Methods of Random Wibration For Inelastic Structures
Methods of Random Wibration For Inelastic Structures
Y K Wen
Department of Civil Engineering, University of Illinois at Urbana-Champaign, Urbana, IL
Structures often become nonlinear and inelastic under severe lateral force pro-
duced by natural hazards such as earthquakes, severe winds and waves. The
restoring force of the structures under such load may become hysteretic and
deteriorate in strength or stiffness, or both. This paper gives an overview of the
major developments in the modeling and response analysis of inelastic structures
under random excitation. It includes: (1) modeling of the hereditary behavior of
inelastic system; (2) methods of solution based on semi-empirical approaches,
Fokker-Planck equation, and equivalent linearization; and (3) applications to
performance and safety evaluation of real structural systems. Limitations of
current methods are mentioned and suggestions of areas for further research are
given.
Appl Mech Rev vol 42, no 2, Feb 1989 39 © Copyright 1989 American Society of Mechanical Engineers
(a) (b)
__———f"~""
o f -—————
' " s^/
o
and u+ and iC are the additional state variables which are the hysteretic components in the two directions, ie zx and z.,
maximum response in the positive and negative directions, satisfy the following coupled differential equations (Park, Wen,
respectively. u + and u~ satisfy the following additional differ- and Ang, 1986):
ential equations:
zx = Aux — $\iixzx\zx — yuxzx — fi\iiyzv\zx — yuvzxzv,
u+=xU(x)U(z~-l), (10)
z =Au ~ P\u z \z - yu zj - P\itxzx\z - yuxzxz ,
u~= -xU(-x)U(-z-l). (9)
The model can be further generalized to reproduce the Takeda where ux and ur are, respectively, the displacements in the x
type hysteresis by introducing additional system parameters and y directions. A, fi, and y as in Eq (3) are system parame-
(Suzuki and Minai, 1987). ters. The above equations give an isotropic hysteretic restoring
2.2.4. Biaxial interaction force. It can be easily demonstrated that both equations reduce
For two-dimensional structures under biaxial excitations, the to the form of Eq (3) for the case of uniaxial loading along the
interaction of the restoring forces in the two directions may arbitrary direction. For an orthotropic system whose stiffness
significantly alter the response behavior. For example, the dam- and strength in the two directions are different, one can intro-
age suffered from oscillations in one direction is likely to duce a simple transformation (scaring) of the response variables
weaken the strength and/or stiffness in the other direction and and still use the same equations (Park, Wen, and Ang, 1986). As
vice versa. The restoring force model given in Eq (3) has been in the uniaxial model, deterioration can be introduced by letting
extended to include such interaction by requiring that the parameters A, ft, and y be functions of time depending on the
- 3 - 2 - 1 O 1 2 - 3 - 2 - 1 O 1 2
DISPLACEMENT UX(T) DISPLACEMENT UX(T)
- 3 - 2 - 1 O 1 2
DISPLACEMENT UY(T) - 3 - 2 - 1 0 1 2
DISPLACEMENT UY(T)
(a) (b)
FIG. 6. Nondegrading biaxial smooth hysteretic restoring force: (a) diamond displacement path; (b) square displacement path
(Ref 47).
or
o
x -400 -
Z 400 -
10
U
a.
o
• eoo
FIG. 7. Force-displacement relationship of analytical model by Powell and Chen (1986); ( ) hinge model; (—) fiber model.
severity of the response, eg, maximum response amplitude or both uniaxial and biaxial models (Sues et al, 1988; Wen and
hysteretic energy dissipation given by Ang, 1987). Other methods such as those based on an extended
Kalman filter method have been applied to the smooth hys-
iT0)=f^Ar)uAr)+zy(r)K(r)]dr (11) teretic restoring force with success (Hoshiya and Maruyama,
1987). Methods for estimating the system parameters of the
or both. distributed element model and the Masing models were pro-
The accuracy and capability of this method of modeling is posed by Peng and Iwan (1987) and Jayakumar and Beck
indicated by comparisons of the force-displacement relation- (1987). The general theoretical aspects of the estimation of
ship with those based on plasticity theory and experimental hysteretic systems have been reviewed by Minai and Suzuki
studies. Figures 6 and 7 show the nondegrading system accord- (1987).
ing to Eq (10) under different displacement paths and the
corresponding analytical solutions by Powell and Chen (1986)
based on plasticity theory. The agreements are surprisingly 3. METHOD OF SOLUTION UNDER
good considering the generally complicated biaxial inelastic RANDOM EXCITATION
stress-strain relationship and the simple and explicit nature of
the model [Eq (10)]. Figure 8 shows the experimental results of The hereditary behavior of the restoring force of the inelastic
an RC column under biaxial load (Takizawa and Aoyama, system makes analytical solution of the response extremely
1976) and the corresponding degrading system based on the difficult. Approximations are almost always necessary. The re-
model. cent progress in this area is summarized as follows.
