Professional Documents
Culture Documents
Testing and Modeling of Soil-Structure Interface
Testing and Modeling of Soil-Structure Interface
Testing and Modeling of Soil-Structure Interface
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.
Abstract: An accurate modeling of soil-structure interfaces is very important in order to obtain realistic solutions of many soil-structure
interaction problems. To study the mechanical characteristics of soil-structure interface, a series of direct shear tests were performed. A
charged-coupled-device camera was used to observe the sand particle movements near the interface. It is shown that two different failure
modes exist during interface shearing. Elastic perfect-plastic failure mode occurs along the smooth interface, while strain localization
occurs in a rough interface accompanied with strong strain-softening and bulk dilatancy. To describe the behavior of the rough interface,
this paper proposes a damage constitutive model with ten parameters. The parameters are identified using data from laboratory interface
shear tests. The proposed model is capable of capturing most of the important characteristics of interface behavior, such as hardening,
softening, and dilative response. The interface behaviors under direct and simple shear tests have been well predicted by the model.
Furthermore, the present model has been implemented in a finite element procedure correctly and calculation results are satisfactory.
DOI: 10.1061/共ASCE兲1090-0241共2004兲130:8共851兲
CE Database subject headings: Soil-structural interaction; Interfaces; Roughness; Strain softening; Dilatancy; Damage; Constitutive
relations; Finite element method.
Introduction Background
The interaction between structures and soils is a critical problem Direct shear tests have often been used to study the behavior of
in geotechnical engineering. The mechanical behavior between soil-structure interfaces. Several factors such as structural mate-
soil and construction materials is of major concern in soil- rial, soil properties, and surface roughness have been investigated
structure interaction problems including retaining structures, deep to better understand their effects on the interface characteristics
foundations, cutoff walls, and earth reinforcement. Study of the 共Potyondy 1961; Coyle and Sulaiman 1967; Kulhaway and Peter-
son 1979兲. Yoshimi and Kishida 共1981兲 utilized a ring torsion
frictional characteristics of soil-structure interface is important in
apparatus for interface testing and observed sand deformation by
terms of both theoretical analysis and practical engineering. A
using x-ray photography. Uesugi and Kishida 共1986a,b兲 used a
limited review of previous work is provided below. Past efforts
simple shear test device to study the interface behavior between
have concentrated on measuring the skin friction coefficient and
several kinds of soils and steel plates of different roughness. Their
many constitutive models have been developed in conjunction comprehensive work concluded that surface roughness had an
with numerical procedures designed to solve such problems. The important influence on the friction coefficient at yielding. Uesugi
material parameters involved in these models need to be deter- et al. 共1988兲 tracked the particle behavior near the interface using
mined from appropriate laboratory and field tests in order to de- close-up photographs and observed the formation of a shear zone
scribe the interface behavior in a realistic manner. with a thickness of 5D 50 within the sand mass along the rough
The objectives of this study are 共1兲 to study the mechanical interface. Yin et al. 共1995兲 conducted a large direct shear test to
characteristics of interfaces with different relative roughness; 共2兲 observe the distribution of relative displacement along the inter-
to develop a constitutive model to describe the behavior of inter- face. Paikowsky et al. 共1995兲 developed a dual interface testing
faces according to test results; 共3兲 to predict the interface behavior apparatus that allowed measurement of friction distribution along
for validation of the model; and 共4兲 to use the proposed model for the interface. Using normalized roughness R n and roughness
numerical simulation. angle ␣, the interface was categorized into three zones:
‘‘smooth,’’ ‘‘intermediate,’’ and ‘‘rough.’’ Evgin and Fakharian
共1996兲 set up an interface simple shear apparatus C3DSSI to per-
1
PhD, Associate Professor, Institute of Geotechnical Engineering, form two-way cyclic tangential-displacement-controlled experi-
Dept. of Hydraulic Engineering, Tsinghua Univ., Beijing 100084, ments with different normal stiffness. The test results showed that
People’s Republic of China. stress path significantly influenced the shear stress-tangential dis-
2
Professor, Institute of Geotechnical Engineering, Dept. of Hydraulic placement and volume change behavior of the interface. Frost
Engineering, Tsinghua Univ., Beijing 100084, People’s Republic of et al. 共1999兲 studied the evolution of the structure of sand adja-
China. cent to the geomembrane, and found that it was directly influ-
Note. Discussion open until January 1, 2005. Separate discussions enced by the surface roughness. A number of studies have shown
must be submitted for individual papers. To extend the closing date by
that surface topography is important in the behavior of soil-
one month, a written request must be filed with the ASCE Managing
Editor. The manuscript for this paper was submitted for review and pos- structure interfaces 共Hryciw and Irsyam 1993; Nakamura et al.
