Академический Документы
Профессиональный Документы
Культура Документы
SUMMARY
This paper presents a method and results of seismic fragility estimation of frame structures with friction
devices and with friction devices and restrictors. The seismic intensity parameter, de"ned as the mean value
of the pseudovelocity spectrum in a speci"ed periods band, is proved to allow the use of linear regression
analysis of the response parameters of the considered non-linear structures on seismic intensity. A simpli"ed
method of fragility estimation is proposed, based on the concept of &mean seismic excitation' and linear
regression of the seismic response parameters on seismic intensity parameter. The key risk contributors for
the system with friction devices and for the system with friction devices and restrictors are identi"ed
on the basis of the fragility analysis, and recommendations for improvement of the seismic response of
the respective systems are derived. The results of the fragility study show that when the initial &bare' frame
is retro"tted by rigidly connected bracings the e!ect is much lower than in the case of connecting the
bracings by friction devices and especially by friction devices and restrictors. Copyright 2000 John
Wiley & Sons, Ltd.
KEY WORDS: seismic fragility; seismic intensity; regression analysis; friction devices; restrictors of displace-
ments; seismic retro"tting
1. INTRODUCTION
* Correspondence to: Silvia L. Dimova, Bulgarian Academy of Sciences, Central Laboratory for Seismic Mechanics and
Earthquake Engineering, P.O. Box 27, So"a 1582, Bulgaria
R E-mail: silvia@bas.bg
retro"tting of buildings. As have been shown recently by Dimova [1] it is possible to improve
substantially the dynamic response of systems with friction devices by introducing restrictors of
the relative displacements of the sliding surfaces. The study [1] have shown that the system with
friction devices and restrictors exhibits equally stable response during both, the short period and
the long period seismic excitations. These positive results motivate probabilistic safety assessment
and comparison of the responses of the system with friction devices and restrictors and of system
with friction devices.
The presence of non-conservative non-linearities, such as hysteresis due to friction devices, does
not allow to apply in seismic fragility analysis the fundamental probabilistic solution approach in
which either time-domain superposition or frequency-domain superposition are often used. In
recent years, simulations are very often used for the analysis of non-linear random vibrations.
According to Yang [2], the basic concept of simulation methods consists of: generating a large
number of sample input excitations based on the speci"ed random excitation process; analysing
the non-linear response corresponding to each sample excitation; analysing the statistics and
probability characteristics of the non-linear structural response on the basis of all the sample
solutions.
During the last decade, simulation methods are developed and extensively used in the
probabilistic safety assessment of nuclear power plants [3}8] and failure probability of
structures with degrading sti!ness [9}11]. In connection with the di!erent structural control
techniques, such methods are developed for base-isolated structures with isolators
behaving inelastically under design earthquake motion, as reported by Hirata et al. [12}15].
Equivalent stochastic linearization method and Monte Carlo simulations are used by Dorka
and Pradlwarter [16] to determine the safety indices of a frame structure retro"tted by
friction devices. Safety of an earthquake retro"tting scheme for large panel buildings by
hysteretic devices is determined by Dorka et al. [17] by the use of Monte Carlo simulations. In
the two above-mentioned studies calculations are performed for the seismic excitation of "xed
intensity.
This paper addresses the seismic fragility (i.e. conditional probability of failure for a given
intensity of the seismic excitation) estimation of frame structures with two types of friction
devices. The simulation method used by Hirata et al. [14] for seismic fragility estimation of
base-isolated structure is applied and extended.
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
SIMPLIFIED SEISMIC FRAGILITY ANALYSIS 1155
Design criteria for use of friction damped bracings are given by Dowdel and Chery [26], Vulcano
and Mazza [27], Antonucci et al. [28].
The problems, associated with existing methods for treatment of friction forces are discussed by
Beucke and Kelly [29], Dimova et al. [23], Dimova [30]. Velocity model, which allows an easy
incorporation of non-uniform friction coe$cient during the sliding phase, gives possibility of
modeling the &pinching' of the hysteresis loops, as recorded during the experimental tests of
steel}steel interfaces. Numerical studies of the seismic response of a multistorey frame structure
with friction devices [31], show that considering the decrease of friction coe$cient with the
growth of the relative velocity of sliding surfaces during the sliding phase leads to signi"cant
increase of the relative displacement of the sliding surfaces (up to 21 per cent) and considerable
increase of the storey drifts (up to 11 per cent). This phenomenon should be taken into account in
designing friction devices and the decrease of the friction coe$cient should be considered when
studying the dynamic behaviour of structures with friction devices.
(a) welding of angle ties to the beam, which restrict the maximum displacement of the top of
the bracing relative to the beam (Figure 1(a)), as proposed by Tzenov [37] for system with
joint connections. In order to prevent the prestressed bolts from shear, the gap size
a between the beam and the top of the bracing should be smaller than the length of the
slotted holes;
(b) introduction of unprestressed &bearing' bolts in slotted holes with a smaller size in parallel
to the prestressed bolts in slotted holes, which corresponds to the maximum relative
displacement of the sliding surfaces a (Figure 1(b)). This way of construction could be
related to the &safety hook' proposed by Roik et al. [38] which prevents the prestressed
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
1156 S. L. DIMOVA AND K. HIRATA
bolts in the considered friction-grip device from shearing, when unexpectedly strong
ground motion occurs.
