Computers and Geotechnics 54 (2013) 16–32

Contents lists available at SciVerse ScienceDirect

Computers and Geotechnics

journal homepage: www.elsevier.com/locate/compgeo

An analytical model of soil–structure interaction with swelling soils

during droughts
Emad Jahangir a,b,⇑, Olivier Deck b, Farimah Masrouri a
a
Laboratoire Environnement Géomécanique & Ouvrages (LAEGO), Université de Lorraine B.P. 40 Rue M. Roubault, 54501 – Vandoeuvre-lès-Nancy, France
b
Georessources, UMR 7359, Ecole des Mines, Université de Lorraine, Campus Artem, CS14234, 54042 Nancy Cedex, France

a r t i c l e i n f o a b s t r a c t

Article history: Lightly loaded structures constructed on expansive soils may develop structural damage as a result of
Received 18 October 2012 changes in the soil’s moisture content. This study investigated an analytical model of soil–structure inter-
Received in revised form 22 March 2013 action to assess the settlement of dwellings built on swelling soils when droughts occur. The building
Accepted 29 May 2013
behavior was investigated with the Euler–Bernoulli beam theory, and the ground behavior was investi-
Available online 1 July 2013
gated with a Winkler-derived model based on the state surface approach. The analytical model results
were compared to those of a finite element analysis using the Barcelona Expansive Model (BExM) per-
Keywords:
formed with Code_Bright.
Soil–structure interaction
Hydro-mechanical behavior of clayey soils
The analytical model was then used to assess the settlement transmission ratio for a typology of clayey
Shrinkage and swelling of clayey soils soils and different parameters of building. The results indicated that the final deflection of the building
Building deflection increased with the building length and soil suction. The building deflection due to the suction variations
was inversely proportional to the load, the rigidity of the building and the embedding depth of the foun-
dation. Increasing these parameters made the building less vulnerable to shrinkage and swelling action.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction interaction (the mechanical part), with a coupling between the

Shrinkage and swelling of clayey soils are known to be costly
geohazards around the world. The study of their impacts on build-
ings for risk management has raised many questions because of
clayey soils’ complex hydro-mechanical behavior and susceptibil-
ity to the soil–structure interaction phenomenon. In moderate cli-
mates such as that of France, the soil is usually close to its
saturated state, and the maximum volume changes in soils occur
during dry seasons, when the changes in water content are great-
est. These volume changes cause differential settlements beneath
foundations due to different variations in moisture content under
the edges of a building and its center.
A better understanding of the behavior of swelling soil and a
building undergoing a drying (or wetting) phase is therefore cru-
cial to the effective design of shallow foundations and buildings
on swelling soils and to assessing existing buildings’ vulnerability.
In unsaturated clayey soils, the ground settlement during a drying
(or wetting) phase is a consequence of both the variation in suction
(negative pressure in soil) due to weather conditions (the hydraulic
part) and the variation of vertical stresses due to soil–structure
⇑ Corresponding author. Tel.: + 33 6 80 94 85 96; fax: + 33 383596300.
E-mail addresses: Emad.jahangir@univ-lorraine.fr (E. Jahangir), Olivier.deck@
univ-lorraine.fr (O. Deck), Farimah.masrouri@univ-lorraine.fr (F. Masrouri).
hydraulic and mechanical parts.
The hydro-mechanical behavior of unsaturated expansive soils
has been studied by several authors (e.g., [2–6]). They have con-
cluded that the swelling behavior of unsaturated expansive clays
can be described as a coupled response of the soil to suction
changes and applied stresses. Few models integrate this coupled
hydro-mechanical response into a unified framework, but one
commonly used model that does is the Barcelona Expansive Model
(BExM), proposed by Alonso et al, with 22 parameters [1]. This
model can be considered as a theoretical reference framework for
the study of the expansive behavior of unsaturated clays. The state
surface approach, which is a simpler but efficient method of linking
the volume change to two independent stress state variables, net
stress (r  ua) and suction (ua  uw), was proposed by Matyas
and Radhakrishna [7] and has been used by other authors [3,5,8–
11]. These researchers have shown that net stress and suction, as
the stress state variables, are necessary to describe the hydro-
mechanical behavior of an unsaturated soil. The state surface ap-
proach has been used as a simplified method for solving practical
problems with a simple stress path [12]. However, because this ap-
proach has a unique constitutive surface, it cannot be used to de-
scribe the effect of stress-path and hysteresis behavior.
Fredlund and Morgenstern [8] performed a series of monotonic
loading stress paths on unsaturated swelling Regina clay and
concluded that a unique state surface can be considered under

0266-352X/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compgeo.2013.05.009
 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32
monotonic loading conditions. However, each cycle requires a sep-
arate state surface in the case of cyclic hydraulic solicitations.
Zhang and Lytton [12] recently presented a modified state surface
approach that considers stress-path dependency and suction-hard-
ening behavior. This approach produced results similar to those
obtained using the BExM model. Some authors [12–14] have re-
ported that estimating volume change using the state surface ap-
proach is adequate for most engineering analysis.
Building damage due to ground movements is most commonly
associated with differential settlement [15,16]. However, differen-
tial settlement is a consequence of soil–structure interaction and
cannot be assessed without accounting for the mechanical behav-
ior of the soil and the structure. Many researchers modeled the
building as a beam resting on the ground [16–19] to assess the
transmission ratio of ground movements to the building. Depend-
ing on the building stiffness, the vertical stresses transmitted by
the building to the ground change during ground settlement
(soil–structure interaction). A flexible building follows the ground
settlement with minor changes in the vertical transmitted stresses,
while a rigid building can resist settlement and cause a redistribu-
tion of the vertical stresses under the foundations. Consequently, a
transmitted deflection ratio D/D0 is defined ([17,19]) as the ratio
between the building deflection D due to the ground movements
with soil–structure interaction and the deflection D0 calculated
using the free-field ground movement without any soil–structure
interaction. D0 is then calculated under the assumption that the
free-field ground movements are integrally transmitted to the
building (Fig. 1).
The differences between free-field ground shrinkage and build-
ing-induced deflection depend on the ground and building stiff-
ness, the building length, the vertical load, the foundation depth
and the swelling capacity of the soil.
Existing soil–structure interaction models for expansive soils
are mostly based on the Winkler model and consider the ground’s
initial mound shape due to shrinkage (or swelling). Nelson and
Miller [20] have summarized these existing models, while other
researchers [21,22] have estimated the shrinkage at the extremi-
ties of foundation slabs with empirical methods and have investi-
gated the effect of the shaped form of the ground (due to
shrinkage) on building behavior using numerical methods. In such
a framework, the influence of vertical stress changes beneath the
foundation during the shrinkage phase on the hydraulic parame-
ters of the soil is not included when calculating the final transmit-
ted settlement.
The focus of this study was on the role of soil–structure interac-
tion in hydro-mechanical coupling. The building’s mechanical
behavior was modeled using Euler–Bernoulli beam theory, and
the hydro-mechanical behavior of the soil was modeled with the
state surface approach. The challenge in the proposed model was
to incorporate the hydro-mechanical behavior of clayey soils
undergoing water content changes in classical soil–structure inter-
action models.
(a) (b)
Fig. 1. Building deflection induced by soil shrinkage during a dry period: (a) Initial state of layers, (b) drying evolution under the building, and (c) final state of layers after
undergoing suction and final building deflection (em: edge moisture change, Za: active depth, D0: free field settlement, D: building deflection).
17
In this study, the deformation due to the loading phase was as-
sumed to have been stabilized before the drying phase [23] to
study the existing building’s behavior. Only the settlement (shrink-
age) caused by the drying phase was considered to study the con-
sequences of drought on the building. The soil was considered to
be a homogenous and isotropic medium, and its variability be-
neath the building was not considered.
The following sections describe the different components of the
model:
– the hydro-mechanical behavior of swelling clays, modeled with
the state surface approach;
– the vertical and horizontal suction change profile in the soil;
– the building stiffness and vertical stress in the ground, modeled
with the Euler–Bernoulli beam theory and Boussinesq relation;
and
– the combination of the different models and solution of the con-
stitutive equation.
2. Description of the model
Fig. 1 shows a schematic of building deflection induced by soil
shrinkage during a drying period.
When a drying period occurs, the suction variation is greatest at
the extremity of a building, where the soil dries easily, and is neg-
ligible at the center of the building [24]. Fig. 1b shows the distance
em under the building undergoing the suction variations. The active
depth za over which the suction varies is not negligible. This
change in suction content leads to a differential settlement of the
ground and the building between its center and its edges (Fig. 1c).
The suction profile of soil is dependent on parameters such as
the soil’s characteristics (nature, structure, particle size, retention
curve, permeability, etc.), meteorological parameters (precipitation
and evaporation rate) and local conditions, such as the groundwa-
ter level, the presence of vegetation, etc. The building can also af-
fect the evolution of a suction (or moisture) profile. This
evolution is discussed further in Section 2.3.
In the proposed model, the ground was divided into several lay-
ers to account for variations in suction and vertical stresses with
depth (Fig. 1). The void ratio variation Dei was calculated at the
middle of each layer with the state surface approach by consider-
ing hydro-mechanical coupling (Section 2.1) and soil–structure
interaction (Section 2.4). The final settlement of each layer Dhi
was then calculated, and the total settlement of the ground surface
was obtained using following equation.
Xn Xn
De i
D¼ Dhi ¼ hi ð1Þ
1 þ e0i
i¼1 i¼1
where e0i is the initial void ratio of layer i, Dei is the variation in void
ratio, and hi is the initial thickness of the ith layer. Fig. 1 shows a
general schematic of the model in its initial state (Fig. 1a) and after
undergoing suction variation (Fig. 1b and c). The model was
(c)
18 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32

