinear with high deformation and dynamic problems.
– The more recent approaches based on the macro-element concept and
generalized variables in an incremental form available in static and dynamic
situations, where a large volume of soil underlying and surrounding the-structure is
represented by a small number of degrees of freedom. This method could be
considered as an improved extension of the stiffness coefficient methods known for
the piles as the t-z and p-y curves.
– The seismology approach with inertia and viscous equivalent damping forces,
in which the SSI is developed between several structures (like buildings in a town)
using the notion of meta-materials. This point of view is especially useful when
analyzing the movement of the buildings induced by a seismic event at the scale of a
town.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 1

Hydro-mechanically Coupled Interface

Finite Element for the Modeling of
Soil–Structure Interactions: Application to
Offshore Constructions

This chapter presents how the multiphysical soil–structure interactions can be

modeled using the finite element method (FEM). This method takes into account
different nonlinearities, due to geometry, material behavior, physical couplings and
interfaces. The FEM allows the volume to be discretized into a finite number of
elements and a finite number of degrees of freedom. The numerical techniques for
solving a nonlinear problem in the framework of the FEM have been extensively
studied and are now very robust.

The behavior of offshore foundations is highly dependent on interface properties.

Upon large overturning moment or lateral load, the soil–structure interaction results
in rotation, sliding or gap opening. Hydro-mechanical couplings have become
particularly important in many applications, as the loading rate and soil permeability
are such that the soil behavior is not completely drained or undrained.

A hydro-mechanically coupled interface element is proposed in this work to

simulate the behavior of suction caissons embedded in sand. This interface element
simulates the mechanical problem with a penalty method and reproduces the friction
at the interface as a simple Coulomb model. The interface element introduces hydro-
mechanical couplings based on the gap opening (storage, longitudinal permeability
evolution) and the definition of an effective contact stress.

Chapter written by Benjamin C ERFONTAINE and Robert C HARLIER.

Copyright Iste 2022 / File for personal use of Robert Charlier only
2 Deterministic Numerical Modeling of Soil–Structure Interaction

1.1. Introduction

The advent of numerical methods and, especially, finite element approaches has
provided engineers with formidable tools to predict the behavior of constructions.
Most of the time, the soil can be considered as a homogeneous medium behaving in
a purely drained or undrained manner. However, in some cases, depending on the soil
permeability and the loading rate, partially drained behaviors should be considered.
In this case, hydro-mechanical couplings must be included in the finite element
formulation for both volume and interface elements.

The development of renewable energy sources is the greatest challenge of the 21st
Century. A large number of offshore wind farms have been developed throughout the
last decade, especially in shallow waters in the North Sea where a sandy seabed is
mostly encountered. The offshore wind industry is expected to grow exponentially,
followed by the development of more recent wave and tidal energy devices. Whether
they are bottom-fixed or floating, their foundations or anchors will have to be designed.
The finite element simulation of anchors and foundations has become more popular
among offshore wind farm designers. However, the modeling of the interface between
the soil and the foundations is still very rudimentary.

From the mechanical point of view, shearing along the interfaces participates in the
strength of the foundation against applied loads in any direction (horizontal, vertical
or moment). When the maximum capacity is overcome, sliding occurs between the
soil and the foundation. Each movement of the foundation results in a fluid flow, since
it lies under the sea. This affects the interface behavior by modifying the maximum
shear stress available or by inducing a suction effect, because of a gap opening (e.g. for
suction caissons). The finite element modeling of these complex interactions at the
interface is the topic of this chapter.

1.1.1. The finite element method (FEM)

The finite element method (FEM) is one of the most popular numerical methods for
solving partial differential equations for boundary value problems [ZIE 00]. The main
concept of the method is to decompose the entire continuum domain to be modeled
into a collection of subdomains, called elements. The continuum field of any physical
variable of interest (e.g. displacement, pore water pressure) is approximated over each
element, based on a finite number of variables (physical unknowns) and pre-defined
interpolation functions. The partial differential equations that must be solved over the
entire continuum domain can then be approximated and discretized for each element.
The set of discrete equations associated with each element can be combined into a
single global system and solved algebraically. The solution of this set of equations
is the one that minimizes the error of the approximation with respect to the actual

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 3

solution. The FEM has many advantages, such as the ability to simulate complex
geometries, incorporate different constitutive laws, involve multiphysical couplings
and analyze stress and internal variables locally.

1.1.2. Review of existing contact formulations

Numerically solving the mechanical contact problem is not a recent topic, and
several books have been dedicated to this issue [JOH 92, WRI 06]. Within the
framework of finite element methods, two general approaches exist in order to manage
contact between two solid bodies, namely the thin layer and the zero-thickness
approach, as shown in Figure 1.1.

The first approach consists of explicitly modeling the contact zone with special
finite elements, designed to encounter large shear or compression deformation
[DES 84, SHA 93, WRI 13, WEI 15]. The second approach, adopted in the following,
involves special boundary elements which have no thickness, namely zero-thickness
elements [GOO 68, CHA 88, DAY 94, HAB 98, WRI 06]. They discretize a potential
zone of contact. A gap between each side of the interface and a probable other solid is
computed at each time step to detect contact. Three main ingredients are necessary to
formulate this approach:
– a criterion to rule the contact detection/loss and contact pressure evolution;
– a constitutive law to describe the shear/normal behavior(s);
– a technique to discretize the contact area between solids and to compute the gap
function gN , namely the distance between two solids.

No contact Contact
Medium 1
Thin layer
elements Medium 3
Medium 2
Thin layer

Medium 1
Boundary
elements
Medium 2
Zero-thickness

Figure 1.1. Comparison between the thin layer and zero-thickness

approaches in the case of Hertzian contact (source: [CER 15])

Copyright Iste 2022 / File for personal use of Robert Charlier only
4 Deterministic Numerical Modeling of Soil–Structure Interaction

The normal contact constraint ensures that two perfectly smooth solids in contact
do not overlap each other. This mathematical criterion relates the gap function gN and
the contact pressure pN . It states that when the gap is null, the solids are in contact
and a contact pressure prevents their overlapping. When the pressure is equal to zero
and the gap is greater than or equal to zero, there is no contact.

The relationship between gap and pressure evolution can be termed as high
precision or low precision. In the first case, this relationship is physically based and
the pressure increases with a gap that can be negative, leading to interpenetration
(gN < 0). This is especially true in rock mechanics, where surfaces in contact are
not perfectly smooth and the normal pressure increases with deformation of asperities
[GEN 90].

A low-precision contact is ensured on a purely geometrical basis. Two surfaces

of bodies in contact are considered to be perfectly smooth and interpenetration is
theoretically not allowed (gN ≤ 0) [WRI 04]. There are only two states: in contact
or not in contact. The Lagrange multiplier method exactly ensures this condition
[BEL 91, WRI 06]. Lagrange multipliers are introduced to enforce the constraint, and
correspond to the contact pressures. On the other hand, the penalty method [HAB 98]
regularizes the constraint by authorizing a limited interpenetration of the solids in
contact. The contact pressure is linearly proportional to this interpenetration by a
coefficient called the penalty coefficient. Both methods are compared in Figure 1.2.

Lagrange multiplier method

Pressure
distribution

Penalty method No penetration

Zoom
Penetration

Figure 1.2. Comparison between Lagrange multiplier

and penalty methods on deformation and distribution of
contact pressures (source: [CER 15])

The penalty method is adopted in the following for its simplicity, since it does
not require us to introduce a variable number of unknowns (Lagrange multipliers) at
each step. Both solutions are theoretically identical if the penalty coefficient tends

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 5

to infinity. However, increasing this coefficient too much creates a badly conditioned
stiffness matrix, which becomes difficult to solve.

The mobilization of shear stress along the interface is very important in many
applications. The maximum shear stress that could be mobilized in a tangential plane
is strongly dependent on the normal pressure. The most basic relationship between
them is the classical Coulomb criterion. However, similar to the normal behavior,
more complex constitutive laws may be defined: for example, to describe rock joints
[ALO 13, ZAN 13] or soil–structure interfaces [LIU 06, STU 16], including critical
state, we use dilative normal behavior and even cyclic degradation [LIU 08].

The normal contact constraint is defined as a continuous condition over the

boundaries of the solids in contact. It must be discretized in the finite element
framework. Computation of the gap (gN ) and enforcement of the contact constraint
are different depending on the method, as shown in Figure 1.3.
Node to node Node to segment

Gap Gap

Segment to segment Contact domain

Penetration
Gap interpolation

Gap

Figure 1.3. Comparison between the discretization methods

of the contact area (source: [CER 15])

The simplest method, namely node-to-node discretization, imposes the constraint

on a pair of nodes [KLA 88], limiting the problems to small relative tangential
displacement. The node-to-segment discretization overcomes this drawback
[HAL 85]. The gap function is computed by the projection of a node of one side,
termed the slave surface, on a segment of the other side, termed the master surface.
This method is sensitive to sudden changes in the projection direction, but smoothing
techniques strongly reduce this issue [WRI 01]. The contact domain discretization
does not involve any projection method [OLI 09, HAR 09]. The gap between the solids
potentially in contact is discretized by a fictitious mesh. Therefore, the gap function
is continuous between them and avoids many discrepancies and loss of unicity due to
projection.

Copyright Iste 2022 / File for personal use of Robert Charlier only
6 Deterministic Numerical Modeling of Soil–Structure Interaction

The segment-to-segment discretization [HAB 98, PUS 04, FIS 06] is based on the
mortar method developed in [BEL 98]. In this case, the contact constraint is applied
over the element in a weak sense. The gap function is computed by the closest-point
projection of a point of the non-mortar surface onto the mortar surface, which is given
more importance. In the following, this method is adopted. The distance is computed at
each integration point of the element and extrapolated over it by means of interpolation
functions.

In many examples, interfaces may be considered as preferential paths for fluid

flows. Therefore, a description of a longitudinal fluid flow, namely parallel to the
interface, must be defined. Boussinesq [BOU 68] first provided a mathematical law
that characterized the laminar flow of a viscous incompressible fluid between two
smooth parallel plates. The total fluid flow was proved to be proportional to the cube
of the aperture between the plates, and this relationship was termed the cubic law. The
applicability of this law to rock mechanics was proven [WIT 80, TSA 81, ORO 98]
under some hypotheses [OLS 01].

The flow problem within interfaces may be treated similarly to the mechanical
contact problem by special volume elements [THI 14] or the zero-thickness method
[GUI 02, SEG 08a]. The second approach is further developed in the following to be
consistent with the mechanical formulation. The flow problem is intrinsically coupled
with the mechanical problem through the opening/closing of the gap. In addition to
the flow along the interface direction, transversal flow also occurs, causing a drop of
pressure between the two sides. Two basic ingredients are necessary to describe this
part:
– a law relating the flow to the gradient of pressure (transversal/longitudinal);
– a technique to discretize the flow within and through the interface.

The single-node discretization of flow is the simplest one, as shown in Figure 1.4.
It simply superposes a discontinuity for fluid flow into a continuous porous medium
[BER 88]. In this case, there is no hydro-mechanical coupling and the opening of the
discontinuity is constant and user-defined. It acts as a pipe.

The double-node discretization considers the longitudinal gradients of pressure on

each side of the interface. The longitudinal flow depends on the average gradient. The
transversal flow is a function of the transversal difference between the pressure and a
parameter called transversal transmissivity [NG 97, GUI 02, SEG 08a, SEG 08b]. In
this case, the mechanical and flow behaviors of the interface are able to coexist and
are coupled.

The last option dissociates the discretization of the longitudinal flow from both
sides of the interface by adding nodes inside the interface. The nodes have only a
single pressure degree of freedom. This method is termed triple-node discretization

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 7

[GUI 02, JHA 14, CER 15]. The underlying hypothesis is that the field of pressure is
homogeneous inside the interface. In this case, there are two drops of pressure between
each side of the interface and inside the interface.
Single node

gN
Finite element mesh

Porous medium
Double node

Discontinuity
gN

Triple node

gN

Figure 1.4. Comparison between the discretization methods of the flow

within and through the interface (source: [CER 15])

1.1.3. Objectives

The first objective of this chapter is to provide a basic formulation of hydro-

mechanically coupled finite elements of the interface. The developments are
sufficiently general to be applied to any finite element code. In addition, any
constitutive law, from the most simple to the most complicated, can be inserted
into this formulation. The governing equations and finite element formulations are
provided in sections 1.2 and 1.3.

The second objective is to highlight the importance of considering hydro-

mechanically coupled interfaces in the soil–structure interaction. An example of
suction caisson modeling is provided in section 1.4 to illustrate the inherently partially
drained (between drained and undrained) behavior of this foundation in the offshore
environment and how interface elements are used to simulate this behavior.

1.2. Governing equations of the interface problem

The interface finite element is modeled using the zero-thickness approach, which
does not require any remeshing technique. The contact constraint is enforced by
a penalty method that is easy to implement and does not require any additional
unknowns. The mechanical constitutive law ruling the shear behavior is the classical
Coulomb model. The fluid flow problem is solved using a three-node discretization.
The couplings arise from the variation of storage and permeability with gap opening,

Copyright Iste 2022 / File for personal use of Robert Charlier only
8 Deterministic Numerical Modeling of Soil–Structure Interaction

as well as the definition of an effective stress. The purpose of this section is to derive
the main equations of the problem and to formulate the continuum interface problem.

1.2.1. Mechanical problem

1.2.1.1. Definition of the mechanical problem and gap function

Let us consider two deformable porous media Ω1 and Ω2 in their current
configurations at time t. The global system of coordinates is termed (E1 , E2 , E3 ). A
2D cross-section of these bodies is illustrated in Figure 1.5. Their evolution is assumed
to be quasi-static. Their boundaries in current configurations are denoted by Γ1 and Γ2 .
Imposed displacement (Dirichlet) and traction (Neumann) boundaries are respectively
denoted by Γiū and Γit̄ .

Γu2 Γ2t Γc1

Ω2 e22 x2
x2
e21 t1 Γc2
t2 e11
Γ1t x1 e12
Ω 1

E2
Γu1 E1

Figure 1.5. Statement of the mechanical problem and cross-section of

the 3D problem in the (E1 , E2 ) plane (source: [CER 15])

Γ1c and Γ2c denote both parts of the boundary where contact is likely to occur. In
this area, a local system of coordinates (e11 , e12 , e13 ) is defined along the mortar side
Γ1c , as shown in Figure 1.5, where e11 denotes the normal to the surface. The closest-
point projection x̄1 of a point of x2 of the boundary Γ2c onto Γ1c is defined such that
[WRI 06]
 
gN = x2 − x̄1 · ē11 , [1.1]

where (ē11 , ē12 , ē13 ) denotes the local system of coordinates at point x̄1 . This function
gN is referred to as the gap function, where the subscript N stands for the normal
direction. If there is no contact between the solids, gN is positive. The contact is
termed ideal if there is no interpenetration of the solids. This can be enforced if
the Lagrange multiplier method is used. However, if the penalty method is used,
interpenetration (negative gap function) is necessary to generate the contact pressure.

The definition of a relative tangential displacement between two points in the plane
of contact has no meaning in the field of large displacement [HAB 98]. Rather, normal

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 9

(N) and tangential (T1 and T2) velocities are defined in the local system of coordinates.
They are gathered into the vector ġ such that

ġ = ġN e11 + ġT 1 e12 + ġT 2 e13 . [1.2]

1.2.1.2. Normal contact constraint

The contact between two solids gives rise to non-zero stress vectors t1 = −t2
along their common boundary, as shown in Figure 1.5. These vectors are described in
the corresponding local system of coordinates at each contact point such that

t1 = −pN e11 + τ1 e12 + τ2 e13 , [1.3]

where pN is the normal pressure, and τ1 and τ2 are the shear stresses in both directions
in the plane of the interface. The ideal contact constraint is summarized into the
Hertz–Signorini–Moreau condition [WRI 06]:

gN ≥ 0, pN ≥ 0 and pN gN = 0. [1.4]

If there is no contact, the gap function gN is positive and the contact pressure pN
is null. In case of contact, the gap function is null and the contact pressure is positive.

This condition is no longer verified if the penalty method is used. In case of contact,
the relation between the pressure and the gap function can be written as

ṗN = −KN ġN , [1.5]

where the minus sign ensures that the contact pressure is positive when inter-
penetration increases, i.e. gN < 0 and ġN < 0.

1.2.1.3. Tangential contact constraint

When solids are in contact, the ideal tangential behavior of the interface
distinguishes between the stick and slip states [WRI 06]. In the former state, two
points in contact cannot move tangentially, i.e. they remain stuck together during
the simulation, i.e. ġT 1 = 0 and ġT 2 = 0. The second condition involves a relative
tangential displacement in the plane of the interface. This is summarized in a condition
similar to equation [1.4]:

ġTsli ≥ 0, f (t, q) ≤ 0 and ġTsli f (t, q) = 0 i = 1, 2 [1.6]

where ġTsli is the variation of the non-recoverable displacement in each tangential

direction. It is related to the variation of tangential displacement

ġT = sign(τ̇1 ) ġTsl1 e12 + sign(τ̇2 ) ġTsl2 e13 . [1.7]

Copyright Iste 2022 / File for personal use of Robert Charlier only
10 Deterministic Numerical Modeling of Soil–Structure Interaction

The stick and slip states are distinguished by the criterion f (t, q). It depends on the
stress state t and a set of internal variables q. The evolution of the stress state within
the interface depends on the constitutive law described hereafter.

The ideal stick state, ġT = 0, is also regularized by the penalty method, i.e. a
relative displacement is allowed. Therefore, the relation between the shear stress and
the tangential variation of displacement is given by
τ˙i = KT ġT i i = 1, 2. [1.8]

1.2.1.4. Constitutive law

The constitutive law adopted is a simple Coulomb model despite the actual
behavior of the soil–structure interface, and rock joints are proved to be more complex
than this simple criterion [HU 04, DEJ 06, HO 11, ZAN 13]. However, the constitutive
law is kept as basic as possible, since the examples presented are complex enough.
On the other hand, any nonlinear relation between stress and strain increments can
be substituted into the equations presented below. The interested reader should refer
to [SHA 97, MOR 02, LIU 08, STU 16] for a deeper insight into more complex
constitutive laws.

The constitutive law depends only on the stress state t within the interface and a
single parameter, the friction coefficient μ. Mathematically, it can be written as

2 2
f (t, μ) = (τ1 ) + (τ2 ) −μ pN . [1.9]
  
τ 

where τ  is the norm of the tangential stresses. The criterion is presented in Figure
1.6. In the absence of contact, the stress state lies on the apex of the criterion. Both
normal pressure and tangential stresses are null, i.e. t = 0. If the combination of
tangential and normal stresses lies below the criterion (f < 0), the tangential state
is considered as stick. Otherwise, if the stress state lies on the criterion (f = 0), the
tangential state is considered as sliding.

The tangential behavior is regularized by two shear penalty coefficients. Therefore,

there is no ideal stick state but, rather, a slight tangential displacement occurs.
The criterion may be interpreted as an elastoplastic relation and the shear penalty
coefficients act as shear moduli, acquiring a physical meaning. Hence, the incremental
relation between variations of stresses ṫ and variations of the gap function ġ can be
written as
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
ṗN −KN 0 0 ġN
⎣ τ̇1 ⎦ = ⎣ 0 KT 0 ⎦ · ⎣ġT,1 ⎦ , [1.10]
τ̇2 0 0 KT ġT,2
  
De

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 11

where De is equivalent to the elastic compliance tensor. When the interface reaches
the slip state, an elastoplastic compliance tensor Dep is defined such that
⎡ ⎤
−KN  0  0
⎡ ⎤ ⎢ τ1
2
τ1 τ2 ⎥ ⎡ ⎤
ṗN ⎢ (τ1 ) ⎥ ġN
⎢−μ KN KT 1 − −K T ⎥
⎣ τ̇1 ⎦ = ⎢ τ  τ 2 τ 2 ⎥ ⎣ ⎦
⎢  ⎥ · ġT,1 .
τ̇2 ⎢ τ2 τ1 τ2
2 ⎥
(τ2 ) ⎦ ġT,2
⎣
−μ KN −KT 2
KT 1 −
τ  τ  τ 2
  
Dep
[1.11]
This tensor was introduced in [CHA 88], which is based on a non-associated flow
rule.
No contact Stick Slip
pN =0 >0 >0
τ  =0 ≥ 0 = μ · pN

Table 1.1. Stress state in the interface

||τ|| f>0
Slip state f=0
f<0
No contact
Stick state
μ
pN

Figure 1.6. Mohr–Coulomb criterion (source: [CER 15])

1.2.1.5. Continuum formulation

Each solid Ωi verifies the classic mechanical equilibrium equations in quasi-static
conditions [BOR 13]. Solving the mechanical contact problem consists of finding the
field of displacement u for all points x ∈ Ωi to verify these equations, subjected to
the contact constraint equations [1.4] and [1.6].

Considering a field of admissible virtual velocities δ ẋ on Ωi , the weak form of the

principle of virtual power can be written as
2    2
   

i
σ :  (δ ẋ) dΩ = f̄ : δ ẋ dΩ + t̄ : δ ẋ dΩ + T : δ ẋ dΓ ,
i=1 Ωi i=1 Ωi Γit̄ Γic
[1.12]

Copyright Iste 2022 / File for personal use of Robert Charlier only
12 Deterministic Numerical Modeling of Soil–Structure Interaction

where f̄ represents the body forces, ū represents the imposed displacements, t̄

represents the imposed tractions, n is the normal to Γit̄ and Ti is the projection of the
local stress tensor ti in global coordinates. The equality of equation [1.12] is enforced
when the contact area Γic is known.

1.2.2. Flow problem

1.2.2.1. Definition of the problem

Let us consider a discontinuity of very thin width embedded in a porous medium
in its current configuration, as shown in Figure 1.7. This could represent, for example,
an open fault within a rock mass or a gap between a foundation and the surrounding
soil. This discontinuity creates a preferential path for fluid flow. Moreover, there is a
transversal fluid flow between the rock mass and the discontinuity.
q
Porous medium Ω2
Γq2 Γq2
Discontinuity Ω3 ̃
Γq1
̃

Fluid flow
Fluid flow
Fluid flow
Fluid flow

Γ1pw E2
Porous medium Ω1 pw E1

Figure 1.7. Definition of the flow problem (cross-section

of the 3D case in the (E1 , E2 ) plane), porous medium, discontinuity
and boundaries (source: [CER 15])

There is a conceptual difference between the treatment of the mechanical and flow
contact problems. The mechanical contact constraint consists of a non-zero pressure
pN applied along the contact zone Γc between the two solids Ω1 and Ω2 .

On the contrary, the opening of the discontinuity creates a gap gN filled with a
fluid. This gap creates a new volume Ω3 in which fluid flow takes place, as shown
in Figure 1.8. It is bounded by the two porous media Ω1 and Ω2 . Their boundary is
termed Γ1q̃ and Γ2q̃ . Therefore, Γq̃ represents a boundary where the solids are close
enough, fluid interactions take place and mechanical contact is likely to occur. It
always includes the contact zone Γc .

Ω3 is modeled as an equivalent porous medium. The fluid flow within it is

described by the cubic law. Fluid flows exist between the inner volume Ω3 and both

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 13

adjacent porous media Ω1 and Ω2 . This flow is a function of the difference of pressure
between them. This is a non-classical boundary condition since it is not an imposed
flux or an imposed pressure.

Finally, imposed flux and pressure boundaries on Ω1 and Ω2 are respectively

denoted by Γq̄i and Γp̄i w .

3

q2 e11

e21 x1
e31

q1 gN

E3
E2
E1

Figure 1.8. Definition of the equivalent interior porous medium Ω3

bounded by Γ1q̃ and Γ2q̃ (source: [CER 15])

1.2.2.2. Fluid flow formulation

A three-node formulation is adopted to describe the fluid flow through and within
the interface, as described in Figure 1.9. Therefore, fluid pressures on each side of the
interface (pw1 and pw2 ) and the inner fluid pressure (pw3 ) are the fluid variables. At
each point within the interface, four fluxes are defined:
– two longitudinal fluxes (fwl1 and fwl2 ) in the local tangential directions (e12 , e13 )
in the plane of the interface;
– two transversal fluxes (fwt1 and fwt2 ) in the local normal direction (e11 ).

The generalized Darcy’s law is assumed to reproduce the local longitudinal fluid
flows fwl1 and fwl2 in the plane of the interface. It reads, in each local tangential
direction (e12 , e13 ):

kl  
fwl(i−1) = − ∇e1i pw3 + ρw g ∇e1i z ρw for i = 2, 3 [1.13]
μw

where ∇e1i is the gradient in the direction e1i , μw is the dynamic viscosity of the fluid,
g is the acceleration of gravity, ρw is the density of the fluid and kl is the permeability.
These equations make the assumption that the fluid pressure is constant over the gap
gN and the gap is fully saturated.

Copyright Iste 2022 / File for personal use of Robert Charlier only
14 Deterministic Numerical Modeling of Soil–Structure Interaction

pw2 on q2

fwl1 fwt2
fwl2 pw3 in
3
fwt1
pw1
on q1 gN

E3
E2
E1

Figure 1.9. Definition of longitudinal and

transversal flows (source: [CER 15])

Each transversal fluid flux is a function of a transversal conductivity Twi and the
drop of pressure across Γiq̃ . They can be written as

fwt1 = ρw Tw1 (pw1 − pw3 ) on Γ1q̃ , [1.14]

fwt2 = ρw Tw2 (pw3 − pw2 ) on Γ2q̃ . [1.15]

1.2.2.3. Continuum formulation

The two actual porous media (Ω1 and Ω2 ) and the equivalent porous medium
Ω3 verify the classic hydraulic equilibrium equations [LEW 98]. Solving the contact
problem consists of finding the pore water distribution on Ωi to verify the equilibrium
equations and satisfy the non-classical boundary conditions [1.14]–[1.15] over Γq̃i .
Considering a field of admissible virtual pore water pressures δpw on Ω, the weak
formulation of the virtual power principle can be written as
3    3
 

Ṡ δpw − fw · ∇ (δpw ) dΩ = Q̄ δpw dΩ + q̄ δpw dΓ
i=1 Ωi i=1 Ωi Γiq̄

 
+ q̃ δpw dΓ [1.16]
Γiq̃

where fw is the fluid flux at point x, Ṡ is the storage term, Q̄ is the imposed volume
source, p̄w is the imposed fluid pressure, i = 1, 2 corresponds to the two porous media
in contact and i = 3 to the volume of the interface. The fluid flow q̃ along the boundary
corresponds to the transversal fluid flows fwti defined in equations [1.14] and [1.15].
The source term Q̄ associated with Ω3 is null.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 15

The mechanical problem gives more importance to the mortar side Γ1c . Similarly,
the integral over Ω3 is transformed into a surface integral over Γq̃1 . This hypothesis is
valid because it is assumed that the inner pressure is constant over the aperture gN of
the interface. Therefore, equation [1.16] for i = 3 can be finally written as
  
Ṡ δpw − fwl1 ∇e12 (δpw ) − fwl2 ∇e13 (δpw ) gN dΓ =
1
Γq̃

ρw Tw1 (pw1 − pw3 ) δpw − ρw Tw2 (pw3 − pw2 ) δpw dΓ, [1.17]
1
Γq̃

where ∇e1i is the gradient in the e1i direction.

In the porous media Ω1 and Ω2 , the storage component Ṡ is coupled with the
deformation of the solid skeleton. The treatment of this component for Ω3 is different
and detailed hereafter.

1.2.3. Couplings between mechanical and flow problems

The couplings between mechanical and flow problems depend strongly on the
gap function gN . The first coupling arises from the definition of the cubic law with
respect to the gap opening (hydraulic and mechanical openings are treated as an
identical variable). Therefore, if contact exists, the gap is negative and permeability
is equal to zero. However, from a physical point of view, there could be a path for
fluid flow through asperities of a rough surface (residual opening). Moreover, from
the numerical point of view, a null permeability may lead to a badly conditioned
problem. Therefore, a very low residual opening D0 is added. Hence, the permeability
is computed according to [OLS 01, GUI 02]
⎧ 2
⎪
⎨ (D0 )
if gN ≤ 0
kl = 12 2 [1.18]
⎪
⎩ (D0 + gN ) otherwise.
12
It is updated during the simulation to take into account the possible gap aperture.

A second coupling is created by the storage component Ṡ. The variation of

the total mass of fluid Ṁf stored in Ω3 comes respectively from the variation of the
fluid density, the opening/closing of the gap and the variation of the surface of the
discontinuity, namely
 
Γ̇q̃
Ṁf = ρ̇w gN + ρw ġN + ρw gn Γq̃ . [1.19]
Γq̃
  
Ṡ

Copyright Iste 2022 / File for personal use of Robert Charlier only
16 Deterministic Numerical Modeling of Soil–Structure Interaction

where Ṡ is the storage term of equation [1.16]. In the following, the fluid is assumed
to be incompressible ρ̇w = 0, and only the geometrical storage is taken into account.
In many applications, the main component of the storage is due to the opening/closing
of the interface ġN .

The mechanical behavior of the interface also depends on the fluid flow within
it. Indeed, the total pressure pN acting on each side Γiq̃ of the interface is defined
according to the Terzaghi principle [TER 25]. It is decomposed into an effective
mechanical pressure pN and a fluid pressure equal to the inner pressure pw3 :

pN = pN + pw3 . [1.20]

In this case, all the developments applied to the mechanical contact constraint and
constitutive laws in sections 1.2.1.2 and 1.2.1.4 must be treated with reference to the
effective pressure pN rather than to the total pressure pN . For example, equation [1.5]
is recast into

ṗN = −KN ġN . [1.21]

1.3. Numerical formulation of the element

The discretization of the governing equations is based on a segment-to-segment

approach. It is suitable for large relative displacements. Fluid flows are discretized
according to the triple-node approach. This method allows the modeling of a drop
of pressure across the interface. This section briefly describes the formulation of the
element in order to implement it in a finite element code. The exact definition of the
force vectors and stiffness matrix components, as well as rotation matrix, can be found
in [CER 15]. The element was implemented in the finite element code LAGAMINE,
developed at the University of Liège.

1.3.1. Space and fluid pressure discretization

The proposed 3D coupled interface finite elements are isoparametric and

quadrangular [ZIE 00]. A complete representation of the interface requires 12 nodes,
as shown in Figure 1.10:
– nodes 1, 2, 3, 4: the first side of the interface Γ1q̃ , three mechanical (coordinates
x, y, z) and a pore water pressure (pw ) degree of freedom per node;
– nodes 5, 6, 7, 8: the inner nodes of the interface Ω3 and a pore water pressure
(pw ) degree of freedom per node;
– nodes 9, 10, 11, 12: the second side of the interface Γ2q̃ , three mechanical
(coordinates x, y, z) and a pore water pressure (pw ) degree of freedom per node.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 17

(x,y,z,pw)
Γq2 11
̃ 7 (pw)
3
ζ2 (x,y,z,pw)

12 ζ1
8 10
4 6
2
E3
E2 9 Actual
5
E1 Γq1 ̃ 1 Element

Parent
Element
(1,1)

η ξ (1,-1)
(-1,1) (0,0)

(-1,-1)

Figure 1.10. Discretization of the interface into isoparametric elements

from convective (ζ1 ζ2 ) to local coordinates (ξ, η). Transformation to the
parent element (source: [CER 15])

Mechanical and hydraulic degrees of freedom are gathered into the vector of
generalized coordinates at each node i such that
 T
ui = xi , ui , zi , piw i = 1, 2, 3, 4, 9, 10, 11, 12 [1.22]

ui = piw i = 5, 6, 7, 8. [1.23]
These coordinates are continuously interpolated over the element using typical
linear interpolation functions φi (ξ, η) related to each node i of the interpolated side.
Continuous generalized velocities u̇ are interpolated over the element accordingly
from nodal values u̇i .

1.3.2. Mechanical problem

1.3.2.1. Local system of coordinates and gap function

The first step of the mechanical formulation is to determine the local system
of coordinates. The rotation matrix R relates the global (E1 , E2 , E3 ) to the local

Copyright Iste 2022 / File for personal use of Robert Charlier only
18 Deterministic Numerical Modeling of Soil–Structure Interaction

(e11 , e12 , e13 ) system of coordinates. This rotation matrix is computed with respect to the
side Γ1q̃ , which is the mortar side. Let us first consider the components in global axes
of two unit non-orthogonal vectors respectively parallel to each edge of an element,
namely
 T
1 ∂x ∂y ∂z
eξ = !" #2 " #2 " #2 , , , [1.24]
∂x ∂y ∂z ∂ξ ∂ξ ∂ξ
+ +
∂ξ ∂ξ ∂ξ
 T
1 ∂x ∂y ∂z
eη = !" #2 " #2 " #2 , , . [1.25]
∂x ∂y ∂z ∂η ∂η ∂η
+ +
∂η ∂η ∂η

The normal to the element is given by the cross product:

e11 = eξ × eη . [1.26]

The first tangential direction e12 is identical to eξ and the second tangential
direction is their cross product:

e13 = e11 × e12 . [1.27]

Therefore, the rotation matrix is the assembling of these vectors:


R = e11 e12 e13 . [1.28]

According to the continuous equation [1.1], the gap function at each point of Γeq̃ is
computed according to
⎡ ⎤ ⎡ 2 ⎤
ġN ẋ − ẋ1
T T
ġ = ⎣ġT,1 ⎦ = [R] · ⎣ẏ2 − ẏ1 ⎦ = [R] · Δẋ. [1.29]
2 1
ġT,1 ż − ż

The norm of the Jacobian of the transformation of the element from the convective
system of coordinates (ζ1 , ζ2 ) to the isoparametric system (ξ, η) can be written as
!" # " #2 " #2 !" #2 " #2 " #2
2
∂x ∂y ∂z ∂x ∂y ∂z
J = + + + + .
∂ξ ∂ξ ∂ξ ∂η ∂η ∂η
[1.30]

1.3.2.2. External energetically equivalent nodal forces

The mechanical contribution of a single interface element to the external virtual
power expression is derived from the continuous equation [1.12]. The energetically
equivalent nodal forces associated with node i of the interface element are computed

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 19

numerically using a Gauss scheme. For example, the mechanical nodal forces acting
on the boundary of Ω1 are computed according to
nIP
 
FiE = R · t φi J W IP
, [1.31]
IP =1

where t is the force vector in local coordinates defined in equation [1.3], φi is the
interpolation function associated with node i and the expression between brackets is
evaluated in each of the nIP integration points, associated with the Gauss weight W.
The reaction forces acting on Ω2 are computed accordingly.

1.3.3. Flow problem

1.3.3.1. Internal energetically equivalent nodal forces

Fluid flow inside the equivalent porous medium Ω3 , along the interface, involves
energetically equivalent internal forces. This component is derived from equation
[1.17]. It is numerically computed according to
nIP 
     
FiI = Ṡ φi − fwl1 ∇e12 φi − fwl2 ∇e13 φi J gN W . [1.32]
IP
IP =1

1.3.3.2. External energetically equivalent nodal forces

Transversal fluid flows between Ω1 , Ω2 and Ω3 provide energetically equivalent
external nodal forces related to fluid degrees of freedom. The contribution to the
external virtual power corresponding to Ω3 is derived from equation [1.17]. For
example, it is numerically computed on the boundary of Ω1 according to
nIP
  
FiE = ρw Tw1 (pw1 − pw3 ) φi − ρw Tw2 (pw3 − pw2 ) φi J W IP
,
IP =1
[1.33]

where pw1 is the fluid pressure on side 1, pw2 on side 2 and pw3 inside. The reaction
forces acting on the boundary of Ω2 are computed similarly.

1.3.4. Time discretization

Internal FiI and external FiE nodal forces defined in equations [1.31], [1.32] and
[1.33] are gathered into the global vectors FI and FE . Therefore, the vector of
out-of-balance forces FOB is defined according to

FOB = FI − FE . [1.34]

Copyright Iste 2022 / File for personal use of Robert Charlier only
20 Deterministic Numerical Modeling of Soil–Structure Interaction

The fluid flow problem within a porous medium is inherently time dependent.
Therefore, modeling its evolution requires the discretization of time. It is assumed
that the media in contact are initially in equilibrium at a given time t, i.e. FOB = 0.
The equilibrium of the discretized system should be verified over a whole time step
Δt such that
 t+Δt
G(t) FOB dt = 0 [1.35]
t

where G(t) is a weighting function. In this work, the weighting function is reduced to
a collocation δ(θ), where δ is the Dirac function. It is proved that a choice of θ ≥ 0.5
leads to an unconditionally stable time scheme [ZIE 00]. In this work, the integration
scheme is implicit, i.e. θ = 1. The equilibrium is then written at the end of the time
step.

1.3.5. Stiffness matrix

The stiffness matrix K related to the interface element is computed analytically

by deriving the out-of-balance forces related to node i with respect to the generalized
degree of freedom j. The extended developments are provided in [CER 15].

1.4. Application

The soil–foundation interaction plays a major role in the resistance of different

types of offshore foundations (e.g. piles, anchors, suction caissons) to combined
horizontal, moment and tension loads. Suction caissons are an emerging technology
for the foundation of offshore wind turbines. Their transient behavior is inherently
linked to hydro-mechanically coupled phenomena. Therefore, they represent a suitable
example of interface element capabilities.

1.4.1. Suction caissons

Ambitious objectives of the European Union to increase the share of renewable

energy in the Union’s energy mix have led to the multiplication of offshore wind
energy projects. The design of a suitable foundation is challenging and may represent
one third of the total project’s cost [BYR 03, SEN 08]. While monopiles represent the
great majority of installed foundations in shallow waters, other types of foundations
such as tripod or jacket structures are used in deeper waters and can be coupled with
different types of foundations.

Among different types of foundations, suction caissons, also termed bucket

foundations or suction anchors, should be highlighted [TJE 90, ISK 02, HOU 05c,

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 21

LI 15]. They consist of a hollow cylinder that is open towards the bottom. Their top
(the lid) can be a stiffer plate or a dome [TRA 05]. They are mainly made of steel and
may have stiffeners added inside to reinforce the structure. They are used as temporary
anchors [SEN 82] or permanent foundations [TJE 90] in sandy [IBS 14] or clayey soils
[GOU 09].

Installation of suction caissons requires only light equipment [HOU 05a]. Initially,
the caisson penetrates the sea bed under its own weight. Afterwards, the water trapped
inside is pumped out through an opening. A differential of water pressure between the
inside and the outside of the caisson digs it into the soil, as shown in Figure 1.11. In
addition, the seepage flow created reduces penetration resistance at the tip and along
the inner skirt, facilitating the installation [AND 08, SEN 09, BIE 18b, RAG 19]. The
foundation may be reused by reversing the process.

Outer pressure

Pumping

Modified
friction Inner pressure
Soil heave

Seepage
flow
Soil

Figure 1.11. Sketch of the installation process

Offshore wind structures are very light but subject to a large overturning moment.
Therefore, attention is paid to the definition of failure criteria under combined
horizontal load and moment in different configurations [BYR 02, GOU 08, BRA 09,
IBS 14, LI 15]. When the superstructure is a multipod, the large moment is converted
into a push-pull vertical load acting on the caissons. In this case, the ability to resist
large tension load is fundamental [HOU 05b, HOU 06, GAO 13, BIE 18b] and the
suction effect may be transiently mobilized for this purpose.

The behavior of suction caissons results from the interaction between the caisson
and the seabed. The soil behavior is the first source of complexity and nonlinearity
[THI 14]. Moreover, the mechanical problem is coupled with a complicated fluid flow
around the caisson. Depending on the geometry, soil permeability and loading rate, the
behavior of the soil is never completely drained or undrained, but most of the time,
partially drained [THI 14, CER 16]. However, in all cases, the strength of the caisson

Copyright Iste 2022 / File for personal use of Robert Charlier only
22 Deterministic Numerical Modeling of Soil–Structure Interaction

depends strongly on the soil–caisson interface behaviour and especially its coefficient
of friction [ACH 13, KOU 14, CER 16].

Modeling of a suction caisson is, therefore, a perfect example to illustrate the

basic working of hydro-mechanically coupled interface elements. The following
simulations involve mechanical contact/loss of contact, friction mobilization, sliding,
transversal/longitudinal fluid flow and storage of water inside a gap. Since this chapter
focuses on the interface behavior, the loading conditions, constitutive law of the
soil (simple elastoplastic) and flow modeling (fully saturated hypothesis, constant
permeability) are kept as basic as possible. The behavior of the caisson is explained
by vertical and lateral imposed displacements at the top of the caisson. The role of
interface elements is highlighted in each simulation.

1.4.2. Problem description

A steel suction caisson is assumed to be embedded in a homogeneous sand

layer, as shown in Figure 1.12a. Vertical and lateral loadings are investigated in the
following using the same 3D mesh for consistency, although axisymmetric conditions
are sufficient to study the vertical loading. The soil is represented by half of a
cylindrical layer. Its radius (Rsoil) is equal to 24 m and its height (Hsoil) is equal to
12 m, as shown in Figure 1.12a. The finite element mesh is composed of 16,058 nodes
and 14,190 elements, including volume (eight nodes, linear interpolation functions),
interface (four nodes) and boundary elements. Geometrical parameters of the mesh
are given in Table 1.2.

The caisson is composed of a horizontal lid at the top and a vertical skirt, as shown
in Figure 1.12b. The ratio of the skirt thickness (tskirt ) to the diameter (2 Rout ) is
larger than the real one [BYR 02, KEL 06] to avoid numerical issues. Indeed, using
thin and elongated elements may lead to numerical instabilities, especially during the
iterative solving of the global matrix. The effect on the results is limited since the
bearing capacity of the tip of the caisson is not mobilized during tension simulations
and only partly mobilized during lateral loading.

The mudline delineates the solid and liquid phases at the bottom of the sea.
The reference level of the sea top is chosen as 10 m above the mudline. However,
this does not matter in this simulation. What matters is only the variations of
pore water pressures (PWP) to balance tension and lateral loading. Above all, no
cavitation is taken into account in the flow modeling. The lateral face (y-z plane)
is considered impervious because of the symmetry. The lower face (x-y plane) is also
deemed impervious and could represent the transition to another layer of much lower
permeability. Pore water pressures are fixed on the other faces and are equal to the
hydrostatic pressure.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 23

Fixed PWP

Hsoil Caisson
Im
Bo perv
un iou
da s
ry

x-fixed
x- y-fixed
2 Rsoil fix
ed Fixed
z PWP
y x
z-fixed
(a) Global sketch of the problem geometry
Inner Imposed displacements
interface (top) Lid Mudline Confinement
(uplift)
(lateral)
Lenght (L)

Inner Outer
interface interface
(skirt) (skirt) σ0 σ0
Initial stresses

Skirt
Radius (Rin)

Radius (Rout)

(b) Description of the caisson’s elements (c) Initial conditions and loading

Figure 1.12. Statement of the problem

Caisson Soil Mesh

Rint Rout L tskirt tlid Rsoil Hsoil Nnodes Nelems
3.8 m 3.9 m 4 m 0.1 m 0.4 m 24 m 12 m 16,058 14,190

Table 1.2. Geometrical parameters: Rint , inner radius; Rout , outer radius; L, length;
tskirt , thickness of the skirt; tlid , thickness of the lid; Rsoil , outer radius of the soil
domain; Hsoil , thickness of the soil layer; Nnodes , number of nodes; Nelems number of
elements

The installation of the caisson into the soil is not modeled. This could have large
implications on the behavior of the caisson, since the stresses are modified and the
soil is disturbed by the installation [RAG 19]. However, for the sake of simplicity,
initial effective stresses (in the soil and interface) corresponds to the weight of the
soil (coefficient of earth pressure at rest K0 is equal to 1). The hydrostatic pore water
pressures corresponding to the depth of water are initialized.

Copyright Iste 2022 / File for personal use of Robert Charlier only
24 Deterministic Numerical Modeling of Soil–Structure Interaction

The mechanical behavior of the caisson is always elastic, but two distinct
hypotheses are considered for the soil: elastic and elastoplastic. The elastic law
considers a relatively high Young modulus of 200 MPa. The elastoplastic law
is an internal friction model, PLASOL, implemented in the finite element code
LAGAMINE [BAR 98]. The Young modulus is equal to the elastic one and
characterizes the low deformation behavior. Its initial friction angle φi is equal to 5°
and may harden up to φf = 30°. The hardening rule is characterized by the parameter
Bφ , indicating that half of the hardening is reached for a deformation equal to Bφ . The
porosity n and the specific mass γs are identical for the soil and the steel in order to
ensure a problem initially in equilibrium. The permeability is equal to 1.E-4 m/s. All
material parameters are summarized in Table 1.3.

The transversal conductivity Tw that characterizes the interface is null between the
caisson and the interface (the caisson is impervious) but not null between the soil and
the interface. The residual hydraulic aperture D0 is equal to 5.E-5 m. Therefore, there
is always a longitudinal fluid flow event in case of contact. The soil caisson friction
coefficient is equal to 0.50. The penalty coefficient is fixed at 1.E10 N/m3 , and the
influence of this parameter is investigated in the following.

Soil
E [MPa] ν [-] n [-] k [m/s] γs [kg/m3 ] φi [°] φf [°] Bφ [-]
2E2 0.3 0.36 1.E-4 2,650 5 30 0.005
Caisson
E [MPa] ν [-] n [-] k [m/s] γs [kg/m3 ]
2E5 0.3 0.36 0 2,650
Interface
KN [N/m3 ] KT [N/m3 ] Tw [m.Pa−1 .s−1 ] D0 [m] μ [-]
1E10 1E10 1.E-8 1.E-5 0.50

Table 1.3. Material parameters: E, Young’s modulus; ν , Poisson’s ratio; n, porosity;

k, permeability; γs , density of solid grains; K0 , coefficient of earth pressure at
rest; KN , KT , penalty coefficients; μ, friction coefficient; Tw , transversal conductivity;
D0 , residual hydraulic aperture

First, the simulations investigate the uplifting behavior of the suction caisson.
A uniform vertical displacement is imposed over the whole upper surface of the
caisson. The tension load is balanced only by the shear along the shaft of the caisson
and suction effect. The total load ΔFtot applied1 to the caisson is the sum of all
the reactions computed at each imposed degree of freedom. ΔFin and ΔFout are
respectively the integral of shear stresses acting on the inner and outer walls of the

1 Results from the simulations are multiplied by 2 to correspond to an entire caisson.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 25

caisson, as shown in Figure 1.13. Finally, the variation of PWP under the top of the
caisson is integrated and summarized into ΔFpw .

ΔFtot ΔFtot

ΔFpw

ΔFin ΔFout

Figure 1.13. Components of reaction balancing a tension

load: shearing inside and outside the shaft (left) and
pore water pressure variation (right)

Second, a lateral displacement is imposed at the center of the lid of the caisson.
In this case, the movement of the caisson is not axisymmetric at all. The caisson
rotates, shear stresses act inside and outside the caisson and passive earth pressures are
mobilized in front of the caisson in the direction of loading, as shown in Figure 1.14.
In this case, ΔFtot is the lateral reaction at the center, but the other components are
kept identical.

The distribution of shear stresses, effective/pore water pressures and p-y curves
will be defined to better understand the behavior of the caisson. These p-y curves result
from the idealization of the laterally loaded pile–soil interaction problem as a beam
supported by nonlinear springs, representing the soil reaction [MCC 56, REE 10].
They relate the resultant total horizontal load per meter length of the pile, applied by
the soil on the caisson, as a function of its lateral displacement y.

