Performance-Based Seismic Engineering: The Need For A Reliable Conceptual Comprehensive Approach

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS
Earthquake Engng Struct. Dyn. 2002; 31:627–652 (DOI: 10.1002/eqe.146)
Performance-based seismic engineering: the need for a reliable

conceptual comprehensive approach
Raul D. Bertero1; ∗; † and Vitelmo V. Bertero2

1 Departamento de Estabilidad; Facultad de Ingeniera; Universidad de BuenosAires; Larrea 1142;
Buenos Aires 1117; Argentina
2 Earthquake Engineering Research Center; University of California at Berkeley; 1301 South 46th Street;
Richmond; CA 94804, U.S.A.
SUMMARY
The main objectives of this paper are: (i) to review brie;y as to what is understood by performance-
based engineering, performance-based seismic engineering, and performance-based seismic design
(PBSD); (ii) to discuss the main requirements that a reliable PBSD should satisfy and why some
approaches that have been proposed fail in satisfying those requirements; (iii) to focus on what is
considered a promising probabilistic PBSD approach; and (iv) to present a conceptual comprehensive
numerical procedure for the PBSD of buildings.
It is shown that to satisfy the objectives of a reliable PBSD philosophy and procedure it is necessary to
start with a multi-level seismic design criteria, to consider a probabilistic design approach, to consider
local structural and non-structural damage and therefore design spectra for buildings (n degrees of
freedom), to take into account the cumulative damage, and to control not only displacements but also
ductility (minimum strength) to limit damage. Finally, it is concluded that a transparent and conceptual
comprehensive preliminary design approach is necessary. Copyright ? 2001 John Wiley & Sons, Ltd.
KEY WORDS: performance-based seismic engineering; performance-based seismic design; conceptual

design; comprehensive design; probabilistic design; performance-based design;
performance-based engineering
1. INTRODUCTION
A review of the lessons learned from signiCcant earthquakes that have occurred recently has
led to the conclusion that the seismic risk in urban areas is increasing rather that decreasing.
One of the most eDective ways to reverse this situation in future signiCcant earthquakes is
through (i) the development of more reliable seismic standards and code provisions than
∗ Correspondence to: Raul D. Bertero, Departamento de Estabilidad, Facultad de IngenierEFa, Universidad de Buenos
Aires, Larrea 1142, Buenos Aires 1117, Argentina.
† E-mail: rbertero@freyreyasoc.com.ar
Received 15 January 2001

Revised 31 August 2001
Copyright ? 2001 John Wiley & Sons, Ltd. Accepted 31 August 2001
628 R. D. BERTERO AND V. V. BERTERO
those currently available, and (ii) their stringent implementation for the complete engineering
of new civil engineering facilities and also for the evaluation of the seismic vulnerability and
upgrading of existing hazardous facilities. A comprehensive approach for development and
implementation of the next generation of standards and codes must include consideration of
all the main aspects involved in the engineering of facilities to resist the eDects of earthquakes.
These main aspects involve not only those concerning the conceptual and numerical designs
but also the proper structural detailing and construction of such conceptual and numerical
designs, and the monitoring of the occupancy (function) and maintenance of the whole (entire)
facility (or system).
To facilitate the stringent implementation of the code or standards for the complete engineer-
ing of new or existing facilities, the strategies and techniques included in these codes and=or
standards should be simple enough so that they can be applied eDectively according to the
education (knowledge) of the professionals involved (designers, public oKcers, contractors,
inspectors, etc.) as well as the owners. However, the simplicity should not compromise the
reliability of such a design. To accomplish these goals, the following two alternatives could
be considered: (a) to improve gradually the reliability of the simpliCed procedure required by
the code and=or standards according to the improvement in the education of the professionals
involved in implementing such simpliCed procedures; and (b) to introduce radical changes
in the present code and/or standards by introducing the most reliable procedure that can be
developed according to the state of the art in seismic engineering.
The above two alternatives are not completely independent, because even if the Crst alter-
native is adopted, the reliability of the simpliCed procedure should be determined. Thus, it is
obvious that for both of these alternatives there is a need to develop what can be considered
the most transparent and reliable procedure or methodology.
A promising approach for such needed development is that which has been proposed as
performance-based engineering (PBE); its application to the case that seismic hazards are
signiCcant is commonly known as performance-based seismic engineering (PBSE) [1; 15].
An important phase of PBSE is the performance-based seismic design (PBSD).
2. OBJECTIVES OF THIS PAPER
Recognizing the fact that for reliable application of PBSE in practice the code regulations
must remain simple but reliable and in accordance with the education in Earthquake (EQ)
Engineering of the practitioners, the following three-step approach could be implemented for
the Cnal formulation of the simple seismic code regulations.
• First step: Based on the state of the art in EQ Engineering, a ‘conceptual performance-
based code’ should be developed, covering all aspects that a seismic code should regulate.
• Second step: In this step, the conceptual code regulations developed in the Crst step
will be applied to the design of building facilities with diDerent regular and irregular
conCgurations, structural layouts and structural systems, which preferably are similar to
those that have been designed and constructed according to current or modern seismic
codes, and whose responses to earthquake ground motions (EQGMs) have been either
recorded or predicted; and to the upgrading of diDerent types of existing hazardous
facilities which preferably are similar to those that have been recently upgraded.
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 31:627–652
PERFORMANCE-BASED SEISMIC ENGINEERING 629
• Third step: From analysis of the results obtained in the second step, a simpliCed concep-
tual performance-based code that can be applied properly by the practitioners should be
developed. It should clearly state all restrictions in sitting and in the selection of conCg-
uration (or form), foundation systems and structural systems for which such simpliCed
code regulations could be used. For complex buildings, a peer review process should be
required in which the conceptual code to be developed in the Crst step could be used.
The main goal of this paper is to cover the Crst step of a rational programme to develop
simple but reliable performance-based seismic codes. In particular, the main objectives of
this paper are: (i) to review brie;y as to what is understood by PBE, PBSE and PBSD;
(ii) to discuss the main requirements that a reliable PBSD should satisfy and why some
approaches that have been proposed fail in satisfying those requirements; (iii) to focus on
what is considered a promising probabilistic PBSD approach; and (iv) to present a conceptual
comprehensive numerical procedure for the PBSD of buildings.
3. DEFINITIONS OF PBE, PBSE AND PBSD
PBE is deCned as consisting of the selection of design criteria, appropriate structural systems,
layout, proportioning, and detailing for a structure and its non-structural components and
contents, and the assurance and control of construction quality and long-term maintenance,
such that at speciCed levels of all the excitations (that can act on the building) and with
deCned levels of reliability, the building or facility will not be damaged beyond certain limit
states. PBE is a process that begins with the Crst concepts of a project and lasts throughout
the life of the building.
PBSE is deCned as the application of PBE to the case that seismic hazard controls the
design. Therefore, PBSE involves the complete design, construction and control (monitoring)
of the maintenance and function of the building to assure that the constructed buildings will
resist the eDects of earthquake ground motions of diDerent severities within speciCed limiting
levels of damage.
PBSD is the subset of activities of PBSE that focus on the design process. Therefore, it
includes identiCcation of seismic hazards, selection of the performance levels and performance
design objectives, determination of site suitability, conceptual design, numerical preliminary
design, Cnal design, acceptability checks during design, design review, speciCcation of quality
assurance during the construction and of monitoring of the maintenance and occupancy (func-
tion) during the life of the building. In this context, the term ‘design’ applies to the whole
building system, including the foundation, non-structural components, contents, equipment and
the utility lines serving the facility.
4. PERFORMANCE OBJECTIVES (POs)
A conceptual framework for PBSE has been developed [1] encompassing the full range of
seismic engineering issues to be addressed in the design of structures for predictable and
controlled seismic performance within established levels of risk. The Crst step is the selection
of the performance design objectives (POs). These objectives are selected and expressed
EARTHQUAKE PERFORMANCE LEVEL