considered in plane strain and assumed to be symmetrical, so only
half of the soil and building are presented in Fig. 1.
Eq. (1) corresponds to oedometric conditions because the results
of suction-controlled tests on expansive soils are mainly obtained
with oedometer tests. Moreover, Lytton et al. [25] showed that
the results of suction-controlled oedometer tests do not underesti-
mate the amplitude of axial settlement under a building because all
of the strains produced by drying are considered to be axial strains
in this test. Thus, the results of the calculations in this study (based
on oedometer test results) were obtained in the interest of safety.
However, this issue could be the subject of further investigations
comparing triaxial, in situ and oedometric conditions.
2.1. Modeling the hydro-mechanical behavior of swelling clays
unknown vertical distribution p(x) that corresponds to the ground
reaction (Fig. 3). I is the second moment of area of the cross section,
with I = BH3/12.
The equation of the beam deflection is a second-order differen-
tial equation that relates the vertical deflection y(x) and the bend-
ing moment M(x) in the x-abscised cross section:
2
MðxÞ d yðxÞ
 ¼ 2
ð3Þ
EI dx
Taking into account the relationship between the shear force
and the bending moment (Eq. (4)) and the relationship between
the shear force and the external load (Eq. (5)), Eq. (3) can be rewrit-
ten as Eq. (6). Fig. 3a shows the applied load (q), ground reaction
(p(x)) and the building deflection (y(x)).

The hydro-mechanical behavior of swelling clays is modeled dMðxÞ

VðxÞ ¼  ð4Þ
using the state surface approach, which describes the void ratio e dx
as a function of the net stress (r  ua) and the suction s = (ua  uw),
where ua and uw are the air pressure and water pressure in the soil, dVðxÞ
¼ q  pðxÞ ð5Þ
respectively, and r is the total stress. dx
Vu [26] proposed different functions to fit the void ratio consti-
4
tutive surfaces of an unsaturated expansive soil and tested these d yðxÞ q pðxÞ
¼  ð6Þ
functions on swelling clay from Regina (Saskatchewan, Canada). dx
4 EI EI
In this study, the first function proposed by Vu [26] was considered
In the final state, the soil and the building are in static equilib-
because of its reasonable number of parameters (3) and its ability
rium, and the sum of the applied loads must be equivalent to the
to model the experimental results:
soil reaction in the first layer p1(x), where the building is in contact
e ¼ a þ b log½1 þ ðr  ua Þ þ cðua  uw Þ ð2Þ with the ground. This point is used to verify the response:
Z L=2
where a is the void ratio at zero net stress and suction, parameter b
controls the total volume change due to suction and stress changes p1 ðxÞdx ¼ qL ð7Þ
L=2
and parameter c controls the rate of volume change during a varia-
tion in suction and may be related to the swelling characteristics of Fig. 3b shows the model parameters and expected beam deflec-
the soil, such as the plasticity index [27]. tion after being subjected to drying action at its extremities. In this
Fig. 2a shows the state surface used for the Regina clay. Hydro- figure, w(x) is the soil settlement, y(x) is the deflection of the beam,
mechanical coupling was considered in this state surface through which is considered negative when the shape is convex (according
the compressibility of soil with respect to the applied load (av1) to beam theory convention), and d is the displacement of the rigid
and the suction (av2). These parameters (av1 and av2) decrease body. Eq. (8) is used to ensure compatibility between w(x), y(x) and
when the respectively applied load and suction increase. The d:
occurrence of this hydro-mechanical coupling under the building
wðxÞ  yðxÞ ¼ d ð8Þ
is explained in Fig. 2b.

2.2. Building model 2.3. Suction profile

The study building was modeled as an elastic Euler–Bernoulli
beam with length L, height H, width B and Young’s modulus E.
The analytical model was developed to describe the relationship
between the beam’s deflection and the applied load and calculate
the final deflection ratio (D/L) due to soil shrinkage that the build-
ing must resist before suffering significant damage. The beam is
loaded by a vertical uniform load q (self-weight in N/m) and an
The final deflection ratio of a building due to the swelling or
shrinkage of a clayey soil depends on the vertical and horizontal
evolution of the suction profile (Fig. 1). This deflection increases
when the suction variation extends deeper into the ground (in-
crease in za) and farther under the building (increase in em). In gen-
eral, the suction change is greatest at the surface, where it can
reach a few megapascals, and decreases with depth [31].

(a) (b)

Fig. 2. (a) State surfaces for expansive clay proposed by Vu [26] and fitted for Regina clay, and (b) hydro-mechanical coupling.
 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32 19

(a) (b) L

em em

Initial position of the beam

Final position w(x)

d
p 1 (x) : ground reaction (Rigid displacement)
-w(y)

Fig. 3. Definition of (a) the building load and ground reaction and (b) the building deflection.

Two relationships are used to consider the suction evolution
with depth and em for vertical and horizontal suction profiles,
respectively.
2.3.1. Vertical profile
Several empirical methods exist in the literature to estimate
vertical suction variation and the active depth beneath buildings.
These methods are based mainly on the diffusion equation ob-
tained by applying Darcy’s law and the conservation of mass prin-
ciple [28]:
@ 2 S @ 2 S @ 2 S f ðx; y; z; tÞ 1 @S
þ þ þ ¼
@x2 @y2 @z2 P a @t
where S is the total suction (kPa), P is the saturated permeability
(m/year), f(x, y, z, t) is the humidity (drying) source with respect
to time and a is the diffusion coefficient of the soil (m2/year).
McKeen and Johnson [29] proposed the following cyclic harmonious
response for a one-dimensional condition (vertical evolution
equation):
Sðz; tÞ ¼ Se
  0;5     0;5  
np np
 DSexp  z cos 2npt  z ð10Þ
a a
where S(z, t) is the suction as a function of depth and time, Se is the
equilibrium suction below the active zone (the average of the sea-
sonal maximum and minimum suction), DS is the amplitude of suc-
tion variation at the surface, n is the frequency of seasonal
variations (cycles per year), t is the time (years) and z is the depth
(m).
Lytton [11] proposed an empirical equation to calculate the dif-
fusion coefficient a, which plays an important role in the determi-
nation of both the vertical and horizontal suction profiles:
a ¼ 0:0029  0:000162ðSsÞ  0:0122ðav 2 Þ
1
where Ss (kPa ) is the slope of the suction–water content curve
and av2 (kPa1) is the coefficient of compressibility with respect
to suction change. Ss can be estimated using Eq. (12) (Lytton 1994):
Ss ¼ 20:29 þ 0:1555ðLL%Þ  0:117ðPI%Þ
þ 0:0684ð%No:200Þ
where LL is the liquid limit in percent, PI is the plasticity index in
percent and%No.200 is the percentage of soil passing a U.S. No.
200 sieve, equivalent to a grain size of 75 lm.
The time-independent envelope of the vertical suction profile
can easily be determined by taking the derivative of Eq. (10) with
respect to time:
  0:5 
np
SðzÞ ¼ Se þ DS exp  z ð13Þ
a
The difference between the maximum and the minimum enve-
lopes for a given depth can be considered the maximum suction
change, DSmax:
  0:5 
np
DSmax ðzÞ ¼ DS exp  z ð14Þ
a
If the maximum suction change DSmax can be determined, its
value gives the depth of the seasonal active zone when substituted
into Eq. (14) as follows:
 