The smooth system has been extended to continuous (plain
strain) systems by Mochio, Samaras, and Shinozuka (1985).
3.1. K - B approximation for bilinear systems
Park (1988) extended the biaxial smooth system model to
describe the two-dimensional stress-strain relationship in an Random vibration of an inelastic system was first investi-
analytical model for shear walls in a building. gated by Caughey (1960) in the context of a bilinear system
2.2.5. System parameter identification under white noise excitation. The response is assumed to have a
To properly model the restoring force of actual structures, slowly varying amplitude and phase and is averaged over one
the force parameters need to be determined from test or field cycle of oscillation according to the hysteresis of the bilinear
data; therefore, an estimation procedure is required for this system. The amplitude is then assumed to follow a Raleigh
purpose. A simple technique based on a least square error distribution and the response statistics are obtained from an
minimization has been developed for the smooth hysteretic equivalent linearization procedure. The equivalent linear system
models given in Eqs (3) and (10) with encouraging results for parameters, ie, the damping and the stiffness coefficients, are
A^\
X-Dlrecllon X-Dlrocllon
i¥~y
\ \\ /
• V___——-~-~~"^
-1.6 -1.6
u (cm) u (cm)
(Test:) (Model)
c
2 ,
Y-Dlractlo(\
m
^
Y-Olractlon
:v£U
r / /
•
(a) (b)
FIG. 8. Comparison of test (after Takizawa and Aoyama, 1976) and model results.
given in integral forms. The counteracting effect of loss of 3.2. Semiempirical method for elastic-plastic systems
stiffness and increase in damping due to hysteresis in an inelas-
tic system was clearly demonstrated. For a truly elasto-plastic system [eg, a = 0 in Eq (2)] under
stationary random excitation, once the yield level is reached,
The slowly varying parameter Krylov-Bogoliubov (K-B)
there is no additional restoring force present in the system to
method, however, implies that the response undergoes displace-
resist further displacement. The response can drift with time
ment reversals in each cycle of oscillation which is certainly not
and increase without bound. This plastic drift or permanent
the case for inelastic systems, which are known to have a
set was investigated by Karnopp and Scharton (1966) and
tendency to drift under random excitation. Iwan and Lutes
Vanmarcke (1976). Their studies were based on a physical
(1968) used electronic-analog techniques to simulate the ran-
observation of the transfer of the kinetic energy of the linear
dom response of bilinear systems. Their results indeed showed
system at the yield threshold to plastic deformation and simula-
that the K - B technique gives good results for systems with
tion results of such systems under the excitation of white noise
small to moderate nonlinearity but may significantly underesti-
or filtered white noise. Assuming that the total plastic deforma-
mate the rms response for nearly elasto-plastic systems. It is
tion D(t), at a given time /, is an cumulation of plastic
shown that for such systems and excitation of certain intensity
deformation increments resulting from yield-level crossings by
the response process is quite wide band which explains why the
the linear system, Vanmarcke showed that D(t) has a zero
K - B technique, essentially a narrow-band assumption, would
mean and a variance which increases linearly with time. The
not work well. They also showed the shifting of the apparent
variance is given by
frequency and the change in the spectral density function of the
yielding systems with the excitation level and the noticeable
non-Gaussian nature of the response. Var[ £>(/)] =2atS2 (12)
where a is the yield level clump uncrossing rate of the corre- hysteretic model, however, allows such linearization to be per-
sponding linear system and 8 is the average amount of inelastic formed in closed form without recourse to the K-B technique,
deformation resulting from a single crossing of the yield level as has been shown by Wen (1980) and Baber and Wen (1981).
