You are on page 1of 13

SOIL-STRUCTURE INTERACTION

<Copyright 1998, Computers and Structures, Inc. All Rights Reserved>
http!!""".csi#er$eley.com!%ech&In'o!soilStructInter.htm
At A Finite Distance from A Structure the Absolute Displacements Must
Approach the Free-Field Displacements
INTRODUCTION
%he estimation o' earth(ua$e motions at the site o' a structure is the most important
phase o' the design or retro'it o' a structure. )ecause o' the large num#er o'
assumptions re(uired, e*perts in the 'ield o'ten disagree #y over a 'actor o' t"o as to
the magnitude o' motions e*pected at the site "ithout the structure present. %his lac$
o' accuracy o' the #asic input motions, ho"ever, does not +usti'y the introduction o'
additional unnecessary appro*imations in the dynamic analysis o' the structure and its
interaction "ith the material under the structure. %here'ore, it "ill #e assumed that the
'ree,'ield motions at the location o' the structure, "ithout the structure present, can #e
estimated and are speci'ied in the 'orm o' earth(ua$e acceleration records in three
directions. It is no" common practice, on ma+or engineering pro+ects, to investigate
several di''erent sets o' ground motions in order to consider #oth near 'ault and 'ar
'ault events.
I' a light"eight 'le*i#le structure is #uilt on a very sti'' roc$ 'oundation, a valid
assumption is that the input motion at the #ase o' the structure is the same as the 'ree,
'ield earth(ua$e motion. %his assumption is valid 'or a large num#er o' #uilding
systems since most #uilding type structures are appro*imately 9- percent voids, and,
it is not unusual that the "eight o' the structure is e*cavated #e'ore the structure is
#uilt. .o"ever, i' the structure is very massive and sti'', such as a concrete gravity
dam, and the 'oundation is relatively so't, the motion at the #ase o' the structure may
#e signi'icantly di''erent than the 'ree,'ield sur'ace motion. /ven 'or this e*treme
case, ho"ever, it is apparent that the most signi'icant interaction e''ects "ill #e near
the structure, and, at some 'inite distance 'rom the #ase o' the structure, the
displacements "ill converge #ac$ to the 'ree,'ield earth(ua$e motion.
SITE RESPONSE ANALYSIS
%he 1980 1e*ico City and many recent earth(ua$es clearly illustrate the importance
o' local soil properties on the earth(ua$e response o' structures. %hese earth(ua$es
demonstrated that the roc$ motions could #e ampli'ied at the #ase o' a structure #y
over a 'actor o' 'ive. %here'ore, there is a strong engineering motivation 'or a site,
dependent dynamic response analysis 'or many 'oundations in order to determine the
'ree,'ield earth(ua$e motions. %he determination o' a realistic site,dependent 'ree,
'ield sur'ace motion at the #ase o' a structure can #e the most important step in the
earth(ua$e resistant design o' any structure.
2or most hori3ontally layered sites a one dimensional pure shear model can #e used to
calculate the 'ree,'ield sur'ace displacements given the earth(ua$e motion at the #ase
o' a soil deposit. 1any special purpose computer programs e*ist 'or this purpose.
S.A4/ 516 is a "ell,$no"n program, #ased on the 're(uency domain solution
method, "hich iterates to estimate e''ective linear sti''ness and damping properties in
order to appro*imate the nonlinear #ehavior o' the site. 7A8/S 596 is a ne" nonlinear
program in "hich the nonlinear e(uations o' motion are solved #y a direct step,#y,
step integration method. I' the soil material can #e considered linear then the
SA:9--- program, using the S;<I= element, can #e used to calculate either the one,
t"o or three dimensional 'ree,'ield motions at the #ase o' a structure. In addition, a
one dimensional nonlinear site analysis can #e accurately conducted using the 2>A
option in the SA:9--- program.
KINEMATIC OR SOIL-STRUCTURE INTERACTION
%he most common soil,structure interaction SSI approach, used 'or three dimensional
soil,structure systems, is #ased on the ?added motion? 'ormulation 5@6. %his
'ormulation is mathematically simple, theoretically correct, and is easy to automate
and use "ithin a general linear structural analysis program. In addition, the
'ormulation is valid 'or 'ree,'ield motions caused #y earth(ua$e "aves generated
'rom all sources. %he method re(uires that the 'ree,'ield motions at the #ase o' the
structure #e calculated prior to the soil,structure interactive analysis.
In order to develop the 'undamental SSI dynamic e(uili#rium e(uations consider the
three dimensional soil,structure system sho"n in 2igure 1A.1.