Fully Operational Operational Life Safety Near Collapse
EARTHQUAKE DESIGN LEVEL

Frequent
(43 years)
Unacceptable
Performance
Occasional
BA (for new construction)
SI
(72 years) C
ES OB
SE JE
NT CT
IA IV
SA L E
Rare FE OB
TY JE
(475 years)
CR CT
IV
IT E
IC
AL
OB
Very Rare JE
(970 years) CT
IV
E
Figure 1. Recommended minimum seismic performance design objectives for buildings [1].
in terms of expected levels of damage resulting from expected levels of earthquake ground
motions [1]. The client makes this selection in consultation with the design professional based
on consideration of the client’s expectations, the seismic hazard exposure, economic analysis
and acceptable risk. POs will range from code minimum requirements (usually based on fully
operational under minor earthquake ground motions and on life safety under a rare earthquake
ground motion) to fully operational in a maximum credible (capable or considered) earthquake
ground motion.
A PO is a coupling of expected performance levels with levels of seismic ground motions.
A performance level represents a distinct band in the spectrum of damage to the structural and
non-structural components and contents, and also considers the consequences of the damage
to the occupants and functions of the facility. Four discrete performance levels are identiCed
in the report [1] (Figure 1), which gives tables that deCne them in terms of the various
components of the building. The seismic hazard at a given site is represented as a set of
earthquake ground motions and associated hazards with speciCed probabilities of occurrence.
For example, the term ‘rare earthquake’ refers to a set of potential earthquake ground mo-
tions that can produce a deCned level of damage with a speciCc mean annual frequency
(e.g. 475 years return period for standard buildings).
The set of earthquake ground motions will vary not only for diDerent seismic regions but
also from site to site within a region because of variations in site conditions (topography
and soil proCle). The four minimum levels of earthquake ground motions corresponding to
the four discrete performance levels identiCed in Figure 1 have been arbitrarily selected for
a California area. Note that the return period, TR , and the exceeding probability in N years,
pN , are two diDerent ways of expressing the same concept and they are related by
1
p1 = ; pN = 1 − (1 − p1 )N (1)
TR
where p1 is the annual probability of exceedance. The validity of these equations assumes
that earthquake occurrences are independent events. This is not strictly correct, but the above
relations are widely used to simplify the discussion of probability.
Performance objectives (POs) typically include multiple goals for the performance of the
constructed building: for example that it be fully operational in the 43-year-event, it oDers
life safety in the 475-year-event, and it will not collapse in the 970-year-event. The selec-
tion of POs sets the acceptability criteria for the design. Design criteria are the rules and
guidelines which must be met to ensure that the usual three major objectives of the design
(i.e. performance of function, safety, economy) are satisCed. The performance levels are
keyed to limiting values of measurable structural response parameters, such as drift and
ductility (monotonic and cumulative), structural damage indexes (DM), storey drift indexes
(IDI), and rate of deformations such as ;oor velocity, acceleration and even the jerk (in
the case of frequent minor earthquake ground motions). When the performance levels are
selected, the associated limiting values become the acceptability criteria to be veriCed in
later stages of the design. Note that once the limit value of the parameter has been se-
lected for a particular earthquake hazard level to completely deCne the design criteria, it
is still necessary to deCne the acceptable conditional probability of going beyond that limit
state (failure probability). An example of the quantiCcation of the POs to control structural,
non-structural and contents damage is shown in Table I. Limiting values of the response
parameters that correlate with the deCned performance levels must be established through
research.
The POs can also be represented as a curve in the ‘performance space’ [2] as shown in
Figure 2(a) for the non-structural damage as deCned in Table I. On the vertical axis, the
probability of exceedance of a particular performance objective over the life-cycle of the
facility (in this case 50 years) is shown on a logarithmic scale. This probability accounts
for both the likelihood of earthquake ground motions occurrences (according to its recurrence
interval) as well as the acceptable probability of failure when these earthquake ground motions
do occur. On the horizontal axis, a measure of structural response is shown (in this case,
the usually so-called interstorey drift index (IDI) when actually a more proper term should
be storey drift index). The POs in Figure 1 are in fact the discrete representations of the
performance objective curve in Figure 2(a).
5. MAIN REQUIREMENTS FOR A RELIABLE PBSD
5.1. The need for multi-level seismic design criteria
During the 1971 San Fernando earthquake (moment magnitude MW = 6:7), the Olive View
Medical Center in Sylmar (California), inaugurated just a few months before, was almost
completely destroyed. The hospital was redesigned in 1976 and rebuilt using RC and steel
shear walls around the perimeter. During the 1994 Northridge earthquake (moment magnitude
MW = 6:7) the new facility resisted ;oor accelerations of 2:8g without any signiCcant struc-
tural damage. However, the hospital had to evacuate all of its patients to other facilities for
several days because of breakage of both sprinkler and chilled water lines [3]. This example
shows very clearly that the design for a life safety performance objective for rare earthquake
ground motions does not necessarily imply that the facility will be operational after frequent
earthquake ground motions.
The situation can be shown in the ‘performance space’ as shown in Figure 2(b), where
typical performance curves of two designs are drawn. Design A, which has been done based
632
Table I. Example of quantiCcation of the performance objectives (POs).
Copyright ? 2001 John Wiley & Sons, Ltd.

EQ performance EQ design level Structural damage Non-structural damage (1) Contents damage (1)
level Return period
(years) Local DM Cond. fail. IDI Cond. fail. Floor accel. Cond. fail.
index prob. (2) prob. (2) prob. (2)
Fully operational 43 0.20 40% 0.003 40% 0:6g 40%
Operational 75 0.40 30% 0.006 30% 0:9g 30%
Life safety 475 0.60 25% 0.015 25% 1:2g 25%
Near collapse 970 0.80 20% 0.020 20% 1:5g 20%
(1) For control of non-structural and contents damage, it may be necessary to limit a combination of IDI, ;oor velocity and ;oor acceleration (and even
jerk for frequent EQGMs).
R. D. BERTERO AND V. V. BERTERO
(2) Conditional probabilities of limit state exceedence given earthquake ground motions with the considered return period occurs.
Earthquake Engng Struct. Dyn. 2002; 31:627–652

Probability of Limit State Exceedence

in 50 years
1
Fully
Operational
Exceedence Probability
Non-acceptable
Operational Design Zone
Performance Objective
0.1 Curve
Acceptable
Design Zone Life Safety
Near
Collapse
0.01
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02
(a) Interstory Drift Index
Probability of Limit State Exceedence

in 50 years
1
Fully Design A
Operational Performance
Exceedence Probability
Non-acceptable
Operational Design Zone
Design B Performance Objective

0.1 Curve
Performance
Acceptable
Design Zone Life Safety
Near
Collapse
0.01
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02
(b) Interstory Drift Index
Figure 2. (a) Performance objectives represented in the ‘performance space’ [2]. (b) Two diDerent
designs in the ‘performance space’ [2].
only on the life safety performance objective, does not satisfy the operational and fully op-
erational performance objectives. On the other hand, Design B, which from the beginning
considered fully operational and life safety performance objectives, is inside the acceptable
design zone.
If the level of direct damage losses during recent earthquakes [Loma Prieta 1989 ($7 bil-
lion); Northridge 1994 ($30 billion); Kobe 1995 ($200 billion)] is considered, it is clear
that such large losses in these relatively frequent moderate events are unacceptable and
1.6
1.4 TR = 72 years ξ = 2 % µ = 1
Seismic Yielding Coefficient, Cy

Serviceable a
1.2
1.0 TR = 20 years ξ = 2 % µ = 1
Serviceable b
0.8
0.6 TR 500 years 5% =3

Life Safety
0.4
0.2
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Period, T(s)
Figure 3. Strength spectra for serviceability and life safety.
that the current code approach of looking only for the life safety performance is not
satisfactory.
Furthermore, it is very important that the designer recognizes, from the very beginning, the
implications of the selected performance objectives. For example, the strength design spectra
for Crm soil at San Francisco (U.S.A.) (Figure 3) have been obtained using the Newmark–
Hall Design Spectra [4], with peak horizontal acceleration according to 20, 72 and 500 years
of return period [5], and the strength-reduction factors proposed by Miranda [6]. It is clear
from Figure 3 that if a serviceability limit state is required for a 72-year-return period (as it
has been suggested for some projects located in California sites with high seismic risks) or
even for 43 or 20 years of return period, not only the serviceability limit state will control
the design but also innovative design approaches should be used to economically satisfy the
performance objectives.
From the examples shown in Figures 2(b) and 3, it is very clear that at least two per-
formance levels should be considered (even for the preliminary design) so that appropriate
design decisions could be made to satisfy the performance objectives. Therefore, the problem
is to deCne as to which two performance levels should be considered. At present, the fully
operational (or serviceability) and life safety levels are the performance levels that should
be selected for design since they are the ones for which there exist data and experience.
Another advantage is that the serviceability limit state can be satisCed by using the very
well-known techniques of elastic analysis. Subsequently, response in the non-linear range
(i.e. the life safety limit state) can be estimated using plastic analysis. Thus, the two dif-
ferent stages of behaviour (linear and non-linear) can be controlled using well-established
procedures.
Currently available information to quantify the requirements for all performance levels is
not reliable. There is an urgent need for an extensive and co-ordinated research programme
to improve the probabilistic deCnition of structural and non-structural damage indexes. Partic-
ularly, there is a need for reliable data (and therefore studies) regarding the interacting eDects
of ;oor acceleration and storey drift index on non-structural components and pipeline damage,
as well as the level of the content damage associated with the ;oor velocities and=or accelera-
tions. There is an urgent need to investigate the reliability of the recommended quantiCcation
of the damage (limit states) for the non-structural components at the diDerent POs, as well
as of the existing guidelines, rules, testing, and requirements regarding the mechanical char-
acteristics and capacities of these components as well as of their quality control, installation
and maintenance.
5.2. The need for a probabilistic design approach
A probabilistic foundation for the development of performance-based guidelines has been