and is approximately equal to a2/2Y, where ax is the mean This greatly facilitates the applicability of the linearization
square response of the linear system and Y is the yield level. method to multi-degree-of-freedom systems and degrading sys-
The same semiempirical approach to the drift analysis was tems. For example, for a nondegrading SDOF system with a
recently used by Iwan et al (1987). The response was separated storing force given by Eq (3) (n = ij = v = 1) under zero-mean,
into an elastic part and a drift part [Eq (5)]. The model of Gaussian excitation, by assuming the response variables to be
cumulation of plastic deformations was retained; however, eval- also jointed Gaussian, Eq (3) can be linearized as
uation of the increments was based on simulations, a Markov
chain model, and a detailed analysis of the crossings of the i + q i + c2z = 0, (13)
response process at the yield level. The computation involved in where
this recently proposed method is quite extensive. The results
compared very well with simulations, including the prediction 'l/^[yE(xz)/ai + Ra:] -A, (14)
of the drift in the system even when the excitation has a BE(xz)/az], (15)
spectral density function which vanishes as the frequency goes
2 1/2 2 1 2
to zero. This is the situation for which the analysis based on tT< = [ £ ( ; t ) ] , oz = [E(z )} / , and £ [ ] = expected value.
equivalent linearization fails to predict the drift, as will be The extension to multi-degree-of-freedom systems and degrad-
discussed in the following. Figure 9 shows a comparison of the ing systems can be found in Baber and Wen (1981). The
solution based on this method with simulation results for such equation of motion of the system together with Eq (13) forms a
an excitation. Earlier, Grossmayer (1981) proposed a somewhat third-order linear oscillator which allows more freedom in the
similar" procedure for bilinear systems with good results. response than the conventional second-order system. For exam-
ple, it can be shown (Wen, 1988) that, under arbitrary excita-
tion, the response of this third-order system consists of two
3.3. Equivalent linearization method for smooth systems parts: an oscillatory part which corresponds to the convolu-
tional integral of a conventional second-order system, and a
As mentioned in the foregoing, the equivalent linearization nonoscillatory (drift) part which is proportional to the time
method has been used by Caughey for bilinear systems under integration of the forcing function. In other words, if the
white noise excitation. Recent work by Iwan (1973) and Spanos excitation has a steady, nonoscillatory component this drift part
(1978) have laid sound theoretical foundation for the method of the response will be greatly amplified. This preserves to some
for stationary as well as nonstationary analysis. An excellent extent one of the most important properties of an inelastic
literature survey of the method was given by Spanos (1981). For system.
the hysteretic system as indicated in the foregoing a major
For excitation that can be modeled by a filtered Gaussian
difficulty is to perform the linearization in closed form because
shot noise the one-time response variable (displacement, veloc-
of the hereditary behavior of the restoring force. The smooth,
ity, and the hysteretic part of the restoring force) covariance
matrix [S] of the linearized system satisfies the matrix ordinary
differential equation
3.0
— NUMERICAL SIMULATION d[S]
— SIMPLIFIED MODEL (EQH. U ) [G][S]+[S][G]' = [B] (16)
-- LINEARIZATION (THIRD ORDER)
dt
2.0
where [G] is the matrix of structural system (including the
linearization coefficients) and excitation (filter) parameters, /
indicates transpose, and [B] is a matrix of the expected values
1.0
of the product of the response vector and the shot noise
excitation. The stationary solution for a nondeteriorating sys-
tem is obtained by solving Eq (16) with d[S]/dt = 0. As in all
s.. 0.0 equivalent Unearization methods, iteration is generally required
because the Unearization coefficients are functions of the re-
sponse variable statistics. In this regard, the algorithm by
Battels and Stewart (1972) is especially efficient for solution of
the equations of this type. For the nonstationary solution, Eq
(16) can be integrated numerically. Note that, in the above
solution procedure, no additional approximations, such as those
impUed in a K - B technique, have been made other than that in
the equivalent linearization and the Gaussian assumption for
the response variables. Therefore, the solution so obtained is the
best linear approximation of the original nonlinear system
under the circumstances.