ΔFtot

ΔFpw

ΔFin ΔFin

ΔFout

Figure 1.14. Components of reaction balancing a lateral load: shearing

inside and outside the shaft and pore water pressure variation

Copyright Iste 2022 / File for personal use of Robert Charlier only
26 Deterministic Numerical Modeling of Soil–Structure Interaction

1.4.3. Tension loading

1.4.3.1. Drained simulation (mechanical problem)

During a drained uplift simulation (elastic soil), the total tension load applied is
balanced only by shearing along the shaft and the weight of the caisson. Figure 1.15
presents the evolution of shear components with the applied vertical displacement.
Variations of inside and outside shearing are almost linear at the beginning, since
variations of shear stress are given by

τ̇ = KT ġT if τ < μ pN . [1.36]

At the very beginning, this criterion is not met anywhere and variation of shear stress
is almost identical all along the shaft. The interface is still in a stick state.
2 ∆Ftot

1.5

∆Fout
Force [MN]

0.5 ∆Fin

∆Fpw
0

−0.5
0 0.2 0.4 0.6 0.8 1
Displacement [mm]

Figure 1.15. Drained uplift simulation of the caisson: components

of reaction for the entire caisson

The force–displacement result is stiffer for the outside shear component ΔFout
than the inside shear component ΔFin . Subsequently, the maximum shear stress is
mobilized first all along the outside wall of the shaft after an imposed displacement
of 0.4 mm. Any increment of the tension load applied is balanced by shearing along
the inside wall, inducing a slight slope breakage in the ΔFtot and ΔFin results. The
maximum total tension is reached after 0.6 mm, when the maximum shear stress is
reached all along the inside wall. There are no variations of pore water pressures since
the simulation is drained (ΔFpw = 0).

The stiffness of shear mobilization is different inside and outside the caisson.
The soil inside the caisson is more confined than that outside, and tends to move
along vertically with the caisson, as shown in Figure 1.16. Subsequently, the relative

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 27

tangential displacement rate ġt along the inner interface is slower than the outside,
which affects the shear rate τ̇ . Finally, the initial slope of ΔFin is lower than ΔFout .

Δz [mm]
9.42
8.64
7.85
7.07
6.28
5.49
4.71
3.92
3.14
2.35
z 1.57
0.78
0.00

Figure 1.16. Drained uplift simulation of the caisson: cross-section of

the vertical displacement, imposed vertical displacement at the top of
the caisson = 1 mm, drawing in the initial configuration

The global behavior of the caisson is closely linked to the shear stress distribution
along its shaft. A vertical cross-section along the 4 m length shaft inside the caisson is
shown in Figure 1.17 for different imposed vertical displacements. The vertical shear
stress (τz ) is shown on the left and the reduced shear stress (τz /pN ≤ μ) on the
right. After the full mobilization of friction (τz /pN = μ = 0.5), the distribution of
shear stress increases with depth, due to an increasing normal pressure pN . Hence,
the maximum shear stress is first mobilized at the top of the caisson (low maximum
stress) and, finally, at its bottom.

In this figure, τz is equal to zero at the top of the caisson after a sufficient
displacement. This indicates the creation of a gap between the caisson and the soil.
In this case, pN is equal to zero and shearing is not possible. This gap is due to the
stress distribution within the soil, resulting from shear stress mobilization along the
interface, which tends to separate the soil from the caisson. The length of the gap
depends on physical parameters (relative stiffness of the soil and caisson) and non-
physical parameters (penalty coefficient, which is detailed in the following). Outside
the caisson, the gap is longer (1.5 m) and occurs faster since the soil is not confined and
is free to move horizontally. The size of this gap is a function of the soil parameters,
as a softer soil would be more prone to collapse and avoid gaping.

The stiffness difference between inside and outside friction mobilization does not
affect the maximum load that can be sustained by the caisson. However, it affects the
small displacement behavior and then serviceability.

Copyright Iste 2022 / File for personal use of Robert Charlier only
28 Deterministic Numerical Modeling of Soil–Structure Interaction

0
Displ [mm] 0
0.02
−0.5 0.425 −0.5
0.625
−1 −1
1
−1.5 −1.5
Z [m]

−2 −2

−2.5 −2.5

−3 −3

−3.5 −3.5

−4 −4
−20 0 20 40 −0.5 0 0.5 1
τ [kPa] τ /p’ [−]
z z n

Figure 1.17. Drained uplift simulation of the caisson: cross-section

along the shaft for different vertical displacements (inside)

As discussed previously, two penalty coefficients (KN and KT ) are introduced

in the formulation of the interface finite element. Therefore, it is important to assess
their influence on the results. Four orders of magnitude of normal penalty coefficients
are compared in Figure 1.18a. The results prove that the behavior is only slightly
affected by these variations. At low displacement, the higher the KN value, the higher
the stiffness. For coefficients lying between 1.E09 and 1.E11, the final tension loads
increase slightly. However, the lowest coefficient leads to a higher final value.

For a drained simulation, the penalty coefficient mainly affects the gap opening
behavior since the normal pressure rate is given by ṗN = KN ġN . A gap is detected
when this pressure is equal to zero. Therefore, reducing the penalty coefficient leads to
less extended gaps and larger zones for shear stress mobilization. However, the normal
pressure distribution is also modified along the shaft due to the global movement of
the soil. Therefore, the final results are almost not affected in this case since these
effects compensate for each other.

It must be noted that soil behavior may have a much greater impact on the results
than penalty coefficients. In purely drained conditions, the maximum tension load is
identical but stiffness is affected by varying the material parameters. In the reference
case, the maximum load is reached after 0.6 mm. It increases up to 6 mm if an
elastic soil (E =20 MPa) is considered instead. However, it is limited to 0.9 mm
if the elastoplastic constitutive law is used, suggesting that plastic zones are very
limited in this case. Therefore, the error introduced by the penalty method is lower
than approximations on the real soil behavior.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 29

1.5

K [N/m3]
∆Ftot [MN]

1 1.E08
1E.09
1.E10
1.E11
0.5

0
0 0.2 0.4 0.6 0.8 1
Vertical displacement [mm]
(a) Influence of the normal penalty coefficient
2

1.5

K [N/m3]
∆Ftot [MN]

1 1.E08
1E.09
1.E10
1.E11
0.5

0
0 0.2 0.4 0.6 0.8 1
Vertical displacement [mm]
(b) Influence of the tangential penalty coefficient

Figure 1.18. Drained uplift simulation of the caisson

The influence of the KT coefficient on the results is assessed in Figure 1.18b.

In this case, the final value is not altered but the stiffness is. Contrary to the KN
coefficient, which is completely non-physical and should be as high as possible, KT
may have a physical signification. In other words, as KT decreases, the displacement
that should be applied to reach the maximum shear stress increases. The actual
interface behavior is more complicated, as it is nonlinear, normal stress dependent
and includes dilatancy. Subsequently, the accurate simulation of this behavior would
require dedicated complex constitutive laws, such as [STU 16].

Copyright Iste 2022 / File for personal use of Robert Charlier only
30 Deterministic Numerical Modeling of Soil–Structure Interaction

3.5
∆F
tot
3
Partially Drained
2.5

Drained
Force [MN]

2
∆F
tot ∆F
1.5 pw

1 ∆Fout
∆F
0.5 in

−0.5
0 1 2 3 4 5
Vertical displacement [mm]
(a) Loading rate 0.05 mm/s, drained total force is added for comparison
7
∆Ftot
6

5
v = 0.5mm/s
∆F
Force [MN]

4 pw
v = 0.05mm/s ∆F
tot
3
∆F
2 out

0 ∆Fin

−1
0 2 4 6 8 10
Vertical displacement [mm]
(b) Loading rate 0.5 mm/s, total load for a rate of 0.05mm/s is added for comparison

Figure 1.19. Partially drained uplift simulation of the caisson:

components of reaction for the entire caisson

As a partial conclusion, it was shown that the drained tension behavior of

a caisson is strongly dependent on interface properties, much more than the
soil volume behavior. Friction mobilization is quite important and could only be
represented by adequate finite elements. The following section introduces hydro-
mechanical couplings and explains why the proposed coupled element is necessary.

1.4.3.2. Partially drained simulation (coupled problem)

The results of a partially drained simulation (elastic soil) are presented in
Figure 1.19a. For comparison, the drained and partially drained total loads ΔFtot

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 31

applied are shown together. The two partially drained shear components are presented
as well as the pore water pressure variation component ΔFpw . The total capacity
is almost doubled, ductility is increased and the response is stiffer than the drained
simulation. The capacity is reached after 0.6 mm in drained conditions and 3 mm in
partially drained conditions.

Detailed results at a faster loading rate (0.5 mm/s) are presented in Figure 1.19b
and compared to the total load evolution at a slower uplifting rate (0.05 mm/s).
Compared to the simulation presented in Figure 1.19a, the suction component ΔFpw
is larger as well as the displacement at the steady state, as PWP have less time to
dissipate. In addition, the inside shear component tends to zero at large displacement,
indicating that the soil is plugged inside the caisson (typical of undrained behavior).
There is then no relative displacement and no shearing. Finally, outside shearing is
increased due to the negative water pressure that tends to keep soil and caisson edges
close enough.

The caisson behavior changes from completely drained to undrained, as the

uplifting rate increases (or the permeability decreases). This rate effect is well known
and was modeled previously by Thieken et al. [THI 14], although the gap formation
is simulated using “water elements”. The more undrained the loading, the higher the
(transient) capacity and stiffness. However, PWP variations become very negative, and
it must be verified whether some hypotheses are still valid, such as the full saturation
of the gap and the absence of water cavitation.

Δpw [kPa]
0.0
-2.6
-5.2
-7.8
-10.4
-13.0
-15.7
-18.3
-20.9
z -23.5
-26.1
-28.8

Figure 1.20. Partially drained uplift (loading rate 0.05 mm/s) simulation of the
caisson: cross-section of pore water pressure variation, imposed vertical
displacement at the top of the caisson = 5 mm, drawing in the initial configuration

Variations of PWP within the soil are responsible for these different behaviors. The
uplifting of the caisson creates a gap under the lid. It is assumed that this gap is fully
saturated with water, inducing a fluid flow from the soil to the gap and resulting in

Copyright Iste 2022 / File for personal use of Robert Charlier only
32 Deterministic Numerical Modeling of Soil–Structure Interaction

the negative variations of PWP, as shown in Figure 1.20 (loading rate is equal to 0.05
mm/s). The most negative variation of pressure is just under the lid of the caisson.
Therefore, the total water pressure over the lid is greater than that under the lid. This
difference of pressure creates the suction effect, transiently balancing a part of the
tension load.

This suction component ΔFpw is the main contributor to the strength

improvement, as shown in Figure 1.19a. However, it is not mobilized simultaneously
to the friction components. At low displacement, shearing along the shaft is the larger
contributor to the tension strength. ΔFpw starts increasing significantly when friction
is mobilized inside and outside. It finally reaches a plateau. This evolution is correlated
with the gap opening under the caisson’s lid. The gap opening rate vg depends on

vg = Δ̇zc − Δ̇zs , [1.37]

where Δ̇zc and Δ̇zs are the velocity of material points of the caisson and the soil,
at the interface under the lid. The gap opening induces negative variations of PWP
and an inverse consolidation process. The soil is uplifting Δ̇zs > 0 but slower than
the caisson. The water flow, soil and caisson movements interact together, but finally
reach a steady state where Δ̇zs → 0. Finally, the gap opening and the caisson velocity
are equal. Therefore, the water flux towards the gap is computed (if the caisson is
deemed completely rigid) according to

fg = ρw vg π R2in . [1.38]

where Rin is the internal radius of the caisson. The evolution of the total flux fg
towards the gap is shown in Figure 1.21. Its evolution is homothetic to the evolution
of ΔFpw . The steady-state water flux computed according to equation [1.38] is
equal to 2.27 kg/s, which is close to the asymptotic value computed numerically in
Figure 1.21.

The increase in tension capacity results from a steady state of water flow (stabilized
variations of PWP). However, it is linked to a continuous gap opening and then
uplifting. From a geotechnical point of view, it can only be mobilized to balance the
transient load. Otherwise, the continuous gap opening would lead to the soil collapse
below the moving tip and to a continuous reduction of the drainage length, eventually
leading to the failure of the caisson. In addition, the longitudinal gap opening, if it is
too large, may modify the results by increasing the dissipation of underpressures. All
of these effects are dependent on the caisson uplifting rate and soil permeability.

The constitutive law of the soil does not affect the load–displacement relationship
in all cases, as shown in Figure 1.22. The initial stiffness is similar for elastic and
elastoplastic constitutive laws. However, the capacity diverges when the flow regimes
approach undrained conditions. Indeed, the drained and low uplift rate behaviors

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 33

are dominated by failure along the interface. On the contrary, undrained or almost
undrained conditions lead to soil plugging. Failure and plasticity occur within the soil.
Consequently, the constitutive law that describes the soil behavior has a great influence
on the results.
2.5
fg = 2.27kg/s

2
Water flux [kg/s]

1.5

0.5

0
0 1 2 3 4 5
Vertical displacement [mm]

Figure 1.21. Partially drained uplift simulation (0.05 mm/s)

of the caisson: total water flux from the soil to the gap under
the top of the caisson

6
vc = 0.5 mm/s

5
∆Ftot [MN]

4
vc = 0.05 mm/s
3

2 Drained

1 Elastic
Elasto−Plastic
0
0 2 4 6 8 10
Vertical displacement [mm]

Figure 1.22. Partially drained uplift simulation of the caisson:

comparison of elastic and elastoplastic constitutive
laws at different rates

Copyright Iste 2022 / File for personal use of Robert Charlier only
34 Deterministic Numerical Modeling of Soil–Structure Interaction

1.4.3.3. Comparison with other authors

In order to validate the interface element and the proposed methodology, the
results presented by Thieken et al. [THI 14] are reproduced in Figure 1.23. Geometry
(L =10 m, Rout = 5 m) and soil parameters (internal friction model) are different,
and the details can be found in [THI 14]. The results presented for the hydro-
mechanically coupled interface are very consistent with those obtained by the “water
element” method.
10 1mm/s
9 Thieken et al. (2014)
This study
8
0.25mm/s
7
ΔFtot [MN]

6
0.001mm/s
5

0
0 5 10 15 20
Vertical displacement [mm]

Figure 1.23. Partially drained simulations for different permeabilities:

comparison of this work and the published results from [THI 14]

1.4.3.4. Loading/unloading cycle

Typical loading applied to offshore foundation is neither monotonic nor
displacement controlled. It consists of a cyclic force that results from the wind and
wave acting on the wind turbine and applied onto the foundation [BIE 18a, BIE 18b].
Modeling the long-term cyclic behavior of the suction caisson is out of the scope
of this study and was partly addressed in [CER 16]. In the following, a single
cycle loading is applied to the caisson and the different components of reaction
are detailed to inform the global behavior of the caisson. The loading is inspired from
the centrifuge tests presented in [BIE 18a, BIE 18b] and a cyclic amplitude of 80 kPa
is applied on the caisson for a zero average load, over a period of 11 s representative
of the offshore environment.

For the reference case, the results are presented in Figure 1.24. The behavior of
the caisson is almost undrained, as most of the total load applied is balanced by the
variations of PWP, and the internal shear stress mobilized, ΔFin , is close to zero.
It is interesting to note that the load–displacement relationship is not completely

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 35

symmetrical. A net tensile displacement can be observed in Figure 1.24a after a single
cycle, the stiffness is not identical in compression or tension and the total shear force
along the interface does not come back to zero inside and outside the caisson after a
cycle. However, there is almost no accumulated pore water pressure. This is consistent
with observations reported in [BIE 18a, BIE 18b].
4
∆ Ftot
3 ∆ Fin
∆ Fout
2
∆ Flid
1 ∆ Fpw
∆ F [MN]

0
S
-1

-2

-3

-4
-3 -2 -1 0 1 2 3
∆ y [mm]
(a) Force versus vertical displacement, S = starting point
4
∆ Ftot
3
∆ Fin
2 ∆ Fout
∆ Flid
1
∆ F [MN]

∆ Fpw
0

-1

-2

-3

-4
0 2 4 6 8 10 12
Time [s]
(b) Force versus time

Figure 1.24. Simulation of a single

compression/tension cycle (k = 1.E-4 m/s)

A second simulation was run, considering a lower permeability (k = 1.E-3 m/s),

to highlight how the drainage regime can affect the overall behavior of the caisson

Copyright Iste 2022 / File for personal use of Robert Charlier only
36 Deterministic Numerical Modeling of Soil–Structure Interaction

(Figure 1.25). In this case, the PWP is still the main contribution to the caisson
strength, but there is a clear contribution of the shear inside the caisson, as well as
the lid (in compression). The lid contribution participates in the non-symmetry of the
load–displacement relationship. Surprisingly, the contribution of PWP at the end of
the loading is not null in this case and indicates a net increase.
3

0

M[N 4 ]

F
S
-0

-1
Mtot
Mlid

Min
Mpw
-2

Mout
-3
-3 -2 -1 -0 F 0 1 2 3

y [mm]
(a) Force versus vertical displacement, S = starting point
4

Ftot
3
Fin

Fout
2

Flid
1
Fpw

F [MN]

-1

-2

-3

-4
0 2 4 6 8 10 12
Time [s]
(b) Force versus time

Figure 1.25. Simulation of a single

compression/tension cycle (k = 1.E-3 m/s)

In both cases, it was shown that the cyclic loading applied is mainly balanced by
the variations of PWP, which are reversible. The effective stress variations within the

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 37

soil, which generate plastic deformation, are limited. The final displacement at the
end of the loading is respectively equal to 0.39 mm (k = 1.E-4 m/s) and 0.67 mm
(k = 1.E-3 m/s). Consequently, the hydro-mechanical couplings can actually decrease
the cyclic-induced settlement of the soil around the caisson, if the caisson is designed
for that purpose, as highlighted in [CER 16].

In both cases, the observed non-symmetrical behavior is mainly due to hydro-

mechanical couplings. However, the asymmetry of the results from [BIE 18a, BIE 18b]
is much stronger than the results presented here. It is expected that a more complex
constitutive law including Lode angle dependency would be more appropriate, as the
stiffness and strength of soils are different in compression and extension.

1.4.4. Lateral loading

In the following, a lateral displacement is imposed at the center of a caisson in the

y-direction while the displacement is free along the vertical direction. This loading
mode corresponds more to anchors than to foundations (usually higher L/D ratio).
However, an identical geometry is used, to ensure the consistency of this chapter. It is
also shown in the following that the nonlinear behavior of the caisson depends mostly
on the plasticity that develops within the soil. Therefore, the soil is only described by
an elastoplastic constitutive law.

1.4.4.1. Drained simulation (mechanical problem)

Upon an imposed lateral displacement, the drained behavior of the caisson
becomes non-symmetrical. Cross-sections of lateral and vertical soil displacements
after an imposed central displacement of 2.5 mm are presented in Figures 1.26 and
1.27 respectively. The displacement of the caisson is the combination of three rigid
body motions: a lateral displacement, a small vertical displacement and a rotation.

The soil in front of the caisson is pushed away along with the lateral displacement
of the caisson, creating a plastic wedge. This is suggested in Figure 1.26 by the
steep gradient of displacement, in front of the caisson (on the left). Plastic strains
are larger near the surface of the soil, where the average stress is low, which in turn
induce larger lateral displacements and eventually a rotational movement, as shown
in Figure 1.27. Consequently, the shear stress developed along the caisson’s shaft has
two components. The horizontal shear stress is due to the lateral relative displacement,
while the vertical component is due to the rotation of the caisson. The soil and caisson
displacements (both lateral and vertical) are not continuous between the inside and
the outside of the caisson. This suggests that a gap is created behind the caisson. No
shear resistance could then be mobilized along this zone. Inside the caisson, the soil
accompanies more or less the movement of the caisson, except along the top. The
difference in vertical displacement along the top surface indicates that a gap is also
created there.

Copyright Iste 2022 / File for personal use of Robert Charlier only
38 Deterministic Numerical Modeling of Soil–Structure Interaction

y Imposed displacement

Δy [mm]
2.23
2.03
1.83
1.62
1.42
1.22
1.01
0.81
0.61
0.40
0.20
0.00

Figure 1.26. Drained lateral displacement of the caisson after an

imposed lateral displacement of 2.5 mm: lateral displacement (Δy),
cross-section in the Y-Z plane, drawing in the initial configuration

y >0 Imposed displacement

Δz [mm]
2.04
1.79
1.53
1.28
>0 1.02
<0 >0 0.76
0.51
0.25
0.00
-0.25
-0.51
<0 -0.76
-1.02

Figure 1.27. Drained lateral displacement of the caisson after an

imposed lateral displacement of 2.5 mm: vertical displacement (Δz),
cross-section in the Y-Z plane, drawing in the initial configuration

The total horizontal load applied to the caisson (ΔFtot ) is balanced directly
by contact pressure and shear along the shaft of the caisson in the horizontal y-
direction (parallel to the lateral face of the mesh). However, due to the rotational
movement, non-negligible vertical shear stresses (inside ΔFin and outside ΔFout )
are also generated within the soil–caisson interface. They are represented similarly to
tension simulations in Figure 1.28. In addition, shear is mobilized at the base of the
caisson. It may lead to a global failure if the shear capacity is reached, which is not
observed here, as the caisson has a large diameter.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 39

The total load–displacement relationship presents a classical shape with a clear

change of slope after an imposed displacement of approximately 2 mm [ACH 10,
BIE 12]. This is mainly due to the development of plasticity within the soil. The
shear components have different magnitudes and signs. The direction of shear forces
is upwards at the front of the caisson and downwards behind it, due to the rotation.
However, the magnitude of shear stress is different in absolute value between the
inside and the outside of the caisson, due to the gap opening (outside at the back).
Maximum friction is reached in both cases since a plateau is reached.
3.5

3 ΔFtot

2.5
Force [MN]

1.5

1
ΔFin
0.5

0 ΔFpw
ΔFout
−0.5
0 5 10 15 20 25 30
Lateral displacement [mm]

Figure 1.28. Drained lateral simulation of the caisson:

components of reaction for the entire caisson

Cross-sections of shear and normal pressure distribution along the shaft of the
caisson are presented radially in Figure 1.29. As these values are vectors, they are
therefore characterized by:
– an application point: the intersection between the radial direction and the trace
of the caisson;
– a direction: always radial;
– a magnitude: the distance between the application point and the trace of the
results;
– a sense: positive (outside the trace of the caisson), negative (inside).

All the results are normalized with respect to the maximum value (Xmax , given in
the figure).

The distribution of normal pressures at different depths of the caisson along the
outside wall of the shaft, are presented in Figure 1.30a after an imposed displacement
of 0.5 mm. Normal pressures indeed increase with depth, and are higher in front of

Copyright Iste 2022 / File for personal use of Robert Charlier only
40 Deterministic Numerical Modeling of Soil–Structure Interaction

the caisson, because of the lateral loading. The lateral displacement also induces a gap
behind the caisson, denoted by the zero lateral pressures at depths of 0.25 and 1.25 m.

y Results

Application
Ma point
gni
tud
e Caisson
>0 <0

Figure 1.29. Normalized radial distribution of the results

The lateral movement induces horizontal shear stress along the outside wall of the
shaft, which is always oriented tangentially to the shaft. Projections of these shear
stresses in the y-direction are presented in Figure 1.31a. The maximum is obviously
located on the side of the caisson where the tangent to the caisson is in the y-direction.
This indicates that friction along the shaft also plays a role in balancing the applied
lateral load.

Vertical shear stresses along the outside wall of the shaft are presented in
Figure 1.32. The sign of these vertical shear stresses change between the front and
the back of the caisson due to the rotational movement. They are of higher magnitude
at the front since the increasing normal pressure allows a higher magnitude of shear
stress. The axis of rotation of the caisson is perpendicular to the plane of symmetry
and has its origin almost at the center of the caisson. It crosses the shaft of the caisson
where the sign of the shear stress changes.

Failure can be defined after an imposed lateral displacement of 2.5 mm, where a
slope breakage occurs in the load–displacement relationship. Normal pressures and
horizontal shear stresses are presented at this moment in Figures 1.30b and 1.31b
respectively.

The gap is completely open all along the back part of the shaft, and the pressure is
equal to zero along this part. However, they are increased by almost 50% at the front.
Once again, the maximum pressures increase with depth, due to the confinement.
Shear stress in the y-direction have also increased, but the maximum has moved to
the front of the caisson. This is due to the gap opening that has decreased the pressure
and then the maximum shear stress available. In addition, it must be kept in mind that
there is an interaction between the two shear stress components since
$
τmax = μ pN = (τ1 )2 + (τ2 )2 . [1.39]

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 41

Depth [m]
|pmax| = 41kPa 0.25
1.25
2.25
3.25

1 −1
0.75 −0.75
0.5 −0.5
0.25 −0.25

(a) Imposed lateral displacement equal to 0.5 mm

Depth [m]
|pmax| = 65.7kPa 0.25
1.25
2.25
3.25

1 −1
0.75 −0.75
0.5 −0.5
0.25 −0.25

(b) Imposed lateral displacement equal to 2.5 mm

Figure 1.30. Distribution of normal effective pressure along

the outer shaft: lateral drained simulation

Therefore, if rotation increases and mobilizes too much vertical shear stresses, then
the lateral shear stress will also be reduced.

P-y curves are a useful tool to understand the basic working of piles that are
laterally loaded and to simulate it as a 1D problem. They provide an insight into the
stiffness, the strength and the displacement history of the caisson at different depths.
Basically, the lateral loads acting on the caisson over a given range of depths are
summed up and expressed as a force per unit length of foundation p. They are shown
as a function of the lateral displacement of the neutral axis of the foundation at the
same depth y. The total load incorporates contact pressure, lateral shear stress and
pore water pressures. These curves are presented in Figure 1.33 at four depths.

Copyright Iste 2022 / File for personal use of Robert Charlier only
42 Deterministic Numerical Modeling of Soil–Structure Interaction

Depth [m]
|τmax| = 12.2kPa 0.25
1.25
2.25
3.25

1 −1
0.75 −0.75
0.5 −0.5
0.25 −0.25

(a) Imposed lateral displacement equal to 0.5 mm

Depth [m]
|τmax| = 23.2kPa 0.25
1.25
2.25
3.25

1 −1
0.75 −0.75
0.5 −0.5
0.25 −0.25

(b) Imposed lateral displacement equal to 2.5 mm

Figure 1.31. Distribution of horizontal shear stress along the outer shaft
(projection into the y-direction): lateral drained simulation

However, a caisson is far from a pile: it has a length-to-diameter ratio of 0.5. There
is then almost no flexion and the caisson cannot be modeled as a beam. Moreover,
loads acting on the shaft are both inside and outside the caisson. Therefore, the p-y
curves are the results of the difference between inside and outside reactions from the
soil onto the caisson, in the y-direction. The y displacement is the displacement of the
mean fiber of the caisson at each depth.

As expected, the lateral load increases while the depth decreases, which is at the
origin of the rotational movement of the caisson. The lowest lateral displacement
occurs at a higher depth. Each p-y curve also exhibits a sharp slope breakage already
observed in the global displacement load curve. However, the lower the depth, the
sooner the slope breakage occurs.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 43

Depth [m]
|τmax| = 14.8kPa 0.25
1.25
2.25
3.25

1 −1
0.75 −0.75
0.5 −0.5
0.25 −0.25

Figure 1.32. Distribution of vertical shear stress along the outer shaft: lateral
drained simulation, imposed lateral displacement equal to 0.5 mm

1.2

0.8
Force [MN/m]

0.6

Depth [m]
0.4
3.25
0.2 2.25
1.25
0 0.25

−0.2
0 5 10 15 20 25 30
Lateral displacement [mm]

Figure 1.33. P-y curves along the shaft of the caisson:

drained simulation

Similar to the tension simulation, the interface element is essential to reproduce

the movement of the caisson. Indeed, rotation and gap opening, resulting from the
lateral movement, can be only reproduced by interface elements. Additional features
induced by the hydro-mechanical couplings are detailed in the following section.

1.4.4.2. Partially drained simulation (coupled problem)

Similar to the tension loading simulation, the partially drained nature of the
caisson’s behavior increases its ductility and strength. The load–displacement
relationships are presented in Figure 1.34 in partially drained conditions. The total

Copyright Iste 2022 / File for personal use of Robert Charlier only
44 Deterministic Numerical Modeling of Soil–Structure Interaction

load drained curve is added for comparison. The difference in resistance after
an imposed lateral displacement of 20 mm is almost 1 MN. Both vertical shear
components ΔFin and ΔFout decrease with respect to the drained simulation,
indicating a modification of shear distribution along the shaft. Finally, the uplifting
movement of the caisson creates a gap under its top surface. Subsequently, a pore
pressure drop occurs and the ΔFpw is not null. This load opposes to the rotational
movement of the caisson.
4 ∆F
tot
3.5
v = 0.05 mm/s
3
∆F
tot
2.5
Force [MN]

Drained
2

1.5

1 ∆Fpw

0.5 ∆F
in

0
∆F
−0.5 out
0 5 10 15 20
Lateral displacement [mm]

Figure 1.34. Partially drained lateral simulation of the caisson

(v = 0.05 mm/s): components of reaction for the entire caisson

The distribution of pore water pressures is no longer symmetric, as shown in

Figure 1.35. Variations of pore water pressures within the caisson are negative due to
the slight uplifting movement. In this figure, they are more negative in the left corner
of the caisson because a gap is also opening vertically in this zone, due to the lateral
movement.

In front of the caisson, water overpressures should be generated since the soil is
under compression. However, these pressures are quickly dissipated, since the material
is very close to the mudline, where pressures are fixed. On the contrary, at the back,
very negative pressures are generated, close to the tip of the caisson. It is thought that
this is due to the longitudinal gap opening behind the caisson. On the one hand, the
gap opening induces a suction effect and a drop of pressure. On the other hand, this
gap increases the vertical fluid flow along the shaft and should decrease the suction
effect. In this case, the first effect seems to be of greater magnitude than the second
effect, but this depends on many variables such as soil permeability/stiffness and
loading rate.

The drained and partially drained p-y curves are presented in Figure 1.36 for
comparison. Each p-y curve includes different components acting on each side of

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 45

the caisson: normal effective pressure, horizontal shear stress and pore water pressure.
The curves close to the surface (depth 0.25 m) have more or less an identical shape.
All the other partially drained curves show an increase in force for an identical
displacement.
y Imposed displacement

Δpw [kPa]
0
-1.7
-3.4
-5.1
-6.8
-8.5
-10.2
-11.9
-13.7
-15.4
-17.1
-18.8

Figure 1.35. Partially drained lateral displacement of the caisson after an

imposed lateral displacement of 20 mm: variation of pore water pressure
(Δpw ), cross-section in the plane Y-Z, drawing in the initial configuration

1.2
2.25m 1.25m
3.25m
1

0.8
Force [MN/m]

0.6 0.25m

0.4

0.2

−0.2
0 5 10 15 20
Lateral displacement [mm]

Figure 1.36. P-y curves along the shaft of the caisson: partially drained
simulation (v=0.05 mm/s)

The distribution of effective stress around the shaft is not really modified, as shown
in Figure 1.37a. Only the gap is less open, since the normal pressure is not equal to

Copyright Iste 2022 / File for personal use of Robert Charlier only
46 Deterministic Numerical Modeling of Soil–Structure Interaction

zero behind the caisson (depth 3.25 m). Therefore, it could be reasonably stated that
the difference between p-y curves is due to the pore pressure distribution.

Depth [m]
|pmax| = 66.1kPa 0.25
1.25
2.25
3.25

1 −1
0.75 −0.75
0.5 −0.5
0.25 −0.25

(a) Distribution of effective normal pressure along the outer shaft

Depth [m]
|Δpwmax| = 10.8kPa 0.25
1.25
2.25
3.25

1 −1
0.75 −0.75
0.5 −0.5
0.25 −0.25

(b) Distribution of pore water pressure variations along the outer shaft

Figure 1.37. Lateral partially drained simulation, with an imposed

lateral displacement equal to 2.5 mm

Variations of pore water pressure with respect to the hydrostatic distribution Δpw
within interface elements outside the caisson are presented in Figure 1.37b at different
depths. The absolute magnitude of pressure variations is lower within the interface
than within the soil because its permeability limits the transversal flow to the interface.
There are almost no overpressures in front of the caisson, where compression of the
soil holds. This is thought to be due to the combination of high sand permeability and
a short drainage path close to the mudline.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 47

Figure 1.38 presents the influence of the displacement rate on the load–
displacement results. The strength and ductility increase with the increasing rate of
loading. However, this increase is less important than when it is observed in tension
simulations. It is not linear, and a very high rate does not involve a very large increase
in loading. Indeed, in laterally loaded cases, negative variations of pore pressures are
generated outside the caisson and dissipate much faster than in the tension case.

Results of simulations where the friction coefficient is equal to zero are also
presented in Figure 1.38. This clearly reduces the resistance to lateral loading. This
reduction arises from the loss of the horizontal shear stress component along the
shaft, which directly opposes the loading. In addition, vertical shear stress no longer
balances the rotational movement.
5

v = 0.5 mm/s v = 0.05 mm/s

4
μ = 0.5

3
∆Ftot [MN]

1 v → 0 mm/s
μ = 0.0
0

−1
0 5 10 15 20 25
Displacement [mm]

Figure 1.38. Partially drained lateral simulation of the caisson:

influence of the loading rate (0 is for a drained simulation)
and friction (μ coefficient equal to 0 or 0.5)

1.5. Conclusion and perspectives

1.5.1. Conclusion

The objectives of this chapter are to describe the formulation of a hydro-

mechanically coupled finite element of interface and to illustrate its importance in
the simulation of the soil–structure interaction in offshore engineering.

The main developments of a zero-thickness element are provided in sections

1.2 and 1.3. The contact constraint is ensured by a penalty method and computed
on a segment-to-segment basis. The flow discretization is taken into account by a
three-node discretization method, namely, the flow inside the interface is discretized

Copyright Iste 2022 / File for personal use of Robert Charlier only
48 Deterministic Numerical Modeling of Soil–Structure Interaction

separately from the flow inside volume elements. Three hydro-mechanical couplings
are introduced:
– definition of a total pressure (pN ) as the sum of effective contact (pN ) and pore
water (pw ) pressures;
– storage component due to gap opening (gN > 0);
– definition of fluid flow along the interface according to the cubic law.

A suction caisson typical of foundations for offshore wind turbines is further

investigated on pure tension and lateral loading. In both cases, the soil–structure
interactions are highly influenced by interface behaviors. Friction mobilization is
crucial in the case of pure traction, and is also involved in the resistance to lateral
loading. The gap creation under the lid and along the shaft of the caisson is also quite
important, since it decreases the contact zone and the possible mobilization of shear
stress. For all of these reasons, the use of a finite element of interface ruling contact
detection and friction mobilization is necessary.

Behavior of suction caissons is neither purely drained nor purely undrained.

Pore water pressures are generated and dissipated progressively during the loading
of the caisson. This partially drained behavior requires hydro-mechanically coupled
simulations. The coupled finite element of interface is developed for that purpose.
The results indicate that the strength of caissons to both tension and lateral loading
is increased by partially drained effects. In addition, the generation of increased
or decreased pore water pressure is a reversible process, while variations of
effective stress within the soil generate plastic strain. Therefore, the hydro-mechanical
couplings can help reduce the settlement of the caisson.

Movement of the caisson in both directions involves gap opening between the lid,
the shaft or both and the soil. In turn, negative variations of pore water pressures are
generated, which involves a suction effect, balancing the loading. This has a stronger
impact on tension than on lateral loading, but is significant in both cases.

1.5.2. Perspectives

A basic interface element was provided in this chapter and its importance was
demonstrated. It is easy to implement and use, but is not free from numerical issues.
Pure contact mechanics is a very dynamic field of research that provides new methods
to avoid oscillations in pressure distribution or discretization issues. The interested
reader should refer to [WRI 06] for an excellent summary of numerical methods
available.

Similarly, a basic constitutive law is described, but more advanced models (e.g.
dilatancy, cyclic degradation) are available, as described in the introduction. The

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 49

literature reporting these developments is mostly concentrated on material behavior,

and rarely on applications in offshore engineering. The main challenge regarding
complex constitutive laws is the multiplication of material parameters that must be
calibrated while in situ and experimental results are, most of the time, limited.

The fast development of offshore foundation technologies (e.g. very large piles,
new anchor types) and the increasing complexity of simulations are likely to
generalize the use of coupled interface elements. Therefore, it is necessary to
popularize these elements and provide guidelines for the choice of numerical
parameters.

1.6. References

[ACH 10] ACHMUS M., T HIEKEN K., “On the behavior of piles in non-cohesive soil under
combined horizontal and vertical loading”, Acta Geotechnica, vol. 5, no. 3, pp. 199–210,
2010.
[ACH 13] ACHMUS M., A KDAG C.T., T HIEKEN K., “Load-bearing behavior of suction
bucket foundations in sand”, Applied Ocean Research, vol. 43, pp. 157–165, 2013.
[ALO 13] A LONSO E., Z ANDARÍN M., O LIVELLA S., “Joints in unsaturated rocks: Thermo-
hydro-mechanical formulation and constitutive behaviour”, Journal of Rock Mechanics and
Geotechnical Engineering, vol. 5, no. 3, pp. 200–213, Elsevier, 2013.
[AND 08] A NDERSEN K., J OSTAD H., DYVIK R., “Penetration resistance of offshore skirted
foundations and anchors in dense sand”, Journal of Geotechnical and Geoenvironmental
Engineering, vol. 134, pp. 106–116, 2008.
[BAR 98] BARNICHON J., Finite element modelling in structural and petroleum geology, PhD
thesis, University of Liège, Belgium, 1998.
[BEL 91] B ELYTSCHKO T., N EAL M., “Contact impact by the pinball algorithm with penalty
and Lagrangian methods”, International Journal for Numerical Methods in Engineering,
vol. 31, pp. 547–572, 1991.
[BEL 98] B ELGACEM F., H ILD P., L ABORDE P., “The mortar finite element method for
contact problems”, Mathematical and Computer Modelling, vol. 28, no. 4, pp. 263–271,
Elsevier, 1998.
[BER 88] B ERKOWITZ B., B EAR J., B RAESTER C., “Continuum models for contaminant
transport in fractured porous formations”, Water Resources Research, vol. 24, no. 8,
pp. 1225–1236, 1988.
[BIE 12] B IENEN B., D ÜHRKOP J., G RABE J., R ANDOLPH M., W HITE D., “Response of
piles with wings to monotonic and cyclic lateral loading in sand”, Journal of Geotechnical
and Geoenvironmental Engineering, vol. 138, no. 3, pp. 364–375, 2012.
[BIE 18a] B IENEN B., K LINKVORT R.T., O’L OUGHLIN C., Z HU F., B YRNE B.W., “Suction
caissons in dense sand, Part I: Installation, limiting capacity and drainage”, Géotechnique,
vol. 11, pp. 1–47, 2018.

Copyright Iste 2022 / File for personal use of Robert Charlier only
50 Deterministic Numerical Modeling of Soil–Structure Interaction

[BIE 18b] B IENEN B., K LINKVORT R.T., O’L OUGHLIN C.D., Z HU F., B YRNE B.W.,
“Suction caissons in dense sand, Part II: Vertical cyclic loading into tension”, Géotechnique,
vol. 68, no. 11, pp. 1–15, 2018.
[BOR 13] B ORJA R.I., Plasticity: Modeling & Computation, Springer Science & Business
Media, Berlin, 2013.
[BOU 68] B OUSSINESQ J., “Mémoire sur l’influence des frottements dans les mouvements
réguliers des fluides”, Journal de Mathématiques Pures et Appliquées, 1868.
[BRA 09] B RANSBY M.F., Y UN G.-J., “The undrained capacity of skirted strip foundations
under combined loading”, Géotechnique, vol. 59, no. 2, pp. 115–125, 2009.
[BYR 02] B YRNE B., H OULSBY G., “Experimental investigations of response of suction
caissons to transient vertical loading”, Journal of the Geotechnical and Geoenvironmental
Engineering, vol. 128, no. 11, pp. 926–939, 2002.
[BYR 03] B YRNE B.W., H OULSBY G.T., “Foundations for offshore wind turbines”,
Philosophical Transactions of the Royal Society A: Mathematical, Physical and
Engineering Sciences, vol. 361, no. 1813, pp. 2909–2930, 2003.
[CER 15] C ERFONTAINE B., D IEUDONNÉ A., R ADU J., C OLLIN F., C HARLIER R., “3D
zero-thickness coupled interface finite element: Formulation and application”, Computers
and Geotechnics, vol. 69, pp. 124–140, 2015.
[CER 16] C ERFONTAINE B., C OLLIN F., C HARLIER R., “Numerical modelling of transient
cyclic vertical loading of suction caissons in sand”, Geotechnique, vol. 66, no. 2, pp. 121–
136, 2016.
[CHA 88] C HARLIER R., C ESCOTTO S., “Modélisation du phénomène de contact unilatéral
avec frottement dans un contexte de grandes déformations”, Journal de Mécanique
Théorique et Appliquée, vol. 7, no. 1, Gauthier-Villars, 1988.
[DAY 94] DAY R., P OTTS D., “Zero thickness interface elements – Numerical stability
and application”, International Journal for Numerical and Analytical Methods in
Geomechanics, vol. 18, pp. 689–708, 1994.
[DEJ 06] D E J ONG J., W HITE D., R ANDOPH M., “Microscale observation and modeling of
soil structure interface behavior using particle image velocimetry”, Soils and Foundations,
vol. 46, no. 1, pp. 15–28, 2006.
[DES 84] D ESAI C., Z AMAN M., L IGHTNER J., S IRIWARDANE H., “Thin-layer element
for interfaces and joints”, International Journal for Numerical and Analytical Methods in
Geomechanics, vol. 8, no. 1, pp. 19–43, 1984.
[FIS 06] F ISCHER K., W RIGGERS P., “Mortar based frictional contact formulation for higher
order interpolations using the moving friction cone”, Computer Methods in Applied
Mechanics and Engineering, vol. 195, no. 37, pp. 5020–5036, Elsevier, 2006.
[GAO 13] G AO Y., Q IU Y., L I B., L I D., S HA C., Z HENG X., “Experimental studies on
the anti-uplift behavior of the suction caissons in sand”, Applied Ocean Research, vol. 43,
pp. 37–45, 2013.
[GEN 90] G ENS A., C AROL I., A LONSO E., “A constitutive model for rock joints formulation
and numerical implementations”, Computers and Geotechnics, vol. 9, pp. 3–20, 1990.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 51

[GOO 68] G OODMAN R., TAYLOR R., B REKKE T., “A model for the mechanics of jointed
rock”, Journal of the Soil Mechanics and Foundations Division, vol. 94, 1968.
[GOU 08] G OURVENEC S., “Effect of embedment on the undrained capacity of shallow
foundations under general loading”, Géotechnique, vol. 58, no. 3, pp. 177–185, 2008.
[GOU 09] G OURVENEC S., ACOSTA -M ARTINEZ H., R ANDOLPH M., “Experimental study
of uplift resistance of shallow skirted foundations in clay under transient and sustained
concentric loading”, Géotechnique, vol. 59, no. 6, pp. 525–537, 2009.
[GUI 02] G UIDUCCI C., P ELLEGRINO A., R ADU J.-P., C OLLIN F., C HARLIER R.,
“Numerical modeling of hydro-mechanical fracture behavior”, Numerical Models in
Geomechanics, pp. 293–299, 2002.
[HAB 98] H ABRAKEN A., C ESCOTTO S., BANNING Q., “Contact between deformable
solids: The fully coupled approach”, Mathematical and Computer Modelling, vol. 28, no. 4,
pp. 153–169, 1998.
[HAL 85] H ALLQUIST J., G OUDREAU G., B ENSON D., “Sliding interfaces with contact-
impact in large-scale Lagrangian computations”, Computer Methods in Applied Mechanics
and Engineering, vol. 51, no. 1, pp. 107–137, Elsevier, 1985.
[HAR 09] H ARTMANN S., O LIVER J., W EYLER R., C ANTE J., H ERNÁNDEZ J., “A contact
domain method for large deformation frictional contact problems. Part 2: Numerical
aspects”, Computer Methods in Applied Mechanics and Engineering, vol. 198, no. 33,
pp. 2607–2631, Elsevier, 2009.
[HO 11] H O T., JARDINE R., A NH -M INH N., “Large-displacement interface shear between
steel and granular media”, Géotechnique, vol. 61, no. 3, pp. 221–234, 2011.
[HOU 05a] H OULSBY G., B YRNE B., “Design procedures for installation of suction caissons
in sand”, Proceedings of the Institution of Civil Engineers – Geotechnical Engineering,
vol. 158, no. 3, pp. 135–144, 2005.
[HOU 05b] H OULSBY G., I BSEN L., B YRNE B., “Suction caissons for wind turbines”,
International Symposium on Frontiers in Offshore Geotechnics, pp. 75–93, 2005.
[HOU 05c] H OULSBY G., K ELLY R., B YRNE B., “The tensile capacity of suction caissons in
sand under rapid loading”, Frontiers in Offshore Geotechnics, no. 1, pp. 405–410, 2005.
[HOU 06] H OULSBY G., K ELLY R., H UXTABLE J., B YRNE B., “Field trials of suction
caissons in sand for offshore wind turbine foundations”, Géotechnique, vol. 56, no. 1,
pp. 3–10, 2006.
[HU 04] H U L., P U J., “Testing and modeling of soil-structure interface”, Journal of
Geotechnical and Geoenvironmental Engineering, vol. 130, no. 8, pp. 851–860, 2004.
[IBS 14] I BSEN L.B., L ARSEN K.A., BARARI A., “Calibration of failure criteria for
bucket foundations on drained sand under general loading”, Journal of Geotechnical and
Geoenvironmental Engineering, vol. 140, no. 7, 2014.
[ISK 02] I SKANDER M., E L -G HARBAWY S., O LSON R., “Performance of suction caissons
in sand and clay”, Canadian Geotechnical Journal, vol. 39, no. 3, pp. 576–584, 2002.

