You are on page 1of 15

This article was downloaded by: [UNAM]

On: 16 February 2011


Access details: Access Details: [subscription number 918399446]
Publisher Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-
41 Mortimer Street, London W1T 3JH, UK

Structure and Infrastructure Engineering


Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713683556

Seismic vulnerability functions of multi-storey buildings: estimation and


applications
Luis Estevaa; Orlando Díaz-Lópeza; Eduardo Ismael-Hernándeza
a
Institute of Engineering, National University of Mexico, Ciudad Universitaria, Mexico, DF, Mexico

First published on: 22 January 2009

To cite this Article Esteva, Luis , Díaz-López, Orlando and Ismael-Hernández, Eduardo(2010) 'Seismic vulnerability
functions of multi-storey buildings: estimation and applications', Structure and Infrastructure Engineering, 6: 1, 3 — 16,
First published on: 22 January 2009 (iFirst)
To link to this Article: DOI: 10.1080/15732470802663755
URL: http://dx.doi.org/10.1080/15732470802663755

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf


This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
Structure and Infrastructure Engineering
Vol. 6, Nos. 1–2, February–April 2010, 3–16

Seismic vulnerability functions of multi-storey buildings: estimation and applications


Luis Esteva*, Orlando Dı́az-López and Eduardo Ismael-Hernández
Institute of Engineering, National University of Mexico, Ciudad Universitaria,
04320 Mexico, DF, Mexico
(Received 10 October 2007; final version received 3 December 2008)

Seismic vulnerability functions of structural systems are presented as expressions relating earthquake intensities with
quantitative measures of their probable consequences on the performance of those systems. The consequences
considered include direct and indirect costs of structural and non-structural damage, either for the condition of
system survival or for the possibilities of partial or total collapse. A review is presented of recent efforts towards the
development of these functions for multi-storey building systems and for their application in life-cycle optimization
studies for the establishment of target safety levels and the corresponding seismic design criteria. Special attention is
given to the evaluation of system reliability with respect to the ultimate capacity (collapse) failure mode. Results are
presented of some studies about the time-dependent process of damage accumulation and reliability evolution in
building frames, as well as of the optimum design criteria and maintenance strategies for structural frames with
hysteretic energy-dissipating devices.
Downloaded By: [UNAM] At: 13:20 16 February 2011

Keywords: seismic vulnerability; secant-stiffness reduction index; expected damage functions; life-cycle optimisation;
energy-dissipating devices; optimum maintenance policies

format where the components of the vulnerability


1. Introduction functions accounting for expected values of
For a given structural system, a seismic vulnerability monetary losses (whether direct or indirect) are
function serves to express the relationship between kept separate from those related to life and limb
the intensity of an earthquake excitation and a potential losses.
quantitative measure of its probable consequences on Emphasis in this article is centred on the
the performance of that system. These consequences estimation of quantitative indicators of system
may include concepts such as the direct and indirect reliability and expected performance required to
costs of structural and non-structural damage, either obtain probabilistic assessments of the types of losses
for the condition of system survival or for the described in the previous paragraph. For this
possibilities of partial or total collapse. In many purpose, a critical review is presented in x2 of
applications related to risk analysis and risk-related some previous research results for the estimation of
engineering decision making, such as studies about expected damage functions of structural elements
life-cycle optimum performance levels, it has been and subsystems, as well as of the seismic reliability
deemed convenient to express vulnerability functions of multi-storey buildings and of the
functions in monetary terms, after recognising the resulting recommendations for the formulation of
difficult problems related to assigning values to performance- and reliability-based criteria and meth-
consequences, such as loss of human lives and ods for the practice of earthquake engineering. This
arousing of social unrest, which cannot be directly is followed by several proposals to fill identified gaps
expressed in those terms. Dealing with these pro- of previous research, or to provide alternatives to
blems is beyond the scope of this article. However, improve available criteria and algorithms for the
considering that risk-related decisions must be estimation of seismic vulnerability functions of multi-
sensitive to socially acceptable risk levels, in addition storey building systems. Some examples are pre-
to those resulting from formal cost-benefit optimisa- sented to illustrate the application of the proposed
tion formulations, here, we have preferred to adopt a new concepts and criteria.

*Corresponding author. Email: lestevam@ii.unam.mx

ISSN 1573-2479 print/ISSN 1744-8980 online


Ó 2010 Taylor & Francis
DOI: 10.1080/15732470802663755
http://www.informaworld.com
4 L. Esteva et al.

lateral distortions for different types of structural sub-


2. Seismic vulnerability functions of multi-storey
assemblages. Reyes (1999) includes, among others,
buildings
reinforced concrete frames with different qualities of
2.1. Failure and damage components of seismic ductility-oriented detailing, flat slabs, isolated and
vulnerability functions coupled reinforced concrete shear walls, steel frames
According to Esteva et al. (2002c), the seismic with different types of seismic details, cross-braced
vulnerability function of a structural system exposed panels and infill walls (masonry, precast elements, roof
to several failure modes, including collapse (exceeding systems and glass). In addition, quantitative indicators
of ultimate capacity), can be expressed as follows: of physical damage that consider maximum values of
local distortions and hysteretic energy dissipation or
dðyÞ ¼ dðyjSÞð1  pF ðyÞÞ þ dF pF ðyÞ; ð1Þ low-cycle fatigue indexes have been developed on the
basis of experimental studies on pseudo-dynamic or
where d(y) is the expected value of the cost of the dynamic tests (Park and Ang 1984). However, no
consequences of the performance attained by the significant efforts have been devoted to apply this
system if subjected to an earthquake of intensity y; knowledge to the estimation of vulnerability functions
pF(y) is the corresponding probability of collapse; and of complex nonlinear systems, such as tall buildings, in
d(yjS) and dF are the expected values of the costs of spite of their relevance for the establishment of
consequences, conditional to survival or to collapse of optimum performance-based seismic design criteria.
the system, respectively. In all cases, the expected cost Similar comments can be made about vulnerability
values are normalised, dividing them by the initial problems related to overturning or other failure modes
Downloaded By: [UNAM] At: 13:20 16 February 2011