0.0 For degrading systems, to keep the analysis tractable, an
0.0 5.0 15.0 20.0 25.0 additional approximation is necessary, ie, A, TJ, and v are
lT
assumed to be slowly varying and can be approximated by their
NORMALIZED DURATION l° mean values fiA, /i , and n„. Therefore, taking expected values
of the time derivatives, for example, of Eq (6), one obtains the
FIG. 9. Transient drift response of elasto-plastic system: (a) w?/w0
:
0.7 and (b) u„/w0 = 1.0, w = ground frequency, co0 = structural nat- time derivative of \xA as a function of the expected energy
ural frequency (Ref 19). ( ) Numerical simulation; (- simpli- dissipation rate, ie, (1 - a)kE[xz]. Since E(xz) is part of the
fied model [Eq (14)]; (---) linearization (third order). solution [S], nA can be updated at each time step to reflect the
system deterioration. Similarly one can update fi^ and /t„. The
above approximation has been verified by extensive Monte aJD
Carlo simulations (Baber and Wen, 1981). ax/D ay/D
The power spectral density matrix of the response variables or ?o = 0 simulation />
(in the case of the stationary response) can be obtained through
(Jy/D
an eigen analysis of the matrix [G]. For a multi-DOF system,
the response power spectral density matrix is given by (Baber £c = 0.05
and Wen, 1981)
W;v(a>)=0[lW-2]~y^*[/<o/-2]~1V', (17)
D/Y
where Wff = constant excitation power spectral density matrix,
<3> and \j/ are the left and right eigenvector matrices of [G],
I = identity matrix and 2 = the diagonal matrix of eigenvalues, i '
t indicates transpose, and the asterisk indicates complex conju- ass,, analytical
, fj.a
gate. Note that [G] is not a symmetric matrix. Although a * simulation
ajax - m
frequency domain approach is possible through iterative solu- or o.e
tions of Eqs (14), (15), and (17), the computational effort can be ax/axco0 /
extensive, since numerical integration is generally required in - OzJOx \ \ y Ox/OxVQ
each iteration. Therefore, a time domain solution for the system
coefficients first is preferred if the power spectral density is
needed. The statistics of maximum response can be obtained by -
I ""•"-—¥ , , , , . , ,
using currently available approximate procedures, eg, that based
on a Poisson outcrossing assumption. A response quantity that D/Y
is particularly useful for predicting potential structural damage
is the total hysteretic energy dissipation eT(t) of Eq (7). The
1 . ( i
mean value of tr(t) is • •' * - ~ -
- •
exzx - / *
E[eT(t)]=(l-a)kf'E[2(r)x(r)]dT. (18) "
•'o or o.e
Qxzx —
x\ analytical
- A simulation -
~
As E{zx) is part of the solution of [S], E[eT(t)] can be easily B
•
V
evaluated. Evaluation of the variance of eT(t) requires the
solution of the covariance matrix between two time instants.
•
/ \ •
\/
Details of the solution procedure can be found in Pires, Wen, - •
strong motion earthquake excitation. In most studies the smooth where D = damage index (positive values represent different
hysteresis model and the equivalent linearization method were degrees of damage and D > 1.0 represents collapse), 8U =
used. The obvious reason is that actual buildings and structures ultimate deformation under static loadings, Q = yield strength,
generally are complex and need to be treated as multi-degree- ft = nonnegative constant, and Qv, 8U, and ft are parameters of
of-freedom systems. Also, deterioration in such systems occurs structural (or member) capacity against damage and collapse. A
as a rule rather than exception when under severe excitation large set of US and Japanese test data of reinforced concrete
such as those due to earthquakes. This particular method of beams and columns tested to failure was analyzed to determine
analysis is, therefore, most suitable under such circumstances. the value of Qv, Su, and /? and to determine the uncertainty in
The applications include response and damage analysis of the ultimate member capacity in terms of D. The overall
reinforced concrete and masomy structures, soil-structural in- damage index DT of a structure can be evaluated from those of
teraction, and the liquefaction of soil deposits. The validity of the members based on an energy consideration. It reflects the
the analysis method has been verified by comparison of results damage concentration at the weakest part of a building (eg, the
with actual field evidence. first story or top story as frequently observed as well as damage
distribution throughout a building).
In terms of the damage index DT, nine reinforced concrete
4.1. Response and damage study of reinforced concrete and buildings that were moderately or severely damage during the
masonry structures 1971 San Fernando earthquake and the 1978 Miyahiken-Oki
earthquake in Japan were evaluated by Park et al (1985). Figure
Reinforced concrete buildings generally show hysteretic and
13 gives the plans and elevations of these buildings; earthquake
degrading response behavior under severe earthquake excita-
excitation is assumed in the longitudinal direction. In calculat-
tion. Sues et al (1985) applied the methodology to a number of
ing DT by the random vibration method, the best estimated
reinforced concrete buildings that were damaged during recent
intensity and frequency content of the excitation at the site
earthquakes. Since considerable uncertainty exists in the struc-
were used. Figure 14 shows the calculated DT vs the observed
tural system itself, in addition to structural response, they also
damage, which provide a basis for defining the significance of
considered the uncertainty of the structural system parameters
DT values. It is found that the index value D, < 0.4 approxi-
(such as stiffness, damping, and hysteretic force parameters)
mately corresponds to reparable damage, whereas DT > 0.4
and sensitivity of the response to changes in the structural
represents damage beyond repair and DT> 1.0 represents total
parameters. They obtained the sensitivity coefficients (the rate
collapse. A damage limiting design procedure based on the
of change of response statistics with the system parameter)
above consideration has also been proposed by Park et al
required in the analysis by solving the following equation,
(1987).
which can be derived from Eq (16):
d I dS dS dG dS dGT dB
G — + —- S • (20)
Tt\~dp~ dp dp dp dp dp '
(A)
j-TTT-3^"-
where dS/dp is the unknown matrix of the sensitivity coeffi-
cient. The above equations are linear and can be solved easily.