Figure 16.1. Soil-
Structure Interaction
Model





Consider the case "here the SSI model is divided into three sets o' node points. %he
common nodes at the inter'ace o' the structure and 'oundation are identi'ied "ith "c"B
the other nodes "ithin the structure are "s" nodesB and the other nodes "ithin the
'oundation are "f" nodes. 2rom the direct sti''ness approach in structural analysis, the
dynamic 'orce e(uili#rium o' the system is given in terms o' the absolute
displacements, , #y the 'ollo"ing su#,matri* e(uation
C1A.1D
"here the mass and the sti''ness at the contact nodes are the sum o' the contri#ution
'rom the structure (s) and 'oundation (f), and are given #y
C1A.9D
In terms o' a#solute motion, there are no e*ternal 'orces acting on the system.
.o"ever, the displacements at the #oundary o' the 'oundation must #e $no"n. In
order to avoid solving this SSI pro#lem directly, the dynamic response o' the
'oundation "ithout the structure is calculated. In many cases, this free-field solution
can #e o#tained 'rom a simple one,dimensional site model. %he three dimensional
'ree,'ield solution is designated #y the a#solute displacements and a#solute
accelerations . )y a simple change o' varia#les it is no" possi#le to e*press the
a#solute displacements and accelerations in terms o' displacements relative to
the 'ree,'ield displacements . ;r,
C1A.@D
/(uation C1A.1D can no" #e "ritten as
C1A.ED
I' the 'ree,'ield displacement is constant over the #ase o' the structure, the term
is the rigid #ody motion o' the structure. %here'ore, /(uation C1A.ED can #e 'urther
simpli'ied #y the 'act that the static rigid #ody motion o' the structure is
C1A.0D
Also, the dynamic 'ree,'ield motion o' the 'oundation re(uires that
C1A.AD
%here'ore, the right,hand side o' /(uation C1A.ED can #e "ritten as
C1A.FD
.ence, the right,hand side o' the /(uation C1A.ED does not contain the mass o' the
'oundation. %here'ore, three dimensional dynamic e(uili#rium e(uations, 'or the
complete soil,structure system "ith damping added, are o' the 'ollo"ing 'orm 'or a
lumped mass system
C1A.8D
"here M, C and K are the mass, damping and sti''ness matrices, respectively, o' the
soil,structure model. %he added, relative displacements, u, e*ist 'or the soil,structure
system and must #e set to 3ero at the sides and #ottom o' the 'oundation. %he terms
and are the 'ree,'ield components o' the acceleration i' the structure is not
present. %he column matrices, , are the directional masses 'or the added structure
only.
1ost structural analysis computer programs automatically apply the seismic loading
to all mass degrees,o','reedom "ithin the computer model and cannot solve the SSI
pro#lem. %his lac$ o' capa#ility has motivated the development o' the massless
'oundation model. %his allo"s the correct seismic 'orces to #e applied to the structureB
ho"ever, the inertia 'orces "ithin the 'oundation material are neglected. %he results
'rom a massless 'oundation analysis converge as the si3e o' the 'oundation model is
increased. .o"ever, the converged solutions may have avoida#le errors in the mode
shapes, 're(uencies and response o' the system.
%o activate the soil,structure interaction "ithin a computer program it is only
necessary to identi'y the 'oundation mass in order that the loading is not applied to
that part o' the structure. %he program then has the re(uired in'ormation to 'orm #oth
the total mass and the mass o' the added structure. %he SA:9--- program has this
option and is capa#le o' solving the SSI pro#lem correctly.
RESPONSE DUE TO MULTI-SUPPORT INPUT MOTIONS
%he previous SSI analysis assumes that the 'ree,'ield motion at the #ase o' the
structure is constant. 2or large structures such as #ridges and arch dams the 'ree,'ield
motion, at all points "here the structure is in contact "ith the 'oundation, is not
constant.
%he approach normally used to solve this pro#lem is to de'ine a quasi-static
displacement that is calculated 'rom the 'ollo"ing e(uation
C1A.9aD