construction cost C0. Function d(yjS) depends on the of equipment, furniture, objects or installations sensi-
probability distribution of the physical damage experi- tive to floor velocities or accelerations, which have
enced by all structural and non-structural components received attention during the last few years, following
of the system, as well as of the direct and indirect an approach proposed by Ishiyama (1982), with
consequences of that damage. Its evaluation must also important results that have hardly transcended into
account for potential damage experienced by equip- practical applications to performance oriented design.
ment, installations or contents that may fail, for This shortcoming may be blamed, partly, at least, on
instance, by overturning or by excessively high the deficient (or zero) interaction between structural
dynamic response. engineers and equipment providers.
Figure 1 shows two usual manners of representing The experimental information provided by Reyes
the seismic vulnerability function d(y). The dashed line (1999) and the references mentioned by him can be
represents failure probabilities, while the dotted line represented by equations of the following form (Esteva
represents expected damage costs. The latter can take et al. 2002c):
values greater than unity, because actual damage
values can be greater than the initial construction cost. dd ðcÞ ¼ 0; c < ccr ð2aÞ
Reyes (1999) presents a list of publications report- and
ing the results of laboratory tests oriented to relating
physical damage levels with maximum amplitudes of dd ðcÞ ¼ 1  exp ðkðc  ccr Þm Þ; c > ccr : ð2bÞ

In these equations, dd is the ratio of the direct cost of


damage on a structural or non-structural element to its
initial construction cost, when it is subjected to a
lateral distortion with an amplitude c; ccr is the value
of that distortion required to initiate cracking; and k
and m are empirically determined constants. For some
practical applications, the forms of Equation (2) can be
simplified by taking ccr ¼ 0.
Ismael-Hernández et al. (2004) present some simple
models to estimate component d(yjS) of the vulner-
ability function d(y), including the contributions of
direct and indirect costs. For this purpose, they adopt
an equation of a form equivalent to:
X
Figure 1. Expected damage and failure probability dðyjSÞ ¼ ðc þ r1 Þ ddi ðyjSÞ; ð3Þ
vulnerability functions. i
Structure and Infrastructure Engineering 5

where ddi (yjS) is the expected value of the direct cost of seismic response of the system of interest, and capable
damage for the ith component of the system; r1 is the of producing accurate estimates of those variables that
ratio of direct to indirect costs; and c, given by determine the reliability functions of interest. This is
Equation (4), depends on the conditional value of the ordinarily achieved using the ordinate of the linear
expected damage and on a pair of variables, a and q, response spectrum for the fundamental period of the
where a 4 q. The q variable takes into account that system. Approximate estimates of second-moment
the direct replacement cost of a given system compo- probabilistic indicators of both variables (response
nent is usually larger than its initial construction cost, demand and deformation capacity) are often obtained
and a takes into account a fixed initial cost attached to with the aid of a simplified reference system, char-
repair costs, associated with the logistic arrangements acterised by mechanical properties determined by
that have to be implemented before the actual repair means of a static pushover analysis of the detailed
work starts. For the calculation of ddi (yjS), it is system. A normalised value of the intensity can then be
necessary to obtain expected values of functions used, such as the expected value of the ductility
similar to those given by Equation (4) with respect to demand on the simplified reference system (Esteva
the probability density function of the corresponding et al. 2002a, b), or as the ratio of that demand to the
distortion, ci. corresponding deformation capacity, expressed in
terms of available ductility.
a qX
n
In spite of the generalised use of deformation
c¼a ddi ðyjSÞ: ð4Þ
n i capacities as a basic variable for the establishment of
failure conditions for multi-storey buildings, it has
Downloaded By: [UNAM] At: 13:20 16 February 2011

In this equation, n is the number of components of the been recognised that their probabilistic estimates
system that may suffer damage during the seismic obtained by means of pushover analysis are tied to
excitation and q is an empirical constant, taken equal severe limitations, as, according to this approach, it is
to 1.2 by Ismael-Hernández et al. (2004). It can easily not possible to account for (a) the influence of
be seen that c tends to q when the direct cost of damage cumulative damage associated with the cyclic response
for the whole system tends to unity, while it tends to a and (b) the dependence of the lateral deformation
when the direct cost tends to zero. capacity on the response configuration of the system
when it approaches failure. Additional limitations are
related to the practical difficulties associated with the
2.2. Probability of reaching the ultimate limit state identification of the critical storeys (those with the
(collapse) highest demand/capacity ratios) and with the fact that,
Displacement-based practical criteria for earthquake at some storeys, deformation capacities cannot be
resistant design are based on the concept that the event reached through conventional pushover analysis, as
of reaching a given performance limit is associated this can be prevented by the occurrence of lateral
with the condition that the structural distortion failure mechanisms at other storeys.
reaches a value equal to the corresponding deforma- Figure 2, taken from Pérez-Martı́nez (2007), illus-
tion capacity. According to this, the estimation of the trates one of the problems related to the definition of
seismic reliability of complex nonlinear systems for the global lateral deformation capacity of a multi-
given values of the ground-motion intensity is ordina- storey system. The figure presents some results obtained
rily based on a measure of the probability that, for an from the step-by-step nonlinear response analysis of a
ensemble of earthquake excitations with a specified five-storey building frame with masonry infill panels; it
intensity, the ratio of the peak absolute value of the shows cyclic shear force versus lateral distortion curves
nonlinear displacement response demand of the system for the fourth storey, where the lateral distortion
to the corresponding deformation capacity is greater reaches its maximum value, as well as for the system
than unity. For multi-storey buildings, both displace- as a whole (base shear versus roof displacement). It is
ment demand and deformation capacity can either be easily observed that, while the curve corresponding to
expressed in terms of the global lateral distortion (that the fourth storey is regular and stable, that correspond-
is, the relative displacement of the roof with respect to ing to the global distortion shows very erratic beha-
the foundation divided by the building height) or in viour, which results from the fact that the lateral
terms of the storey distortion where the ratio of the configuration of the system does not remain propor-
displacement demand to the deformation capacity is tional throughout time, but keeps changing from
largest during the dynamic response. instant to instant, thus affecting the ratio of base shear
When defining a seismic vulnerability function, the to global distortion for a given value of the latter.
earthquake intensity must be measured by an adequate Alternative criteria to determine seismic vulner-
indicator, having a high statistical correlation with the ability functions have been proposed that do not
6 L. Esteva et al.

