Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
European School for Advanced Studies in Reduction of Seismic Risk (ROSE School),
University of Pavia, Via Ferrata, 27100 Pavia, Italy
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
The state-of-the-art earthquake loss estimation techniques make use of pushover analy-
sis to define the performance of structures under earthquake loading, represented by a
demand spectrum whose ordinates reflect the inelastic response. The performance point
defined in this way is then used as input to the fragility or loss curves. This rigorous
approach represents the earthquake actions by a parameter that is known to have a
good correlation with damage and also takes into account the dynamic characteristics
of different buildings. However, for many applications, the available data on key input
parameters, such as soil conditions and the type and distribution of the exposed buil-
ding stock, are limited to the extent that the rigorous capacity spectrum approach can
become either impractical or unjustified. It is for such case scenarios that a much sim-
pler deformation-based assessment methodology, presented herein, has been developed.
The proposed method possesses all the inherent advantages of a displacement-based
approach, whilst reducing the required calculations by one or more orders of magnitude
when compared to conventional vulnerability assessment procedures.
1. Introduction
1.1. Preamble
Recent developments across the world relating to the devastating impacts of earth-
quakes in both the social and economic arenas have resulted in an increased aware-
ness amongst populations, industry, commerce and governments, of seismic hazard
and the vulnerability of the built environment. Greater interest has thus been placed
on the research of methods with which the potential risk from earthquakes can be
assessed and, eventually, mitigated.
More recently, this interest has taken the form of investments of both time and
knowledge by private companies and government institutions, especially within the
reinsurance industry. The result of such efforts has been commendable. Programmes
such as HAZUS [FEMA, 1999; Whitman et al., 1997] have not only provided a re-
liable and detailed method for the prediction of the effects of potential earthquakes
within urban areas or across large regions (including both the nature and the distri-
bution of physical damage, and estimations of the total direct and indirect economic
107
May 5, 2003 16:51 WSPC/124-JEE 00098
loss), but have also led to a far greater understanding of the problem, highlighting
the most efficient ways in which this risk can be mitigated.
The downside is that such comprehensive vulnerability assessment methodolo-
gies often require considerable amounts of time and research to collect and com-
pile the detailed data necessary to perform a complete study, and then significant
amounts of computing power to run (the size of which being usually out of reach
to all but governments and large institutions).
At the core of any loss estimation methodology is a process that takes a given
building inventory with known properties and geographical distribution and con-
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
volutes it with a representation of the seismic demand calculated across the same
geographical region using ground-motion prediction equations. This representation
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
of seismic demand has previously taken the form of PGA, PGV or Intensity. How-
ever, new developments in the area of displacement-based design [Moehle, 1992;
Calvi and Pavese, 1995; Kowalsky et al., 1995] and its increasing popularity as the
most logical and rational means of structural design have naturally led to displace-
ments becoming also the major parameter in building assessment [e.g. Priestley,
1997; Calvi, 1999].
In the realm of vulnerability assessment, this trend has seen the introduction
of the Capacity Spectrum method [Freeman, 1975; Freeman, 1998; Fajfar, 1999],
in which a curve of spectral acceleration versus spectral displacement is used to
represent earthquake demand and pushover curves characterise structural capacity.
This method, when used in vulnerability assessment programs such as HAZUS,
allows the inclusion of the effective period of structural vibration and the effect of
variable amounts of inelastic deformation (ductility) in the estimation of damage
states. It is this most crucial part of the procedure, involving the derivation of these
demand and capacity curves and the identification of their intersection (known
as the performance point, SD ), that is especially demanding in terms of time,
computing power, and required input data.
The rationale underlying this study is that an earthquake loss estimation model
should be consistent in the level of refinement in all its elements and in its method of
calculation. There will no doubt be many situations in which the available data on
the soil conditions and building stock will be limited, and in which constraints of
time and cost will not permit the investment required to gather more complete
information; it has been estimated that as much as two years is required to com-
pile a detailed loss estimation inventory for a new study area using the HAZUS
methodology (FEMA, 1999). This may be particularly true for studies developed
at regional or national levels [e.g. Bommer et al., 2002; Spence et al., 2002]. In such
cases, the computational complexity of the capacity spectrum approach may not
be justified, but this should not necessarily result in a reversion to the use of PGA
or intensity to represent the ground motion hazard. A simplified approach, initially
proposed by Pinho et al. [2002], which retains the elements of the effective period of
vibration of structures and structural displacements as the most suitable indicator
of damage, is therefore presented and discussed in this paper.