recently proposed [7]. The suggested foundation assumes that the basis for assessing the ad-
equacy of the structure will be a vector of certain key ‘Decision Variables’, DV, such as the
exceedence of one limit state (e.g. collapse). These can only be predicted probabilistically.
Therefore, the speciCc objectives of engineering assessment analysis are in eDect quanti-
ties such as the mean annual frequency of collapse
. The proposed strategy [7] involves
the expansion of the mean annual frequency in terms of the structural ‘Damage Measures’,
DM (e.g. storey drift index) and ground motion ‘Intensity Measures’, IM (e.g. peak ground
acceleration—PGA), which can be written symbolically as

(DV) = G(DV=DM) dG(DM=IM) d
(IM) (2)
Here, G(DV=DM) is the probability that the decision variable exceed speciCed values
given that the engineering damage measures (e.g. the maximum storey drift index) are equal
to particular values (‘fragility curves’). G(DM=IM) is the probability that the damage mea-
sure exceed these values given that the intensity measure (e.g. PGA) equals particular values.
(IM) is the mean annual frequency of the intensity measure. The following simpliCed ap-
proach is based on the fact that due to the large uncertainties in the demand, the in;uence of
the Crst term in the integral of Equation (2) could be neglected. Therefore, it is assumed that
(DV) =
(DM) (e.g. the mean annual frequency of the storey drift index) can be computed
by integrating on all the earthquakes that might occur at the site. For example, for each limit
state,
(DM) are the inverses of the return periods deCned in Table I. Once the mean annual
frequency or return period and the conditional failure probability of Table I are selected, the
designer could develop the design using a traditional load and resistance factor design (LRFD)
format [8]. An LRFD format that can be used for strength-, displacement- or damage-based
design is developed as explained below.
In order to obtain a guide to develop this probabilistic design approach, it is necessary to
review the uncertainties of the diDerent variables involved in the earthquake design problem.
Means, coeKcient of variations (COVs), and probability distributions for structural resistances
have been determined from test data on the strength of materials and on dimensions of struc-
tural members, from laboratory tests of full-scale members under idealized load conditions
and, in some cases, where a clearly deCned analytical model exists, through Monte Carlo
simulation. A representative sampling of these data, which summarizes results of numerous
research programmes, can be found in Reference [8]. The COVs on resistance of structural
steel and reinforced concrete members have values between 0.11 and 0.19, depending on
material and failure mode. Much less data exist regarding the COVs on displacement ca-
pacity of structural members. Experimental data usually show larger scatter in the ultimate
displacement than in the ultimate resistance and therefore a COV of slightly larger than
0.2 could be expected in general for the displacement and local rotation (i.e. members) ca-
pacities. On the other hand, seismic hazard studies [9] indicate that the COV of the PGA
is site dependent and can range from 0.56 to 1.38. The COV of the demand depends not
only on the uncertainty in PGA, but also on the uncertainties in the dynamic characteris-
tics of the earthquake ground motion time histories and of the structure. For example, the
COV of strength demands normalized by PGA for ground motions records have been com-
puted by Miranda [6]. Although the values obtained by Miranda are period dependent, a
COV of about 0.60 could be used for a wide range of periods. Combining the 0.56 COV
in PGA with the 0.60 COV in the strength demand for a given PGA, a COV of about
0.80 is obtained for the seismic strength demand. Therefore, a COV = 0:80 could be used
to represent the scatter in the earthquake strength, plastic hinge rotation, IDI and damage
demand.
Since a COV of about 0.20 could be expected for the capacities and a COV of about 0.80
for the earthquake demand, a simple probabilistic approach could be used for design. This
simple approach is based on the fact that due to the dominant uncertainties in the demand, it
is possible to consider all the random variables as deterministic (and equal to the mean value)
in the design with the exception of the earthquake demand. Therefore, the design equation
for the mean capacity of each parameter X , Cx (where X is any design target parameter like
yield strength Cy , storey drift index IDI, damage index DM, etc.) could be reduced to a load
factor design using [10]:
Cx ¿Dx (1 + COVDx )= Dx + Dx (3)
where Dx is the mean demand for the design parameter X , Dx = Dx COVDx , the standard
deviation of the demand for the design parameter X , and is a parameter used to measure
the target failure probability, Pf , so that
Pf = W(−) (4)
where W( ) is the cumulative standard normal distribution. For example, for Pf = 0:20, =
0:84. Note that normal distribution functions are implicit in using Equations (3) and (4).
If necessary, other distribution functions could be adopted following the same approach but
using equivalent normal variables obtained with the Rosenblatt transformation [11].
Therefore, if a conditional failure probability of 0.20 is speciCed for one PO and a COV of
0.80 is assumed for the earthquake demand parameter under study, the mean value (not the
nominal or speciCed value) of the capacity parameter should be larger than (1+ COVDx ) = 1+
0:84 × 0:80 times the mean demand. If a nominal value of the capacity is used, the demand
ampliCcation factor should be reduced according to the ratio between mean and nominal
capacity. Note that = 0:84 (Pf = 0:20) is very close to the usual rule for designing with
(mean +) spectra. However, no reduction factor is necessary with respect to the mean value
of the capacity, and the standard deviation, Dx , must include all the demand uncertainties
(such as earthquake magnitude, type of fault, focus, directional eDects, topography, soil proCle
and structural response).
When a multi-level design criterion is adopted, it is important to consider not only the
expected levels of earthquake ground motions but also the levels of other types of seismic
hazards as well as the other possible types of excitations [such as those due to gravity (includ-
ing snow), wind, temperature, etc.]. Furthermore, it is necessary to have a reliable estimation
of all the possible critical combinations among these diDerent excitations at each of the dif-
ferent levels of severity that are considered. The same performance-based approach used for
the earthquake excitations (forces) could be used for each diDerent source of excitation. How-
ever, since the experience with other excitations using the current code approach seems to
be satisfactory, it is reasonable to continue using it. For forces such as those produced by
wind and snow, the ‘design load level’ is typically deCned on the basis of a mean recurrence
interval of 50 years, and =3:0 (i.e. Pf = 1:35 × 10−3 ) is the target conditional probability
of exceeding the local ultimate strength given the occurrence of that design load level. In
addition, a set of load and force combinations and of their factors were recommended for use
by the individual material speciCcation writers [8].
The problem is to deCne the load combination factors for use with each of the earthquake
performance objectives previously deCned. The load combinations are needed not only for
strength-based design but also for displacement-based and damage-based designs. This is so
because gravity loads can aDect the storey drift index and plastic hinge rotations because of
the P–X eDect and the formation of collapse mechanisms with plastic hinges in the central
regions of beams. The gravity load factors can be easily selected if we apply the Turkstra rule
for modelling the load combination [12]. The rule is based on the observation that when one
excitation component reaches an extreme value (in this case, the earthquake excitations), the
other excitation components are often acting at their instantaneous or arbitrary-point-in-time
(apt) values. Since the mean values of arbitrary-point-in-time loads are the same for all the
earthquake levels, and if it is considered that their COVs are much smaller than the COV
of the earthquake demand, the load combination for each performance objective i could be
deCned with an LFRD format as:
Dapt Lapt Sapt
(1 + i COVEQi )EQ + Dn + Ln + Sn (5)
Dn Ln Sn
Statistical data on loads [13] have shown for dead load, Dapt = 1:05Dn , and for live load,
Lapt = 0:24Ln . The temporal characteristics of snow forces vary widely with local clima-
tology. From a study of snow and earthquake excitation combinations by Ellingwood and
Rosowsky [14], Sapt = 0:20Sn , could be used. Therefore, considering for all performance ob-
jectives COVEQ = 0:80 and, as an example, the conditional failure probabilities of Table I, the
following excitation factors could be used:
Frequent EQGMs; 1:20EQfreq + 1:05Dn + 0:24Ln + 0:20Sn
Occasional EQGMs; 1:40EQoccas + 1:05Dn + 0:24Ln + 0:20Sn
(6)
Rare EQGMs; 1:50EQrare + 1:05Dn + 0:24Ln + 0:20Sn
Very rare EQGMs; 1:70EQv:rare + 1:05Dn + 0:24Ln + 0:20Sn
Note that the earthquake demand deCned by Formulas (6) should be compared with the
mean capacity without reduction factors (since all the uncertainty has been allocated in the
demand).
Formulas (6) are valid for strength, displacement or damage-based design. Since signiCcant
non-linear behaviour is expected for performance levels beyond the fully operational level,
the combination Formulas (6) must not be interpreted as a linear combination of load eDects.
Formulas (6) indicate the gravity loads that should be considered acting simultaneously with
the EQGM. Owing to the P–X eDects, these gravity loads aDect the IDI spectra of the building
presented in Section 5.3. Local plastic hinge rotation and local damage are also aDected by
the level of the gravity loads acting on beams, especially on long beams of top storeys [15].
Note that these eDects cannot be properly considered using single-degree-of-freedom (SDOF)
systems-based spectra.
5.3. The need for design spectra for buildings (n degree of freedoms)
Most of the current seismic design methodologies are based on design spectra for SDOF
systems (i.e. based on global parameters). However, since the performance levels are keyed
to limiting values of local structural response parameters, such as local structural damage
indexes (DM), storey drift indexes (IDI) and ;oor accelerations, local design spectra of
such parameters should be available for the preliminary design. It is obvious that several
assumptions must be made to build these spectra for buildings, but in any case these spectra
and the knowledge of these assumptions will provide a better design guide than the present
SDOF spectra. The development of design spectra for buildings is explained in the following
sections.
5.3.1. Development of elastic and inelastic IDI spectra. Let us Crst consider that the building
can be modelled like a one-dimensional linear elastic system with distributed mass
(Figure 4(a)). The formulation of the earthquake response for systems having continuously
distributed parameters can be carried out by procedures that are completely analogous to those
of lumped-mass systems [16]. Using modal analysis, the decoupled normal-co-ordinate equa-
tions of motion take the same form as that for the lumped-mass system and they may be
expressed as
Li
YYi (t) + 2i !i Ẏi (t) + !i2 Yi (t) = − uYg (t) (7)
Mi
where Yi is the response amplitude of the mode i, !i the frequency of mode i, i the
viscous damping of mode i, uYg (t) the earthquake ground motion acceleration, Mi the
generalized mass and Li the modal earthquake excitation factor. Mi and Li are asso-
ciated with the distributed mass m(x) and the shape of the i mode ’i (x) as shown
below:
H H
Li = m(x)’i (x) dx; Mi = m(x)[’i (x)]2 dx (8)
0 0
The displacement relative to the base, w(x; t), and the distortion, w (x; t), at each storey can
be computed as