Figure 2 Cyclic shear force versus lateral distortions for a multi-storey building frame (from Pérez-Martı́nez 2007): (a) fourth
storey distortions and (b) global distortions.

depend on the use of the concept of deformation In this paper, an alternative approach is explored,
Downloaded By: [UNAM] At: 13:20 16 February 2011

capacity (Esteva 1992, Shome and Cornell 1999, based on the use of a secant-stiffness reduction index
Alamilla and Esteva 2006); according to these refer- for the estimation of reliability functions relating the
ences, system collapse is assumed to take place when reliability index b (Cornell 1969) with the ground
the displacements predicted by the dynamic response motion intensity, including the influence of cumulative
analysis become infinitely large and non-reversible. damage and avoiding the need to obtain probabilistic
The stiffness matrix then becomes ill conditioned. definitions of lateral deformation capacities. This
According to this, the safety factor with respect to this approach is based on the concept of failure intensity,
type of failure for a given ground motion time history mentioned above.
is obtained as the scale factor that has to be applied to
that time history in order to reach the minimum value
3. Secant-stiffness reduction index as a basis for
of the intensity required to produce system collapse.
seismic reliability and vulnerability functions
This intensity will be denoted ‘failure intensity’.
Because the determination of the required scaling 3.1. Probabilistic relationships between intensity and
factor has to be attained by means of an iterative stiffness-reduction index
procedure, it may call for excessive demands of The possibility of estimating the seismic reliability
computer time. function of a structural system in terms of a probabil-
The method of incremental dynamic analysis (IDA; istic description of the reduction of the secant lateral
Vamvatsikos and Cornell 2002) offers both possibilities stiffness for the maximum amplitude of the response of
for the estimation of probabilistic indicators of seismic the system of interest to an earthquake ground motion
reliability for given ground motion intensities: either of specified intensity was first proposed by Esteva and
on the basis of deformation capacities or using the Ismael (2003) and further refined by Esteva and Dı́az-
concept of failure intensity. According to this method, López (2006). In the following, this criterion is reviewed
the lateral deformation capacity of the system of and some recently developed concepts intended to
interest is defined as the value of the peak roof improve its efficiency are presented.
displacement when the slope of the incremental Consider a multi-storey building with a cyclic
dynamic response function (spectral acceleration ver- response curve similar to that shown in Figure 3,
sus peak roof displacement) becomes lower than a when subjected to a horizontal earthquake ground
specified value; this point also serves to define the motion. In this figure, the x-axis represents the relative
failure intensity. It offers the possibility to visualise lateral displacements u of the roof of the building with
the evolution of the lateral roof displacement as respect to its base, while the instantaneous values of
the intensity grows, as well as to observe the values the base shear forces V are represented by the y-axis.
reached by the ground motion intensity before the At the instant when the lateral relative displacement
occurrence of unbounded peak roof displacements. is largest, the secant global stiffness K ¼ V/u has
However, these advantages are often tied to excessive suffered a reduction from its initial value K0. The
computer time demands (Dolsek and Fajfar 2004). collapse of the system can be expressed by the
Structure and Infrastructure Engineering 7

Figure 3. Cyclic response curves for base-shear versus roof lateral displacement for a system with tri-linear behaviour.

condition K ¼ 0. Therefore, a secant-stiffness reduc-


Downloaded By: [UNAM] At: 13:20 16 February 2011

tion indicator ISSR ¼ (K0 – K)/K0 can be defined, so


that the failure condition is expressed as ISSR ¼ 1.0.
Figure 4 shows values of Q, the natural logarithm
of ISSR of a multi-storey building system for different
values of m0, the ductility demand of the corresponding
simplified reference system, which may be interpreted
as a value of the intensity normalised with respect to a
simple indicator of the seismic capacity of the system.
In consistence with its definition, Q cannot take values
higher than 0, which leads to a discrete concentration
of its probability density function at that value.
Therefore, for a given value of the intensity (y), the
probability density function of Q ¼ ln ISSR is equal to
fQ(q), which is continuous for q 5 0 and includes a
discrete concentration at q ¼ 0, which is equal to
pF(y) ¼ P(Q ¼ 0jy), the failure probability for an Figure 4. Safety margins in a multi-storey building system
intensity equal to y. In order to permit the joint versus ductility demands in a simplified reference system.
analysis of information including cases with values of
Q  0, Esteva and Ismael (2003) introduced an
auxiliary variable U, such that Q ¼ U for Q 5 0
and Q ¼ 0 for U  0 (i.e. for the failure condition).
The probability density function of U is designated by
fU(), and the associated cumulative distribution
function by FU() (see Figure 5). These functions
are characterised by a set of parameters {k} that
determine their basic properties, such as mean,
standard deviation and skewness. (For instance, in
the case of a Gaussian distribution, {k} may be made Figure 5. Introduction of auxiliary variable to represent the
by the mean and the standard deviation of U.) These probability density function of Q(y).
parameters can, in turn, be expressed as functions of y,
G(yja), with given form and unknown parameters, a.
It is also assumed that we count with a set of m þ n step time history dynamic response analysis. Under the
pairs of values of the normalised intensity m0i, and the assumption that these pairs of values are ordered so
corresponding logarithmic damage index Q ¼ qi, that Q is smaller than 0 (survival condition) for the first
where the latter is obtained by means of a step-by- m cases, and equal to 0 for the remaining n cases, the
8 L. Esteva et al.