May 5, 2003 16:51 WSPC/124-JEE 00098
shaking scenario is found from the intersection of the demand spectrum (spectral
acceleration versus displacement), representing the ground motion, and the capacity
spectrum (pushover curve), representing the horizontal displacement of the struc-
ture under increasing lateral load. This displacement then provides the input for
the fragility curves of the same building type from which it is possible to read the
expected proportions (PI , PII , PIII ) that will be in each damage state (I, II, III).
This procedure is illustrated schematically in Fig. 1.
This method allows the inclusion of the effective period of structural vibration
and the effect of variable amounts of inelastic deformation (ductility) in the esti-
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
III
Capacity Curve
PI
PII
PIII
SD displacement SD Demand
(a) (b)
Fig. 1. HAZUS methodology: (a) demand/capacity and (b) fragility curves.
May 5, 2003 16:51 WSPC/124-JEE 00098
establish upper and lower bound estimates of the material and geometric variables
of the building stock. These minima and maxima values then led to the development
of ranges of possible capacities and corresponding periods.
These capacity ranges define areas that can then be in some way intersected by
a given response spectrum chosen to represent the demand, as shown in Fig. 2. The
probability of attaining the associated limit state is then computed by integrating
the volume of a Joint Probability Density Function (JP DF ) between capacity
and period above and below the spectral line and within the defined ranges. This
JP DF was assumed as constant in the aforementioned publication, most likely
due to the significant complexity of deriving a more realistic and statistically valid
distribution, noting the compounding effect of the multiple pairs of maximum and
minimum assumptions made in the formulation of the possible ranges.
However, this pioneering work of Calvi [1999], whereby the concept of using
a fully displacement-based philosophy in the assessment of seismic vulnerability of
classes of buildings was for the first time proposed and used, possesses an important
shortcoming. The variation of effective height with ductility was neglected and,
more fundamentally, the existence of a derivable relationship between displacement
capacity, material properties, building height and, through that, period, was over-
looked. In the methodology proposed within the current work, described in the
following section, such ductility-height relationship is explicitly treated, leading to
a more complete, transparent and theoretically robust deformation-based method
for seismic vulnerability assessment of classes of buildings.
May 5, 2003 16:51 WSPC/124-JEE 00098
ing factor (ηLSi ) in correspondence to the ductility demand (µLSi ), and its corre-
sponding drift capacity curves, indicate the periods (TLSi ) that mark the boundaries
of the various limit states (LSi). That is, the intersection of the demand and ca-
pacity curves indicate periods at which capacity (defined by a given limit state)
exactly matches the demand. Thus, all periods to the left of this intersection rep-
resent buildings for which the demand exceeds the capacity, and all periods to its
right stand for buildings where capacity exceeds demand (in this particular case).
These periods are then transformed into their equivalent heights (HLSi ), using the
previously mentioned period-height relationships, and plotted across a cumulative
distribution function (CDF) of buildings with height to find the proportions of the
exposed building stock failing each limit state [Fig. 3(b)]. It is noteworthy that this
procedure neglects the inherent dispersion in the vulnerability of any individual
class of building; this issue is addressed in Sec. 4.3.
Thus, by identifying these relationships between natural period, height and dis-
placement capacity it is possible to use this alternative approach to the capacity
cumulative
displacement
ηLS3
PLS2
Demand
Spectra
PLS3
0
TLS3 TLS2 TLS1 effective HLS3 HLS2 HLS1 Height
period
(a) (b)
Fig. Fig.
3. 3. Proposedmethodology:
Proposed methodology: (a)
(a)Displacement
displacementcapacity/demand curves
capacity/demand (b) CDF
curves (b)ofCDF
building stock
of building
stock.