n 9w(x; t) n
w(x; t) = ’i (x)Yi (t); w (x; t) = = ’i (x)Yi (t) (9)
i=1 9x i=1
n
w( x, t ) = ∑ ϕ i ( x)Yi (t )
m( x ) w( x, t ) i =1
∂w( x, t ) n
IDI ( x, t ) = w ( x, t ) = = ∑ ϕ i ( x)Yi (t )
x ∂x i =1
IDI ( x , t )
(a)
Elastic IDI (Los Gatos EQGM) Elastic IDI (SCT EQGM)

ξ =5% ξ =5%
0.05 0.05
H=30 m H=90 m H=30 m
0.04 H=60 m 0.04
Story Drift Index, IDI

H=60 m
0.03 H=120 m 0.03

H=90 m
0.02 0.02
H=120 m
0.01 0.01
0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0 0.5 1 1.5 2 2.5 3 3.5 4
(b) Period, T (s) Period, T (s)
Inelastic IDI (Los Gatos EQGM) Inelastic IDI (SCT EQGM)

ξ =5% ξ =5%
0.06 0.06
0.05 µ = 4.0 0.05

µ = 3.0 µ = 4.0
0.04 0.04
µ = 2.0 µ = 3.0
0.03 µ = 1.0 0.03
µ = 2.0
0.02 0.02
µ = 1.0
T = 0.0731H 3/ 4
H [ m]
0.01 0.01
T = 0.0731H 3 / 4 H [ m]
0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0 50 100 150 200
(c) Period, T (s) Building Height, H (m)
Figure 4. (a) Building modelled like a one-dimensional linear elastic system with distributed mass. (b)
Maximum elastic IDI for buildings of several heights as a function of the Crst mode period T for
Los Gatos (Loma Prieta, 1989) and SCT (Mexico, 1985) EQGMs. (c) Maximum non-linear IDI for
Los Gatos (Loma Prieta, 1989) and SCT (Mexico, 1985) EQGMs.
If it is assumed that the building can be modelled like a uniform shear beam, the modal
parameters are as follows:

2i − 1 x
i Ti 1
’i (x) = (−1) sin ; =
2 H T1 2i − 1
H
2 mH
Li = m’i (x) dx = (−1)i
0 2i − 1
H
H
Mi = m[’i (x)]2 dx = m (10)
0 2
In Equations (10), m is the mass per unit height of the building (assumed as a constant),
H the height of the building and Ti the period of mode i. Since w (x; t) can be considered an
estimate of the IDI at diDerent levels during the response of the structure, the maximum IDI
can be computed as
2
n
Li
max |w (x; t)| = ’i (x) Sd (Ti ; i ) (11)
i=1 Mi
where Sd (Ti ; i ) is the spectral relative displacement of an SDOF system. If we assume that
the maximum IDI occurs at the base of the structure (as is expected for a shear beam) and
considering that for all the modes

Li 2i − 1 ((2=(2i − 1))(mH=)) 2
’i (0) = = (12)
Mi 2 H mH=2 H
the maximum elastic IDI can then be computed as

2 n
IDI = Sd (Ti ; i )2 (13)
H i=1
Figure 4(b) shows the maximum elastic IDI for buildings with diDerent heights as a function
of the Crst mode period T using Equation (13) for Los Gatos (Loma Prieta, 1989) and SCT
(Mexico, 1985) earthquake ground motions as representative of severe pulse and periodic type
of earthquake ground motion, respectively.
If the structure response is non-linear, an ampliCcation of the IDI occurs because of the
concentration of damage in one of the storeys. This ampliCcation is larger as the structure
response goes deeper in the non-linear range (i.e. as the displacement global ductility in-
creases). Since there are very few systematic studies of this problem, comprehensive studies
of the local concentration of damage during the non-linear response in real and idealized
buildings are urgently needed. Until these results are available, the empirical factor 0:414
recommended by Hwang and Jaw [17] for estimating the ampliCcation of maximum inter-
storey displacement due to the concentration of plastic rotations at one storey, could be used.
Therefore, the maximum inelastic IDI can be estimated with