likelihood function can be expressed as follows in estimated using a sample including m values of
terms of a: ISSR 5 1.0 and n values of ISSR ¼ 1.0. For the
observed sample, the likelihood function of (a 1, a 2)
Y
m Y
n    
LðaÞ ¼ fU ðui jyi ; aÞ 1  F U 0 yj ; a : ð5Þ can be expressed as:
i¼1 j¼1
Y
m Y
n    
Lða1 ; a2 Þ ¼ f Z ð z i j ui ; a1 ; a2 Þ 1  FZ zj uj ; a1 ; a2 :
The vector of parameters a is determined by means of a i¼1 j¼1
maximum likelihood approach applied to the sample
ð7Þ
of pairs of values of Y and Q.
This article follows an approach proposed by
Esteva and Dı́az-López (2006) for the determination In order to get sufficiently accurate estimates of the
of the reliability function b(y), starting from a sample mean and the standard deviation of ZF, without an
of pairs of values of ISSR and Y. Here, b is the safety excessively large computational effort, it is convenient
index proposed by Cornell (1969) and y is a given value to generate samples containing a large proportion of
of the ground motion intensity. If the sample includes its points with values of ISSR slightly smaller than 1.0.
only cases with ISSR 5 1.0, the reliability function can This can be achieved by means of an algorithm that
be obtained by means of a regression analysis; if cases includes two basic steps: (a) an approximate probabil-
with ISSR ¼ 1.0 are also included, a maximum like- istic relationship between ISSR and Z is first obtained
lihood analysis must be performed. Instead of for- from a small sample of those variables and (b) a new
mulating the problem as that of obtaining an indicator sample of values of ISSR is determined for a set of
Downloaded By: [UNAM] At: 13:20 16 February 2011

of the probability that ISSR 5 1.0 (survival) for a given simulated ground motion records, with their intensities
intensity, attention is focused on the determination of chosen in such a manner that they have high
second moment indicators of the probability distribu- probabilities of producing damage levels in the
tion of ZF ¼ lnYF, where YF is the minimum value required intervals. This probabilistic relationship just
of the intensity required to lead to the condition mentioned may be as simple as the expected value of
ISSR ¼ 1.0 (collapse). For an earthquake with intensity Z ¼ lnY as a function of ISSR.
equal to y, the safety margin ZM is defined as the
natural logarithm of the ratio YF/y. The reliability
function can then be expressed as: 3.2. Efficient scheme for Monte Carlo simulation of
pairs of values of ISSR and Z
bðyÞ ¼ ðEðZF Þ  ln yÞ=sðZF Þ; ð6Þ Consider a multi-storey reinforced concrete wall–frame
dual system similar to that shown in Figure 6, built at a
where E() and s() stand for expected value and soft soil site in the Valley of Mexico, and designed in
standard deviation, respectively. accordance with the 2004 edition of the Mexico City
For the structural system of interest, a sample of Building Code (NTC 2004). In order to estimate its
pairs of random values of Z and the stiffness reduction seismic reliability function, a probabilistic model of the
index, ISSR, can be used to estimate means and mechanical properties and of the gravitational loads
standard deviations of Z(u), the latter defined as the acting on the system was formulated, and a sample of
natural logarithm of the random intensity Y that possible realisations of the vector of those properties
corresponds to ISSR ¼ u. The values of Z in the sample was generated by Monte Carlo simulation, as de-
that correspond to values of ISSR ¼ 1.0 are upper scribed in Esteva et al. (2002b), using the statistical
bounds of ZF ¼ lnYF, where YF is the minimum value properties reported by Meli (1976) and Mirza and
of Y required to produce system failure. Therefore, the McGregor (1979). The model proposed by Wang and
points corresponding to these values of ZF cannot be Shah (1987) was adopted to represent the constitutive
incorporated into the sample used to estimate E(Z(u)) functions describing the bending behaviour of the
and s(Z(u)) by means of conventional minimum- critical sections at the ends of beams, columns and
squares regression analysis. However, they can be walls. For each system in the sample, a nonlinear
included in the estimation process, through a max- analysis was made of its response to a randomly
imum likelihood analysis slightly different from that chosen member of a sample of artificial ground motion
proposed by Esteva and Ismael (2003) and described time histories with statistical properties representative
above. of those expected at the site of interest. Figures 7 and 8
The expected value and the standard deviation of show respectively a typical ground motion time history
Z(u) will now be represented by the functions and a set of pseudo-acceleration response spectra
E(Z(u)ja 1) and s(Z(u)ja 2), with a given form and included in the sample, in all cases for a damping
unknown vectors, a 1 and a 2, of the parameters to be ratio equal to 0.05. The Monte Carlo simulation was
Structure and Infrastructure Engineering 9

Figure 6. Reinforced concrete dual wall–frame multi-storey building structure.


Downloaded By: [UNAM] At: 13:20 16 February 2011

S-wave earthquake magnitude, R the site-to-source


distance (km) and SCT identifies a soft soil site in
Mexico City (Secretarı́a de Comunicaciones y
Transportes). In Figure 8, y/g is the spectral accelera-
tion as a fraction of gravity and T is the natural period
of the structure (seconds). The response results are
presented in Figure 9, where Z ¼ Sdl/uF is a normalised
intensity, Sdl is the ordinate of the linear displacement
response spectrum and uF is the deformation capacity,
Figure 7. Artificial ground motion acceleration time history
at a soft soil site in the Valley of Mexico. determined by means of a conventional static pushover
analysis. An empirical equation of the form E(Z) ¼
a þ b(1–u) þ c(1–u)2 was fitted to the data. For the
variance, the following form was adopted: var (Z) ¼
a1 þ b1(u). The resulting reliability function, deter-
mined by means of Equation (6), appears in Figure 10.
For values of the intensity leading to linear re-
sponse amplitudes of the simplified reference system
approximately equal to the deformation capacity
estimated through a static pushover analysis, b is
approximately equal to 2.0; for Z equal to 0.7, b is
approximately equal to 4.5. Under the assumption
that Z is normally distributed, these values correspond
to failure probabilities of 0.023 and 0.34610–4,
respectively.
The information shown in Figure 9 was also used
to make a rough assessment about efficient criteria for
Figure 8. Pseudo-acceleration response spectra for samples choosing the sampling range of Z, intended to produce
of synthetic ground motion records at a soft soil site in the a large proportion of values of Z slightly smaller than
Valley of Mexico.
1.0, thus leading to more accurate estimates of the
expected value and the standard deviation of ZF, the
performed according to the algorithm proposed by minimum value of the intensity required to produce
Grigoriu et al. (1988). In Figure 7, a is the ground collapse. The results of this analysis are shown in
acceleration, as a function of time t (seconds), M is the Figures 11 and 12. Figure 11 shows all cases with
10 L. Esteva et al.