May 5, 2003 16:51 WSPC/124-JEE 00098
The first step in the process of determining the capacity curves is the separation
of the exposed building stock into its various classes, defined as a function of the
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
building’s constituent material (steel, concrete, masonry, etc.) and structural system
(i.e. frame, wall or dual systems). In the current presentation, only the case of
reinforced concrete moment resisting frames (M RF ) is considered. Nonetheless,
the methodology described in the following sections can be extended, with varying
degrees of difficulty, to any other building class. It is also noteworthy that other
sub-divisions usually adopted in vulnerability assessment type of work, such as
height, age, level of seismic design and degree of maintenance, are already explicitly
or implicitly incorporated within the formulation proposed below.
Fig. 4. Deformation profiles for beam-sway and column-sway frames [Priestley, 1997].
anisms; the former features a height-dependent shape whilst the latter consists of
a constant linear profile. In fact, at the pre-yield stage, when a plastic deformation
mechanism has not yet been formed, it would seem more logical to assume equal
deformation profiles for both beam- and column-sway cases, both varying with
total height of the building. However, whilst the post-yield deformation profiles of
beam-sway mechanisms may be assumed to follow the same shape as their pre-yield
counterparts (i.e. variable height dependent shape) the same cannot be presupposed
for the case of column-sway structures, where a clearly height-independent defor-
mation shape forms upon yielding at the soft-storey level. If such variation in the
deformation profiles of column-sway mechanisms (from height-dependent at pre-
yield to height-independent in post-yield ranges) were to be taken into account, a
significant increase in the complexity of the capacity equations would result, with
little returns in terms of accuracy. For this reason, the height-independent column-
sway profiles suggested by Priestley [1997], for both elastic and inelastic phases,
have been adopted in the current work.
HT
Deformation profile of
actual structure
HCSF
H SDOF
ef h =
HT
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
HSDOF
Deformation profile of
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
θ equivalent SDOF
of seismic force (Hcsf ), and the total height of the original structure (HT ), as
schematically shown in Fig. 5. Clearly, efh is a function of deformed shape and,
through that, a function also of failure mode, level of ductility, and building height.
As gathered from the discussion in Sec. 2.1, for the case of beam-sway frames,
Hcsf is a function of total height, varying from a value of 0.67HT to 0.61HT as the
number of storeys increases from 4 to 20. Such variation, however, is clearly not
significant, and it was thus suggested by Priestley [1997] that, for regular structures,
displacements be determined considering an average Hcsf = 0.64HT , irrespective
of building height (and also ductility, since elastic and inelastic deformation profiles
have the same shape). Therefore, the effective height coefficient efh for beam-sway
frames can be computed using Eq. (1), defined below as a function of the number
of storeys n.
sake of simplicity, that Eq. (2a) is replaced by Eq. (3a), where instead of ductility,
the steel strain εs(Lsi) corresponding to a given limit state is used to define the
deformation-dependent elastic component of the effective height coefficient. Use of
this alternative representation of efh effectively fixes its value within any given limit
state, minimising considerably the complexity of the resulting equations whilst still
enabling it to vary between limit states and thus allowing for a close approximation
of the resulting capacity curves to their true shape.
εS(Lsi) − εy
efh = 0.67 − 0.17 (3a)
εS(Lsi)
1
efhp = . (3b)
n
rotation at yield and is equivalent to the base rotation of the equivalent SDOF
substitute structure.
∆LSy = efh HT θy . (4)
To compute θy , one must first determine the curvature at yield (Φy ) as a func-
tion of the yield strain of steel (εy ) and the beam section height (hb ) or column
depth (hc ), for the cases of beam- and column-sway mechanisms, respectively. Ref-
erence was made to the work of Priestley [1993, 1998] to find estimates of these
curvatures, given in Eqs. (5a) and (6a), which are then used to define, assuming
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
linearly varying curvature distribution with a zero value at member midspan, the
yield chord rotation (θby or θcy ) expressions shown in Eqs. (5b) and (6b), where lb
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
stands for beam length. The final frame rotation values are obtained after multiply-
ing θby or θcy with empirical coefficients that aim at introducing the effects of flexure
and shear flexibility of joint and framing members. The resulting expressions are
shown by Eqs. (5c) and (6c), for the case of beam- and column-sway mechanisms,
respectively.