2
n
IDI = IDI 0:414 = Sd (Ti ; i )2 0:414 (14)
H i=1
Equation (14) should be used very carefully, since the ampliCcation factor has a large coef-
Ccient of variation and is very sensitive to small changes of storey capacities (the probabilistic
analysis should include this uncertainty). Larger values should be used when a weak storey
cannot be avoided. On the other hand, smaller values are possible if a careful conceptual
performance-based design is followed.
Figure 4(c) shows the maximum inelastic IDI for = 1:0, 2.0, 3.0 and 4.0 as a function
of the height of typical RC frame buildings using Equation (14) for Los Gatos and SCT
earthquake ground motions. For each period T , the height of the building was computed
using the NEHRP [18] code formula for RC frames (T = CT H 3=4 , where CT = 0:0731 and H
is in metres). Note how these spectra show very clearly that to control the IDI for earthquake
ground motions like Los Gatos and SCT, a stiDer structure than usual should be used. Also,
to avoid intolerable concentration of IDI, a small global ductility must be considered.
5.3.2. Development of local DM spectra. To satisfy the deCnition of PBSE, there is a need
to numerically compute diDerent levels of structural damage for speciCc levels of earthquake
ground motions. Structural damage during an earthquake is due to excessive deformations
and accumulated damage sustained under inelastic deformation reversals. The earliest and
simplest measures of damage were ductility-based, and so failed to take any account of the
possible cumulative eDect of repeated cycles of deformation [19]. More recently, a number
of researchers have proposed damage indexes that take into account the cumulative eDects
by including the hysteretic energy. Although it cannot accurately reproduce all possible load–
deformation paths and damage mechanisms, the best known and most widely used of all the
cumulative damage indexes is that of Park and Ang [20]. This consists of a simple combination
of normalized deformation and plastic energy dissipation. The Park and Ang index, modiCed
to obtain DM = 0 before yielding, can be written for an SDOF system as
! − !y EH ! − (!= ) !
DM = +b = + b% 2 (15)
!umon − !y Fy !umon !umon − (!= ) !umon
where ! is the maximum displacement during the earthquake ground motion, !y the yield
displacement, Fy the yield strength of the system, !umon the ultimate value of displacement
at impending collapse under monotonically increasing excitation, EH the hysteretic energy
dissipated by plastic deformation, the maximum displacement ductility during the earthquake
ground motion, b a parameter controlling strength deterioration as a function of the amount
of dissipated energy by plastic deformation that depends on the designer’s decision about the
type of connection used in the beam–column joints, detailing, level of axial load and shear
at critical regions (plastic hinges), and aspect ratio of members, and the parameter % deCned
by Fajfar [21] as

EH EH =m
%= = (16)
k! 2 !!
where k is the structural stiDness, m the structural mass, and ! the circular frequency of the
structure.
It is possible to express the damage index in terms of the average rotation instead of
displacement. The resulting expression is
' − 'y EH ' − ('= ) '

DM = +b = + b% 2 (17)
'umon − 'y My 'umon 'umon − ('= ) 'umon
Note that the only changes with respect to Equation (15) are that the maximum structural
displacement, !, was replaced by ' the maximum average rotation, the ultimate displacement
!umon was replaced by 'umon the ultimate rotation under monotonic deformation, and the yield
strength of the system Fy was replaced by My the yield bending moment at the plastic hinge.
Equation (17) has an advantage over Equation (15) in that DM is expressed in terms of plastic
hinge parameters where the plastic deformation energy is dissipated (and therefore where the
damage occurs). Note that 'umon is a parameter that can be speciCed as a general requirement
for structural members; this is not the case for !umon .
Instead of an SDOF oscillator, consider the case of a real building. If it is assumed that all
the plastic hinges that transform the structure in a mechanism occur at the same time and that
they have the same 'umon , then all plastic hinges will have the same local damage deCned by
Equation (17). If it is considered now that the rotations in the plastic hinges in each storey
are approximately equal to the IDI, and that the concentration of plastic hinge damage can
be considered with the same factor proposed by Hwang and Jaw [17] [Equation (14)], the
maximum local (storey) damage can be computed as

IDI − (IDI= ) IDI IDI − (IDI= ) IDI
DML = + b% 2 0:414
= + b% 2 0:414
IDI umon − (IDI= ) IDI umon 'umon − (IDI= ) 'umon
(18)
Figure 5 shows the local damage for the Los Gatos and SCT (Mexico, 1985) records
obtained using Equation (18) for typical RC frame buildings. For each period T , the height
of the building was computed using the NEHRP [18] code formula for RC frames (T = CT H 3=4
where CT = 0:0731 and H is in metres). In Figure 5, the Crst two graphs (at the top) show
the DM due to the consideration of IDI only. The second group of two graphs shows the
damage due to the cumulative EH , and Cnally the two graphs at the bottom show the total
local damage scaled by the empirical factor of 0:414 as shown in Equation (18).
5.4. The need for a preliminary design procedure that considers a cumulative damage
index
According to Figure 5, the cumulative eDect of damage for = 3 could be responsible not
only for more than 70 per cent of the damage index in the case of a long duration of periodic
strong motion like SCT (T = 2 s or H =80 m), but could also be responsible for about 40
per cent of the local damage for severe pulse type of ground motions like Los Gatos for
speciCc values of structural period (T = 0:7 s or H =20 m). From Figure 5, it is very clear
that for a long-duration periodic type of earthquake ground motion, the cumulative eDect is
larger than the damage due to the maximum IDI only (or displacement for an SDOF). The
cumulative eDect of damage could also be important if signiCcant aftershocks are considered.
5.5. The need to control not only displacements but also ductility (minimum strength) to
limit damage
To control damage, it is necessary to control not only the displacement but also the maximum
global ductility (i.e. minimum strength of building). This fact is apparent from Figure 5,
DM due to maximum IDI (Los Gatos EQGM) DM due to maximum IDI (SCT EQGM)
ξ = 5 % θ umon = 0.05 b = 0.20 ξ = 5 % θumon = 0.05 b = 0.20

1.2 1.2
IDI IDI
IDI − T = 0.0731H 3/4 H [m] IDI −
T = 0.0731H 3/4 H [m] µ µ
1.0 1.0 IDI
IDI θ umon −
θ umon − µ
Damage Index, DM
µ
Damage Index, DM
0.8 0.8
0.6 0.6
µ = 3.0 µ = 3.0
0.4 0.4 µ = 2.0
µ = 2.0
0.2
0.2
µ = 1.5 µ = 1.0 µ = 1.5 µ = 1.0
0.0 0.0
50 100 150 200 0 50 100 150 200
0
Building Height, H (m) Building Height, H (m)
DM due to cumulative hysteretic energy (Los Gatos EQGM) DM due to cumulative hysteretic energy (SCT)
ξ =5% θ umon = 0.05 b = 0.20 ξ = 5 % θ umon = 0.05 b = 0.20
1.2 1.2
T = 0.0731H 3/4 H [ m] T = 0.0731H 3/4 H [ m] IDI
IDI bγ 2µ
bγ 2µ 1.0 θ umon
1.0 θ umon µ = 3.0
Damage Index, DM
Damage Index, DM
0.8 0.8
0.6 µ = 3.0 0.6
µ = 2.0 0.4 µ = 2.0

0.4
µ = 1.5
0.2 µ = 1.0 0.2 µ = 1.5
µ = 1.0
0.0 0.0
50 100 150 200 0 50 100 150 200
0
IDI
Maximum Local DM (Los Gatos EQGM) Maximum Local DM (SCT EQGM) IDI −
µ IDI
IDI + bγ µ µ 0.414
IDI − ξ = 5 % θ umon = 0.05 b = 0.20
ξ = 5 % θ umon = 0.05 b = 0.20 µ 2 IDI
+ bγ µ µ 0.414 θ umon−
IDI θ umon
IDI θ umon 1.2 µ
1.2 θ umon−
µ T = 0.0731H 3/4 H [m] µ = 3.0
1.0 1.0
µ = 3.0
Damage Index, DM
Damage Index, DM
0.8 µ = 2.0
0.8
µ = 2.0
0.6 0.6
µ = 1.5 µ = 1.5
0.4 0.4
0.2 0.2
µ = 1.0 3/4
µ = 1.0
T = 0.0731H H [m]
0.0 0.0
50 100 150 200 0 50 100 150 200
0
Figure 5. Local damage for Los Gatos and SCT (Mexico, 1985) EQGMs for typical RC frame buildings.
when the local damage is computed using Equation (18), the increase in the global ductility
increases the Crst and second terms of the equation as well as the common factor outside
the parenthesis. In the Crst term, this happens because, for the same maximum displacement,
smaller ductility means larger yielding displacement so that in the limit it could be possible to
obtain maximum displacement without damage if the structure remains elastic (i.e. =1). The
ductility increases the damage even for an SDOF system with no consideration of cumulative
damage. On the other hand, on average the second term of Equation (18) tends to increase
almost linearly with . Finally, the concentration of damage represented by the common factor
outside the parenthesis is larger as the structure response goes deeper in the non-linear range.
As a result, the global ductility has very important eDects on the level of local damage as
shown in Figure 5.
If the Tinitial (i.e. the elastic T ) is controlled by serviceability, then to control damage for life
safety or collapse prevention, it is necessary to limit not only the displacement but also the
ductility, and therefore a minimum strength is necessary. If the design does not consider the
serviceability performance limit, the displacement ductility can be decreased by decreasing
the stiDness k [see designs A and B in Figure 6(a)]. This clearly reveals the need for two
POs (i.e. designing for serviceability besides life safety).
5.6. The need for a conceptual comprehensive preliminary design approach for PB-ERD
Several simpliCed design methods have been considered for performance-based earthquake
resistant design (PB-ERD): (a) strength-based design, (b) displacement based design, and
(c) energy-based design [1]. A discussion about the weakness of strength-based design can
be found in Reference [22]. Brie;y, since PBSD needs to limit damage, it is clear that dis-
placement-based design should be a better approach because damage is more sensitive to
displacement (strain) than to strength (stress).
Among the procedures that are related to displacement-based design there are two groups.
In the Crst group [23; 24], displacements are used mainly to check a preliminary design
based on strength using serviceability or code forces. The second group [22] attempts to use
a simpliCed approach to obtain a preliminary design using displacements, which has been
termed as ‘direct displacement-based design’. Although this simple procedure uses only linear
elastic spectra, its reliability for general application can be questioned because it is based
on the use of an elastic substitute structure and an equivalent eDective structural damping
which is estimated based on the implied ductility level = !=!y using derived relationships
between and an equivalent eDective damping e . Since damping involves dissipation of
energy without damage, meanwhile ductility involves dissipation of energy with associated
damage, it is not possible to deCne just a unique equivalent damping e for all the response
parameters that need to be considered.
The shortcomings of any approach that tries to reduce the real inelastic non-linear behaviour
using an equivalent eDective damping e arise from the governing equations for an inelas-
tic system. Consider the following non-linear and linear equations of motion for an SDOF
system [16]:
uY u̇ fs (u; u̇) muY g (t)