The numerical simulations show that such system of seismic protection exhibits equally
stable response during both, the short-period and the long-period seismic excitation [1]. The
action of restrictors gives the possibility to reduce moment response in the columns thus
preventing inelastic deformations, while in the initial &bare' frame, in the frame with "xed bracing
and in the system with friction devices inelastic deformations occur during considered seismic
excitations.
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
SIMPLIFIED SEISMIC FRAGILITY ANALYSIS 1157
where x and x are the displacements of the upper and of the lower sliding surfaces, k is the
maximum value of the friction coe$cient k (see Figure 3(a)), F "Pk is the maximum
friction force, P is the normal force acting on the friction interface, c "Pk e\, c*"tga,
c*'0, according to the denotation of the angle a given in Figure 3(a) and e is a positive constant
chosen su$ciently small in order to describe as better as possible the sticking phase.
Numerical simulations in this study are performed for e"0.001 m/s, c*"tga"0.05 s/m and
k "0.15 in accordance with the studies by Dimova et al. [23] and Zinoviev [40]. The normal
force acting on the friction device stems partly from the dead load of the frame and mainly from
pressing forces introduced by the prestressed bolts. Maximum friction force mobilized is con-
sidered to be 200 kN in all four friction devices.
The action of restrictors is modeled by uniaxial link "nite element. Whenever the relative
displacement of the sliding surfaces exceeds a preliminary "xed value a , initial zero sti!ness of
the element increases to a value of k as shown in Figure 3(b). The magnitude of the sti!ness of the
J
restrictors is taken to be very high (k "2.0;10 kN/m), thus practically allowing no relative
J
displacement of the sliding surfaces. In the following numerical simulations the value of
a "0.005 m is considered to permit the development of energy dissipation through friction.
Also, such a value of a gave good results in the study of the e!ectiveness of the system with joint
connections [33].
Viscous damping in the structure is considered by introducing a Rayleigh proportional
damping matrix [C]"t[K], where [K] is the sti!ness matrix and t"¹ m/n with ¹ the natural
period of the frame. A typical viscous damping coe$cient for steel structures m"0.02 is used.
Further, in the comparative study of the seismic response and fragility of all the possible
structural modi"cations (as outlined in Figure 4), the following abbreviations are introduced:
FBS (#exible basic system)*initial &bare' frame with "rst natural period of 1.0 s;
RBS (rigid basic system)*frame with "xed bracing, with "rst natural period of 0.31 s.
FRS (friction system)*braced frame with friction devices without restrictors;
FRSR (friction system and restrictors)*braced frame with friction devices and restrictors.
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
1158 S. L. DIMOVA AND K. HIRATA
Axial force in the bracings of the FRS is not considered to be a critical parameter, because it
remains almost constant during the sliding of the friction device (the contribution of the inertial
force due to the bracings mass is negligible). This way, when properly designed the braces of the
FRS practically cannot su!er buckling failure even for considerably strong seismic intensities.
The fragility of restricting devices (bolts or angle ties) is not considered, because they are
assumed to be designed for a shear force larger than that causing the buckling of bracings.
Since the fragility analysis in this paper is directed to the potentiality of the considered
structures to withstand seismic excitations without damage of the main frame members, the
considered seismic capacities (shown in Table I) correspond to the elastic limit. The seismic
capacity for storey drifts is set to be 0.0122 m (1/250 of the storey height) according to the
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
SIMPLIFIED SEISMIC FRAGILITY ANALYSIS 1159
kN m kN m kN kN
provisions of the Bulgarian Codes for design of buildings and structures in seismic regions [49].
The seismic capacity of columns is related to the yield moment M . The values of M for yielding
stress p "2.85;10 kN/m are calculated taking into the account the axial force by the dead
load. The axial force capacity values N in the bracings of FRSR are calculated considering the
hinge connection to the connecting plate in order to provide proper conditions for the slotted
bolted connections. For this reason Nb for FRSR are smaller than the corresponding axial force
capacities for RBS where rigid connection of the bracings is considered (Table I). Axial force
capacities are calculated taking into account the axial forces by the dead load, too.
In the seismic fragility analysis of nuclear power plants usually the conditional probability of
failure is expressed as a function of the peak ground acceleration A which is considered to
describe well the ground motion intensity}response relationship, as reported by Kennedy and
Ravindra [3]. For base isolated fast breeder reactor structures Hirata et al. [12, 15] expressed the
seismic fragility as a function of the peak ground velocity < . Hirarta et al. evaluated the
conditional probability of failure for base-isolated reactor building using spectral pseudovelocity
S (¹"2.0 s) as the ground motion intensity parameter [14]. The considered ¹"2.0 s is the
fundamental period of the reactor corresponding to tangential sti!ness of the bi-linear
load}displacement relationship of isolators after yielding, which is considered to be equivalent
sti!ness in the region where the ductility factor is considerably large.