sible publication on December 20, 2000; approved on October 6, 2003. 1999; Dove and Harpring 1999; Dove and Jarrett 2002兲. This
This paper is part of the Journal of Geotechnical and Geoenvironmental current study attempts to investigate interfacial shear and friction
Engineering, Vol. 130, No. 8, August 1, 2004. ©ASCE, ISSN 1090- strength based on the micromechanism between soil particles and
0241/2004/8-851– 860/$18.00. a solid surface using theoretical and experimental approaches.
D 50 D 10 D 60 ␥ max ␥ min
共mm兲 共mm兲 共mm兲 Cu Cc Gs 共kN/m3兲 共kN/m3兲 e max e min 共deg兲
1.0 0.59 1.20 2.1 0.87 2.65 16.02 13.08 1.026 0.654 34.4
The mechanical characteristics of interfaces between structural structure interface have been reported, various interface behaviors
and geological materials and accurate modeling are very impor- such as strain softening and dilative behavior are not adequately
tant if numerical methods are to be used in predicting soil- simulated, or large numbers of the parameters increase the com-
structure interaction. In this context, numerical methods such as plexity and minimize further application of the models. A simple
finite element methods, special interface elements, and constitu- model with fewer parameters and strong simulation capability has
tive models are often used to account for relative motion and yet to be developed.
deformation modes. One of the commonly used interface ele-
ments in soil-structure interaction is based on the joint element
proposed by Goodman et al. 共1968兲. The element formation was Interface Test
derived from the relationship of relative nodal displacements and
stresses of the interface element. For two-dimensional analysis, A series of sand-steel interface direct shear tests were performed
the constitutive relation was expressed as to obtain the stress-strain relationship. Yongdinghe sand, which is
再 冎冋 册再 冎
a coarse silica sand with subangular particles, was used in the
d k nn 0 dv tests. The physical properties of Yongdinghe sand are presented in
d ⫽ 0 k ss du (1)
Table 1. Low-carbon steel plates with five different values of
where and ⫽normal and shear stresses; v and u⫽relative nor- surface roughness were used to make the interface. The surface
mal and shear displacements of the interface, respectively; and k ss was grooved by lathe in the direction normal to that of tangential
and k nn ⫽tangential and normal stiffness of the interface. For ap- load and the surface profile was measured by a stylus profilome-
plication to soil-structure interaction problems, the thickness of ter. The quantification of surface roughness in terms relative in-
the element is often assumed to be zero. Zienkiewicz et al. 共1970兲 terface roughness R n , as reported by Uesugi and Kishida
used a thin isoparametric finite element formulation for soil- 共1986b兲, was defined by measuring R max 共vertical distance be-
structure interfaces, which was treated essentially like a solid el- tween the highest peak and the lowest trough兲 along a profile
ement. Ghaboussi et al. 共1973兲 proposed an interface formulation, length L equal to the mean grain size D 50 and then normalizing it
which was derived by considering relative motions between sur- by D 50 , that is
rounding solid elements as independent degrees of freedom. Ka- R n ⫽R max共 L⫽D 50兲 /D 50 (3)
tona 共1983兲 derived an interface model from the virtual work
principle modified by appropriate constraint conditions. Desai The roughness of steel plates is listed in Table 2.
