Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Comput. Methods Appl. Mech. Engrg.

193 (2004) 59–68


www.elsevier.com/locate/cma

Sensitivity evaluation in seismic reliability analysis


of structures
Abhijit Chaudhuri 1, Subrata Chakraborty *

Department of Civil Engineering, Bengal Engineering College (A Deemed University), Howrah 711103, India
Received 3 March 2003; received in revised form 15 August 2003; accepted 2 September 2003

Abstract

Sensitivity evaluation of response under static and dynamic load is proved to be an essential part of the optimization
and reliability analysis of structure. Most of these works concentrate on sensitivity analysis of static and dynamic
structural responses under deterministic forcing function. The present paper deals with the important issue of response
sensitivity evaluation of structures in seismic reliability evaluation. The formulation has been developed in double
frequency domain to tackle non-stationary earthquake motion for obtaining the analytical sensitivity statistics of
various dynamic response quantities with respect to structural parameters. A multistoried building frame has been
studied to elucidate the proposed algorithm.
 2003 Elsevier B.V. All rights reserved.

Keywords: Non-stationary earthquake; Double Fourier analysis; Dynamic response and reliability; Sensitivity

1. Introduction

Sensitivity is one of the ways to evaluate the performance of structures when the design variables are
under the influences of parametric changes. The sensitivities of structural responses represent an essential
ingredient for gradient-based optimization methods needed in structural reliability analysis, optimization,
identification etc. As structural response measure, the displacements, stresses, buckling loads or natural
frequencies are considered, whereas the design variables are usually the material properties (i.e. YoungÕs
modulus and density etc.), size and shape of the structure. In order to evaluate the reliability of structures, a
key ingredient is the response gradients or sensitivities with respect to loading and system parameters.
The analysis and design sensitivity computations are very simple for static problems and from early
seventies, analytical formulations for design sensitivity analysis are being reported in the literature [1–5].
Though, the sensitivity analysis of structure under transient dynamic loads [6–9] are studied but restricted

*
Corresponding author. Present address: Department of Engineering, University of Cambridge, Trumpington Street, Cambridge
CB1 2PZ, UK. Tel.: +91-033-2668-4561/62/63; fax: +91-033-2668-4564/668-2916.
E-mail address: schak@civil.becs.ac.in (S. Chakraborty).
1
Presently with Department of Civil Engineering, IISc Bangalore, India.

0045-7825/$ - see front matter  2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.cma.2003.09.007
60 A. Chaudhuri, S. Chakraborty / Comput. Methods Appl. Mech. Engrg. 193 (2004) 59–68

to deterministic dynamic load. However, not much research [10,11] seems to have been carried out for
computation of design derivative under stochastic dynamic environment. But, in reality major dynamic
loads on structures are random in nature i.e. earthquake, wind, wave etc. When a structure is subjected to
such uncertain loading the responses are also stochastic in nature, which are well documented in the field of
random vibration theory [12–15].
The objective of this paper is to present an algorithm for evaluating the analytical sensitivity of dynamic
responses of structures subjected to stochastic earthquake load. In doing so, the random ground motion is
modelled as a fully non-stationary sigma oscillatory process. The proposed formulation computes the
analytical sensitivity gradient of reliability with respect to the structural design parameters. A multi-
storeyed building frame idealized as a space frame in finite element modelling is considered to study the
sensitivity of dynamic responses. The depths of different beams or columns of the space frame are con-
sidered as design variables.