In the present study the following ground motion parameters are considered as seismic
intensity parameters, i.e. peak ground acceleration A , peak ground velocity < , spectral
pseudovelocity S and spectral acceleration S with damping h 5 per cent of critical. Since the
considered MDOF non-linear systems (FRSR and FRS) exhibit "rst periods of free vibrations
which depend on the amplitude of vibration and vary between 0.31 s (the natural period of RBS)
and 1.0 s (the natural period of FBS), the average values S and S in the periods band of
0.3}1.0 s of S and S , respectively, are considered as seismic intensity parameters of FRSR and
FRS. The seismic intensity parameters for FBS and RBS are the S and S values corresponding
to their "rst natural periods.
Twenty observed earthquake records (see Table II) are used to investigate the correlation
between the response of the considered systems, and the above-mentioned seismic intensity
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
1160 S. L. DIMOVA AND K. HIRATA
parameters. Records 1}12 are scaled to 10 and 20 times as large as the original records while
records 13}20 are used with their original intensity. In this way, wide range of values of seismic
intensity parameters, such as A , < , S (S ) and S (S ) is considered.
Linear regression analysis is performed for the seismic response parameters of the considered
structures on these seismic intensity parameters. The goal of the regression analysis is to derive
a prediction equation that describes the functional relationship between dependent variable
> and independent variable X (i.e. >I "f [X], where >I is the predicted value of > for a given
value of X). Although the prediction equation for the regression could be presented by higher-
order functions, these expressions will di!er signi"cantly from each other and will not lead to
a general result. The aim of linear regressions is to create a low-cost model of the &real situation',
which duplicates its essential properties and could be easily used as a predictive tool. As
a measure of the goodness of "t the coe$cient of determination R is used. It is de"ned as
R"
L (>I !>M )
G G
L (> !>M )
G G
(3)
where > is the value of the seismic response parameter obtained from the ith simulation, >I is
G
the predicted value of > for a given value of seismic intensity parameter, and >M is the mean
value of > .
G
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
SIMPLIFIED SEISMIC FRAGILITY ANALYSIS 1161
Figure 5. (a) Regression of the second storey drift of FRSR and FRS on peak ground acceleration; (b)
regression of the second storey drift of FRSR and FRS on mean spectral pseudovelocity.
The coe$cient of determination represents the proportion of the total variation in > that is
accounted for by the regression line [41]. R could be treated also as a parameter proportional to
the amount of scatter in the sample points about the regression line and thus is a measure of the
strength of bivariate association [41].
In Figures 5(a) and 5(b) are illustrated regression results for the second storey drift of FRSR
and FRS on A and S , respectively. In both cases of FRSR and FRS, response correlates
withS better than with A . Coe$cients of determination, obtained from the linear regression
analysis are presented in Table III. R of the regression between response parameters of FBS and
RBS on S are not presented in the table, because they coincide with those on S , since for elastic
spectra S "uS , where u is the natural frequency. The coe$cients of determination obtained for
RBS ("rst natural period of 0.31 s) show that its response parameters correlate well with A and
do not correlate with < . On the contrary, the response parameters of FBS ("rst natural period of
1.0 s) correlate well with < and do not correlate with A. These results con"rm that in the
low-period band the response spectra should be normalized to A and in the middle-period band
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
1162 S. L. DIMOVA AND K. HIRATA
Table III. Coe$cients of determination for linear regression analysis for observed accelerograms.
A < S S A < S
Storey 1 0.7995 0.7925 0.9018 0.8710 0.7468 0.4868 0.9701
drift 2 0.8187 0.8004 0.9003 0.8694 0.7191 0.4547 0.9631
D 3 0.8093 0.8227 0.8608 0.8358 0.6949 0.4346 0.9536
G
4 0.6981 0.7893 0.7530 0.7478 0.6879 0.4268 0.9461
Max 1 0.7972 0.7921 0.9018 0.8713 0.7494 0.4898 0.9704
moment 2 0.6851 0.7558 0.8712 0.8634 0.6824 0.5044 0.9188
in 3 0.7907 0.8132 0.8487 0.8228 0.6945 0.4332 0.9532
col. M 4 0.6764 0.7753 0.7328 0.7296 0.6879 0.4262 0.9453
G
Axial 1 0.4587 0.7751 0.8953 0.8591 0.6503 0.4868 0.9701
force 2 0.4903 0.7785 0.8942 0.8568 0.6344 0.4547 0.9631
in 3 0.4871 0.7891 0.8801 0.8422 0.6123 0.4346 0.9536
brac. N 4 0.4793 0.7974 0.8324 0.8124 0.6038 0.4268 0.9461
G
FRS FBS
Storey 1 0.5875 0.7605 0.8306 0.8975 0.5702 0.7807 0.9629
drift 2 0.5529 0.7132 0.8242 0.8871 0.5602 0.7671 0.9548
D 3 0.5112 0.6706 0.7967 0.8612 0.5886 0.7655 0.9610
G
4 0.5259 0.6395 0.7982 0.8665 0.6320 0.7784 0.9615
Max 1 0.5954 0.7671 0.8338 0.9004 5777 0.7822 0.9624
moment 2 0.5685 0.7173 0.8342 0.8950 0.5566 0.7640 0.9599
in 3 0.5125 0.6737 0.7979 0.8616 0.5950 0.7648 0.9613
col. M 4 0.5293 0.6317 0.7992 0.8661 0.6413 0.7786 0.9587
G
they should be normalized to < . On the basis of statistical study, Nau and Hall [42] determined
the low-period band from 0.11 to 0.5 s and middle period band from 0.5 to 5 s. The response
characteristics of both RBS and FBS correlate much better with the corresponding values of the
seismic intensity parameter S (and, respectively, with S ), than with the seismic excitation
characteristics A and < .