For beam-sway frames,
εy
Φy = 1.7 (5a)
hb
1 1b
θby = Φy 1b = 0.283εy (5b)
6 hb
1b
θy = 1.75 × θby = 0.5εy . (5c)
hb
For column-sway frames,
εy
Φy = 2.14 (6a)
hc
1 hs
θcy = Φy hs = 0.357εy (6b)
6 hc
hs
θy = 1.2 × θcy = 0.43εy . (6c)
hc
It is noteworthy that whilst the empirical flexibility/shear augmenting coefficient
introduced in Eq. (5c) follows the proposal of Priestley [1998], its column-based
counterpart [see Eq. (6c)] has been assumed by the authors, for consistency and
conceptual soundness, considering a much lower relative effect of joint and beam
deformability over the calculated column chord rotation. Evidently, future work
should target on the verification, and possible updating, of such scaling factor.
Finally, substituting Eqs. (5c) and (6c) into Eq. (4) results in the expressions for
yield displacement capacity of beam- and column-sway frames, respectively Eqs. (7)
and (8), below.
1b
∆LSy = 0.5efhHT εy (7)
hb
May 5, 2003 16:51 WSPC/124-JEE 00098
hs
∆LSy = 0.43efhHT εy . (8)
hc
It is also important to note that although the two yield displacement expressions
derived above appear somewhat different, they are conceptually identical, reflecting
the fact that at pre-yield stages, when a failure mechanism has not yet been formed,
the displacement capacity of a structure should be obtained using an unique set of
principles and rules for both beam- and column-sway type of buildings. The only
difference between these two equations being that in one it is the characteristics of
the beams that control the capacity, whilst in the other it is the characteristics of
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
of the validity of this parameter within the framework of the current research is an
expected outcome of future investigations. Alternatively, this empirical coefficient
could also be considered as an input parameter to be defined by users on a case-by-
case basis, a solution not adopted here so that the input required for application
of the proposed method is confined to the definition of material and geometrical
properties.
The structural displacement for a given post-yield limit state (∆LSi ) and fail-
ure mechanism, is obtained by adding together the yield and post-yield deforma-
tion components, resulting in the beam- and column-sway inelastic displacement
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
As discussed in further detail in Sec. 5.1, it is clear that the use of such simplistic
relationship between period and height implies a considerable degree of approxima-
tion. Indeed, recent studies [e.g. Priestley, 1998; Priestley and Kowalsky, 2000] have
shown that period of buildings is also highly dependent on strength. However, the
incorporation in the proposed vulnerability assessment methodology of such period-
strength dependence requires extensive additional investigation, clearly beyond the
scope of the current presentation, for which reason the adoption of Eq. (14) was
deemed as appropriate and justifiable.
It is also important to note that the value of parameter α adopted here is
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
modified from that suggested in EC8 [CEN, 2001], where an α value of 0.075 is
used for estimation of period of uncracked buildings. In the present application, the
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
limit state i. These are given below in Eqs. (17) and (18) for beam- and column-sway
mechanisms, respectively.
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
4/3
HT (TLSi ) = C3 TLSi (17)
2
C2 C1
HT (TLSi ) = − (18)
3 efh C2
where the constants C1 , C2 and C3 are defined as follows:
εC(LSi) + εS(LSi) − 2.14εy
C1 = hc (19a)
0.86εy
s 1/3
3
2 C 1 4
C2 = 1350 · TLSi +3 3 + 202500 · TLSi (19b)
efh
states. Similarly, the inclusion of Eqs. (17) and (18) into Eqs. (10) and (11) provides
a similar outcome for post-yield limit states. The resulting capacity curves for both
pre- and post-yield limit states are provided below in Eqs. (20) and (21) for beam-
and column-sway frames, respectively.