2. Ground motion model

The ground motion due to earthquake is characterized by a sudden rise and slow decay i.e. mean and
auto-correlation function or power spectral density function (PSDF) are not time invariant. Though the
motion becomes stationary after some time and can be suitably modelled as such. But, sudden rise of
ground acceleration may become critical for satisfactory performance and safety of various structures. To
incorporate this non-stationary character of ground motion, it is described as a sigma-oscillatory process as
suggested by Conte and Peng [16]. In this model the oscillatory processes are pair wise independent and
expressed as,
Xp Xp
ug ðtÞ ¼
€ Xk ðtÞ ¼ Ak ðtÞSk ðtÞ; ð1Þ
k¼1 k¼1

where, p is the number of component process; Sk ðtÞ is kth Gaussian stationary process and Ak ðtÞ is the time
modulating function of kth sub-process Xk ðtÞ defined as
b
Ak ðtÞ ¼ ak ðt  nk Þ k eck ðtnk Þ H ðt  nk Þ; ð2Þ
where ak and ck are positive constant; bk is a positive integer; nk is the ‘‘arrival time’’ of the kth sub-process
Xk ðtÞ, H ðtÞ is Heaviside unit-step function. The kth zero-mean, stationary Gaussian process Sk ðtÞ is char-
acterized by its autocorrelation function RSk Sk ðsÞ ¼ emk jsj cosðgk sÞ and the associated PSD function
" #
mk 1 1
/Sk Sk ðxÞ ¼ þ ; ð3Þ
2p m2k þ ðx þ gk Þ2 m2k þ ðx  gk Þ2
where, mk and gk are two free parameters representing the frequency bandwidth and the predominant
frequency of the process Sk ðtÞ. The autocorrelation function of ground motion U € g ðtÞ can be expressed as
X
p X
p
R€ug €ug ðt1 ; t2 Þ ¼ R€ug €ug ðt1 ; sÞ ¼ Ak ðt1 ÞAk ðt2 ÞRSk Sk ðsÞ ¼ Ak ðt1 ÞAk ðt2 ÞRSk Sk ðt2  t1 Þ
k¼1 k¼1
X
p
¼ Ak ðt1 ÞAk ðt2 Þ emk jt2 t1 j cosðgk ðt2  t1 ÞÞ: ð4Þ
k¼1

And the evolutionary PSD function will be


X
p
2
/€ug €ug ðt1 ; xÞ ¼ jAk ðt1 Þj /Sk Sk ðxÞ: ð5Þ
k¼1
A. Chaudhuri, S. Chakraborty / Comput. Methods Appl. Mech. Engrg. 193 (2004) 59–68 61

The generalized non-stationary cross-spectral density matrix for ground motion can be represented in
double frequency domain as a function of two frequencies [17] i.e.
Z 1 Z 1
1
S€ug €ug ðx1 ; x2 Þ ¼ 2 R€u €u ðt1 ; t2 Þ eiðx1 t1 x2 t2 Þ dt1 dt2
4p 1 1 g g
Z 1 Z 1 X !
p
1
¼ 2 Ak ðt1 ÞRSk Sk ðt1  t2 ÞAk ðt2 Þ eiðx1 t1 x2 t2 Þ dt1 dt2
4p 1 1 k¼1
p Z Z Z 1 
1 X 1 1
¼ 2 Ak ðt1 Þ /Sk Sk ðxÞ eiðxt1 xt2 Þ dx Ak ðt2 Þ eiðx1 t1 x2 t2 Þ dt1 dt2
4p k¼1 1 1 1

Xp Z 1  Z 1   Z 1 
1 iðxx1 Þt1 1 iðxx2 Þt2
¼ Ak ðt1 Þ e dt1 /Sk Sk ðxÞ Ak ðt2 Þ e dt2 dx
k¼1 1 2p 1 2p 1
Xp Z 1
¼ Mk ðx  x1 Þ/Sk Sk ðxÞMk ðx  x2 Þ dx: ð6Þ
k¼1 1

In above,
Z 1
1 ak eixnk
Mk ðxÞ ¼ Ak ðtÞ eixt dt ¼ ðbk þ1Þ
Cðbk þ 1Þ: ð7Þ
2p 1 ðck  ixÞ