! * isr „ . . .
J-—-, W Direction of Analysis
a,b
Note that typical sensitivity analyses require repeated solutions ' • • • |-4-4—|j
tfct Direction of Pholo, a (orb)
of the original nonlinear system.
The variance due to parameter uncertainty was then com-
bined with that due to random vibration to arrive at the overall ,a,b \
(B) { c )
COLLAPSE
P A R T I A L COLLAPSE
-HHJ3I,
WI2 HP27
HHC8
SEVERE
HHC3
DAMAGE W14 HHJ30
o
D5
HYI9
WI3
MODERATE o
W4 HHJ33
DAMAGE
W2
J2A*.
HHJ29
Y21 HP23
o
MINOR HHCH
DAMAGE m D2
YiO
W18 HHC13
o o
SLIGHT W17
o HHCH
o
DAMAGE W7 W8
o o
Y6 W10
_1_
0.4 06
on
DAMAGE INDEX
FIG. 15. Comparison of calculated mean damage index with observed damage of masonry buildings (Ref 25).
i - 1
A = 1.0 0 = 0.417
n = 0.50 7 - 83.4
320.
Iff '
E
I
en 160. -
0.
//
en
160. -
/
in
CO
320. - -
-480. 1
-a x )z x k x
*o(t)
M(t)
-A/VWVW- -H(t)
U_ e ( t )
which is basically a softening system. The superstructures (the FIG. 19. Effect of soil structural interaction and system nonlinear-
building and foundation) were modeled as inelastic structures. ity: ( ) linear system; (• • •) nonlinear structural with linear founda-
Extensive parametric studies were carried out to examine the tion; ( ) nonlinear structure with foundation uplifting; ( — — ) a
nonlinear structure, soil, with foundation uplifting (Ref 12).
effect of soil-structure interaction and the importance of vari-
ous nonlinear effects under random excitations. Typical results
are shown in Fig. 19 where the ratio of the rms displacement of
a coupled system to that of a fixed based system as a function effects were found to be generally small. Details can be found in
of the structure/soil stiffness ratio is shown. Different curves Chu et al (1984).
refer to different combinations of structural and soil model Under cyclic excitation, the shear stress may deteriorate at a
(linear or nonlinear) and whether uplifting is included, r is the fast rate in a saturated sand deposit because of the excess pore
structural height-to-base aspect ratio. It is seen that the pressure rise, resulting in failure of the soil, known as liquefac-
soil-structure interaction may increase or decrease the rms tion. Pires et al (1983) used the smooth hysteretic model and
response depending on the structural height-to-base aspect ra- random vibration method to study the liquefaction of sand
tio. Also, the nonlinearity in the superstructure contributes a deposits during earthquakes in which the random nature of the
dominant role in determining the structural response. Nonlin- seismic ground motions and the nonlinear soil behavior are
earities in the soil material and due to uplifting are relatively properly accounted for. It was assumed that Uquefaction will
unimportant as far as displacement response is concerned. The occur when the excess pore pressure becomes equal to the initial
response of reinforced concrete and steel frame buildings of effective vertical stress, ie, when the sand stiffness has deterio-
various heights designed according to the Uniform Building rated to zero. The excess pore pressure rise is governed by a
Code Specification (1979) were investigated. The interaction differential equation involving the hysteretic energy dissipated
and amplitude of the shear stress earthquakes in Nigata in 1964, 1902, and 1887 and in
Hachinohe in 1968. The results are shown in Fig. 20, in which
\-r„X(t)iT(t) Q,r = coefficient of variation of shear stress ratio indicating the
(22)
dt AH (l)cos( 7TrJ2)sm2e~ l(
/
irrjl) variability of the capacities of the sand deposit against liquefac-
tion. It is seen that the probabilities predicted with the analyti-
where ru = the excess pore pressure ratio, e7-(r) = the shear- cal method are consistent with the observed results.
strain energy dissipated given by Eq (7), X(t) = an equivalent
weighing function depending primarily on the stress amplitude,
and AlF(l) is the soil deposit energy dissipation capacity. The 5, SUMMARY AND AREAS FOR FURTHER RESEARCH
degradation of the soil system is also a function of /;,.