%he trans'ormation matri* allo"s the corresponding (uasi,static acceleration in
the structure to #e calculated 'rom
C1A.9#D
/(uation C1A.ED can #e "ritten as
C1A.1-D
A'ter su#stitution o' /(uations C1A.AD and C1A.9D, /(uation C1A.1-D can #e "ritten as
C1A.11D
%he reduced structural sti''ness at the contact sur'ace is given #y
C1A.19D
%here'ore, this approach re(uires a special program option to calculate the mass and
sti''ness matrices to #e used on the right,hand side o' the dynamic e(uili#rium
e(uations. >ote that the loads are a 'unction o' #oth the 'ree,'ield displacements and
accelerations at the soil,structure contact. Also, in order to o#tain the total stresses and
displacements "ithin the structure the (uasi,static solution must #e added to the
solution. At the present time, there is not a general,purpose structural analysis
computer program that is #ased on this ?numerically cum#ersome? approach.
An alternative approach is to 'ormulate the solution directly in terms o' the a#solute
displacements o' the structure. %his involves the introduction o' the 'ollo"ing change
o' varia#les
C1A.1@D
Su#stitution o' this change o' varia#les into /(uation C1A.1D yields the 'ollo"ing
dynamic e(uili#rium e(uations in terms o' the a#solute displacement, , o' the
structure
C1A.1ED
A'ter the 'ree,'ield response, /(uation C1A.AD, is removed the dynamic loading is
calculated 'rom the 'ollo"ing e(uation
C1A.10aD
%his e(uation can #e 'urther simpli'ied #y connecting the structure to the 'oundation
"ith sti'' massless springs that are considered as part o' the structure. %here'ore, the
mass o' the structure at the contact nodes is eliminated and /(uation C1A.10aD is
reduced to
C1A.10#D
It is apparent that the sti''ness terms in /(uation C1A.10#D represent the sti''ness o' the
contact springs only. %here'ore, 'or a typical displacement component Cn G *, y or 3D,
the 'orces acting at point ?i? on the structure and point ?+? on the 'oundation are given
#y
C1A.1AD
"here is the massless spring sti''ness in the nth direction and is the 'ree,'ield
displacement. .ence, points ?i? and ?+? can #e at the same location in space and the
only loads acting are a series o' time,dependent, concentrated, point loads that are
e(ual and opposite 'orces #et"een the structure and 'oundation. %he spring sti''ness
must #e selected appro*imately three orders,o',magnitude greater than the sti''ness o'
the structure at the connecting nodes. %he spring sti''ness should #e large enough so
the 'undamental periods o' the system are not changed, and small enough not to cause
numerical pro#lems.
%he dynamic e(uili#rium e(uations, "ith damping added, can #e "ritten in the
'ollo"ing 'orm
C1A.1FD
It should #e pointed out that concentrated dynamic loads generally re(uire a large
num#er o' eigenvectors in order to capture the correct response o' the system.
.o"ever, i' <=R vectors are used, in a mode superposition analysis, the re(uired
num#er o' vectors is reduced signi'icantly. %he SA:9--- program has the a#ility to
solve the multi,support, soil,structure interaction pro#lems using this approach. At the
same time, selective nonlinear #ehavior o' the structure can #e considered.
ANALYSIS OF GRAVITY DAM AND FOUNDATION
In order to illustrate the use o' the soil,structure interaction option several earth(ua$e
response analyses o' the :ine 2lat =am "ere conducted "ith di''erent 'oundation
models. %he 'oundation properties "ere assumed to #e the same properties as the
dam. =amping "as set at 'ive percent. %en Rit3 vectors, generated 'rom loads on the
dam only, "ere used. .o"ever, the resulting appro*imate mode shapes, used in the
standard mode superposition analysis, included the mass inertia e''ects o' the
'oundation. %he hori3ontal dynamic loading "as the typical segment o' the <oma
:rieta earth(ua$e de'ined in 2igure 10.1a. A 'inite element model o' the dam on a
rigid 'oundation is sho"n in 2igure 1A.9.
Figure 16.2. Finite Element Model of Dam only
%he t"o di''erent 'oundation models used are sho"n in 2igure 1A.@.