DS
ln DSmax
za ¼ pnffiffiffiffip ð15Þ
a
The framework of Eq. (10) is defined in terms of two suction lev-
els, Ds and Se. It is assumed that Se is the average of the wet and dry
ð9Þ
extreme suctions Swet and Sdry. This assumption is consistent with a
climatic pattern consisting of wet and dry periods of approxi-
mately equal duration. However, the equilibrium suction Se is
much closer to the wet extreme of the suction, Swet, in a humid cli-
mate and is much closer to the dry extreme, Sdry, in an arid climate.
Using a time-weighted average of the measured surface suction
in McKeen and Johnson’s database [29], Aubeny and Long [30] pro-
posed a climate parameter (r) that characterizes the suction
boundary conditions as follows:
r ¼ t dry n ¼ ðSe  Swet Þ=ðSdry  Swet Þ ð16Þ
where tdry is the duration of the drying period (years) and n is the
frequency of wetting–drying cycles during the year. Considering
the boundary values of Swet and Sdry in the literature, Aubeny and
Long [30] obtained typical values for the climate parameter
r = 0.25, 0.5, and 0.75, representing the characteristics of humid,
semi-arid, and arid climates, respectively.
By substituting the Se value calculated using Eq. (16) into Eqs.
(10) and (13), the final equations for the vertical time-dependent
suction profile and envelope can be expressed by Eqs. (17) and
(18), respectively.
ð11Þ
Sðz; tÞ ¼ ½ðt dry  nÞðSdry  Swet Þ þ Swet 
  0;5     0;5  
np np
 DSexp  z cos 2npt  z ð17Þ
a a
  0:5 
np
SðzÞ ¼ ½ðt dry  nÞðSdry  Swet Þ þ Swet  þ DS exp  z ð18Þ
ð12Þ a
To avoid underestimation of suction variation and, conse-
quently, building deflection, an envelope of the vertical suction
profile that estimated its lower and upper boundaries for wetting
and drying profiles (Eq. (18)) was used in this study. An example
of the application of Eqs. (17) and (18) can be found in [34].
2.3.2. Horizontal profile
The amplitude of the differential settlement under the building
is directly dependent on the edge moisture variation distance (em),
which is defined as the distance measured inward from the edge of
the shallow foundation along which the moisture content of soil
varies due to drying or wetting around the perimeter (Fig. 1b).
20 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32

The amplitude of suction at the extremities beneath the founda- During the first phase, the building load is applied and the con-
tion is equal to S(z), evaluated using Eq. (19), depending on the tact pressure p0(x) is calculated (using the procedure explained be-
foundation’s embedding depth. This amplitude is zero beyond a low). Then, using Boussinesq’s stress diffusion equation (Eq. (21)),
distance of em under the foundation. The suction variation shape the stress in the ground can be calculated for any position {x, z}:
is unknown throughout this distance (L/2  em < x < L/2). The shape Z B
of this variation was modeled with a parabolic function (Eq. (19)) 2z3
pi ðx; zi Þ ¼ p0 ðxÞ dB ð21Þ
because such a trend was observed in the literature [32,33,48] 0 pðx2 þ z2i Þ2
and because it was confirmed by comparison with numerical re-
sults (Section 3). where pi(x, z) is the stress state at a depth of zi, x is the distance to
the point of application, p0(x) is the contact pressure, and B is the
8 width of the building. The suction in the soil is assumed to remain
> SðzÞ when x <  2L and x > 2L
>
> constant during this first phase. The total settlement w1(x), which is
>
> SðzÞ 2
< e2  ðx  2L þ em Þ when  2L 6 x 6  2L þ em the sum of the individual layer settlements (i 2 {1, n}) due to load-
m
Sðx; zÞ ¼ ð19Þ ing, can be calculated as follows: where c is the soil unit weight,
>
> 0 when  2L þ em 6 x 6 2L  em
>
> and S0(x, z) is the initial suction of each layer, which corresponds
>
: SðzÞ 2
e2m
 ðx  2L þ em Þ when L
2
 em 6 x 6 2L to Shumid at the surface and is calculated using Eqs. (18) and (19)
at depth zi. The initial stress pi0(x, y) corresponds to the in situ stress

Xn Xn
De i
w1 ðxÞ ¼ wi ðx; zi Þ ¼ hi
i¼1 i¼1
1 þ e0
X n 
log½1 þ pi1 ðx; zi Þ=B þ c  S0i ðx; zi Þ  log½1 þ cðhi =2Þ þ c  S0i ðx; zi Þ  hi
w1 ðxÞ ¼ b
i¼1
1 þ a  log½1 þ cðhi =2Þ þ c  S0i ðx; zÞ ð22Þ
0 3
1
2zi
X log½1 þ p 2 =B þ c  S0i ðx; zi Þ  log½1 þ cððzi  zi1 Þ=2Þ þ c  S0i ðx; zi Þ  ðzi  zi1 Þ
n
B 01 pðxþzi Þ C
¼b @ A
i¼1
1 þ a  log½1 þ cððzi  zi1 Þ=2Þ þ c  S0i ðx; zi Þ

where S(z) is the amplitude of the suction at a depth (z), which is
calculated using Eq. (18), L is the building length and em is the edge
moisture change, which is evaluated analytically using the method
proposed by El-Garhy and Wray [32]. These researchers studied
edge moisture variation distance under foundations for seven dif-
ferent sites with different climatic conditions in the US and Austra-
lia. They presented the following empirical equation (Eq. (20)) for
the edge moisture variation distance:
 DS
at the middle of each layer i. The rigid displacement of the beam (d
in Fig. 3b) is the sum of the individual layer settlements under the
edges of the building:
d1 ¼ w1 ðL=2Þ ¼ w1 ðL=2Þ ð23Þ
According to the beam theory convention used in Section 2.2
(Eq. (8)), settlement and rigid displacement may be related to the
deflection of the beam during the first phase:
C2
em ¼ C1 log 0:1
C1 ¼ 2:204ðaÞ0:0697
C2 ¼ 0:633ðaÞ0:0778
where DS is the amplitude of the change in suction on the surface, a
is the diffusion coefficient of the soil and C1 and C2 are constants.
Further information and applications of this suction profile
determination method can be found in Jahangir et al. [34].
2.4. Constitutive equation of the model and resolution procedure
y1 ðxÞ ¼ w1 ðxÞ  d1 ð24Þ
ð20Þ
By substituting Eqs. (22)-(24) into the beam equation (Eq. (6)),
p01 becomes the unknown of the first phase equation (Eq. (25)) and
can be calculated by applying the boundary conditions (the resolu-
tion procedure is explained in the next section). Once this pressure
is known, it is possible to calculate the ground reaction in each
layer by accounting for the distribution of stress (Eq. (21)) in the
soil, and the values of y1(x), w1(x) and d1 are obtained by backcal-
culation. A positive deflection is expected at this stage:

d4 ðw1 ðxÞd1 Þ
The calculation of the soil’s and building’s final settlements at dx4
¼ EIq  p01EIðxÞ
4
ð25Þ
static equilibrium (the final state shown in Fig. 3b) can be obtained d ðw1 ðxÞw1 ðL=2ÞÞ
¼ EIq  p01EIðxÞ
dx4
by combining the expressions for the state surface (Eq. (2)), the
vertical stresses (Eq. (21)), the beam equation (Eq. (6)), the suction This equation is solved using the finite difference method, as
profile (Eqs. 20) and the general settlement equation for a multi- shown in the next section. In the second step, a drying phase
layer soil system (Eq. (1)). The final product of this combination (S1(x)) is applied based on the suction profile. The stress state in
is shown in Eqs. (22) and (26). The constitutive equation of the first the soil p01 calculated in the first step is considered in the second
phase (loading) and second phase (drying) are described below. phase as follows:

0 h i h i1
Xn log 1 þ pi2 ðx; zi Þ þ c  s1i ðx; zi Þ  log 1 þ pi1 ðx; zi Þ þ c  s0i ðx; zi Þ
w2 ðxÞ ¼ b @ A  hi
i¼1
1 þ a  log ½1 þ pi1 ðxÞ þ c  s0i ðxÞ  hi
0 h  . i h i1 ð26Þ
2z3i
Xn
B log 1 þ p02 pðxþzi Þ2 B þ c  s 1i ðx; z i Þ  log 1 þ p i1 ðxÞ þ c  s 0i ðx; z i Þ C
¼b @ A  hi
i¼1
1 þ a  log ½1 þ pi1 ðxÞ þ c  s0i ðx; zi Þ
 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32 21

d2 ¼ w2 ðL=2Þ ¼ w2 ðL=2Þ ð27Þ The system of equations is overdetermined because six bound-
ary conditions exist for a system of equations with four unknowns.
y2 ðxÞ ¼ w2 ðxÞ  d2 ð28Þ A checkpoint was used on the results obtained by applying differ-
ent boundary conditions to choose the most appropriate among
4
d ðw2 ðxÞ  w2 ðL=2ÞÞ q p02 ðxÞ them.
4
¼  ð29Þ Verification of the results was performed by (i) checking the
dx EI EI
shear force at the middle of the beam, which should be zero in
where S1(x, z) is the drying suction value that corresponds to Sdry at equilibrium; (ii) checking the soil reaction integral throughout
the surface, which is calculated using Eqs. (18) and (19) for a certain the contact distance, which should be equal to the applied load
depth, and p02(x) is the new stress state, which is unknown in this on the beam (Eq. (7)); and (iii) controlling the symmetric form of
step. Substituting Eqs. (26)-(28) into Eq. (6) produces the final equa- the bending moment. Eqs. (33) and (34) are the relevant boundary
tion (Eq. (29)), and once this equation is resolved, the new stress conditions that satisfy the required checkpoints.
state in the soil is determined, allowing for the calculation of
w2(x), y2(x) and d2. The final settlement due to suction variation is
the difference between y1(x) and y2(x): 2.6. Results

YðxÞ ¼ y1 ðxÞ  y2 ðxÞ ð30Þ The building deflection is estimated by applying the relevant
The final building deflection ratio (D/L) is then calculated using boundary conditions to Eqs. (25) and (29). Example results are
equation: illustrated in Fig. 4, with an emphasis on the building loading ef-
fect. In addition, following a sensitivity analysis of the discretiza-
D=L ¼ ½YðL=2Þ  Yð0Þ=L ð31Þ tion interval, checking the integral of the load q and the ground
It should be noted that if the contact pressure p(x) is negative, a reaction p(x) under the building (Eq. (7)), an interval Dx equal to
theoretical loss of contact between the soil and the foundation can 0.03 m was used in this study for all calculations.
be considered. The deflection of a building on the Regina soil [26] as a result of
the loading and drying phases was calculated for EI = 1 GN m2,
2.5. Resolution by the finite difference method L = 15 m, q = 50 kN/m and 100 kN/m, a suction variation
DS = 1 MPa and state surface parameter values a = 1.186,
The analytical solution of Eqs. (25) and (29) is complex, so the b = 0.092 and c = 0.610. Fig. 4 shows that the building deformation
finite difference method was used to transform the fourth-order due to the load increases and the deformation due to the suction
partial differential equations into a set of linear equations associ- variation decreases when the building load q increases. The final
ated with the boundary conditions. Eq. (32) presents the corre- deflection is the difference between these two phases and de-
sponding algebraic equations for different orders of differential creases as the building load increases (Eq. (31)). Therefore, the final
equations using a forward operator. deflection considered in assessing the building damage corre-
8  9 sponds to the difference between the building deformation due
yjþ1 yj
> dy > to the load (the first phase) and the final deformation due to the
> dx j ¼ dx
>
>
>
>
>
>
>   >
> load and suction variation (the second phase).
>
> d2 y y 2y þy >
>
>
< dx2 ¼ jþ2 dx2jþ1 j >
=
j
ð32Þ 3. Comparison of the model with BExM
>
>
3 y 3y þ3yjþ1 yj >
>
>
> ðddx3yÞ ¼ jþ3 jþ2
dx3 >
>
>
>  4 j >
>
>
> >
> This section compares the proposed analytical model with a
>
: d 4y ¼ jþ4 jþ3 dx4jþ2 jþ1 j >
y 4y þ6y 4y þy
;
dx j numerical model studied with Code_Bright finite element software
The finite difference method consists of the discretization of the and the BExM model [1] for expansive soil behavior. This model
length of beam (L) into n points. The number of unknown values is was implemented in Code_Bright by Mrad [35]. The BExM model
n. y(x) was chosen as the unknown function. When applying this considers the micro-structural features of swelling soils to esti-
method, one of the discretized points is removed at each derivation mate the hydro-mechanical response for both cyclic (wetting–dry-
according to Eq. (32). Therefore, (n  4) independent equations can ing) and monotonic solicitations.
be written for the fourth derivative, and four boundary conditions The developed analytical model is intended to estimate the vol-
are required to solve the problem. Six boundary conditions, similar umetric hydro-mechanical response in a general way without deal-
to those used by Deck and Singh [19], can be defined for a beam in ing with micro- and macro-structure contributions to the volume
static equilibrium under a distributed load q and reaction p(x), change during a hydraulic solicitation.
when there is no loss of contact: The reference soil used in this section is Le Deffend (France).
This soil was subjected to suction-controlled compression testing
 Deflection at the extremities: the building deflection is equal to by Nowamooz et al. [36]. Both the analytical state surface-based
zero at the beam extremities when the beam displacement is model and the BExM model were calibrated to these test results
equal to the rigid vertical displacement: to establish the model parameters.
The hydraulic and mechanical parameters of the Le Deffend soil
yðL=2Þ ¼ yðL=2Þ ¼ 0 ð33Þ were characterized for the BExM model by Nowamooz et al. [36].
Table 1 presents the hydraulic parameters of this soil, essentially
 Bending moment at the extremities: the bending moment at the for the hydraulic conductivity (using Mualem’s model [37]) and
beam extremities is equal to zero because the beam is in static soil–water retention curve (using Van Genuchten’s model [38]).
equilibrium under load q and reaction p(x): Table 2 presents the mechanical and yield surface parameters that
yð2Þ ðL=2Þ ¼ yð2Þ ðL=2Þ ¼ 0 ð34Þ describe the evolution of the pre-consolidation pressure, based on
 Shear at the extremities: the shear force at the beam extremi- the applied suction in the BExM model described by equation:
ties is equal to zero because the beam is in static equilibrium  kð0Þj
under loads q and p(x) with no concentrated forces: p0 kðsÞj
ð3Þ ð3Þ p0 ¼ pc ð36Þ
y ðL=2Þ ¼ y ðL=2Þ ¼ 0 ð35Þ pc
22 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32

Fig. 4. (a) Building deflection for both loading and drying phases and (b) final deflection for drying phase.

where j is the slope of the unloading–reloading line associated
with the mean net stress, p0 is the pre-consolidation stress at suc-
tion s, pc is the reference stress, p0 is the saturated pre-consolidation
pressure and k(s) is the slope of the virgin compression line at suc-
tion s, which can be calculated using equation:
The values of the state surface parameters were obtained by
curve fitting of the suction-controlled compression test results
Table 2
Mechanical parameters of the material used [36].

kðsÞ ¼ kð0Þ½r þ ð1  rÞesb  ð37Þ Parameter Description Value

p0 Initial saturated preconsolidation stress 600 kPa
where b and r are the two model parameters that control the soil pc Reference stress 200 kPa
stiffness related to the suction level. More details about this model j Elastic compression index 0.04
k(0) Saturated virgin compression index 0.19
can be found in Alonso et al. [1].
r Parameter that defines the limiting value of the 0.70
compression index for high suctions
b Parameter that defines the curvature of the LC 0.60 MPa1
Table 1 yield curve
Hydraulic parameters characterizing the material used [36]. M Slope of the critical state 0.57

Parameter Value
Saturated hydraulic conductivity, ks 8  1012 m/s
Parameters of the soil–water retention curve Se ¼ ½1 þ ðasÞn m  Table 3
(van Genuchten, 1980) Soil and building parameters.
1
Model parameter a 0.013 MPa
Model parameter n 3.010 Parameter Value
Model parameter m = 1  1/n 0.668 L 10 m
pffiffiffiffiffi B 1m
Parameters of the hydraulic conductivity k ¼ ks Se ½1  ð1  Se1=m Þm 2
curve (Mualem, 1976) S S
EI 1 GN m2
r rðresÞ
With Se ¼ SrðsatÞ SrðresÞ q 100 kN/m
Model parameter m 0.668 Foundation depth 0m
Residual degree of saturation, Sr(res) 0.1 Le Deffend soil State surface parameters (a = 1.361, b = 0.091, c = 0.085)
Maximum degree of saturation, Sr(sat) 1 Suction profile Sdry = 1000 kPa, Swet = 10 kPa, Se = 180 kPa
 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32 23