3. Stochastic dynamic analysis

3.1. Equilibrium equation in frequency domain

The dynamic equilibrium equation for a multi-degree of freedom system subjected to ground excitation
can be written as
½Mf€
uðtÞg þ ½Cfu_ ðtÞg þ ½KfuðtÞg ¼ ½M½Lf€ug ðtÞg; ð8Þ

where ½M, ½C and ½K are the global mass, damping and stiffness matrix, respectively. fuðtÞg is the dis-
placement due to ground motion f€ ug ðtÞg and jth column of the influence coefficient matrix ½L represents the
pseudo-elastic response in all degrees of freedom due to unit translation of support motion at the ÔjthÕ
degree of freedom.
The displacement of the linear finite element model subjected to unit amplitude ground motion €ug ðtÞ ¼
eixt can be assumed as fuðtÞg ¼ fHu ðxÞ  eixt g where fHu ðxÞg is the complex frequency response function.
Thus, the equation of motion transform to
€ g ðxÞ i:e: ½DðxÞfHu ðxÞg ¼ fF ðxÞg;
ð½K  x2 ½M þ ix½CÞfHu ðxÞg ¼ ½MfLgU ð9Þ

where, ½DðxÞ is the dynamic stiffness matrix and fF ðxÞg is the forcing vector. All these matrix and vectors
are the functions of either of any design variable or in combination, which may be YoungÕs modulus ðEÞ
and/or PoissonÕs ratio, mass density ðmÞ, cross-sectional area, boundary conditions etc. In present work,
depths fdg of beams and sides of columns are considered as design variables. Thus Eq. (9) can be explicitly
re-written as,
1
½Dðx; fdgÞfHu ðx; fdgÞg ¼ fF ðx; fdgÞg or fHu ðx; fdgÞg ¼ ½Dðx; fdgÞ fF ðx; fdgÞg: ð10Þ
62 A. Chaudhuri, S. Chakraborty / Comput. Methods Appl. Mech. Engrg. 193 (2004) 59–68

3.2. Power spectral density (PSD) function of the responses and spectral moments

When the forcing function on the right-hand side of Eq. (8) are non-stationary, the cross correlation
matrix of responses can be expressed as [17]
Z 1 Z 1
T
½Ruu ðt1 ; t2 ; fdgÞ ¼ fHu ðx1 ; fdgÞgS€ug €ug ðx1 ; x2 ÞfHu ðx2 ; fdgÞg eiðx1 t1 x2 t2 Þ dx1 dx2 : ð11Þ
1 1

Thus, for a linear system with known frequency response function, the PSD function ½Suu ðx1 ; x2 ; fdgÞ of
any response variable fuðtÞg for known PSD function of ground motion can be readily obtained as
T
½Suu ðx1 ; x2 ; fdgÞ ¼ fHu ðx1 ; fdgÞgS€ug €ug ðx1 ; x2 ÞfHu ðx2 ; fdgÞg : ð12Þ
The PSD function for ground motion are described elaborately in Section 2. From that PSD function of
displacement, the PSD function of any stress or combination of stresses can be evaluated simply utilizing
the constitutive relationship i.e.
ff ðtÞgi ¼ ½ki ½T i fuðtÞgi ; ð13Þ
where, ½ki , ½T i and fuðtÞgi are stiffness, transformation matrix and displacement vector respectively for ith
element. In frequency domain the above equation will be
fF ðxÞgi ¼ ½ki ½T i fU ðxÞgi ; ð14Þ
where, fF ðxÞgi , fU ðxÞgi are the Fourier transform of ff ðtÞgi , fuðtÞgi respectively. Thus, the PSD function
for force can be readily obtained as,
T T
½Sff ðx1 ; x2 ; fdgÞi ¼ ½ki ½T i ½Suu ðx1 ; x2 ; fdgÞi ½T i ½ki ; ð15Þ
where, ½Suu ðx1 ; x2 ; fdgÞi is the PSD function of displacement vector of ith element, obtained from the
associated PSD function of global displacement vector.
If xðtÞ is any response of the structure, the autocorrelation function of xðtÞ is
Z 1 Z 1 Z 1
Rxx ðt1 ; t2 ; fdgÞ ¼ Sxx ðx1 ; x2 ; fdgÞ eiðx1 t1 x2 t2 Þ dx1 dx2 ¼ /xx ðt1 ; x; fdgÞ eixðt2 t1 Þ dx; ð16Þ
1 1 1