Figure 16.. Models of Dam !it" Small and #arge Foundation
Selective results are summari3ed in %a#le 1A.1. 2or the purpose o' comparison, it "ill
#e assumed that Rit3 vector results, 'or the large 'oundation mesh, are the re'erenced
values.
$able 16.1. Selecti%e &esults 'f Dam-Foundation (nalyses


=A1 7I%.
>;
S1A<<
2oundation
<ARH/
2oundation
2oundation
%;%A< 1ASS l#,sec
9
!in 1,8F- 1@,90- FF,@A-
:/RI;=S seconds -.@@0 -.108 -.E-E -.91- -.E00 -.@F1
1a*. =isplacement
inches
-.A0 1.98 1.@1
1a* I 1in Stress $si ,@F to J@8@ ,E9- to J989 ,019 to J99F
%he di''erences #et"een the results o' the small and large 'oundation models are very
close "hich indicates that the solution o' the large 'oundation model may #e nearly
converged. It is true that the radiation damping e''ects in a 'inite 'oundation model are
neglected. .o"ever, as the 'oundation model #ecomes larger, the energy dissipation
due to normal modal damping "ithin the massive 'oundation is signi'icantly larger
than the e''ects o' radiation damping 'or transient earth(ua$e type o' loading.
THE MASSLESS FOUNDATION APPROXIMATION
1ost general purpose programs 'or the earth(ua$e analysis o' structures do not have
the option o' identi'ying the 'oundation mass as a separate type o' mass on "hich the
earth(ua$e 'orces do not act. %here'ore, an appro*imation that has commonly #een
used is to neglect the mass o' the 'oundation completely in the analysis. %a#le 1A.9
summari3es the results 'or an analysis o' the same dam,'oundation systems using a
massless 'oundation. As e*pected, these results are similar. 2or this case the results
are conservativeB ho"ever, one cannot #e assured o' this 'or all cases.
$able 16.2. Selecti%e &esults 'f Dam )it" Massless Foundation (nalyses


=A1 7I%.
>; 2oundation
S1A<<
2oundation
<ARH/
2oundation
%;%A< 1ASS l#,sec
9
!in 1,8F- 1,8F- 1,8F-
:/RI;=S seconds -.@@0 -.108 -.E-- -.190 -.E10 -.9-F
1a*. =isplacement
inches
-.A0 1.9F 1.E@
1a* I 1in Stress $si ,@F to J@8@ ,E8- to J989 ,00- to J@@-