ISSR 5 1.0, while Figure 12 shows the probability that


ISSR will be contained in the interval 0.8 5 ISSR 5 1.0,
as a function of the normalised width of the
sampling interval (D/ZX), for different values of ZX
(the central value of Z). It can be easily seen that this
probability remains practically constant, about 0.6, for
1.3 5 ZX 5 1.5 and 0.10 5 D/ZX 5 0.8.

3.3. Applications
The concepts described in the previous sections have
been applied to the study of two systems with different
types of lateral-force-resisting structural arrange-
ments: a frame system (Figure 13) and a wall–frame

Figure 9. Values of normalised intensities for given levels of


the stiffness reduction index ISSR.
Downloaded By: [UNAM] At: 13:20 16 February 2011

Figure 12. Probabilities of generating values of ISSR in the


interval (0.8, 1.0], in terms of ZX and D/ZX.
Figure 10. Reliability function for system in Figure 6.

Figure 11. Sample of values of Z and ISSR for the system in


Figure 6. Figure 13. Multi-storey frame building studied in x3.3.
Structure and Infrastructure Engineering 11

system (Figure 14). They are square in-plan 10-storey a non-stationary Gaussian stochastic process with
buildings, regular along their height. They are sup- evolutionary spectral density. Family F2 is simulated
posed to be built at the SCT site in Mexico City (soft in accordance with the procedure described by
soil). They possess the same number of lateral-resisting Alamilla (1997), while Family F3 is that mentioned
frames in the two orthogonal directions. The general in x3.2; it was generated by the algorithm proposed
properties of each system selected are schematically by Grigoriu et al. (1988). The statistical properties of
presented in the mentioned figures. The systems were all simulated records are similar (except for a scaling
designed to comply with the 1993 edition of the factor) to those of the East-West component of the
Mexico City Building Code, taking into account the SCT record of the earthquake of 19 September 1985.
influence of both accidental torsion and soil–structure Pseudo-acceleration response spectra for Families 1
interaction. Some results of the study are discussed in and 2 are shown in Figures 15 and 16, respectively;
the following paragraphs; other results, not presented those corresponding to Family 3 were shown in
here, were reported by Dı́az-López and Esteva (2006). Figure 8.
Two main objectives were established: (a) compar- Figures 17a and b show plots of computed values
ing the values of the reliability index b produced by the of the normalised intensity measures ln(y/g) versus
criterion proposed in xx3.1 and 3.2 with those resulting values of ISSR, for the frame and the dual wall–frame
from the incremental dynamic analysis proposed by system of Figures 13 and 14, respectively. Here, y is
Vamvatsikos and Cornell (2002) and (b) studying the equal to the spectral pseudo-acceleration for the
sensitivity of the results to the criteria used to generate fundamental period of the system and g is the
the samples of recorded and synthetic ground motion acceleration of gravity. Differences ranging from 0.2
Downloaded By: [UNAM] At: 13:20 16 February 2011

time histories used for the Monte Carlo simulation of to 0.25 can be observed between the mean values of
structural response and performance. Variables taken
as random include dead and live loads, cross sections
of structural members and parameters that determine
the stress–strain properties of concrete and reinforcing
steel.
As mentioned by Dı́az and Esteva (2006), three
different families of seismic ground motion records
were selected to represent seismic excitation: one
corresponds to a set of actual acceleration records
obtained at the site where the structures are assumed to
be located (Family F1); the other two are integrated by
synthetic time histories (Families F2 and F3). The
latter were generated by two different simulation
algorithms. Both consider the ground acceleration as

Figure 15. Family F1: response spectra for recorded


ground motion time histories.

Figure 14. Multi-storey dual wall–frame system studied in Figure 16. Family F2: response spectra for synthetic
x3.3. ground motion records.
12 L. Esteva et al.
Downloaded By: [UNAM] At: 13:20 16 February 2011

Figure 17. Relation between normalised intensities and values of the secant-stiffness reduction index (ISSR) for: (a) a frame and
(b) a wall–frame.

ln(y/g) for the ground motion families F3 and F2; this


means a factor ranging from 1.22 to 1.28 for the
median values of y/g. These differences arise from
those associated with the detailed representation of the
evolutionary statistical properties of the instantaneous
variances and frequency contents of the ground
motion accelerations, as assumed by Alamilla (1997)
and Grigoriu et al. (1988) models; the former is based
on a much simpler, but less accurate model of the
evolution of the mentioned variables.
Figure 18 shows that the reliability index b varies in
a much faster way with y/g for the dual wall–frame
system than for the frame system. This may be due to
the fact that, in the former case, the lateral strength
Figure 18. Reliability functions determined with the aid of
and stiffness are concentrated in the wall, while they incremental dynamic analysis (IDA) and the secant-stiffness
are distributed among different structural members in reduction index (ISSR).
the other case. Differences in the values of b for both
families of ground motion time histories are shown to approximately 0.2 to 0.7. This systematic trend is
be small for b, and are approximately equal to 3.0, consistent with the way failure intensities are deter-
but can be greater than 1.0 (in absolute value) for mined according to each approach: while the criterion
higher values of this index. Figure 18 also shows that based on the IDA assumes that failure occurs when
values of b determined with the aid of the index ISSR the slope of the intensity versus lateral distortion
are systematically larger than those determined in curve reaches a sufficiently low (but not zero) pre-
accordance with the IDA approach. Differences are established value, use of the failure criterion based on
not too large, however, they range from a value of ISSR equal to zero corresponds to the limit
Structure and Infrastructure Engineering 13