where /xx ðt1 ; x; fdgÞ represents the evolutionary PSD function of the non-stationary process xðtÞ. The time
varying spectral moments which are used in reliability evaluation are obtained from the evolutionary PSD
function as described by Eq. (17),
Z 1 Z 1 Z 1
j j
kj ðt1 ; fdgÞ ¼ jxj /xx ðt1 ; x; fdgÞ dx ¼ jxj Sxx ðx1 ; x2 ; fdgÞ eiðx1 t1 x2 t1 Þ dx1 dx2 ; ð17Þ
1 1 1

where ÔjÕ is the order of the spectral moment. The RMS value of the response and its derivative are
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rx ðt; fdgÞ ¼ k0 ðt; fdgÞ and rx_ ðt; fdgÞ ¼ k2 ðt; fdgÞ: ð18Þ

3.3. Reliability of the structure

Assuming linear structural behaviour and Gaussian input ground motion, the output responses will be
also Gaussian and the reliability with respect to the response xðtÞ is [14]
Rt
 aðx;s;fdgÞ ds
Lðx; t; fdgÞ ¼ Ae 0 ; ð19Þ
where A is the probability that the process is below the barrier level at time t ¼ 0. For non-stationary or
zero start random process A ¼ 1. x is the barrier level and aðx; t; fdgÞ is the so called hazard function i.e.
A. Chaudhuri, S. Chakraborty / Comput. Methods Appl. Mech. Engrg. 193 (2004) 59–68 63

crossing rate of D-type barrier level xðtÞ ¼ x at time ÔtÕ. The better approximation of the phenomenon of
barrier crossings i.e. the crossings obey a two-state Markov process [18] is
0 !1
rffiffiffi
p 
x qðt; fdgÞ
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiB
B 1  exp  2 r ðt; fdgÞ C
C
1 k2 ðt; fdgÞB B x C
aðx; t; fdgÞ ¼ 0 ! 1 C: ð20Þ
p k0 ðt; fdgÞB B 2 C
1 
x C
@ exp@ A  1A
2 rx ðt; fdgÞ

In which qðt; fdgÞ is the shape parameter of the spectral density function of the process, a measure of the
narrowness of the PSD function defined as,
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
k21 ðt; fdgÞ
qðt; fdgÞ ¼ 1  : ð21Þ
k0 ðt; fdgÞk2 ðt; fdgÞ

4. Sensitivity of random responses

4.1. Sensitivity of FRF of response

If the sensitivity computation is performed with respect to depth of kth member as the only design
variable, differentiation of Eq. (10) with respect to depth of kth member dk gives,
 
o 1 o o
fH ðx; fdgÞg ¼ ½Dðx; fdgÞ fF ðx; fdgÞg  fH ðx; fdgÞg ½Dðx; fdgÞ : ð22Þ
odk odk odk
o o
The depths or side of different members being independent, the derivatives ½kl ðfdgÞ and ½ml ðfdgÞ
o odk odk
vanish for k 6¼ l. Thus ½Dðx; fdgÞ becomes a function of dk only.
odk

4.2. Sensitivity of PSD function of response and spectral moments

The sensitivity of PSD function of displacements at nodes with respect to depth Ôdk Õ is obtained directly
from Eq. (12) i.e.