et al. 共1984兲 used a thin layer element for interfaces and joints, A schematic layout of the experimental setup is presented in
which can consider relative motions such as slippage, debonding, Fig. 1. The interface between steel and sand was 60 mm in length
rebonding, and separation as occurring in the thin zone. The thin and 53 mm in width. The dimension of the steel plate was
layer element can be ameliorated to prevent interpenetration be- 110 mm⫻125 mm, and the area of interface can remain constant
tween soil and structure. Various constitutive laws have been de- during shearing due to the length of the plate. The sand was
veloped to describe interface behavior. The elastic perfect-plastic air-pulviated using a sand rainer so as to yield uniform sand
model can be used for the soil-structure interface, in which k ss in specimens at 90% relative density. The sand sample was placed in
Eq. 共1兲 is set equal to zero when the shear stress exceeds the shear a transparent plexiglas shear box with inside graduated grids to
strength. The nonlinear elastic model developed by Clough and facilitate visual observation of sand particle displacement near the
Duncan 共1971兲 has been used widely in the analysis of soil- interface. A SANYO VPC-Z400 digital charged-coupled-device
structure interface behavior. The hyperbolic relationship of shear 共CCD兲 camera with a resolution of 1600 共H兲 by 1200 共V兲 pixels
stress and relative displacement is obtained from the interface was introduced to record the microdisplacements of the sand par-
direct shear test. The coefficient of tangential stiffness k ss can be ticles. In each series of test, the normal stress was maintained at
expressed as 50, 100, 200, and 400 kPa, respectively, and the steel plate moved
冉 冊
at a horizontal displacement rate of 0.6 mm/min. The test was
R f s • 2
terminated at 10 mm of total horizontal displacement. The shear
k ss ⫽K s ␥ 共 n / P a 兲 n s 1⫺ (2)
n tan ␦ s ⫹c s stress, tangential displacement, and normal displacement were
where ␥ w ⫽unit weight of water; and P a ⫽atmospheric pressure.
There are five parameters in this model. The model is incapable of
capturing restrained dilation that is the phenomenon consistently Table 2. Properties of Structure Material
observed in dense sand-structure interface. Several plastic models Surface roughness Relative roughness
were proposed to consider the coupling of normal and shear be- Plate number R max (mm) Rn
havior 共Ghaboussi et al. 1973; Boulon and Nova 1990; Gens
1 0.01 0.01
et al. 1990兲. The disturbed state concept 共DSC兲 was proposed to
2 0.05 0.05
describe the coupling behavior of soil-structure interfaces and
3 0.10 0.10
rock joints and predict interface test results 共Desai and Ma 1992;
4 0.20 0.20
Navayogarajah et al. 1992; Pal and Wathugala 1999; Fakharian
5 0.50 0.50
and Evgin 2000兲. Although several constitutive models for soil-
Fig. 1. Scheme of direct shear apparatus for interface test 共unit: mm兲
关 De兴⫽ 冋 Ds
0 Dn
0
册 (5)
再 冎
d i
d ⫽ 关 D e
再 冎
兴
d es
d ev
(4)
Fig. 5. Scheme of damage model
where the elastic matrix is
D ep
ns Dn
ns ⫽0
D ep (17)
1⫺ i /•tan ␦ 0
ep
D ss ⫽ (18)
sp /•tan ␦ 0 ⫹1/D s
sp /
sn ⫽D n
D ep (19)
sp /•tan ␦ 0 ⫹1/D s
Fig. 6. Yield surfaces in stress space of intact state Reference State II: Critical Material
关 D ep 兴 ⫽ 关 D e 兴 ⫺
再 冎再 冎
关 De兴
Q
˜
F
˜
T
关 De兴T
tion D is defined as a function of plastic shear strain sp
再 冎 再 冎
(15)
F T
Q F
关 De兴 ˜ ⫺ p where ⫽ 兰 (d sp •d sp ) 1/2 is the trajectory of plastic shear strain;
˜ s
a,b⫽empirical coefficients, which can be determined from the
The stiffness matrix can be expressed as test results.