The response characteristics of FRSR exhibit slightly bigger coe$cients of determination on
the mean value of spectral acceleration S , than on the mean value of spectral pseudovelocity
S . On the contrary, the responses of FRS correlate better with S , than with S . The choice of
S as seismic intensity parameter for fragility analysis of linear systems such as RBS and FBS and
S as a seismic intensity parameter for fragility analysis of non-linear systems such as FRSR and
FRS is supported also by the following reasons:
(i) S is a basis of the Housner's spectrum intensity S [43], de"ned as the area under the
'
spectrum curve between periods of 0.1 and 2.5 s. Nau and Hal [42] proved that S is a very
'
good normalizing factor as for elastic response spectra, as well as for inelastic ones, in the
periods range of 0.5}5 s. The seismic intensity parameter S could be considered as a kind
of S calculated for the period band *¹ (from 0.31 to 1.0 s in the considered case) and
' G
normalized to *¹ . In this way, S relates S and also gives information for the shape of
G
'
S in the considered period band *¹ .
G
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
SIMPLIFIED SEISMIC FRAGILITY ANALYSIS 1163
(ii) for zero damping oscillator the Fourier amplitude spectrum (FAS) of the accelerogram is
the upper limit of S [44]. Since FAS may be interpreted as a measure of the total energy in
the end of the earthquake within the undamped oscillator, thus S and S also re#ect this
quantity for small values of damping.
(iii) FAS, divided by 2n, is the ground acceleration amplitude intensity at a given circular
frequency of - per unit of - as mentioned by Clough and Pensien [45]. This way S and
S for small values of damping directly relate the ground acceleration amplitude charac-
teristics, too.
A system with friction devices (FRS) could be considered as a system with degrading sti!ness in
which yielding is concentrated in the friction devices. The good correlation between the seismic
response of FRS and S gives reasons for further investigation of this seismic intensity estimate
for implementation in fragility analysis and as a normalizing parameter for linear and non-linear
response spectra. As to determination of an appropriate period band *¹ for the estimation of
G
S for systems with degrading sti!ness, it could be taken between the "rst natural period of the
initial (undamaged) structure and the "rst natural period of the structure with sti!ness, equivalent
to some appropriate sti!ness degradation of critical members.
The present study is based on a method for the estimation of reliability for elasto-plastically
behaving structures under strong earthquake motions, using reliability index, as "rst proposed by
Kanda [46, 47] and developed by Hirata et al. [12}15] for base-isolated structures. According to
this method safety margin of a member of the structure for the present level of seismic intensity is
de"ned as a ratio of its seismic capacity to its seismic response. Safety margin S (e) concerning
some seismic response parameter for a given seismic intensity e is de"ned as the ratio of its seismic
capacity R to the seismic response parameter S(e), i.e.
S (e)"R/S(e) (4)
P (e)"Prob(R/S(e)(1) (5)
Under the assumption that R and S are log-normally distributed, Equation (5) leads to the
following expression for fragility as a function of the seismic intensity parameter e:
ln S (e)!ln R
P (e)"
(6)
(b#b
where P (e) is the probability of failure as a function of the seismic intensity parameter e,
is the
cumulative function of standard normal distribution, S (e) and R are the median values of the
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
1164 S. L. DIMOVA AND K. HIRATA
seismic response S(e) and seismic capacity R, respectively, and b and b are the log-normal
standard deviations of R and S(e), respectively.
For a variate a which follows log-normal distribution, the median value a and lognormal
standard deviation ba could be expressed by its mean k and coe$cient of variation d as [48]
? ?
k
a " ? (7)
(1#d
?
and
b"ln(1#d ) (8)
? ?
Thus having the regression equation of the seismic response S(e) on the chosen seismic intensity
parameter e (see Section 4) one can express the seismic fragility of the structural members by the
use of Equation (6). It gives the possibility to consider the randomness inherent in the structural
response to seismic excitations (accounted by b ) and the randomness in structural capacity
(accounted by b ). The e!ect of modelling uncertainty herewith is not taken into account. When
needed, it could be expressed in the manner, for example as a present con"dence interval [3, 14].