o o
½Suu ðx1 ; x2 ; fdgÞ ¼ fH ðx1 ; fdgÞgS€ug €ug ðx1 ; x2 ÞfH ðx2 ; fdgÞgT
odk odk

o T
þ fH ðx1 ; fdgÞgS€ug €ug ðx1 ; x2 Þ fH ðx2 ; fdgÞg : ð23Þ
odk
The sensitivity of the spectral moments is evaluated from the sensitivity of corresponding PSD function i.e.
Z 1 Z 1
o j o
ðkj ðt1 ; fdgÞÞ ¼ jxj ðSxx ðx1 ; x2 ; fdgÞÞeiðx1 t1 x2 t1 Þ dx1 dx2 : ð24Þ
odk 1 1 odk
The sensitivity of the RMS value of the response and its derivative are obtained as
o o
ðk ðt ; fdgÞÞ ðk ðt ; fdgÞÞ
o 1 odk 0 1 o 1 odk 2 1
ðrx ðt; fdgÞÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi and ðrx_ ðt; fdgÞÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : ð25Þ
odk 2 ðk0 ðt1 ; fdgÞÞ odk 2 ðk2 ðt1 ; fdgÞÞ
64 A. Chaudhuri, S. Chakraborty / Comput. Methods Appl. Mech. Engrg. 193 (2004) 59–68

4.3. Sensitivity of reliability of response

Finally, the sensitivity of reliability with respect to variable dk can be easily obtained by differentiating
Eq. (19) with respect to dk i.e.
 Z t  Z t 
oLðx; t; fdgÞ o
¼ exp  aðx; s; fdgÞ ds aðx; s; fdgÞ ds
odk 0 odk 0
 Z t  Z t !
oaðx; s; fdgÞ
¼  exp  aðx; s; fdgÞ ds ds ; ð26Þ
0 0 odk

where the sensitivity of expected crossing rate with respect to dk is obtained as follows:
 
oaðx; t;fdgÞ orx_ ðt; fdgÞ 1 orx ðt; fdgÞ 1
¼  aðx; t; fdgÞ
odk odk rx_ ðt; fdgÞ odk rx ðt; fdgÞ
rffiffiffi( )
p oðqðt; fdgÞÞ xðfdgÞ qðt; fdgÞ oðxðfdgÞÞ qðt; fdgÞxðfdgÞ oðrx ðt;fdgÞÞ
þ 
2 odk rx ðt; fdgÞ rx ðt; fdgÞ odk ðrx ðt; fdgÞÞ2 odk
þ rffiffiffi ! aðx;t; fdgÞ
p qðt; fdgxðfdgÞÞ
exp  1
2 rx ðt; fdgÞ
( )
oðxðfdgÞÞ xðfdgÞ oðrx ðt; fdgÞÞ ðxðfdgÞÞ2

odk ðrx ðt;fdgÞÞ2 odk ðrx ðt; fdgÞÞ3
 0 !2 1 aðx; t; fdgÞ; ð27Þ
1 
xðfdgÞ
1  exp@  A
2 rx ðt; fdgÞ

where

oðqðt; fdgÞÞ 1 ðk1 ðt; fdgÞÞ2


¼
odk qðt; fdgÞ k0 ðt; fdgÞk2 ðt; fdgÞ
 
oðk0 ðt; fdgÞÞ 1 oðk2 ðt; fdgÞÞ 1 oðk1 ðt; fdgÞÞ 1
 þ  : ð28Þ
odk k0 ðt; fdgÞ odk k2 ðt; fdgÞ odk k1 ðt; fdgÞ