D ss ⫽⫺
冉 D•h⫹ 共 1⫺D 兲 •D ep
sn /D n ⫺ 共 ⫺ 兲 •g•
c i
D ep
D s •D n
sn
冊 • cv •g
1 D epsn
⫹D• f ⫺g•
Dn D s •D n
冉
⫻ 1⫺
D ep
Ds
s
冊 ⫹ 共 1⫺D 兲 •D ep
s ⫹ 共 ⫺ 兲 •g• 1⫺
c i
冉 D ep
Ds
s
冊
(27)
D ep
sn
D•h⫹ 共 1⫺D 兲 •D ep
sn /D n ⫺ 共 ⫺ 兲 •g•
c i
D s •D n
D sn ⫽ (28)
1 D ep
sn
⫹D• f ⫺g•
Dn D s •D n
D ns ⫽⫺
1
⫹D• f ⫺g•
cv •g
D epsn
冉
• 1⫺
D ep
s
Ds
冊 (29)
Dn D s •D n
1
D nn ⫽ (30)
1 D epsn
⫹D• f ⫺g•
Dn D s •D n
Fig. 7. Relationship of shear stress and normal displacement versus
relative shear displacement ( n ⫽50, 100, 200, and 400 kPa; D r where
⫽90%; R n ⫽0.5) s
i⫽ (31)
1/D s ⫹ s /•tan ␦ 0
Stress-Strain Relationship g⫽ 共 1⫺D 兲 •a•b• b⫺1 (32)
According to DSC, the shear zone is assumed to be a mixture of
f ⫽ 0v •exp共 ⫺K• 兲 •K (33)
intact and critical state materials. Within the smeared zone, the
observed stresses and strains can be considered as follows: h⫽ i •tan ␦ r (34)
v ⫽ ev ⫹ vp ⫽ ev ⫹D• cv Considering the interface thickness t, the strain increments can
be expressed by the relative displacements
s ⫽ es ⫹ sp ⫽ is ⫽ cs
⫽ i ⫽ c
(23)
再 冎 再 冎
d s 1 du
d v ⫽ t d v (35)
再 冎
Parameters of Damage Model and Method of
再 冎
d d s
d ⫽ 关 D 兴 d v
ep
(25)
Identification
There are 10 parameters in the proposed model, which are listed
The stiffness matrix 关 D ep 兴 of the damage model is in Table 3. They can be easily determined from the results of the
冋 册
interface shear test.
D ss D sn
关 D ep 兴 ⫽ (26) The parameters D r , 0v , and R n can be obtained from the
D ns D nn physical properties of the sand and the structural material. Critical
Table 4. Damage Model Parameters for Direct Shear 共DS兲 and Simple Shear 共SS兲 Test
Test Rn Dr tan ␦0 tan ␦r D s (MPa) D n (MPa) a b 0v K
DS 0.50 90% 1.39 0.60 4.00 4000 3.17 1.60 0.22 0.005
SS 0.04 84% 1.08 0.60 4.00 4000 3.00 1.00 0.19 0.094
friction coefficient tan ␦r can be obtained from the critical strength comparisons between test results and predictions are presented in
of the interface shear test, and tan ␦0 is about 30–50% larger than Fig. 8. Although the initial contraction cannot be predicted, the
the peak friction coefficient. Incorporating the interface thickness peak strength, residual strength, and dilative response are well
t⫽5D 50 , the testing data of shear stress and relative displace- estimated in all predictions.
ments can be transformed to the stress-strain curve. Then the The comparisons between model predictions and experimental
shear modulus D s can be derived from the initial slope of the data show the very good predictive capability of the proposed
shear stress-strain curve, and the normal modulus D n usually damage model.
adopts a large value 共e.g., 1,000D s ). K can be derived from the
relationship of ultimate dilatancy and normal stress.