For beam-sway frames,
0.5efh1b εy
∆LSy (TLSy ) = (10 · TLSy )4/3 (20a)
hb
4/3 1b
∆LSi (TLSi ) = 0.5efhC3 TLSi εC(LSi) + εS(LSi) − 1.7 − εy . (20b)
hb
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
0.43efhhs εy
∆LSy (TLSy ) = (10 · TLSy )4/3 (21a)
hc
2
0.43efhhs εy C2 C1
∆LSi (TLSi ) = −
hc 3 efh C2
+ 0.5(εC(LSi) + εS(LSi) − 2.14εy )hs . (21b)
It is perhaps worth noting that the efh term included in these capacity equations
relates to the effective height coefficient, defined by Eqs. (1) and (3a), and has not
been explicitly introduced in the expressions given here so as avoid large equation
sizes, unpractical for this type of presentation.
80
Post-yield Capacity
Yield Capacity
60
Displacement [cm]
40
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
20
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
0
0 20 40 60 80
Height [m]
80
20
n>
60
Displacement [cm]
20
n<
4<
40
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
20
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
4
<
n
0
0 20 40 60 80
Height [m]
20
Post-yield Capacity
Yield Capacity
15
Displacement [cm]
10
0
0 20 40 60
Height [m]
for an easier perception of the differences between the yield and post-yield capacity
curves, the current example considers the not-so-common case of a base column
with reduced axial loads.
It is observed that a linear variation of capacity with height is featured by
both limit states. This is to be expected since Eqs. (8) and (11) clearly introduce
height as a linear operator. However, and unlike in the previously discussed beam-
sway case, the post-yield capacity curve does not start at the origin, but rather
May 5, 2003 16:51 WSPC/124-JEE 00098
results that such reduction can only be justified by a decrease of the contribution of
its yield counterpart. As inconsistent as this might seem, it is actually the correct
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
behaviour, particularly if one recalls the fact that, as discussed in Sec. 2.2, the
displacement capacities being computed here are those of the equivalent SDOF
structure, and not those of the actual building.
Observing Eq. (2a) and noting that efh diminishes with increasing ductility,
which in turn decreases with increasing building height or period (see Eq. (13) and
discussion in Sec. 3.1), we see that efh in fact increases with building height or
period, beginning with an initial value of 0.5 at a period of 0 s and approaching
asymptotically a value of 0.67 at larger periods. Since yield displacement component
in Eq. (11) is directly proportional to the effective height coefficient efh , it results
that this elastic component of the post-yield capacity will always be less than the
yield capacity of the same structure, diverging away from it as periods increase from
0 to around 10 s before turning to asymptotically approach it as periods increase
above this value, (see Fig. 9).
Also shown in Fig. 9 is the capacity curve for post-yield limit states assuming
efh as a function of ε(Lsi) , that is, considering efh constant within a given limit
∆
LSi, efh=f(µ)
LSi, efh=f(ε) LSi - inelastic
component
LSy
Displacement
LSi - elastic
component
≈T=5s Height H
state i. This curve, on the other hand, is linear and the figure illustrates the general
range in which this approximation is valid (i.e. from 0 to 5 s), thus fully justifying
its adoption as an alternative for the more elaborate form of Eq. (2a), where ef h is
defined as a function of µ instead.
resented, it is suggested that the demand and capacity curves, defined as a function
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
The equivalent viscous damping values obtained through Eq. (22), for different
ductility levels, can then be combined with Eq. (23), proposed by Bommer et al.
[2000] and currently implemented in EC8 [CEN, 2002], to compute a reduction
May 5, 2003 16:51 WSPC/124-JEE 00098
30
25
20
Damping [%]
15
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
10
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
0
0 2 4 6 8 10
Period [s]
Fig. 10. Variation of damping with period, for the column-sway mechanism.
80
40
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
20
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
0
0 2 4 6 8 10
Period [s]
Fig. 11. Yield and post-yield displacement response spectra for the Erzincan earthquake (Turkey,
1992).
one, due to the decrease in level of damping, as opposed to the beam-sway case
where the inelastic to elastic spectra ratio is always constant.