This review highlights the major developments in the area of
The smooth hysteresis model was shown to give a secant random vibration analyses for inelastic systems and their appli-
shear modulus and equivalent viscous damping ratios as func- cations to real structural systems. The methodology is seen to
tions of shear-strain which agree well with test results. So was have reached a certain degree of maturity that it is capable of
the resultant excess pore pressure rise as a function of cycles of realistically modeling structural hysteresis, degradation, and
loading (Pires et al, 1983). biaxial interaction and giving response statistics that can be
Under random earthquake excitation, the method of equiva- used in assessing structural performance and damage. Compar-
lent linearization gives the mean and standard deviation of the isons with field evidence generally indicate that such a random
energy dissipation. These statistics can be used to calculate vibration method is a viable and useful tool in the study of
reliability against liquefaction. For example, for an earthquake structural reliability under dynamic hazards. The limitations
with given intensity and duration, failure occurs when are; (1) except for those based on the equivalent linearization
AW(l) -AH/<0, (23) technique, most solution procedures are restricted by the com-
T 7
putational effort requirement that they are suitable only for the
where &.W = JOX(T)ZT( ) dr. The statistics of AW are obtained analysis of single-degree-of-freedom systems; and (2) the lin-
from the random vibration analysis, whereas those of Ajl^l) earized systems are deficient in terms of preserving the drift due
can be determined from data on tests carried out under uniform to the high frequency component of the excitation. Therefore,
cyclic stress. The probability of liquefaction can then be calcu- there is an urgent need for further developments alleviating
lated. these limitations. Also, most research is primarily concerned
The probabilities of liquefaction were predicted with the with the development of methodology with little interest in the
proposed method and compared with field evidence for sand demonstration of the practicality of the methods developed.
deposits during several past earthquakes in Japan, namely, More effort in this area is needed as the ultimate goal and the
justification of any engineering investigation is the eventual
application to practical engineering problems.,Lastly, although
efforts have already started on continuous systems, develop-
ment is still in the preliminary stage and much more needs to be
done, in particular, in the area of developing realistic and yet
tractable constitutive equations and efficient discretization
schemes and computational methods.
ACKNOWLEDGMENT
This study is supported by the National Science Foundation
under Grant CES-861917 monitored by Dr S C Liu. This
support is gratefully acknowledged. Comments of the reviewer
are sincerely appreciated.
REFERENCES
5 10 15 20 25 30
.STANDARD PENETRATION RESISTANCE CORRECTED TO 1. Baber, T T, and Noori, M N, Random vibration of pinching, hysteretic
AN EFECTIVE VERTICAL PRESSURE OF 1 tonfsq f t , N, systems, J Eng Mech Div ASCE 110(7), 1984, 1036-1049.
2. Baber, T T, and Wen, Y K, Random vibration of hysteretic degrading
systems, / Eng Mech Div ASCE 107(EM6), 1981, 1069-1087.
FIG. 20. Predicted liquefaction probabilities and comparison with 3. Baber, T T, and Wen, Y K, Stochastic response of multistory yielding
historical d a t a (Ref 33): ( • ) liquefaction (good acceleration data); ( 0 ) frames, Earthquake Eng Struct Dyn 10, 1982, 403-416.
no liquefaction (good acceleration data); ((D) no-liquefaction (estimated 4. Bartels, R H, and Stewart, G W, Solution of the matrix equation
accelerations). AX + XB = C, Commun ACM 15(9), 1972, algorithm 432.
5. Bergman, L A, and Spencer, B F, On the solution of several first passage
Case Pr obabiliti es of iquefaction (v / ° v 0 ) boundary problems in nonlinear stochastic dynamics, Proceedings, WTAM sym-
Qf = 0.30 posium on nonlinear stochastic dynamic engineering systems, Innsbruck,
history 0.3" <J2 f < 0.40 K v c / ° V o ) point Austria, Jun 1987, pp 479-493.
6. Bouc, R, Forced vibration of mechanical systems with hysteresis, ab-
1 0.611 0.64 0.5 stract, Proceedings, 4th conference on nonlinear oscillation, Prague,
2 0.34 0.32 0.9 Czechoslovakia, 1967.