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
SIMPLIFIED SEISMIC FRAGILITY ANALYSIS 1165
System Storey Floor slab Storey drift Max moment Axial force in
displacement D in columns M bracings N
G G G
1 0.05525 0.05525 0.04793 0.16840
FRSR 2 0.08571 0.08453 0.08353 0.21110
3 0.13030 0.39712 0.46281 0.19690
4 0.13571 0.36909 0.30935 0.18656
1 0.11150 0.11150 0.10755 3.8873.10\
FRS 2 0.14021 0.15892 0.16056 7.3478.10\
3 0.14895 0.30381 0.25927 4.3061.10\
4 0.14495 0.05455 0.04221 0.04254
1 0.11558 0.11558 0.11484 0.11558
RBS 2 0.14889 0.12217 0.12190 0.12208
3 0.12095 0.12779 0.12794 0.12779
4 0.12193 0.11773 0.11622 0.11774
1 0.20200 0.20200 0.20058
FBS 2 0.20280 0.20507 0.20444 No
3 0.20250 0.19968 0.19956 bracings
4 0.20228 0.15725 0.20435
when the storey drift in the next storey and the respective #oor slabs displacements are con"ned
by the action of restrictors and/or the sticking in friction devices. The maximum moment in the
columns in a given storey depends on the storey drift and this way shows similar c.o.v.
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
1166 S. L. DIMOVA AND K. HIRATA
for some of the seismic response parameters of FRSR and all of the seismic response parameters
of FRS have a negative free member b . The analysis of the zero crossings of these lines of the
seismic intensity axis have shown that these points are in very low seismic intensities where no
exceedance of seismic capacity is expected. This fact gives reason to accept zero seismic response
values in the case where the seismic response value is calculated negative.
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
SIMPLIFIED SEISMIC FRAGILITY ANALYSIS 1167
Storey 1 2 3 4
Storey drifts D
G
FRSR 0.99544 0.99219 0.97848 0.86781
FRS 0.97947 0.97183 0.97461 0.91478
Max moments M
G
FRSR 0.99490 0.98749 0.94532 0.85857
FRS 0.98049 0.96938 0.98118 0.90497
Axial forces in bracins N
G
FRSR 0.99790 0.99219 0.99835 0.96552
The above proposed concept of &mean seismic excitation' and linear regression of the response
of FRSR and FRS on S is proved by setting some value of S higher than the initial one, for
example S "1.0 m/s. Using the linear regression equations the predicted response is found.
The generated twenty arti"cial accelerograms are scaled to S "1.0 m/s and the mean responses
are obtained from the 20 calculated responses of FRSR and FRS, respectively. The calculated and
the predicted mean values are compared in Table VI.
The results obtained show very good agreement except for the prediction for the fourth storey
response of FRS, where the friction device starts to act for intensities of S "0.8 and higher, so
the regression line is derived for di!erent quantities of response in the considered intensities band
and exhibit comparatively low coe$cient of determination, as shown in Table V. These results
could be improved by a more appropriate choice of the range of seismic intensities in which the
regression analysis will be performed. For example, if the regression lines for the fourth storey
drift and moment of the FRS are obtained for seismic intensity S '0.8 m/s, the predicted values
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
1168 S. L. DIMOVA AND K. HIRATA
Table VI. Comparison of calculated and predicted mean values of the response of FRSR and FRS.
become D4"0.78531 cm and M4"47.985 kN m and the respective errors are #23.60 and
#39.17 per cent thus reducing the prediction error more than 2.6 times.
As will be shown in Section 6.3, almost all key risk contributors for each of the considered
structures are related to their response in the "rst or the second storey, so the obtained
comparative results between the predicted and calculated mean values show, that it is possible to
predict well the structural response of FRSR and FRS by the proposed concept. Moreover, an
error in the predicted mean response was already introduced with the choice of wave No. 16
which "tted not exactly with the calculated mean responses of all the considered structures in all
the storeys. In this context, the proposed concept which allows a considerable reduction of the
number of calculations, gives the possibility to concentrate the calculational e!orts in obtaining
more exact &mean seismic excitation' by considering a higher number of accelerograms. In this
way the coe$cients of variation of structural response parameters, needed for fragility estimation,
will be calculated more precisely, too.
P "0.06 for the "rst story drift D1 and P less than 10\ for the axial force in the "rst storey
bracings N1.
The key risk contributors for FRS (Figure 11) are the "rst storey drift and the second storey
drift. Di!erent from the case of FRS, the vulnerable seismic response parameters of FRSR (Figure
12) are the axial force in the "rst storey bracings and the axial force in the second storey bracings
for P '0.14. For P (0.14 the key risk contributors of FRSR become the storey drifts in the
fourth and third stories. This result is a consequence of the high values of the c.o.v. (up to 0.397)
for the third and fourth storey drifts. FRSR and FRS will withstand the seismic excitation with
the response spectrum parameter SM (¹"0.31}1.0 s, h"5 per cent)"0.4285 m/s considered in
Section 6.1 with P "0.03 and P "0.1;10\, respectively.