80
60
Displacement [cm]
40
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
0
0 20 40 60 80
Height [m]
For the sake of simplicity, the outcome of the vulnerability assessment exercised
is given by means of a height range, ∆H, defined as the difference between the two
boundary heights where capacity meets demand (see Fig. 13). Buildings with a total
height within these two values, feature a displacement capacity that is smaller than
the respective demand, effectively meaning that they fail to meet the requirements
of the limit state under consideration. If one considers a uniform distribution of
buildings within a target area, then the higher the value of ∆H, the larger the
number of buildings at risk.
In this short study, the Erzincan record (Turkey, 1992) is used throughout, albeit
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
scaled to varying levels, where convenient. The ensuing results are merely indicative,
serving the single purpose of stimulating discussion on the characteristics of the
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
50
∆H=40 m
40
Displacement [cm]
30
20
ε y = 0.0020
ε y = 0.0016
10
∆H=53 m
0
0 20 40 60 80
Height [m]
Fig. 13. Beam-sway yield capacity/demand curves for different values of steel yield strain (record
scaled down by 50%).
May 5, 2003 16:51 WSPC/124-JEE 00098
the grade of reinforcement steel, the smaller the number of buildings failing a given
limit state.
In Fig. 14, the ultimate concrete strain for the unconfined beams controlling
the deformation capacity of the beam-sway frame is varied between the values of
0.005 and 0.004 (the corresponding steel strains, resulting from section analysis,
are 0.015 and 0.010, respectively). The decrease in deformation capacity of the
concrete material causes, in this particular case, a 30% increase of the number of
buildings not being able to comply with the corresponding demand, confirming the
well-acknowledged fact that older buildings, where little confinement detailing is
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
should not, however, constitute a major difficulty in the application of the proposed
vulnerability assessment methodology, since yield and ultimate strains of reinforce-
ment steel and concrete, respectively, can usually be well estimated, being typically
a discrete function of age and location. Hence, their approximation across a given
structural class is usually a valid and non-problematic operation.
As for the column-sway case, in Fig. 15 the ultimate limit state of the case study
frame is studied for two different levels of axial load (P = 0.2fcAg and P = 0.4fc Ag )
to which correspond, for a constant ultimate concrete strain of 0.005, steel strain
values of 0.010 and 0.005, respectively. It is observed that, despite such significant
variation in ultimate steel strain, the changes in the range of buildings failing the
limit state are minimal, in clear contrast to what has been noted for beam-sway
80
∆H=37 m
60
Displacement [cm]
40
20
ε c = 0.005
εc = 0.004
∆H=49 m
0
0 20 40 60 80
Height [m]
Fig. 14. Beam-sway inelastic capacity/demand curves for different ultimate concrete strain
values.
May 5, 2003 16:51 WSPC/124-JEE 00098
20
∆H=54 m
15
Displacement [cm]
10
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
εs = 0.005
ε s = 0.010
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
∆H=57 m
0
0 20 40 60 80
Height [m]
Fig. 15. Column-sway inelastic capacity/demand curves for different steel strain values (record
scaled down by 75%).
cases. This is not entirely unexpected, as the beam-sway mechanism (being the
preferred mode of inelastic deformation) is a far more rational and ductile means of
inelastic deformation than the column-sway mechanism and as such makes better
use of any addition material ductility.
To be able to introduce a noticeable decrease in the vulnerability of reinforced
concrete buildings responding in a column-sway mode, one would require very high
values of concrete ultimate strain, a feature often unseen in structures designed
and built more than 10/20 years ago. It is therefore concluded that no significant
variations in the assessment results of columns-sway frames are expected to be
caused by the usually small variability in input material properties.
50
∆H=53 m
40
Displacement [cm]
30
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
20
lb = 6.0 m
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
lb = 5.0 m
10
∆H=60 m
0
0 20 40 60 80
Height [m]
Fig. 16. Beam-sway yield capacity/demand curves for different beam lengths (record scaled down
by 50%).
25
∆H=45 m
20
Displacement [cm]
15
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
10
hc = 0.45 m
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
hc = 0.60 m
5
∆H=57 m
0
0 20 40 60 80
Height [m]
Fig. 17. Column-sway yield capacity/demand curves for different column section depths (record
scaled down by 75%).
culties without the necessity for a more refined building type classification, although
increasing significantly the theoretical complexity of the method. In Sec. 5.2, this
option is discussed further.