The damage function D can be derived from the data of inter- Finite Element Method Analysis of Interface
face shear tests
The proposed damage model has been incorporated in the FEM
i ⫺ program, and both two- and three-dimensional formulations have
D⫽ (36) been developed to simulate soil-structure interaction problems
i ⫺ c
共Hu and Pu 2002a, 2003兲. This section describes two simple and
Then the relationship of ln关⫺ln(1⫺D)兴 versus gives values practical applications in soil-structure interaction.
of parameter a and b
ln关 ⫺ln共 1⫺D 兲兴 ⫽ln a⫹b ln (37) Example 1: One-Element Problem
Therefore, all the parameters can be determined from a series A single interface element was idealized as a plane strain problem
of interface shear tests with different normal stresses. The details and analyzed by FEM. The constant normal stress is 50, 100, 200,
of parameter identification can be found in Hu 共2000兲. and 400 kPa for each case. The mean diameter of the sand particle
is 1 mm, so the thickness of the interface is assumed to be 5D 50 ,
i.e., 5 mm. The parameters of the damage model are those of the
Verification of Damage Model direct shear test in this paper, which are listed in Table 4. The
relative tangential displacement is controlled to increase gradually
In this section, the experimental results of direct shear and simple from 0 to 25 mm.
shear tests are predicted with the proposed model. Fig. 9 shows the computed results for the shear stress and two
displacement components and the comparison with the theoretical
Direct Shear Test solution from the damage model. It is obvious that the two solu-
tions are very close and the developed FEM program incorporates
The results of a direct shear test in this paper are predicted by the the damage model correctly.
proposed model. Fig. 7 shows the model predictions and corre-
sponding experimental data of the R n ⫽0.5 test. The line with data
markers indicates test results and the line without data markers
indicates model prediction. The damage model parameters for the
direct shear 共DS兲 test are listed in Table 4.
The predictions seem to be in good agreement with the experi-
mental results. The calculation results of the shear stress–relative
displacement curve indicate that the peak strength, residual
strength, and corresponding displacements are predicted very
well. The dilative response of the interface is also well predicted
by the proposed model. However, the sand contraction at the
onset of shearing is not described well. The reason is that the
compressive mode is controlled by the constitutive relationship of
the intact state, while the proposed model assumes that the intact
state has no plastic volumetric strain.
the combination of intact state and critical state behaviors v ⫽ relative normal displacement;
with the weight of the damage function. The model has ten ␥ max ,␥ min ⫽ maximum and minimum dry unit weight of
parameters that can be determined on the basis of interface soil;
shear tests. The predictions of the proposed damage model ␥ w ⫽ unit weight of water;
agree well with the direct and simple shear test results. The ␦ r ⫽ friction angle of critical state;
hardening, softening, and dilative response of the rough in- ␦ 0 ⫽ ultimate friction angle of intact state;
terface between sand and steel plate can be well described s ⫽ shear strain in interface;
and the peak and residual strengths are adequately predicted. v ⫽ volumetric strain in interface;
3. The interface element of a damage model was integrated into 0v ⫽ maximum volumetric strain;
the FEM program correctly, and satisfactory numerical solu- ⫽ trajectory of plastic shear strain;
tions were obtained for two simple examples. The proposed ⫽ normal stress;
model can be used for soil-structure interaction problems ⫽ shear stress;
through numerical simulation. ⫽ internal friction angle; and
⫽ relative shear displacement in interface.
Subscripts
Acknowledgments n ⫽ normal response;
s ⫽ shear response; and
The writers would like to thank Mr. Yin Kunting and Mr. Qiu v ⫽ volumetric response.
Tong for their assistance in the laboratory experiment and Mr. Li Superscripts
Haikun for help with the numerical analysis described in this c ⫽ critical state;
paper. Financial support from the National Science Foundation of e ⫽ elastic;
China 共Project Nos. 50099620 and 50379032兲 and the Kiso-jiban ep ⫽ elastoplastic;
Consultant Company are gratefully acknowledged. i ⫽ intact state; and
p ⫽ plastic.
Notation
References
The following symbols are used in this paper:
a,b ⫽ coefficients of damage function; Boulon, M., and Nova, R. 共1990兲. ‘‘Modelling of soil-structure interface
Cc ⫽ coefficient of curvature; behavior: A comparison between elastoplastic and rate type laws.’’