The above identi"cation of the key risk contributors gives the possibility to clarify the ways
of improving the structural seismic safety. For FRS it would be optimization of the value of
maximum friction force in the di!erent storeys. FRSR exhibits risk contributors in the range of
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
1170 S. L. DIMOVA AND K. HIRATA
the relatively small fragilities due to the big scatter of the seismic response parameters which
normally do not contribute to the vulnerability of the structure}storey drifts and related to them
moments in the upper storeys. This result also naturally calls for optimization of the design values
of the maximum friction force in the height of the structure to provide simultaneous action of the
restrictors and/or friction devices in all the storeys for a given seismic intensity. Regarding
the axial forces in the "rst and second storey bracings, the seismic behaviour of FRSR could be
improved by designing the restricting elements (bolts or angle ties) to break before the axial force
in the bracings will reach its critical value. This way even after the break of restrictors FRSR will
have redundant seismic resistance by behaving like FRS and will keep its ability to dissipate
energy.
In order to compare the fragilities of the di!erent structures their &envelope' fragility curves, i.e.
the fragility curves of the most vulnerable seismic response parameters are compared in Figure 13.
The envelope fragility of FRS consists of the fragility curve of the "rst storey drift till
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
SIMPLIFIED SEISMIC FRAGILITY ANALYSIS 1171
S "0.57 m/s and from the second storey drift fragility curve for S '0.57 m/s. The envelope
fragility of FRSR consists of the fragility curve of the third storey drift till S "0.63 m/s and
from the axial force in the "rst storey bracings fragility curve for S '0.63 m/s. Since the mean
values of the seismic intensity parameters (see Section 6.1) of the generated accelerograms are
di!erent for the di!erent structures, the seismic intensity for each system is normalized as
S "S /S (10)
C G
where S is the normalized seismic intensity, S comprises the values of intensity considered in
C
fragility estimation of each structure, and S are the mean values of the seismic intensity
G
parameters calculated for those generated from the design response spectrum waves (Figure 6), as
follows:
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
1172 S. L. DIMOVA AND K. HIRATA
Figure 14. Comparison of fragilities of the "rst storey bracing of FRSR on normalized intensity for recorded
and generated accelerograms.
Figure 15. Comparison of fragilities of the second storey drift of FRS on normalized intensity for recorded
and generated accelerograms.
When using the mean seismic excitation concept, regression analysis of seismic response
parameters on seismic intensity is performed for one accelerogram. The used seismic intensity
parameters SM and SM (described in Section 6.1) for only one accelerogram are linearly propor-
tional to its peak acceleration A . In this case the fragility on the normalized seismic intensity
S calculated on SM and SM coincides with the fragility on the normalized seismic intensity
S calculated on A , as shown in Figures 14 and 15 for the fragility of the "rst storey bracing of
FRSR (N1) and second storey drift of FRS (D2), respectively. Herewith arises the question what is
the meaning of introduction of the proposed seismic intensity parameters SM and SM . To answer
this question fragility of the above considered seismic response parameters N1 and D2 is
calculated for 20 recorded accelerograms used in Section 4 with di!erent spectral content and
intensity by use of linear regression on S and A . It is assumed that the coe$cients of variation
of the response parameters coincide with those for generated accelerograms given in Table IV.
The comparison of the fragilities on the normalized intensity conclusively shows that the fragility
calculated for the speci"ed target response spectrum gives an estimate of the fragility for the
recorded accelerograms with di!erent spectral content, when the latter is calculated on S (see
Figures 14 and 15).
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
SIMPLIFIED SEISMIC FRAGILITY ANALYSIS 1173
7. CONCLUSIONS
The main conclusions from the presented estimation of seismic fragility of structures with two
types of friction devices could be summarized as follows:
1. Seismic intensity parameter, de"ned as the mean value of the pseudovelocity spectrum in the
periods band of the "rst natural periods of the initial &bare' frame and frame with "xed
bracings, gives possibility to use linear regression analysis of the response parameters of the
considered non-linear structures on seismic intensity.
2. Simpli"ed method of fragility estimation, based on the concept of &mean seismic excitation'
in the estimation of the mean response value and linear regression of the seismic response
parameters on seismic intensity parameter allows to reduce considerably the computational
time. A comparison between the mean values of the response parameters predicted by this
concept and calculated mean responses shows very good agreement. This comparison shows
that the seismic response of the system with friction devices and restrictors is more
predictable, than those of the system with friction devices.
3. The fragility analysis gave the possibility to identify the key risk contributors for the
system with friction devices and for the system with friction devices and restrictors. On
the basis of these results the seismic resistance of these systems could be increased by
optimization of the value of maximum friction force in the di!erent storeys. The seismic
behaviour of the system with friction devices and restrictors could be improved also by
designing the restricting elements (bolts or angle ties) to break before the axial force in the
bracings will reach its critical value. Even after the break of restrictors the system will have
a redundant seismic resistance and will keep its ability to dissipate energy in the friction
devices.
4. The comparative fragility analysis, based on normalized seismic intensity, shows that when
the initial &bare' frame is retro"tted by rigidly connected bracings, the e!ect is much lower,
than in the case of connecting the bracings by friction devices and especially by friction
devices and restrictors.