3 0.42 0.41 0.6 7. Cai, G Q, and Lin, Y K, A new solution technique for randomly excited
10 0.14 0.09 1.3 hysteretic structures," Report CAS 88-4, Florida Atlantic University,
11 0.0062 0.0023 2.0 Boca Raton FL, May 1988.
5 0.33 0.30 1.0 8. Casciati, F, Nonlinear stochastic dynamics of large structural systems by
0.94 equivalent linearization, Proceedings, 5th international conference on ap-
6 0.92 0.3
plication of statistics and probability in soil and structural engineering
9B 0.26 0.23 1.2 ICASP 5, Vancouver, BC, Canada, May 1987, vol II, pp 1165-1172.
9T 0.11 0.08 1.5 9. Casciati, F, and Faravelli, L, Stochastic equivalent linearization in 3-D
: T
/< hysteretic frames, Proc 9th SMiRT M20-4, 1987.
10. Caughey, T K, Random excitation of a system with bilinear hysteresis, Structural Dynamics, 14, 1986, 543-557.
J Appl Mech, Trans ASME 27 (Dec) 1960, pp 649-652. 32. Peng, C Y, and Iwan W D, Identification of hysteretic structural behav-
11. Chu, C T, Wen, Y K, and Ang, A H-S, Nonlinear random vibration of ior froill strong, Inotion accelerog,rarns, Proceedings, structural safety
building foundation system, Civil Engineering Studies, Structural Re- evaluation based 011 system identification approaches. Lambrecht. CJcr-
search Series, University of Illinois at Urbana-Champaign, Urbana IL, many, Jul 1987, Friedr. Vieweg & Sohn., BraunchweigjWiesbaden
1984. F.R.G. pp 103-117.
12. Clough, R W, and Johnston, S B, Effect of stiffness degradation on 33. Pires, J E A, Wen, Y K, and Ang, A H-S, Stochastic analysis of
earthquake ductility requirements, Proceedings, 2nd Japan earthqaake liquefaction under earthquakc loading. Civil engineering studies. SRS no
engineering symposium, Tokyo, Japan, Oct 1966, pp 227-232. 504, University of Illinois at Urbana-Champaign, Urbana IL, Apr 1983.
13. Grossmayer, R L, Approximate stochastic analysis of elasto-plastic sys- 34. Powell, G H, and Chen, P F-S, 3-D beam-column element with general-
tems, J Eng Mech Diu ASCE 107(EM1), 1981, 97-116. ized plastic hinges, J Eng Mech, ASCE 112(76), 1986 pp 627-641.
14. Hoshiya, M, and Maruyama, 0, Identification of nonlinear structural 35. Roberts, J B, Application of averaging methods of randomly excited
systems, Proceedings, ICASP 5, Vancouver BC, Canada, 1987, vol I, pp hysteretic syslen1s, Proceedings, IUTA Nt jJ'nlfJosium 011 nonlinear
182-189. stochastic ((Vllwnic engiHeering systems, Innsbruck. Austria, Jun 1987, pp
15. Iwan, W D, A distributed-element model for hysteresis and its steady- 361-380.
state dynamic response, J Appl Mech 33(4), 1966, 893-900. 36. Spanos, P D, Stochastic linearization in structural dynamics, Appl Mech
16. Iwan, W D, A generalization of the concept of equivalent linearization, Rev 34(1), 1981.
Int J Nonlinear Mech 8, 1973, 279-287. 37. Spanos, P D, and Iwan, W D, On the existence and uniqueness of
17. Iwan, W D, Application of nonlinear analysis techniques, Proceedings, solution generated by equivalent linearization, 1111 J Nonlinear Mech 13,
applied mechanics in earthquake engineering, Winter Annual Meeting of 1978,71- 78.
ASME, New York, Nov 1974, pp 135-161. 38. Spencer, B F, and Bergman, L A, On the reliability of a simple hysteretic
18. Iwan, W D, and Lutes, L D, Response of the bilinear hysteretic system system, J Eng Mech Div ASCE III, 1985, 1502-1514.
to stationary random excitation, J Acoust Soc Am 43(3), 1968, 545-552. 39. Sues, R H, Wen, Y K, and Ang, A I-I-S, Stochastic evaluation of seismic
19. Iwan, W D, Moser, M A, and Pararizos, L G, The stochastic response of structural performance, J Struct Eng ASCE III(6), 1985, pp 1204-1218.