Copyright Iste 2022 / File for personal use of Robert Charlier only
52 Deterministic Numerical Modeling of Soil–Structure Interaction

[JHA 14] J HA B., J UANES R., “Coupled multiphase flow and poromechanics: A
computational model of pore pressure effects on fault slip and earthquake triggering”, Water
Resources Research, vol. 50, pp. 3376–3808, 2014.
[JOH 92] J OHNSON K., Contact Mechanics, Cambridge University Press, New York, 1992.
[KEL 06] K ELLY R., H OULSBY G., B YRNE B., “A comparison of field and laboratory tests
of caisson foundations in sand and clay”, Géotechnique, vol. 56, no. 9, pp. 617–626, 2006.
[KLA 88] K LARBRING A., B JÖRKMAN G., “A mathematical programming approach to
contact problems with friction and varying contact surface”, Computers & Structures,
vol. 30, no. 5, pp. 1185–1198, 1988.
[KOU 14] KOURKOULIS R., L EKKAKUS P., G ELAGOTI F., K AYNIA A., “Suction caisson
foundations for offshore wind turbines subjected to wave and earthquake loading: Effect of
soil–foundation interface”, Géotechnique, vol. 64, no. 3, pp. 171–185, 2014.
[LEW 98] L EWIS R., S CHREFLER B., The Finite Element Method in the Static and Dynamic
Deformation and Consolidation of Porous Media, John Wiley & Sons, New York, 1998.
[LI 15] L I D., Z HANG Y., F ENG L., G AO Y., “Capacity of modified suction caissons in marine
sand under static horizontal loading”, Ocean Engineering, vol. 102, pp. 1–16, 2015.
[LIU 06] L IU H., S ONG E., L ING H., “Constitutive modeling of soil-structure interface
through the concept of critical state soil mechanics”, Mechanics Research Communications,
vol. 33, pp. 515–531, 2006.
[LIU 08] L IU H., L ING H., “Constitutive description of interface behavior including cyclic
loading and particle breakage within the framework of critical state soil mechanics”,
International Journal for Numerical and Analytical Methods in Geomechanics, vol. 32,
no. 12, pp. 189–213, 2008.
[MCC 56] M C C LELLAND B. et al., “Soil modulus for laterally loaded piles”, Journal of the
Soil Mechanics and Foundations Division, vol. 82, no. 4, pp. 1–22, 1956.
[MOR 02] M ORTARA G., B OULON M., G HIONNA V., “A 2-D constitutive model for cyclic
interface behaviour”, International Journal for Numerical and Analytical Methods in
Geomechanics, vol. 26, no. 11, pp. 1071–1096, 2002.
[NG 97] N G K., S MALL J., “Behavior of joints and interfaces subjected to water pressure”,
Computers and Geotechnics, vol. 20, no. 1, pp. 71–93, 1997.
[OLI 09] O LIVER J., H ARTMANN S., C ANTE J., W EYLER R., H ERNÁNDEZ J., “A contact
domain method for large deformation frictional contact problems. Part 1: Theoretical basis”,
Computer Methods in Applied Mechanics and Engineering, vol. 198, no. 33, pp. 2591–
2606, Elsevier, 2009.
[OLS 01] O LSSON R., BARTON N., “An improved model for hydromechanical coupling
during shearing of rock joints”, International Journal of Rock Mechanics and Mining
Sciences, vol. 38, no. 3, pp. 317–329, 2001.
[ORO 98] O RON A., B ERKOWITZ B., “Flow in rock fractures: The local cubic law assumption
reexamined”, Water Resources Research, vol. 34, no. 11, pp. 2811–2825, 1998.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Hydro-mechanically Coupled Interface Finite Element 53

[PUS 04] P USO M., L AURSEN T., “A mortar segment-to-segment contact method for large
deformation solid mechanics”, Computer Methods in Applied Mechanics and Engineering,
vol. 193, no. 6, pp. 601–629, 2004.
[RAG 19] R AGNI R., B IENEN B., S TANIER S., O’L OUGHLIN C., C ASSIDY M.,
“Observations during suction bucket installation in sand”, International Journal of Physical
Modelling in Geotechnics, pp. 1–49, 2019.
[REE 10] R EESE L., VAN I MPE W., Single Piles and Pile Groups under Lateral Loading, 2nd
edition, CRC Press, Boca Raton, 2010.
[SEG 08a] S EGURA J., C AROL I., “Coupled HM analysis using zero-thickness interface
elements with double nodes. Part I: Theoretical model”, International Journal for
Numerical and Analytical Methods in Geomechanics, vol. 32, no. 18, pp. 2083–2101, 2008.
[SEG 08b] S EGURA J., C AROL I., “Coupled HM analysis using zero-thickness interface
elements with double nodes. Part II: Verification and application”, International Journal
for Numerical and Analytical Methods in Geomechanics, vol. 32, no. 18, pp. 2103–2123,
2008.
[SEN 82] S ENPERE D., AUVERGNE G., “Suction anchor piles – A proven alternative to
driving or drilling”, Proceedings of the 14th Offshore Technology Conference, pp. 486–493,
1982.
[SEN 08] S ENDERS M., Suction caissons in sand as tripod foundations for offshore wind
turbines, PhD thesis, University of Western Australia, 2008.
[SEN 09] S ENDERS M., R ANDOLPH M., “CPT-based method for the installation of suction
caissons in sand”, Journal of Geotechnical and Geoenvironmental Engineering, vol. 135,
pp. 14–25, 2009.
[SHA 93] S HARMA K., D ESAI C., “Analysis and implementation of thin-layer element for
interfaces and joints”, Journal of Engineering Mechanics, vol. 118, no. 12, pp. 2442–2462,
1993.
[SHA 97] S HAHROUR I., R EZAIE F., “An elastoplastic constitutive relation for the soil-
structure interface under cyclic loading”, Computers and Geotechnics, vol. 21, no. 1,
pp. 21–39, 1997.
[STU 16] S TUTZ H., M ASIN D., “Hypoplastic interface models for fine-grained soils”,
International Journal for Numerical and Analytical Methods in Geomechanics, vol. 41,
no. 2, pp. 284–303, 2016.
[TER 25] T ERZAGHI K., Erdbaumechanik auf Bodenphysikalischer Grundlage [The
Mechanics of Earth Construction Based on Soil Physics], Deuticke, Leipzig, 1925.
[THI 14] T HIEKEN K., ACHMUS M., S CHRÖDER C., “On the behavior of suction buckets in
sand under tensile loads”, Computers and Geotechnics, vol. 60, pp. 88–100, 2014.
[TJE 90] T JELTA T., A AS P., H ERMSTAD J., A NDENAES E., “The skirt piled Gullfaks C
platform installation”, Proceedings of the Offshore Technology Conference, 1990.
[TRA 05] T RAN M., Installation of suction caissons in dense sand and the influence of silt and
cemented layers, PhD thesis, The University of Sydney, 2005.

Copyright Iste 2022 / File for personal use of Robert Charlier only
54 Deterministic Numerical Modeling of Soil–Structure Interaction

[TSA 81] T SANG Y., W ITHERSPOON P., “Hydromechanical behavior of a deformable rock
fracture subject to normal stress”, Journal of Geophysical Research: Solid Earth, vol. 86,
no. B10, pp. 9287–9298, 1981.
[WEI 15] W EISSENFELS C., W RIGGERS P., “A contact layer element for large deformations”,
Computational Mechanics, vol. 55, no. 5, pp. 873–885, 2015.
[WIT 80] W ITHERSPOON P., WANG J., I WAI K., G ALE J., “Validity of cubic law for fluid
flow in a deformable rock fracture”, Water Resources Research, vol. 16, no. 6, pp. 1016–
1024, 1980.
[WRI 01] W RIGGERS P., K RSTULOVIC -O PARA L., KORELC J., “Smooth C1-interpolations
for two-dimensional frictional contact problems”, International Journal for Numerical
Methods in Engineering, vol. 51, no. 12, pp. 1469–1495, 2001.
[WRI 04] W RIGGERS P., Z AVARISE G., “Computational contact mechanics”, in S TEIN E., D E
B ORST R., H UGHES T. (eds), Encyclopedia of Computational Mechanics, Volume 2: Solids
and Structures, pp. 195–226, John Wiley & Sons, Chichester, 2004.
[WRI 06] W RIGGERS P., Computational Contact Mechanics, 2nd edition, John Wiley & Sons,
Chichester, 2006.
[WRI 13] W RIGGERS P., S CHRÖDER J., S CHWARZ A., “A finite element method for contact
using a third medium”, Computational Mechanics, vol. 52, no. 4, pp. 837–847, 2013.
[ZAN 13] Z ANDARIN M., A LONSO E., O LIVELLA S., “A constitutive law for rock joints
considering the effects of suction and roughness on strength parameters”, International
Journal of Rock Mechanics and Mining Sciences, vol. 60, pp. 333–344, Elsevier, 2013.
[ZIE 00] Z IENKIEWICZ O.C., TAYLOR R.L., Finite Element Method: Volume 1, 5th edition,
Butterworth-Heinemann, Oxford, 2000.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 2

DEM Approach of the Modeling

for Geotechnical Structures in
Interaction with Reinforcements

The sophistication of geotechnical structures (new technologies and innovative

materials in civil engineering) implies the use of increasingly sophisticated
numerical models for their design. The ever-increasing improvement in the
computational capabilities, and the emergence of new methods of calculation, make
possible now what was impossible to envisage a few decades ago, particularly in
terms of duration and number of elements of the numerical simulations. Among the
numerical methods in full development are discrete methods. Although often used
for research purposes, these methods are increasingly used in engineering, in view of
the certain advantages that they represent.

2.1. Introduction

The SSI problem addressed in this chapter is studied using the DEM (Discrete
Element Method). The basic principle of discrete methods is to consider a material
as a set of particles that interact at their contact points. Therefore, these methods
ideally apply to granular materials due to their discrete nature and make it possible –
considering a limited number of parameters – to easily reproduce the behavior of
soil (load transfer mechanisms, arching effects, expansion, cracking or collapse).
These methods apply to quasistatic problems and see their main interest in problems
that involve large deformations or collapsed areas, as well as to cyclic or dynamic
applications, easily modeled given the specific mathematical formulation used in
DEM. Note that the use of discrete element modeling does not necessarily imply

Chapter written by Pascal VILLARD.

Copyright Iste 2022 / File for personal use of Robert Charlier only
56 Deterministic Numerical Modeling of Soil–Structure Interaction

having to faithfully represent either the real form or the number of grains present in
a geo-material. In any case, this approach would not be feasible from the point of
view of the present abilities of the computers. In fact, the shape, the distribution and
the number of granular elements are chosen for the simple purpose of having a
numerical material able to reproduce the macroscopic mechanical behavior of the
granular material. While the DEM is highly relevant for the modeling of discrete
materials such as soils, it is naturally less apt to reproduce the behavior of
reinforcement elements that are, by nature, continuous. To overcome this problem,
specific elements and contact law were implemented in DEM modeling to restore
the behavior of deformable reinforcements, and to take account of a realistic
description of the interface behavior between the soil and the reinforcement.

After presenting briefly the discrete element method and the theoretical concepts
used to take account of the soil–inclusion interaction, we will present different
illustrations of this method and, in particular, the applications to geotechnical
structures in interaction with rigid piles (quasistatic and cyclic loadings) or with
flexible and deformable reinforcements (geosynthetic sheets).

2.2. Discrete modeling

2.2.1. General concepts of the discrete modeling approach

The various discrete methods differ mainly in the resolution scheme employed,
the type of elements used, and the kind of interaction laws defined between the
elements in contact. Two main families of methods are usually employed:
the approaches resulting from contact dynamics and those derived from molecular
dynamics. The main differences between these two methods result from the
deformability tolerated or not between two elements at the contact points. In the
contact dynamics approach, initiated by Moreau and Jean [JEA 92, 99], particles
cannot interpenetrate and are considered non-deformable even at the contact points,
so that contact interactions are controlled by shock laws, involving temporal
discontinuities of velocities and forces.

The dynamic molecular method is also based on non-deformable elements but

those which are able to overlap slightly at the contact points. Based on molecular
dynamics, the DEM was first introduced by Cundall [CUN 71], for applications to
rock mechanics problems, and then applied to civil engineering by Cundall and
Strack [CUN 79]. Contact laws defined at the microscopic level make it possible to
establish the interaction forces taking into account the overlapping and relative
displacement between the two elements in contact. The numerical scheme used to
solve the quasistatic or dynamic problem is an iterative process that successively
alternates the resolution of Newton’s second law of motion, applied to the discrete

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 57

elements, and the actualization of the interaction forces at each contact point. The
motion equations are integrated using an explicit centered finite difference
algorithm, involving a time step Δt. This numerical scheme allows very large
displacements, alternative movements between the elements, shocks and dynamics
behaviors.

The basic elements used in the discrete element method are generally spherical
in shape since these forms have the advantage of facilitating the contact detection
process. Nevertheless, in an assembly of spheres, the rolling between particles is
such that it is difficult to simulate a natural soil presenting a high friction angle value.
In order to limit grain rotations, non-spherical elementary particles (polyhedron or
sphero-polyhedron) or clumps (agglomerates of rigid and non-breakable spherical
particles, which make it possible to optimize the contact detection) are most often
used. In some cases, rolling resistance laws can be introduced to limit the excessive
rotation of grains [IWA 98]. To avoid regular assemblies of particles that have
singular behaviors, elements of different sizes are almost systematically introduced.
The size distribution of elements, their shape and their arrangement have a
preponderant role in the macroscopic behavior of the granular material [SAL 09, 11].
Particular attention must therefore be paid when placing the elements, in order to
ensure the production of a homogeneous and isotropic assembly of particles at a
fixed density. For the following applications to reinforced structures, and in order to
realize numerical granular assemblies with various porosities, the particles were put
in place using the ERDF (Expansion Radius and Decrease Friction [CHA 05])
methodology. This methodology makes it possible to produce homogeneous and
isotropic particle assemblies.

Due to the fact that there is no explicit relation between the microscopic and
macroscopic parameters (Young’s modulus E, Poisson’s ratio ν, peak friction angle
ϕp and residual friction angle ϕr), a specific calibration methodology, based on the
trial-and-error process, is required. A classical way to obtain the micromechanical
parameters is to simulate laboratory tests, performed in well-controlled conditions
(as triaxial tests, for example) and compare the numerical results to the experimental
ones [SAL 09].

Classically, the interaction laws between two elements are most often expressed
as a function of interpenetration and relative displacement between the particles,
and are generally defined by a stiffness model, a slip model or possibly a cohesion
model. Numerous contact models have been developed, ranging from the simplest
based on linear elastic considerations, to the most advanced: nonlinear elasticity,
elastoplasticity [LUD 05] and viscosity [IWA 00]. For the following applications to
reinforced structures, a classical linear elastic contact law [DON 95], characterized by
a friction coefficient δ and two normal and tangential contact stiffness coefficients

Copyright Iste 2022 / File for personal use of Robert Charlier only
58 Deterministic Numerical Modeling of Soil–Structure Interaction

kn and ks, was used to define the contact behavior. The normal contact force can be
written as a function of the overlap between two elements using equation [2.1]:
 
{ Fn } = kn { Un } [2.1]


{ } the normal contact force between the elements and {U } the overlap in

with Fn n

the normal direction of the contact.

In the tangential direction, an incremental contact law, equation [2.2], is used to

link the increase in tangential contact force to the increase of tangential relative
displacement, calculated by considering the translation and the rotation of the two
particles in contact.
   
{ } { }
d Ft = k s d U t with Ft ≤ Fn tg ( δ ) [2.2]

{ } the tangential contact force and {U } the tangential displacement.


with Ft t

The normal contact stiffness kn (or the tangential contact stiffness ks) between
two spheres of radius Ri and Rj, expressed in N/m, is a function of the normal
rigidity Knij (or the tangential rigidity Ksij) of the two constitutive materials in
contact, expressed in N/m2, as defined in equation [2.3].

kn = Knij (Ri * Rj)/(Ri + Rj) [2.3]

Interaction laws, defined at a local scale, make it possible to reproduce the global
macroscopic behavior of the particle assembly, considering various loading paths
[SIB 19]. The macroscopic behavior of a set of particles strongly depends on
the density of the granular medium, the shape of the particles and the value of the
micromechanical contact parameters (normal and tangential stiffness, contact
friction coefficient, grading, particle shape and density of the numerical sample).
Usually, the macroscopic behavior restored leads to a granular material whose
macroscopic friction angle differs from the value of the microscopic friction angle
defined at the contact level, given the interlocking of the grains, which contributes to
increasing its mechanical characteristics.

The stress tensor σij within a volume V of the granular assembly can be
computed by equation [2.4] considering the contact forces acting at all contact
points included in the volume V [WEB 66]. Nc is the number of contact points in V,
f i is the projection of the contact force f on the i-axis and l j is the projection of the

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 59

branch vector l on the j-axis with i = x, y, z and j = x, y, z. The branch vector l is

defined by the vector linking the mass centers of the elements in contact.

Nc
1
σij =
V  f αi lαj [2.4]
α =1

2.2.2. Specific interaction between discrete particles and reinforcement

elements

The interface between the reinforcement elements and the soil is the privileged
area that allows the transmission of interaction forces from one element to another
(a completely smooth and non-frictional interface induces non-transmission of
frictional forces). The care taken in modeling the interface is therefore essential in
hoping to highlight the real behavior of structures in reinforced soil. The use of
discrete modeling makes it possible to very easily manage the interface behavior
by introducing appropriate force–displacement interaction laws. There are two
strategies for modeling boundary conditions or integrating reinforcement elements.

The first is to use a group of spheres associated together to form a wall or a

reinforcing element that can deform according to the constitutive interaction laws
defined between the spheres. The disadvantage of this method comes from the
fact that there is a microroughness between the group of spheres and the particles
of the soil that can be modified during the numerical simulation according to the
extensibility of the group of spheres. It is therefore difficult to control the interface
friction since uncontrolled structural friction is introduced during the simulation by
the microroughness between the elements. Another important point lies in the fact
that the discrete element method most often considers that the contact between
two elements is lost as soon as the elements separate. It is therefore very difficult
to ensure a regular transmission of the friction forces, from the particles of soil
towards the reinforcement constituted by a sphere assembly, due to the contact
microroughness that can involve a loss of contact when the relative displacements at
the interface are very large.

The second strategy is to use plane walls as boundary conditions and to develop
smooth specific elements to represent reinforcements as accurately as possible.
This technique is the most relevant because it does not introduce structural friction
(the macroscopic friction angle of the interface corresponds to the microscopic
friction angle defined at the contact) and ensures the continuity of the contact even
when the relative displacements at the interface are very large. This technique makes

Copyright Iste 2022 / File for personal use of Robert Charlier only
60 Deterministic Numerical Modeling of Soil–Structure Interaction

it possible to take account of very complex mechanisms acting at the interface as the
rolling mechanisms of soil particles, or the expansion or decrease in volume due to
the shearing forces.

Similar laws to those established between two basic particles, equations [2.1]
and [2.2], may be used to define the interface behavior between a planar element
(wall or reinforcement) and a soil particle. The normal stiffness kni at the contact
prevents the interpenetration of the elements in contact, whereas the tangential
stiffness ksi directly influences the rate of mobilization of the friction at the interface,
for the case of small relative displacements and until the maximum value of the
friction is obtained. The amplitude of the relative displacements at the interface and
the contact tangential stiffness directly influence both the intensity and the
orientation of the contact forces. The classical Mohr–Coulomb friction law, similar
to those presented Figure 2.1, is usually used.

τ τmax = σn tan δ i

E0
τ = σn tan δ i UM/U0

UM U0 U

Figure 2.1. Friction interface criterion

τ and σn denote the tangential and normal stresses acting at the interface, U the
relative displacement between soil particles and the planar element, U0 the value of
the minimal relative displacement necessary to mobilize the friction fully and E0 the
tangential stiffness modulus of the interface expressed in N/m3 and defined by
equation [2.5], where S represents the influence area of the contact (S = π d2/4 for a
spherical particle of diameter d).

E0 = ksi/S [2.5]

Note that δi, the microscopic friction angle of the interface, is equal to the
macroscopic friction angle that can be measured by a classical friction test. In
practice, U0 does not exceed a few millimeters. The intensity and the orientation of
the friction force depend strongly on the value retained for U0, especially for small
relative displacements [VIL 09]. As an example, the friction response (blue lines)
induced by the displacement along an elliptical path (red dashed line) of a spherical
particle, in contact with a fixed plane, is presented in Figure 2.2. A vertical force

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 61

keeps the contact persistent. We note that the smaller the value of U0, the sooner full
mobilization of the friction occurs so that the friction forces between the soil particle
and the plane are increasingly tangential to the trajectory given to the sphere.

U0 = 0.5 mm U0 = 1.25 mm

Maximal friction force Maximal friction force

3.5 mm

5 mm

Figure 2.2. Intensity and orientation of the tangential friction force for different
values of U0. For a color version of this figure, see www.iste.co.uk/grange/soil.zip

2.2.3. Numerical strategy for geotechnical structure modeling using

DEM

As mentioned, and due to the fact that a specific calibration methodology –

based on trial and error – was used to calibrate the micromechanical parameters, the
modeling of true-scale structures using DEM does not need to reproduce perfectly
the grain morphology, the particle size distribution and the particle number, as this
would imply a prohibitive calculation duration. However, to roughly mimic the
shape of the soil particles and to approach the macroscopic mechanical behavior of
real soil (particularly the internal friction angle at the peak) as well as possible, the
granular material is modeled by means of an assembly of clusters of different sizes.
The main results expected for applications to geotechnical structures are: the
displacements of the discrete particles, the network of contacts within the granular
assembly, the interaction forces between the soil and reinforcement, the stress
tensor, the change in structure of the granular material and the mechanism of
collapse.

2.3. Application of the DEM to geotechnical structures in interaction

with rigid piles

Soil improvement by rigid inclusions is a technique used in many countries

to limit surface settlements in highly compressible areas. This technique is based

Copyright Iste 2022 / File for personal use of Robert Charlier only
62 Deterministic Numerical Modeling of Soil–Structure Interaction

on the use of a network of concrete piles (assumed to be rigid due to their low
deformability, compared to that of the compressible soil) and a granular load
transfer layer positioned over the piles. This makes it possible to reduce the loads
acting on the compressive subsoil by redirecting one part of the vertical loads related
to the weight of the embankment and overloads to a rigid bedrock. There are two
common applications of soft soil reinforcement, based on the use of a network of
rigid piles: reinforcement under civil engineering buildings or embankments (road or
railway embankments). For civil engineering applications that require very low
differential settlements, a rigid slab is commonly employed.

The discrete element method was used to investigate the load transfer
mechanisms within the embankment and to take into account the significant
settlements of the compressible supporting soil, as well as the large deformations
within the granular embankment and the soil–pile interaction mechanisms. Some of
the advantages of the DEM compared, for example, to continuous methods, lie in the
fact that it is possible to take account of complex mechanisms such as the punching
of the granular embankment by the piles, the evolution of the mechanical behavior
of the granular soil following the grains rearrangement and that this method
integrates very large deformations.

Figure 2.3. Geometry of the network of piles [CHE 11]. For a

color version of this figure, see www.iste.co.uk/grange/soil.zip

Two applications will be proposed: the first concerns the behavior under
monotonic loading of a granular layer over piles for the two following applications,
under embankment and under building. The second concerns the study of the
evolution of load transfer mechanisms under cyclic behavior of a granular layer

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 63

above piles loaded by a rigid slab (wind turbine foundations, for example). The
typical geometry of the proposed simulations is shown in Figure 2.3. For reasons
of symmetry, a square mesh of rigid inclusions was considered. The pile caps
are assumed as having a square section of side a. The distance between two piles is
denoted s, and the covering rate of the network of piles (ratio of the surface of the
pile cap to the area of an elementary cell of the network of piles) is defined as α.

The typical discrete numerical sample used is composed of (Figure 2.3):

– a granular embankment with a height hm, made of a large amount of clusters.
Depending on the cases, the soil elements of the numerical sample have been placed
at different density states, according to the ERDF method, [CHA 05, SAL 09];
– for building applications, a rigid slab sited at the top surface of the
embankment making it possible to apply monotone or cyclic loadings. This slab
consists of two beds of spheres of the same diameter rigidly associated. Vertical
forces applied at the upper part of the slab make it possible to obtain a rather
uniform vertical stress;
– a compressible horizon whose action on the granular layer has been taken into
account by vertical springs of stiffness Kc (expressed in MPa/m) making it possible
to apply, at the base of the granular embankment, a vertical stress proportional to the
settlement of the supporting soil (Winkler linear model). Thus, the relation between
the settlement of the supporting soil Δh and the vertical stress σv applied to it is
given by Δh = σv/Kc, where Kc is the coefficient of compressibility of the soil,
which can be connected, for a subsoil layer of thickness H, to the oedometric
modulus of the soil by the relation: Kc = Εoedo/H. The value of Kc is chosen so that
centimeter displacements will be obtained during loading cycles. For example, a
compressibility coefficient Kc of 0.75 MPa/m reflects the behavior of a compressible
soil, 4 m thick, having an oedometric modulus of 3 MPa. Under these conditions, an
unreinforced soil would undergo a settlement of 0.1 m under an overload of 75 kPa.
It follows from this modeling that the soil is considered elastic and therefore the
compressible horizon may, during the loading–unloading cycles, be compacted or
disorganized freely, which would correspond to a configuration for which delayed
settlements of the supporting soil are excluded;
– boundary conditions to take account of the symmetry conditions by way of
frictionless vertical walls at the sides of the model. The piles are considered strongly
rigid. As a result, the vertical displacements at the interface, between the granular
material and the head of the piles, are fixed in the vertical direction. A Coulomb-
type contact law was taken into account at the soil–pile interface to reproduce real
friction mechanisms.

Copyright Iste 2022 / File for personal use of Robert Charlier only
64 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 2.4. Typical geometry of the numerical samples [CHE 11]

Parameters investigated are the intensity of the load distributions on the piles and
on the compressible soil, the surface settlements, the displacements of the granular
particles and the load transfer mechanisms, which can be quantified by two specific
parameters: the stress reduction rate, SRR, equation [2.6], defined as the ratio of the
load Fp applied on one pile to the total vertical external loads (w + q) acting on an
elementary cell of the network of piles, with w the slab and embankment weights
and q the overload, and the ability G of the load transfer layer to postpone the
overloads to the piles, equation [2.7], defined as the ratio of the incremental load
ΔFP acting on the piles to the overload applied q.

Fp
SRR = [2.6]
w+q

Δ Fp
G= [2.7]
q

The results expected are the subsoil settlement, the displacements of the discrete
particles, the network of contact forces within the granular embankment, the stress states
at different points of the granular layer and the contact forces between the granular soil
particles and the piles, in order to determine the efficiency of the pile network.

2.3.1. Load transfer mechanisms within granular embankments over a

network of piles

This work was performed in the framework of the PhD of Bastien Chevalier
[CHE 08], in collaboration with Gaël Combe, and of the Master’s thesis of Quoc
Anh Tran [TRA 15] in collaboration with Daniel Dias. Thanks to them.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 65

Two applications will be presented successively: the first showing the arching
effect that can be developed within a granular layer subjected to a uniform loading,
and the second, the load transfer mechanisms obtained resulting from the use of a
rigid loading slab [CHE 10, 11]. For these two applications, only a 2.5 m by 2.5 m
elementary cell was considered. The pile horizontal cross-section is 0.375 m by
0.375 m. Thus, the covering rate represents 2.19%. The load transfer embankments
are made up of an assembly of clumps, generated at the minimal porosity of 0.355
using the ERDF method [CHA 05, SAL 09].

Clumps are made of two imbricate spheres (of the same diameter d) whose centers
have a distance of 0.95 d between them. The shape ratio of the greater to smaller
particles is 4. Two thicknesses of the granular layer were considered: a load transfer
layer of 0.5 m in height made up of 16,000 particles and the other 1.0 m in height
made up of 32,000 particles. The micromechanical parameters and the corresponding
macromechanical parameters are given in Table 2.1. A friction angle δi of 30° was
considered at the interface between the granular elements and the piles. Four values
of Kc were used to determine the influence of the subsoil compressibility on the load
transfer mechanisms: Kc = 0.25, 0.5, 0.75 and 1.00 MPa/m. Several successive values
of the uniform vertical overload q were applied: 12.8 kPa, 25.5 kPa, 46.8 kPa and
68 kPa. The expected load transfer mechanisms are due to the differential settlements
produced by the contraction of the compressible soft soil.

Normal contact stiffness coefficient (N/m2) Knij 1.0 x 109

Ratio of tangential to normal contact stiffness (-) Ktij/Knij 0.75
Micromechanical friction coefficient (-) μ = tan δ 0.577
Young’s modulus (MPa) E 257
Poisson’s ratio (-) ν 0.08
Peak friction angle (°) ϕp 44.8
Residual friction angle (°) ϕr 30.1
3
Dry apparent density (kN/m ) γd 17.7

Table 2.1. Micro- and macromechanical parameters of the discrete model

The numerical results obtained for SRR and G are presented in Figure 2.5 for
the two studied granular embankment heights and for a subsoil compressibility
coefficient Kc of 0.75 MPa/m. It can be seen in this figure that SRR increases with
the total applied load (qt = w + q) to reach a threshold value, dependent on the height
of the granular embankment, that is much higher than the covering rate of 2.19%.
At the same time, it is observed that the capacity G of the granular mattress to
transfer efforts towards the piles is rather constant for given values of the granular

Copyright Iste 2022 / File for personal use of Robert Charlier only
66 Deterministic Numerical Modeling of Soil–Structure Interaction

embankment height (approximately 75% and 35% for embankment heights of 1.0 m
and 0.5 m, respectively).
100 100
90 SRR [%] G [%]
h m = 1.0 m 90
80 h m = 0.5 m 80
70 70
60 60
50 50
40 40
30 30
h m = 1.0 m
20 20
10 10 h m = 0.5 m
qt [kPa] q [kPa]
0 0
0 20 40 60 80 100 0 20 40 60 80

Figure 2.5. Load transfer efficiencies expressed as the SRR

and G ratios for Kc = 0.75 MPa/m and two values of hm [CHE 11]

The interaction mechanisms between the piles and the load transfers
embankment are shown in Figure 2.6, by way of the intensities of the displacements
of the particles of the granular material in a cross-section between two piles. In this
figure, we clearly distinguish two areas: the first located above the piles where the
movements are very low, the second above the compressible soil, where the
displacements are rather similar to the surface settlement. As it can be seen, these
two zones can be distinguished from each other by an inclined line, whose slope is
directly correlated to the peak friction angle of the granular material.

Figure 2.6. Displacements of the granular particles in a cross-section sited

between two piles for two values of hm and Kc = 0.75 MPa [CHE 10].
For a color version of this figure, see www.iste.co.uk/grange/soil.zip

Looking now into the case of the load transfer mechanisms when using a rigid
slab, we present in Figures 2.7 and 2.8 the SRR reduction rates and the capacity G of

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 67

the granular material to transmit the overloads to the piles. Due to the use of a rigid
slab, only a 0.5 m-thick granular layer is tested. In these figures, it can be clearly
seen that when the overload is applied on the granular embankment by a rigid slab,
the SRR reduction ratio and the capacity G to retransmit the external forces to the
piles are greater than in the previous case, where the overload is applied directly on
the top surface of the granular embankment.
100 100
SRR [%] h m = 0.5 m G [%]
90 h m = 0.5 m
90
80 80
70 70
60 60
50 50
40 Kc = 1.00 MP a/m
Kc = 1.00 MP a/m 40
Kc = 0.75 MP a/m
30 Kc = 0.75 MP a/m 30 Kc = 0.50 MP a/m
20 Kc = 0.50 MP a/m 20 Kc = 0.25 MP a/m
10 Kc = 0.25 MP a/m 10
qt [kP a] q [kP a]
0 0
0 20 40 60 80 100 0 20 40 60 80

Figure 2.7. Load transfer efficiencies when using a rigid slab expressed as
the SRR and G ratios for different values of Kc and for hm = 0. 5 m [CHE 11]

If we compare the movements and displacements of the particles of the granular

embankment for the two cases studied, with or without rigid slab (Figure 2.8), a
significant difference can be seen in terms of mechanical behavior. For example, the
most sheared soil zone, in the case of loading slab, is highly localized and is
essentially limited to the soil cylinder above the head of the piles. This soil cylinder
is highly compressed and the surface settlements obtained are strongly conditioned
by its mechanical behavior.

Figure 2.8. Displacements of the granular particles in a cross-section sited

between two piles for Kc = 0.75 MPa: (a) embankment and (b) under a rigid slab

Copyright Iste 2022 / File for personal use of Robert Charlier only
68 Deterministic Numerical Modeling of Soil–Structure Interaction

In order to better understand the influence of the soft soil compressibility on the
efficiency, we compare, in Figure 2.9, the results obtained with and without a rigid
slab for several values of the compressibility Kc. For very soft soils (Kc = 0.25 MPa/m
and 0.5 MPa/m), and without a rigid slab, the efficiency increases with the total load
and then decreases for greater values of the load.
50 100
SRR [%] Kc = 0.25 MP a/m (a) SRR [%] (b)
45 Kc = 0.50 Mpa/m 90
h m = 0.5 m
40 Kc = 0.75 Mpa/m 80
Kc = 1.00 Mpa/m
35 70
30 60
25 50
20 40
Kc = 1.00 MP a/m
15 30 Kc = 0.75 MP a/m
h m = 0.5 m 20 Kc = 0.50 MP a/m
10
5 10 Kc = 0.25 MP a/m
q [kP a] q t [kP a]
0 0
0 20 40 60 80 0 20 40 60 80 100

Figure 2.9. Comparison of the values of SRR versus the total load
applied qt for Hm = 0.5 m and different values of the subsoil
stiffness Kc: (a) embankment and (b) under a rigid slab

For soft soils with a greater stiffness (Kc = 0.75 MPa/m and 1.0 MPa/m) and
without a rigid slab, the load transfers increase regularly with the total load and then
reach a threshold. For very soft soils (Kc = 0.25 MPa/m and 0.5 MPa/m) and without
a rigid slab, the efficiency increases with the total load and then decreases for greater
values of the load. Note that, in this case, the maximum efficiency increases with
the increase of soft soil stiffness. In the case where the overload is transmitted to the
granular material, by means of a rigid slab, the observations are different: the more
the soil is compressible, the more transfers of loads to the piles are important. In
fact, this is due to the ability of the stiffer subsoil to withstand the overload when a
constant surface settlement is imposed by the rigid slab.

In order to compare the discrete element method and the finite difference
method, particular care was taken to perform numerical simulations in an equivalent
manner. Therefore, in the continuum approach, the cap yield model, accounting
for friction hardening and softening, has been used to capture the behavior of
the granular material. The main common parameters are: s = 3 m, a = 0.6 m,
hm = 0.75 m, 1.5 m, 2.25 m, 3 m and Kc varying between 0.05 to 1.0 MPa. Thus,
the covering rate is equal to 4%. Only a quarter of the elementary mesh was
considered for the two models, as presented in Figure 2.10. The DEM numerical
samples were made of an assembly of clusters of various sizes (dmax/dmin = 4),

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 69

composed of two overlapped spheres of diameter d with an angularity of 0.8 d;

16,000, 32,000, 48,000 and 64,000 clumps were used for the four considered
values of hm.

Figure 2.10. Numerical samples used: (a) DEM and (b) FDM [TRA 19].
For a color version of this figure, see www.iste.co.uk/grange/soil.zip

The DEM numerical sets of micromechanical parameters retained make it possible to

restore the behavior of loose (L), medium (M) and dense (D) granular materials, as
presented in Table 2.2. For the continuum model, the macromechanical parameters,
achieved by comparison with the DEM triaxial tests, make it possible to reproduce rather
similar behaviors of the three DEM numerical granular materials retained.

Numerical sample L M D
Numerical porosity 0.41 0.38 0.34
Knij (N/m²) 1.0 x 107 1.0 x 107 1.0 x 107
Ktij/Knij 1 1 1
μ = tan δ 0.84 0.84 0.84
φp (°) 34 40 46
φr (°) 26 26 26
3
γd (kN/m ) 1.47 1.55 1.65

Table 2.2. Micro- and macromechanical parameters of the discrete model

First comparisons between the two numerical models were made in Figure 2.11a,
considering the efficacy SRR of the load transfers versus the soft soil stiffness (Kc).

Copyright Iste 2022 / File for personal use of Robert Charlier only
70 Deterministic Numerical Modeling of Soil–Structure Interaction

For each case tested, a maximal value of the efficacy can be found, for which the
intensity is a function of the subsoil stiffness. For high values of the subsoil
stiffness, the load resulting from the weight of the granular embankment is
transferred mainly to the subsoil, strong enough to support the load applied. When
the stiffness of the subsoil decreases, the load transmitted to the piles increases to a
maximum value, characteristic of the maximum load transfer. For very low values of
the subsoil stiffness, the efficacy reduces dramatically due to the punching of
the granular embankment by the pile. In this case, some discrepancies between
discrete and continuum models were observed, due to the large displacement at the
soil–pile interface during the punching of the granular material. In this case, the
DEM model seems to be more relevant to take account of the large displacements of
the embankment, especially in the vicinity of the corner of the pile cap. The
maximal efficacy as a function of the density of the granular material, characterized
by its peak friction angle, is presented in Figure 2.11b.

Figure 2.11. Efficacy versus subsoil stiffness for the material density M and
maximal efficacy obtained for all the numerical simulations performed [TRA 19]

As can be seen in this figure, the comparison between DEM and FDM shows
similar trends and values for the maximal efficacy, especially in the case of dense
material, regardless of the embankment height. As can be expected, for a fixed
value of Kc, the maximal efficacy increases with the density and the thickness of
the embankment. Load transfer mechanisms can also be seen when looking at the
curve representing the efficacies SRR versus the shear rate (Figure 2.12), defined as
the ratio between the maximum vertical displacement of the granular embankment
and the embankment height hm. It can be seen that the efficiency increases with the
shearing ratio until the maximal efficacy value was obtained. When large shear
mechanisms are obtained, the efficacy decreases as a function of the density of
the granular embankment: a dense material leads to a greater decrease than a loose
material. These trends can be observed for both the FDM and DEM, despite the
disparity obtained between the two numerical models.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 71

Figure 2.12. Efficacy versus shear rate obtained for all the numerical
simulations performed for the material densities L and D [TRA 19]

The considerable reduction of the load transfer efficacy, observed in the discrete
model when dealing with great values of the shearing strain, cannot be reproduced
by the continuum model, especially for loose material. This can be attributed to the
ability of the DEM model to take account of the change in the mechanical behavior
of soil when large shearing and expansion of the granular soil occur. Moreover, the
corner of the pile cap represents a singularity for the FDM that prevents sliding of
the soil near the pile and introduces force concentration, which is not observed
with DEM.

2.3.2. Load transfer mechanisms within granular embankments over a

network of piles under cyclic loadings

This work was performed in the framework of the engineer internship of Agathe
Furet [FUR 13] in collaboration with Stéphane Grange. Thanks to them.

The potentiality of discrete element models to account for load transfer

mechanisms has already been tested and validated under monotonic loading. The
possible application of the numerical model to cyclic solicitations results from the
fact that the discrete element method explicitly takes into account the mechanisms
that develop at the micromechanical level, such as the irreversible displacements
between particles and the granular assembly rearrangement at each cycle, without
the need to implement complex interaction laws. For this reason, a simple
Coulomb-type frictional model is used in this study. To avoid dynamic phenomena,
the loading speeds (in force or in displacement), which can lead to rather long
simulation durations, must be sufficiently weak, especially when a large number of
loading cycles occur.

The work presented here [FUR 13] focuses on the evolution of the load transfer
mechanisms within a granular embankment loaded by a rigid slab and subjected to a

Copyright Iste 2022 / File for personal use of Robert Charlier only
72 Deterministic Numerical Modeling of Soil–Structure Interaction

large variety of cycles. Note that the interaction contact law used leads – when an
inversion of the relative displacement between the grains is initiated – to a gradual
decrease in tangential contact forces, until a complete change of orientation is
obtained when the relative displacements between grains increase significantly.
Moreover, when the maximum shear is reached and the normal contact forces
between the soil particles decrease, which will be the case during loading–unloading
cycles, there is logically a reduction in shear forces. Under these conditions, it is
clear that loading–unloading cycles will generate a large variation in the normal and
tangential contact forces between the grains and modify the load transfer mechanisms.

For this study, the compressible soil is elastic, and because of this, during
unloading cycles, can be discharged, which will generate changes in the structure
of the granular assembly and thus disrupt the mechanisms of load transfer. The
applications targeted by this type of modeling correspond to cyclic loadings cases,
for which delayed settlements are excluded. In the present study, only one subsoil
stiffness and several loading cycles of different amplitudes were tested.

The geometry of the network of piles refers to a square mesh of rigid inclusions
for which the distance between two adjacent piles is s = 2.5 m. The piles have a
square section of side a = 0.416 m. The covering rate of the network of piles is
2.77%. The embankment height is 0.5 m. For reasons of symmetry, only one
elementary cell of 2.5 m by 2.5 m was considered for modeling. The discrete
numerical sample is composed of 16,000 clusters made of two imbricate spheres of
the same diameter d spacing of 0.95 d, as defined previously. A compressibility
coefficient Kc of 0.75 MPa/m was retained for the subsoil stiffness, and a friction
angle of 30° was taken into account at the soil–pile interface. The micromechanical
and macromechanical parameters are similar to those defined previously (Table 2.1).

The numerical simulations were conducted according to the following processes:

– application of gravity to the particles of the granular material and of the
loading slab with the assumption of zero vertical displacement at the base of the
embankment. The density of the spheres is such that the initial vertical stress applied
by the slab on the granular material is 7 kPa;
– slow gradual displacement of the particles of the granular material due to
gravity until their stabilization without dynamic effect on the compressible soil;
– constant-amplitude cycles, from 0 to 60 kPa (load case A) and from 60 to
120 kPa (load case B), are imposed in successive increments on the slab. The main
objective of these simulations is to evaluate the influence of the loading rate of the
embankment on the evolution of the load transfer mechanisms. About 150 loading
cycles were performed. Several loading speeds have been tested to verify that within
the selected speed range, the dynamic effects are minimized.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 73

The results obtained in terms of efficiency of the load transfer mechanisms are
shown in Figures 2.13, 2.14 and 2.15. qt is the total stress applied comprising the
weight of the granular material, the weight of the loading slab and the overload, and
ds is the vertical displacement of the slab during the loading–unloading cycles.
Figure 2.13 shows the evolution of the vertical displacements of the slab during the
loading–unloading cycles. It can be seen that the displacements of the slab are rather
small but that they change over time during the loading cycles, and this in a similar
way for the two load cases studied. This increase in displacement of the slab during
loading cycles reflects a loss of efficiency of the load transfer mechanisms.
180
q t (kPa)
Loading process B, K c=0.75 MPa/m
160
Loading process A, K c=0.75 MPa/m
140

120

100

80

60

40

20
ds (m)
0
0 0,005 0,01 0,015 0,02 0,025 0,03 0,035 0,04

Figure 2.13. Evolution of the vertical displacements of the slab (ds) as a function
of the total load applied for the two selected loading processes (Kc = 0.75 MPa/m).
For a color version of this figure, see www.iste.co.uk/grange/soil.zip

The load transfer rates SRR are presented in the following figures, according to
the displacement of the loading slab (Figure 2.14) and to the number of cycles
(Figure 2.15). In Figure 2.14, we see that 50% of the loads resulting from the weight
of the embankment and of the loading slab are redirected to the piles before the
cyclic loading process. Subsequently, the influence of the loading slab on the load
transfer is clearly highlighted since the percentage of the total load redirected to the
piles increases progressively during the first loading, until reaching a maximum
value of 85% for the loading process used.

At the first loading cycle, it can be seen that the efficiency of the load transfer
system (granular embankment and loading slab) decreases during the unloading
phase as a result of the change in structure of the granular material and of the
network of contact forces following the elastic unloading of the compressible soil.
Unloading the granular embankment implies the reduction or cancellation of the

Copyright Iste 2022 / File for personal use of Robert Charlier only
74 Deterministic Numerical Modeling of Soil–Structure Interaction

tangential contact forces between the particles, following rearrangement and relative
displacements of the grains during unloading.

For loading process B (cycles from 60 to 120 kPa), this decrease in efficiency
remains low (approximately 8%). For loading process A (cycles from 0 to 60 kPa),
this decrease is much larger considering that the granular layer is at each cycle
completely discharged. After the first loading cycle, the efficiency rate SRR obtained
is, in this case, about 40%. This value is to be compared with the value of 50%,
corresponding to the efficiency of the load transfer system before setting up the
loading cycles. The load transfer mechanisms generated within the granular material
have therefore been altered in the first cycle and will continue to degrade over all
loading cycles, given that the efficiency of the load transfer system never reaches the
level of performance obtained during the phase related to the first loading.
100
SRR (%) Loading process B, K c=0.75 MPa/m
90
Loading process A, K c=0.75 MPa/m
80

70

60

50

40

30

20

10
ds (m)
0
0 0,005 0,01 0,015 0,02 0,025 0,03 0,035 0,04

Figure 2.14. Evolution of the efficiency of the load transfer system SRR as a
function of the displacements of the slab ds for the two loading processes.
For a color version of this figure, see www.iste.co.uk/grange/soil.zip

In Figure 2.15, we see (loading processes A and B) a gradual decrease in the

maximum efficiency of the load transfer system over the first 150 cycles studied.
This loss is about 10% over the entire loading–unloading cycles, which remains
acceptable given the high efficiency rates obtained. In the unloading phase, the
contrast in behavior between the two loading processes studied is very important.
This is related to the fact that in loading process B, the granular embankment is not
completely discharged, whereas in loading process A, it is completely unloaded.

The almost complete cancellation of the efficiency after unloading in case A means
that the granular embankment is highly unstructured during the loading–unloading

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 75

cycles and that the load transfer mechanisms that have been set up in the
embankment, during the first phase of the numerical process (application of the
gravity), were completely annihilated. On the other hand, after reloading the
embankment with an overload of 60 kPa, it can be seen that the efficiencies obtained
in loading process A are very close to those obtained after the unloading cycles in
the case of loading process B (overload of 60 kPa).

100
SRR (%) Loading process B, K c=0.75 MPa/m
90
Loading process A, K c=0.75 MPa/m
80
70
60
50
40
30
20
10
Number of Cycles
0
0 20 40 60 80 100 120 140 160 180

Figure 2.15. Evolution of the efficiency of the load transfer system SRR
as a function of the number of loading cycles for the two loading processes.
For a color version of this figure, see www.iste.co.uk/grange/soil.zip

To better distinguish the role of the deformation of the granular embankment on

the load transfer mechanisms, we have shown, in Figure 2.16, the efficiency curves
G of the load transfer system as a function of the displacement of the slab at each
loading cycle. Recall that G represents the capacity of the granular material to
postpone the overloads to the piles (and not the total load). It is defined as the ratio
of the load increment ΔFP acting on the piles to the overload increment applied Δq
and therefore ignores the load transfer mechanisms that occur during the set up of
the embankment and the loading slab.

It can be seen from this figure that for an equivalent deformation of the granular
material (ds unchanged), the capacity of the granular material to redirect the forces
towards the piles is identical to and independent of the two loading processes
studied. For a given deformation state, the capacity of the granular layer to mobilize
the load transfers is therefore unchanged. It will also be noted that the loading–
unloading cycles gradually disorganize the granular material, resulting in a loss of
efficiency G, which varies from 90% to 80% for all the cycles performed. It is
concluded that it is the changes in the granular structure and the rearrangement of
the grains that condition the load transfer mechanisms.

Copyright Iste 2022 / File for personal use of Robert Charlier only
76 Deterministic Numerical Modeling of Soil–Structure Interaction

100
G (%)
90
80
70
60
50
Loading process B, K c=0.75 MPa/m
40
Loading process A, K c=0.75 MPa/m
30
20
10
d s (m)
0
0 0,005 0,01 0,015 0,02 0,025 0,03 0,035 0,04

Figure 2.16. Efficiency of the load transfer system G to redirect overloads

to the piles according to slab displacements ds for the two selected loading
processes. For a color version of this figure, see www.iste.co.uk/grange/soil.zip

The numerical results show that, for the value of the subsoil stiffness tested and
for the levels of the amplitudes of the loading cycles carried out, the efficiency of
the granular embankment gradually decreases during the loading–unloading cycles
following a disorganization of its structure and modifications of the tangential
contact forces between particles. The efficiency values obtained at each loading
cycle nevertheless remain significant, even after 150 cycles. When the loading
cycles lead to the complete suppression of the overload (loading process A), it is
demonstrated that after a series of loading–unloading cycles, the load transfer
mechanisms that have developed in the load transfer system (following the set up of
the granular material and the loading slab) are completely annihilated, which means
that after a certain number of cycles, the total weight of the granular material and the
slab is supported by the compressible soil.

2.4. Application of the DEM to geotechnical structures in interaction

with flexible and deformable reinforcement – comparison with
experiment results

Due to their ease of implementation and their very competitive cost, geosynthetic
reinforcements are widely used for the reinforcement of geostructures.
Geosynthetics are most often in the form of a plane sheet made of woven fibers,
knitted or bonded together and which can mobilize, by their tensioning, tensile
forces that cannot be supported by the soil. To model these deformable
reinforcements, thin triangular flat elements, [GIR 97, VIL 98], defined by three
nodes, were implemented in the DEM computation code, [LEH 06, VIL 09]. To
ensure the continuity and progressiveness of the frictional forces from one element

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 77

to another during stretching of the sheet, cylinders and spheres have been positioned,
respectively, on the edges and at the nodes of each triangular element (Figure 2.17).

Figure 2.17. Finite elements used to ensure the

geometric continuity of the geosynthetic sheet

The behavior of the geosynthetic elements is governed by a stiffness matrix

[VIL 98] that takes into account each orientation and behavior of the fibers. The
basic assumptions of the formulation of these elements are:
– each element consists of a set of fibers with various orientations, initially
forming a plan;
– there is no slipping between the fibers (presence of connection points between
wires);
– the tensile forces acting in each fiber are oriented in the direction of this fiber
after deformation (large displacements).

The numerical parameters required for modeling the geosynthetic layer are: the
tensile stiffness of the geosynthetic in each fiber direction and the interface
parameters (kni, ksi and δι), as previously defined, between a soil particle and a plane.
The behavior of the geosynthetic elements is managed, as for conventional discrete
elements, by the alternation of Newton’s second law of motion applied to the nodes
of the elements, and the determination of the interaction forces. In the case of a
continuous geosynthetic sheet, the interaction forces that need to be taken into
account are those acting between the sheet elements (following their tensioning) and
those generated by the contact with the soil particles.

Two applications to soil structures reinforced by geotextiles will be presented,

for which the interaction mechanisms between the soil and the reinforcement
are essential to their survival. The first refers to retaining walls whose facing is

Copyright Iste 2022 / File for personal use of Robert Charlier only
78 Deterministic Numerical Modeling of Soil–Structure Interaction

made of geotechnical tubes filled with a granular material. The second concerns road
and rail embankments, reinforced by geosynthetic sheets subjected to localized
sinkholes.