5. The seismic response of the systems with friction devices could be substantially improved by
introducing restrictors of the relative displacements of the sliding surfaces. The system with
friction devices and restrictors exhibits higher seismic safety, than the system with friction
devices and could be proposed for further development and application.
6. The calculated fragility by the proposed simpli"ed method for accelerograms com-
patible with a given seismic response spectrum on the normalized intensity gives also an
estimate of the fragility for seismic excitations with di!erent spectral content, when the
latter is calculated on the above proposed mean value of the pseudovelocity spectrum.
The use of the peak ground acceleration as seismic intensity parameter for seismic
excitations with di!erent spectral content does not give precise enough estimation of the
fragility.
ACKNOWLEDGEMENTS
The "rst author gratefully acknowledges the "nancial support from the Japan Science and Technology
Corporation under STA fellowship.
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
1174 S. L. DIMOVA AND K. HIRATA
REFERENCES
1. Dimova SL. Seismic protection of frame structures by friction devices with restrictors. Proceedings of the 11th ECEE,
Paris, 1998; (on CD).
2. Yang CY. Random <ibration of Structures. Wiley: New York, 1986.
3. Kennedy RP, Ravindra MK. Seismic fragilities for nuclear power plant risk studies. Nuclear Engineering and Design
1984; 79: 47}68.
4. US NRC (Nuclear Regulatory Commission). An approach to the quanti"cation of seismic margins in nuclear power
plants. N;REG/GR-4334, 1985.
5. IAEA, International Atomic Energy Agency. The role of probabilistic safety assessment and probabilistic safety
criteria in NPP safety. Safety Series, No 106, Vienna, Austria, 1992.
6. IAEA, International Atomic Energy Agency. Probabilistic safety assessment (INSAG-6). Safety Series, No. 75,
Vienna, Austria, 1992.
7. Lederman L, Niehaus F, Tomic B. Probabilistic safety assessment past, present and future. An IAEA perspective.
Nuclear Engineering Design 1996; 160: 273}285.
8. Kafka P. Probabilistic safety assessment: quantitative process to balance design manufacturing and operation for
safety of plant structures and systems. Nuclear Engineering and Design 1996; 165: 333}350.
9. Penzien J, Liu SC. Nondeterministic analysis of nonlinear structures subjected to earthquake excitations. Proceedings
of the 4th =CEE, vol. 1, Section A-1, Santiago, Chile, 1969.
10. Ruiz P, Pensien J. Probabilistic study of behaviour of structures during earthquakes. EERC Report No 69-3,
Earthquake Engineering Research Center, University of California at Berkeley, CA, 1969.
11. SchueK ler GI (ed.). Structural Dynamics. Recent Advances. Springer: Berlin, 1991.
12. Hirata K, Kobayashi Y, Kameda H, Shiojiri H. Fragility of seismically isolated FBR structure. Nuclear Engineering
and Design 1991; 128: 227}236.
13. Hirata K, Shirahama K, Somaki T. Structural reliability of seismic isolation system. From Probabilistic Mechanics
and Structural and Geotechnical Reliability, 6th Speciality Conference, EM, ST, GTDiv/ASCE, 57}60, Denver, CO,
1992.
14. Hirata K, Ootori Y, Somaki T. Seismic fragility analysis for base-isolated structure. Journal of Structural and
Constructive Engineering, AIJ 1993; 452: 11}19 (in Japanese).
15. Hirata K, Somaki T. Fragility estimation of an isolated FBR structure considering the ultimate state of rubber
bearings. Nuclear Engineering and Design 1994; 147: 183}196.
16. Dorka UE, Pradlwarter HJ. Reliability based retro"tting of RC frames with hysteretic devices, In Structural Safety
& Reliability. SchueK ller, Shinozuka, Yao (eds), Balkema: Rotterdam, 1994; 1545}1552.
17. Dorka UE, Ji A, Dimova S. Earthquake safety of large panel buildings retro"tted with hysteretic devices. Proceedings
of ICOSSAR, Kyoto, Japan, 1998; 1533}1538.
18. Kelly JM. Aseismic base isolation: review and bibliography. Soil Dynamics and Earthquake Engineering 1986; 5(4):
202}217.
19. Buckle IC, Mayes RL. Seismic isolation: history, application and performance * a world overview. Earthquake
Spectra 1990; 6(2): 161}202.
20. Nagarajaiah S, Reinhorn AM, Constantinou MC. Nonlinear dynamic analysis of three-dimensional base isolated
structures (3D-BASIS). Report No NCEER-89-0019, National Center for Earthquake Engineering Research, State
University of New York at Bu!alo, 1989.
21. Nagarajaiah S, Reinhorn AM, Constantinou MC. 3D-BASIS: nonlinear dynamic analysis of three-dimensional base
isolated structures: Part II. Report No NCEER-91-0005, National Center for Earthquake Engineering Research, State
University of New York at Bu!alo, 1991.
