SOIL-STRUCTURE INTERACTION
By M d . Musharraf-uz Z a m a n , 1 A. M . ASCE,
C h a n d r a k a n t S. Desai, 2 M . ASCE, a n d
Eric C. D r u m m , 3 A. M . ASCE
Downloaded from ascelibrary.org by Ryerson University on 01/05/19. Copyright ASCE. For personal use only; all rights reserved.
INTRODUCTION
8ody 2 Body 2
Downloaded from ascelibrary.org by Ryerson University on 01/05/19. Copyright ASCE. For personal use only; all rights reserved.
C:
/ Body 1
M" r
A A
" c 4> A
Body 1
c -I
A = Total Area
•vfe—TJP—^—?m 7W 1
=75 >SS *W
FIG. 2.—Modes of Deformation at Interface: (a) Stick or No-Slip; (b) Slip; (c) De-
bonding; (d) Rebonding
THIN-LAYER ELEMENT
B = average width of
side of element
ing appropriate constitutive law for the element. In other words, the
thin-layer element is treated essentially like any other solid (soil, rock
or structural) element, but its constitutive relations are denned differently.
Constitutive Relation for Thin-Layer Element.—The constitutive re-
lation matrix, [C,], of the thin-layer element is expressed as
L*" nn J i L*-* ns J i
[C,] = (i)
-\S-$n\i V-ss\i.
in which, [C„„] = normal component; [Css] = shear component; and [Cm],
[Csn] represent coupling effects between normal and shear behaviors. At
this time, the coupling components are not considered. The subscript,
i, denotes that the quantities are related to interface.
Normal Behavior.—In most of the previous models related to the be-
havior of interface, a high value was assigned for the normal sitffness
to avoid interpenetration of neighboring solid elements. Selection of this
high value has often been arbitrary. Since the interface is surrounded
by the structural and geological materials, its normal properties during
the deformation process should be dependent upon the characteristics
of the thin interface zone, as well as the state of stress and properties
of the surrounding elements. Based on these considerations, it was pro-
posed to express the normal stiffness as (2)
[C„„],- = [C„„(aL,Pxm,7m)] (2)
in which a'm/ (3f,, y„ (m = 1, 2, ...) = the properties of the interface,
geological (soil or rock), and structural elements, respectively. For nu-
merical application, Eq. 2 needs to be written explicitly in a suitable
functional form as
[C„„]i = Xi[C„„], + A.2[C„„]g + X.3[C nn\st • • (3)
in which [C„„], = normal behavior of thin layer; subscripts g and sf =
geological and structural materials, respectively, and Xx, X.2 / a n d X3 are
the participation factors, which may vary from 0 to 1. One of the sim-
plifications of Eq. 3 would be to assume X2 = ^ 3 = 0 and \ 2 = 1, implying
that the normal component is based on the normal behavior of the thin-
layer element alone, evaluated just as the adjacent soil element. It can
be possible to arrive at appropriate values for \ , based on a trial and
1260
9 3
> <4>
in which t' = thickness of the interface element; and the superscript, 0,
= total quantity. Relating 6 to the tangent shear modulus, G,, of the
interface yields
G, = (5)
Body 1
Body 2
Interface
H
(b) Deformations at the Interface
FIG. 4.—Shear Behavior of Interface: (a) Direct Shear Test; (b) Deformations at
Interface
1261
G,= (10)
du° o-f}„,N=constant
Eq. 10 can be used to estimate G, for elements in stick and sliding modes.
It may be noted here that several previously developed models (9,10,17)
assumed arbitrary and small values for shearing modulus, G,, for ele-
ments in sliding mode, and often, selection of this (low) value did not
have a logical basis. In the present formulation, this assumption is not
necessary. Use of appropriate values for coefficients a, and p, (Eq. 9)
automatically yields small values for G, under sliding modes.