[36]. The only direct link between BExM’s parameters and those of The building height was 0.5 m, and the Young’s modulus was
the used state surface is expressed by the following equation: 96 GPa, yielding the same flexural rigidity that was used for the
 analytical model (EI = 1 GN m2). The soil and the building were dis-
@e  b cretized with four-node-quadrilateral finite elements, and the final
j¼ ¼ ð38Þ
@ rs;r!0 log10 mesh was made up of 1700 elements and 1786 nodes. The vertical
The state surface parameters for Le defend soil are: a = 1.361, mesh size was smaller near the surface (5 mm), where the rate of
b = 0.091 and c = 0.085(Table 3). The parameter j evaluated using flow (infiltration or evaporation) was greater, and increased up
Eq. (39) for Le Deffend soil is equal to 0.0399 (0.434 * 0.091). The to 0.25 m in depth for greater accuracy.
same value is reported in table 2 for the BExM model. According to the hydraulic boundary conditions (Fig. 5), the flux
The same building characteristics were used for both models: a causing the drying phase was imposed on the upper free surface of
10-m-long building with q = 100 kN/m lying on Le Deffend expan- the ground. Fig. 6 shows the suction profiles after applying a suc-
sive soil and subjected to a drying period of 3 months. The suction tion of 1 MPa at the surface for 90 days (3 months). It should
variation profile was calculated using Eqs. (18) and (19). The be noted that the soil was initially almost saturated (S0 = Swet = 10
parameters of the analytical model are summarized in Table 3. kPa). This figure compares the vertical and horizontal profiles of
The numerical model was two-dimensional with a plane–strain suction produced by Code_Bright and those produced by the ana-
assumption. The geometry, mesh, and hydraulic and mechanical lytical model described in this study (Eqs. (18) and (19)).
boundary conditions are shown in Fig. 5. Considering the model’s Fig. 6 shows that changes in the suction become negligible
symmetry, only half of the ground and building were modeled. beyond a horizontal distance of 2.5 m under the building’s

Fig. 5. Geometry, mesh and boundary conditions of the numerical model.

Fig. 6. Comparison of (a) vertical and (b) horizontal suction profiles from the numerical model and the analytical responses.

Table 4
Database of expansive soils and normalized parameters of the state surface.

Soil Reference PI LL a/e0 b/e0 c a (m2/an) em (m) za (m) Swelling a v2

Indian Head Till [39] 19.7 36.1 1 0.026 0.022 12 2.65 3.46 Intermediate 0.040
Jossigny Silt [40] 21 37 1 0.230 0.036 12.11 2.64 3.48 Intermediate 0.057
Champenoux [41] 19.6 51 1 0.128 0.070 10.94 2.60 3.5 High 0.057
Boom Clay [42] 27 56 1 0.184 0.39 11.42 2.62 3.3 High 0.045
Le Deffend [41] 32 85 1 0.066 0.085 9.06 2.55 3 Very high 0.055
Regina Clay [27] 38 70 1 0.077 0.610 10.6 2.39 2.95 Very high 0.06
FoCa Clay [43] 62 112 1 0.133 0.611 4.67 2.47 2.8 Critical 0.13
Bentonite from Cortijo [43] 49 102 1 0.114 0.905 5.53 2.43 3.12 Critical 0.138
Montmorillonite [44] 60 170 1 0.17 2.30 2.32 2.19 2.02 Critical 0.1
24 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32
extremities. This result is consistent with the value of 2.35 m for
the parameter em for the Le Deffend soil obtained using the analyt-
ical model (Table 4). The parabolic function used in Eq. (19) was
also confirmed.
Fig. 7 summarizes the results obtained with the Code_Bright
numerical model and compares them to the analytical model’s re-
sults. The models yield fairly similar results, with a final building
deflection of approximately 7 mm. The analytical model’s results
are slightly dependent on the layer thicknesses used. A minimum
layer thickness of 0.1 m was used to obtain results close to those
obtained with the numerical model.
3.1. Discussion
Considering the applied loads in this section corresponding to
the building load (100 kPa) and the saturated preconsolidation
pressure of Le Deffend soil (600 kPa in Table 2), the LC (Load Col-
lapse) surface of the BExM model was not reached, and the soil re-
mained in the elastic domain. Under these conditions, the
volumetric behavior in the BExM model is governed essentially
by the j slope, which is directly related to the state surface param-
eters (Eq. (38)). Moreover, in the BExM model, the shear strain is
also considered, and the total strain is the sum of the volumetric
and shear strains.
The BExM model is fundamentally based on isotropic stress
paths. However, for extension to anisotropic stress states in elastic-
ity, this model considers the macro-structural elastic strain as an
additional distortion component given by classical isotropic
elasticity:
dq
deeqM ¼ deq ¼
3G
where q is the deviatoric stress and G is the shear modulus. The
micro-structural elastic strain remains volumetric.
Deflection (y 1) due to mechanical load q
(a)
(c) Deflection (y2-y1) due to suction variation only
Fig. 7. Results of the Code-Bright numerical model compared to the analytical model: (a) building deflection after applying the load, (b) building deflection after applying
suction and (c) final building deflection due to the suction variation only.
The final elastic deformation is the sum of the volumetric (ev)
and deviatoric (eq) deformations:
dee ¼ dee v þ dee q
ð40Þ
dee v ¼ j1þe
ðsÞ dp
p
þ j1þe
ðs;pÞ ds
sþ0:1
This equation shows that the suction variation does not create
any shear strain under isotropic conditions. In our study, after
applying the building loads, the stress state in the soil may become
anisotropic before applying the suction variation.
The small difference between the numerical and analytical solu-
tions for step 2 (Fig. 7b), where the suction variation is applied,
may be attributed to shear deformations, which are not considered
in the analytical model. However, the good agreement of the final
results (step 3) indicates that the analytical model based on the
state surface approach could be an interesting tool for volume
change analysis in expansive soils under lightly loaded structures.
However, further investigations are needed to examine the shear
strain ratio comparing to the volumetric strain for a given suction
change in an unsaturated soil.
4. Applications
The developed model was used to determine the transmission
ratio D/D0 of differential settlement (where D0 is the free-field
shrinkage and D is the differential settlement of a building) for var-
ious building types and different swelling soils during dry periods.
The results were synthesized by introducing a relative stiffness
ð39Þ parameter that combines the building and soil stiffnesses and were
then used to estimate building damage by comparing the final
deflection ratio D/L to different threshold values [16]. Soil and
building typologies were developed to investigate D/D0.
Deflection (y 2) due to mechanical load
(b)
and suction variation
 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32 25

Non Low Intermediate High very high Critical

swelling swelling swelling swelling swelling swelling 4&6

7
3 4

2
9

2
1
8
5
3 1
10

Line A: Ip = 0.73(WL-20)

1 Intermediate Swelling LL(%)

2 High Swelling 1 : Boom Clay 6 : FoCa Clay
2 : Sol de Regina 7 : FEBEX
3 : Indian Head Till 8 : Sol du Def f end 1
3 Very High Swelling
4 : Montmorillonite Calcique grecque 9 : Sol du Def f end 2
5 : Jossigny Silt 10 : Sol de Champenoux
4 Critical Swelling

Fig. 8. Classification of swelling soils with the Casagrande diagram, with superimposition of the 10 collected soils.