5. Numerical example

The horizontal ground motion along 45 with X -axis is described by the PSD function, as shown in Fig.
1 which corresponds to the El Centro 1940 earthquake, the detail parameters to describe the one sided PSD
function of the same may be found in [16]. A two storied unsymmetrical RC building idealized as a space
frame (shown in Fig. 2) subjected to above ground motion is taken to elucidate the proposed formulation.
The sizes of the columns and beams are 0.35 m · 0.35 m and 0.25 m · 0.40 m respectively. The density of the
beam material is assumed to be ten times that of columns to account for typical additional masses e.g. slabs,
cross beams, floors, ceilings etc. Density and modulus of elasticity of concrete members are taken as 2400
kg/m3 and 2 · 107 kN/m2 , respectively.
The reliability and associated sensitivity based on maximum top displacement (i.e. barrier level of 15 mm)
of the frame along X -axis are shown in Fig. 3a and b and that along Y -axis are shown in Fig. 4a and b. The
reliability and its sensitivity based on maximum base shear criteria (i.e. barrier level of 500 kN/m2 ) are
A. Chaudhuri, S. Chakraborty / Comput. Methods Appl. Mech. Engrg. 193 (2004) 59–68 65

Fig. 1. Generalized non-stationary power spectral density correspond to 1940 El Centro Earthquake.

Fig. 2. A two storied unsymmetrical building frame.

presented in Figs. 5a,b, and 6a,b with respect to X - and Y -axis, respectively. It can be observed from the
figures that when the depth of beam at top is increased the stiffness of the structure to lateral movement is
66 A. Chaudhuri, S. Chakraborty / Comput. Methods Appl. Mech. Engrg. 193 (2004) 59–68

Fig. 3. (a) Reliability of structure based on displacement along X -axis at node 16. (b) Sensitivity of reliability of structure based on
displacement along X -axis at node 16. (––) Side of column (1), (- - -) side of column (2), (– –) depth of beam (7), (– - –) depth of beam
(12), ( ) depth of beam (21).

Fig. 4. (a) Reliability of structure based on displacement along Y -axis at node 16. (b) Sensitivity of reliability of structure based on
displacement along Y -axis at node 16. (––) Side of column (1), (- - -) side of column (2), (– –) depth of beam (7), (– - –) depth of beam
(12), ( ) depth of beam (21).

slightly increased and mass of structure is affected with large scale and consequently structural responses
become high leading to lesser safety margin of the structure. Thus the sensitivity of reliability becomes
negative as revealed in figures. It is also seen that larger size of columns make the structure more safe with
respect to deflection at top as well as shear force at column base. It is obvious that the overall stiffness of
structure is increased with the increment of columns dimension and response is mainly controlled by
the stiffness of structure. This also leads to better safety margin for the structure i.e. sensitivity becomes
positive.
A. Chaudhuri, S. Chakraborty / Comput. Methods Appl. Mech. Engrg. 193 (2004) 59–68 67

Fig. 5. (a) Reliability of structure based on shear force along X -axis at the base of column 2. (b) Sensitivity of reliability of structure
based on shear force along X -axis at base of column 2. (––) Side of column (1), (- - -) side of column (2), (– –) depth of beam (7), (– - –)
depth of beam (12), ( ) depth of beam (21).

Fig. 6. (a) Reliability of structure based on shear force along Y -axis at the base of column 2. (b) Sensitivity of reliability of structure
based on shear force along Y -axis at base of column 2. (––) Side of column (1), (- - -) side of column (2), (– –) depth of beam (7), (– - –)
depth of beam (12), ( ) depth of beam (21).

6. Conclusion

An evaluation procedure to compute the gradient i.e. analytical sensitivities in seismic reliability analysis
with respect to the structural parameters was developed in double frequency domain. The method consists
of differentiating exactly the numerical finite element algorithm with respect to desired design variables. The
procedure can be carried out for several design parameters and extension to uncertain design parameters is
straightforward utilizing proper probabilistic description of random design parameter. The analysis in
double frequency domain is developed as it takes care of the transient nature of response under non-sta-
tionary input. When the non-stationary input is described as a function of time ÔtÕ and frequency ÔxÕ the
68 A. Chaudhuri, S. Chakraborty / Comput. Methods Appl. Mech. Engrg. 193 (2004) 59–68

transient nature of response can only be included through convolution integral. This integration can be
avoided if the analysis is done in double frequency domain. The first passage failure probability or reli-
ability is determined using the expression given by Vanmarcke based on assumption that the crossings by
the random process is a two state Markov process. In general, for narrow band process crossings may occur
in clump i.e. crossing does not behave always as a Poisson process. It is found that sensitivities of reliability
with respect to some design variables (i.e. member properties) are negative and it indicates the adverse effect
of increment of size while positive value of sensitivity indicates that increment of size of those members will
improve the safety.