Another important feature incorporated in the thin-layer element is
loading, unloading, and reverse loading behavior of interfaces (Fig. 5).
The loading and reverse loading portions are simulated by using Eq. 10,
whereas a constant modulus is used for the unloading behavior. It may
be also noted that when separation or debonding occurs, the shear stiff-
ness of the interface tends toward zero. This aspect is incorporated in
Unloading
U - constant
satisfactory simulation can be obtained for t'/B ratios (Fig. 3) in the range
of 0.01-0.1 Here, B is the average width of the adjoining elements.
Evaluation of a, and p;.—As mentioned previously, coefficients a, and
P; associated with the shear behavior of thin-layer element can be de-
termined from the results of appropriate laboratory tests. In this study,
a series of cyclic tests were performed with interfaces between Ottawa
sand and concrete, by using a cyclic multi-degree-of-freedom shear de-
vice (2,6). This device can be essentially viewed as a large direct shear
device in which tests can be performed under static or cyclic loading.
To perform a cyclic test, a normal load is applied first to simulate in situ
stress, and then a cyclic displacement or shear load is applied. Average
interface shear stress, T°, and relative displacement, u°, histories are
recorded at selected number of cycles, N. Fig. 5 schematically shows T°
versus u°r. The coefficients a, and p, (Eqs. 7 and 9) are determined from
such plots by using a curve fitting procedure. Typical values of these
coefficients obtained from sand-concrete interface tests are given in Ta-
ble 1. As an approximation, these values were used to analyze the model
structure described here.
1263
mode if
{(a°„)' + A<r„„} > 0 (compressive) (11)
and Fs = (sgn) {(T°)< + AT} - [c. + {(«r°,)' + Acr„„} tan 8] < 0 (12)
in which ca = activated adhesion; 8 = activated friction angle; and (sgn)
= 1 if {(T 0 )' + AT} > 0, and (sgn) = - 1 if {(T 0 )' + AT} < 0. Fs in Eq. 12
is called a slip function, and is shown in Fig. 6. The element can go into
slip mode if {(cr°„ + ACT„„} > 0 (compressive)
and Fs = (sgn) {(T 0 )' + AT} - [c. + {(*"„)' + A<r„„} tan 8] > 0 (13)
Iterations are not required if Fs = 0. Otherwise, iterations are performed
within the time step, At, until Fs ss 0 is satisfied. At each iteration, the
residual value of Fs, which has the dimension of stress, is converted
into forces and applied to the system as self-equilibrating nodal loads.
The element can go into separation or debonding mode if
{ « ) ' + Aa„„} < 0 (tensile) (14)
In the separation mode, neither normal nor shear traction can be trans-
mitted through the interface. Assume that for an element in the sepa-
ration mode, it is found that
{(<T°m)< + Acr„„} = R, (15)
0
and {(T )' + AT} = R2 (16)
Slip Cone
, Z _ ^ L \.— Rb -|x
VERIFICATION
TABLE 2.—Cap Model Parameters for Materials for Soils and Elastic Parameters
for Concrete (18,19)
Parameters Soil 1 Soil 2 Backfill Concrete
(1) (2) (3) (4) (5)
E psi 46,435 22,113 13,100 4 x 106
(kPa) (3.2 x 105) (1.5 X 105) (9.0 x 104) (2.7 X 107)
V 0.3 0.3 0.26 0.2
a psi 470.0 470.0 470.0
(kPa) (3,240.2) (3,240.2) (3,240.2)
e 0.0 0.0 0.0
P Vpsi 0.165 0.165 0.165
(1/kPa) (0.024) (0.024) (0.024)
7 psi 390.0 390.0 390.0
(kPa) (2,688.7) (2,688.7) (2,688.7)
D 1/psi 0.0007 0.0007 0.00065
(1/kPa) (0.000102) (0.000102) (0.000094)
W 0.06 0.06 0.06
z 0.0 0.0 0.0
R 2.5 2.5 2.5
-yw lb/in. 3 0.0637 0.0637 0.0637 0.058
(kg/cm3) (0.00176) (0.00176) (0.00176) (0.0016)
1267
:
;;V.\;>v,: *7 :'•''-'$& V'-i v; : '?.*:U-
U •ilfi KKVK A
Downloaded from ascelibrary.org by Ryerson University on 01/05/19. Copyright ASCE. For personal use only; all rights reserved.