4.1. Soil typology
There are several methods for classification of expansive soils in
the literature. These methods are based mainly on the Atterberg
limits (the liquid limit LL, the plastic limit PL and the plasticity in-
dex PI), the clay activity, the clay fraction, etc. [3]. The Casagrande
diagram (Fig. 8) is one of these methods that is widely used.
The typology of expansive soils used in this study was based on
the Casagrande diagram, along with suction-controlled compres-
sion test results for different swelling soils reported in the litera-
ture, to consider hydro-mechanical coupling. Table 4 presents the
data for ten different swelling soils. Four different zones associated
with different swelling capacities were identified on the Casa-
grande diagram (Fig. 8), and the state surface parameters were
approximated for each class of soil by fitting the previously dis-
cussed state surface (Eq. (2)) to the suction-controlled compression
test results.
The parameters of the surface state (Eq. (2)), the diffusion coef-
ficient (Eq. (11)), the active depth (Eq. (15)) and the parameter em
(Eq. (20)) are presented in Table 4 for the soils studied.
The swelling of soils is dependent on their initial density (void
ratio). A normalization step was performed on the parameters of
the state surface to compare soils with the same swelling charac-
teristics. Therefore, these parameters were defined independently
of their initial void ratios (Eq. (41)). The normalization consisted
of dividing the state surface function by the initial void ratio for
each soil in Table 4:
e a b b ing’s age, the presence of openings in the walls such as windows
¼ þ  log10 ð1 þ r þ cSÞ ¼ 1 þ  log10 ð1 þ r þ cSÞ
e0 e0 e0 e0
The equation of the state surface was transformed into Eq. (41)
for all soils considered (Table 4). Note that parameter c remained
invariable.
After the normalization step, the state surfaces of different ini-
tial void ratios of each soil were superimposed. It was then de-
duced that parameter b controlled the compressibility of the soil
and increased with the initial void ratio. The value of b/e0 vary from
0.026 to 0.23 and are not correlated to the swelling capacity. The
parameter c is related to the swelling capacity (as discussed previ-
ously) and appears to be constant for different initial conditions.
The value of c varied from 0.022 to 2.3 and was strongly correlated
to the swelling capacity of the soil.
Three ground classes were selected based on the collected data.
Table 5 presents the mean values of the classes, which correspond
to a soil of intermediate swelling capacity (zone 1), a soil of high
swelling capacity (zone 2) and a soil of very high or critical swell-
ing capacity (zones 3 and 4). Zones 3 and 4 were combined due to
lack of data.
The mean values of the parameters presented in Table 5 show
that most swelling soils are characterized by a greater parameter
c value, which is associated with a smaller diffusion coefficient
(a) that results in a smaller active depth (za) and a smaller horizon-
tal moisture variation distance em.
4.2. Building typology
A building typology was required to investigate the deflection
transmission ratio in the developed analytical model. The typology
was based on the mechanical parameters of a beam (EI, L, q).
In general, the evaluation of the exact stiffness of a masonry
building is very complex and depends on many factors, including
the mechanical characteristics of the joints and blocks, the build-
ð41Þ
and doors, the geometry of the building, the contribution of out-
of-plane walls to the stiffness, etc. [45]. Deck and Singh [19] used
a range of possible building stiffnesses to overcome this difficulty.
The same strategy was adopted in this study.

Table 6
Table 5 Variability of the selected building mechanical
Average parameter values of the selected expansive soils for different classes. characteristics.

Swelling a/e0 b/e0 c b*c a (m2/an) em (m) za (m) Parameter Value

Intermediate 1 0.128 0.030 0.004 12.05 2.64 3.47 Load q (kN/m) 30; 50; 100; 200; 300
High 1 0.115 0.25 0.028 10.45 2.55 3.15 Rigidity EI (GN m2) 0.5; 1; 25; 50
Very high/critical 1 0.139 1.27 0.191 4.17 2.36 2.65 Length L (m) 10; 20; 30
26 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32
Fig. 9. (a) Typical results of the deflection transmission ratio for the three soil classes (Table 5) and different mechanical properties of the building (Table 6), (b) pair-wise
correlation analysis.
To estimate the range of stiffness variation, the lower limit of
the flexural rigidity (presented in Table 6) corresponded to a strip
footing with a wall height of 3 m, and the upper limit corresponded
to several parallel walls 10 m high, along with their foundation.
Young’s modulus for masonry was assumed to be in the range of
1–20 GPa [46]. In Table 6, the minimum stiffness value corre-
sponds to a wall with a height of 3 m, a width of 0.25 m and a
Young’s modulus of 1 GPa (EI = 0.25 * 33/12).
The beam loading represents the building self-weight and the
service loading. A value of 100 kN/m is equivalent to the weight
of a 5-m-high wall with a thickness of 0.5 m plus the service load-
ing. A value of 200 kN/m is equivalent to the weight of a 10-m-high
wall with a thickness of 0.5 m plus the service loading. A value of
300 kN/m is equivalent to the weight of a building with three 5-
m-high walls with thicknesses of 0.25 m plus the service loading.
4.3. Deflection transmission ratio for different suction variations
The final deflection transmission ratio is defined as the ratio of
the final deflection of the building D to the shrinkage D0 of free-
field soil (Fig. 1c). The global results of the deflection transmission
ratio for 180 cases, based on the soil typology (Table 5) and the
building parameters (Table 6) for different suction values at the
surface (between 500 kPa and 3000 kPa), were synthesized using
D/D0. A typical result is presented in Fig. 9 (the foundation embed-
ding depth is zero in this section).
Fig. 9 shows that when the building rigidity increases, the set-
tlement transmission ratio decreases. However, the transmission
ratio increases when the suction and swelling capacity increase.
It can be observed that the maximum ratio of settlement trans-
mission increases to approximately 80% for a building 30 m long
with a rigidity of 5 GN m2, loaded with q = 50 kN/m, constructed
Table 7
Parameters used to quantify the relative stiffness.
Reference q* Parameters
[47] ðEJÞs (EJ)s: flexural rigidity of the structure per meter of the
Es L3
building [kN m/m]
L: length of the foundation [m]
Es: soil elastic modulus [kN/m2]
[17] 16EI (EI): flexural rigidity of the structure [kN m2]
Eg L4
L: length of the foundation [m]
Eg: soil elastic modulus [kN/m2]
[19] EI (EI): flexural rigidity of the structure [kN m2]
KBL4
L: length of the foundation [L]
K: soil reaction modulus (Winkler) [kN/m3]
B: width of the foundation [m]
on soil with a very high swelling capacity and undergoing a suction
augmentation of 3 MPa.
The settlement transmitted to the building becomes negligible
for a rigid building (EI = 5 GN m2) constructed on an intermediately
swelling soil.
4.4. Deflection transmission ratio for different relative stiffnesses
A parameter for the relative stiffness of the building and the soil
is commonly used to investigate the soil–structure interaction phe-
nomenon. This section defines a relative stiffness parameter that
summarizes the deflection transmission curves.
The relative stiffness q* is a dimensionless parameter that de-
scribes the relative stiffness of the soil and the structure. Table 7
shows some definitions used for this parameter in the literature
that are based mainly on the Young’s modulus of the soil Es, the
building stiffness EI and the building length L. These definitions
are inappropriate for swelling soils because they account only for
mechanical ground stiffness and not for hydraulic variations.
Fredlund and Rahardjo [3] proposed two independent stiffness
parameters for the hydraulic and mechanical parts in the two-
dimensional analysis of unsaturated soils:
E ¼ ð1þ lÞð12lÞ
ð1lÞav 1
ð42Þ
ð1þlÞ
H ¼ ð1 lÞav 2
where E is the mechanical modulus, H is the hydraulic modulus, l
is Poisson’s ratio, av1 is the compressibility of the soil with respect
to stress changes when the suction is zero (Fig. 2a), and av2 is the
compressibility with respect to suction changes when the stress is
zero (Fig. 2a). These two moduli do not account for hydro-
mechanical coupling, which means that H is independent of the
applied load and E is independent of the suction. Consequently,
the hydraulic rigidity (kh) and mechanical rigidity (km) were intro-
duced using Eq. (43). These two moduli are similar to E and H
when q = 0 and s = 0, so they characterize the mechanical and
hydraulic ground elastic stiffness for vertical stress and suction
values:
kh ¼ 1=ð@e j
@s r!q;S!Sdry
Þ
ð43Þ
km ¼ 1=ð@@er jr!q;S!Sdry Þ
where q is the building load, and Sdry is the applied suction at the
surface. The most relevant parameter for characterizing the soil
stiffness by incorporating both mechanical stiffness and swelling
capacity is now required. It is possible to consider an average value
of the two rigidities calculated using Eq. (43), so the arithmetic
average (Eq. (44)) was used to consider both rigidities equally:
k ¼ ðkh þ km Þ=2 ð44Þ
 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32 27

Fig. 10. Deflection transmission ratio versus relative stiffness.