2.4.1. Numerical and experimental behavior of geosynthetic tubes filled

with granular material

This work was performed in the framework of the PhD of Joanna Gorniak
[GOR 13]. Thanks to her.

The need to build retaining structures in areas with weak mechanical

characteristics led to the testing and construction of reinforced geostructures, made
of lightweight aggregates, encapsulated within geosynthetic tubes. These structures
are often used in geotechnical engineering mainly to reduce the bearing pressure
acting on the subgrades, while providing a high stability and significantly reducing
the settlements [WAT 04]. A quasi-vertical slope of the front of the reinforced
structures can be realized using stacked geosynthetic tubes, filled with lightweight
aggregates (Figure 2.18).

Figure 2.18. Construction phase of a retaining wall with a facing

made of geosynthetic tubes filled with granular material [GOR 15].
For a color version of this figure, see www.iste.co.uk/grange/soil.zip

These tubes are filled quickly, at a constant speed, using a specific process
developed in the framework of the TeMaSi research program, enabling a
homogeneous and uniform shape of the tube cross-section. The compaction process

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 79

needs only three or four passes of a compactor and provides a regular and constant
height of the tube. During these tests, it has been observed that the filling rate of the
tube is also dependent on the pressure from the blowing truck and the developed
filling device.

In situ experiments were carried out to analyze the behavior of these structures
under different types of loading. In particular, the geosynthetic tubes, used for the
facing of the retained structures, were the subject of specific instrumented
experiments of loading and unloading (Figure 2.19). The main concept of this
technique is that the geosynthetic tube provides a confining pressure to the
encapsulated granular clay material consecutively to the application of a vertical
loading, allowing the mobilization of tensile forces in the geosynthetic. To complete
this study, discrete numerical models (Figure 2.20) were used to better understand
the contribution of each of the components of the granular filled tubes, and to allow
the establishment of a design method.

Figure 2.19. Loading test performed on a geosynthetic tube filled

with granular material [GOR 16]. For a color version of this figure,
see www.iste.co.uk/grange/soil.zip

The experimental loading tests are performed on a 6 m-long geosynthetic tube

filled with lightweight granular materials. Several types of wrap-knitted geotextiles,
produced from different polymers, were used (referred to as PP-PVA, PVA, PP-1,
PP-2 and PES) with rather similar mechanical properties, i.e. an initial stiffness in
the radius direction of 59 kN/m and an initial stiffness in the longitudinal direction
of 43 kN/m.

Copyright Iste 2022 / File for personal use of Robert Charlier only
80 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 2.20. Geometry of the numerical DEM sample [GOR 13a].

For a color version of this figure, see www.iste.co.uk/grange/soil.zip

The aggregates are expanded clay aggregates, produced on the principle of

expansion of soft clay (LWA), showing a grain size distribution of 10–20 mm and
an internal friction angle between 35° and 38° [GOR 16]. Two main filling rates
(85% and 100%) were obtained depending on the materials and filling device used.
The filling rate of 100% was considered for the first group of tubes, constructed
of PP-PVA, PVA and PP-1 for a bulk density of 540–560 kg/m3, while
the second group, represented by tubes PP-2 and PES, has a bulk density of
490–498 kg/m3. A steel plate, positioned in the central part of the tube, makes it
possible to apply a vertical load, which was measured by load cell transducers. The
total compressive load applied to the geosynthetic tube equals 31 kN/m. The cross-
sectional geometry and the height of the tube were obtained from 3D scanner
measurements after each loading step.

The numerical loading is performed on 1.0-meter-long tubes with initial radius

of R0 = 0.3 m. Smooth vertical, rigid plates at both ends of the tube ensure the plane
strain boundary condition. To reproduce the macroscopic behavior of expanded clay
aggregates, 12,000 clusters made of two overlapping spheres of diameter d are used.
The distance between the centers of the two spheres of one cluster is equal to 0.5 d.
A ratio of 2 between the maximal and minimal diameters was used. The particles of
the granular material are generated at different porosities η using the expansion
radius and decrease of friction process (ERDF), in order to simulate different rates
of filling of the granular material included within the geosynthetic tubes. Values of
the micromechanical parameters and corresponding macromechanical parameters
are given in Table 2.3 for the five values of the tested porosities. E is the initial
Young’s modulus, φp is the friction angle at the peak, φr the residual friction angle at
the threshold, γpeak is the angle of dilatancy defined between 2% and 3% of vertical

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 81

strain and α is the residual volumetric deformation obtained at the threshold. A total
of 768 thin three-node triangular finite elements, forming a continuous fabric, are
used to describe the tensile and membrane behavior of the fabric. The interaction
between the discrete particles and the elements of the geosynthetic tube is taken into
account by a friction contact law, characterized by: E0 = 5 MN/m3 and a friction
parameter φi equal to 0° or 30° depending of the case studied. The stiffness of the
geotextile tube is, according to those of the experiment, 59 kN/m in the radial
direction and 43 kN/m in the longitudinal direction. Loading forces are imposed by
the regular vertical displacements of two rigid horizontal plates that interact with the
geosynthetic fabric on the top and at the bottom of the tube.

Micromechanical parameters of the numerical sample

η Dr Knij Ksij μ = tg δ
(-) (%) (N/m²) (N/m²) (-)
Soil-1 0.34 78.2
Soil-2 0.36 60.1
Soil-3 0.38 43.5 1.0 108 1.0 108 1
Soil-4 0.40 26.1
Soil-5 0.42 8.7
Macromechanical parameters of the numerical granular assembly
Dr E φp φr γpeak α ν
(%) (MPa) (°) (°) (°) (%) (-)
Soil-1 78.2 29 46 23 15 0.081
Soil-2 60.1 24 42 19 11 0.091
Soil-3 43.5 20 38 26 16 7.2 0.113
Soil-4 26.1 15 34 15 3.2 0.139
Soil-5 8.7 11 28 8 -0.5 0.162

Table 2.3. Micro- and macromechanical parameters

of the numerical granular materials used

To compare the experimental results to the numerical ones, a sensitivity study

was performed, based on the influence of the filling rate of the granular material and
its density and thus, its ability to expand or contract during shearing. This was
motivated by the fact that the experimental filling rates were not well known, even if
the experimental loading tests can be distinguished into two groups: tubes filled at
100% maximum rate and tubes filled at 85%. As a consequence, five numerical
materials (soil-1 to soil-5) set up at different porosities were used as granular filling
material. First, the own weight of the granular material was neglected regardless of

Copyright Iste 2022 / File for personal use of Robert Charlier only
82 Deterministic Numerical Modeling of Soil–Structure Interaction

the high intensity of the overload applied and the friction parameters φi between the
loading plate and the geosynthetic tube, and between the soil and the geosynthetic,
were set to zero for the sake of simplicity and due to the lack of relevant
measurement; the influences of these parameters will be investigated later.

The experimental data obtained after the maximal loading (Fv = 34.5 kN/m) are
compared in Figure 2.21 to the numerical results (loading curves defined as the
height of the geosynthetic tube versus the compressive force). As can be seen,
a relatively good correlation between the experimental and numerical results can be
found. A clear distinction between the two degrees of filling can be made. It can
be seen that the geosynthetic tube filled with the numerical granular material set up
with a high density (soil-1) is significantly more rigid than the geosynthetic tube
constituted by the granular material (soil-5), which has a high porosity. It can be also
seen that the experimental results obtained with the first group of tubes made of
PP-PVA, PVA and PP-1 for a bulk density of 540–560 kg/m3 are in accordance with
the numerical results obtained with dense numerical materials (soil-2 to soil-4),
while the experimental results obtained with the second group of tubes, represented
by tubes PP-2 and PES, for a bulk density of 490–498 kg/m3, are logically well
restored by the loose numerical material (soil-5).

Figure 2.21. Comparison between the experiment results obtained for

Fv = 34.5 kN/m and the numerical force–displacement curves (soil-1 to soil-5)
[GOR 16]. For a color version of this figure, see www.iste.co.uk/grange/soil.zip

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 83

In order to better understand the role of the geosynthetic–soil interaction, we

perform new loading tests following a protocol quite similar to the one previously
used, which consists of a vertical compression up to 30% of axial deformation,
decompression until a zero-compressive force and reloading up to 50% of the
vertical deformation.

Figure 2.22. Network of contact forces between the granular particles

at various stages of the loading–unloading numerical process [GOR 16].
For a color version of this figure, see www.iste.co.uk/grange/soil.zip

The networks of contact forces obtained at different stages of the numerical

loading–unloading–reloading process are shown in Figure 2.22. At the very
beginning of the loading test, it can be seen that the interaction forces between the
soil particles are highly concentrated in the vicinity of the contact points between
the loading plates and the geosynthetic tube. This zone expands rapidly following
the increase in loading; the central rectangular part of the filling tube being subjected
to rather uniform vertical and horizontal stresses. The stress tensor in this area can
be computed using equation [2.4]. From that, it can be seen that the horizontal and
vertical stresses are the minor and minimum stresses, respectively. The ratio of the
horizontal to vertical stresses is equal to 0.32 at 40% of the vertical deformation that
corresponds to the active pressure coefficient of the numerical granular material.
After unloading, this ratio equals 2.75, which refers to the passive pressure

Copyright Iste 2022 / File for personal use of Robert Charlier only
84 Deterministic Numerical Modeling of Soil–Structure Interaction

coefficient of the numerical soil. The ability of the geosynthetic tube to withstand
vertical forces is obviously related to the rigidity of the geosynthetic reinforcement,
which by confinement effect makes it possible to maintain a high lateral pressure.
During unloading, the horizontal stresses remain, thanks to the beneficial action of
the geosynthetic, whereas the vertical stresses logically vanished. The reloading
of the tube leads to a state of stress similar to that obtained during the first loading.
We will already underline the ability of the numerical model to account for the very
large deformations of the geosynthetic tube and of the granular particles assembly.

Additional numerical simulations were carried out in order to better understand

the influence of the friction effects (not taken into account previously) at the
interface between the geosynthetic fabric and the filling soil, and between the
loading plate and the supporting soil, using a friction coefficient set at 30°. As can
be seen in Figure 2.23 (a – without friction, b – with friction between the fabric and
the filling soil, c – with friction between the fabric and the loading plate and the
foundation soil and d – with friction between the fabric and the soil, the loading
plate and the foundation soil), the friction between the geosynthetic fabric and the
granular material can be neglected without any significant influence on the results.
The influence of the friction between the geosynthetic fabric and the supporting soil
on the loading plate is very important.

Figure 2.23. Comparison of the numerical loading curves assuming

different friction parameters between supporting soil, loading plate,
geosynthetic fabric and granular material [GOR 13a]

The influence of the friction on the distribution of the circumferential tensile

force in the geosynthetic fabric, at 50% of the axial deformation, is presented

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 85

in Figure 2.24. It can be seen that without friction between the loading plate, the
supporting soil and the geosynthetic fabric (cases a and b), the distribution of tensile
stresses is rather constant along the circumference of the tube. The presence of
friction between the geosynthetic tube, the loading plate and the supporting soil
(cases c and d) has both a significant effect on the loading capacity of the
geosynthetic tube and on the tensile force distribution along the circumference of the
tube. As can be seen, the friction forces have a beneficial effect on the loading
capacity of the geosynthetic tube when subjected to vertical loading.

Figure 2.24. Comparison of the numerical tensile forces in the geosynthetic

fabric assuming different friction parameters between supporting soil,
loading plate, geosynthetic fabric and granular material [GOR 13a]

An analytical design method was developed on the basis of the true scale
experiments and of the numerical results, in order to obtain a practical tool that
can be used in engineering. This design method will be used for comparison and
validation of the numerical model using simplifying assumptions. Particular
attention was paid to estimate the loading capacity of the tube and to the values of
the average tensile forces in the geosynthetic. The main assumptions used for the
establishment of the loading force–displacement relation are:
– the geosynthetic tube is a three-dimensional infinite long elastic membrane
of initial radius R0 subjected to vertical external forces;
– the geosynthetic fabric is considered as a sufficiently thin membrane with
no compressive and bending stiffness. The tensile behavior of the geosynthetic
is governed by an elastic law (T = J ε);
– the filling material is a frictional granular soil with a friction angle ϕ and an
expansive coefficient at large deformations denoted α. Taking into consideration
a dilatant soil leads to the assumption that the increment of tube’s volume is
proportional to the initial volume. During the loading and unloading process, the
granular soil is assumed to be at the limit state of its stability;

Copyright Iste 2022 / File for personal use of Robert Charlier only
86 Deterministic Numerical Modeling of Soil–Structure Interaction

– the supporting soil and the loading plate are modeled as frictionless and
rigid wall;
– the unit weight of the structure is neglected in view of the intensity of the
applied vertical load, as well as the friction forces between the filling soil and
the geosynthetic.

From these assumptions, and by studying the static equilibrium of any part of the
loaded tube, it can be demonstrated [GOR 15] that:

T=
(
J 4R 02 (1 + α ) + H 2 − 4R 0 H ) [2.8]
4 H R0

Fv =
(
π J 4 R 02 (1 + α ) − H 2 ) ( 4 R 02 (1 + α ) + H 2 − 4 R 0 H ) [2.9]
8 H3 R 0 K a

where T is the average tensile force in the geosynthetic, equation [2.8], Fv is the
load-bearing capacity of the filled tube, equation [2.9], H is the height of the
deformed tube and Ka is the coefficient of the active soil pressure.

The comparison between the results of the numerical model and of the analytical
formulation is represented as load–settlement curves (Figure 2.25). Similar
assumptions were used in both cases, so that the weight of the granular material was
neglected and the friction between the plate, the supporting soil, the geosynthetic
fabric and the granular material was set to zero.

Figure 2.25. Comparison between the numerical and analytical

curves in terms of compressive force versus settlement [GOR 16]

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 87

The five numerical granular materials (soil-1 to soil-5) were considered in order
to test the ability of the numerical model to reproduce the loading behavior of the
granular material filling tube, using different soil behaviors (expansive or
contractive soils). Due to the large deformation of the geosynthetic tube, the
comparison was established using the characteristics of the numerical granular soils
obtained under large displacements (φ = φr = 26°). The expansion coefficients α
used for the analytical design method are those presented in Table 2.3.

From the obtained results, it was shown that the loading capacity of the
geosynthetic tube is a function of the expansive coefficients of the granular material.
Except for very large values of the expansion coefficient (soil-1), we can observe a
very good correlation between the numerical and experimental results. In fact, rather
high geosynthetic stiffness disturbs and prevents the expansion of the soil at very
large deformation rates. Nevertheless, the good agreement between the numerical
and analytical results demonstrates the relevance of the proposed analytical design
method and the capability of the discrete element model to take account of the
interactive phenomenon between the soil and the reinforcement.

2.4.2. Numerical and experimental behavior of granular embankments

reinforced with geosynthetic in areas prone to subsidence

This work was performed in the framework of the PhD of Audrey Huckert
[HUC 14] in collaboration with Laurent Briançon. Thanks to them.

The progressive development of the territory leads to the exploitation of new

zones, currently neglected, because they present risks for the safety of users. This is
the case of areas of potential subsidence that are related to the presence of
underground cavities. The recognition techniques of these cavities generally do not
make it possible to detect them exhaustively, and required filling works are often
very expensive, without being able to systematically provide the guarantee of a
durable solution. In recent years, the development of the geosynthetic reinforcement
technique positioned at the base of the embankments has made it possible to provide
economical and technical solutions for temporarily limiting the consequences of a
localized sinkhole by limiting the surface deflections to acceptable values, necessary
for the continuation of the traffic. The main mechanisms involved in this issue
are the flexion behavior of the geosynthetic, the load transfer mechanisms within
the part of the granular embankment located above the cavity and the progressive
stretching of the geosynthetic obtained by friction in the anchoring zones, located on
either side of the cavity.

Although significant progress had been made in recent decades to improve the
design methods of reinforced geosynthetic embankments above cavities [BRI 08,

Copyright Iste 2022 / File for personal use of Robert Charlier only
88 Deterministic Numerical Modeling of Soil–Structure Interaction

VIL 08], some scientific challenges had to be removed. The use of a discrete
numerical model, coupled with full-scale experiments, has proved to be an
appreciable help to clarify different assumptions such as the shape of the distribution
of the vertical load acting on the geosynthetic layer or to better understand the
phenomenon of expansion of the soil cylinder located above the cavity. These new
results made it possible to modify the existing analytical methods towards more
realistic design methods [HUC 16, VIL 16, CHA 19].

The experiments carried out in the framework of the FUI Geo-inov project
[HUC 14] consist of simulating the gradual opening of a circular cavity under a
granular embankment, reinforced by a geosynthetic sheet. Three experimental tests
using different types of soil and geosynthetics have been realized. The instrumentation
carried out made it possible to measure, at each stage of the opening process of the
cavity and for each loading phase, the surface settlements, the displacements within
the granular embankment, the geosynthetic strains and deflections, as well as the load
transfer mechanisms within the embankment, using pressure sensors positioned at the
edges of the cavity. The mechanical parameters of the granular material and of the
interface between the soil and the geosynthetic were determined by laboratory
characterization tests (shearing and friction tests using 0.3 m x 0.3 m boxes). We will
then focus on the full-scale experiments carried out using a granular soil embankment
(20/40 mm sized gravel) with a thickness hm of 1.0 m and a PET geosynthetic
consisting of a non-woven support reinforced by wires in one direction.

Figure 2.26. Geometry of the reference numerical sample [HUC 14a]

The discrete numerical model was used to take into account the large
deformations in the granular embankment during expansion, shearing or compaction,
to follow the evolution of the contact forces between the particles during the
opening of the cavity, and to integrate the complex interaction mechanisms acting at
the interface between the soil and the geosynthetic as rolling or friction. The
expected results are the displacements of each particle of the granular material, the

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 89

deformations and tensions of the geosynthetic layer, as well as the interaction forces
at the soil–geosynthetic interface. Several geometries of the numerical model, each
adapted to the diameter of the cavities tested, were used. Due to the symmetry, the
cavity was positioned in one corner of the numerical sample (Figure 2.26).

The reference numerical sample used for comparison with the experimental
results consists of:
– a set of discrete particles set up at a fixed porosity. Each numerical element,
whose shape is close to that of the real forms of grain, consists of clumps made of
two spheres of the same diameter d juxtaposed to one another. For the reference
numerical model (4 m x 4 m), 32,000 elements of different sizes were used, the
larger particles having, just as for the experimental material, a size twice as large as
the size of the smaller particles;
– a continuous geosynthetic sheet consisting of three-node triangular elements
[VIL 98]. Different tensile stiffnesses, in the direction of reinforcement and in the
perpendicular direction, were considered to restore the anisotropic behavior of the
geosynthetic used during the experiments. For the reference numerical sample,
12,800 triangles were used;
– a set of spheres regularly arranged and positioned at the base of the numerical
model to simulate the action of an elastic subsoil. The positions of these spheres are
modified to simulate the opening of the cavity. Two opening mechanisms were
studied: a gradual opening of the cavity by increasing its diameter (until obtaining a
diameter D, process A) and a progressive lowering of all the spheres of the
supporting soil, located under the cavity for a fixed diameter cavity D (process B).
For the reference numerical sample, 12,800 spheres were used to model the subsoil;
– frictionless walls positioned at the sides of the numerical sample and boundary
conditions in displacement (on the horizontal directions) for the nodes sited at the
periphery of the geosynthetic layer, to ensure the symmetry of the problem.

During the numerical process, the granular embankment is subjected to gravity

and then to the sinkhole test by displacement of the spheres of the subsoil located
above the cavity, following process A or B. The microscopic parameters of the soil
particles, presented in Table 2.4, have been calibrated [SAL 09] to account for the
mechanical behavior of a low-density granular material with an angle of friction at
the peak of 36.5°, a friction angle at rest of 31° and a Young’s modulus of
19 MPa. The interface behavior between the elements of the granular material and
the geosynthetic is restored by a Coulomb-type contact law using parameters given
in Table 2.4. The bidirectional behavior of the geosynthetic used in the experiments
is taken into account in the numerical model using two specific tensile stiffnesses,
one in the direction of reinforcement (Jx = 3,000 kN/m) and the other in the
perpendicular direction (Jy = 250 kN/m). The results of the numerical model are

Copyright Iste 2022 / File for personal use of Robert Charlier only
90 Deterministic Numerical Modeling of Soil–Structure Interaction

confronted here with the experimental results considered to be the most

representative. In this case and at any point of view, the numerical process is similar
to the experimental protocol, namely a gradual opening of the cavity diameter up to
a maximum of nominal diameter of 2.2 m.

Numerical parameters allowing us to define the granular material behavior

Micromechanical friction parameter: μ = tan(δ) (-) 0.53
Normal contact stiffness: Knij (N/m²) 1.0 x 108
Tangential contact stiffness: Ktij (N/m²) 1.0 x 108
Numerical parameters of the soil–geosynthetic interface
Friction angle for the upper interface: ϕi-sup (°) 23
Friction angle for the lower interface: ϕi-inf (°) 40
Tangential contact stiffness: E0 (MN/m3) 15

Table 2.4. Numerical parameters used to reproduce the mechanical behavior of the
granular material and the interface behavior between the soil and the geosynthetics

In Figures 2.27 and 2.28, the surface settlements, the vertical displacements and
the strain of the geosynthetic sheet, for a cavity opening diameter of 2.2 m, are
compared. As can be seen in Figure 2.27, the surface settlements and the vertical
displacements of the geosynthetic are fairly well reproduced by the numerical
model. It can also be noted that the geometry of the numerical and experimental soil
areas involved in the collapse is located mainly above the cavity. By comparison
between the volumes of soil before (cylinder above the cavity) and after collapse
(delimited by the surface deflection and the deformation of the geosynthetic), it is
possible to estimate the coefficient of expansion of the numerical soil at 1.048,
which remains quite close to the experimental value of 1.037. This result is very
satisfactory if we consider that the shape of the particles and the initial density of the
embankment granular material were only very roughly taken into account in the
numerical modeling.

If we compare (Figure 2.28) the numerical values of the strains of the

geosynthetic sheet with the values recorded by the Bragg gratings, we can also see,
given the level of precision that can be achieved in situ, a fairly good match between
the results. In order to analyze the load transfer mechanisms within the granular
embankment, it is possible to numerically determine the vertical stresses in the
vicinity of the cavity by considering all of the vertical forces acting on the
geosynthetic sheet around the periphery of the cavity (annular surface defined by
internal and external radii of 1.1 m and 1.5 m) and compare this value to the
measurements provided by the total pressure sensors. At the end of the process of

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 91

opening the cavity at D = 2.2 m, the stress increase obtained with the numerical
model on the edges of the cavity is 34% for an experimental value of 49%. Again,
given the potential uncertainties related to measurements (proximity to the edges of
the cavity and possible tilting of the sensors), the results obtained are satisfactory.

Figure 2.27. Comparison between numerical and experimental

results of geosynthetic surface settlements and deflections [VIL 16]

1,4
Numerical results
1,2
Experimental results
Geosynthetic strain (%)

0,8

0,6

0,4

0,2

0
-4 -3 -2 -1 Cavity 1 2 3 4

Linear abscise from the center of the cavity (m)

Figure 2.28. Comparison between the numerical and experimental

results of the strains in the geosynthetic sheet [VIL 16]

The numerical tool makes it possible to measure specific quantities and to

investigate complex phenomena which are difficult to quantify or to apprehend by in
situ measurements. For example, it is possible to numerically determine the load

Copyright Iste 2022 / File for personal use of Robert Charlier only
92 Deterministic Numerical Modeling of Soil–Structure Interaction

transfer mechanisms that are related to a gradual change in the orientations and
intensities of the interaction forces between grains of the granular embankment,
following a modification of the boundary conditions. These load transfer mechanisms
are highly dependent on the loading history and thus the opening mode of the cavity.

In a simple way, the efficiency of the load transfer mechanisms can be defined as
the ratio between the vertical loads transmitted at the edges of the cavity and
the weight w of the soil cylinder sited above the cavity. Knowing the interaction
force Fg between the particles of the granular material above the cavity and the
geosynthetic, it is possible to quantify the efficiency Eff of the load transfer using the
relation Eff = (w - Fg)/w. Similarly, knowing the interaction forces acting at any
point of the sheet, it is possible, taking account of annular areas centered on the
cavity, to accurately determine the geometry of the distribution of vertical stresses
on the sheet.

The influence of the opening mode of the cavity on the efficiency of the
load transfers is highlighted in Figure 2.29. It should be emphasized that for this
study, the size of the numerical models and the number of elements have been
adapted to the cavity diameters: 0.5 m x 0.5 m x 1.0 m mesh for cavity diameters
of 0.1–0.2 m (model A), mesh size of 1.0 m x 1.0 m x 1.0 m for cavity diameters of
0.2–0.5 m (model B) and meshes of 4.0 m x 4.0 m x 1.0 m for cavity diameters
of 0.5–2.5 m (reference model).

100
90
80 Process B (Models A, B, C)
70
Efficacy (%)

60
50
40
30 Process A (Model A)
20 Process A (Model B)
10 Process A (Model C)
D/hm
0
0 0,4 0,8 1,2 1,6 2 2,4 2,8 3,2 3,6 4

Figure 2.29. Efficiency of the load transfer according to the ratio D/hm
for the two modes of cavity opening [VIL 16]. For a color version
of this figure, see www.iste.co.uk/grange/soil.zip

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 93

For each cavity diameter tested (process B), two points are plotted in Figure 2.29.
The first characterizes the maximum efficiency of the load transfer obtained during
the opening process for small deformations of the granular embankment, and the
second characterizes the residual efficiency obtained after the complete opening of
the cavity. The difference between these two points increases with the size of the
cavity and the rate of deformation of the granular embankment. In Figure 2.29, it
can also be seen that the values of the efficiency of the load transfers, from one
mode of opening to the other, are quite close and whatever the cavity diameter
considered (the values of the efficiency of the load transfers obtained with process
A being between the two values obtained during process B).

On the other hand, the values of the surface settlements and the displacements
of the geosynthetic sheet, presented for a particular value of the ratio D/hm in
Figure 2.30, are very different from one opening mode to another, which suggests
that the geometry of the load distributions on the geosynthetic sheet varies
depending on the mode of opening of the cavity. Moreover, the values of the
expansion coefficients of the granular material sited above the cavity (1.048 for
process A and 1.036 for process B) are slightly different from one opening mode
to the other.

Logically, as can be seen in Figure 2.31, the expansion coefficient obtained with
process A is greater, since the expanded and sheared area during the opening process
is greater because of the gradual increase in the diameter of the cavity.

Cavity
Cavité (D=2,2 m)
(D = 2,2 m)
1,2
1
0,8
0,6
hH
m= m
=1m
0,4
0,2
0
-0,2
-0,4
-3 -2 -1 0 1 2 3
Linear abscise
Distance fromau
par rapport thecentre
center
deof the cavity
la cavité (m) (m)
ProcessB
Procédure B ProcessA A
Procédure

Figure 2.30. Comparison between the surface settlements and the vertical
displacements of the geosynthetic sheet depending on the opening mode
of the cavity (D/hm = 2.2 m) [VIL 16]

Copyright Iste 2022 / File for personal use of Robert Charlier only
94 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 2.31. Change in porosity within the granular embankment

depending on the opening mode of the cavity (D/hm = 2.2 m) [VIL 16].
For a color version of this figure, see www.iste.co.uk/grange/soil.zip

To explain the difference in behavior obtained as a function of the cavity

opening mode, Figure 2.32 shows the stress distributions on the geosynthetic sheet
using processes A and B for D/hm = 2.2. This figure shows a very clear difference in
the geometry of the load distributions on the sheet, namely: a conical distribution for
process A (with load concentrations in the center of the cavity)
and a rather constant or inverted load distribution for process B (with a higher
concentration of force in the vicinity of the edges of the cavity).
2,5
(Δ σ / σ0) Process A
2
Process B
1,5

0,5
D/hm = 2.2
0
0 1 2 3 4
Linear abscise from the center of the cavity (m)

Figure 2.32. Comparison of the geometries of the load distribution acting on

the geosynthetic sheet, depending of the opening process mode [HUC 14a]

This is due to the fact that the load transfers, towards the edges of the cavity, are
systematically modified by the gradual increase of the diameter of the cavity

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 95

(process A), whereas they are disturbed little throughout process B. At the end of the
opening processes, the shear forces at the periphery of the collapsed soil cylinder are
quite similar from one cavity opening mode to the other. This explains why the load
transfer efficiencies are very close in the two cases.

The networks of the contact forces obtained, depending on the cavity opening
mode for three cavity diameters, are shown in Figure 2.33. Again, it is found that the
opening process B of the cavity to a fixed diameter, allows a direct transfer of loads
due to the weight of the embankment to the edges of the cavity. For process
A, however, these load transfer mechanisms are systematically challenged by the
gradual increase in the diameter of the cavity.

Figure 2.33. Comparison of the contact force distributions depending on the cavity
opening mode for three cavity diameters [VIL 16]. For a color version of this figure,
see www.iste.co.uk/grange/soil.zip

This study made it possible to show that, during the formation of a cavity under a
granular embankment reinforced by geosynthetic, the load distribution acting on the
sheet is not uniform and is largely influenced by the opening mode of the cavity.
In particular, a conical load distribution, for which the load is greater in the center of
the cavity, has been obtained when the diameter of the cavity increases gradually.
This is related to the fact that the load transfers that take place at
the beginning of the formation of the cavity, are constantly challenged during the
gradual increase of its diameter. On the contrary, and as shown by the contact force

Copyright Iste 2022 / File for personal use of Robert Charlier only
96 Deterministic Numerical Modeling of Soil–Structure Interaction

networks, when the supporting soil under the embankment moves progressively, the
load transfer mechanisms, towards the edges of the cavity, generated at the very
beginning of the opening process, persist so that the load acting on the geosynthetic
sheet is greater near the edges of the cavity. Despite these mechanisms, the overall
load transferred from the collapsed zone to the stable zones seems to be not much
influenced by the opening mode of the cavity. Moreover, these results also showed
that the soil expansion mechanism was not homogeneous in the volume of collapsed
soil and that it was strongly influenced by the opening mode of the cavity.

2.5. Conclusion

Discrete modeling has proved to be a relatively relevant numerical tool for

describing the interaction mechanisms between a soil and a rigid or deformable
reinforcement. The advantage of DEM compared to other numerical methods lies in
the fact that, with a rather small number of parameters, it is possible to describe the
complex behavior of granular materials, such as the contracting behavior of loose
materials, or the expansive behavior of dense materials. The discrete nature of the
numerical model makes it possible to take into account the large deformations,
the evolution of the behavior of the soil following changes of densities, the swells,
the collapses and other modes of rupture. Shear and expansion mechanisms at the
interface are also very well restored by the numerical model when surface texture of
the reinforcement, and shape of the granular particles, are well taken into account.
Unloading–loading cycles are handled naturally without the need to add complexity
to the contact laws. The applications treated have demonstrated the interest of the
discrete model to highlight the mechanical behavior of the reinforced structures and
to develop design methods. It can be emphasized that the use of the DEM requires
expertise and a consequent learning phase and that, for applications in operational
companies, sometimes leads to substantial computation durations.

2.6. References

[BRI 08] BRIANÇON L., VILLARD P., “Design of geosynthetic-reinforced platforms spanning
localized sinkholes”, Geotextile Geomembrane, vol. 26, no. 5, pp. 416–428, 2008.
[CHA 05] CHAREYRE B., VILLARD P., “Dynamic spar elements and DEM in two dimensions
for the modeling of soil-inclusion problems”, Journal of Engineering Mechanics, vol. 131,
no. 7, pp. 689–698, 2005.
[CHA 19] CHALAK C., BRIANÇON L., VILLARD P., “Coupled numerical and experimental
analyses of load transfer mechanisms in granular-reinforced platform overlying cavities”,
Geotextiles and Geomembranes, vol. 47, no. 5, pp. 587–597, 2019.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 97

[CHE 08] CHEVALIER B., Etudes expérimentale et numérique des transferts de charge dans les
matériaux granulaires – Application aux renforcements des sols par inclusions rigides,
Doctoral thesis, Université Joseph Fourier, Grenoble, France, 2008.
[CHE 10] CHEVALIER B., BRIANÇON L., VILLARD P. et al., “Prediction of load transfers in
granular layers used in rigid inclusions technique: Experimental and discrete element
method analysis”, Proceedings of the GeoFlorida 2010, West Palm Beach, USA,
pp. 1718–1726, 2010.
[CHE 11] CHEVALIER B., VILLARD P., COMBE G., “Investigation of load transfer mechanisms
in geotechnical earth structures with thin fill platforms reinforced by rigid inclusions”,
International Journal of Geomechanics, vol. 11, no. 3, pp. 239–250, 2011.
[CUN 71] CUNDALL P.A., “A computer model for simulating progressive large scale
movements of blocky rock systems”, Proceedings of the Symposium of the International
Society of Rock Mechanics, vol. 1, pp. 132–150, 1971.
[CUN 79] CUNDALL P.A., STRACK O.D.L., “A discrete element model for granular
assemblies”, Géotechnique, vol. 29, no. 1, pp. 47–65, 1979.
[DON 95] DONZÉ F.V., MAGNIER S.A., “Formulation of a three dimensional numerical model
of brittle behavior”, Geophysical Journal International, vol. 122, pp. 790–802, 1995.
[FUR 13] FURET A., La modélisation numérique du renforcement des sols par inclusions
rigides, Engineer internship, Ecole Nationale d’ingénieurs de Saint Etienne, France, 2013.
[GIR 97] GIRAUD H., Renforcements des zones d’effondrement localisé - Modélisations
physique et numérique, Doctoral thesis, University of Grenoble, France, 1997.
[GOR 13] GORNIAK J., Geosynthetic tubes filled with lightweight aggregate in new
geotechnical structures: First studies, Doctoral thesis, Université Pierre et Marie Curie,
Sorbonne Universités, Paris, France, 2013.
[GOR 13a] GORNIAK J., VILLARD P., DELMAS Ph., “Analytical design method for predicting
the vertical loading capacity of geosynthetic tubes filled with lightweight granular
material”, International Symposium on Design and Practice of Geosynthetic-Reinforced
Soil Structures, Bologna, Italy, pp. 406–415, 2013.
[GOR 15] GORNIAK J., VILLARD P., BARRAL C. et al., “Experimental and analytical studies of
geosynthetic tubes filled with expanded clay lightweight aggregate”, Geosynthetics
International, vol. 22, no. 3, pp. 235–248, 2015.
[GOR 16] GORNIAK J., VILLARD P., DELMAS Ph., “Coupled discrete and finite element
modeling of geosynthetic tubes filled with granular material”, Geosynthetics
International, vol. 23, no. 5, pp. 362–380, 2016.
[HUC 14] HUCKERT A., Approches numérique et expérimentale du dimensionnement de
renforcements géosynthétiques sur cavités et inclusion rigides, Doctoral thesis, University
of Grenoble, France, 2014.
[HUC 14a] HUCKERT A., BRIANÇON L., VILLARD P., “Experimental and numerical approaches
of the design of geosynthetic reinforcements overlying voids”, 23rd European Young
Geotechnical Engineers Conference, EYGEC, Barcelona, Spain, pp. 133–136, 2014.

Copyright Iste 2022 / File for personal use of Robert Charlier only
98 Deterministic Numerical Modeling of Soil–Structure Interaction

[HUC 16] HUCKERT A., BRIANÇON L., VILLARD P. et al., “Load transfer mechanisms in
geotextile-reinforced embankments overlying voids: Experimental and analytical
approaches”, Geotextiles and Geomembranes, vol. 44, no. 3, pp. 442–456, 2016.
[IWA 98] IWASHITA K., ODA M., “Rolling resistance at contacts in simulation of shear band
development by DEM”, Journal of Engineering Mechanics, vol. 124, no. 3, pp. 285–292,
1998.
[IWA 00] IWASHITA K., ODA M.,“Micro-deformation mechanism of shear banding process
based on modified distinct element method”, Powder Technology, vol. 109, pp. 192–205,
2000.
[JEA 92] JEAN M., MOREAU J.J., “Unilaterality and granular friction in the dynamics of rigid
body collections”, Proceedings of the Contact Mechanics International Symposium,
pp. 31–48, 1992.
[JEA 99] JEAN M., “The non-smooth contact dynamics method”, Computer Methods in
Applied Mechanics Engineering, vol. 177, pp. 235–257, 1999.
[LEH 06] LE HELLO B., VILLARD P., NANCEY A. et al., “Coupling finite elements and
discrete elements methods, application to reinforced embankment by piles and
geosynthetics”, Proceedings of Numerical Methods in Geotechnical Engineering, Graz,
Austria, pp. 843–848, 2006.
[LUD 05] LUDING S., “Shear flow modelling of cohesive and frictional fine powder”, Powder
Technology, vol. 158, pp. 45–50, 2005.
[SAL 09] SALOT C., GOTTELAND Ph., VILLARD P., “Influence of relative density on granular
materials behavior: DEM simulations of triaxial tests”, Granular Matter, vol. 11, no. 4,
pp. 221–236, 2009.
[SIB 19] SIBILLE L., VILLARD P., DARVE F., ABOUL HOSN R., “Quantitative prediction of
discrete element models on complex loading paths”, International Journal for Numerical
and Analytical Methods in Geomechanics, vol. 43, no. 5, pp. 858–887, 2019.
[SZA 11] SZARF K., COMBE G., VILLARD P., “Polygons vs. clumps of discs: A numerical
study of the influence of grain shape on the mechanical behaviour of granular materials”,
Powder Technology, vol. 208, no. 2, pp. 279–288, 2011.
[TRA 15] TRAN Q.A., Numerical modelling of soil arching between piles, Master Mécanique,
Energétique et Ingénieries, Université Joseph Fourier – Grenoble INP, France, 2015.
[TRA 19] TRAN Q.A., VILLARD P., DIAS D., “Discrete and continuum numerical modeling of
soil arching between piles”, International Journal of Geomechanics, vol. 19, no. 2, 2019.
[VIL 98] VILLARD P., GIRAUD H., “Three-dimensional modelling of the behaviour of
geotextile sheets as membrane”, Textile Resarch Journal, vol. 68, no. 11, pp. 797–806,
1998.
[VIL 08] VILLARD P., BRIANÇON L., “Design of geosynthetic reinforcements for platforms
subjected to localized sinkholes”, Canadian Geotechnical Journal, vol. 45, no. 2,
pp. 196–209, 2008.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 DEM Approach of the Modeling of Geotechnical Structures 99

[VIL 09] VILLARD P., CHEVALIER B., LE HELLO B. et al., “Coupling between finite and
discrete element methods for the modelling of earth structures reinforced by
geosynthetics”, Computers and Geotechnics, vol. 36, no. 5, pp. 709–717, 2009.
[VIL 16] VILLARD P., HUCKERT A., BRIANÇON L., “Load transfer mechanisms in geotextile-
reinforced embankments overlying voids: Numerical approach and design”, Geotextiles
and Geomembranes, vol. 44, no. 3, pp. 381–395, 2016.
[WAT 04] WATN A., ØISETH E., AABØE R., “Reinforced slope of road embankment with light
weight aggregates”, Proceedings of the 3rd European Conference on Geosynthetics,
Munich, Germany, vol. A-5, pp. 175–179, 2004.
[WEB 66] WEBER J., “Recherches concernant les contraintes inter-granulaires dans les
milieux pulvérulents”, Bulletin de Liaison des Ponts et Chaussées, no. 20, 1966.

Copyright Iste 2022 / File for personal use of Robert Charlier only
Copyright Iste 2022 / File for personal use of Robert Charlier only
 3

SSI Analysis in Geotechnical

Engineering Problems Using a
Finite Difference Method

Soil–structure interaction (SSI) problems in geotechnical engineering are often

complex problems, with common interactions at different levels and between several
parts of the studied problem, both under static or dynamic loading conditions. These
complex conditions justify the use of adapted numerical tools to take account of the
global behavior of the geotechnical structure. The presence of different materials in
terms of nature and mechanical behavior calls for the need for a high modularity of
the computation code used. Moreover, the geo- and structural materials usually
exhibit a highly nonlinear behavior and irreversible deformations. In the family of
the numerical modeling in a continuum, the finite difference method is quite well
adapted to address these issues.

3.1. Introduction

The SSI problems addressed in this chapter are studied using the finite difference
method, implemented in numerical models in a continuum. These methods thus
belong to the family of direct methods to take account of the SSI. In this kind of
method, the soil mass and the structural elements are explicitly simulated and are
part of the same model. Their interactions are thus taken directly into account
(contrarily to substructure or macroelement approaches). Even though the
simulations can be time-consuming, this tool appears to be reliable and versatile, as
well as ideal for parametric studies and for the calibration of simplified methods.

Chapter written by Daniel DIAS and Orianne JENCK.

Copyright Iste 2022 / File for personal use of Robert Charlier only
102 Deterministic Numerical Modeling of Soil–Structure Interaction

The implementation of finite difference methods allows the resolution of

stress–strain problems in a continuum medium, discretized into a finite difference
mesh (numerical grid). The stress and strain tensors are determined on each point/zone
of the medium. This allows the visualization of the involved phenomena (and thus
their analysis), simultaneously in the ground mass and in the structural elements.

Several case studies will be addressed in this chapter, all of which are related to
geotechnical engineering issues exhibiting strong SSIs: numerical studies of
reinforced retaining walls, tunneling in soft ground and soft soil improvement using
rigid piles, which are detailed and analyzed. However, this list of examples of SSI
geotechnical problems is obviously not exhaustive. Emphasis is mainly on static
loading conditions and some dynamic studies are also presented. Other types of
problems are not considered in this work.

3.2. The finite difference method using an explicit scheme

Finite difference methods are numerical methods for solving differential

equations with initial and/or limit conditions. Each derivative of the set of equations
is replaced by an algebraic expression in terms of the variation of the field variables
(stress or strain for classical geotechnical problems). These variables are defined at
discrete points in the space (on the numerical grid point) and are undefined
elsewhere (i.e. within the model elements). The finite difference method described
here is the one implemented in the software FLAC (Fast Lagrangian Analysis of
Continua) from Itasca [ITA 09], i.e. an explicit method performing a Lagrangian
analysis, using a time-marching scheme and a mixed discretization procedure
[CUN 82, MAR 82, BIL 93].

The time-marching scheme makes it so that even when dealing with static
loading conditions, dynamic equations of motions are used in the formulation. In
reality, part of the deformation energy stored by the system is converted into
kinematic energy which will propagate and dissipate in the medium. In this explicit
resolution scheme, the unbalance force induced in a zone will propagate in the
whole model. Figure 3.1 explains the calculation sequence during one time step of
duration Δt: from the forces/stresses, the Newton equation of movements is
implemented to compute the new velocities and thus the displacements and strain
rates. From the strain rates, the implementation of the constitutive equations gives
the new stress state. On each box, each variable value is updated from the previous
time step value, which remains fixed during the period Δt (i.e. explicit resolution
scheme). To justify this, the time step duration should be sufficiently small to ensure
that the velocity of the “calculation wave” is always higher than the physical wave
propagation: two adjacent elements do not influence one another during the time lap Δt,
as the information has no time to be transmitted. After a sufficient number of cycles,

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 103

the information (disturbances) will propagate across the model elements, as they
would propagate physically. The time step is sufficiently small if smaller than the
critical time step. Instead of determining a global critical time step, which would
necessitate the determination of the whole system eigenperiod (which is
unpractical), the time step calculation is based on the local critical time step (with a
factor of safety, as it is only an estimation), depending on the element size and
element mass and stiffness (conditioning the p-wave speed). As such, stiff or small
model zones may control the time step value.

This approach ensures the stability of the numerical model even if the system is
physically unstable (slope sliding, for example) but necessitates that the user knows
exactly what he/she is currently modeling (no black box giving the result!).

Figure 3.1. Resolution scheme during one calculation time step

In this type of modeling, there is no global rigidity matrix to be stored, as the

equations are formulated at each time step. No iteration process is needed to
compute the stresses from the strains, regardless of the constitutive model of the
material. The explicit scheme is particularly well adapted for nonlinear problems,
such as those encountered when dealing with SSI problems. However, the
implementation of this method is counterproductive when dealing with linear
problems in small deformations: in any way, the resolution will necessitate a large
number of calculation cycles to reach the solution, whereas the result would be
immediate if a finite element method is used.

Lagrangian analysis means that the node coordinates can be updated at each time
step of the computation, the grid can thus distort with the material it represents and
large deformation problems can be treated (e.g. in problems involving extensive
reinforcements such as geosynthetic, in which the tension load depends on the axial
strain, computed from the node positions). This is trivial because no global matrix of
stiffness is used.

Copyright Iste 2022 / File for personal use of Robert Charlier only
104 Deterministic Numerical Modeling of Soil–Structure Interaction

In this looping process, for static analysis, the movements need to be damped in
order to achieve a stationary state (equilibrium or permanent flow) in a minimum
number of cycles. In the described method, a damping force is applied to the model
nodes (local non-viscous damping), whose direction is such that the energy is
always dissipated. For dynamic simulations, more realistic damping – such as
Rayleigh or hysteretic damping – should be used (see below).

In this looping process, the stationary state (equilibrium or permanent flow) is

supposed to be obtained when the maximum unbalanced force on all model nodes
is considered small enough, then the node velocities are very small (unless for the
permanent flow case, where velocity vectors of all nodes will have a prevailing
direction and will not tend to decrease during the cycling process). These aspects of
convergence definition require that the user is aware of what he/she is currently
doing/modeling.

In a dynamic analysis, the process of computation is the same, with the mass of
the points constituting the model mesh determined from the unit mass of the
surrounding zones to solve the equations of motion, relating the nodal force to the
node acceleration. The damping of the waves in the soil is due to its characteristics
in terms of viscosity, friction and the development of plasticity. The damping in the
dynamic numerical process should reproduce the real energy loss in the system.
However, this is not always possible for numerical limitations. Different sorts of
damping, such as Rayleigh, hysteretic or local damping, are usually implemented.
The use of local damping is simpler than the Rayleigh damping because it does not
require the specification of a frequency. In order to take account of a semi-infinite
medium and avoid wave reflexion in a numerical model, specific boundary
conditions are assigned such as quiet (or absorbing) boundary or free-field boundary
conditions. The distance to the area of interest of these latter boundaries depends on
the material damping (the higher the damping, the closer the boundaries).

Moreover, it should be noted that a built-in programming language (FISH

language) in the software FLAC/FLAC3D gives a high level of flexibility to the user,
allowing him/her to perform the desired simulation in terms of loading conditions
(load or displacement conditions, servo-control routines, etc.), to automate parameter
studies, to give an unusual distribution of properties, to generate a specific grid, etc.

3.3. Application of the finite difference method to soil–structure

interaction problems

It is essential to first have a conceptual idea of the problem and the estimated
behavior of the global system to build a numerical model. Moreover, the questions
that should arise are: could the problem be considered two-dimensional

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 105

(possibly axisymmetric) or is it truly three-dimensional? What should be the size of

the model (geotechnical influence zone, with a special attention when dealing with a
dynamic problem) and thus what are the boundary conditions to be applied?
Where are the interesting parts of the system where the mesh should be
refined? What are the loading conditions (including cyclic or dynamic loading
conditions)? What are the hydraulic conditions (drained or undrained conditions,
hydro-mechanical coupled phenomena, i.e. the consolidation process to be taken
into account)? What is the optimal constitutive model complexity, according to the
level of accuracy targeted and the available material properties available, in
combination with the type of loading (monotonic, quasi-static cyclic or dynamic
loading)? What are the most appropriate structural elements and are interface
elements between the ground mass and these elements pertinent? The response to
most of these questions also determines the calculation time.

3.3.1. Structural elements

When dealing with an SSI geotechnical problem, structural elements are often
needed. However, to represent a structural element, volumetric elements (zones, so
as for the ground mass) can also be implemented with the material properties of
the structure, but it can be less convenient than real structural elements such as those
described below due to time calculation, which is usually higher. Moreover, the use
of structural elements allows direct access to forces and moment values. For 3D
analysis, typical structural elements are beam, cable, pile, shell, geogrid and liner
elements (Figure 3.2). These elements comprise nodes, with a maximum of six
degrees of freedom per node (three translational and three rotational components),
the mass associated with the structural element being lumped at the nodes.