Cu ⫽ uniformity coefficient (D 60 /D 10); Comput. Geotech., 17共9兲, 21– 46.
c s ,␦ s ⫽ parameters of interface shear strength for Clough, G. W., and Duncan, J. M. 共1971兲. ‘‘Finite element analyses of
nonlinear elastic model; retaining wall behavior.’’ J. Soil Mech. Found. Div., 97共12兲, 1657–
D ⫽ damage function; 1673.
Coyle, H. M., and Sulaiman, I. 共1967兲. ‘‘Skin friction for steel piles in
Dn ⫽ normal elastic modulus;
sand.’’ J. Soil Mech. Found. Div., 93共6兲, 261–270.
Dr ⫽ relative density;
Desai, C. S. 共1987兲. ‘‘Future on unified hierarchical models based on
Ds ⫽ shear elastic modulus; alternative correction or damage approach.’’ Rep., Dept. of Civil En-
D ss , etc. ⫽ element of stiffness matrix for damage model; gineering and Engineering Mechanics, Univ. of Arizona, Tucson, Ariz.
ep
D ss , etc. ⫽ element of stiffness matrix for intact state; Desai, C. S., and Ma, Youzhi 共1992兲. ‘‘Modeling of joints and interfaces
D 10 ⫽ diameter of particles corresponding to 10% using the disturbed-state concept.’’ Int. J. Numer. Analyt. Meth. Geo-
finer; mech., 16, 623– 653.
D 50 ⫽ mean diameter of sand particle; Desai, C. S., Zaman, M. M., Lightner, J. G., and Siriwardane, H. J.
D 60 ⫽ diameter of particles corresponding to 60% 共1984兲. ‘‘Thin-layer element for interface and joints.’’ Int. J. Numer.
finer; Analyt. Meth. Geomech., 8, 19– 43.
Dove, J. E., and Harpring, J. C. 共1999兲. ‘‘Geometric and spatial param-
e ⫽ void ratio;
eters for analysis of geomembrane/soil interface behavior.’’ Proc.,
e max ,e min ⫽ maximum and minimum void ratio;
Geosynthetics ’99, Industrial Fabrics Association International,
F ⫽ yield function; Boston, Balkema, Rotterdam, The Netherlands, 575–588.
Fu ⫽ function of failure surface; Dove, J. E., and Jarrett, J. B. 共2002兲. ‘‘Behavior of dilative sand interface
Gs ⫽ specific gravity; in a geotribology framework.’’ J. Geotech. Geoenviron. Eng., 128共1兲,
K ⫽ volume dilatancy coefficient; 25–37.
Ks ⫽ coefficient of tangential stiffness; Evgin, E., and Fakharian, K. 共1996兲. ‘‘Effect of stress paths on the be-
k nn ⫽ normal stiffness of interface; havior of sand-steel interfaces.’’ Can. Geotech. J., 33共6兲, 485– 493.
k ss ⫽ tangential stiffness of interface; Fakharian, K., and Evgin, E. 共2000兲. ‘‘Elasto-plastic modeling of stress-
L ⫽ surface profile length; path-dependent behavior of interfaces.’’ Int. J. Numer. Analyt. Meth.
Pa ⫽ atmospheric pressure; Geomech., 24, 183–199.
Frost, J. D., Lee, S. W., and Cargill, P. E. 共1999兲. ‘‘The evolution of sand
Q ⫽ potential function;
structure adjacent to geomembranes.’’ Proc., Geosynthetics ’99,
R cr ⫽ critical relative roughness; Industrial Fabrics Association International, Boston, Balkema, Rotter-
R f ,n s ⫽ parameters for nonlinear elastic model; dam, The Netherlands, 559–573.
R max ⫽ vertical distance of between highest peak and Gens, A., Carol, I., and Alonso, E. E. 共1990兲. ‘‘A constitutive model for
lowest trough along surface profile length L; rock joints formulation and numerical implementation.’’ Comput.
Rn ⫽ relative interface roughness; Geotech., 9, 3–20.