25
20
Displacement [cm]
15
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
10
α = 0.15
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
α = 0.075
5
0
0 30 60 90 120
Height [m]
Fig. 18. Column-sway yield capacity/demand curves for different α values (record scaled down
by 75%).
80
∆H=37 m
60
Displacement [cm]
40
20 efp = 0.50
efp = 0.25
∆H=55 m
0
0 20 40 60 80
Height [m]
Fig. 19. Beam-sway post-yield capacity curves for different plastic hinge lengths.
May 5, 2003 16:51 WSPC/124-JEE 00098
ticularly hard to quantify value, even for a single building, let alone for entire urban
areas. The relevance of such difficulty comes magnified by the fact that changing
the value of efp between limits often observed in the field, may introduce signif-
icant modifications in the outcome of the assessment exercise, as seen in Fig. 19,
where the adoption of a plastic hinge length equal to a quarter of the section depth,
instead of its half, results in a 50% increase in the number of buildings failing to
meet the limit state under consideration.
Similar comments can be made with regards to the effect of variability in the
joint flexibility/shear coefficients introduced in Eqs. (5c) and (6c), thus highlighting
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
5. Discussion
5.1. Application to design
The most significant weakness in the simplified method described within the current
work is the limitation imposed by assuming a simple and approximate relationship
between period and height. As previously discussed, it is clear that the elastic
period of even regular frames is a function of more than height alone (indeed, most
buildings will even have different natural periods of vibration about their two major
axes), and it is this limitation that prevents one from using the above relationships
and curves for the purposes of design.
In fact, establishing a dependence of period on height alone implies that build-
ings of equal or similar heights and mass, irrespective of material properties and
dimensions, will possess the same elastic stiffness. Thus, as one increases the elastic
displacement capacity (by reducing member depth), one is also increasing the base
shear at yield (due to a constant elastic stiffness based solely on height). Taking
it to the extreme, it observed that this would imply that the smaller the member
depths, the higher the design force; an evidently absurd proposition.
A similar inconsistency is found when looking instead at inelastic periods and
secant stiffness. One of the most significant advantages of direct displacement-based
design, as proposed by Priestley [1997], is its insensitivity to the poorly defined or
unknown elastic stiffness. In the case of the derived capacity curves for ductile
limit states, it is clear to see that ductile effective periods are in fact a function
of elastic stiffness, implicitly determined in the assumed relationship between pe-
riod and height. Thus, base shear is shown once again to be a function of poorly
approximated elastic properties, rather than inelastic displacement and effective
period.
Although the above outlines a significant failing hidden within the proposed
methodology, it is a failing that is only highlighted when its equations are applied
to the assessment of a single structure or when such an application is with intent to
May 5, 2003 16:51 WSPC/124-JEE 00098
design. It must therefore be recalled that the motivation behind the development
of this simplified method was not one of design on a case-by-case basis, but one of
assessment on a large and non site-specific scale. As such, the use of such a simple
relationship between period and height can be judged as fully justified.
firstly, the existence of a known and deterministic relationship between period and
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
height and, secondly, that the various input parameters can be well approximated
over a given building stock. The first of these two assumptions allows the construc-
tion of the capacity-period curves derived in Sec. 2, whilst the second then enables
these curves to be applied indiscriminately across a given building stock.
However, both these assumptions are simplifications of the reality, and neglect
the existence of known uncertainties. Indeed, relationships between elastic period
and height are far from deterministic. Such equations are typically derived from re-
gression analysis with height as the only independent variable and, as such, possess
significant scatter. Whilst it will often be hard to determine single value approxima-
tions of the input parameters that will result in a meaningful estimation of capac-
ity across the entire building stock, the use of statistical distributions could often
capture in a much more credible manner the actual properties of these variables.
Characterising instead the input parameters (including height) as random vari-
ables, defined by known statistical distributions, and including in any period-height
relationship a measure of the scatter, the equations present in the current work could
be used to describe a correlated Joint Probability Density Function (JP DF ) be-
tween capacity and effective period. This JP DF would then be intersected by any
given demand curve in much the same way as are current capacity-period curves.