In the geotechnical engineering field, is it usually sufficient to consider structural

elements with intrinsic linear elastic properties, sometimes with perfectly plastic
yielding or plastic hinge between the elements, and in interaction with the
surrounding soil with a frictional behavior (with usually a constant friction angle).
Nevertheless, with the objective of achieving more advanced SSI studies, more
realistic behavior of the material constituting the structure (concrete, synthetic
materials, etc.) could be implemented.

Beam elements are straight, two-node finite elements – with six degrees of
freedom per node (three translational and three rotational), and usually have a linear
elastic behavior without failure. A collection of beam elements is used to obtain
a beam structure, with the possibility of introducing a limiting plastic moment or a
plastic hinge between unit elements. These types of elements are used, for example,
to model framed structures or struts.

Copyright Iste 2022 / File for personal use of Robert Charlier only
106 Deterministic Numerical Modeling of Soil–Structure Interaction

Cable elements are finite, straight elements with two nodes – with one axially
oriented translational degree of freedom per node, which can yield in tension or
compression, but without any bending moment resistance. A pretension load can be
applied. Cables can be in frictional interaction with the grid representing the
surrounding soil, to take account of grouting, for example. These types of elements
are used to simulate structural elements with high tensile capacity and negligible
bending effects, such as bolts and tiebacks.

Pile elements resemble beam elements, but with additional frictional interaction
with the grid, representing the surrounding soil, to take account of skin-friction
effects. The end-bearing effect can also be modeled. These elements are mainly used
to simulate piles and rockbolt reinforcements.

Shell elements are flat elements with three nodes. Depending on the type of
shell, they can either resist bending loading, membrane loading or both. A real shell
structure is simulated by a collection of unit shell elements. If in contact with the
surrounding soil, the elements are rigidly bound to the grid node (no relative
displacement possible). These elements are implemented to simulate the structural
support provided by a thin shell, when relative tangential displacements or their
effects are negligible (structure slab, raft, tunnel liner, etc.).

Geogrid elements are similar to shell elements but with only membrane
resistance (no flexural rigidity) and in frictional interaction with the grid
representing the surrounding soil. They are used to simulate geotextiles and
geogrids, which are flexible membranes in shear interaction with the soil.

Liner elements are similar to shell elements with additional frictional interaction
with the surrounding ground mass (possibly on both sides of the liner). These are
implemented, for example, to represent shotcrete lining or retaining walls.

Figure 3.2. Structural elements and degrees of freedom for

three-dimensional modeling (source: from [ITA 09])

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 107

In two-dimensional analysis, the classical structural elements are beam, liner,

cable, pile, rockbolt and strip elements. Strip elements can yield in tension or
compression but cannot bear a bending moment. The interface with the soil is
defined by a nonlinear shear failure envelope. Strip elements are used in reinforced
earth retaining walls. Rockbolt elements are similar to strip elements, but they can
also transmit bending moments to the grid.

3.3.2. Interfaces

In SSI problems, interface elements are usually used between the various parts of
the system (in particular between the soil mass and the structural elements, even
though simulated with volume zones), in order to allow relative displacements and
sliding. In three-dimensional analysis, interfaces are a collection of triangular
elements (defined by three interface nodes) attached to a zone surface face, in
interaction with another grid surface when in contact. In two-dimensional analysis,
interfaces are segmental elements defined by two interface nodes.

In the more generalized case, interfaces have the properties of friction, cohesion,
dilation, normal and shear stiffness, tensile and shear bonding (Figure 3.3). During
each computation time step, the normal penetration and the relative shear velocity
are calculated for each interface node and its contacting zone surface face. The
interface constitutive model then gives the normal and shear forces. The interface
constitutive model is usually linear (constant shear stiffness ks and normal stiffness kn),
with a Coulomb shear strength criterion that limits the shear force according to the
normal force, cohesion and friction angle of the interface. The dilation (D) causes an
increase in the effective normal force when the shear strength is reached. For a
bonded interface, the behavior is elastic until the tensile or shear bond strength is
reached.

The effect of pore water pressure is taken into account using the effective
stresses for the slip conditions.

Figure 3.3. Components of a bounded interface (source: from [ITA 09])

Copyright Iste 2022 / File for personal use of Robert Charlier only
108 Deterministic Numerical Modeling of Soil–Structure Interaction

For a real interface between the soil mass and a structural element, the interface
properties have physical significance and should be derived from tests or data of
similar materials. Concerning the stiffness values, if the physical values are very
high (much more than 10 times the equivalent stiffness of adjacent zones), the
computation time will be very long, as the calculation scheme performs mass scaling
based on stiffness. If the normal stiffness is very small, the problem of grid
interpenetration could occur. If no information is accessible on the stiffness values,
or if the real values are judged as not important concerning the treated problem, the
calculation duration is optimized when kn and ks are set to 10 times the equivalent
stiffness of the stiffest neighboring zone.

3.3.3. Constitutive models for soil

The simplest constitutive model for soil is the linear elastic model. In the
geotechnical engineering field, this should only be used when dealing with complex
dynamic problems which cannot easily afford for more complex features such as
plasticity, or when considering rigid structures (concrete, steel). The use of the
elastic–perfectly plastic model with Mohr–Coulomb failure criteria is still very
frequently used in the practice of geotechnical engineering [MES 04]. The
parameters for this model have a clear meaning to all geotechnical engineers. In the
field of research studies and for more advanced numerical models, more complex
models are implemented, to simulate nonlinearity (sometimes even before yielding:
influence of small strain stiffness), shear yielding (strain hardening or softening),
compression yielding, strain accumulation, creep, etc. Some models out of the
elastoplasticity framework also exist, such as hypoplastic models. To choose
the most appropriate constitutive model according to the studied problem, the reader
should refer to the specialized literature on the constitutive modeling of soils in the
framework of the continuum mechanics [PUZ 12, HIC 08]. When using an existing
code, the choice of the constitutive model is also generally restricted to the available
built-in models (refer to the code user’s guide), unless implementing
the constitutive equations (which is feasible using the software FLAC). The choice
of the complexity constitutive model should also be made in accordance with
the available experimental soil data.

The influence of the complexity of the constitutive model on the global system
behavior has been studied for different SSI problems, in order to determine the
optimum model complexity.

[DO 13] compared the basic linear elastic–perfectly plastic model with
Mohr–Coulomb failure criteria with a strain hardening constitutive model, which

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 109

takes account of the hyperbolic behavior of the soil, the Cap-Yield soil (CYsoil)
model, on the tunnel lining behavior and displacement field around the tunnel. They
obtained a large effect of the constitutive model with higher structural forces and
ground settlement when implementing the CY model.

[JAN 14] studied the face stability of shallow tunneling using a three-
dimensional numerical model, implementing successively (1) the basic linear
elastic–perfectly plastic model with Mohr–Coulomb failure criterion (MC model)
and (2) an elastoplastic model with two isotropic hardening mechanisms (CJS2). In
particular, this model allows the simulation of dilatancy before failure for dense or
overconsolidated soils. Model parameter calibration was performed on triaxial and
oedometric compression tests (Figure 3.4). Comparison of the tunnel face extrusion
and surface settlement numerical results with experimental results clearly displayed
a good agreement when using the CJS2 model, highlighting the shortcomings of the
MC model for this application (Figure 3.5). The numerical modeling in the case of
tunnel face bolting reinforcement was thus performed with the CJS2 model
(see section 3.2.2).

Figure 3.4. Triaxial (top) and oedometric (bottom) test results (source:
from [JAN 14]), CJS2 model surfaces in the deviatoric plane

Copyright Iste 2022 / File for personal use of Robert Charlier only
110 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 3.5. Tunnel face extrusion (left) and surface settlement (right) due to the
decrease in the pressure applied on the tunnel face (source: from [JAN 14])

[JEN 09] highlighted that the numerical behavior of a simple piled embankment
system (see section 3.4.3) under monotonic loading was almost equivalent, either
implementing a constitutive model with a nonlinear elastic part and perfectly plastic
failure criterion for the load transfer platform (LTP) material or using a more
advanced elastoplastic model with isotropic shear hardening (CJS2 model). This is
attributed to the fact that the elastic part in the first model was nonlinear:
the stiffness depended on the current stress state.

[ABD 11] compared the Duncan & Chang hyperbolic model (with Mohr–
Coulomb failure criterion), the elastic–perfectly plastic model with Mohr–Coulomb
failure criterion and the CJS2 model for the soil modeling in the case of a
mechanically reinforced retaining wall (see section 3.4.1), considering several types
of reinforcing strip elements (see also section 3.4.1). They highlighted that the
existence of dilatancy before failure in the CJS2 model led to higher tensile loads on
the reinforcing strips at the top of the wall, which was more realistic knowing that
the part of dilatancy was important when the stresses were lower. Their study also
confirmed that a nonlinear constitutive model for the soil was required to
appropriately take account of the shear displacement and tensile loads in the
reinforcing strips, in the critical zones of the wall.

[DO 14a] studied the impact of considering soil plasticity in the case of a lined
tunnel in seismic conditions (see section 3.4.2). They compared dynamic
simulations performed with an elastic model to simulations completed with an
elastic–perfectly plastic model. They showed that the impact of the constitutive
model depends on the amplitude of the seismic signal: for a low seismic excitation,
implementing a plastic constitutive model appeared useless. Nevertheless, the effect
of a nonlinear elastic part and the effect of small strain stiffness were not
investigated in this study.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 111

3.3.4. Dimension of the problem

Simple geotechnical problems can usually be considered as two-dimensional,

either with the assumption of plane strain or axisymmetric conditions. In the case of a
complex SSI problem, this simplification is not always straightforward as the problem
could be fully three-dimensional, in particular due to the interaction of several
elements. However, despite recent advances in computing capacities, a full 3D
analysis is usually complex and requires important resources, both in terms of
storage and time. Some SSI studies focus on numerical processes to simplify
full 3D problems in 2D calculations and highlighted the relevance or limits of 2D
calculations.

Tunnel excavation is typically three-dimensional, especially around the tunnel

face. [DO 17] compared 2D deconfinement methods (to take account of the
displacement of the soil surrounding the tunnel before installation of the tunnel
structure) to a full 3D numerical analysis (Figure 3.6), taking account of most of the
components of a shield machine tunneling process. They used error functions to
compare the 2D versus 3D results in terms of surface settlements, horizontal
displacements on several points of the ground mass and internal forces in the lining
(bending moment and normal force). They showed that the 2D “convergence
confinement” method is the most appropriate to compute the soil surface settlements
or structural forces in the tunnel lining, whereas the 2D “volume loss” method is
more pertinent when focusing on the lateral movements of the soil.

Figure 3.6. 2D and 3D numerical modeling procedures (source: from [DO 17]).
For a color version of this figure, see www.iste.co.uk/grange/soil.zip

Copyright Iste 2022 / File for personal use of Robert Charlier only
112 Deterministic Numerical Modeling of Soil–Structure Interaction

3.3.5. Monotonic, quasi-static cyclic and dynamic loadings

The finite difference continuum modeling is adapted to perform static or

dynamic loading simulations. Static simulations can be monotonic or cyclic (under
quasi-static conditions). In the latter, an appropriate constitutive model should be
used, taking into account the specific properties of the simulated materials under
several unloading–reloading conditions.

The earthquake-resistant design of structures, as well as their deformations,

depend on the SSIs during seismic excitation. When performing a dynamic analysis
using a finite difference method, two important features should be carefully
considered: the damping characteristics of the materials (in particular the soil mass)
and the boundary conditions (see section 3.2), to avoid wave reflexion. Due to
computation time, these types of studies are often performed with an elastic
behavior of the soil, but the consideration of soil plasticity could lead to different
results [LOP 17], in particular for high-amplitude seismic signals [DO 14a].

3.4. Some application examples in the geotechnical engineering field

This section relates to some selected case studies of SSI problems in the
geotechnical engineering field using the finite difference method implemented in
the commercial software FLAC or FLAC3D from Itasca [ITA 09]. The aim of this
section is to give insight into the type of issues that can be treated, with a focus on
some research works. Some additional references are given for the reader who wants
to go beyond on a specific topic. Three different types of geotechnical problems are
addressed in this section: reinforced retaining walls, tunneling in soft ground and
rigid pile improvement. This is a non-exhaustive list of geotechnical structures
involving SSI. Moreover, these examples only deal with mechanical SSI and do not
treat coupled phenomenon issues such as hydro-mechanical (flow or consolidation
process), thermo-mechanical (energetic geostructures) or creep behavior problems.

3.4.1. Reinforced retaining walls

Reinforced retaining walls comprise a high variety of techniques, all of them

exhibiting strong mechanical interactions between the reinforcing elements
(geotextile, geogrid, metallic or synthetic strips, etc.), the soil mass and sometimes
the facing elements, leading to the effective working of the global structure. The
finite difference method has been used to study the behavior of such geotechnical
structures, both under static [DAM 13, YU 15a, YU 15b] or seismic loading
conditions [KRI 17, CHA 17].

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 113

Mechanically stabilized earth (MSE) walls highlight strong SSI between the
reinforcing strips and the soil mass, the presence of the strips improving the overall
mechanical properties of the soil, due to their tensile strength and adherence or bond
at the interface with the surrounding soil. [ABD 11] performed a numerical analysis,
in order to better understand the behavior of such structures and contribute to the
optimization of the design methods. Moreover, new types of MSE walls are
reinforced with synthetic strips, generally more flexible than the traditional metallic
strips, and existing design methods usually do not consider the deformation state of
the structure and appear excessively conservative concerning the ultimate limit state.
The stress–strain analysis, as allowed by the finite difference approach. It is
therefore a powerful tool to handle this issue. The main aim of the study was to
highlight the influence of the soil, the reinforcement and the soil/structure
parameters on both the ultimate limit state (failure of the structure) and
serviceability limit state (deformation of the structure). In order to perform an
intensive parametric study, a 2D model was derived from the 3D configuration of
the wall (Figure 3.7).

Figure 3.7. MSE wall and 2D equivalent model (source: from [ABD 11])

The behavior of the soil composing the reinforced backfill was successively
simulated with various elastoplastic constitutive models (see section 3.3.3), the
parameters being calibrated on triaxial compression test results; the reinforcement
elements were simulated with strip elements (see section 3.3.1). Three types of
existing synthetic or metallic strips were considered; the parameters of the interface
between the strips and the soil were determined from laboratory pull-out tests, which
were simulated in a numerical model dedicated to this calibration (Figure 3.8). To
take account of the influence of the confining pressure on the strip/soil interface
shear behavior, a model with a nonlinear shear failure envelope was used. In fact,
this made it possible to consider the increase in the interface friction coefficient due
to the constrained dilatancy effect, which increased the vertical stress. This feature
was particularly significant at low vertical stress, i.e. at the top of the wall.

Copyright Iste 2022 / File for personal use of Robert Charlier only
114 Deterministic Numerical Modeling of Soil–Structure Interaction

The facing concrete panels were modeled with beam elements, with interface
elements to consider the frictional interaction with the backfilling soil.
The construction stages of the wall were considered in this modeling.

Figure 3.8. 2D numerical model of a laboratory pull-out test (source: from [ABD 11])

The proposed numerical model first allowed a precise analysis of the failure
modes and deformation of the wall (Figure 3.9). Failure was obtained by applying
the c–φ reduction process to the backfill soil, which consisted of reducing the
cohesion and friction angle of the soil until failure was reached (the interface friction
parameter was also reduced accordingly). The factor of safety was then defined as
the ratio between the initial strength and the strength at failure. The reference case
was with synthetic strips (2.5 GPa elastic modulus, 100 kN/m tensile limit strength,
12% strain at tensile limit strength; initial soil/strip interface coefficient equal to 1.2
on top of the wall and 0.6 at the wall base, 0.22 MN/m²/m shear stiffness at the
soil/strip interface). The factor of safety was equal to 1.51 and the maximum
displacement was equal to 78 mm, which was a relatively high value as the strips
presented a low stiffness. The authors mentioned that this modeling did not consider
the horizontal displacement correction which could be made on a real wall
construction work, achieved by varying the facing slope.

Figure 3.9. Ultimate limit state of the reference MSE wall of 6 m height (source: from
[ABD 11]). For a color version of this figure, see www.iste.co.uk/grange/soil.zip

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 115

The parametric study performed on the soil, soil/strip interface and strip behavior
parameters (Figure 3.10) showed that the soil friction angle and cohesion and the
soil/strip interface friction coefficient had the most important influence on the wall
factor of safety, whereas the soil cohesion and strip elastic modulus are predominant
concerning the wall deformations. The soil/strip interface parameters influenced
both the deformation and the limit strength of the wall, showing that a good
estimation of these parameters based on experimental results seems to be essential
for each type of reinforcement; this can be achieved with laboratory pull-out tests.

Figure 3.10. Main behavior parameters influencing the ultimate limit state
(ULS) and serviceability limit state (SLS) (source: from [ABD 11])

3.4.2. Tunneling

3.4.2.1. Tunneling in soft ground and interaction with adjacent structures

In urban areas, the construction of shallow tunnels (usually excavated using

tunnel boring machines – TBMs) creates deformations and displacements of the
surrounding soil which may affect existing structures and thus can be a source of
unacceptable dysfunctions. A simplistic approach is to apply the greenfield
deformations caused by the TBM passage to the foundations of the structure
(i.e. soil movements obtained without considering the presence of the structure).
However, the presence of the structure contributes to the stiffening of the system and
consequently to the reduction of soil deformation and stresses in the structure. To
take account of the full SSI and the tunnel excavation process, a three-dimensional
model is necessary, representing the ground mass and the surface structure. The
relevance of using 3D finite difference continuum modeling to take account of the
tunneling process using TBM in soft soil has notably been highlighted by [DIA 13],
who compared numerical results to field measurements of the Lyon subway tunnel.

An example of a numerical study of the tunneling–surface structure interaction

was given by [JEN 04], where the framed surface structure was simulated by beams

Copyright Iste 2022 / File for personal use of Robert Charlier only
116 Deterministic Numerical Modeling of Soil–Structure Interaction

and shell elements. The tunnel excavation process under the structure was performed
step by step and reproduced the volume loss inside the tunnel, due to the annular
void between the TBM and the surrounding soil, due to its conical shape and grout
consolidation (Figure 3.11). The software’s built-in programming language made
it possible to successfully apply the desired conditions inside the tunnel
(free convergence of the tunnel walls until a limit imposed by the presence of the
TBM and setting of concrete rings). The tunnel face deconfinement was not taken
into consideration.

Figure 3.11. 3D numerical model (left) and sequential tunnel

excavation procedure (right) (source: from [JEN 04])

The ground mass was multi-layered, with soil properties taken from the Lisboa
subway project. The soil behavior was drained and simulated with an
elastic–perfectly plastic model. The surface building was simplified, and structure
elements were modeled with beam elements for the columns and shell elements for
the foundation raft and the slabs, so the structure had an elastic behavior
(with elastic properties of reinforced concrete), which could be considered as a
model limitation from a structural engineering point of view.

Several simulations varying the volume loss in the tunnel and the type of surface
structure highlighted the SSI between the ground mass and the surface structure. The
presence of the structure increased the surface settlement for a volume loss in
the tunnel higher than 1%, in comparison to the greenfield case, and, due to the
presence of a raft with a great axial stiffness and the fact that the raft is bound with
the soil, the horizontal displacements were negligible under the structure in
comparison to the greenfield case (Figure 3.12). Therefore, it would be conservative
for the structure behavior prediction to apply the greenfield conditions to its
foundations. With this full numerical model, the forces and bending moments could
also be studied within the structure elements.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 117

Distance to tunnel axis (x/D)

Distance to tunnel axis (x/D) 0 1 2 3 4 5 6 7 8 9
0 1 2 3 4 5
0
0

Horizontal displacement
Building
Surface settlement (cm)

-1 -0,5

(cm)
-2 Greenfield
-1
With structure
-3 Greenfield
-1,5 With structure
-4

-5 -2

Figure 3.12. Surface displacements (source: from [JEN 04])

Another big issue of the interaction of tunneling with adjacent structures is the
pile–tunnel interaction or piled structure–tunnel interaction. The interaction
mechanisms are not fully understood yet and greatly differ considering the pile
length and location, the pile group effect, the pile stiffness, etc. This problem has
been addressed using finite difference numerical modeling, sometimes as a tool to
calibrate simplified methods (see [LOG 01, LEE 12]).

3.4.2.2. Tunnel face reinforcement

In urban areas, it is necessary to control the ground movements caused by

shallow tunneling, to avoid damages on neighboring structures. One of the
techniques consists of preconfining the core ahead of the tunnel face with glass fiber
or steel bolts. The deformation prediction of the bolted ground mass is required in
order to optimize the reinforcement technique. The soil mass and the bolts are in
strong interaction, and this interaction highly determines the global behavior.

[JAN 14] proposed a three-dimensional numerical modeling of the tunnel

digging with bolted face, based on experimental tests performed on a centrifuge 1/50
scale down model. The modeling procedure aimed to reproduce the physical testing
procedure, which was straightforward using the finite difference method as
implemented in FLAC3D, as it was highly versatile. However, only half of the
whole physical model placed in a rigid container needed to be simulated, due to
symmetry conditions. The tunnel lining was made of steel and then assumed to be
non-deformable. In the numerical model, the sliding between the lining and the soil
was possible by using an interface element with a friction angle equal to two thirds
of the soil mass friction angle. A latex membrane under air pressure was placed on
the tunnel face, with the initial geostatic value equal to 200 kPa and the decrease in
air pressure simulated the tunnel excavation, until failure occurred. In the numerical
model, normal stresses were applied on the external side of the tunnel face soil
zones to reproduce these conditions. The numerical model was subjected to a 50 g
vertical acceleration, as in the centrifuge tests. The soil mass was Fontainebleau

Copyright Iste 2022 / File for personal use of Robert Charlier only
118 Deterministic Numerical Modeling of Soil–Structure Interaction

sand and its behavior was simulated alternatively with the linear elastic–perfectly
plastic model with the Mohr–Coulomb failure criterion or with the elastoplastic
CJS2 model, with the parameters being calibrated on triaxial and oedometric tests.
The bolts were sand-coated PVC pipes and simulated with pile structural elements,
with a frictional interface with the surrounding sand (friction angle equal to two
thirds of the soil friction angle); the interface behavior properties being one of the
key parameters controlling the bolting influence on displacement and tunnel face
stability, as demonstrated by previous studies.

Figure 3.13. Physical model cross-section (dimensions in mm), 3D numerical model

composed of 56,000 soil zones and tunnel face bolting (source: from [JAN 14])

The comparison between the experimental and numerical results highlighted a

good agreement in terms of displacement and tunnel face stability, both in the
unreinforced and bolt-reinforced cases, when using the CJS2 elastoplastic model for
the soil mass behavior, as shown in Figure 3.14. Moreover, the numerical model
allowed an in-depth analysis of the failure mechanisms, by representing the contour
of velocity magnitude and the soil zones that had plasticized (Figure 3.15), and
comparison with existing theoretical approaches could be made.

Figure 3.14. Comparison between experimental and numerical results in terms of

axial face displacement and surface settlements (source: from [JAN 15])

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 119

Figure 3.15. Numerical plastic zones (left) and iso-settlement lines

obtained in the centrifuge test (right) (source: from [JAN 15])

The developed model was then used to analyze the influence of additional
parameters, such as the bolting density, bolt length and bolt material. In fact, in the
experimental campaign, the number of performed tests was limited.

3.4.2.3. Tunnels under seismic loading

The SSI phenomenon occurring in tunneling is still an open question when

subjected to seismic loading. [DO 14a] studied the segmental tunnel lining behavior
under seismic loading using 2D finite difference modeling (Figure 3.16), mainly in
order to assess the tunnel response in terms of loads induced in the lining and to
discuss the design of segmental tunnel lining. Most of the numerical studies consider
seismic loads as pseudo-static load, but the literature review shows that it could lead
to smaller forces in the tunnel lining than with a truly dynamic analysis. This option
was thus adopted in this study, even though the calculation time was much longer.

Ground parameters (clayey sand) and tunnel features from the Bologna–Florence
high speed railway line tunnel were used. Two constitutive models for the soil were
successively considered: linear elastic and elastoplastic with Mohr–Coulomb failure
criterion. The tunnel structure behavior was assumed to be linear elastic and
simulated with liner elements, with no sliding at the ground mass interface.
Segmental tunnel linings were often implemented in seismic areas, as they could
undergo deformations with less or no damage compared to continuous linings, due
to their flexibility. The segment elements from the tunnel lining were connected
with joints, considered as elastic pins with rotational, axial and radial stiffness
(Figure 3.17), whose values were calibrated on a procedure developed by [DO 14b].

Copyright Iste 2022 / File for personal use of Robert Charlier only
120 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 3.16. 2D plane strain numerical model (source: from [DO 14a])

Figure 3.17. Connection between lining segment elements (source: from [DO 14a])

Before the dynamic loading application, the steady state of the model was
obtained under static conditions (leading to initial forces in the tunnel lining). A real
seismic signal with a maximum acceleration of 0.35g (or 0.0035g for comparison
purposes) and a duration of 21 seconds was then given as a velocity input to the
numerical model. Dynamic calculations were performed using Rayleigh damping,
and quiet model boundaries were used to avoid wave reflection. Forces induced in
the tunnel lining due to the seismic loading could be studied.

Even though only one single tunnel geometrical configuration and a single
homogeneous ground mass was considered, a comparison with a continuous lining
could be performed and bending moments reached during the earthquake were
smaller in the case of segmental lining; the importance of taking into account soil
plasticity under high seismic loading was underlined, as the irreversible behavior of
the soil modified the tunnel loads during and after the earthquake (Figure 3.18),
exhibiting the strong SSI of the studied problem; the assumption that an equivalent
static solution led to smaller tunnel loads was confirmed (Figure 3.19).

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 121

Figure 3.18. Impact of the soil constitutive model on the bending

moments in the tunnel lining (source: from [DO 14a])

Figure 3.19. Bending moments in the tunnel lining obtained with a dynamic
calculation and a pseudo-static calculation (source: from [DO 14a])

3.4.3. Soft soil improvement using vertical rigid piles

Improvement using vertical rigid piles (Figure 3.20) is one of the techniques to
build structures or infrastructures on soft soil stratums. This technique is increasingly
used worldwide as it is of rapid implementation and leads to large (total and
differential) settlement reduction. A three-dimensional grid of rigid piles (usually
unreinforced concrete) is driven through the soft soil layer down to a more competent
stratum. This technique differs from those of classical piles as the piles are not
structurally connected to the surface structure: a load transfer platform (LTP – usually
made of granular material) is placed between the improved layer and the surface

Copyright Iste 2022 / File for personal use of Robert Charlier only
122 Deterministic Numerical Modeling of Soil–Structure Interaction

structure. An additional basal reinforcement (reinforcement geotextile or geogrid)

can be placed at the platform base. Shearing mechanisms occurring both in the LTP
and in the reinforced soil layer lead to the transfer of part of the surface load to the
piles and to surface settlement reduction and homogenization. The different parts
constituting the system are in complex mechanical interaction, making the problem
difficult to handle with simple analytical methods. The numerical modeling
approach using the finite difference method is one of the adapted tools to study the
behavior and performance of the system, under static, cyclic and dynamic loading
conditions, taking into account the behavior of each part of the system and their
interactions in a unique model.

Figure 3.20. Schematic vertical cross-section

[JEN 07a] performed a two-dimensional preliminary numerical study, based on

experimental results obtained on a laboratory small-scale model (Figure 3.21). The
physical model was reduced to two dimensions and used analogical materials.
Indeed, even though it is a simplification of the real system, it was developed to
emphasize the mechanisms developing in the LTP, in interaction with the lower part
of the system (presence of the soft soil and the piles), and to highlight the important
controlling parameters by performing an intensive parametric study. Results were
obtained both in terms of load distribution to the piles and the “soft soil” and
displacement field in the LTP, implementing a digital image correlation (DIC)
method, which is very relevant for numerical validation purposes. The main
objectives of the numerical work were to assess the possibility of the modeling to
reproduce the physical behavior under monotonic loading and to go further into the
analysis of the load transfer and settlement reduction coupled mechanisms. In the
numerical model, the piles were considered to be infinitely rigid (by fixing the nodes
at the pile edges): in the small-scale model, their rigidity was actually much larger
than the material simulating the soft material. However, in a non-idealized case,

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 123

it is known that SSI phenomena are also governed by the relative stiffness of the
various parts of the system. Interface elements were integrated to allow for relative
displacements between the pile–soft material and the platform.

Figure 3.21. 2D schematic view of the laboratory small-scale

model (left) and numerical model (one quarter of the
physical model setup) (source: from [JEN 07a])

The LTP behavior was successively simulated with an elastic–perfectly plastic

Mohr–Coulomb type model using nonlinear elasticity and a two-mechanism
elastoplastic model with isotropic hardening, namely the CJS2 model. The
parameters were calibrated on biaxial compression tests. The foam elements
representing the soft soil had a nonlinear elastic behavior, capturing exactly the
results obtained with simple compression tests. The purpose was to assess the
impact of the platform constitutive model complexity on the system behavior. It
appeared that there was no big difference between results obtained with both
models, mainly because nonlinear elasticity was considered for the elastic–perfectly
plastic model. The comparison with the experimental results showed a good
agreement in terms of displacements in the whole model; however, the numerical
model overestimated the proportion of the load which was transmitted to the pile
through the LTP by developing the arching effect. This divergence has been
overcome by implementing a numerical model using the discrete element method
(DEM) [JEN 09a]. In fact, in this approach of modeling, the soil particles are
simulated by discrete elements, so the soil mass can largely deform at the pile head
corners, due to soil element rearrangement, whereas this aspect is hidden in a
continuum approach. An intensive numerical parametric study was undertaken to
highlight the role of the geotechnical parameters, both in the continuum and discrete
modeling. Both approaches showed that the shear strength of the LTP material
(macroscopic friction angle and apparent cohesion for the continuum model, friction
coefficient and bond between particles in the DEM) was the preponderant feature
governing the load distribution and settlement reduction.

Copyright Iste 2022 / File for personal use of Robert Charlier only
124 Deterministic Numerical Modeling of Soil–Structure Interaction

A major interest in performing numerical modeling based on experimental

results is its possibility to obtain additional information, not accessible in the
experiment, as the stress distribution in the LTP. This can help the development of
simplified methods by analyzing more precisely the morphology of soil arching
within the LTP, identified as the main phenomenon of load redistribution and
settlement homogenization on the LTP surface. Figure 3.22 shows that the mean
principal stresses are oriented towards the pile head and the earth pressure
coefficient, i.e. the ratio between horizontal and vertical stresses is equal to 0.4 on
top of the pile and 2.3 at the crown of the vault (corresponding to the reaching of the
limit equilibrium states with a friction angle equal to 24°).

a – Numerical model output b – Schematic representation

Figure 3.22. Principal stress orientation around the pile

in the continuum model (source: from [JEN 09a])

Further work was undertaken by performing full three-dimensional analysis on a

realistic case, using the same numerical modeling strategy [JEN 09b] and the
modified Cam-clay model for the soft soil. Although not compared and validated on
experimental data, this work proposes a numerical procedure to model piled
embankments. In this modeling, an infinitely rigid substratum and rigidly embedded
piles are considered, by fixing the node of the lower model boundary. A rigid pile
grid usually contains tens or hundreds of regularly placed piles; then, due to
symmetry conditions, only one quarter of a pile and its zone of influence has been
simulated (Figure 3.23): the grid points of the vertical boundaries were fixed in the
out-of-plane horizontal direction. One of the objectives was to perform parametric
studies on various parameters. In particular, the impact of the soft subsoil

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 125

compressibility has been highlighted, in interaction with the embankment properties

(Figure 3.24), whereas the subsoil compressibility is currently not taken into account
in simplified design methods. As for the 2D analysis, the dominant impact of the
embankment material shear strength on the load transfer has been highlighted
[JEN 07b]. Another aim was to assess the behavior of a current embankment section
including lateral slopes (Figure 3.25), where the piles are then subjected to
horizontal loading and bending moments, which are important aspects to be taken
into account in the pile design.

Figure 3.23. 3D numerical model of the pile grid

unit element (source: from [JEN 09a])

Figure 3.24. Parametric study results for two soft deposit compressibilities (S1 and
S2, with S1 being more compressible than S2) and two embankment materials
(M1 and M2, with M1 being stiffer and having a higher shear strength than M2)
(source: from [JEN 09a]). The depicted surface settlement is the maximum value due
to the application of the next 0.5 m-thick layer

Copyright Iste 2022 / File for personal use of Robert Charlier only
126 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 3.25. 3D numerical model of an embankment current section and

horizontal displacements at the top of the piles (source: from [JEN 09a])

The question of the relevance of simulating the load transfer onto the piles by the
3D arching effect in the embankment material using a continuum approach as the
finite difference method has been addressed by [TRA 19], by comparing numerical
modeling results in a continuum (FDM) and implementing the DEM (Figure 3.26).
In the continuum approach, the embankment material behavior was simulated with
the cap yield model to take account of friction hardening and softening. The
comparison of both approaches is addressed in detail in Chapter 5 of this book.
Some discrepancy between discrete and continuum models was observed when there
was a relative sliding between the granular layer and the pile cap in the DEM. In
fact, the compatibility conditions in the continuum model prevented this sliding and
induced a stress concentration at the pile corner. Moreover, the continuum approach
was not able to take account of porosity change that occurred in the DEM for a loose
embankment material. This was due to the simplicity of the constitutive model used.

Figure 3.26. System efficacy according to the embankment material shear

ratio for a dense (left) and a loose (right) material, for several
embankment heights (source: from [TRA 19])

Even though a realistic case of a piled embankment is considered, one major

limitation of this modeling is the lack of validation on experimental results. Thus,
[NUN 13] performed a numerical study of full-size experiments at the Chelles

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 127

(France) test site (Figure 3.27). The test site was a 5 m-high embankment built on
8 m-thick compressive alluvial soil. The site was divided into four zones: three with
pile reinforcement (two also with horizontal geosynthetic reinforcements –
geotextile or two layers of geogrid) and one unimproved, to obtain the reference
behavior.

Figure 3.27. Chelles experimental site (source: from [NUN 13]). For a color
version of this figure, see www.iste.co.uk/grange/soil.zip

The performance of the embankment support system was assessed by monitoring

data (total stresses, horizontal and vertical displacements). This fully instrumented
study was ideal for validating the finite difference numerical modeling approach,
also as the site geotechnical characterization allowed the model parameter
calibration. The lateral friction limit of the piles was determined using an
instrumented axially loaded test performed on the site on a floating pile, which
allowed the input of the interface element parameters between the piles and the soft
ground layer in the numerical model. The site instrumentation indicated a rapid
consolidation of the alluvial soft soil, and most of the deformations were recorded
during the embankment construction, justifying a numerical modeling performed in
drained conditions.

Two numerical models were considered: an elementary cell, considering only

one pile, and a global model, simulating the whole experimental site and interaction
between adjacent zones (Figure 3.28). To fully take the global system behavior into
account, several parts of the system were explicitly simulated: the several layers of
soft soil, the piles, the interface between the piles and soil, the platform and/or the
embankment, the geosynthetic layer(s) when present and also the substratum (sandy
gravel layer) underneath the reinforced soft layer into which the piles are embedded.
The global model comprised 445,000 volume elements (using a coarser grid than for
the elementary cell). The platform and/or the embankment behavior were simulated
with an elastic–perfectly plastic model with Mohr–Coulomb failure criterion, as well
as the substratum, the soft layer behavior was represented by the modified
Cam-clay model (see the model parameters in Table 3.1) and the piles had a linear

Copyright Iste 2022 / File for personal use of Robert Charlier only
128 Deterministic Numerical Modeling of Soil–Structure Interaction

elastic behavior. The calculations were performed in large strain mode, in order to
activate the membrane effect in the geosynthetic layers, simulated with geogrid
elements.

Figure 3.28. 3D numerical models of the Chelles experimental site: elementary

cell, model for pile loading test and global model (source: from [NUN 13]).
For a color version of this figure, see www.iste.co.uk/grange/soil.zip

Table 3.1. Model parameters (source: from [NUN 13])

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 129

The first numerical validation step was the comparison of the results obtained on
the non-reinforced zone and on the embedded pile loading test, to validate the
modeling hypothesis of the soft soil layers and an embedded pile behavior
(including shaft and toe resistance). Comparisons were then performed for the
reinforced zones, in terms of stress efficacy (proportion of the total load transmitted
to the piles; Figure 3.29) and settlements (Figure 3.30). The proposed numerical
model (in particular, the global model, taking into account the influence of adjacent
zones) made it possible to obtain results close to the site instrumentation results,
showing the ability of the numerical modeling to reproduce the phenomena leading
to the system performance. A comparison with simplified analytical methods
showed that the system performance was underestimated, as they did not take into
account all the features of such a complex system (in particular the soil
compressibility and the pile–soil skin friction). However, some discrepancies
between experimental and numerical results were underlined for some cases. In
particular, the role of the geosynthetic reinforcement could not be accurately taken
into account.

Figure 3.29. Comparison of the numerical (CE = elementary cell;

MG = global model) and experimental results in terms of
stress efficacy (source: from [NUN 13])

Another instrumented case study of soil improvement with rigid piles and
comparison to a finite difference numerical modeling was reported by [BRI 15]. The
project consisted of constructing an industrial building on a thick compressible clay
layer: the hard sandstone bedrock is located 50 m down the soil surface. In this case,
soil improvement consisted of 1 m-in-diameter-piles installed along a 6 m square
grid and a 2 m-thick granular platform located between the piles and the surface
1.5 m-thick concrete slab to ensure the load transfer. As the aptitude of the thin
alluvial layer located between the clay layer roof and the head of the piles was not

Copyright Iste 2022 / File for personal use of Robert Charlier only
130 Deterministic Numerical Modeling of Soil–Structure Interaction

verified to ensure a good compaction of the LTP material, 0.5 m of the alluvial layer
was removed and replaced by a cement gravel layer. A geotextile was placed around
the pile at 0.5 m to limit the hanging effect of the treated soil layer on the piles. The
work site was largely monitored: a total of 100 sensors were installed in the LTP,
soft soil and concrete piles.

Figure 3.30. Comparison of the numerical and experimental results in

terms of settlements (source: from [NUN 13]). For a color version
of this figure, see www.iste.co.uk/grange/soil.zip

The finite difference numerical model comprised all the elements described above
(Figure 3.31), from the hard bedrock up to the surface concrete slab, on which an
additional loading was applied up to 150 kPa (vertical stress applied by the building and
exploitation loadings). On the site, the grid contained 292 rigid piles, covering an area of
3,624 m². Only one unit cell was represented in the numerical model, considering a
current grid rigid pile. The soil and the pile were modeled by volume elements, and
interface elements were considered along the pile, with a Mohr–Coulomb failure
criterion. For the pile–soil interfaces, an adherence (cohesion) value and a null friction
angle were used, to obtain a constant friction limit. The effect of pile installation was not
taken into account: the vertical piles and the soft soil were set up in only one phase,
which constituted the initial state, and static equilibrium under self-weight was reached.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 131

The numerical procedure then reproduced the successive work stages, with calculation
until static equilibrium at each stage. The calculation was performed in drained
conditions and corresponded to the observed conditions of quasi-instantaneous pore
pressure dissipation. The clay deposit was composed of five successive clay layers, and
their behavior was simulated by the modified Cam-clay model. Due to the lack of data
on its constitutive behavior, an elastic–perfectly plastic model with a Mohr–Coulomb
type failure criterion was used to simulate the granular material, the treated soil, the
alluvia and the bedrock behavior. The slab and the piles were made of concrete and were
assumed to behave elastically. The numerical model consisted of approximately 5,436
zones and 6,508 grid points. The time requested for the FLAC3D analysis of the model
was about 15 hours when using a 2.40 GHz core i7 CPU computer.

Figure 3.31. Schematic cross-section of the upper part of the improved system (left,
top), top view of the pile grid (left, bottom) and numerical model (right) (source: from
[BRI 15]). For a color version of this figure, see www.iste.co.uk/grange/soil.zip

Numerical results were compared to experimental measurements in terms of

stress distribution at two levels in the platform: at the base and at mid-height
(to assess the load transfer within the platform), differential settlement in the

Copyright Iste 2022 / File for personal use of Robert Charlier only
132 Deterministic Numerical Modeling of Soil–Structure Interaction

platform and pile strain, at various stages of the construction process. For all
construction stages, the numerical results concurred with the experimental
measurements (Figure 3.32). The numerical model appeared to be able to effectively
represent the mechanisms developed inside the LTP, in interaction with the pile–soft
soil system, so the behavior of the system has been predicted for the additional
150 kPa storage load. The numerical models then provided a stress value equal to
1 MPa on the pile head, an average stress equal to 110 kPa on the soft soil surface
and a differential settlement equal to 0.3 mm just under the concrete slab. These
results are very useful for validating or optimizing the pile and concrete slab design.

Figure 3.32. Stress at the platform base (left) and differential settlement at
various heights in the platform (source: from [BRI 15]). For a color version
of this figure, see www.iste.co.uk/grange/soil.zip

The good agreement between in situ measurements and finite difference

numerical modeling, as shown by [NUN 13] and [BRI 15], justifies the use of this
latter to predict the behavior of future pile-improved works, but it is influenced by
several features:
– good quality and complete geotechnical site investigation, in order to give
relevant input values for the stratigraphy and for the constitutive model parameters,
and to assess the hydraulic conditions (drained or undrained conditions);
– pile load test realization on the same site, in order to assess the pile–soil
interface behavior;
– implementation of a numerical model taking explicitly the various parts of the
system into account, including interfaces at the pile shaft and the presence of the
substratum and, if necessary, global modeling of the site;
– use of appropriate constitutive models: an elastic–perfectly plastic model with
a Mohr–Coulomb failure criterion for the platform and the modified Cam-clay

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 133

model for the compressible layer seem appropriate for monotonic loading.
The structural parts (piles, surface slab) can behave simply elastic (influence of their
stiffness);
– implementation of a numerical modeling procedure reproducing the successive
working stages. However, it seems that not considering the pile installation process
is not a limitation.

The finite difference model presented previously could quite truly represent a
pile improvement system under monotonic loading. However, in some applications,
quasi-static cyclic loading is applied on the system, such as for storage tanks and
heavy load storage areas. Under such loading conditions, the durability of the
mechanisms leading to the performance of the system is still an open question and
no guideline exists yet. To address this issue, [HOU 17] performed experiments on a
laboratory small-scale model and the numerical modeling back analysis of
the behavior under cyclic loading was then completed using the finite difference
modeling approach. The laboratory model consisted of a rigid box of 1 m2 square
section, containing 20 rigid piles (Figure 3.33). The scale reduction on the length
compared to a real site was 1/10. Around the piles, the box was filled with a
compressible material with Cc/(1+e0) equal to 0.34, on a thickness equal to 0.4 m, to
represent the soft soil layer. A granular platform (LTP) composed of granular
material is placed on top of the pile-compressible layer system, on a thickness equal
to 50 or 100 mm. A uniform vertical load (surface loading) was applied on the
surface, either directly using a soft membrane under pressure covering the entire
model surface or by first placing metallic plates at the platform surface, to examine
the case of a rigid slab on the surface. Low frequency vertical cyclic loading was
studied by applying pressure-controlled cycles. The system was instrumented with
load and displacement sensors, and four semi-piles were placed along a model
border, two of them behind a Plexiglas window, allowing for photographs to be
taken during the experiments and then obtaining the displacement field using a
digital image correlation (DIC) method. This allowed for the analysis of the
mechanisms developing in the granular platform and their evolution during the
cyclic loading, and for the comparison with the numerical modeling approach, in
addition to the results given by the instrumentation, located in the central part of
the setup. This experimental study had some shortcomings compared to a real
pile-improved structure, mainly due to the fact that the similitude between small
scale and real scale was not entirely satisfied (the model has been scaled on the
length, but not on the stress level, as it is under normal gravity), and because
analogue material was used. So, the results cannot be directly extrapolated to a real
system. Nevertheless, it constituted a powerful tool for validating numerical
modeling approaches and their ability to represent the behavior observed physically,
on a laboratory model where the experimental conditions were well controlled.

Copyright Iste 2022 / File for personal use of Robert Charlier only
134 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 3.33. Photograph of the laboratory small-scale model and schematic

horizontal cross-section at the pile head level. For a color version of this
figure, see www.iste.co.uk/grange/soil.zip

The finite difference numerical model (Figure 3.34) resembles the model
previously presented in [JEN 09b], with the dimensions of the 3D small-scale
model. As the surface loading was uniform, only one quarter on a unit pile grid cell
was represented. If explicit paths of loading–unloading are considered, it is usually
recommended using an elastoplastic model with kinematic hardening (which is able
to take account of plastic deformation under unloading of the cyclic loading history).
For repeated cyclic loading (traffic load type), an equivalent viscoplastic constitutive
model can be considered, but the precise behavior under the loading–unloading
cycles cannot then be studied. In this first numerical study, an elastoplastic with
isotropic hardening for the shear mechanism with a nonlinear elastic part (cap yield
model, available as a built-in model in FLAC3D) was implemented, in order to
assess the pertinence of this modeling to reproduce the observed behavior, before
considering a more complex constitutive model. The parameter was determined
based on cyclic triaxial tests performed on the LTP granular material. The soft
material behavior was simulated with the modified Cam-clay model and parameters
were calibrated based on an oedometric test result. The pile was represented by
volume elements with elastic properties of the pile material. Interface elements were
placed between the various parts of the system. In order to reduce the computation
time, which can be high due to the high number (50) of loading–unloading cycles, a
numerical procedure implicitly taking the presence of the rigid slab was developed.
In fact, the presence of rigid structural elements, in contrast to the soil stiffness,
considerably increases the computation time. The presence of the rigid slab on
the surface was taken into account by imposing a uniform vertical displacement
on the grid points on top of the LTP and a routine controlled the corresponding
average vertical stress on the whole surface, to reproduce the experimental loading
procedure.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 135

Figure 3.34. Numerical model (left) and vertical displacement field in the model
for a monotonic loading (30 kPa vertical stress applied on the surface), without
rigid slab. For a color version of this figure, see www.iste.co.uk/grange/soil.zip

The numerical and experimental results were first compared under monotonic
loading and a good agreement was obtained both in terms of load transmitted to the
piles and displacement field in the LTP (comparison could be made using the results
obtained by the DIC method applied in the laboratory tests), for both surface
boundary conditions (uniform vertical stress or uniform settlement). The results
obtained under cyclic loading are shown in Figure 3.35 (the efficiency is the
proportion of the total vertical load which is transmitted to the pile heads).
The numerical back analysis of the tests performed in the laboratory showed that the
numerical model was able to effectively reproduce some observed features, such as
the basal settlement accumulation measured during the cyclic loading and the load
efficiency of the system for the case without rigid slab. These encouraging results
could be obtained even with an elastoplastic with isotropic hardening mechanism
model, due to the nonlinearity of the model combined with the stress redistribution
in the LTP during the cycles. Nevertheless, the proposed model was not able to
accurately simulate the load reparation in the LTP in the case of the rigid slab on
the surface: the numerical modeling underestimated the efficiency; this could be
attributed to the non-explicit numerical modeling of the rigid slab and/or to the
punching of the piles through the LTP that might have occurred in the experimental
tests. Moreover, the settlements recorded under the post-cyclic loading were

Copyright Iste 2022 / File for personal use of Robert Charlier only
136 Deterministic Numerical Modeling of Soil–Structure Interaction

overestimated by the numerical model, as the constitutive model used was not able
to take account of the soil rigidifying process under cyclic loading.

Figure 3.35. Comparison of experimental and numerical results in terms of load

efficiency and settlement at the LTP base, in the center of the model, for the cases
without rigid slab (WR) and with rigid slab (R) (reference state in after model
construction, just before applying the external load) (source: from [HOU 17]). For a
color version of this figure, see www.iste.co.uk/grange/soil.zip

In addition, the proposed numerical modeling can give access to additional

information, useful in the practice, and difficult to obtain in the experimental part,

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 137

as the surface total and differential settlements and the surface stress distribution
under the slab, required a design of the structures that we will place on the improved
system. Moreover, additional configurations can be simulated. For example, the
influence of the surface slab stiffness: only the two extreme cases were investigated
experimentally (infinitely soft or infinitely stiff), but the study demonstrated the
strong influence of the surface boundary condition, so the strong interaction between
the improvement system and the type of structure placed on the surface. This study
would nevertheless necessitate taking explicitly this part of the system into
account by implanting structural elements (shell elements) or volume elements with
the corresponding slab stiffness – even if this would considerably increase the
computation time. Moreover, the implementation of a more complex constitutive
model for the LTP would be required to take the observed behavior into account
even better.