All of the results presented in Section 4.3 may be synthesized by

introducing the relative stiffness q* (Table 7, proposed by [19])
using Eq. (44). Fig. 10 shows the obtained results. The overall
shapes of the deflection transmission ratio curves versus the
relative stiffness parameter are similar to those calculated for other
ground movements, such as tunneling and mining subsidence
([17,19]).
To prioritize the importance of building parameters and the soil
typology on the deflection transmission ratio, a correlation analysis
was conducted. The correlation matrix is presented in Table 8. The
building stiffness, the building length, the soil type, the suction Fig. 11. Performance of the proposed design equation, (a) without Tanh function
variation at the surface and the building load have the highest cor- and (b) with Tanh function.
relations to D/D0. The building stiffness and load are negatively
correlated to the D/D0 ratio, which is consistent with the expected
results. The parameter k is the soil hydro-mechanical stiffness (Eq.
(44)), which is also negatively correlated to the D/D0 ratio. This D
¼ 0:36610ð1  Tanh½1:86554  5:39436ð0:2696
means that the deflection transmission ratio decreases in stiff soil. D0
It is worth noting that according to Eq. (44), stiff soil in this study  0:000435q þ 0:0000288  s þ 0:9547  ðb  cÞ
refers to soil with low compressibility with respect to the applied  0:1423  log10 ðq ÞÞ ð45Þ
load and suction.
It should be noted that in this study, the building parameters
(EI, L, and q) were assumed to be independent, although they could where q is the building load, b and c are the state surface parame-
be correlated for real buildings. This possible correlation was not ters (soil typology, Section 4.1), s is the suction at the surface
taken into account. (kPa), and q* is the relative stiffness (proposed by [19] and shown
Further statistical analysis of the results was conducted to de- in Table 7) based on Eq. (45).
velop a design equation based on the soil and building parameters. The performance of the proposed equation is presented in
A stepwise linear regression analysis was first performed and an R2 Fig. 11b, in which the estimated D/D0 values obtained using Eq.
of 0.71 was obtained. However, the equations of the analytical (45) are compared to the analytical values. The estimated results
model involve nonlinear relations, and the residual values of the are aligned along the bisector line without residual form. The
linear regression appeared to be significantly correlated residuals’ standard deviation SD is equal to 0.057. To reduce SD,
(Fig. 11a). Consequently, the nonlinear regression was then ad- more parameters could be considered, but our statistical analysis
justed by introducing a Tanh function (Eq. (45)). This adjustment showed that doing so does not significantly improve the results.
resulted in a better R2 (0.91) with less correlation in the residual Eq. (45) is considered a good compromise between the number
values (Fig. 11b). of used parameters and the obtained accuracy.

Table 8
Correlation matrix.
28 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32

L= 30m
q= 30 kN/m
q= 50 kN/m
L= 20m q=100 kN/m
q=150 kN/m
q=200 kN/m
q=300 kN/m

L= 10m

q= 30 kN/m
q= 50 kN/m
q=100 kN/m
q=150 kN/m
q=200 kN/m
q=300 kN/m

q= 30 kN/m
q= 50 kN/m
q=100 kN/m
q=150 kN/m
q=200 kN/m
q=300 kN/m

Fig. 12. Deflection transmission ratio for a soil with very high swelling capacity.

The deflection transmission ratio curves obtained for the three 14 show the results for the three foundation embedding depths (0,
types of swelling soils (Table 5) and the building parameters (Ta- 0.5 and 1 m) that were considered.
ble 6) using the analytical model can also be analyzed case by case, Figs. 12–14 show the influence of soil type, building stiffness,
with the three foundation embedding depths considered. Figs. 12– building length, applied load and foundation depth on the
 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32 29

q= 30 kN/m
q= 50 kN/m
q=100 kN/m
q=150 kN/m
q=200 kN/m
q=300 kN/m

q= 30 kN/m
q= 50 kN/m
q=100 kN/m
q=150 kN/m
q=200 kN/m
q=300 kN/m

q= 30 kN/m
q= 50 kN/m
q=100 kN/m
q=150 kN/m
q=200 kN/m
q=300 kN/m

Fig. 13. Deflection transmission ratio for a soil with high swelling capacity.

deflection transmission ratio when a suction variation of 1 MPa is
applied at the surface. The soil–structure interaction intensity is
low for a flexible building on rigid ground (high kh value) and the
D/D0 ratio approaches 1. This means that the reaction pressure
in the soil remains constant under the building and the building
acts as a uniformly distributed load on the ground, following the
ground movements without modifying the stress distribution in
the soil.
When the building is rigid and the soil is more compressible
(hydraulically meaning more expansive, or mechanically meaning
30 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32

q= 30 kN/m
q= 50 kN/m
q=100 kN/m
q=150 kN/m
q=200 kN/m
q=300 kN/m

q= 30 kN/m
q= 50 kN/m
q=100 kN/m
q=150 kN/m
q=200 kN/m
q=300 kN/m

q= 30 kN/m
q= 50 kN/m
q=100 kN/m
q=150 kN/m
q=200 kN/m
q=300 kN/m

Fig. 14. Deflection transmission ratio for a soil with intermediate swelling capacity.