• ; • ! •
^ i
; - , ••',
• ; : . • • . ' '
w
FIG. 8.—Mesh Used in Simulation of Soil-Structure Interaction Due to SIMQUAKE
II
I 5-0
puted and measured response is found to be very good for the first cycle
of strong shaking, approximately from t = 0.0-0.8 s. During this cycle,
the modeling of interface behavior as frictional or bonded does not sig-
nificantly influence the response. For the second phase of strong mo-
tion, the computed values underestimate the measured response. This
difference is particularly noticeable after t = 1.35 s. However, the overall
correlation given by the frictional interface condition is considered to be
better. After t = 1.8 s, the peak response from the bonded case is lower
in magnitude than that from the frictional case. However, after t =
2.2 s, the bonded case yields higher magnitude of the response.
Fig. 11 shows the computed horizontal displacement history at the top
of structure for both bonded and frictional interface conditions. Since
measurements were not available, only computed values are compared.
It can be noticed that for the horizontal displacement, the bonded con-
dition predicts the lower response. Particularly for t > 1 s, the horizontal
response is underestimated by the bonded interface assumption.
Distribution of Contact Stress.—In the SIMQUAKE test, normal con-
tact stress histories at selected points on the interface region were re-
corded (18). A typical comparison between the calculated and measured
contact stress histories is presented in the following description.
Fig. 12 compares measured and calculated normal contact stresses be-
neath the bottom left corner of the structure for frictional and bonded
interface conditions. In the finite element simulation with the frictional
interface (dynamic), normal tensile stress is not allowed to exceed 10.5
psi (72.4 kPa). This value corresponds to initial static stresses induced
by the self-weight of the structure. It is observed that computed and
measured stresses for the frictional case correlate well up to t = 1.0 s.
After 1.0 s, the predicted values differ significantly from the measured
values in terms of magnitude as well as phase. Peak (compressive) com-
puted responses are generally higher than the measured response. A
1269
- Fractional Interface
-Bonded Interface
1 inch = 2.54 cm
Time (sec)
reverse trend is observed for tensile contact stress. The computed in-
terface stresses represent average stresses at integration points of an in-
terface element. Thus, the computed and measured stresses do not ex-
actly correspond to same location. The difference between the computed
and predicted responses can be partly attributed to this reason. The pre-
dictions from the bonded case are quite erratic and show significant dif-
ferences in phase. In general, predictions by frictional interface condi-
1270
Based on the comparisons for the field problem and the verification
the writers believe that the proposed interface model can provide a sat-
isfactory and consistent representation of the interface behavior.
CONCLUSIONS
ACKNOWLEDGMENTS
APPENDIX.—REFERENCES
1. Bathe, K. J., and Wilson, E. L., Numerical Methods in Finite Element Analysis,
Prentice-Hall, Inc., Englewood Cliffs, N.J., 1976.
2. Desai, C. S., "Behavior of Interfaces Between Structural and Geologic Me-
dia," International Conference on Recent Advances in Geotechnical Earthquake En-
gineering and Soil Dynamics, St. Louis, Mo., 1981.
3. Desai, C. S., and Christian, J. T. (eds.), Numerical Methods in Geotechnical En-
gineering, McGraw-Hill Book Co., Inc., New York, N.Y., 1977.
4. Desai, C. S., Eitani, I. M., and Haycocks, C , "An Application of Finite Ele-
ment Procedure for Underground Structures with Nonlinear Materials and
1271
1273