Concerning the behavior of the piled structure under seismic loading, only a few
studies have considered the nonlinearity of the soil response and most have limited
to a single pile. [LOP 17] performed a dynamic numerical analysis using FLAC3D
(Figure 3.36). Linear elastic and elastoplastic constitutive models were used to
represent the soil behavior and different support (pile head connection and
embedment) conditions were taken into account. The dynamic loading was a
sinusoidal acceleration wave applied at the base of the model in the horizontal
direction. Lateral absorbent boundaries were placed to minimize wave reflection.
Rayleigh damping for the soil and local damping for the superstructure and rigid
elements were used. The analysis of the internal forces and the displacement in the
vertical reinforcements during the dynamic loading was performed for the several
cases studied. The embedment conditions of the piles affected the shear forces and
bending moments (in particular at the toe level), which were higher in piles
anchored in the bedrock compared to the case of piles placed on the bedrock or
floating piles, due to extra kinematic forces induced by the lateral deformation of the
surrounding soil. Compared to the case of the piles connected to the upper structure,
the system with rigid inclusions allowed the reduction of the head level efforts, due
to the presence of an LTP between the structure and the rigid inclusions. In fact, for
the connected pile system, the connection with the structure increased the inertial
forces at this level.

Another study of the seismic ground response in the presence of rigid inclusions
performed using the finite difference method was presented in [MÁN 16]. It used a
viscoelastic model for the soil and showed that the dynamic characteristics of the
structure at the surface were the main parameters affecting the ground response of
the system.

Copyright Iste 2022 / File for personal use of Robert Charlier only
138 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 3.36. Numerical model (source: from [LOP 17]). For a color version of this
figure, see www.iste.co.uk/grange/soil.zip

3.5. Conclusion

The finite difference numerical method addressed in this chapter is one of the
direct methods to study the SSI problem in the field of geotechnical engineering.
It allows the explicit modeling of both the ground mass and the structural element in
the same model in a continuum, taking the interaction between the various parts
fully into account. However, it necessitates the accurate modeling of the behavior of
several parts and their interfaces; thus, it also necessitates the knowledge of the
appropriate material properties.

In this chapter, it was shown that this type of numerical modeling is an

interesting and efficient tool to help for the development or improvement of design
methods, for current or new types of geotechnical structures. For this purpose, the
comparison with experimental results is compulsory (using physical modeling
and/or field testing). In particular, by easily achieving parametric studies, the
numerical modeling method makes it possible to clearly identify the controlling
parameters, which should be inevitably taken into account even in simplified design
methods.

However, in the field of geotechnical engineering, this method is adapted to

accurately predict the behavior of geotechnical structures, both in terms of ultimate
and serviceability states, by giving access to the stress and displacement fields
throughout the structure. Nevertheless, this necessitates the preliminary accurate
knowledge of the stress–strain behavior of the constituting materials, given by an
appropriate site investigation and interpretation. The implementation of numerical
models also makes it possible to optimize the envisaged structure, by simulating
various possibilities.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 139

3.6. References

[ABD 11] ABDELOUHAB A., DIAS D., FREITAG N., “Numerical analysis of the behaviour of
mechanically stabilized earth walls reinforced with different types of strips”, Geotextiles
and Geomembranes, vol. 29, pp. 116–129, doi:10.1016/j.geotexmem.2010.10.011, 2011.
[BIL 93] BILLAUX D., CUNDALL P., “Simulation des géomatériaux par la méthode des
éléments Lagrangiens”, Revue française de géotechnique, vol. 63, pp. 9–21, 1993.
[BRI 15] BRIANÇON L., DIAS D., SIMON C., “Monitoring and numerical investigation of a rigid
inclusions – Reinforced industrial building”, Canadian Geotechnical Journal, vol. 52,
pp. 1–13. dx.doi.org/10.1139/cgj-2014-0262, 2015.
[CHA 17] CHAUDHARY B., HAZARIKA H., KRISHAN A.M., “Effect of backfill reinforcement on
retaining wall under dynamic loading”, in HAZARIKA, H., KAZAMA, M., LEE, W.F. (eds),
Geotechnical Hazards from Large Earthquakes and Heavy Rainfalls, Springer Japan,
2017.
[CUN 82] CUNDALL P.A., “Adaptive density-scaling for time-explicit calculations”, 4th
International Conference on Numerical Methods in Geomechanics, Edmonton, pp. 23–26,
1982.
[DAM 13] DAMIANS I.P., BATHURST R.J., JOSA A. et al., “Comparison of finite element and
finite difference modelling results with measured performance of a reinforced soil wall”,
Geo Montréal, 2013.
[DIA 13] DIAS D., KASTNER R., “Movements caused by the excavation of tunnels using face
pressurized shields: Analysis of monitoring and numerical modeling results”, Engineering
Geology, vol. 152, pp. 17–25, available at: http://dx.doi.org/10.1016/j.enggeo.2012.10.002,
2013.
[DO 13] DO N.A., DIAS D., ORESTE P. et al., “3D modelling for mechanized tunnelling in soft
ground – Influence of the constitutive model”, American Journal of Applied Sciences,
vol. 10, no. 8, pp. 863–875, doi: 10.3844/ajassp.2013.863.875, 2013.
[DO 14a] DO N.A., DIAS D., ORESTE P., “2D seismic numerical analysis of segmental tunnel
lining behaviour”, Bulletin of the New Zealand Society for Earthquake Engineering,
vol. 47, no. 3, pp. 206–216, 2014.
[DO 14b] DO N.A., DIAS D., ORESTE P. et al., “2D tunnel numerical investigation: The
influence of the simplified excavation method on tunnel behaviour”, Geotechnical
and Geological Engineering, vol. 32, no. 1, pp. 43–58, doi: 10.1007/s10706-013-9690-y,
2014.
[DO 17] DO N.A., DIAS D., “A comparison of 2D and 3D numerical simulations of tunnelling
in soft soils”, Environmental Earth Sciences, vol. 76, no. 3, 2017.
[HIC 08] HICHER P.Y., SHAO J.F., Constitutive Modeling of Soils and Rocks, ISTE Ltd,
London and John Wiley & Sons, New York, 2008.

Copyright Iste 2022 / File for personal use of Robert Charlier only
140 Deterministic Numerical Modeling of Soil–Structure Interaction

[HOU 17] HOUDA M., JENCK O., EMERIAULT F., “Rétro-analyse numérique du comportement
d’un massif de sol renforcé par inclusions rigides sous chargement cyclique (Numerical
back analysis of the behaviour of soft soil improved by rigid piles under cyclic loading)”,
Proceedings of the International Conference on Soil Mechanics and Geotechnical
Engineering, pp. 12–18, Seoul, September 2017.
[ITA 09] ITASCA CONSULTING GROUP., Flac 3D Version 4.0, User’s guide, 2009.
[JAN 14] JANIN J.P., DIAS D., “Tunnel face reinforcement by bolting – Numerical modelling
of centrifuge tests”, Soils and Rocks, vol. 37, no. 1, 2014.
[JEN 04] JENCK O., DIAS D., “Analyse tridimensionnelle en différences finies de l’interaction
entre une structure en béton et le creusement d’un tunnel à faible profondeur”,
Géotechnique, vol. 8, no. 54, pp. 519–528, 2004.
[JEN 07a] JENCK O., DIAS D., KASTNER R., “Two-dimensional physical and numerical
modeling of a pile-supported earth platform over soft soil”, Journal of Geotechnical and
Geoenvironmental Engineering, vol. 133, no. 3, pp. 295–305, doi: 10.1061/(ASCE)1090-
0241(2007)133:3(295), 2007.
[JEN 07b] JENCK O., DIAS D., KASTNER R. et al., “Three-dimensional finite-difference
modeling of a piled embankment on soft ground”, Proceedings of the 5th International
Workshop on Application of Computational Mechanics in Geotechnical Engineering,
Guimares, Portugal, 2007.
[JEN 09a] JENCK O., DIAS D., KASTNER R., “Discrete element modelling of a granular
platform supported by piles in soft soil – Validation on a small scale model test and
comparison to a numerical analysis in a continuum”, Computers and Geotechnics,
vol. 36, no. 6, pp. 917–927, doi: 10.1016/j.compgeo.2009.02.001, 2009.
[JEN 09b] JENCK O., DIAS D., KASTNER R., “Three-dimensional numerical modeling of a piled
embankment”, International Journal of Geomechanics, vol. 9, no. 3, doi: 10.1061/(ASCE)1532-
3641(2009)9:3(102), 2009.
[KRI 17] KRISHNA A.M., BHATTACHARJEE A., “Behavior of rigid-faced reinforced
soil-retaining walls subjected to different earthquake ground motions”, International
Journal of Geomechanics, vol. 17, no. 1, 2017.
[LEE 12] LEE C.J., “Three-dimensional numerical analyses of the response of a single pile
and pile groups to tunnelling in weak weathered rock”, Tunnelling and Underground
Space Technology, vol. 32, pp. 132–142, 2012.
[LOG 01] LOGANATHAN N., POULOS H.G., XU., K.J., “Ground and pile-group responses due to
tunnelling”, Soils and Foundations, vol. 41, no. 1, pp. 57–67, 2001.
[LOP 17] LOPEZ JIMENEZ G.A., DIAS D., OKYAY U.S. et al., “Dynamic behavior of the
soil-pile-structure interaction in soft soils. Influence of the pile type”, International
Journal of Geomechanics, April 2017.
[MÁN 16] MÁNICA-MALCOM M.A., OVANDO-SHELLEY E., BOTERO JARAMILLO E., “Numerical
study of the seismic behavior of rigid inclusions in soft Mexico City clay”, Journal of
Earthquake Engineering, vol. 20, no. 3, pp. 447–475, 2016.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 SSI Analysis in Geotechnical Engineering Problems Using a Finite Difference Method 141

[MAR 82] MARTI J., CUNDALL P.A., “Mixed discretization procedure for accurate solution of
plasticity problems”, International Journal of Numerical and Analytical Methods in
Geomechanics, vol. 6, pp. 129–139, 1982.
[MES 04] MESTAT, P., BOURGEOIS, E., RIOU, Y., “Numerical modelling of embankments and
underground works”, Computer and Geotechnics, vol. 3, pp. 227–236, 2004.
[NUN 13] NUNEZ M.A, BRIANÇON L., DIAS D., “Analyses of a pile-supported embankment over
soft clay: Full-scale experiment, analytical and numerical approaches”, Engineering
Geology, vol. 153, pp. 53–67, available at: http://dx.doi.org/10.1016/j.enggeo.2012.11.006,
2013.
[PUZ 12] PUZRIN A., Constitutive Modelling in Geomechanics, Springer, Berlin, 2012.
[TRA 19] TRAN Q.A., VILLARD P., DIAS D., “Discrete and continuum modelling of soil
arching between piles”, International Journal of Geomechanics, vol. 19, no. 2, available
at: https://doi.org/10.1061/(ASCE)GM.1943-5622.0001341, 2019.
[YU 15a] YU Y., BATHURST R.J., MIYATA Y., “Numerical analysis of a mechanically
stabilized earth wall reinforced with steel strips”, Soils and Foundations, vol. 55, no. 3,
pp. 536–547, 2015.
[YU 15b] YU Y., DAMIANS I.P., BATHURST R.J., “Influence of choice of FLAC and PLAXIS
interface models on reinforced soil–structure interactions”, Computers and Geotechnics,
vol. 65, pp. 164–174, 2015.

Copyright Iste 2022 / File for personal use of Robert Charlier only
Copyright Iste 2022 / File for personal use of Robert Charlier only
 4

Macroelements for
Soil–Structure Interaction

4.1. Introduction

The influence of soil–structure interaction (SSI) on the dynamic response of

structures – such as bridges, towers or viaducts – is the object of a significant number
of recent studies. In the current practice, SSI is often disregarded, because of the
difficulty in modeling it, and also because it is often assumed that the increase in the
fundamental period of the soil–structure system (due to the deformability of the soil
and foundation) along with the increase in the damping coefficient (due to hysteretic
behavior and radiation damping) have beneficial effects on the structural behavior
[CIA 95, ELN 96]. Recent works [GAZ 98, MYL 00, JER 04], however, have revealed
that stresses in the structure can be amplified in the presence of SSI depending on
the characteristics of the loading and of the soil–structure system. Ignoring SSI can
therefore lead to a non-conservative assessment of the safety level.

The nonlinear finite element method can be adopted to simulate complex dynamic
SSI phenomena. It necessitates, however, discretizing the structure, the foundation
and the underlying soil volume, and adopting constitutive models capable of
reproducing the soil and the structural material (e.g. concrete and steel) behavior under
cyclic/dynamic conditions. Several examples can be found in the literature, see, for
example, [ELG 08, JER 09, KOT 13, CHO 19]. The main disadvantage of the approach
lies in the necessary dimension of soil volume interacting with the foundation,
particularly under seismic conditions, leading to a significant number of degrees of
freedom for the soil discretization. The method also requires a detailed geotechnical
characterization of the soil deposit and of the structural materials (properties and the

Chapter written by Diana S ALCIARINI, Stéphane G RANGE , Claudio TAMAGNINI and

Panagiotis KOTRONIS.

Copyright Iste 2022 / File for personal use of Robert Charlier only
144 Deterministic Numerical Modeling of Soil–Structure Interaction

exact position of steel reinforcement) and absorbing boundaries to accurately simulate

the energy radiation in a deformable medium. The resulting computational cost is
prohibitive in everyday engineering practice and often requires massive parallelization
numerical tools, especially when strong nonlinearities are present in the structure and
in the soil.

On the other hand, simplified tools, such as the Winkler-type models (distributed
1D springs (and dashpots) across the interface between the soil and the footing)
should not be preferred in view of all the uncertainties underlying the choice of their
parameters [FAR 15]. Their values and distribution vary with frequency, and there is
no unique distribution that preserves the global foundation stiffness for all degrees
of freedom; in that case, the rocking stiffness is not correctly matched [FAR 15].
The same limitations apply, perhaps even to a larger extent, to pile models developed
within the same approach [LI 16]. Winkler-type models require, as input data, a series
of soil pressure–displacement (p–y) curves, which are very difficult to select in the
absence of data from instrumented lateral pile load tests. For example, Murchison and
O’Neill [MUR 84], in a study comparing four proposed procedures for selecting p–y
curves with data from field tests, showed that errors in pile head deflection predictions
could be as large as 75%. It is also uncertain how the p–y curves are affected by pile
head constraints and the relative stiffness of the pile and soil [LI 16].

A significant step forward in the development of efficient, robust and accurate

calculation tools for SSI is the so-called macroelements, a term coined for the first
time by [NOV 91], for the description of the soil–foundation system response by
a single constitutive equation that describes the overall behavior of the system. A
macroelement can be viewed as an “upscaled” version of classical macroscopic
models considering the soil as a continuous (albeit sometimes multiphase) medium
where the global behavior of the foundation and of the soil volume interacting with
it is “lumped” into a single, integral, constitutive equation that links the evolution
of the resultant loads/moments on the foundation to the corresponding displacement/
rotation histories [LI 16]. A similar idea in structural engineering to model plastic
hinge areas in beam–column joints can be found in [ELA 92] and [DAV 98].

The key element for modeling some of the fundamental features of the global
behavior of the soil–foundation system (nonlinearity, irreversibility, dependence on
previous load history, coupling between the different degrees of freedom of the
system) consists of formulating the constitutive equations of the macroelement in
incremental form, that is, in the form of evolution laws for the variables of the
system. Therefore, it is not surprising that the first applications of the macroelement
concept are based on the principles of hardening plasticity [NOV 91] to reproduce
the hysteretic response of a shallow soil–foundation system. Macroelements for
shallow foundations for monotonic loads are among others proposed by [MON 97,
GOT 99, MAR 01, LEP 01, HOU 02, BIE 06] and for cyclic/dynamic loads by
[CRÉ 01, CRÉ 02, PRI 03, PRI 06, CHA 08] and [GRA 08a, GRA 08b, GRA 09].

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 145

An alternative approach to the theory of elastoplasticity was used by [SAL 09],

who derived the constitutive equations of an inelastic macroelement from the basic
principles of the generalized hypoplasticity theory [NIE 02]. To extend the basic
formulation of the macroelement to the cyclic/dynamic load conditions, the model was
enriched by a suitable kinematic internal variable (called the internal displacement),
adapting the approach proposed by [NIE 97] for continuous media. The hypoplastic
macroelement is characterized by the absence of an elastic domain in the space of the
generalized stresses and by a continuous variation of the stiffness matrix of the system,
depending on the direction of the generalized velocity (incremental nonlinearity).
Following the work of [SAL 09], macroelements based on hypoplasticity are
developed for single vertical piles [LI 16], single inclined piles [LI 18], pile groups
[PER 20] and caisson foundations [JIN 19b].

The aim of this chapter is, on the one hand, to show the potentiality offered by
the macroelements for the quantitative evaluation of SSI effects and, on the other, to
compare the two alternative approaches in the formulation of macroelements (based on
elastoplasticity and hypoplasticity). For this purpose, the two macroelements proposed
by [GRA 09] and [SAL 09] are used in the numerical simulation of the seismic
response of a three-pier viaduct, taken as a benchmark for structures of relevant
practical interest.

The chapter is organized as follows. The concept of generalized forces is presented

in section 4.2. The general formulation of the incremental constitutive equations,
provided by the theories of elastoplasticity and hypoplasticity for the inelastic
macroelements, is given in section 4.3. The details of the constitutive equations of
the two macroelements are provided in section 4.4. Section 4.5 presents a case study.
To compare the two macroelements, particular attention is paid to the calibration of
their parameters in section 4.6. The main results of the numerical simulations are
discussed and compared in section 4.7. Conclusive remarks are finally summarized in
section 4.8.

4.2. The concept of generalized forces: Eurocode 8 recommendations

In the formulation of the two macroelements considered in this chapter, the

constitutive equations are written in terms of generalized forces and conjugated
generalized displacements. This is motivated by the well-known definition of the
bearing capacity of foundations, Nu = Sqmax , where S is the footing area and
qmax is the ultimate compression stress of the soil under a vertical centered load (see
[DAV 73, MAT 79, PHI 03, RAN 04]), given by
as
qmax = γBNγ + q0 Nq + bs cNc [4.1]
2
where q0 is the vertical effective stress at the bottom of the foundation; Nγ , Nq and
Nc are the bearing capacity factors depending on the soil friction angle ϕ – see, for

Copyright Iste 2022 / File for personal use of Robert Charlier only
146 Deterministic Numerical Modeling of Soil–Structure Interaction

example, the relations provided by [CAQ 66] and [RAN 04]; γ is the soil unit weight;
B is the footing dimension; and c is the soil cohesion. as and bs are the shape factors
defined in Table 4.1.

Circular Rectangular Strip (A → ∞)

 
as 0.6 1 − 0.2 B
A
1
 B

bs 1.3 1 + 0.2 A 1

Table 4.1. Shape factors for the bearing capacity of circular,

rectangular and strip foundations (note: A, B = the length
and width of the rectangular footing)

More recently, several authors provided the ultimate bearing capacity failure
conditions by means of suitable failure criteria defined in the space of generalized
forces applied to the foundation (see Figure 4.1, where V = NEd is the vertical force,
H = VEd is the horizontal force and M = MEd is the moment applied at the center
of the foundation).

Figure 4.1. Generalized forces for a shallow foundation

In the work by [NOV 91], the shape of the failure locus of a shallow foundation,
expressing a failure criteria in graphical form, was, for the first time, experimentally
defined under a combination of generalized forces for cohesive or frictional soils.
[BUT 94, CAS 02, CAS 04, MAR 94, GOT 99] and recently [CHA 07] also provided
failure criteria for shallow foundations (see also [LI 14] for pile and [JIN 19a] for
caisson foundations). Figure 4.2 shows a comparison between several failure criteria
for shallow foundations when projected in different planes.

Chapter 5 of Eurocode 8 [EUR 04], related to SSI problems under seismic loading
conditions, provides its own criterion for shallow foundations in terms of generalized
forces (equation [4.2]).
 cT  cT   cM  cM
1 − eF̄ β V̄ 1 − f F̄ γ M̄
 k b +  k d − 1 ≤ 0 [4.2]
N̄ a 1 − mF̄ k − N̄ N̄ c 1 − mF̄ k − N̄

where the dimensionless variables N̄ = γRd NEd

Nmax , V̄ =
γRd VEd
Nmax and M̄ = γRd MEd
BNmax
are defined in terms of the vertical design load (NEd ), the design horizontal load

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 147

(VEd ), the design moment (MEd ), the maximum vertical load (Nmax ), the soil unit
weight (γ) and the footing dimension (B). The parameter F̄ represents the horizontal
inertial force of the soil, and a, b, c, d, e, f , m, β, γ, cM and cT are the parameters
depending on the cohesive or frictional character of the soil. Figure 4.3 represents the
failure surface in the M − V space, which is described by equation [4.2] for a purely
cohesive soil (φ = 0) and a purely frictional soil (c = 0).

0.15 0.2
Nova Nova
Butterfield 0.15 Butterfield

0.1 0.15 Pecker Pecker

0.1
0.1
0.05
0.05 0.05
0
0 0
-0.05
-0.05 -0.05
-0.1
-0.1
-0.1
Nova
-0.15 -0.15
Butterfield
-0.15
Pecker
-0.2 -0.2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2

Figure 4.2. Comparisons between different failure surfaces plotted with dimensionless
variables: strip foundations for [NOV 91] and [PEC 97] and circular footings for
[BUT 94]. For a color version of this figure, see www.iste.co.uk/grange/soil.zip

(a) purely cohesive soil (b) purely frictional soil

Figure 4.3. Representation of the bearing capacity for a shallow foundation from
Eurocode 8 (a) for a purely cohesive soil and (b) for a purely frictional soil.
For a color version of this figure, see www.iste.co.uk/grange/soil.zip

Failure criteria are needed to evaluate the resistance of a foundation under a

combination of vertical and horizontal forces or moments. However, if displacements
have to be estimated, the ultimate failure conditions are not sufficient. For this reason,
nonlinear models based, for example, on plasticity mechanisms and failure criteria,
are developed to describe the reversible and irreversible load–displacement behavior
of the foundation–soil system, under cyclic or dynamic loadings.

Copyright Iste 2022 / File for personal use of Robert Charlier only
148 Deterministic Numerical Modeling of Soil–Structure Interaction

4.3. Macroelements for shallow foundations

4.3.1. Generalities

The response of the soil–foundation system under given loading conditions is

described by means of a single constitutive equation that links the vector of the
generalized load applied to the foundation:
 T
T := V , Hx , My /Bx , Hy , Mx /By [4.3]

to the history of the generalized displacement, collected in the vector

 T
U := Uz , Ux , Θy Bx , Uy , Θx By [4.4]

Here, Bx and By represent two characteristic dimensions of the footing, in the

horizontal x− and y−directions, as shown in Figure 4.4. The torsion moment Mz and
the rotation Θz are neglected; the extension to the six-dimensional degrees of freedom
is discussed in [BIE 06].

Figure 4.4. Components of T e U : (a) generalized forces and

(b) generalized displacements

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 149

To reproduce correctly some important features of the experimentally observed

behavior of the foundation–soil system such as nonlinearity, irreversibility and
dependence on the past loading history, the constitutive equation for the macroelement
must be formulated in rate form, as an evolution law of the generalized loads:

Ṫ = F T , Q, U̇ [4.5]

starting from a known state, which is defined by T and Q. The latter is a vector that
collects the internal variable, accounting for the effect of the previous loading history
on the system response.

It is worthwhile expressing the generalized forces and the displacements in the

following non-dimensional form [GRA 08a, GRA 08b]:
 T 1
t = v , h x , my , h y , mx := T [4.6]
Vf
 T 
u = u z , u x , θy , u y , θx := U [4.7]
A
where Vf is the footing bearing capacity under a vertical centered load; A = Bx By is
the footing area, and  = Bx2 + By2 is a characteristic length accounting for the size
of the footing.

With the definitions of [4.6] and [4.7], the constitutive equation for the macro-
element can be written in the following non-dimensional form:

ṫ = f (t, q, u̇) [4.8]

where q is the non-dimensional internal variable vector.

Since the properties of f (t, q, u̇) are selected according to the basic features of
observed behavior, f must depend on the loading direction to reproduce an inelastic
behavior [KOL 91], and equation [4.8] can have the following alternative form:

ṫ = K (t, q, η) u̇ [4.9]

where K is the tangent stiffness matrix of the system, t is the function of the current
loading state, q is the previous loading state and the unit vector:
u̇
η :=
u̇
which describes the loading direction.

To construct the function K, two different strategies are elaborated hereafter: the
theories of elastoplasticity and hypoplasticity.

Copyright Iste 2022 / File for personal use of Robert Charlier only
150 Deterministic Numerical Modeling of Soil–Structure Interaction

4.3.2. Macroelements formulated in the framework of hardening

elastoplasticity

For macroelements developed in the framework of the theory of elastoplasticity

[NOV 91, MAR 01, CRÉ 01, GRA 08b], the constitutive equation is built starting
from the fundamental assumptions of:
– elastic and plastic decomposition of the generalized velocity:

u̇ = u̇e + u̇p [4.10]

– existence of an elastic law:

ṫ = K e (t) u̇e [4.11]

where K e is the elastic stiffness matrix of the soil foundation system;

– existence of a yield function fy (t, q) for each mechanism in the generalized load
space;
– existence of a plastic potential function g (t, q), evaluated by means of the
following flow rule:
∂g
u̇p = γ̇ (t, q) γ̇ ≥ 0 [4.12]
∂t
where γ̇ is the plastic multiplier;
– existence of a suitable hardening law for the internal variables

q̇ = γ̇h (t, q) [4.13]

– enforcement of Prager’s consistency condition.

The resulting constitutive equation in rate form can then be written as


T
H (γ̇) ∂g ∂fy
ṫ = K ep (t, q) u̇ K ep := K e − Ke Ke
Kp ∂t ∂t
[4.14]

T
H (γ̇) ∂fy
q̇ = H (t, q) u̇ H := h Ke [4.15]
Kp ∂t
where Kp > 0 is a scalar quantity given by
T T
∂fy ∂g ∂fy
Kp := Ke − Hq [4.16]
∂T ∂T ∂T

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 151

and H(γ̇) is the Heaviside step function, equal to one if the plastic multiplier:
T
1 ∂f
γ̇ := K e u̇
Kp ∂t
is positive and zero otherwise.

4.3.3. Macroelements formulated in the framework of hypoplasticity

In the approach based on the theory of hypoplasticity, the constitutive function f

is constructed starting from the following general expression:

ṫ = L (u̇ − Y m u̇) [4.17]

where the matrix (5 × 5) L(t, q), the unit vector m(t, q) and the scalar quantity Y (t)
(varying between 0 and 1) are constitutive equations, built from observed behavior of
the system, as better described in the following.

The application of the Euler theorem to the second term of equation [4.17] can be
written as

ṫ = K hp u̇ K hp := L − Y mη T [4.18]

where K hp (t, q, η) is the tangent stiffness matrix of the system. Note that, unlike the
theory of hardening plasticity, where the stiffness matrix can assume only two distinct
values (elastic and elastoplastic), in the theory of hypoplasticity, the tangent stiffness
of the system is a continuous function of η. This property is named incremental
nonlinearity by [DAR 78].

The physical meaning of the three constitutive equations L, m and Y can be

explained assuming that the scalar function Y (loading function) has values close to
zero within the failure surface Sf , and monotonically increases when the distance from
Sf to the current loading state decreases, up to reach the unit value for t = tf ∈ Sf .

When the current state is that Y → 0, equation [4.17] reduces to:

ṫ  Lu̇

and the response of the system is approximately linear. In this condition, the matrix L
can be assimilated to the elastic stiffness matrix K e of the theory of elastoplasticity.

If the soil–foundation system is subjected to displacements having constant

velocity module and direction (u̇ = u̇ η = Cη constant), in failure conditions
(Y = 1), equation [4.17] becomes:

ṫ = CL

Copyright Iste 2022 / File for personal use of Robert Charlier only
152 Deterministic Numerical Modeling of Soil–Structure Interaction

Hence, the vector m represents the direction at failure, associated with the loading
state tf that is asymptotically reached by the system.

To correctly describe the response of the soil–foundation system under

cyclic loads, [SAL 09] introduced the following modification to the constitutive
equation [4.18], starting from the extension proposed by [NIE 97] for the hypoplastic
constitutive model of [WOL 96].

Let us introduce an internal variable δ, called the internal displacement, with the
following evolution law:
 
I − ρβr η δ η Tδ u̇ (η δ · η > 0)
δ̇ = [4.19]
u̇ (η δ · η ≤ 0)

where ηδ := δ/ δ is the unit vector, which provides the direction of the internal
displacement; ρ ∈ [0, 1] is a normalized measure of the module of δ; βr is a model
constant, and I is the identity matrix (5 × 5). The evolution law [4.19] is constructed
to have internal displacements that tend to be tangent to the path of the generalized
displacements (Figure 4.5).

Figure 4.5. Effect of the displacement history on the system response, for a
given loading state, varying the direction of the generalized velocity u̇

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 153

When including the internal displacement δ, the constitutive equation of the

macroelement [4.18] modifies as follows ([SAL 09]):

ṫ = K hp (t, q, δ) u̇ [4.20]

where:
 (t, q, δ)
K hp (t, q, δ) := [ρχ mT + (1 − ρχ ) mR ] L (t, q) + K [4.21]

and

ρχ (1 − mT ) (Lη δ ) η Tδ − ρχ Y mη δ T (η δ · η > 0)
 (t, q, δ) :=
K χ
ρ (mR − mT ) (Lη δ ) η Tδ (η δ · η ≤ 0)
[4.22]

with mR , mT and χ being model constants.

Due to the evolution law [4.19], the internal displacement δ takes into account the
displacement history which occurred during the previous loading stages. Varying the
generalized velocity direction with respect to the internal displacement direction, the
constitutive equation [4.21] can assume the following forms:
a) For proportional loading, ρ → 1 and η δ → η (Figure 4.5a); in these conditions,
from equation [4.21] and equation [4.22]:

ṫ = L (u̇ − Y m u̇)

i.e. equation [4.17] of the original hypoplastic model.

b) For a loading reversal, starting from a hypoplastic state (i.e. η = −η δ and ρ =
1), we have:

K hp = mR L [4.23]

and the system response is locally (hypo)elastic.

c) For a loading path orthogonal to the internal displacement direction, starting
from a hypoplastic state (i.e. η · ηδ = 0 and ρ = 1), we have:

K hp = mT L [4.24]

and the system response is again locally (hypo)elastic, but characterized by a different
stiffness (typically mT ≤ mR ).

In all the intermediate cases, the expression of the tangent stiffness of the system
is obtained by interpolation.

Copyright Iste 2022 / File for personal use of Robert Charlier only
154 Deterministic Numerical Modeling of Soil–Structure Interaction

4.4. The considered macroelements

Two macroelements for shallow foundations are considered and compared:

the elastoplastic macroelement with isotropic hardening proposed by [GRA 08a,
GRA 08b, GRA 09] for the SSI analysis in seismic conditions, and the hypoplastic
macroelement by [SAL 09]. In the following, a brief description of the relevant
aspects of the two formulations is presented; complete details are given in
the cited references.

4.4.1. The elastoplastic macroelement

The elastoplastic macroelement by [GRA 09] considers the following decom-

position of the generalized velocity:
u̇ = u̇e + u̇p + u̇up [4.25]
p up
The velocities u̇ and u̇ refer to two different mechanisms of plastic deformation:
the first mechanism is related to the irreversible deformations that develop in the
soil when the applied loads reach the failure conditions of the foundation; the
second mechanism is due to the partialization of the contact area between soil and
foundation (phenomenon defined as “uplift” by [GRA 09]). In the following, this
second mechanism is not considered.

The elastic component of the generalized velocity is provided by the elastic

law [4.11], where the stiffness matrix is given by
K e = diag {kvv ; khx ; kmy ; khy ; kmx } [4.26]
with
√
A G 1
kα = βα [4.27]
 qf (1 − ν)
and α ∈ {vv ; hx ; my ; hy ; mx}. In equation [4.26], G and ν are the shear modulus
and the Poisson ratio of the soil, under the hypothesis of linear elastic, homogeneous
and isotropic half-space; qf := Vf /A is the ultimate bearing capacity of the
foundation under a vertical centered load, and the non-dimensional coefficients βα
are given by [PHI 03] as function of the ratio between the footing size Bx /By .

The elastic domain of the macroelement is defined starting from the following
yield function:
 2  2
hx τhx my τmy
fy (t, τ , γ, ρ) = − + −
ρav c (γ − v)d ρ ρbv e (γ − v)f ρ
 2  2
hy τhy mx τmx
+ − + − − 1 = 0 [4.28]
ρav c (γ − v)d ρ ρbv e (γ − v)f ρ

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 155

where a, b, c, d, e and f are the model constants, while τ , ρ and γ are the internal
variables controlling the position, size and shape of the failure surface.

The failure surface can be obtained by equation [4.28], putting τ = 0 and

ρ = γ = 1:

 2  2
hx my
fc (t) = +
av c (1 − v)d bv e (1 − v)f
 2  2
hy mx
+ + − 1 = 0 [4.29]
av c (1 − v)d bv e (1 − v)f
A plot of a possible yield surface and the failure surface is given in Figure 4.6.

Figure 4.6. Elastoplastic model by [GRA 09]: evolution of the yield

surfaces with the plastic deformation and failure surface

In plastic conditions (f = 0 and f˙ = 0), the plastic component of the generalized

velocity is given by equation [4.12], with the following plastic potential function:
 2
hx (κγ + v0 )d τhx
g(t, τ , γ, ρ, v0 ) = c d d
−
ρav0 (γ − v) (κγ + v) ρ
 2
my (ξγ + v0 )f τmy
+ e −
ρbv0 (γ − v)f (ξγ + v)f ρ
 2
hy (κγ + v0 )d τhy
+ c d d
−
ρav0 (γ − v) (κγ + v) ρ
 2
mx (ξγ + v0 )f τmx
+ e − −1=0 [4.30]
ρbv0 (γ − v)f (ξγ + v)f ρ

Copyright Iste 2022 / File for personal use of Robert Charlier only
156 Deterministic Numerical Modeling of Soil–Structure Interaction

where κ and ξ are the two model constants and v0 is the vertical force, normalized
with respect to the current state, see [GRA 08a, GRA 08b] for further details. The
internal variables τ , γ and ρ control, respectively, the position of the center of the yield
surface, its size in the v-direction and its shape. Their evolution laws equation [4.13]
derive from the experimental observations by [GOT 99] and from the results of finite
element numerical simulations by [CRÉ 01].

The calibration of the elastoplastic macroelement requires the determination of 16

constants described in Table 4.2. The constants Bx and By define the geometry of the
foundation; kvv , khx , kmy , khy and kmy are the components of the normalized elastic
stiffness matrix; a, b, c, d, e and f control the shape of the yield surface; ξ and κ control
the direction of the plastic flow in the (hx , v)-plane, (hy , v)-plane, (mx , v)-plane
and (my , v)-plane, while Vf is the footing bearing capacity under centered vertical
load.

Constant Unit Description

Bx [ L]
footing size
By [ L]
kvv [– ]
khx [– ]
kmy [– ] elastic stiffness matrix components
khy [– ]
kmx [– ]
a [– ]
b [– ]
c [– ]
yield function parameters
d [– ]
e [– ]
f [– ]
ξ [– ]
plastic potential function parameters
κ [– ]
Vf [ F] bearing capacity under vertical centered load

Table 4.2. Description of the constants of the

elastoplastic macroelement

4.4.2. The hypoplastic macroelement

In the formulation of the hypoplastic macroelement, [SAL 09] relies on the model
by [NOV 91] to define both the loading function Y and the direction of the plastic
flow m.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 157

In the space of non-dimensional loads, the failure surface proposed by [NOV 91]
can be generalized as follows:
 2  2  2  2
hx mx hy my
fc (t) = + + + − v 2 (1 − v)2β = 0
μ ψ μ ψ
[4.31]

where μ, ψ and β are the model constants. For any “stable” loading state within the
failure domain, a surface that is omothetic to the failure surface can be defined as
 2  2  2  2 2β
hx mx hy my v
f ∗ (t, γ) = + + + − v2 1 − =0
μ ψ μ ψ γ
[4.32]

passing for t and intercepting the v-axis at v = ξ (Figure 4.7). When the loading
state varies, the parameter γ varies between (0, 1] and reaches the value of 1 when the
loading state is on the failure surface. Owing by this, the following loading function
has been adopted:

Y (t) = {γ (t)}κ [4.33]

where κ is a model constant.

Figure 4.7. Definition of the surface f = 0 and γ

The plastic flow direction is evaluated as

1 ∂g
m(t) = [4.34]
∂g/∂t ∂t

Copyright Iste 2022 / File for personal use of Robert Charlier only
158 Deterministic Numerical Modeling of Soil–Structure Interaction

starting from the generalization of the plastic potential function defined by the model
of [NOV 91]:
 2  2  2  2 2β
hx mx hy my v
g(t, γ) = + + + − v2 1 −
λh μ λm ψ λh μ λm ψ γg
[4.35]
In equation [4.34], λh and λm are the model constants, while γg is a dummy
variable, obtained considering the condition g(t, γg ) = 0 at the current loading state.
The geometrical meaning of γg is given in Figure 4.8.

Figure 4.8. Definition of the surface g = 0 and γg

Considering that, after a full reversal of the loading path, the hypoplastic macro-
element responds similarly to a linear elastic system, the L matrix of the hypoplastic
model can be written as
1 1
L= Ke = diag {kvv ; khx ; kmy ; khy ; kmx } [4.36]
mR mR
where kα are provided by equation [4.27].

To complete the hypoplastic macroelement description, the normalized internal

displacement ρ, appearing in equation [4.21] and equation [4.22], has to be defined.
The following expression has been adopted by [SAL 09]:

δW δT W δ
ρ= := [4.37]
R R
where:
W := diag {1 ; wh ; wm ; wh ; wm } [4.38]

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 159

is a diagonal matrix of weights, accounting for the geometry of the failure surface,
while R represents the size region within the displacement space where the stiffness
of the system is high and the behavior is quasilinear (“elastic nucleus”), see [SAL 09]
for further details.

The hypoplastic macroelement is characterized by 19 constants, which are

described in Table 4.3. As in the elastoplastic model, Bx and By define the footing
size, while kvv , khx , kmy , khy and kmy are the components of the normalized
elastic stiffness matrix. μ, ψ and β control the shape of the failure domain in the
generalized loading space. In particular, μ can be interpreted as the friction coefficient
at the soil–foundation interface, while ψ controls the failure condition when vertical
eccentric loads are present [NOV 91]. λh and λm control the plastic flow direction
along the planes (hx , v), (hy , v), (mx , v) and (my , v), while κ controls the degree
of nonlinearity of the system response. Finally, mR , mT , R, βr and χ control the
evolution of the stiffness of the system at varying of the loading path direction and
history of previous loading conditions.

Constant Unit Description

Bx [ L]
footing size
By [ L]
kvv [– ]
khx [– ]
kmy [–] components of the stiffness matrix mR L
khy [– ]
kmx [– ]
μ [– ]
ψ [–] parameters of the function F (t)
β [– ]
λh [– ]
parameters of the function g(t)
λm [– ]
κ [–] parameter of the loading function Y
mR [–] stiffness coefficient for η · η δ = −1
mT [–] stiffness coefficient for η · η δ = 0
R [L] size of the zone at high stiffness in the generalized
displacement space
βr [–] coefficient of the evolution law for δ
χ [–] coefficient of the interpolation law for K hp
Vf [F] bearing capacity under centered vertical load

Table 4.3. Description of the constants of the hypoplastic macroelement

Copyright Iste 2022 / File for personal use of Robert Charlier only
160 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 4.9. Viaduct scheme (out of scale; the dimensions reported

refer to the small-scale prototype)

4.5. Case study: seismic response of a reinforced concrete viaduct

The potentialities offered by the two macroelements considered in the analysis

of SSI phenomena are hereafter evaluated by simulating the seismic response of a
four-span viaduct founded on shallow foundations, resting on medium-dense sands.
In particular, the following three cases have been considered:
– soil–foundation system modeled by means of the elastoplastic macroelement
by [GRA 09] (case “MEP”);
– soil–foundation system modeled by means of the hypoplastic macroelement by
[SAL 09] (case “MHP”);
– foundation on infinitely rigid soil (case “RIG” – no SSI).

The goal of this simulation program is twofold: on the one hand, quantifying
the influence of the macroelement mathematical formulation on the prediction of
the global system performance by comparing MEP and MHP case solutions; on the
other hand, evaluating the effect of SSI on the distribution of structural stresses and
displacements by comparing the MEP and MHP case solutions with the one obtained
from the RIG case (in the absence of interaction).

4.5.1. Features of the viaduct

The considered structure is a four-span viaduct, with a pre-compressed reinforced

concrete deck and reinforced concrete piers with rectangular hollow sections
(Figure 4.9). The viaduct has been analyzed at the ELSA laboratory (JRC Ispra) using
small-scale physical models (1÷2.5) for the three piers and a finite element model for
the deck [PIN 96].

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 161

The geometric characteristics of the piers and the beam sections are provided
in Table 4.4. In the following, the viaduct is simulated by considering a length
of the deck of 80 m, a height of the lateral piers (P1 and P3) of 5.6 m, and a
height of the lateral pier (P2) of 8.4 m. The piers are arranged symmetrically with
respect to the center of the deck, which is considered bounded with a simple roller
constraint at the head of the piers.

A Ix Iy Iz J
(m2 ) (m4 ) (m4 ) (m4 ) (m4 )
Deck 1.11 0.13 – 2.26 2.39
Piers 0.66 0.056 0.19 – 0.20

Table 4.4. Geometric characteristics of the structural

elements of the viaduct

4.5.2. The finite element model of the viaduct and its foundations

For the numerical simulations, the finite element Matlab toolbox FEDEASLab
[FIL 04] has been used. The finite element model of the structure is shown in
Figure 4.10.

To simulate the behavior of the viaduct piers, subjected to bending and shear
stresses, nonlinear Timoshenko beam elements with multi-fiber formulation [KOT 05]
have been adopted. The mechanical behavior of the concrete fibers has been described
using the damage model by [LAB 91], while a modified version of the model
[MEN 73] has been adopted for the steel fibers of the reinforcing cage. For the deck,
elastic linear beam elements have been used.

Figure 4.10. Finite element model of the viaduct. The black circles
represent the concentrated masses

In the adopted finite element spatial discretization, six nonlinear multi-fiber

Timoshenko beam elements have been used for the P1 and P3 piers, while nine
nonlinear multi-fiber Timoshenko beam elements have been used for the central
highest pier. The discretization has been refined at the base of the piers where a higher

Copyright Iste 2022 / File for personal use of Robert Charlier only
162 Deterministic Numerical Modeling of Soil–Structure Interaction

concentration of stresses is expected. Some details of the multi-fiber discretization of

the sections of the nonlinear Timoshenko beams are illustrated in Figure 4.11. For
each section, 40 concrete fibers and 80 steel fibers, representing the reinforcing bars
in their actual position, have been used.

Figure 4.11. Details of the pier discretization

and the different cross-sections

The inertial characteristics of the structural elements have been simulated through
masses concentrated at the element nodes, as shown in Figure 4.10. The values of the
constants that characterize the materials of the different structural elements used in the
simulations are provided in [GRA 08a].

The two macroelements used to describe the soil–foundation system in MEP and
MHP analyses have been implemented in FEDEASLab using different integration
algorithms, according to the different nature of their constitutive equations. For the
elastoplastic macroelement by [GRA 09], a predictor–corrector strategy has been
adopted, based on the use of the implicit backward Euler algorithm for the projection
phase on the yield surface [GRA 08a]. For the hypoplastic macroelement by [SAL 09]
– which does not have a yield surface to impose the consistency condition – an explicit
Runge–Kutta–Fehlberg algorithm of third order with adaptive step and error check has
been adopted [STO 93].

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 163

In the formulation of the two macroelements, the introduction of dissipative

elements capable of reproducing the damping phenomenon by radiation has not been
provided. This introduces a simplification that may have an effect on the computed
response of the system, but this is accepted considering the observations by [JER 04]:
– radiation damping shows its effects primarily at high frequencies and low values
of the mechanical damping for hysteresis;
– if the foundation undergoes a partial detachment from the soil, the radiation
damping is greatly reduced;
– the simplified approaches used, for example, by [SPY 92] or [ZHA 02], which
use dampers with frequency-independent response, tend to produce a generalized
reduction of stresses in the structure.

In addition, studies conducted by Bielak et al. [BIE 03, YOS 03] indicate that,
in some cases, radiation damping can be ignored without significant variations in the
structure response.

4.5.3. Seismic input

The seismic input adopted in the finite element simulations is shown in Figure
4.12. This is an artificial accelerogram, applied along the x-direction, compatible with
the response spectrum provided by Eurocode 8 for a Class B soil and a damping of
5%, characterized by a peak acceleration of 0.35 g. To comply with the conditions
of similarity in the 1÷2.5 scale model used in the simulations, the accelerations have
been multiplied by 2.5, while the time scale was divided by the same coefficient.
The same seismic input has been applied to the abutment of the bridge and to the
foundations of the three piers, neglecting differences that may be due to topographic
amplification effects.

Figure 4.12. Scaled accelerogram applied to the foundations

and abutment of the bridge

Copyright Iste 2022 / File for personal use of Robert Charlier only
164 Deterministic Numerical Modeling of Soil–Structure Interaction

4.6. Calibration of the macroelements

The width Bx and the length By of the foundations have been assumed equal
to 2.1 m and 4.1 m, respectively. The procedures for determining the remaining 14
constants of the elastoplastic macroelement are described in detail in [GRA 08a]. For
the purposes of this chapter, the values reported in Table 4.5 have been adopted, as
they can be considered appropriate for foundations on medium-dense sands.

kv khx kmy khy kmx ξ κ

(–) (–) (–) (–) (–) (–) (–)
84.85 69.42 30.87 69.42 30.87 0.0 0.0

a b c d e f Vf
(–) (–) (–) (–) (–) (–) (MN)
0.48 0.33 1.00 0.95 1.00 0.95 11.26

Table 4.5. Constants of the elastoplastic macroelement adopted

in the numerical simulations

To compare the response provided by the two macroelements, a calibration of the

MHP has been conducted on the following stress paths:
– phase a): centered vertical load V monotonously increasing up to 1200 kN;
phase b): horizontal shift Ux monotonously increasing up to 0.02 m;
– phase a): centered vertical load V monotonously increasing up to 1200 kN;
phase b): rotation Θy monotonously increasing up to 0.004 rad;
– phase a): centered vertical load V monotonously increasing up to 1200 kN;
phase b): cyclic horizontal displacement:

Ux (t) = Ux,ref λ̂(t) 0≤t≤T

applied at a distance h = 0.775 m from the baricenter of the foundation. In the

previous expression, Ux,ref = 0.001 m, while the load multiplier λ̂(t) is represented in
Figure 4.13 as a function of the ratio t/T .

The constants that characterize the foundation size and its pseudoelastic behavior,
as well as the vertical load in ultimate conditions Vf , have the same meaning of the
corresponding constants of the elastoplastic model, and can therefore be assumed
identical to those found in Table 4.5.

By comparing the expressions provided by equations equation [4.29] and [4.31],

observing that for the elastoplastic macroelement, we assumed c = e = 1 and d = f ,

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 165

and choosing μ = a, ψ = b and β = d, the collapse domains of the two models are
coincident.

The numerical values of the remaining constants (λh , λm , κ, mR , mT , R,

βr and χ) have been determined by attempts, comparing the response of the two
models (MHP, MEP) along the loading paths (1)–(3). This comparison is graphically
illustrated in Figures 4.14–4.19. The final result of the calibration procedure of the
hypoplastic macroelement is summarized in Table 4.6.