where the IDA curve would tend for increasing values control parameters for the required values of the
of the intensity. lateral stiffness and strength; z is the viscous damping
The differences observed in Figures 17 and 18 ratio; and g is a net discount rate, which is used to
demonstrate the requirement for systematic studies transform future benefits and costs into their present
oriented to (a) formulating recommendations regard- values. The last term in the second member of
ing criteria for the generation of samples of synthetic Equation (10) must be evaluated in accordance with
ground motion records and (b) developing simplified Equations (8) and (9). Once the optimum values of c
design rules for engineering practice intended to lead to and T are determined, they can be transformed into an
consistent reliability levels for different families of equivalent design criterion, with a conventional format
structural arrangements. expressed in terms of two performance requirements:
ultimate capacity and serviceability limit conditions.
The latter specifies a maximum acceptable distortion
4. Use of seismic vulnerability and risk functions in the
for a moderate intensity seismic excitation; the
establishment of seismic design criteria based on
former is presented in the form of the pseudo-
optimum performance levels
acceleration response spectrum for the ultimate
4.1. Life-cycle risk analysis for the condition of no capacity performance requirement, together with a
damage accumulation base-shear reduction factor intended to account
If the seismic hazard at the site is represented by nY(y), for structural over-strength and nonlinear dynamic
the annual rate of occurrence of earthquake ground response. An illustrative example of this was presented
motions with intensities equal to or higher than y, the by Ismael-Hernández et al. (2004).
Downloaded By: [UNAM] At: 13:20 16 February 2011

seismic risk sustained by the system can be represented


by D, the expected value of the annual cost of damage
and failure consequences, given by Equation (8), in 4.2. Influence of damage accumulation and
terms of the vulnerability functions d(yjS), dF and maintenance strategies
pF(y): Esteva (2006) presents a Markov model for the process
of damage accumulation on a composite structural
D ¼ ðDS þ DF Þ C0 ; ð8Þ system made of a conventional frame and a set of
energy-dissipating devices, taking into account main-
where tenance and repair actions. The structural response of
the system to a given earthquake, and therefore the
Z  
dnY ðyÞ damage increment at each individual member or critical
DS ¼  dðyjSÞð1  pF ðyÞÞdy ð9aÞ
dy  section, is a function of both the characteristics of the
earthquake ground motion and of the residual damage
and throughout the system at the start of that event. Repair
Z   actions on the frame members and preventive-replace-
dnF ðyÞ
DF ¼  dF pF ðyÞdy: ð9bÞ ment measures on the energy dissipation devices
dy  (EDDs) are undertaken whenever empirical evidence
or theoretical assessment lead us to consider that
These equations are based on the assumption that DS current damage levels at individual elements or
and DF are constant in time; that is, they do not depend portions of the system may have reached critical
on the damage that may have accumulated in the acceptance thresholds. In general, damage on the frame
system as a consequence of its response to previous members is apparent in the form of local cracking and/
earthquakes or of other potential damage sources. or crushing, while that on the EDDs may be apparent
They imply the adoption of a maintenance strategy or has to be theoretically inferred on the basis of their
such that all damage is detected and repaired estimated responses to the earthquakes experienced.
immediately after its occurrence. In accordance with the above process, the vulner-
According to Ismael-Hernández et al. (2004), ability function of the system varies in time in a
optimum seismic design criteria should be established random manner. As a consequence, DS and DF
in accordance with an objective function given by the (Equation (9)) are not constant in time, and the
following equation: objective function to be minimised, previously given by
Equation (10), is now expressed as:
Dðc; T; zÞ
Uðc; T; zÞ ¼ C0 ðc; T; zÞ þ ; ð10Þ
g " #
X
1
where T and c are, respectively, the fundamental U ¼ C0 þ E LgT
i
i
; ð11Þ
natural period and the base-shear ratio, taken as i¼1
14 L. Esteva et al.

where C0 and g were as defined above; E[] stands for values corresponding to the end of the jth earthquake,
expected value, Ti, i ¼ 1, . . ., ?, are the (random) it is necessary to calculate the joint probability
times of occurrence of earthquakes that may affect the density function of the waiting time to the (j þ 1)th
system and Li, i ¼ 1, . . ., ?, are the losses associated earthquake and its intensity, and to determine the
with those earthquakes – they include damage and damage states D’fi and D’di of the system’s compo-
failure consequences as well as repair and maintenance nents after carrying out the operations of repairing
actions. The latter are expressed in terms of the the conventional frame members and/or replacing the
threshold values of local accumulated damage for EDDs.
repair or replacement of structural members. An These concepts have been applied by Esteva et al.
analytic formulation for the probabilistic description (1999) to the study of the time-dependent process of
of this process was previously proposed by Esteva and damage accumulation and reliability evolution in
Dı́az (1993). building frames. They were also employed by the
A practical option for the determination of the same authors for the study of optimum design criteria
optimum combination of the values of the design and maintenance strategies for structural frames with
parameters and the repair and maintenance strategies hysteretic energy-dissipating devices. Because of the
consists of the use of Monte Carlo simulations to complexity of the probability transition matrices
determine values of U for several alternative combina- involved, extensive use has been made of Monte Carlo
tions of these parameters and strategies. The following simulations. One of the cases studied corresponds to a
paragraphs summarise an approach proposed by two-bay, 15-storey frame system with hysteretic energy
Esteva (2006) for the time-history description of the dissipating devices. The system was supposed to be
Downloaded By: [UNAM] At: 13:20 16 February 2011