Non-linear equation + 2!n + !n2 = −!n2 (19a)
uy uy fy fy
uY u̇ u muY g (t)
‘Equivalent’ linear equation + 2 e !n + !n2 = −!n2 (19b)
uy uy uy fy
where u(t) is the SDOF displacement, uy the yielding displacement, fy the yielding strength
and fs (u; u̇) the resisting force for an elastoplastic system. Although diDerent ways (methods)
(a) Seismic Yielding Coefficient, Cy (Los Gatos EQGM) Maximum Local DM (Los Gatos EQGM)
ξ =5% ξ = 5 % θ umon = 0.05 b = 0.20
2.0 1.2
1.8
1.0 µ = 3.0
1.6
H=60 M
Damage Index, DM
1.4
0.8
1.2 V
A µ = 2.0
1.0 µ = 1.0 0.6
B
0.8 B B A
µ = 1.5 µ = 1.5
0.4
0.6 µ = 2.0
0.4 A
IDI
0.2
0.2 µ = 3.0 µ = 1.0
0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Period, T (s) Period, T (s)
(b) Seismic Yielding Coefficient, Cy (Los Gatos EQGM) Seismic Yielding Coefficient, Cy (Los Gatos EQGM)
ξ =5% µ = 1.0
3.5 3.5

3.0 3.0
ξ =5%
2.5 µ = 1.0 2.5
2.0 2.0
ξ = 15 %
1.5 1.5
1.0 µ = 2.0 1.0
0.5 0.5
µ = 4.0 ξ = 30 %
0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Period, T(s) Period, T(s)
(c) Displacement, Sd (Los Gatos EQGM) Displacement, Sd (Los Gatos EQGM)

ξ =5% µ = 1.0
2.5 2.5
µ = 1.0
2.0 2.0 ξ =5%
µ = 4.0
Displacement, Sd (m)
Displacement, Sd (m)
1.5 1.5 ξ = 15 %
µ = 2.0
1.0 1.0
ξ = 30 %
0.5 0.5
0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Period, T(s) Period, T(s)
Figure 6. (a) Structural damage control for life safety decreasing the stiDness. (b) Strength SDOF
spectra for Los Gatos EQGM. (c) Displacement SDOF spectra for Los Gatos EQGM.
have been suggested and used to estimate the ‘equivalent’ viscous damping (e ) to be used in
Equation (19b) the authors believe that it will be very diKcult to establish reliable e for PB-
ERD. The main reason for this is that the dissipated energy arises from diDerent terms in the
two equations so that the maximum response (and therefore the damage) of an SDOF can be
signiCcantly diDerent for both equations. From the interpretation of the above two equations,
it becomes clear that while the energy dissipated by the viscous damping (e ) occurred during
the entire response of the structure, i.e. in its elastic as well inelastic response, the energy
dissipated by plastic deformation ( ) takes place only when the structure undergoes plastic
deformations. Thus, the value of the equivalent e will be very sensitive to the type (the
entire time history) of the earthquake ground motions.
Reasons for supporting the above statements are shown in Figure 6(b) and 6(c), where
the spectra for strength and displacements for Los Gatos earthquake ground motion using
5 per cent of damping and 1.0, 2.0 and 4.0 ductilities, and using elastic behaviour and
5, 15 and 30 per cent damping are plotted. Note that the strength demand Cy decreases
with increasing (with constant ) or increasing (with elastic behaviour). However, the
displacement demand Sd always decreases with increasing (with elastic behaviour) but can
increase or decrease when is increased (with constant ).
From the results shown in Figure 6(c), it is clear that damping is much more eDective in
reducing displacements than ductility (note the importance of the designer having an under-
standing that ductility and damping are not equivalent in terms of displacement reduction).
If the prediction of the actual inelastic response (strength and deformations) is replaced by
predicting such response through a linear elastic analysis considering an equivalent damping
coeKcient e obtained by equating EH to EHe , this analysis cannot lead to a reliable estima-
tion of the actual response (strength and deformations) since only one of the main response
parameters (strength and deformation) can be Ctted. A discussion about the diDerences in the
relative eDects of yielding and damping can be found in Reference [16].
Although the application of the elastic substitute structure procedure is straightforward, the
use of an elastic substitute structure and equivalent damping that fuses two diDerent types
of mechanical characteristics in one could be misused by designers. There is a need for
a comprehensive design procedure for PBSD that can be used to calibrate this simpliCed
approach to avoid its misuse, i.e. to Cnd its limitations.
A few simpliCed procedures have been developed for using energy-based design [25]. Again,
although the input energy and hysteretic energy are probably the best parameters to select
the design earthquake ground motions, the energy-based design approaches that have been
developed until now satisfy the energy equation using basically a monotonically increasing
deformation approach that can hardly represent the actual seismic response and damage. Most
of the methods that have been proposed have sacriCced some important concept for the sake
of simplicity. However, with the amount of speciCc software, spreadsheets, and mathematical
packages available today, simplicity should be redeCned. A numerical procedure is not simpler
because an equation has fewer terms or some important parameter is ignored. A numerical
procedure is simple when it is easily understood and when the designer can go from the
performance objectives to the design values in an explicit and transparent way. A design
procedure, based on a probabilistic multi-level seismic design criteria, that from the beginning
explicitly considers the performance objectives in terms of displacement, strength, energy
dissipation, and damage for structural and non-structural elements and contents, is called a
comprehensive design. A conceptual comprehensive design procedure has been developed and
applied to the design of two reinforced concrete buildings, one of 30 storeys [26] and another
of 10 storeys [27].
6. CONCEPTUAL COMPREHENSIVE APPROACH FOR THE PBSD OF BUILDINGS
A comprehensive design procedure involves several steps: (i) selection of the performance
objectives, (ii) site suitability and seismic hazard analysis, (iii) conceptual overall seismic
design, and (iv) comprehensive numerical seismic design. The last step is divided into three
main phases: (a) establishment of the design earthquake ground motions; (b) numerical pre-
liminary design procedure; and (c) the Cnal design. In this paper, only the comprehensive
numerical seismic design will be discussed. In order to arrive at the desired Cnal design, it
is necessary to start with a preliminary numerical design procedure, whose main objectives
are twofold. First, the procedure is based on fundamental principles of structural dynamics
considering the real mechanical behaviour of the entire building system and second, it leads
to a preliminary design that is as close as possible to the desired Cnal design. The numerical
preliminary design phase consists of three main groups of steps: (i) Statement of the prob-
lem and its preliminary analysis, (ii) Preliminary sizing and detailing and (iii) Acceptability
checks of the preliminary design.
6.1. Statement of the problem and its preliminary analysis
This step is discussed under a ‘Given–Required–Solution’ format as follows:

Given: Function of building and desired performance design objectives; general conCgura-
tion, structural layout, structural system, structural materials and non-structural components
and contents (which should have been selected using the guidelines developed for conceptual
overall seismic design [28]); gravity, wind, snow and other possible loads or excitations; and
displacement, strength and dissipated hysteretic energy due to and (i.e. EH and EH ) de-
sign spectra for diDerent damping and ductility for frequent minor and rare major earthquake
ground motions.
Required: Establishment of design criteria (acceptable damage levels under the established
earthquake ground motion levels and the levels of other signiCcant excitations), minimum
lateral stiDness (or maximum period T ) and minimum strength of the building capable of
controlling the damage, the design seismic forces and the critical load combinations.
Solution: To use a comprehensive approach that from the beginning takes into account the
fact that the building structure is a multi-degree-of-freedom (MDOF) system and there can
be important torsional eDects even under service earthquake ground motions (i.e. in the linear
elastic response); that for safety earthquake ground motions these eDects can be diDerent; and
that it is also necessary to consider the desired damage index (control of damage) taking into
account the eDects of the cumulative damage (or cumulative ) due to the inelastic hysteretic
behaviour of the critical regions of members and connections, and the ductility ratio that can
be used, as well as the expected overstrength.
Using the given data, the designer must prepare (using the methodology presented in
Section 5.3 of this paper) design spectra for local structural damage (DML ) and non-structural
damage (IDI) like those of Figure 7. If it is foreseen that (due to the severity of the earth-
quake ground motions at the service level as well as the life safety or near collapse levels)
it will not be economical to use the traditional or conventional structural systems, it will be
necessary to consider the use of innovative systems [29]. In this case, to facilitate the selection
of the most eKcient innovative strategy and technique, it will be necessary to prepare spectra
similar to those shown in Figure 7 but for diDerent values of rather than just and for
possible combinations of and .
The spectra shown in Figure 7 were developed and considered for the design of a building
having a height of 90 m and located on Crm soil. The PGAs for frequent and rare earthquake
were chosen as 0:19g and 0:60g. The SDOF spectra were obtained using the Newmark–Hall
2.0 0.020
µ = 3.0
µ = 2.0
Life safety
µ = 1.5 IDI saf

ξ = 5%
1.5 0.015
µ = 1.5

µ = 2.0
µ = 3.0
1.0 0.010
Serviceability
TR = 20 years ξ = 2 % µ = 1.0
0.5 0.005
IDI ser
0.0 0.000
0.0 0.5 1.0 1.5 2.0 2.5 3.0
T ser T saf Maximum periods to satisfy Service
Period, T(s) and Safety Performance Objectives
1.0
µ = 3.0
0.8 Life safety

ξ = 5%
µ = 2.0
Damage Index, DM
DM saf
0.6
µ = 1.5
0.4
0.2
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Period, T (s)
Figure 7. Design spectra for a 90 m building on Crm soil at high seismic risk zone.
Design Spectra [4] and the strength-reduction factors proposed by Miranda [6] rather than
those that can be obtained from an average (or mean + ) of a suit of recorded earthquake
ground motions. The design spectra for local structural damage (DML ) and non-structural
damage (IDI) were obtained using Equations (18) and (14). Using Figure 7, and considering
the speciCed POs for structural damage (DMsaf ), and non-structural damage (IDIser , IDIsaf )
for service and life safety limit states, the designer can select the minimum period and the
maximum ductility from an explicit consideration of those performance objectives. Once the
design period, ductility, and damping have been selected the design forces can be obtained
using the strength spectra.
To illustrate the application of the spectra shown in Figure 7 to the design of the above-
mentioned building with a height of 90 m, consider the following two basic POs: the ser-
vice performance level (selecting as acceptable an IDIser 60:004 and a local DMser = 0:0);
and the life safety performance level (limiting the IDIsaf 60:015 and the acceptable local
DMsaf 60:60 and that the corresponding required global should be 63:0). From the IDI
spectra of Figure 7, it becomes clear that to control the IDIser and the IDIsaf to values
smaller that the above acceptable values, a maximum Crst mode period T = 1:7 s is needed
(note that serviceability controls the design in this particular case). Once the target period
has been selected, the preliminary design and detailing can be done as in the following
sub-sections.
6.2. Preliminary sizing and detailing
This step starts with the preliminary design for stiDness that requires the solution of three
problems: (a) given a target period and Crst mode shape to obtain the required storey stiD-
ness; (b) given the required storey stiDness to obtain the required member sizes; and (c) given
the required member sizes to obtain the eDective moment of inertia of members. It can be
shown that using a stick model of the building, from the selected Crst mode period, modal
shape and storey mass, the required storey stiDness can be directly estimated starting from
the top storey down to the bottom one. Then, using a sub-structure model for each storey,
problems (b) and (c) can be solved, and the preliminary sizes of the members to satisfy
the required stiDness are obtained [26; 27]. Using plastic analysis and linear programming to
optimize the amount of reinforcement, the beam ;exural reinforcement of an RC structure
could be obtained storey by storey satisfying simultaneously service and life safety require-
ments. Then, using capacity design concepts, the columns and shear reinforcement are selected
[26; 27].
6.3. Acceptability checks of the preliminary design
A discussion about the use of the diDerent kind of analyses available for the acceptability
checks has been presented in Reference [26], and that about the reliability aspects and the
number of analysis required can be found in Reference [10].
7. SUMMARY, CONCLUSIONS AND RECOMMENDATIONS
It has been shown in this paper that to satisfy the objectives of a reliable PBSD, it is necessary:
• to start with a multi-level seismic design criteria: At least two performance levels should
be considered (even for the preliminary design) so that appropriate design decisions can
be made to satisfy the performance objectives.
• to consider a probabilistic design approach: The ideal approach would be to develop
a reliability-based performance oriented methodology for design and evaluation, which
will permit the designer to calculate the conCdence that the building will satisfy the POs.
Since a COV of about 0.20 could be expected for the capacities and a COV of about
0.80 for the earthquake demand, a simple probabilistic approach (using load factors)
could be used for design.
• to consider local structural and non-structural damage and therefore design spectra for
buildings: Elastic and inelastic IDI spectra as well as local damage spectra for buildings
can be developed from the basic equations of the earthquake modal response of n degree
of freedom systems.
• to account for cumulative damage: It has been shown that for a long duration periodic
type of earthquake ground motion, the cumulative eDect can be larger than the damage
due to the maximum IDI only and could even be responsible for an important portion
of the local damage for severe pulse type of ground motions for speciCc values of
structural period. Also, the cumulative eDect of damage could be important if signiCcant
aftershocks are considered.
• to control not only displacements but also ductility (minimum strength) to limit damage:
It has been shown that the global ductility (i.e. for a given stiDness, the strength of the
building) has a very important eDect on the level of local damage. Increasing the ductility
not only increases the damage of SDOF system, but it also increases the concentration
of local damage in a building as the structure response goes deeper in the non-linear
range.
• to develop a conceptual comprehensive preliminary design approach: It is shown that
some approaches that have been proposed fail to satisfy the requirement of a reliable
PB-ERD. In particular, it has been concluded that conceptually if the prediction of the
actual inelastic responses (strength and deformations) is replaced by predicting such
responses through a linear elastic analysis considering an equivalent damping coeKcient
e obtained by equating EH to EHe , this analysis cannot lead to a reliable estimation
of the actual responses (strength and deformations) since only one of the main response
parameters (strength or deformation) could be Ctted.
Finally, a conceptual comprehensive preliminary design approach is proposed. This
methodology is transparent, i.e. based on well-established fundamental principles of
structural dynamics, mechanical behaviour of real buildings and comprehensive design,
and in compliance with the worldwide-accepted philosophy for seismic design. The
design procedure starts with the speciCcation of desired performance objectives for the
entire structural system, given the hazardous environment in which it is to be constructed, and
then provides a direct rational path by which the structure may be designed to attain these
goals.
Its main advantages are: (i) it leads to a transparent numerical design procedure that con-
siders and checks the selected or desired performance objectives; (ii) in spite of the great
uncertainties in the quantiCcation of some of the concepts involved in its codiCcation, such
quantiCcations can be improved as new more reliable data become available without changing
the philosophy and particularly, the format of this codiCed methodology; (iii) such formu-
lation can be used as a basis for improving the education of architects and engineers, as
well as for the establishment of the much-needed prioritization and programme of the focused
research needed to improve seismic design. In particular, it is urgent to improve the available
information regarding the following areas:
• to quantify the requirements for all performance levels: There is a need for an extensive
and co-ordinated research programme to improve the probabilistic deCnition of structural
and non-structural damage indexes. Particularly, there is a need for reliable data (and
therefore studies) regarding the interacting eDects of ;oor acceleration and IDI on non-
structural components and pipeline damage, as well as the level of the content damage
associated with ;oor velocities and=or accelerations.
• to develop more reliable structural damage indexes that can accurately reproduce the
eDects of all types of excitations that can act simultaneously and all possible load–
deformation paths and damage mechanisms and take into consideration the possible ef-
fects of previous ground motions as well as those due to possible aftershocks and future
earthquakes.
• to develop more reliable ampliCcation factors for taking into account the concentration
of damage in buildings including the eDects of gravity loads.
REFERENCES
1. SEAOC Vision 2000 Committee. Performance-based seismic engineering. Report prepared by Structural
Engineers Association of California, Sacramento, CA, 1995.
2. Collins K, Stojadinovic B. Limit states for performance-based design. Proceedings of 12 WCEE, Auckland,
New Zealand, 2000.
3. EERI. Northridge Earthquake, 17 January 1994. Preliminary Reconnaissance Report. Earthquake Engineering
Research Institute, 94-01, 1994.
4. Newmark NM, Hall WJ. Procedures and criteria for earthquake resistant design. Build. Res. Series No.46, Nat.
Bureau of Standards, U.S. Depts of Commerce, Washington, DC, 1973; 209 –236.
5. Der-Kiureghian A, Ang AHS. A fault-rupture model for seismic risk analysis. Bulletin of the Seismological
Society of America 1977; 67(4):1173–1194.
6. Miranda E. Site-dependent strength-reduction factors. Journal of Structural Engineering, ASCE 1993;
119(12):3503 –3519.
7. Cornell A, Krawinkler H. Progress and challenges in seismic performance assesment. PEER Center News 2000;
3(2):1–3.
8. Ellingwood B, MacGregor J, Galambos T, Cornell A. Probability based load criteria: loads factors and load
combinations. Journal of the Structural Division, ASCE 1982; 108(ST5).
9. Algermissen ST, et al. Probabilistic estimates of maximum acceleration and velocity in rock in the contiguous
United States. Open File Report No. 82-1033, U.S. Geological Survey, Reston, VA, 1982.
10. Bertero R. Acceptability checks for performance-based design. Proceedings of The EERC-CUREe Symposium
in Honor of Vitelmo V. Bertero, 31 January–1 February 1997, Berkeley, CA, USA, 1997.
11. Ang A, Tang W. Probability Concepts in Engineering Planning and Design, vol. II. Wiley: New York, 1984.
12. Nowak A, Collins K. Reliability of Structures. McGraw-Hill: New York, 2000.
13. Galambos T, Ellingwood B, MacGregor J, Cornell A. Probability based load criteria: assessment of current
design practice. Journal of the Structural Division, ASCE 1982; 108(ST5):959–967.
14. Ellingwood B, Rosowsky D. Combining snow and earthquake loads for limit states design. Journal of the
Structural Division, ASCE 1996; 122(ST11):1364–1368.
15. Bertero VV. Performance-based seismic engineering: a critical review of proposed guidelines. Proceedings of the
International Workshop on Seismic Design Methodologies for the Next Generation of Codes. A.A. Balkema:
Rotterdam, 1997.
16. Chopra A. Dynamics of Structures. Theory and Applications to Earthquake Engineering. Prentice-Hall:
Englewood CliDs, NJ, 1995.
17. Hwang H, Jaw J. Statistical evaluation of de;ection ampliCcation factors for reinforced concrete structures.
Proceedings of the Fourth U.S. National Conference on Earthquake Engineering, vol. 2. EERI: El Cerrito,
CA, USA, 1990; 937 –944.
18. NEHRP Recommended Provisions for the Development of Seismic Regulations for new Buildings, 1994.
19. Williams M, Sexsmith R. Seismic damage indexes for concrete structures: a state-of-the-art review. Earthquake
Spectra 1995; 11(2):319 –349.
20. Park YJ, Ang AHS. Mechanistic seismic damage model for reinforced concrete. Journal of Structural
Engineering, ASCE 1985; 111(4):722–739.
21. Fajfar P. Equivalent ductility factors, taking into account low-cycle fatigue. Earthquake Engineering and
Structural Dynamics 1992; 21(10):837– 848.
22. Priestley MJN. Performance based seismic design. Proceedings of 12 WCEE. Auckland, New Zealand, 2000.
23. Mohele, JP. Displacement-based design of RC structures subjected to earthquakes. Earthquake Spectra 1992;
8(3):403 – 428.
24. ATC 40. Seismic Evaluation and Retro@t of Concrete Buildings. ATC: Redwood City, CA, 1997.
25. Leelataviwat S, Goel S, Stojadinovic B. Drift and yield mechanism based seismic design of structures.
Proceedings of the Sixth U.S. National Conference on Earthquake Engineering. EERI: Oakland, CA, 1998.
26. Bertero R, Bertero V. Tall reinforced concrete buildings: development of conceptual seismic-resistant design
methodology. Report UCB=EERC 92=6, U.C. Berkeley, USA, August 1992.
27. Bertero R. Diseño Sismorresistente basado en la Performance. Ph.D. Thesis, Universidad de Buenos Aires,
Argentina, 2001 (in Spanish).
28. Bertero V. Lessons learned from structures damaged in recent earthquakes. Proceedings of 7 WCEE, Tomo 4,
1980; 257 –264.
29. Bertero V. Performance-based seismic engineering: conventional vs. innovative approaches. Proceedings of 12
WCEE, Auckland, New Zealand, 2000.