Figure 4.13. Load multiplier adopted in the cyclic calibration test no. 3

μ ψ β λh λm κ
(–) (–) (–) (–) (–) (–)
0.48 0.33 0.95 1.75 1.50 0.25

mR mT R βr χ Vf
(–) (–) (mm) (–) (–) (MN)
1.1 1.05 5.0 1.0 1.5 11.26

Table 4.6. Constants of the hypoplastic macroelement adopted

in the numerical simulations

The figures show that the response of the two models in the three calibration tests
is generally very similar, even for cyclic load paths with variable amplitude, such as
the one used in test no. 3. The only significant difference concerns the evolution of the
vertical displacement calculated during the cyclic phase of this last test, qualitatively
similar but quantitatively different (Figure 4.19). Indeed, the maximum settlement
of the hypoplastic macroelement at the end of the test is greater than that of the
elastoplastic macroelement, of about 130%.

Copyright Iste 2022 / File for personal use of Robert Charlier only
166 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 4.14. Calibration test no. 1 – phase a): normalized vertical force
v versus normalized vertical displacement w. For a color version of this
figure, see www.iste.co.uk/grange/soil.zip

Figure 4.15. Calibration test no. 1 – phase a): normalized horizontal

force hx versus normalized horizontal displacement ux . For a color
version of this figure, see www.iste.co.uk/grange/soil.zip

Figure 4.16. Calibration test no. 2 – phase b): normalized moment

my versus normalized rotation θy . For a color version of this figure,
see www.iste.co.uk/grange/soil.zip

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 167

Figure 4.17. Calibration test no. 3 – phase b): normalized horizontal

force hx versus normalized horizontal displacement ux . For a color
version of this figure, see www.iste.co.uk/grange/soil.zip

4.7. Results of the numerical simulations

Several results of the numerical simulations on the behavior of the reinforced

concrete viaduct subjected to seismic action are shown in the following, with a
comparison of the predictions of both macroelements and a “reference” solution,
obtained by constraining displacements and rotations at the base of the piers (where
infinitely rigid soil is assumed), in which the effects of SSI are completely ignored.

4.7.1. Forces and displacements in the structural elements

Figures 4.20 and 4.21 show the trends of Hx : Ux and My : Θy curves for the
three pier foundations. Due to the symmetry of the problem, the responses of P1 and
P3 piers are identical. The figures show that the predictions of the two macroelements
are very similar, although the hypoplastic model provides slightly smaller maximum
horizontal displacements. It is interesting to observe that, at greater amplitude cycles,
the resistance (to sliding and rotation) of the foundations is fully mobilized, giving rise
to irreversible displacements and rotations, associated with relevant hysteresis cycles
with related energy dissipation.

Copyright Iste 2022 / File for personal use of Robert Charlier only
168 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 4.18. Calibration test no. 3 – phase b): normalized moment my

versus normalized rotation θy . For a color version of this figure, see
www.iste.co.uk/grange/soil.zip

Figure 4.19. Calibration test no. 3 – phase b): normalized vertical

displacement v versus time factor t/T . For a color version of this
figure, see www.iste.co.uk/grange/soil.zip

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 169

Figure 4.20. Horizontal force versus horizontal displacement at the

foundations: a) P1 and P3 piers and b) P2 pier. For a color version
of this figure, see www.iste.co.uk/grange/soil.zip

The time evolution of the shear force and bending moment at the head of
the pier foundations is shown in Figures 4.22 and 4.23 respectively. For the two
macroelements, the minimum and maximum values of the shear force Hx and the
bending moment My have an upper limit represented by the ultimate conditions for
sliding and rotation of the foundation; this is not the case for an infinitely rigid soil.
While the predictions obtained using the two macroelements are very similar, the
reactions calculated in the hypothesis of an infinitely rigid soil are significantly higher,
with increments in peak values of about 100%, both in terms of shear forces and
bending moment. This result is in agreement with what was observed by, for example,
[WOT 10] and [ALG 10] and depends both on the increase in the fundamental
period of the structure associated with the deformability of the soil–foundation
system and on the energy dissipation due to the hysteretic behavior of the two
macroelements.

Copyright Iste 2022 / File for personal use of Robert Charlier only
170 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 4.21. Bending moment versus rotation at the foundations:

a) P1 and P3 piers and b) P2 pier. For a color version of this figure,
see www.iste.co.uk/grange/soil.zip

4.7.2. Displacements of the abutment and the foundations

The horizontal displacements at the head of the two piers are shown in Figure 4.24.
As seen previously, the solutions obtained with the two macroelements are almost
identical. Conversely, the displacements computed for the case of infinitely rigid soil
are characterized by a significantly different time evolution.

Contrary to the forces, the peak values of the displacements at the center of the
deck – computed considering SSI – are only slightly higher than those obtained in the
case of rigid soil. This is due to the fact that, in the latter case, the forces induced
by the seismic action applied at the base of the piers are high enough to cause
yielding of the pier itself, as shown by the bending moment versus curvature diagram
(Figure 4.25). In the analyses conducted considering SSI, the pier remains in the
elastic domain for the entire duration of the seismic event.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 171

Figure 4.22. Time evolution of the shear force at the head of the
foundations: a) P1 and P3 piers and b) P2 pier. For a color version
of this figure, see www.iste.co.uk/grange/soil.zip

The settlements accumulated by the three foundations during the seismic event are
shown in Figure 4.26. The responses provided by the two macroelements show some
differences. While the simulation with the elastoplastic macroelement provides very
small vertical displacement values (with a maximum of 1 mm at the end of the seismic
event) and monotonous increase in time, the settlements accumulated in the simulation
with the hypoplastic macroelement are significantly higher (with a maximum of about
6 mm for the P2 pier), although they remain somewhat limited in absolute terms.
This difference is likely due to the kinematic feature of the surface defined by the
plastic potential function in the model by [GRA 09], whereas in the hypoplastic model,
the irreversible deformations are determined from the plastic potential function by
[NOV 91], characterized by an isotropic evolution.

Copyright Iste 2022 / File for personal use of Robert Charlier only
172 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 4.23. Time evolution of the bending moment at the head of the
foundations: a) P1 and P3 piers and b) P2 pier. For a color version of
this figure, see www.iste.co.uk/grange/soil.zip

4.8. Concluding remarks

Two macroelements for shallow foundations subjected to cyclic/dynamic loads,

recently developed in the context of the theories of hardening plasticity and
hypoplasticity, have been used to evaluate the effects of soil–structure interaction on
a pre-compressed reinforced concrete viaduct in seismic conditions.

The simulations carried out using the two macroelements provided very similar
results, both in terms of horizontal forces/moments and horizontal displacements/
rotations, despite the different mathematical formulations of the implemented
constitutive equations. This is probably due to the presence, in both formulations, of an
appropriate internal vectorial variable (the internal force τ which in the elastoplastic
model assumes the role of “back stress” and the internal displacement δ in the
hypoplastic model) which keeps the memory of the previous loading history to allow
a description of the hysteretic behavior of the soil–foundation system.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 173

Figure 4.24. Time evolution of the horizontal displacement of the pier

heads: a) P1 and P3 piers and b) P2 pier. For a color version of this
figure, see www.iste.co.uk/grange/soil.zip

The only significant difference is on the magnitude of the settlements accumulated

by the foundations under the effect of cyclic loading, which is significantly greater
for the case of the hypoplastic macroelement. This is probably due to the different
functions of the two macroelements that define the direction of the irreversible
component of the velocity (u̇p or m); further studies are, however, currently underway
to assess this hypothesis.

The comparison between the macroelement results and those obtained considering
an infinitely rigid soil allows the quantitative evaluation of the influence of SSI. For
the considered case study, the effect of SSI is beneficial in terms of shear forces
and bending moments, which decrease about 50% compared to the rigid soil case.
Furthermore, the increase in the deformability of the system does not have a significant
impact on the deck displacements. The reason is that the higher forces obtained in the
case of rigid soil cause yielding at the base of the piers, resulting in a significant
reduction of the structural stiffness.

Copyright Iste 2022 / File for personal use of Robert Charlier only
174 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 4.25. Bending moment versus curvature: a) P1 and P3 piers

and b) P2 pier. For a color version of this figure, see
www.iste.co.uk/grange/soil.zip

As observed by [MYL 00], considering soil deformability in the soil–foundation

system can lead either to a reduction or to an increase in the computed forces on the
structure, depending on the characteristics of the response spectrum of the considered
seismic input. Ignoring the effect of SSI in the design stage can therefore lead to
inaccurate assessment of the system behavior in a non-conservative side.

The chosen case study illustrates that the two macroelements represent a signi-
ficant step forward in the development of simple, robust and accurate calculation
tools for the quantitative evaluation of the effects of soil–structure interaction in
seismic conditions, not only for research purposes but also for structural design.
The macroelement concept combines the computational advantages typical of
conventional approaches to the ability of reproducing the behavior of soil–foundation
systems on complex stress paths, comparable only to that obtained using full 3D finite
element models, whose computational cost is prohibitive in everyday practice.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 175

Figure 4.26. Time evolution of the vertical displacements of the

foundations: a) P1 and P3 piers and b) P2 pier. For a color version of
this figure, see www.iste.co.uk/grange/soil.zip

4.9. References

[ALG 10] A LGIE T.B., P ENDER M.J., O RENSE R.P., W OTHERSPOON L.M., “Dynamic
field testing of shallow foundations subject to rocking”, 2010 New Zealand Society for
Earthquake Engineering Conference, Wellington, New Zealand, 26–28 March, 2010.
[BIE 03] B IELAK J., L OUKAKIS K., H ISADA Y., YOSHIMURA C., “Domain reduction
method for three-dimensional earthquake modeling in localized regions. Part I: Theory”,
Bulletin of the Seismological Society of America, vol. 93, pp. 817–824, 2003.
[BIE 06] B IENEN B., B YRNE B.W., H OULSBY G.T., C ASSIDY M.J., “Investigating six-
degree-of-freedom loading of shallow foundations on sand”, Géotechnique, vol. 56,
pp. 367–379, 2006.
[BUT 94] B UTTERFIELD R., G OTTARDI G., “A complete three-dimensional failure envelope
for shallow footings on sand”, Géotechnique, vol. 44, no. 1, pp. 181–184, 1994.
[CAQ 66] C AQUOT A., K ÉRISEL J., Traité de mécanique des sols, Gautiers-Villars, Paris,
1966.

Copyright Iste 2022 / File for personal use of Robert Charlier only
176 Deterministic Numerical Modeling of Soil–Structure Interaction

[CAS 02] C ASSIDY M., B YRNE B., H OULSBY G., “Modelling the behaviour of circular
footings under combined loading on loose carbonate sand”, Géotechnique, vol. 52, no. 10,
pp. 705–712, 2002.
[CAS 04] C ASSIDY M., M ARTIN C., H OULSBY G., “Development and application of force
resultant models describing jack-up foundation behaviour”, Marine Structures, vol. 17,
pp. 165–193, 2004.
[CHA 07] C HATZIGOGOS C., Comportement sismique des fondations superficielles : vers
la prise en compte d’un critère de performance dans la conception, PhD thesis, Ecole
Polytechnique, 2007.
[CHA 08] C HATZIGOGOS C.T., P ECKER A., S ALENCON J., “Macroelement modeling of
shallow foundations”, Soil Dynamics and Earthquake Engineering, vol. 29, pp. 765–781,
2008.
[CHO 19] C HOINIÈRE M., PAULTRE P., L ÉGER P., “Influence of soil-structure interaction
on seismic demands in shear wall building gravity load frames”, Engineering Structures,
vol. 198, pp. 109–259, 2019.
[CIA 95] C IAMPOLI M., P INTO P.E., “Effects of soil–structure interaction on inelastic seismic
response of bridge piers”, ASCE Journal of Structural Engineering, vol. 121, pp. 806–814,
1995.
[CRÉ 01] C RÉMER C., P ECKER A., DAVENNE L., “Cyclic macro–element for soil–structure
interaction: Material and geometric non–linearities”, International Journal for Numerical
and Analytical Methods in Geomechanics, vol. 25, pp. 1257–1284, 2001.
[CRÉ 02] C RÉMER C., P ECKER A., DAVENNE L., “Modelling of nonlinear dynamic
behaviour of a shallow strip foundation with macro-element”, Journal of Earthquake
Engineering, vol. 6, pp. 175–211, 2002.
[DAR 78] DARVE F., Une formulation incrémentale des lois rhéologiques. Application aux
sols, PhD thesis, INP Grenoble, 1978.
[DAV 73] DAVIS E., B OOKER J., “The effect of increasing strength with depth on the bearing
capacity of clays”, Géotechnique, vol. 23, no. 4, pp. 551–563, 1973.
[DAV 98] DAVENNE L., B RENET C., Macro-éléments de poutres en béton armé, Report, LMT
Cachan, no. 210, 1998.
[ELA 92] E LACHACHI S., Sur l’élaboration d’une méthode simplifiée d’analyse des
structures de Génie Civil par macro-éléments adaptés aux constructions composites et
endommageables, PhD thesis, Université Paris VI, 1992.
[ELG 08] E LGAMAL A., YAN L., YANG Z., C ONTE J.P., “Three–dimensional seismic
response of Humboldt Bay bridge-foundation-ground system”, ASCE Journal of Structural
Engineering, vol. 134, pp. 1165–1176, 2008.
[ELN 96] E LNASHAI A.S., M C C LURE D.C., “Effect of modelling assumptions and input
motion characteristics on seismic design parameters of RC bridge piers foundations”,
Earthquake Engineering and Structural Dynamics, vol. 25, pp. 435–463, 1996.
[EUR 04] E UROPEAN C OMMITTEE FOR S TANDARDIZATION, Eurocode 8: Design of
structures for earthquake resistance, European standard, EN 1998-1, 2004.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 177

[FAR 15] FARDIS M.N., C ARVALHO E.C., FAJFAR P., P ECKER A., Seismic Design of
Concrete Buildings to Eurocode 8, CRC Press, Boca Raton, 2015.
[FIL 04] F ILIPPOU F.C., C ONSTANTINIDES M., FedeasLab Getting Started Guide and
Simulations Examples, Department of Civil & Environmental Engineering, University of
California, Berkeley, 2004.
[GAZ 98] G AZETAS G., M YLONAKIS G., “Seismic soil–structure interaction: New evidence
and emerging issues”, in DAKOUKAS P., Y EGIAN M.K., H OLTZ R.D. (eds), Geotechnical
Earthquake Engineering and Soil Dynamics III, ASCE, Reston, 1998.
[GOT 99] G OTTARDI G., H OULSBY G.T., B UTTERFIELD R., “The plastic response of
circular footings on sand under general planar loading”, Géotechnique, vol. 49, pp. 453–
470, 1999.
[GRA 08a] G RANGE S., Modélisation simplifiée 3D de l’interaction sol-structure: application
au génie parasismique, PhD thesis, INP Grenoble, 2008. Available at: http://tel.archives-
ouvertes.fr/tel-00306842/fr.
[GRA 08b] G RANGE S., KOTRONIS P., M AZARS J., “A macro-element for a circular
foundation to simulate 3D soil–structure interaction”, International Journal for Numerical
and Analytical Methods in Geomechanics, vol. 32, pp. 1205–1227, 2008.
[GRA 09] G RANGE S., KOTRONIS P., M AZARS J., “A macro-element to simulate 3D soil–
structure interaction considering plasticity and uplift”, International Journal of Solids and
Structures, vol. 46, pp. 3651–3663, 2009.
[HOU 02] H OULSBY G.T., C ASSIDY M.J., “A plasticity model for the behaviour of footings
on sand under combined loading”, Géotechnique, vol. 52, pp. 117–129, 2002.
[JER 04] J EREMI Ć B., K UNNATH S., X IONG F., “Influence of soil–foundation–structure
interaction on seismic response of the I-880 viaduct”, International Journal for Engineering
Structures, vol. 26, no. 3, pp. 391–402, 2004.
[JER 09] J EREMI Ć B., J IE G., P REISIG M., TAFAZZOLI N., “Time domain simulation of soil–
foundation–structure interaction in non-uniform soils”, Earthquake Engineering Structural
Dynamics, vol. 38, no. 5, pp. 699–718, 2009.
[JIN 19a] J IN Z., Y IN Z.-Y., KOTRONIS P., L I Z., “Advanced numerical modelling of
caisson foundations in sand to investigate the failure envelope in the HMV space”, Ocean
Engineering, vol. 190, pp. 106–394, 2019.
[JIN 19b] J IN Z., Y IN Z.-Y., KOTRONIS P., L I Z., TAMAGNINI C., “A hypoplastic
macroelement model for a caisson foundation in sand under monotonic and cyclic
loadings”, Marine Structures, vol. 66, pp. 16–26, 2019.
[KOL 91] KOLYMBAS D., “An outline of hypoplasticity”, Archive of Applied Mechanics,
vol. 61, pp. 143–151, 1991.
[KOT 05] KOTRONIS P., M AZARS J., “Simplified modelling strategies to simulate the
dynamic behaviour of R/C walls.”, Journal of Earthquake Engineering, vol. 9, no. 2,
pp. 285–306, 2005.

Copyright Iste 2022 / File for personal use of Robert Charlier only
178 Deterministic Numerical Modeling of Soil–Structure Interaction

[KOT 13] KOTRONIS P., TAMAGNINI C., G RANGE S. (eds), Soil-Structure Interaction,
Alert Doctoral School 2013, Aussois, France, 3–5 October, 2013. Available at: http://
alertgeomaterials.eu/publications.
[LAB 91] L A B ORDERIE C., Phénomènes unilatérales dans un matériau endommageable:
modélisation et application à l’analyse des structures en béton, PhD thesis, Université Paris
VI, 1991.
[LEP 01] L E PAPE Y., S IEFFERT J.-G., “Application of thermodynamics to the global
modelling of shallow foundations on frictional material”, International Journal for
Numerical and Analytical Methods in Geomechanics, vol. 25, pp. 1377–1408, 2001.
[LI 14] L I Z., KOTRONIS P., E SCOFFIER S., “Numerical study of the 3D failure envelope of a
single pile in sand”, Computers and Geotechnics, vol. 62, pp. 11–26, 2014.
[LI 16] L I Z., KOTRONIS P., E SCOFFIER S., TAMAGNINI C., “A hypoplastic macroelement
for single vertical piles in sand subject to three-dimensional loading conditions”, Acta
Geotechnica, vol. 11, no. 2, pp. 373–390, 2016.
[LI 18] L I Z., KOTRONIS P., E SCOFFIER S., TAMAGNINI C., “A hypoplastic macroelement
formulation for single batter piles in sand”, International Journal for Numerical and
Analytical Methods in Geomechanics, vol. 42, no. 12, pp. 1346–1365, 2018.
[MAR 94] M ARTIN C., Physical and numerical modelling of offshore foundations under
combined loads, Master’s thesis, DPhil thesis, University of Oxford, 1994.
[MAR 01] M ARTIN C.M., H OULSBY G.T., “Combined loading of spudcan foundations on
clay: Numerical modelling”, Géotechnique, vol. 51, pp. 687–700, 2001.
[MAT 79] M ATAR M., S ALENÇON J., “Capacité portante des semelles filantes”, Revue
Française de Géotechnique, vol. 9, pp. 51–76, 1979.
[MEN 73] M ENEGOTTO M., P INTO P.E., “Method of analysis of cyclically loaded reinforced
concrete plane frames including changes in geometry and non-elastic behaviour of elements
under combined normal force and bending”, IABSE Symposium on Resistance and Ultimate
Deformability of Structures Acted on by Well-Defined Repeated Loads, Lisbon, Portugal,
1973.
[MON 97] M ONTRASIO L., N OVA R., “Settlements of shallow foundations on sand:
Geometrical effects”, Géotechnique, vol. 47, pp. 49–60, 1997.
[MUR 84] M URCHISON J.M., O’N EILL M.W., “Evaluation of P-Y relationships in
cohesionless soils”, Analysis and Design of Pile Foundations, ASCE, pp. 174–191, 1984.
[MYL 00] M YLONAKIS G., G AZETAS G., “Seismic soil–structure interaction: Beneficial or
detrimental?”, Journal of Earthquake Engineering, vol. 4, pp. 227–301, 2000.
[NIE 97] N IEMUNIS A., H ERLE I., “Hypoplastic model for cohesionless soils with elastic
strain range”, Mechanics of Cohesive-frictional Materials, vol. 2, pp. 279–299, 1997.
[NIE 02] N IEMUNIS A., Extended hypoplastic models for soils, Habilitation thesis, Bochum
University, 2002.
[NOV 91] N OVA R., M ONTRASIO L., “Settlements of shallow foundations on sand”,
Géotechnique, vol. 41, pp. 243–256, 1991.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Macroelements for Soil–Structure Interaction 179

[PEC 97] P ECKER A., “Analytical formulae for the seismic bearing capacity of shallow strip
foundations”, in Seismic Behaviour of Ground and Geotechnical Structures, S ECO E P INTO
P.S. (ed.), Taylor & Francis Group, London, 1997.
[PER 20] P EREZ -H ERREROS J., Interaction dynamique sol-structure des fondations sur
pieux : étude expérimentale et numérique, PhD thesis, Ecole Centrale de Nantes, 2020.
[PHI 03] P HILIPPONNAT G., H UBERT B., Fondations et ouvrages en terre, Eyrolles, Paris,
2003.
[PIN 96] P INTO A.V., V ERZELETTI G., P EGON P., M AGONETTE G., N EGRO P., G UEDES J.,
Pseudo-dynamic testing of large-scale R/C bridges, Research report EUR 16378 EN, Joint
Research Centre, Ispra, 1996.
[PRI 03] DI P RISCO C., N OVA R., S IBILIA A., “Shallow footing under cyclic loading:
Experimental behaviour and constitutive modelling”, in M AUGERI M., N OVA R. (eds),
Geotechnical Analysis of the Seismic Vulnerability of Historical Monuments, pp. 99–121,
Pàtron Editore, Bologna, 2003.
[PRI 06] DI P RISCO C., M ASSIMINO M.R., M AUGERI M., N ICOLOSI M., N OVA R.,
“Cyclic numerical analysis of Noto Cathedral: Soil–structure interaction modelling”,
Rivista Italiana di Geotecnica, vol. 48, 2006.
[RAN 04] R ANDOLPH M., JAMIOLKOWSKI M., Z DRAVKOVI L., “Load carrying capacity
of foundations”, Proceedings of the Skempton Memorial Conference on Advances in
Geotechnical Engineering, vol. 1, pp. 207–240, London, United Kingdom, 2004.
[SAL 09] S ALCIARINI D., TAMAGNINI C., “A hypoplastic macroelement model for shallow
foundations under monotonic and cyclic loads”, Acta Geotechnica, vol. 4, no. 3, pp. 163–
176, 2009.
[SPY 92] S PYRAKOS C.C., “Seismic behavior of bridge piers including soil–structure
interaction”, Computers & Structures, vol. 43, no. 2, pp. 373–384, 1992.
[STO 93] S TOER J., B ULIRSCH R., Introduction to Numerical Analysis, 2nd edition, Springer
Verlag, New York, 1993.
[WOL 96] VON W OLFFERSDORFF P.A., “A hypoplastic relation for granular materials with a
predefined limit state surface”, Mechanics of Cohesive-frictional Materials, vol. 1, pp. 251–
271, 1996.
[WOT 10] W OTHERSPOON L.M., P ENDER M.J., “Effect of shallow foundation modeling on
seismic response of moment frame structures”, in O RENSE R.P., C HOUW N., P ENDER
M.J. (eds), International Workshop on Soil–Foundation–Structure Interaction, CRC Press,
Boca Raton, 2010.
[YOS 03] YOSHIMURA C., B IELAK J., H ISADA Y., “Domain reduction method for three–
dimensional earthquake modeling in localized regions. Part II: Verification and examples”,
Bulletin of the Seismological Society of America, vol. 93, pp. 825–840, 2003.
[ZHA 02] Z HANG J., M AKRIS N., “Seismic response analysis of highway overcrossing
including soil–structure interaction”, Earthquake Engineering and Structural Dynamics,
vol. 31, pp. 1967–1991, 2002.

Copyright Iste 2022 / File for personal use of Robert Charlier only
Copyright Iste 2022 / File for personal use of Robert Charlier only
 5

Urban Seismology: Experimental

Approach to Soil–Structure Interaction
Towards the Concept of Meta-city

5.1. Introduction

Since the beginning of the 19th Century, a large number of authors have been
interested in the phenomenon of coupling between soil and foundations under dynamic
loading. They addressed these processes by means of empirical, and later analytical,
approaches [HAD 93]: for example, Lamb’s works in 1904 [LAM 04], which studied
the vibrations of a semi-infinite linear elastic medium; Reissner’s works in 1936
[REI 36], which highlighted the phenomenon of dissipation of vibration energy in soil
in the form of waves; or those of Hsieh in 1962 [HSI 62] and especially Lysmer in 1965
[LYS 65], which for the first time introduced the idea of replacing the soil–foundation
system by a simple oscillator, whose stiffness and damping values are constant with
frequency.

With the advent of civil nuclear power at the end of the 1950s, the analysis and
understanding of SSI took on another dimension and led to a new approach. Unlike the
first analyses focusing on vibrating machines, seismic interaction considers harmonic
loading that originates in soil. While for light structures, SSI appeared at first glance
to be negligible, rigid or massive structures have shown important interactions, in
particular, during deformations in the horizontal directions of structures, and no longer
only in the vertical direction.

In the case of a structure based on rocky soil, in other words considered infinitely
rigid, the loading wavelength is very large compared to the size of the foundation,

Chapter written by Philippe G UÉGUEN, Philippe ROUX and Andrea C OLOMBI.

Copyright Iste 2022 / File for personal use of Robert Charlier only
182 Deterministic Numerical Modeling of Soil–Structure Interaction

such that motion at depths, at the soil–foundation contact level, and at the surface
are in phase. The motion at the base will therefore directly refer to the center of
mass in the form of forces of inertia, causing the deformation of the structure. This
deformation will naturally lead to the appearance of a bending moment and a shear
force that represent the reaction of the non-deformable rock. No soil deformation is to
be expected at the base of the foundation and free-field motion is not disturbed by the
presence of the foundation. It will then be said that the structure is fixed or rigid.

In contrast, we will refer to a flexible-base structure when the latter has its
foundations in soil, soft enough to allow a relative shift of the foundation with
respect to soil. This situation is comparable to the case where heterogeneities can
cause perturbations in the wave field, from the moment their impedance contrast
with soil becomes significant. For example, irregularities located at the level of
soil-substratum contact can cause interactions between surface waves and volume
waves [HIL 84, LEV 85, CHA 90], as well as at the soil surface [WIR 89, CHA 90].
Since the foundation can be considered to be rigid compared to the soil, the kinematic
compatibility between soil and foundation deformation will induce a phase shift
between incident and diffracted motion, all the more significant as the wavelength
of the incident field will be at least of the same order of magnitude as the size of the
foundation [KAU 10]. The interaction introduces scattering and resonance phenomena
along the length of the foundations [STE 98], which can pollute the incident wave
field. This effect is exacerbated by the presence of a strong impedance contrast
between soil and foundations [TRI 72, WON 77, MOS 87].

On the contrary, the inertia developed in the structure causes translation and
rocking motion of the foundation (and therefore of the structure) relative to the soil.
Given that soil is flexible, it will deform under the action of the efforts transmitted by
the foundation. As a result, an additional deformation field will propagate into the soil
in the form of seismic waves [JEN 70, KAN 71, GUÉ 00, COR 04, KIM 01].

This shows that significant couplings occur between soil and structure, as soon
as they are subjected to seismic load. Because of the kinematic or inertial effects,
the incident wave field is perturbed and moves away, with these characteristics, from
the seismic motion that would be observed over a space free of any constructions. The
diffracted wave field at the foundation level, the seismic motion caused by structure
vibration, and the interactions between structures, suggest that the seismology of
farmlands appears very different from the seismology of cities (or urban seismology).

In this chapter, we will illustrate with observations, the phenomenon of coupling

that may exist between soil and structure and its consequences on the dynamic
response of the soil–structure system and the nearby seismic ground motion. First,
we will show some examples of meaningful interactions. Then, we will describe the
interactions observed between structures under dynamic loading. Finally, we will

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Urban Seismology 183

illustrate at the city scale the generalization of site–city effect by introducing the
concept of metamaterials applied to urban seismology.

5.1.1. Observation of soil–structure interaction under weak and strong

seismic loading

5.1.1.1. Analytic model for a 1-DOF system

Among the simplified methods, Richart and Whitman [RIC 67] in 1967 extended
Hsieh’s [HSI 62] and Lysmer’s works [LYS 65] to structural vibrational motion in
general. Concerning the translation modes in all three directions and the rotation
modes around the three main axes, and following the same philosophy as for the
macroelements of Chapter 4, they proposed static stiffness and damping formulas.
A pivotal period in the evaluation of SSI corresponds to the publication, at about the
same time, of two articles, [LUC 71, VEL 71], which expressed the solution to the
dynamic response of a circular foundation based on an infinite and elastic half-space,
subjected to an external harmonic force. They established this solution for a massless
foundation, using an expression that connects displacements to efforts, for vertical,
horizontal, flexion and torsion harmonic forces: the impedance functions K are the
functions that depend on elastic parameters of soil and on the geometric dimensions
of the foundation.

Different methods for calculating impedance functions are presented in

[GAZ 91]. They are grouped into four categories: analytical or semi-analytical
methods, dynamic methods by numerical methods, hybrid methods combining
analytical and numerical approaches, and approximated methods that make it possible
to simply solve singularities that cannot be rigorously obtained. A wealth of literature
is available regardless of the types of soil and foundation being considered e.g.
[VEL 71, LUC 71, KAU 74, LUC 74, GAZ 83, APS 87]. Another approach consists
of defining analytical cone models that represent the impedance conditions of a
foundation on soft soil [WOL 85].

A 1D model is commonly used to represent the response of the soil–structure

system [LUC 88], (Figure 5.5). Under seismic loading, the total horizontal movement
of the system xtot is given by

xtot = x1 + x0 + Hφ + xg [5.1]

where xg is the soil displacement, x0 is the horizontal movement of the foundation

relative to the soil, Hφ is the horizontal displacement at the top produced by the
rocking motion of the foundation around the axis perpendicular to the horizontal
direction, with H being the height of the structure and x1 being the bending motion of
the fixed-base structure. In this equation, the vertical displacement of the structure is
overlooked, and a uniform rocking motion is assumed over the height of the structure.

Copyright Iste 2022 / File for personal use of Robert Charlier only
184 Deterministic Numerical Modeling of Soil–Structure Interaction

By neglecting the coupling between the translation motion and rocking, and assuming
that the soil does not have rotation motion, the system response is expressed by
[GUÉ 05]:

m1 (ẍ1 + ẍ0 + H φ̈y ) + c1 ẋ1 + k1 x1 = −m1 ẍg

m0 ẍ0 + ch ẋ0 + kh x0 + c1 ẋ1 − k1 x1 = −m0 ẍg [5.2]
J0 φ̈y + cr φ̇y + kr φy − Hc1 ẋ1 − Hk1 x1 = 0

with m1 , m0 and J0 being the masses of the structure and the foundation, and the
moment of rotational inertia of the foundation; c1 and k1 being the viscous damping
and the rigidity of the fixed-base structure; ch , kh , cr and kr being the damping
and stiffness coefficients of the impedance functions for translation h and rocking r
movements.

A conventional approach for identifying the response of a soil–structure system is

to determine from equation [5.2] the resonance frequencies and dampings associated
with the vibration modes of the system. Stewart and Fenves [STE 98] summarized
the input/output pairs of the system (Table 5.1) that ultimately represent the
information required for its experimental identification: we need recordings of free-
field motion, of foundation motion and those at the top of the construction to enable
the components of the motion of the system to be directly obtained, similarly to
[GUÉ 05].

System Input Output

Flexible base xg xg + x0 + Hφ + x1
Pseudo-flexible base xg + x0 xg + x0 + Hφ + x1
Fixed base xg + x0 + Hφ xg + x0 + Hφ + x1

Table 5.1. Components of the motion of the soil–structure system,

measured for experimental evaluation, according to [STE 98]

The effect of the stiffness contrast between soil and structure is illustrated with
a simple analytical model, which solves analytical equation [5.2] by considering the
expression of impedance functions for a superficial foundation resting on different
types of soil (Figure 5.1). The results illustrate observations made on numerous real
buildings [STE 99], which showed that in most cases the natural frequencies of
structures based on soft soil (flexible base) are lower than in those on rigid soil (fixed
base). In some cases, it is also found that the frequency of the soil–structure system is
located below the maximum energy of seismic load, resulting in a decrease in efforts
in the structure. Finally, the efforts at the base of the foundation are exacerbated in the
case of rigid soil.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Urban Seismology 185

In this model, the kinematic interaction is considered as being a second-order

effect. It will only have consequences for exceptional structures with long, deeply
buried foundations. We also find that the damping of a structure increases when it is
on a soft soil. In the case of rigid soil, damping is caused by the loss of energy during
cycles due to internal friction and nonlinearity. On the contrary, in the case of soft
soil, some of the vibration energy is dissipated in the form of seismic waves emitted
in soil at the soil–foundation contact level, being the result of inertial interaction. This
additional damping has a positive effect on the safety of the structure, by reducing the
maximal amplitude of deformations and the duration of oscillations.

5.1.1.2. Frequency analysis of motion components of the soil–structure

system
Based on equation [5.1], Stewart and Fenves [STE 98] have interpreted the
soil–structure system response, considering several configurations. If the structure
and soil are flexible, the fundamental frequency of vibrations observed at the top
of the building corresponds to the response of the system fsys . This frequency is
different from the frequency of the fixed-base structure f1 , which corresponds to a
flexible structure on rigid soil. On the rigid soil, the fixed-base structure and the system
vibrate at the same frequency, namely fsys = f1 . In this case, the soil–foundation
anchoring condition is perfect, and soil and foundations move in the same way. Luco
et al. [LUC 88] then proposed approximate equations linking the components of the
soil–structure system, i.e.
1 1 1 1
2
= 2+ 2+ 2 [5.3]
fsys f1 fφ f0

with f0 and fφ being the frequencies of the horizontal translation motion and
foundation rocking in relation to soil. Equation [5.3] indicates that the system
parameters depend on the fixed parameters and on the impedance of the foundation
in translation and rocking. This equation also implies that if the building is very rigid
(f1 = ∞), the system frequency fsys is equivalent to the frequency of the rigid body.
If soil is very rigid, fsys is thus equal to the fixed base frequency f1 . An exceptional
situation also exists for a partial flexibility condition of the foundation, representing
rocking only, i.e. assuming very rigid foundations in the horizontal direction. This
so-called apparent frequency fapp (or pseudo-flexible system) is given by

1 1 1
2
= 2 + 2 [5.4]
fapp f1 fφ

The parameters of the flexible-base structure can thus be approached by those of

the pseudo-flexible model. According to this model, and in the case of a rigid building
on soft soil, f1 is always larger than fsys . This implies that the ratio of structure
rigidity to soil is the key factor controlling the lengthening of the frequency [BIE 75,
AVI 96].

Copyright Iste 2022 / File for personal use of Robert Charlier only
186 Deterministic Numerical Modeling of Soil–Structure Interaction

Flexion X Translation X0
15 1

0.8
Amplitude

10
0.6

0.4
5
0.2

0 0
0 0.5 1 1.5 0 0.5 1 1.5

Rocking Phi Total diplacement X0+X+H.Phi

0.1 15

0.08
Amplitude

10
0.06

0.04
5
0.02

0 0
0 0.5 1 1.5 0 0.5 1 1.5

9 Shear base force F 10 Rocking momentum M

x 10 x 10
2 6

1.5
Amplitude

4
1
2
0.5

0 0
0 0.5 1 1.5 0 0.5 1 1.5
Frequency
Hz Frequency
Hz

Figure 5.1. Example of the effect of soil–structure interaction on the response of a

structure with a vibration frequency of 0.75Hz, according to the shear wave velocity
of the soil. The motions and the efforts correspond to those of a model with one
degree of freedom [GUÉ 05], including a structure, foundation and soil described by
equation [5.2]. Efforts are obtained directly from the classic relation connecting them
to foundation displacements via impedance functions. Black continuous line: Vs=100
m/s; black dashes: 150 m/s; black bold: 200 m/s; thin continuous red: 400 m/s; red
dashes: 600 m/s; red continuous bold: 800 m/s; blue: 1000 m/s. For a color version of
this figure, see www.iste.co.uk/grange/soil.zip

By analyzing the variations in frequencies (in particular, f1 and fsys ), the

contribution of SSI can be observed on the system response, as well as nonlinear
phenomena that appear according to loading levels. For example, Trifunac et al.
[TRI 10] observed, on a reinforced concrete building, a stable frequency f1 for
moderate loading levels, while fsys varied from about 10-15%, indicating a strong
sensitivity of the interaction with the loading level. This observation is also reported

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Urban Seismology 187

by Michel et al. [MIC 10b] with the Grenoble City Hall, France, where the presence
of SSI is identified by analyzing the difference between f1 obtained under weak
earthquakes and fsys obtained by processing recordings of ambient vibrations.
However, by compiling data from different earthquakes of varying significance,
classified according to the maximal structural drift produced between the top and
base of the building, and unlike Trifunac et al. [TRI 10], Michel et al. [MIC 10b]
observed for deformations limited to 10−5 variations in frequency f1 of the order
of 2%, indicating a structural nonlinear effect under weak loading. This variation is
due to the opening and closing during loading of pre-existing cracks, which locally
alter stiffness, as other authors also reported [CLI 06, MIC 10a, GUÉ 16b]. Still
others have also suggested elastic nonlinear (ENL) behaviors to explain variations
in elastic properties under low deformation levels. Characterized at the laboratory
scale with rock samples [GUY 99, JOH 05], this ENL behavior is the signature of
a heterogeneity rate existing in the system under test (in this case, different rock
samples). In a California building, this ENL behavior has recently been demonstrated
through the analysis of variations in elastic properties (frequency and damping) over
time [GUÉ 16b]. This observation reflects physical processes that are currently poorly
constrained, occurring in soil, in the structure or at the interface, which must be better
defined to interpret these variations, in terms of structural health.

Todorovska et al. [TOD 08] analyzed the ratio of stiffnesses f1 and fsys under
earthquakes of stronger magnitude (e.g. San Fernando or Northridge) observed in
the Van Nuys building in Los Angeles. Their analysis showed that during the San
Fernando earthquake, f1 decreased by about 40% and by about 22% during the
Northridge earthquake; at the same time, they also found that, although the first
frequency of the system fsys was always smaller than f1 , the difference between
these two characteristic frequencies varied, depending on the magnitude of the seismic
loading. Nonetheless, the model under consideration for extracting the components of
the SSI was linear, reflecting a particular behavior at the soil–structure interface.

In this study, they applied an alternative approach to the analysis of the system
response in the time domain, based on the analysis of the propagation of deformation
[ŞAF 98, ŞAF 99, SNI 06]. By calling it Seismic Interferometry by Deconvolution
(SIbyD), Snieder and Safak [SNI 06] considered the structure as a one dimensional
layer, in which a pulse wave obtained by deconvolution is propagated. In other words,
it is tantamount to performing the deconvolution of the top signal by the bottom signal.
This results in the structural impulse response, i.e. its frequency transfer function.
Mehta et al. [MEH 07], Nakata et al. [NAK 12], Chandra et al. [CHA 15] and
Guéguen [GUÉ 16a] applied the SIbyD method successfully with columns of soil
instrumented by vertical accelerometer networks to obtain elastic properties and soil
response (particularly the shear wave velocity V s).

This method is particularly accurate for detecting small fluctuations in the building
properties. Moreover, Guéguen et al. [GUÉ 17b] identified using this method, the

Copyright Iste 2022 / File for personal use of Robert Charlier only
188 Deterministic Numerical Modeling of Soil–Structure Interaction

effect of changing the anchoring conditions of a structure, by analyzing the ratio

of stiffnesses between the structure and the soil–structure system (Figure 5.2). The
change was the consequence of a period of extreme cold, and although the depth
of penetration of temperature diffusion remains superficial, they showed that the
soil–structure system became stiffer when the top layer of the soil was frozen. In this
case, we can observe a transition from the structure with flexible behavior to a pseudo-
flexible behavior, which can have moderate effects during earthquakes, but affects the
assumption of invariance over time, often applied to SSI, with important implications
for the development and adjustment of realistic models for damage detection and for
the physical interpretation of frequency variations observed in actual buildings.

Figure 5.2. a) Presentation of the Grenoble Town Hall (France) and the permanent
instrumentation used in this study. b) Seasonal variation in the frequency of the
system fsys (in red) and in the pseudo-flexible frequency fapp for 100 days in
early 2012 in the longitudinal direction (HN2 component) observed at the OGH4
and OGH6 stations located at the top of Grenoble City Hall (France). c) Seasonal
variation in frequency variation Δf (in %) calculated between fsys and fapp (Δf =
(fapp − fsys )/fapp ), according to [GUÉ 17b]. For a color version of this figure, see
www.iste.co.uk/grange/soil.zip

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Urban Seismology 189

5.1.1.3. Cross-interaction between soil and structure

The experimental analysis of the structure and soil response generally does
not make it possible to easily capture the deformation induced by shear located
under the surface foundations. However, a soil-induced nonlinear response can be
expected when buildings are subjected to heavy loading and when the constraint
under cyclic loading increases in the upper layers under the foundations. To
analyze this phenomenon, one solution consists of reproducing soil–structure
systems in centrifuge. These dynamic experiments have proved to be effective in
analyzing the propagation of waves in soil [SEM 98] and the structure–soil and
structure–soil–structure coupling under seismic excitation [GHO 06, MAS 13,
CHA 03, TRO 13, TRO 14]. Furthermore, Bray and Dashti [BRA 14] have recently
used centrifuge models to analyze the vibration-induced effect of the structure on
liquefaction localized just below the foundations. At the same time, again using the
SIbyD method, Chandra et al. [CHA 16] evaluated the nonlinear response of the soil
by analyzing variation in shear wave velocity and associated deformation, using a
network of accelerometric sensors buried in the centrifugated soil medium.

Petrovic and Parolai [PET 16] applied the SIbyD method to recordings acquired by
sensors installed in nearby buildings and boreholes to study the interactions between
soil and structure, by analyzing wave propagation through the building-soil layers. In
this study, the analysis shows complicated patterns of wave propagation in the soil,
resulting from coupling between soil and building. By deconvolution, they separated
the various components of the total wave field, in particular by separating the incident
wave from the backscattered wave in soil. They thus estimated that the energy radiated
by the structure at different depths, immediately under the foundation, could be in the
order of 10% of the energy of the incident wave field. This observation confirms the
results of Guéguen et al. [GUÉ 00] and Guéguen and Bard [GUÉ 05] who, for the first
time, interpreted the energy ratios between the vibration of the structure and the wave
backscattered in soil. They noted that:
– the time damping of the induced soil motion is proportional to the damping
of the soil–structure system, at its fundamental frequency. As a result, part of the
vibration energy of the structure is backscattered into soil in the form of a seismic
wave, which is consistent with what is known about the effects of SSI;
– the spatial decay of the wave field is successively proportional to the geometric
damping
√ of volume waves (namely 1/r in the immediate vicinity) and surface waves
(1/ r) from a certain distance, equal to about five times the dimensions of the
foundation;
– the vibration of the structure is detectable up to about 10 times the size of the
foundation, at which distance the ground motion still represents 5% of the motion
at the base of the building. At a distance of twice the size of the foundation, this
proportion reaches 25%.

Copyright Iste 2022 / File for personal use of Robert Charlier only
190 Deterministic Numerical Modeling of Soil–Structure Interaction

The energy released into soil in the form of seismic waves can therefore be very
significant. The waves thus emitted can alter soil motion near the structure, interacting
with these neighbors and questioning the representativeness of free-field motion for
applications in urban environments.

Figure 5.3. Models of structures tested in centrifuges. The black arrow indicates the
position of the sensors used in the medium to calculate the response of soil column by
SIbyD, according to the different configurations (either in free-field or with the different
structures being tested), and in the structure to calculate the components of the motion
of the soil–structure system, according to [CHA 19]

Resuming the centrifuge tests, Chandra and Guéguen [CHA 19] tested, also
by SIbyD, cross-interactions between soil and structure, for stand-alone structures
of different configurations (rigid, flexible, with two degrees of freedom, etc. –
Figure 5.3), and depending on the loading level. Their main conclusion is as follows:
– It seems that the errors observed between the frequencies obtained by the
analytical equations of Luco et al. [LUC 88] (equations [5.3] and [5.4]) and the
experimental approach are all the more important because the structure is rigid in
relation to soil (Figure 5.4b). In this case, the interaction increases, notably due to
the increasing rocking mode: the cyclical nature of the load, nonlinearity in soil
and at the soil–structure interface induce a bias in the identification of the SSI. The
same observation is made when the fixed-base frequency is higher, which increases
the impedance contrast between soil and the structure, and thus the SSI effect. The
presence of buildings changes the response of the site, as well as its nonlinear
response (Figure 5.4a). This change is illustrated by the reduction of the nonlinear
response of the soil when buildings are present. This reduction may originate due to

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Urban Seismology 191

rocking motion, which cyclically increases the containment pressure and reduces soil
deformation. This phenomenon is more pronounced for rigid buildings.
– The rocking motion dominates in the total structure motion and is the most
sensitive component to nonlinearity.
– It is observed that soil nonlinearity is one of the parameters affecting the
nonlinear response of the system, with a more significant impact on the rocking
motion.
– Nonlinear SSI is more pronounced for rigid buildings than for flexible buildings.

All these observations challenge the validity of soil models derived from
free-field conditions for urbanized areas, as well as modifications in site conditions
after construction. The effects can be non negligible: the presence of buildings can
reduce by 2 the frequency offset in the soil transfer function between weak and strong
motion. They can also call into question the reliability of the nonlinear soil model,
which is defined without taking into account the presence of buildings. In order to
improve the prediction of the building response for design purposes, the site response
must incorporate the effects due to the presence of structures into the nonlinear soil
response and soil–structure interaction. Nonetheless, soil dynamic loading produced
by the motion of the building itself due to rocking must also be integrated, which
can increase cyclic loading in the top layer. In the presence of saturated soil, the
response could be different, since rocking may accentuate liquefaction [BRA 14].
This observation also suggests that in addressing the problems of nonlinear SSI,
soil and structure should be considered as a single system, presenting multiple and
cross-interactions.

5.1.1.4. Interactions between structures

The heterogeneity of urbanizations also very often cause several types of buildings,
of different sizes, foundation systems and structural elements, to exist side by side.
The distribution of the building then generates large lateral variations and significant
contrasts between neighboring structures. Imposing masses are thus built next to
lighter structures. In dynamics, it is legitimate to consider whether the presence of
large vibrating masses near light masses does not create behavioral changes or even
dynamic interactions between buildings.

Kitada et al. [KIT 99] studied the interaction effects between buildings at an
experimental site. The study consisted of causing a reference structure to vibrate using
an exciter placed at the top. This experiment was reproduced in centrifuge by Chazelas
et al. [CHA 03], as well as by using computers in the context of this study (Figure 5.5).
In these three examples, the analysis consisted of assessing the response of a structure,
with or without the presence of a nearby building, excited by a load applied at its top
(Table 5.2).

Copyright Iste 2022 / File for personal use of Robert Charlier only
192 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 5.4. Results of centrifuge trials, according to [CHA 19]. a) Soil column
response (SIbyD method), under light (in black) and heavy (in red) loading for different
configurations, with or without buildings. The frequency values correspond to the
resonance frequency of the soil. b) Comparison of the fixed-base frequency extracted
by deconvolution (SIbyD method, f1∗ ) or by the equations of the model equation [5.3]
(f1 ). The continuous line corresponds to the 1:1 relation. For a color version of this
figure, see www.iste.co.uk/grange/soil.zip

In all these experimental examples, numerical studies or in centrifuge, the same

observations are reproduced. First, it can be noted that there is a strong coupling
between the active structure and the passive structure, especially when the resonance

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Urban Seismology 193

between the soil and the active building exists and when the buildings have similar
resonance frequencies (Figure 5.5). The essential effect is to cause the passive building
to oscillate, in the order of 10% of the one calculated at the top of the active building,
even for significant distances (in the order of the height of the building). It can also
be noted that this phenomenon exists for the higher modes of the structure. Since
the frequency of the active building and that of the passive building are close to
one another, the shape of the beatings is characteristic of the coupling phenomena
of resonant systems with almost identical frequencies. In frequency (Figure 5.6), this
is reflected by a decoupling of the vibration frequency of the structure, as shown by
Kitada et al. [KIT 99].