However, in this instance, the volume of the JP DF above the demand curve pro-
vides the probability of any building in the exposed stock chosen at random in
failing a given limit state, and through that the proportion of buildings in similar
condition.
In this way, the proposed methodology would clearly begin to approach, philo-
sophically at least, the method suggested by Calvi [1999], previously discussed.
However, in this case, the inclusion of sound fundamental relationships between
capacity and height, and between height and period, would provide a conceptually
sounder, more accurate, and ultimately more flexible solution.
6. Conclusions
In this paper, the first steps in the development of an innovative simplified method
for earthquake loss estimation of classes of buildings are presented. The proposed
May 5, 2003 16:51 WSPC/124-JEE 00098
will follow, whereby case study applications used in sensitivity analyses will pursue
the identification of the effect that the variability in the quality of input data, and
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com
the complexity of the modelling adopted, have on the outcome of a loss estimation
exercise.
Further, the paper outlines a deterministic method for displacement-based
earthquake loss assessment across a given region. The next step in the development
of this simplified displacement-based procedure, other than its obvious extension
to other building classes and structural systems (e.g. masonry structures, wall sys-
tems, infilled frames, shear failure mechanisms), must be the incorporation of a
probabilistic representation of both the period-height relationships and the input
variables to the capacity curves.
Acknowledgments
The authors are especially grateful to Dr. Julian Bommer, from Imperial College
of London, for his contributions to the development of this work. The basic idea
of the simplified approach presented in this paper was born of long and interesting
discussions with him, and his enthusiastic support and insightful suggestions, as
well the thorough reviews of parts of the manuscript, were extremely valuable for
the completion of the work presented herein. The authors would also like to thank
Professor Nigel Priestley for his helpful comments and suggestions, and Ms. Helen
Crowley, for her help in the verification of the derived formulae.
References
Bommer, J. J., Elnashai, A. S and Weir, A. G. [2000] “Compatible acceleration and
displacement spectra for seismic design codes,” Proc. 12th World Conference on
Earthquake Engineering, Auckland, New Zealand.
Bommer, J., Spence, R., Erdik, M., Tabuchi, S., Aydinoglu, N., Booth, E., del Re, D. and
Peterken, O. [2002] “Development of an earthquake loss model for Turkish catastrophe
insurance,” J. Seism. 6(3), 431–446.
Calvi, G. M. [1999] “A displacement-based approach for vulnerability evaluation of classes
of buildings,” J. Earthq. Engrg. 3(3), 411–438.
Calvi, G. M. and Pavese, A. [1995] “Displacement based design of building structures,”
Proc. of the Fifth SECED Conference on European Seismic Design Practice, Chester,
United Kingdom, pp. 127–132.
May 5, 2003 16:51 WSPC/124-JEE 00098
CEB [1994] Behaviour and Analysis of Reinforced Concrete Structures Under Alternate
Action Inducing Inelastic Response. Bulletin No. 220, Comité Euro-International du
Béton, Lausanne, Switzerland.
CEN [2002] European Prestandard ENV 1998-1-4: Eurocode 8 — Design provisions for
earthquake resistance of structures, Part 1–4: Strengthening and repair of build-
ings. Draft No. 5, May 2002, Doc CEN/TC250/SC8/N317, Comite Europeen de
Normalisation, Brussels.
Faccioli, E., Pessina, V., Calvi, G. M. and Borzi, B. [1999] “A study on damage scenarios
for residential buildings in Catania,” J. Seism. 3, 327–343.
Fajfar, P. [1999] “Capacity spectrum method based on inelastic demand spectra,” Earthq.
Engrg. & Struct. Dyn. 28(9), 979–993.
by UNIVERSITY OF CALIFORNIA @ SAN DIEGO on 06/05/15. For personal use only.
and mid-rise buildings in Japan,” Earthq. Engrg. & Struct. Dyn. 9, 93–115.
Shinozuka, M., Chang, S. E., Eguchi, R. T., Abrams, D. P., Hwang, H. H. M. and
J. Earth. Eng. 2003.07:107-140. Downloaded from www.worldscientific.com