process of damage accumulation and maintenance built at a site in Mexico City where the seismic hazard
actions for a single-storey system with one EDD. was represented by a Poisson process characterised by
Consider the case of a combined single-storey the function relating seismic intensity with annual
system, made by a conventional rigid frame and an exceedance rate. The state of damage at the end of each
EDD. Just after the occurrence of the jth earthquake, earthquake was measured by the maximum value
the damage accumulated on that device equals Ddj, attained at any storey by the index ISSR, defined above.
wile that affecting the structural frame is equal to Dfj. Searching for a life-cycle optimum solution, several
After the (j þ 1)th event, these values become respec- options were explored regarding the seismic design
tively Dd(jþ1) ¼ DDj þ dd(jþ1) and Df(jþ1) ¼ Dfj þ coefficient c and the threshold values of ISSR adopted
df(jþ1), where dd(jþ1) and df(jþ1) are the correspond- as a condition for repair of the main frame and for
ing damage increments. If Dfj exceeds a given thresh- replacement of the energy dissipating devices (Drc and
old, designated here as Drf, the frame is repaired in Drd, respectively). Values of the negative utility
such a manner as to eliminate the damage accumu- function U were obtained for each option, in accor-
lated, thus restoring its initial strength and stiffness, Rf dance with Equations (8) and (9). The resulting values
and Kf. It is assumed that the damage level on the of U, normalised with respect to the initial construc-
frame can be assessed on the basis of the evidence of tion cost for the main frame designed for gravitational
physical deterioration, while that on the EDDs is loads only, are depicted in Figure 19. The first section
inferred from the estimated value of the low-cycle- corresponds to the plain conventional frame, while the
fatigue index. This information is used to implement
the preventive strategy of replacing the EDD after the
occurrence of a number of high-intensity earthquakes,
on the basis of a threshold value Drd.
Whether the process of occurrence of earthquake
ground motions with different characteristics considers
some kind of correlation with previous history or
ignores it, the levels of damage accumulated Dfj and
Ddj, j ¼ 1, . . . , ?, at the end of the jth earthquake
occur as events of a Markov process. The transition
probabilities from (Dfj, Ddj) to (Df(jþ1), Dd(jþ1))
are obtained from the probability density functions
of df(jþ1) and dd(jþ1), which depend on Dfj, Ddj and on
the probability density function of Yjþ1, the intensity
of the (j þ 1)th event.
In order to determine the conditional probability Figure 19. Utility functions for different design base-shear
density functions of Df(jþ1) and Dd(jþ1), given the ratios and maintenance strategies.
Structure and Infrastructure Engineering 15

other three sections contain energy dissipating devices the secant-stiffness reduction index (ISSR). They both
that contribute 75% of the lateral strength and stiffness assume that system collapse takes place when the
of each storey. It can be observed that the negative displacements predicted by the dynamic response
utility function is sensitive to both the seismic design analysis become infinitely large and non-reversible.
coefficient and the repair and replacement strategies. The effective values of the elements of the resulting
The large values that resulted for the initial and long- stiffness matrix are then infinitely small. Both criteria
term costs of the systems with EDDs are probably due permit the estimation of the mean value and the
to the excessively large values assumed for the standard deviation of the failure intensity. They both
contribution of those devices to the lateral strength call for excessively long computer time demands, as
and stiffness of the combined system. This fact may compared with criteria based on the concept of
also be responsible for the high values shown by the deformation capacity. These demands are lower for
optimum threshold damage level for repair of the the approach based on the ISSR; they can be
conventional frame. Because no analysis has been additionally reduced with the aid of an algorithm
made of the sensitivity of these results to the presented to improve its efficiency through a scheme
constitutive functions, the damage–response models, proposed for the determination of the optimum
the repair and replacement costs, and the consequences sampling range for the ground motion intensities to
of ultimate failure, their value is limited to their role be used for its application.
for the purpose of illustrating the application of the Both the IDA and the SSRI criteria were applied
proposed life-cycle optimisation criteria. to the determination of seismic vulnerability func-
Results of a parametric study about expected tions of two multi-storey buildings of similar global
Downloaded By: [UNAM] At: 13:20 16 February 2011

utility functions associated with different maintenance dimensions, but with two different structural systems:
strategies on reinforced concrete frames with a rigid frame and a dual wall–frame system. The
hysteretic EDDs were presented by Campos (2005) results show that the reliability index varies in a
and Esteva (2006). Variables covered include dis- much faster way with the intensity for the dual
tribution of EDDs along the height of the building, wall–frame system than for the frame system; this
deformation capacity of conventional frame, econom- may be due to the fact that, in the former case, the
ic value of collapse and damage threshold values for lateral strength and stiffness are concentrated in the
repair and replacement. wall, while they are distributed among different
structural members in the other case.
Differences in the values of the reliability index for
5. Concluding remarks a given intensity, for two different families of ground
The vulnerability function for a multi-storey building motion time histories depend on the value of that
system has been presented as the expected value of the index. The values determined with the aid of the SSRI
cost of the consequences of the performance attained were systematically larger than those determined in
by the system if subjected to an earthquake of a given accordance with the IDA approach. Differences are
intensity, including the potential events of both not too large, and their trend is consistent with the way
collapse and survival. Its evaluation should account failure intensities are determined according to each
for the physical damage experienced by all structural approach.
and non-structural components of the system, as well The results mentioned above demonstrate the
as of the direct and indirect consequences of that requirement for systematic studies oriented to (a)
damage. It must also account for potential damage formulating recommendations regarding criteria for
experienced by equipment, installations or contents the generation of samples of synthetic ground motion
that may fail, for instance, by overturning or by records and (b) developing simplified design rules for
excessively high dynamic response. engineering practice intended to lead to consistent
Deformation-based criteria for the determination reliability levels for different families of structural
of seismic reliability levels of multi-storey building arrangements.
systems have been shown to be affected by significant
limitations, associated with the concepts of: (a) References
sensitivity of the deformation capacity to the system Alamilla, J., 1997. Simulación de registros en el valle de
configuration at the instant of impending collapse and Me´xico mediante procesos estocásticos modulados en
(b) influence of strength and stiffness degradation amplitud y frecuencia. Thesis (Masters). School of
resulting from the process of damage accumulation. Engineering, National University of Mexico, Mexico.
Alamilla, J. and Esteva, L., 2006. Seismic reliability
Two alternative criteria were examined that do not functions for multistory frame and wall–frame systems.
depend on the use of the concept of deformation Earthquake Engineering and Structural Dynamics, 35 (15),
capacity: the incremental dynamic analysis (IDA) and 1899–1924.
16 L. Esteva et al.