Acknowledgements

The financial support receive in OYS Scheme No. SR/FTP/ET-40/2000 dated 3.10.2000 from the
Department of Science and Technology, Govt. of India is gratefully acknowledged.

References

[1] O.C. Zienkiewcz, J.S. Campbell, Shape optimization and sequential linear programming, in: R.H. Gallaghar, O.C. Zienkiewcz
(Eds.), Optimum Structural Design, Theory and Application, John Wiley, London, 1977, pp. 109–126.
[2] S.Y. Wang, Y. Sun, R.H. Gallaghar, Sensitivity analysis in shape optimization of continuum structures, Comp. Struct. 20 (1985)
855–867.
[3] T. Haftka, H.M. Adelman, Recent development in structural sensitivity analysis, NASA Technical Memorandum, 100668, 1988.
[4] M. Kleiber, T.D. Hien, The Stochastic Finite Element Method, John Wiley & Sons, 1992.
[5] M. Klieber, H. Antunez, T.D. Hien, P. Kowalczyk, Parameter Sensitivity in Nonlinear Mechanics, John Wiley & Sons, 1997.
[6] C.C. Hsieh, J.S. Arora, Design sensitivity analysis and optimization of dynamic response, Comput. Methods Appl. Mech. Engrg.
43 (1984) 195–219.
[7] C.C. Hsieh, J.S. Arora, An efficient method for dynamic response optimization, AIAA 23 (9) (1985) 1454–1456.
[8] A. Dutta, C.V. Ramakrishnan, Accurate computation of design sensitivities for structures in transient dynamic loads using time
marching scheme, Int. J. Numer. Methods Engrg. 41 (1998) 977–999.
[9] J.P. Conte, Finite element response sensitivity analysis in earthquake engineering, in: Spencer, Hu (Eds.), Proc. of China–US
Millennium Symposium of Earthquake Engrg., 8–11 November 2000.
[10] J.P. Conte, P K. Vijalapura, M. Meghella, Consistent finite element sensitivities in seismic reliability analysis, in: 13th ASCE
Engrg. Mech. Div. Conf., Johns Hopkins University, 13–16 June 1999.
[11] S. Benfratello, S. Caddemi, G. Muscolino, Gaussian and non-Gaussian stochastic sensitivity analysis of discrete structural system,
Comp. Struct. 78 (2000) 425–434.
[12] N.C. Nigam, S. Narayanan, Application of Random Vibration, Narosa Publishing House, 1994.
[13] Y.K. Lin, G.O. Cai, Probabilistic Structural Dynamics: Advanced Theory and Application, Mc Graw-Hill, 1995.
[14] L.D. Lutes, S. Sarkani, Stochastic Analysis of Structural and Mechanical Vibration, Prentice-Hall Inc., 1997.
[15] E.H. Robert, C.H. Gary, Nonstationary response of structural systems, ASCE J. Engrg. Mech. 100 EM2 (1974) 415–431.
[16] J.P. Conte, B.F. Peng, Fully nonstationary analytical earthquake ground motion model, ASCE J. Engrg. Mech. 123 (1) (1997) 15–
24.
[17] R.E. Holman, G.C. Hart, Nonstationary response of structural systems, ASCE J. Engrg. Mech. 100 EM2 (1974) 415–431.
[18] R.B. Corotis, E.H. Vanmarcke, C.A. Cornell, First passage of nonstationary random process, ASCE J. Engrg. Mech. 98 EM2
(1972) 401–415.

You might also like