softer), the value of k decreases, the influence of the soil–structure
interaction becomes more important, and the D/D0 ratio ap-
proaches 0. The building does not follow the ground movements,
and depending on its rigidity, it modifies the reaction pressure in
the soil (meaning that gaps may appear at the building
extremities).
In all three Figs. 12–14, three building lengths are considered
(10, 20 and 30 m), as depicted in Fig. 12. The upper limits of the
settlement transmission rate occur for the longest building with
the lowest applied load and rigidity. The deflection transmission
ratio is negligible for a building with a length of 10 m built on a soil
with an intermediate swelling capacity.
 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32
The amplitudes of shrinkage and building settlement are higher
in more expansive soils. Therefore, the settlement transmission ra-
tio increases overall, from a soil with intermediate swelling capac-
ity to a soil with a very high swelling capacity of up to 65%.
Figs. 12–14 also show the important effect of foundation
embedding depth on the deflection transmission ratio. A greater
foundation embedding depth makes the foundation less exposed
to suction variations and decreases the deflection transmitted to
the building, making it less vulnerable.
These figures show that increasing the foundation embedding
depth from 0 to 1 m decreases the mean deflection transmission
ratio by up to 40%.
5. Conclusions
An analytical model was developed to investigate the effects of
the shrinkage of clayey soils on buildings through soil–structure
interaction analysis. The hydro-mechanical behavior of clay soils
was modeled using the concept of a state surface integrated into
a soil–structure interaction model that is based on the idealized
Winkler model for the soil and an elastic Euler–Bernoulli beam
model the building. The model can be used to calculate the final
deflection of a building as a function of the suction imposed at
the surface, the resulting evolution of the vertical and horizontal
suction profiles, the mechanical properties of the structure and
the hydro-mechanical properties of the soil.
The analytical model’s results were compared with those ob-
tained using a numerical model performed with Code_Bright in
which the elasto-plastic model BExM for swelling unsaturated soils
was implemented. The results showed good agreement between
the models.
The analytical model was then used to investigate the deflection
transmission ratio for different building types on three types of
expansive soils undergoing different possible suction variations.
A parameter for the relative stiffness of the ground and the
building was introduced. This parameter accounts for both the
mechanical and hydraulic ground stiffnesses and was used to syn-
thesize the results of the deflection transmission ratio, as has been
performed for other ground movements, such as ground move-
ments due to tunneling and mine subsidence.
The obtained curves can be used by engineers to predict the
approximate transmitted deflection of a building without analyti-
cal calculations. Based on these results and a statistical analysis,
a design equation was proposed.
The results showed that the final deflection of a building in-
creases with the length of the building and the imposed suction.
However, the building settlement is inversely proportional to the
load, the rigidity of the building and the embedding depth of the
foundation, which makes the building less vulnerable to shrinkage
and swelling action.
This research could be extended in the future using the esti-
mated relative deflection of the building to compare the threshold
values of classical damage categories and to assess building dam-
age. The developed model could be improved by accounting for
the shear deformation of the soil and the building, with a more
complete soil database that would give a more precise soil typol-
ogy. It would also be useful to study other state surfaces and iden-
tify the impact of the choice of state surface on the final results. In
addition, the state surface-based model proposed by Kodikara [49]
for recently compacted soils could be combined with the method-
ology proposed in this paper for new construction cases.
References
[1] Alonso EE, Vaunat J, Gens A. Modeling the mechanical behaviour of expansive
clays. Eng Geol 1999;54:173–83.
31
[2] Alonso EE, Gens A, Josa A. A constitutive model for partially saturated soils.
Géotechnique 1990;40(3):405–30.
[3] Fredlund DG, Rahardjo H. Soil Mechanics for unsaturated soils. A Wiley-
Interscience Publication; 1993.
[4] Gens A, Sanchez M, Sheng D. On constitutive modelling of unsaturated soils.
Acta Geotech 2006;1:137–47.
[5] Sheng D, Fredlund DG, Gens A. A new modeling approach for unsaturated soils
using independent stress variables. Can Geotech J 2008;45:511–34.
[6] Yigzaw Z, Audiguier M, Cojean R. Analyse du comportement d’un sol
argileux sous sollicitations hydriques cycliques. Eng Geol Environ 2009;68:
421–36.
[7] Matyas EL, Radhakrishna HS. Volume change characteristics of partially
saturated soils. Géotechnique 1968;18(4):432–48.
[8] Fredlund DG, Morgenstern NR. Stress state variables for unsaturated soils. J
Geotech Eng 1977;103:447–66.
[9] Lloret A, Alonso EE. State surfaces for partially saturated soils. In: Proceedings
11th international conference on soil mechanics and foundation engineering,
vol. 2, San Francisco; 1985. p. 557–62.
[10] Gatmiri B, Delage P. A new void ratio state surface formulation for the
nonlinear elastic constitutive modeling of unsaturated soil-Code UDAM. 1st
Int. Conf. on unsaturated soils, Paris; 1995. p. 1049–56.
[11] Lytton R. Prediction of movement in expansive clays. ASCE 1994:1827–45.
Geotechnical Special Publication.
[12] Zhang X, Lytton RL. Modified state-surface approach to the study of
unsaturated soil behaviour – Part II. Can Geotech J 2009;46:553–70.
[13] Vu HQ, Fredlund DG. Challenges to modeling heave in expansive soils. Can
Geotech J 2006;43:1249–72.
[14] Zhang X, Briaud JL. Improved approach to construct constitutive surfaces for
stable-structured soils covering both saturated and unsaturated conditions. J
Geotech Geoenviron Eng, ASCE 2008;134(6):876–82.
[15] Skempton AW, Mcdonal DH. The allowable settlements of buildings. Proc Inst
Civ Eng Part III 1956;5:726–68.
[16] Burland JB. Assessment of risk of damage to buildings due to tunneling and
excavation. Earthquake geotechnical engineering, éditions Ishihara, Balkema;
1995. p. 1189–201.
[17] Potts DM, Addenbrooke TI. A structure’s influence on tunnelling-induced
ground movements. Proc Incln Civ Eng 1997;125:109–25.
[18] Franzius JN, Potts DM, Burland JB. The influence of soil anisotropy and K-0 on
ground surface movements resulting from tunnel excavation. Géotechnique
2005;55:189–99.
[19] Deck O, Singh A. Analytical model for the prediction of building deflections
induced by ground movements. Int J Numer Anal Meth Geomech 2010. http://
dx.doi.org/10.1002/nag.993.
[20] Nelson J, Miller D. Expansive soils: problems and practice in foundation and
pavement engineering. New York: John Wiley & sons Ltd; 1992.
[21] Abdelmalak R. Soil structure interaction for shrink-swell soils. A new design
procedure for foundation slabs on shrink-swell soils. Ph.D. thesis of Texas A&M
University; 2007.
[22] Viet Do Q, Mebarki A, Vivien J, Heck R, Saada A. Vulnérabilité des bâtis de
maisons individuelles sur sols argileux. Symposium international Sécheresse
et construction; 2008.
[23] Gould SJF, Kodikara J, Rajeev P, Zhao XL, Burn S. A new mathematical equation
for void ratio-water content-net stress surface of expansive soils. Can Geotech
J 2011;48(6):867–77.
[24] Overton DD, Chao KC, Nelson JD. Time rate of heave prediction in expansive
soils. GeoCongress: geotechnical engineering in the information technology
age, ASCE, Reston; 2006.
[25] Lytton R, Aubeny C, Bulut R. Design procedure for pavement on expansive
soils, vol. 1. Texas Transportation Institute, Report 0–4518; 2005.
[26] Vu HQ. Uncoupled and coupled solutions of volume change problems in
expansive soils. Ph.D. thesis, University of Saskatchewan; 2003.
[27] Vu HQ, Fredlund DG. Challenges to modelling heave in expansive soils. Can
Geotech J 2006;43:1249–72.
[28] Mitchell PW. The structural analysis of footing on expansive soil. Research
Report No.1, Adelide, South Australia; 1979.
[29] McKeen RG, Johnson LD. Climate-controlles soil design parameters for mat
foundations. J Geotech Eng 1990;116(7):1073–94.
[30] Aubeny C, Long X. Moisture diffusion in shallow clay masses. J Geotech
Geoenviron Eng 2007;133(10):1241–8.
[31] Vincent M, Fleureau JM, Masrouri F, Oppenheim E, Heck JV, Ruaux N, et al.
Etude des mécanismes de déclenchement du phénomène de retrait-
gonflement des sols argileux et de ses interactions avec le bâti. Rapport
final. BRGM/RP-54862-FR; 2006.
[32] El-Garhy BM, Wray WK. Method for calculating the edge moisture variation
distance. J Geotech Geoenviron Eng 2004;130:945–55.
[33] Al Qadad A. Influence de la Sécheresse sur les Structures: Modélisation
de l’Interaction Sol-Atmosphère-Structure. Thèse de Doctorat. Université
des Sciences et Technologies de Lille, Laboratoire de Mécanique de Lille;
2009.
[34] Jahangir E, Deck O, Masrouri F. Estimation of the ground settlement beneath
foundations due to the shrinkage of clayey soils. Can Geotech J 2012;49(7):
835–52.
[35] Mrad M. Modélisation de comportement hydromécanique des sols gonflants
non saturés. Ph.D. thesis, Institut National Polytechnique de Lorraine, Nancy,
France; 2005.
32 E. Jahangir et al. / Computers and Geotechnics 54 (2013) 16–32
[36] Nowamooz H, Mrad M, Abdallah A, Masrouri F. Experimental and numerical
studies of the hydromechanical behaviour of a natural unsaturated swelling
soil. Can Geotech J 2009;46:1–18.
[37] Mualem Y. A new model for predicting the hydraulic conductivity of
unsaturated porous media. Water Resourc Res 1976;12:513–22.
[38] Van Genuchten MTH. A closed-form equation for predicting the hydraulic
conductivity of unsaturated soils. Soil Sci Soc Am J 1980;44:892–8.
[39] Pham HQ. A volume-mass constitutive model for unsaturated soils. PhD thesis,
University of Saskatchewan, Canada; 2005.
[40] Fleureau JM, Verbrugge JC, Huergo PJ, Gomes Correia A, Kheirbek-Saoud S.
Description and modelling of the drying and wetting paths of compacted soils.
Can Geotech J 2002;39:1341–57.
[41] Nowamooz H. Retrait/gonflement des sols argileux compactés et naturels,
Thèse de Doctorat, INPL, Nancy, 2007.
[42] Romero E, Gens A, Lloret A. Suction effects on a compacted clay under non-
isothermal conditions. Géotechnique 2003;53(1):65–81.
[43] Tang AM. Effet de la temperature sur le comportement des barrièrs de
confienement, PhD thesis of Ecole Nationale des Ponts et Chaussées, Paris;
2005.
[44] Fleureau JM, Soemitro R, Taibi S. Behaviour of an expansive clay related to
suction. In: Proceedings 7th international conference on expansive soils, vol. 1,
Dallas, Août; 1992. p. 173–8.
[45] Son M, Cording EJ. Estimation of building damage due to excavation-induced
ground movements. J Geotech Eng ASCE 2005;131(2):162–77.
[46] Dimmock PS, Mair RJ. Effect of building stiffness on tunneling-induced ground
movement. Tunnell Undergr Space Technol 2008;23:438–50.
[47] Eurocode 2 (EN 1992–1-1). Calcul des structures en béton – Partie 1–1: règles
générales et règles pour les bâtiments; October 2005.
[48] Jahangir E, Abdallah A, Masrouri F. Evaluation du profil hydrique saisonnier
dans un sol gonflant sous un fondation. JNGG, Bordeaux, France: Journées
Nationales de Géotechnique et de Géologie de l’Ingénieur; 2012. p. 585–92.
[49] Kodikara J. New framework for volumetric constitutive behavior of compacted
unsaturated soils. Can Geotech J 2012(49):1227–43.