Campos, D., 2005. Diseño sı´smico óptimo de edificios con Esteva, L., Dı́az-López, O., Garcı́a-Pérez, J., Sierra, G., and
disipadores de energı´a. (Optimum seismic design of Ismael, E., 2002c. Life-cycle optimization in the estab-
buildings with energy dissipators). Thesis (PhD). Gradu- lishment of performance-acceptance parameters for
ate Program in Engineering, National University of seismic design. Structural Safety, 24, 187–204.
Mexico, Mexico. Grigoriu, M., Ruiz, S.E. and Rosenblueth, E., 1988.
Cornell, C.A., 1969. A probability-based structural code. Nonstationary models of seismic ground acceleration.
Journal of the American Concrete Institute, 66 (12), Earthquake Spectra, 4 (3), 551–568.
974–985. Ishiyama, Y., 1982. Motions of rigid bodies and
Dı́az-López, O. and Esteva, L., 2006. Seismic reliability criteria for overturning by earthquake excitations.
analysis of complex nonlinear systems using secant- Earthquake Engineering and Structural Dynamics, 10
stiffness-reduction indicators. In: Proceedings of first (5), 635–650.
European conference on earthquake engineering and Ismael-Hernández, E., Dı́az-López, O., and Esteva, L., 2004.
seismology, Geneva, Switzerland, paper 1068. Seismic vulnerability analysis for optimum design of
Dolsek, M. and Fajfar, P., 2004. IN2 – a simple multistory reinforced concrete buildings. In: Proceedings
alternative for IDA. In: Proceedings of 13th world of 13th world conference on earthquake engineering,
conference on earthquake engineering, Vancouver, Vancouver, Canada, paper 514.
Canada, paper 3353. Meli, R., 1976. Bases para los criterios de diseño estructural
Esteva, L., 1992. Nonlinear seismic response of soft-first- del proyecto del Reglamento de Construcciones para el
story buildings subjected to narrow-band accelerograms. distrito federal (Bases for earthquake resistant design
Earthquake Spectra, 8 (3), 373–389. criteria for the draft of Mexico City Building Code).
Esteva, L., 2006. Reliability and optimization as design basis Mexico: Institute of Engineering, National University of
for systems with energy-dissipating devices. In: Keynote Mexico, report 375.
Lecture, Proceedings of 5th international workshop on life- Mirza, S.A. and McGregor, J., 1979. Variation in dimensions
cycle cost analysis and design of civil infrastructure of reinforced concrete members. Journal of Structural
systems, Seoul, Korea. Division, ASCE, 105 (ST4), 751–766.
Downloaded By: [UNAM] At: 13:20 16 February 2011

Esteva, L. and Dı́az, O., 1993. Optimum decisions related to NTC, 2004. Normas te´cnicas complementarias para diseño por
design and replacement of seismic energy dissipators. In: sismo, Reglamento de construcciones para el distrito
Proceedings of 6th international conference on structural federal (Complementary technical norms for earthquake
safety and reliability, ICOSSAR 93, Innsbruck, Austria, resistant design, Mexico City Building Code). Mexico
1, 653–660. City.
Esteva, L. and Dı́az-López, O., 2006. Seismic reliability Park, Y.J. and Ang, H.S., 1984. Mechanistic seismic damage
functions for complex systems based on a secant-stiffness model for reinforced concrete. Journal of Structural
reduction index. In: Proceedings of 13th IFIP WG7.5 Division, ASCE, 110, 722–739.
working conference, Kobe, Japan. Pérez-Martı́nez, R., 2007. Vulnerability analysis of
Esteva, L. and Ismael, E., 2003. A maximum likelihood multistory building frames with masonry infill panels.
approach to system reliability with respect to seismic Research work in the development of Ph D Thesis.
collapse. In: Proceedings of 11th IFIP WG7.5 working Institute of Engineering, National University of Mexico,
conference, Banff, Canada. Mexico.
Esteva, L., Dı́az-López, O. and Garcı́a-Pérez, J., 1999. Reyes, C., 1999. The serviceability limit state in the earth-
Life cycle optimization of structures with seismic quake resistant design of buildings. Thesis (PhD). School
energy-dissipating devices. In: D. Frangopol, ed., Case of Engineering, National University of Mexico, Mexico
studies in optimal design and maintenance planning of civil (in Spanish).
infrastructure systems. American Society of Civil Shome, N. and Cornell, C.A., 1999. Probabilistic seismic
Engineers. demand analysis of nonlinear structures. Stanford,
Esteva, L., Alamilla, J., and Dı́az-López, O., 2002a. Failure California, USA: Department of Civil Engineering,
models, significant variables and reference systems in the Stanford University, report no. RMS.35.
reliability-based seismic design of multistory buildings. Vamvatsikos, D. and Cornell, C.A., 2002. Incremental
In: Proceedings of 7th US national conference on earth- dynamic analysis. Earthquake Engineering and Structural
quake engineering, Boston, MA, USA. Dynamics, 31 (3), 491–514.
Esteva, L., Dı́az-López, O., Garcı́a-Pérez, J., Sierra, G., and Wang, M.L. and Shah, S.P., 1987. Reinforced
Ismael, E., 2002b. Simplified reference systems in the concrete hysteresis model based on the damage concept.
establishment of displacement-based seismic design Earthquake Engineering and Structural Dynamics, 15,
criteria. In: Proceedings of 12th European conference on 993–1003.
earthquake engineering, London, UK, paper 419.

You might also like