22. Warburton GB. Reduction of <ibrations. ¹he third Mallet}Milne ¸ecture. Wiley: New York, 1992.
23. Dimova SL, Meskouris K, Kraetzig WB. Numerical technique for dynamic analysis of structures with friction devices.
Earthquake Engineering and Structural Dynamics 1995; 24: 881}898.
24. De Angelis M, Ciampi V. E!ectiveness of dissipative connections on improving earthquake response of adjacent
buildings. Proceedings of the 10th ECEE, vol. 3. Balkema: Rotterdam, 1995; 1891}1896.
25. Dorka UE, Conversano GA. Seismic retro"t of allstate building. Report of the IABSE Symposium, San Francisco,
1995; 145}150.
26. Dowdel DJ, Cherry S. Optimal seismic response control of friction damped structures. Proceedings of the 10th ECEE.
Balkema: Rotterdam, 1995; 2051}2056.
27. Vulcano A, Mazza F. Earthquake damage mitigation of R.C. frames by using damped steel braces. Proceedings of the
EERC-C;RE Symposium in honor of <.<. Bertero, ;CB/EERC-97/05, 1997; 199}203.
28. Antonucci R, Giacchetti R, Di Matteo G. Design parameters optimization for R.C. structures incorporating energy
dissipation bracings. Proceedings of the EERC-C;RE Symposium in honor of <.<. Bertero, ;CB/EERC-97/05, 1997;
211}215.
29. Beucke KE, Kelly JM. Equivalent linearizations for practical hysteretic systems, International Journal of Non-¸inear
Mechanics 1985; 23(4): 211}238.
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175
SIMPLIFIED SEISMIC FRAGILITY ANALYSIS 1175
30. Dimova SL. Comparison and veri"cation of two techniques for friction forces representation. Computers & Structures
1996; 60(1): 11}19.
31. Dimova SL. Modeling of friction forces with decreasing friction coe$cient during the sliding phase. Proceedings of the
Mouchel Centenary Conference on Innovation in Civil and Structural Engineering, Cambridge, England, 1997; 330}336.
32. Tzenov L. Reduction of the seismic vulnerability of structures. Doctor1s ¹hesis, CLSMEE-BAS, So"a, 1983 (in
Bulgarian).
33. Dimova SL. System with joint connections subjected to dynamic excitations. PhD ¹hesis, BAS, So"a, 1987 (in
Bulgarian).
34. Olariu I. Passive control and base isolation: state-of-the-art lecture. Proceedings of the 10th ECEE, Balkema:
Rotterdam, 1995; 703}713.
35. Dimova SL, Tzenov L. Analysis of a system of special seismic protection to real strong ground motion. Proceedings of
the 9th ECEE, 1990; 7-B: 50}55.
36. Pall AS, Marsh C. Response of friction damped braced frames. Journal of Structural Division, ASCE 1982; 108:
1313}1323.
37. Tzenov L. Seismically Protected Structure. License No. 26752, INRA, So"a, 1982 (in Bulgarian).
38. Roik K, Dorka UE, Dechent P. Vibrational control of structures under earthquake loading by three stage friction
grip elements. Earthquake Engineering and Structural Dynamics 1988; 16: 501}521.
39. Robinson RH. Inelastic design of steel frames to resist seismic loads. ESSB Report No. 11, Massachusetts Institute of
Technology, Cambridge, MA, 1977.
40. Zinoviev V. (ed.). Short ¹echnical Handbook, Part 1. State Publishers of Technical and Theoretical Literature,
Moscow, 1952 (in Russian).
41. Size WB. (ed.). ;se and Abuse of Statistical Methods in the Earth Sciences. Oxford University Press: New York, 1987.
42. Nau JM, Hall WJ. Scaling methods for earthquake spectra. Journal of Structural Engineering, ASCE 1984; 110(7):
1533}1548.
43. Housner GW. Behaviour of structures during earthquakes. Journal of the Engineering of Mechanical Division, ASCE
1959; 85(EM4): 109}129.
44. Clough RW, Pensien J. Dynamics of Structures (1st edn). McGraw Hill: New York, 1975.
45. Clough RW, Pensien J. Dynamics of Structures (2nd edn). McGraw Hill: New York, 1993.
46. Kanda J. Probability-based seismic margin index for inelastic members of reactor buildings. ¹ransactions of the 8th
SMiR¹, M1K2/5, 1985; 353}359.
47. Kanda J, Iwasaki R, Sunohara H. Stochastic evaluation of inelastic seismic response of a simpli"ed reactor building
model. ¹ransactions of the 9th SMiR¹, K1, 1987; 403}408.
48. Ang ZH-S, Tang WH. Probability Concepts in Engineering Planning and Design, Basic Principles, vol. 1. Wiley: New
York, 1975.
49. Bulgarian Committee for Area and Village Arrangement and Bulgarian Academy of Sciences, Regulations for design of
buildings and structures in seismic regions. So"a, 1987 (in Bulgarian).
50. Architectural Institute of Japan. Recommendations for limit state design of steel structures (draft), 1998.
Copyright 2000 John Wiley & Sons, Ltd Earthquake Engng Struct. Dyn. 2000; 29:1153}1175