strongly nonlinear systems with Coulomb damping elements, Proceed- 40. Suzuki, Y, and Minai, R, Application of stochastic dilfercntial equation
ings IUTAM synlposium on nonlinear stochastic dynamic engineering j}'S- to seismic feliabili ty analysis of hysteretic structures, Proceedings,
tems, Innsbruck, Austria, Jun 1987, pp 455-466. US-Japall joint seminar 011 stochastic approaches in earthquake engineer-
20. Jayakumar, P, and Beck, J L, System identification using nonlinear ing, Boca Raton FL, May 1987, pp 334-356.
stnlctural rnodels, Proceedings, structural safety evaluatioll based on .~ys 41. Takeda, T, Sozen, M A, and Nielson, M N, Reinforced concrete re-
tem identification approaches, Lambrecht, Germany, 1987, pp 82-102. sponse to simulated earthquakes, J Struct Diu A SCE 96, 1976,
21. Jennings, P C, Periodic response of a general yielding structure, J Eng 2557-2573.
Mech Diu ASCE 90 (Apr), 1964, 131-165. 42. Takizawa, H, and Aoyama, H, Biaxial efrect in modclling earthquake
22. Karnopp, D, and Scharton, T D, Plastic deformation ion random vibra- response of RjC structures, J Earthquake Eng Struct Dyn 4, 1976,
tion, J Acoust Soc Am 39, 1966, 1154-1161. 523-552.
23. Kobori, T, Reliability-based anti-seismic analysis of structures, Proceed- 43. Uniform Building Code, Whittier, CA, International Conference of
ings, ICOSSAR '85, Kobe, Japan, May 1985, vol Ill, pp 1-22. Building, Oflicials, 1979.
24. Kobori, T, Minai, R, and Suzuki, Y, Stochastic seismic response of 44. Vanmarcke, E H, Structural response to earthquakes, Seismic risk and
hysteretic structures, Bull Disaster Prevention Res Inst 26 (part 1), 1976, engineering decision. Lonmitz and Rosenblueth, Eds, Elsevier. New York,
pp 55-70. 1976.
25. Kwok, Y H, Ang, A-H S, and Wen, Y K, Seismic damage analysis and 45. Wen, Y K, Method for random vibration of hysteretic systems, J Eng
design of unreinforced masonry buildings, Civil engineering studies, SRS Mech Div ASCE I02(EM2), 1976, 249-263.
no 536, University of Illinois, Jun 1987. 46. Wen, Y K, Equivalent linearization for hysteretic systems under random
26. Minai, R, and Suzuki, Y, Stochastic estimates of nonlinear dynamic excitations, J Appl Mech ASME 47(1),1980,150-154.
systeIl1S, Proceedings, US-Japanjoinl seminar Oil stochastic approaches in 47. Wen, Y K, On response of smooth hysteretic systems to low frequency
earthquake engineering, Boca Raton, FL, May 1987, pp 204-230. cxcitation, J Prohabilistic Eng Mech, 1988 submitted.
27. Mochio, T, and Shinozuka, M, Stochastic equivalent linearization in a 48. Wen, Y K, and Ang, A H-S, Inelastic modeling and system identifica-
finite element-based reliability analysis, Proceedings, 4th international tion, Proceedings, structural safety evaluation hosed 011 .\},·stem identifica-
conference on structural safety and reliability, Kobe, Japan, May 1985. tion approaches, International Workshop, Lambrecht, Germany, Ju11987,
28. Park, Y J, Ang, A H-S, and Wen, Y K, Seismic damage analysis of pp 142-160.
reinforced concrete buildings, J Struct Eng ASCE III(4), 1985, 740-757. 49. Wen, Y K, and Yeh, C-H, Bi-Axial and torsional response of inelastic
29. Park, Y J, Ang, A H-S and Wen, Y K, Damage-limiting asiesmic design structures ':Jnder random excitation, Proceedings, .\ymposium on stochas-
of buildings, Earthquake Spectra 3(1), Feb 1987, 1-26. tic struclUral dynamics, University of Illinois, Urbana IL, Oct 1988.
30. Park, Y J, Stochastic formulation of plasticity theory, private communi- 50. Zhu, W-Q, and Lei, Y, Stochastic averaging of energy envelope of
cation, Jun 1987. bilinear hysteretic systems, Proceedings, IUTAM symposium of nonlinear
31. Park, Y J, Wen Y K and Ang A H-S, Two-Dimensioal Random Vibra- stochastic dynamic engineering systems, Innsbruck, Austria, Jun 1987, pp
tion of Hysteretic Structures, Journal of Earthquake Engineering and 381-392.