Performance-Based Seismic Engineering: The Need For A Reliable Conceptual Comprehensive Approach

Uploaded by

Copyright:

Available Formats

Performance-Based Seismic Engineering: The Need For A Reliable Conceptual Comprehensive Approach

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Performance-Based Seismic Engineering: The Need For A Reliable Conceptual Comprehensive Approach

Uploaded by

Copyright:

Available Formats

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2002; 31:627–652 (DOI: 10.1002/eqe.146)

Performance-based seismic engineering: the need for a reliable

Raul D. Bertero1; ∗; † and Vitelmo V. Bertero2

Richmond; CA 94804, U.S.A.

KEY WORDS: performance-based seismic engineering; performance-based seismic design; conceptual

Received 15 January 2001

2. OBJECTIVES OF THIS PAPER

3. DEFINITIONS OF PBE, PBSE AND PBSD

4. PERFORMANCE OBJECTIVES (POs)

EARTHQUAKE PERFORMANCE LEVEL

EARTHQUAKE DESIGN LEVEL

5. MAIN REQUIREMENTS FOR A RELIABLE PBSD

5.1. The need for multi-level seismic design criteria

Table I. Example of quantiCcation of the performance objectives (POs).

Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2002; 31:627–652

Probability of Limit State Exceedence

Probability of Limit State Exceedence

Design B Performance Objective

(b) Interstory Drift Index

Seismic Yielding Coefficient, Cy

0.6 TR 500 years 5% =3

Figure 3. Strength spectra for serviceability and life safety.

5.2. The need for a probabilistic design approach

A probabilistic foundation for the development of performance-based guidelines has been

Cx ¿Dx (1 +  COVDx )= Dx +  Dx (3)

Elastic IDI (Los Gatos EQGM) Elastic IDI (SCT EQGM)

Story Drift Index, IDI

0.03 H=120 m 0.03

(b) Period, T (s) Period, T (s)

Inelastic IDI (Los Gatos EQGM) Inelastic IDI (SCT EQGM)

0.05 µ = 4.0 0.05

Story Drift Index, IDI

' − 'y EH ' − ('= ) '

ξ = 5 % θ umon = 0.05 b = 0.20 ξ = 5 % θumon = 0.05 b = 0.20

0.6 µ = 3.0 0.6

µ = 2.0 0.4 µ = 2.0

uY u̇ fs (u; u̇) muY g (t)

Seismic Yielding Coefficient, Cy

1.0 µ = 2.0 1.0

(c) Displacement, Sd (Los Gatos EQGM) Displacement, Sd (Los Gatos EQGM)

6. CONCEPTUAL COMPREHENSIVE APPROACH FOR THE PBSD OF BUILDINGS

6.1. Statement of the problem and its preliminary analysis

This step is discussed under a ‘Given–Required–Solution’ format as follows:

Seismic Yielding Coefficient, Cy

Story Drift Index, IDI

0.8 Life safety

6.2. Preliminary sizing and detailing

6.3. Acceptability checks of the preliminary design

7. SUMMARY, CONCLUSIONS AND RECOMMENDATIONS

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Cx ¿Dx (1 + COVDx )= Dx + Dx (3)