Building B (m) H (m) D (m) f1 ζ1 (%)

B1E 10 20 10 1.64 5.2
B1S 0 1.67 5.8
B2E 20 40 10 0.88 5.4
B2S 0 0.90 5.8
B3E 40 40 10 1.25 6.2
B3S 0 1.32 6.5
B4E 40 50 10 0.93 5.8
B4S 0 0.98 6.1
B5E 40 60 10 0.70 12.0
B5S 0 0.74 12.7

Table 5.2. Characteristics of modeled structures. B , H and D respectively represent

the width, height and recess of the buildings. f1 and ζ1 are respectively the frequency
and damping of fixed-base structures, inferred from modeling

Profile Layer βs (m/s) ρ(t/m3 ) E(108 P a) ν z(m)

I 1 100 1.50 0.38 0.25 0
III 1 200 1.50 1.50 0.25 0
2 250 1.80 2.81 0.25 25
3 300 2.00 4.50 0.25 45
4 330 1.95 5.31 0.25 85
5 460 2.10 11.0 0.25 130
6 600 2.00 18.0 0.25 195

Table 5.3. Characteristics of the 1D stratified half-space used in the model. β , ρ, E and
ν respectively correspond to the velocity of shear waves, to the density, the elasticity
modulus (or the Young modulus) and to the Poisson ratio of the different layers. z
represents the position of the roof of the layer

Copyright Iste 2022 / File for personal use of Robert Charlier only
194 Deterministic Numerical Modeling of Soil–Structure Interaction

Figure 5.5. Examples of displacements u(t) calculated at the top of buildings

B1E-B1S, B2E-B2S and B5E-B5S for the cases where they have foundations on
(a) type-I soil and (b) type-III soil (Table 5.2). Three distances were tested, which
correspond to L1 = Ha /4, L2 = Ha /2 and L3 = Ha . The classes of buildings
presented correspond to the situations where f˜p > f˜a (B1E-B1S), f˜p ≈ f˜a (B2E-B2S)
and f˜p < f˜a (B5E-B5S). For example, on the right, we can see the displacement u(t)
of the active building, for the case of the passive building B2E based on soil I (on top)
and III (at the bottom)

In 2014, an earthquake of magnitude 4.9 occurred 100 km south of Grenoble

[COU 14]. An accelerometer station in the French permanent accelerometer
network [PÉQ 08], located in the vicinity of three large buildings in Grenoble
(Figure 5.7), constituting a cluster, recorded a horizontal soil acceleration of
2.5cm/s2 . The Y-shaped sedimentary basin of the Grenoble Valley, composed of
thick glaciolacustrine deposits (up to 1 km) systematically amplifies seismic motion
[LEB 01, GUÉ 07] between 0.3 and 5.0 Hz, with the result that the inhabitants located
in the three towers felt the vibrations more strongly. This was particularly the case
during the 2014 earthquake. An investigation into what inhabitants had experienced
was immediately conducted and revealed that within the cluster, the inhabitants in the
central tower had hardly felt the vibrations of the earthquake, compared to those of
the two towers at the extremity of the cluster. Guéguen and Colombi [GUÉ 17a] thus

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Urban Seismology 195

showed that, while the three towers were identical and with foundations in the same
soil, not only was the modal response of the middle structure different from that of
the other two, but also the seismic response of the structure alone or included in the
cluster differed considerably. This difference was particularly sensitive to the position
of the seismic source, in relation to the orientation of the cluster (Figure 5.7).

Figure 5.6. Spectral responses of soil–structure systems

presented in Figure 5.5

This observation ultimately validates previous centrifugal and modeling tests,

and confirms the observations made by Kitada et al. [KIT 99]. It also questions the
meaning of the response of a structure generally defined without taking into account
its nearby environment and the ability of a structure to interact with these neighbors,
under seismic loading. What has been observed in the elastic domain can have an
exacerbated impact when stronger motions occur, influencing the relation between
the parameter characterizing the engineering seismic demand parameter EDP (e.g.
inter-stage deformation) and the intensity measure IM of seismic motion (e.g. the
maximal value of acceleration). This relation between EDP and IM is essential in
seismic design based on performance [PRI 00]. The difference in energy between the
case of a stand-alone structure, or located within a cluster, can reach 20%, depending
on the azimuth of seismic loading, amplifying or mitigating the building motion. This
example only concerns three buildings; however, it is reasonable to imagine that global
coupling at the urban scale can lead to an even more complex response that can be

Copyright Iste 2022 / File for personal use of Robert Charlier only
196 Deterministic Numerical Modeling of Soil–Structure Interaction

characterized analogously to a regionally-sized geophysical material. Composed of

a group of resonators, it can reveal a new type of physics, which in our case could
control wave propagation within designated frequency ranges. When observing the
very structured organization of certain exposed cities (Figure 5.8), the question of
how these resonators composed of cities would react, in the event of an earthquake,
ought to be raised.

Figure 5.7. Variation in the difference of vibration energy of the central tower (MB*)
between the case where it is isolated and the case where it is inside the cluster of the
three towers, for different azimuth values of the source θ with respect to the alignment
of the cluster. + and - indicate the positive or negative energy differential. For a color
version of this figure, see www.iste.co.uk/grange/soil.zip

5.1.2. Contamination of urban seismic motion by the vibration of

buildings

In the literature, which is abundant about the phenomena of soil–structure

interaction and site effects, only a few examples highlight the effect of urban
constructions on seismic motion. Jennings [JEN 70], Kanamori et al. [KAN 71],
Guéguen et al. [GUÉ 00], Cornou et al. [COR 04], and Kim et al. [KIM 01] identified
waves generated by the vibration of structures due to the inertial interaction of
soil–structure. These waves are diffracted back into the soil and superimposed onto
the incident seismic motion. In these studies, the structures were excited by internal or
external forces applied at the top, and the wave diffused into soil was easily identifiable
in the free-field recording station. This is not a local phenomenon, because Kanamori
et al. [KAN 71], Favela et al. [FAV 02], and Kim et al. [KIM 01] observed it at
distances greater than 50 km. For example, during the terrorist attacks on the twin
towers of the World Trade Center, a seismological station managed by the Lamont-
Doherty seismological network (Colombia University), situated about 40 kilometers
from the impact site, not only showed the effect of the complete destruction of the
towers, but also the effect of the vibration of the structures during the impact of the
aircrafts (Figure 5.9).

Since then, a few observations have been made reutilizing the scheme originally
considered by Guéguen et al. [GUÉ 00] at the Volvi test site in Greece, but applied to

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Urban Seismology 197

other soil–structure configurations [MUC 03, GAL 06], which confirm the existence
of coupling between soil and structure motion. These observations have also been
confirmed by simplified numerical modeling [WIR 96] and/or generalized to a group
of buildings simulating a city [KHA 06, ISB 15], reproducing, in some cases, the
monochromatic nature and duration of signals observed in Mexico City in 1985 during
the Michoacan earthquake [AND 86]. This effect is not negligible and yields the
question of how the response of the sites can be contaminated by the redistribution
of seismic energy to the city surface.

Figure 5.8. Aerial images of cities exposed to seismic hazard and

presenting a macroscopic organization of resonating elements. For a
color version of this figure, see www.iste.co.uk/grange/soil.zip

In Japan, as early as 1935, Sezawa and Kanai [SEZ 35] noted that the attenuation
of earthquake effects with distance seemed more significant on the outskirts of cities,
and they attributed this observation to the local coupling between the city and soil.
Guéguen et al. [GUÉ 02] simulated the contribution of the Roma Norte district in
Mexico City to the total seismic motion of soil. They showed that the evaluation of
site effects in cities cannot be freed from integration into the process of the urban layer
of buildings. They called this phenomenon the site–city interaction (SCI).

Copyright Iste 2022 / File for personal use of Robert Charlier only
198 Deterministic Numerical Modeling of Soil–Structure Interaction

41.5
PAL station - East component - 2001/11/09 - 12:46
41.4

41.3

41.2
0 10 20 30 40
41.1
Longitude

Palisasades, NY
seismological station
41 LCSN network

40.9

40.8
New-York City
WTC
40.7

40.6

40.5
-75 -74.5 -74 -73.5 -73
Latitude

Figure 5.9. Recording of soil motion 40 km from the WTC excited by the impact
of aircrafts during the terrorist attacks of September 11, 2001, according to
[KIM 01]. For a color version of this figure, see www.iste.co.uk/grange/soil.zip

Wirgin and Bard [WIR 96] are certainly the first to have conducted numerical-
based studies for analyzing the coupling between soil and structure motion when
these structures are subjected to earthquakes. This idea was mainly motivated by the
signature recordings of soil motion over the lake-bed area of Mexico City. In fact,
as shown in Figure 5.10, the motion in the lake area certainly shows considerable
amplitude amplification, due to the site effects of the surface soft layer, as well as
remarkable lengthening and monochromatic beats. If we consider the 1D response of
the Mexico City soil column, which has been well documented since the Michoacan
earthquake of 1985, the surface motion can be calculated based on the motion recorded
at the rock. It can be observed that the amplitude is well rendered, but not at all the
monochromatic oscillations or the duration of the ground motion. On the contrary, in
this area, there is a dense and massive habitat built on a flexible and superficial soil
formation that favors the phenomena of soil–structure interaction. The contribution of
structure motion to the observed soil motion was also encouraged by the similarity of
the vibration frequencies of the structures and the lake surface layer, which meant that
strong coupling had been established during the earthquake.

Guéguen et al. [GUÉ 02] demonstrated, with a simple analytical model, that in
the time domain, motion of the wave field radiated by all structures was of the same
order of magnitude as the incident seismic motion, thus demonstrating the importance
of the site–city effect in the case of Mexico City. Despite some differences, mainly
due to the simplicity of the model used, the calculated total field (direct + radiated

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Urban Seismology 199

by the structures) presents similarities with the records observed for the earthquake
in question, particularly with regard to the lengthening and characteristic beats in the
Mexico City area. This phenomenon does not seem to increase seismic motion, but
rather perturb it with direct consequences for seismic risk. Indeed, Guéguen et al.
[GUÉ 02] reached the conclusion that the estimation of site effects using the spectral
ratio method [BOR 70] and observed on data could only be explained by integrating
the total field for the calculation of site effects.

Tacu rock station : Vs>800m/s

M 7.3 earthquake (September, 14 1995)

Co56 soil station : Vs=65m/s

0 100 200 300

Times (s)

Figure 5.10. Accelerometer recordings in a station at the rock (on top) and in
the lake area of Mexico City (at the bottom) during a significant earthquake.
For a color version of this figure, see www.iste.co.uk/grange/soil.zip

An analysis of the energy ratios between the urban layer and the soil layer
allowed Guéguen et al. [GUÉ 02] to evaluate urban configurations that could produce
a site–city interaction phenomenon. The maximum kinetic energy Es of all the
sediment filling and the maximal kinetic energy Eb of the n buildings are successively
calculated. Guéguen et al. [GUÉ 02] thus propose the following simplified relation:

n
Eb Sbi Hbi fs2
=2 2 [5.5]
Es i=1
Ss Hs fbi

which indicates that this phenomenon is essentially controlled by three terms that are:
– the urban density that represents the area of land occupied by buildings,
expressed as the ratio between the total built surface area and the area of the site
under consideration, i.e. SSbis ;
– the mass contrast, and therefore of geometry, between the city and soil,
characterized by the ratio between the average height of the buildings that compose
the city and the thickness of the sedimentary filling, i.e. Hbi
Hs ;

– the coincidence between the resonance frequencies of the soil layer and of the
buildings, which favors the trapping of surface waves emitted by the vibration of
f2
buildings, namely f s2 .
bi

Copyright Iste 2022 / File for personal use of Robert Charlier only
200 Deterministic Numerical Modeling of Soil–Structure Interaction

Groby and Wirgin [GRO 04] also demonstrated, using modeling, that identical rigid
blocks regularly distributed over half a space significantly altered the amplitude and
duration of seismic motion. Kham et al. [KHA 06] observed a perturbation outside the
city, exacerbated when the urban density is high and there is resonance between the
city and the basin. They also digitally showed that:
– in a periodic city, the ISV effects are beneficial: the soil motion is reduced;
– this reduction increases with urban density and is maximal when construction
and soil frequencies coincide. Under optimal conditions (the highest density and
perfect resonance), the reduction reaches 50%;
– the density effect can, however, be significant even when the frequencies do not
coincide. For example, the energy of the radiated field in relation to the incident field
is of the same order (67%) in “low density – resonance” and “strong density – no
resonance” situations;
– these reduction effects decrease significantly when the regularity of the building
is broken. The reduction for non-periodic cities does not exceed 15%, which can be
explained by the low number of buildings at 2 Hz and/or the low group effect due to
the irregular layout.

Nevertheless, unlike Groby and Wirgin [GRO 04], Kham et al. [KHA 06]
showed that the spatial consistency of seismic movement in the urban field was very
perturbed and that the motion of structures and soil was reduced due to soil–structure–
soil effects, and the greatest perturbations in soil motion related to the presence
of structures appearing on the periphery of the city. This simulation resumes the
observations made by Sezawa and Kanai as early as 1935 [SEZ 35].

At the scale of a city composed of a multitude of resonant structures, it is

easy to imagine strong coupling between multiple structures and soil, redistributing
the incident seismic energy. This raises questions about the validity of seismic
evaluations of soil motion in urban areas, the understanding and interpretation
of the site effects observed during earthquakes, and the way to design structures
without integrating their urban environment in the process. This collective effect
of coupled resonators has already been described in physics at the laboratory scale
through the concept of locally resonant metamaterials, for example for a plate wave
interacting with a dense collection of metal rods perpendicularly attached to it
[RUP 14]. Applied to geophysics, where the earth replaces the plate for surface
waves and a dense distribution of buildings would play the role of resonant metal
rods, these metamaterials would cause a redistribution of energy of seismic waves
[BRÛ 14, NIC 14, COL 16] and frequency band-gaps could appear. We might then
ask to what extent the urban organization can contribute to the emergence of these
band-gaps, in which the seismic energy is controlled and reduced, ultimately allowing
seismic-resistant developments to be devised, i.e. structures that, organized according
to a particular scheme, could create anti-seismic barriers.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Urban Seismology 201

Figure 5.11. Aerial and forest view of the deployment (yellow dots and yellow flags)
of seismic sensors in METAFORET, at the interface between a field and a forest. The
curves below represent the propagation of recorded waves on the vertical component
of a line of seismic sensors during a test with a vibrator placed in the field. The black
line represents the forest-field boundary. All three figures correspond to frequency
bands applied by filtering to the recorded data. A clear transition can be seen around
50 Hz that corresponds to the vertical resonance frequency of the trees (according to
[ROU 18]). For a color version of this figure, see www.iste.co.uk/grange/soil.zip

We are still a long way from that, since the relationship between the vibration
mode of the structures and the wavelengths involved is yet to be clarified. However,
after laboratory experiments [RUP 14] and numerical modeling, a geophysical-scale
experiment was conducted. It consisted of measuring the propagation of surface waves
at the border between a field and a dense forest using a network of 1000 sensors
[ROU 18]. The distance between the seismic sensors was equal to the distance between
the trees (Figure 5.10). Early analyses clearly show prohibited frequency bands, with
couplings between surface waves, also very present in sedimentary fillings of large
exposed urban centers, and trees that resonate according to their vertical and horizontal
mode. This dense and irregular forest is ultimately an analogue of a city, the concept
of meta-forest can extend to that of meta-city, which would allow us to consider the
control of seismic motion within the urban center and for certain frequency bands.

Copyright Iste 2022 / File for personal use of Robert Charlier only
202 Deterministic Numerical Modeling of Soil–Structure Interaction

5.1.3. Conclusion

From this compilation of observations, specific experiments and numerical

computations, we can draw some conclusions about the effects of constructions on
seismic motion in urban areas. These results first indicate that the effect exists:
soil motion is significantly perturbed in the immediate vicinity of a building and
centrifuge trials indicate that the buildings “talk” to each other through soil. Computer
simulations support the existence of strong interaction under favorable conditions
(presence of resonance between soil and constructions and high urban density).
The physical origin is multiple, mixing kinematic and inertial interactions: waves are
generated in soil at the base of the vibrating buildings and they are trapped in the
surface layers when the contrast is sufficient. Group effects also occur when distances
between buildings are short. In the presence of high urban densities and/or large
buildings, the inertia effect produces some (slight) frequency shifts. The phenomenon
is particularly exacerbated when there is coincidence between soil and building
frequencies: the ingredients for a significant SCI are therefore the simultaneous
presence of a thin layer of soft sediments and dense urbanization with homogeneous
buildings exhibiting similar frequencies.

In situations of strong multiple interaction, the effects appear to be generally

beneficial: soil motion in urban areas is decreased for particular frequency bands.
However, local amplifications cannot be completely eliminated. The overall beneficial
effect should, nonetheless, be balanced by the fact that the SCI significantly increases
the variability of seismic motion in urban areas. Some significant amplifications
can locally occur, currently unpredictable because they are highly dependent on the
incident wave field (frequency and phase). On the contrary, areas at the edges of dense
and homogeneous urban centers are exposed to increased soil motion as a consequence
of the effects of the SCI outside the urbanized area.

From a seismological point of view, it seems important to analyze seismic

recordings taking into account the urban environment, in addition to the effects of
sources, propagation and site effects. This issue is particularly important when dense
networks are deployed in cities for seismic wave field analyses. The observation and
analysis of the distribution of damage should also take into account the SCI as damage
may not be due solely to vulnerability variability or site effects: from the transition
from site–city interaction to the concept of meta-city.

From a seismic risk perspective, the main lesson to be drawn is that structural
and site response may undergo modifications due to the urban environment. This
could lead to many unpredictable developments, for example in urban planning (by
trying to design “optimal land use” in order to reduce soil motion) and in the time
evolution of risk (risk may change according to constructions and demolitions).
Nevertheless, before confirming these consequences, the next necessary step is to
obtain indisputable experimental evidence of the occurrence of these effects in actual

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Urban Seismology 203

cities subjected to earthquakes. It will therefore be necessary to specifically equip

cities following particular instrumental methods and to use new sophisticated signal
processing techniques to separate incident waves from the waves induced by the
structures.

While there is still a need for additional 3D computations, these new results
confirm that full-scale observation data in actual buildings are much more reliable than
the information provided by sophisticated laboratory or modeling experiments, due to
the in situ evaluation of the structure, its environment and boundary conditions. There
are now signal processing algorithms that can provide new and valuable information
about unexpected couplings or behaviors; these algorithms could be used to test and
adjust new models and methods to improve structural health monitoring strategies.

This chapter was supported by funds from The METAFORET project funded
by the French National Research Agency (ANR), the LabEx OSUG@2020
(Investissements d’avenir-ANR10LABX56), and the URBASIS-EU ITN project from
the European Union’s H2020 research and innovation program under the Marie
Sklodowska-Curie Grant, Agreement Number 813137.

5.2. References

[AND 86] A NDERSON J.G., B ODIN P., B RUNE J.N. et al., “Strong ground motion from the
Michoacan, Mexico, earthquake”, Science, vol. 233, no. 4768, pp. 1043–1049, 1986.
[APS 87] A PSEL R., L UCO J., “Impedance functions for foundations embedded in a
layered medium: An integral equation approach”, Earthquake Engineering and Structural
Dynamics, vol. 15, pp. 213–231, 1987.
[AVI 96] AVILES J., P EREZ -ROCHA L., “Evaluation of interaction effects on the system
period and the system damping due to foundation embedment and layer depth”, Soil
Dynamics and Earthquake Engineering, vol. 15, no. 1, pp. 11–27, 1996.
[BIE 75] B IELAK J., “Dynamic behavior of structures with embedded foundations”, Journal
of Earthquake Engineering and Structural Dynamics, vol. 3, no. 3, pp. 259–274, 1975.
[BOR 70] B ORCHERDT R.D., “Effects of local geology on ground motion near San Francisco
Bay”, Bulletin of the Seismological Society of America, vol. 60, pp. 29–61, 1970.
[BRA 14] B RAY J., DASHTI S., “Liquefaction-induced building movements”, Bulletin of
Earthquake Engineering, vol. 12, no. 3, pp. 1129–1156, 2014.
[BRÛ 14] B RÛLÉ S., JAVELAUD E.H., E NOCH S. et al., “Experiments on seismic
metamaterials: Molding surface waves”, Physical Review Letters, vol. 112, no. 3,
pp. 133–901, 2014.
[CHA 90] C HAVEZ -G ARCIA F., BARD P.-Y., “Surface ground motion modifications by the
presence of a thin resistant layer. Applications to Mexico City”, 9th European Conference
on Earthquake Engineering, vol. 4-B, Moscow, pp. 37–46, 1990.

Copyright Iste 2022 / File for personal use of Robert Charlier only
204 Deterministic Numerical Modeling of Soil–Structure Interaction

[CHA 03] C HAZELAS J., G UÉGUEN P., BARD P. et al., “Modélisation de l’effet site-ville en
modèles réduits centrifugés”, Actes du 6ème colloque National AFPS, vol. I, Palaiseau,
France, pp. 245–252, 2003.
[CHA 15] C HANDRA J., G UÉGUEN P., S TEIDL J. et al., “In-situ assessment of the G-i3 curve
for characterizing the nonlinear response of soil: Application to the Garner Valley Downhole
Array (GVDA) and the Wildlife Liquefaction Array (WLA)”, Bulletin of the Seismological
Society of America, vol. 105, no. 1A, pp. 993–1010, 2015.
[CHA 16] C HANDRA J., G UÉGUEN P., B ONILLA F., “On the use of the seismic interferometry
technique for testing PGV/Vs as a proxy for predicting nonlinear soil response”, Soil
Dynamics and Earthquake Engineering, vol. 85, pp. 146–160, 2016.
[CHA 19] C HANDRA J., G UÉGUEN P., “Nonlinear response of soil–structure systems using
dynamic centrifuge experiments”, Journal of Earthquake Engineering, vol. 23, no. 10,
pp. 1719–1741, 2019.
[CLI 06] C LINTON J.F., B RADFORD S.C., H EATON T.H. et al., “The observed wander of the
natural frequencies in a structure”, Bulletin of the Seismological Society of America, vol. 96,
no. 1, pp. 237–257, 2006.
[COL 16] C OLOMBI A., ROUX P., G UENNEAU S. et al., “Forests as a natural seismic
metamaterial: Rayleigh wave bandgaps induced by local resonances”, Scientific Reports,
vol. 6, no. 19238, 2016.
[COR 04] C ORNOU C., G UÉGUEN P., BARD P.-Y. et al., “Ambient noise energy bursts
observation and modeling: Trapping of harmonic structure-soil induced-waves in a topmost
sedimentary layer”, Journal of Seismology, vol. 8, pp. 507–524, 2004.
[COU 14] C OURBOULEX F., D ELOUIS B., D UJARDIN A. et al., “The two events of
Barcelonette (French Alps), 2012 (MW 4.1) and 2014 (MW 4.9) - The role of directivity
on ground motions, macroseismic intensities and site”, Second European Conference on
Earthquake Engineering and Seismology, Istanbul, Turkey, 25–29 August 2014.
[FAV 02] FAVELA J., H EATON T.H., TANIMOTO T., “Far-field energy radiation from a
building under harmonic excitation”, AGU Fall Meeting Abstracts, no. A1169, http://
adsabs.harvard.edu/abs/2002AGUFM.S12A1169F (last accessed September 2016), 2002.
[GAL 06] G ALLIPOLI M.R., M UCCIARELLI M., P ONZO F. et al., “Buildings as a seismic
source: Analysis of a release test at Bagnoli”, Bulletin of the Seismological Society of
America, vol. 96, no. 6, pp. 2457–2464, 2006.
[GAZ 83] G AZETAS G., “Analysis of machine foundation vibrations: State of the art”, Soil
Dynamics and Earthquake Engineering, vol. 2, pp. 2–42, 1983.
[GAZ 91] G AZETAS G., “Foundation vibrations”, in FANG H.-Y. (ed.), Foundation
Engineering Handbook, 2nd edition, Van Nostrand Reinhold, New York, 1991.
[GHO 06] G HOSH B., M ADABHUSHI S., “Centrifuge modeling of seismic soil structure
interaction effects”, Nuclear Engineering Design, vol. 237, no. 8, pp. 887–896, 2006.
[GRO 04] G ROBY J.-P., W IRGIN A., “Seismic motion in urban sites consisting of blocks
in welded contact with a soft layer overlying a hard half-space”, Geophysical Journal
International, vol. 172, no. 2, pp. 725–758, 2004.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Urban Seismology 205

[GUÉ 00] G UÉGUEN P., BARD P., S OUSA O LIVEIRA C., “Experimental and numerical
analysis of soil motions caused by free vibrations of a building model”, Bulletin of the
Seismological Society of America, vol. 90, no. 6, pp. 1464–1479, 2000.
[GUÉ 02] G UÉGUEN P., BARD P.-Y., C HAVEZ -G ARCIA F., “Site-city interaction in Mexico
City-like environments: An analytical study”, Bulletin of the Seismological Society of
America, vol. 92, no. 2, pp. 794–811, 2002.
[GUÉ 05] G UÉGUEN P., BARD P.-Y., “Soil-structure and soil-structure-soil interaction:
Experimental evidence at the Volvi test site”, Journal of Earthquake Engineering, vol. 9,
no. 5, pp. 657–693, 2005.
[GUÉ 07] G UÉGUEN P., C ORNOU C., G ARAMBOIS S. et al., “On the limitation of the H/V
spectral ratio using seismic noise as an exploration tool: Application to the Grenoble valley,
a small apex ratio basin”, Pure and Applied Geophysics, vol. 164, no. 1, pp. 1–20, 2007.
[GUÉ 16a] G UÉGUEN P., “Predicting nonlinear site response using spectral acceleration vs.
PGV/Vs30: A case history using the Volvi-test site”, Pure and Applied Geophysics, vol. 173,
no. 6, pp. 2047–2063, 2016.
[GUÉ 16b] G UÉGUEN P., J OHNSON P., ROUX P., “Nonlinear dynamics induced in a structure
by seismic and environmental loading”, The Journal of the Acoustical Society of America,
vol. 140, no. 1, pp. 582–590, 2016.
[GUÉ 17a] G UÉGUEN P., C OLOMBI A., “Experimental and numerical evidence of the
clustering effect of structures on their response during an earthquake: A case study of three
identical towers in the city of Grenoble, France”, Bulletin of the Seismological Society of
America, vol. 106, no. 6, pp. 2855–2864, 2017.
[GUÉ 17b] G UÉGUEN P., L ANGLAIS M., G ARAMBOIS S. et al., “How sensitive are site
effects and building response to extreme cold temperature? The case of the Grenoble’s
(France) City Hall building”, Bulletin of Earthquake Engineering, vol. 15, no. 3, pp. 889–
906, 2017.
[GUY 99] G UYER R. A., J OHNSON P. A., “Nonlinear mesoscopic elasticity: Evidence for a
new class of materials”, Physics Today, vol. 52, pp. 30–6, 1999.
[HAD 93] H ADJIAN A., “Seismic soil-structure interaction: A full circle”, Memorias X
Congreso Nacional de Ingenieria Sismica, Puerto Vallarta, Mexico, pp. 1–16, 1993.
[HIL 84] H ILL N., L EVANDER A., “Resonances of low-velocity layers with lateral variations”,
Bulletin of the Seismological Society of America, vol. 74, pp. 521–537, 1984.
[HSI 62] H SIEH T., “Foundation vibrations”, Proceedings of Institution of Civil Engineers,
vol. 22, pp. 211–225, 1962.
[ISB 15] I SBILIROGLU Y., TABORDA R., B IELAK J., “Coupled soil-structure interaction
effects of building clusters during earthquakes”, Earthquake Spectra, vol. 31, no. 1,
pp. 463–500, 2015.
[JOH 05] J OHNSON P., S UTIN A., “Slow dynamics and anomalous nonlinear fast dynamics
in diverse solids”, The Journal of the Acoustical Society of America, vol. 117, pp. 124–30,
2005.

Copyright Iste 2022 / File for personal use of Robert Charlier only
206 Deterministic Numerical Modeling of Soil–Structure Interaction

[JEN 70] J ENNINGS P.C., “Distant motion from a building vibration test”, Bulletin of the
Seismological Society of America, vol. 60, no. 6, pp. 2037–2043, 1970.
[KAN 71] K ANAMORI H., M ORI J., A NDERSON D. et al., “Seismic excitation by the space
shuttle Columbia”, Nature, vol. 349, pp. 781–782, 1971.
[KAU 74] K AUSEL E., Forced vibrations of circular foundations on layered media, Research
Report no. R77-3, MIT, Boston, USA, 1974.
[KAU 10] K AUSEL E., “Early history of soil–structure interaction”, Soil Dynamics and
Earthquake Engineering, vol. 30, no. 9, pp. 822–832, 2010.
[KHA 06] K HAM M., S EMBLAT J.-F., BARD P.-Y. et al., “Seismic site–city interaction: Main
governing phenomena through simplified numerical models”, Bulletin of the Seismological
Society of America, vol. 96, no. 5, pp. 1934–1951, 2006.
[KIM 01] K IM W.Y., S YKES L., A RMITAGE J. et al., “Seismic waves generated by aircraft
impacts and building collapses at World Trade Center, New-York City”, Eos, Transactions
American Geophysical Union, vol. 82–47, p. 565, 2001.
[KIT 99] K ITADA Y., K INOSHITA M., I GUCHI M. et al., “Soil-structure interaction effect on
an NPP reactor building. Activities of Nupec: Achievements and the current status”, in
C ELEBI M., O KAWA I. (eds), Proc. UJNR Workshop on Soil-Structure Interaction, Paper
no. 18, Menlo Park, California, September 22–23, 1999.
[LAM 04] L AMB H., “On the propagation of tremors over the surface of an elastic solid”,
Philosophical Transactions of the Royal Society, vol. A-203, pp. 1–42, 1904.
[LEB 01] L EBRUN B., H ATZFELD D., BARD P., “A site effect study in urban area:
Experimental results in Grenoble (France)”, Pure and Applied Geophysics, vol. 158,
pp. 2543–2557, 2001.
[LEV 85] L EVANDER A.R., H ILL N., “P-SV resonances in irregular low-velocity surface
layers”, Bulletin of the Seismological Society of America, vol. 75, pp. 847–864, 1985.
[LUC 71] L UCO J., W ESTMANN R., “Dynamic response of circular footing”, Journal of
Engineering Mechanics, ASCE, vol. 97, no. EM 5, p. 1381, 1971.
[LUC 74] L UCO J., “Impedance functions for a rigid foundation on a layered medium”,
Nuclear Engineering and Design, vol. 31, pp. 204–217, 1974.
[LUC 88] L UCO J., T RIFUNAC M., W ONG H., “Isolation of soil structure interaction effects
by full scale forced vibration tests”, Earthquake Engineering and Structural Dynamics,
vol. 16, no. 1, pp. 1–21, 1988.
[LYS 65] LYSMER J., Vertical motions of rigid footings, PhD thesis, University of Michigan,
1965.
[MAS 13] M ASON H., T ROMBETTA N., C HEN Z. et al., “Seismic soil-foundation-structure
interaction observed in geotechnical centrifuge experiments”, Soil Dynamics and
Earthquake Engineering, vol. 48, pp. 162–174, 2013.
[MEH 07] M EHTA K., S NIEDER R., G RAIZER V., “Downhole receiver function: A case
study”, Bulletin of the Seismological Society of America, vol. 97, no. 5, pp. 1396–1403,
2007.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Urban Seismology 207

[MIC 10a] M ICHEL C., G UÉGUEN P., “Time-frequency analysis of small frequency variations
in civil engineering structures under weak and strong motion”, Structural Health
Monitoring, vol. 9, no. 2, pp. 159–171, 2010.
[MIC 10b] M ICHEL C., G UEGUEN P., E L A REM S. et al., “Full scale dynamic response of an
RC building under weak seismic motions using earthquake loadings, ambient vibrations and
modelling”, Earthquake Engineering and Structural Dynamics, vol. 39, no. 4, pp. 419–441,
2010.
[MOS 87] M OSLEM K., T RIFUNAC M., “Spectral amplitudes of strong earthquake
accelerations recorded in buildings”, Soil Dynamics and Earthquake Engineering, vol. 6,
no. 2, pp. 100–107, 1987.
[MUC 03] M UCCIARELLI M., G ALLIPOLO M., P ONZO F. et al., “Seismic waves generated
by oscillating buildings: Analysis of a release test”, Soil Dynamics and Earthquake
Engineering, vol. 23, pp. 255–262, 2003.
[NAK 12] NAKATA N., S NIEDER R., “Estimating near-surface wave velocities in Japan
by applying seismic interferometry to KiK-net data”, Journal of Geophysical Research,
vol. 117, no. B01308, 2012.
[NIC 14] N ICOLETTI O., “Seismic cloaks”, Nature Materials, vol. 13, no. 5, pp. 428, 2014.
[PÉQ 08] P ÉQUEGNAT C., G UÉGUEN P., H ATZFELD D. et al., “The French Accelerometric
Network (RAP) and National Data Centre (RAP-NDC)”, Seismological Research Letters,
vol. 79, no. 1, pp. 79–89, 2008.
[PET 16] P ETROVIC B., PAROLAI S., “Joint deconvolution of building and downhole strong-
motion recordings: Evidence for the seismic wavefield being radiated back into the
shallow geological layers”, Bulletin of the Seismological Society of America, vol. 106,
pp. 1720–1732, 2016.
[PRI 00] P RIESTLEY M. J.N., “Performance based seismic design”, Bulletin of the New
Zealand Society for Earthquake Engineering, vol. 33, no. 3, pp. 325–346, 2000.
[REI 36] R EISSNER E., “Stationare, axialsymmetrische, durch eine shut-telnde masse erregte
shwimgugen eines homogenen elastischen halbraumesn”, Ingenieur-Archiv, vol. 7, p. 381,
1936.
[RIC 67] R ICHART F., W HITMAN R., “Comparison of footing vibrations tests with theory”,
Journal of the Soil Mechanics and Foundations Division, vol. 93, pp. 143–168, 1967.
[ROU 18] ROUX P., B INDI D., B OXBERGER T. et al., “Toward seismic metamaterials:
The METAFORET project”, Seismological Research Letters, vol. 89, pp. 582–593,
2018.
[RUP 14] RUPIN M., L EMOULT F., L EROSEY G. et al., “Experimental demonstration of
ordered and disordered multiresonant metamaterials for lamb waves”, Physical Review
Letters, vol. 112, no. 23, pp. 234–301, 2014.
[ŞAF 98] Ş AFAK E., “New approach to analyzing soil-building systems”, Soil Dynamics and
Earthquake Engineering, vol. 17, no. 7, pp. 509–517, 1998.

Copyright Iste 2022 / File for personal use of Robert Charlier only
208 Deterministic Numerical Modeling of Soil–Structure Interaction

[ŞAF 99] Ş AFAK E., “Wave-propagation formulation of seismic response of multistory

buildings”, Journal of Structural Engineering, vol. 125, no. 4, pp. 426–437, 1999.
[SEM 98] S EMBLAT J., L UONG M., “Wave propagation through soils in centrifuge testing”,
Journal of Earthquake Engineering, vol. 2, no. 1, pp. 147–171, 1998.
[SEZ 35] S EZAWA K., K ANAI K., “Decay in seismic vibration of a simple or tall building
structure by dissipation of their energy into the ground”, Bulletin of the Earthquake
Research Institute, vol. XIII, no. 3, pp. 681–697, 1935.
[SNI 06] S NIEDER R., S AFAK E., “Extracting the building response using seismic
interferometry: Theory and application to the Millikan Library in Pasadena, California”,
Bulletin of the Seismological Society of America, vol. 96, no. 2, pp. 586–598, 2006.
[STE 98] S TEWART J., F ENVES G.L., “System identification for evaluating soil–structure
interaction effects in buildings from strong motion recordings”, Earthquake Engineering
and Structural Dynamics, vol. 27, no. 8, pp. 869–885, 1998.
[STE 99] S TEWART J., F ENVES G., S EED R., “Seismic soil-structure interaction in buildings.
I: Analytical methods”, Journal of Geotechnical and Geoenvironmental Engineering,
ASCE, vol. 125, no. 1, pp. 26–37, 1999.
[TOD 08] T ODOROVSKA M.I., T RIFUNAC M.D., “Impulse response analysis of the Van Nuys
7-storey hotel during 11 earthquakes and earthquake damage detection”, Structural Control
and Health Monitoring, vol. 15, no. 1, pp. 90–116, 2008.
[TRI 72] T RIFUNAC M., “Interaction of a shear wall with the soil for incident plane SH
waves”, Bulletin of the Seismological Society of America, vol. 62, pp. 63–83, 1972.
[TRI 10] T RIFUNAC M.D., T ODOROVSKA M.I., M ANI Ć M.I., B ULAJI Ć B.Ä., “Variability
of the fixed base and soil structure system frequencies of a building. The case of Borik 2
building”, Structural Control and Health Monitoring, vol. 17, no. 2, pp. 120–151, 2010.
[TRO 13] T ROMBETTA N., M ASON H., C HEN Z. et al., “Nonlinear dynamic foundation and
frame structure response observed in geotechnical centrifuge experiments”, Soil Dynamics
and Earthquake Engineering, vol. 50, pp. 117–133, 2013.
[TRO 14] T ROMBETTA N., M ASON H., C HEN Z. et al., “Nonlinear soil-foundation-structure
and structure-soil-structure interaction: Centrifuge test observations”, Journal of
Geotechnical and Geoenvironmental Engineering, vol. 140, no. 5, pp. 04013057-(1-11),
2014.
[VEL 71] V ELETSOS A., W EI Y., “Lateral and rocking vibration footings”, Journal of the Soil
Mechanics and Foundations Division, ASCE, vol. 97, pp. 1227–1248, 1971.
[WIR 89] W IRGIN A., “Amplification résonante du tremblement d’une chaîne de montagne
cylindrique soumise à une onde SH”, Compte rendu de l’Académie des Sciences, vol. 311,
pp. 651–655, 1989.
[WIR 96] W IRGIN A., BARD P.-Y., “Effects of buildings on the duration and amplitude of
ground motion in Mexico City”, Bulletin of the Seismological Society of America, vol. 86,
pp. 914–920, 1996.

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Urban Seismology 209

[WOL 85] W OLF J., Dynamic Soil-Structure Interaction, Prentice Hall, Englewood Cliffs,
New Jersey, 1985.
[WON 77] W ONG H., T RIFUNAC M., W ESTERMO B., “Effects of surface and subsurface
irregularities on the amplitudes of monochromatic waves”, Bulletin of the Seismological
Society of America, vol. 67, pp. 353–368, 1977.

Copyright Iste 2022 / File for personal use of Robert Charlier only
Copyright Iste 2022 / File for personal use of Robert Charlier only
 List of Authors

Benjamin CERFONTAINE Philippe GUEGUEN

School of Science and Engineering Earth Science Institute (ISTerre)
University of Dundee Grenoble Alpes University
UK CNRS
IFSTTAR
Robert CHARLIER France
Dpt. ArGEnCo
University of Liège Orianne JENCK
Belgium 3SR Laboratory
Grenoble Alpes University
Andrea COLOMBI France
Institute of Structural Engineering
Swiss Federal Institute of Panagiotis KOTRONIS
Technology GeM
Zurich Centrale Nantes
Switzerland France

Daniel DIAS Philippe ROUX

3SR Laboratory Earth Science Institute (ISTerre)
Grenoble Alpes University Grenoble Alpes University
France CNRS
IFSTTAR
Stéphane GRANGE France
GEOMAS Laboratory
INSA Diana SALCIARINI
Lyon Department of Civil and
France Environmental Engineering
University of Perugia
Italy

Copyright Iste 2022 / File for personal use of Robert Charlier only
212 Deterministic Numerical Modeling of Soil–Structure Interaction

Claudio TAMAGNINI
Department of Civil and
Environmental Engineering
University of Perugia
Italy

Pascal VILLARD
3SR Laboratory
Grenoble Alpes University
France

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Index

C, D, F L, M, N

contact mechanics, 48 lateral loading, 1, 22, 25, 37,

cyclic loading, 56, 63, 71–73 40–42, 47, 48
discrete element method, 55–57, 59, 61, macroelement, 143–145, 148–151,
62, 68–71, 76, 80, 96 153, 154, 156, 158–160, 162–165,
dynamic loading, 101, 105, 112, 120, 167, 169–174
122, 137 metamaterials, 183, 200
finite nonlinear behavior, 143–145, 147, 149,
difference method, 101, 102, 104, 151, 159, 161, 162
112, 117, 122, 126, 137 numerical modeling, 101, 109, 111,
difference model, 119, 133 117, 122, 124, 126, 127, 129, 132,
element modeling, 2 133, 135, 136, 138

G, H, I O, P, R, S

geosynthetic, 56, 76–96 offshore foundations, 1, 20, 34, 49

geotechnical partially drained behavior, 2, 7, 48
engineering, 101, 102, 105, 108, 112, piles, 56, 61–68, 70–73, 75, 76
138 vertical rigid, 80, 121
structure, 55, 56, 61, 76 plasticity, 144, 147, 151, 172
hydro-mechanical couplings, 1, 2, 6, 30, reinforcement, 55, 56, 59–62, 76, 77, 84,
37, 43, 48 87, 89, 96
hypoplasticity, 145, 149, 151–154, 156, retaining walls, 102, 106, 107, 110, 112
158–160, 162, 165, 167, 171–173 seismic risk, 199, 202
interface, 1–3, 5–11, 13–20, 22–24, shallow foundation, 144, 146–148, 154,
26–30, 32–35, 38, 43, 46–49, 56, 160, 172
59, 60, 63, 65, 70, 72, 77, 84, site–city interaction (SCI), 197, 199,
88–90, 96 202

Copyright Iste 2022 / File for personal use of Robert Charlier only
214 Deterministic Numerical Modeling of Soil–Structure Interaction

soil improvement, 102, 121, 129 T, U, W

soil–structure interaction (SSI), 101–105,
tunnels, 102, 106, 109–112, 115–121
107, 108, 111–113, 115, 116, 119, 120,
urban seismology, 181, 182, 183
123, 138
waves, 181, 182, 185–187, 189, 190, 193,
static loading, 102, 119
196, 198–203
structural dynamic, 143
suction
anchors, 2, 20, 21, 37, 49
caisson, 1, 2, 7, 20–22, 24,
34, 48

Copyright Iste 2022 / File for personal use of Robert Charlier only
 Other titles from

in
Numerical Methods in Engineering

2021
GENTIL Christian, GOUATY Gilles, SOKOLOV Dmitry
Geometric Modeling of Fractal Forms for CAD
(Geometric Modeling and Applications Set – Volume 5)

2020
GEORGE Paul Louis, ALAUZET Frédéric, LOSEILLE Adrien, MARÉCHAL Loïc
Meshing, Geometric Modeling and Numerical Simulation 3: Storage,
Visualization and In Memory Strategies
(Geometric Modeling and Applications Set – Volume 4)
SIGRIST Jean-François
Numerical Simulation, An Art of Prediction 2: Examples

2019
DA Daicong
Topology Optimization Design of Heterogeneous Materials and Structures

Copyright Iste 2022 / File for personal use of Robert Charlier only
GEORGE Paul Louis, BOROUCHAKI Houman, ALAUZET Frédéric,
LAUG Patrick, LOSEILLE Adrien, MARÉCHAL Loïc
Meshing, Geometric Modeling and Numerical Simulation 2: Metrics,
Meshes and Mesh Adaptation
(Geometric Modeling and Applications Set – Volume 2)
MARI Jean-Luc, HÉTROY-WHEELER Franck, SUBSOL Gérard
Geometric and Topological Mesh Feature Extraction for 3D Shape Analysis
(Geometric Modeling and Applications Set – Volume 3)
SIGRIST Jean-François
Numerical Simulation, An Art of Prediction 1: Theory

2017
BOROUCHAKI Houman, GEORGE Paul Louis
Meshing, Geometric Modeling and Numerical Simulation 1: Form
Functions, Triangulations and Geometric Modeling
(Geometric Modeling and Applications Set – Volume 1)

2016
KERN Michel
Numerical Methods for Inverse Problems
ZHANG Weihong, WAN Min
Milling Simulation: Metal Milling Mechanics, Dynamics and Clamping
Principles

2015
ANDRÉ Damien, CHARLES Jean-Luc, IORDANOFF Ivan
3D Discrete Element Workbench for Highly Dynamic Thermo-mechanical
Analysis
(Discrete Element Model and Simulation of Continuous Materials Behavior
Set – Volume 3)
JEBAHI Mohamed, ANDRÉ Damien, TERREROS Inigo, IORDANOFF Ivan
Discrete Element Method to Model 3D Continuous Materials
(Discrete Element Model and Simulation of Continuous Materials Behavior
Set – Volume 1)

Copyright Iste 2022 / File for personal use of Robert Charlier only
JEBAHI Mohamed, DAU Frédéric, CHARLES Jean-Luc, IORDANOFF Ivan
Discrete-continuum Coupling Method to Simulate Highly Dynamic
Multi-scale Problems: Simulation of Laser-induced Damage in Silica Glass
(Discrete Element Model and Simulation of Continuous Materials Behavior
Set – Volume 2)
SOUZA DE CURSI Eduardo
Variational Methods for Engineers with Matlab®

2014
BECKERS Benoit, BECKERS Pierre
Reconciliation of Geometry and Perception in Radiation Physics
BERGHEAU Jean-Michel
Thermomechanical Industrial Processes: Modeling and Numerical
Simulation
BONNEAU Dominique, FATU Aurelian, SOUCHET Dominique
Hydrodynamic Bearings – Volume 1
Mixed Lubrication in Hydrodynamic Bearings – Volume 2
Thermo-hydrodynamic Lubrication in Hydrodynamic Bearings – Volume 3
Internal Combustion Engine Bearings Lubrication in Hydrodynamic
Bearings – Volume 4
DESCAMPS Benoît
Computational Design of Lightweight Structures: Form Finding and
Optimization

2013
YASTREBOV Vladislav A.
Numerical Methods in Contact Mechanics

2012
DHATT Gouri, LEFRANÇOIS Emmanuel, TOUZOT Gilbert
Finite Element Method
SAGUET Pierre
Numerical Analysis in Electromagnetics

Copyright Iste 2022 / File for personal use of Robert Charlier only
SAANOUNI Khemais
Damage Mechanics in Metal Forming: Advanced Modeling and Numerical
Simulation

2011
CHINESTA Francisco, CESCOTTO Serge, CUETO Elias, LORONG Philippe
Natural Element Method for the Simulation of Structures and Processes
DAVIM Paulo J.
Finite Element Method in Manufacturing Processes
POMMIER Sylvie, GRAVOUIL Anthony, MOËS Nicolas, COMBESCURE Alain
Extended Finite Element Method for Crack Propagation

2010
SOUZA DE CURSI Eduardo, SAMPAIO Rubens
Modeling and Convexity

2008
BERGHEAU Jean-Michel, FORTUNIER Roland
Finite Element Simulation of Heat Transfer
EYMARD Robert
Finite Volumes for Complex Applications V: Problems and Perspectives
FREY Pascal, GEORGE Paul Louis
Mesh Generation: Application to finite elements – 2nd edition
GAY Daniel, GAMBELIN Jacques
Modeling and Dimensioning of Structures
MEUNIER Gérard
The Finite Element Method for Electromagnetic Modeling

2005
BENKHALDOUN Fayssal, OUAZAR Driss, RAGHAY Said
Finite Volumes for Complex Applications IV: Problems and Perspectives

Copyright Iste 2022 / File for personal use of Robert Charlier only
Copyright Iste 2022 / File for personal use of Robert Charlier only