APPROXIMATE RADIATION BOUNDARY CONDITIONS
I' the 'oundation volume is large and the modal damping e*ists, it "as demonstrated
in the previous section that a 'inite 'oundation "ith 'i*ed #oundaries can produce
converged results. .o"ever, the use o' energy a#sor#ing #oundaries can 'urther
reduce the si3e o' the 'oundation re(uired to produce a converged solution.
In order to calculate the properties o' this #oundary condition consider a plane "ave
propagating in the *,direction. %he 'orces, "hich cause "ave propagation, are sho"n
acting on a unit cu#e in 2igure 1A.E.
Figure 16.*. Forces (cting on +nit ,ube
2rom 2igure 1A.E the one dimensional e(uili#rium e(uation in the *,direction is
C1A.18D
Since the one dimensional partial di''erential e(uation is "ritten in
the 'ollo"ing classical "ave propagation 'orm
C1A.19D
"here is the "ave propagation velocity o' the material and is given #y
C1A.9-D
in "hich is the mass density and is the #ul$ modulus given #y
C1A.91D
%he solution o' /(uation C1A.1@D, 'or harmonic "ave propagation in the positive *,
direction, is a displacement o' the 'ollo"ing 'orm
C1A.99D
%his e(uation can #e easily veri'ied #y su#stitution into /(uation C1A.18D. %he
ar#itrary 're(uency o' the harmonic motion is . %he velocity, , o' a particle at
location * is
C1A.9@D
%he strain in the *,direction is
C1A.9ED
%he corresponding stress can no" #e e*pressed in the 'ollo"ing simpli'ied 'orm
C1A.90D
%he compression stress is identical to the 'orce in a simple viscous damper "ith
constant damping value e(ual to per unit area o' the #oundary. %here'ore, a
#oundary condition can #e created, at a cut #oundary, "hich "ill allo" the "ave to
pass "ithout re'lection and allo" the strain energy to ?radiate? a"ay 'rom the
'oundation.
Also, it can #e easily sho"n that the shear "ave ?radiation? #oundary condition,
parallel to a 'ree #oundary, is satis'ied i' damping values are assigned to #e per
unit o' #oundary area. %he shear "ave propagation velocity is given #y
C1A.9AD
"here is the shear modulus.
%he 2>A method can #e used to solve structures, in the time domain, "ith these types
o' #oundary conditions. In later editions o' this #oo$, the accuracy o' these #oundary
conditions appro*imation "ill #e illustrated "ith numerical e*amples. Also, it "ill #e
used "ith a 'luid #oundary "here only compression "aves e*ist.
USE OF SPRINGS AT THE BASE OF A STRUCTURE
Another important structural modeling pro#lem, "hich must #e solved, is at the
inter'ace o' the ma+or structural elements "ithin a structure and the 'oundation
material. 2or e*ample, the de'ormations at the #ase o' a ma+or shear "all in a #uilding
structure "ill signi'icantly a''ect the displacement and 'orce distri#ution in the upper
stories o' a #uilding 'or #oth static and dynamic loads. Realistic spring sti''ness can
#e selected 'rom separate 'inite element studies or #y using the classical hal',space
e(uations that are given in %a#le 1A.@.
It is the opinion o' the author that the use o' appropriate site,dependent 'ree,'ield
earth(ua$e motions and selection o' realistic massless springs at the #ase o' the
structure are the only modeling assumptions re(uired to include site and 'oundation
properties in the earth(ua$e analysis o' most structural systems.
%a#le 1A.@ also contains e''ective mass and damping 'actors that include the
appro*imate e''ects o' radiation damping. %hese values can #e used directly in a
computer model "ithout any di''iculty. .o"ever, considera#le care should #e ta$en in
using these e(uations at the #ase o' a complete structure. 2or e*ample, the e''ective
earth(ua$e 'orces must not #e applied to the 'oundation mass.
Table 16.3. Properties Of Rigid Circular Plate On Surface Of Half-Space

=IR/C%I;> S%I22>/SS =A1:I>H 1ASS
8ertical
.ori3ontal
Rotation
%orsion
plate radiusB shear modulusB :oissonKs ratioB mass density
Source Adapted 'rom ?2undamentals o' /arth(ua$e /ngineering, #y >e"mar$ and Rosen#lueth, :rentice,.all, 19F1
SUMMARY
A large num#er o' research papers and several #oo$s have #een "ritten on structure,
'oundation,soil analysis and site response due to earth(ua$e loading. .o"ever, the
ma+ority o' these pu#lications have #een restricted to the linear #ehavior o' soil,
structure systems. It is possi#le, "ith the use o' the numerical methods presented in
this #oo$, to conduct accurate earth(ua$e analysis o' real soil,structure systems in the
time domain, including many realistic nonlinear properties. Also, it can #e
demonstrated that the solution o#tained is converged to the correct soil,structure
interactive solution.
2or ma+or structures on so't soil one dimensional site response analyses should #e
conducted. Lnder ma+or structural elements, such as the #ase o' a shear "all, massless
elastic springs should #e used to estimate the 'oundation sti''ness. 2or massive
structures, such as gravity dams, a part o' the 'oundation should #e modeled #y three
dimensional S;<I= elements in "hich SSI e''ects are included.
REFERENCES
?S.A4/ , A Computer :rogram 'or the /arth(ua$e Response 'or .ori3ontally
<ayered Sites?, #y :. Schna#el, M. <ysmer and .. Seed, //RC Report >o. F9,9,
Lniversity o' Cali'ornia, )er$eley, 2e#ruary 19F-.
M. .art. and /. 7ilson ?7A8/S , An /''icient 1icrocomputer :rogram 'or >onlinear
Site Response Analysis?, >ational In'ormation Center 'or /arth(ua$e /ngineering,
=avis .all, Lniversity o' Cali'ornia, )er$eley, %el. N CE10D AE9,011@.
R. Clough, and M. :en3ien, Dynamics of Structures, Second /dition, 1cHra",.ill,
Inc., IS)> -,-F,-11@9E,F, 199@.