Professional Documents
Culture Documents
G R A H A M H. POWELL* A N D RAKESH A L L A H A B A D I ~
Department of Civil Engineering, Unioersity of California, Berkeley, CA 94720, U.S.A.
SUMMARY
Although the prediction of seismic damage is primarily a probabilistic problem, deterministic analysis can neverthelessbe a
valuable tool. This paper reviews the concepts which have been used in damage prediction, and identifies and classifies
several specific techniques. Techniques of two types are considered, the first based on the balance between some demand on
the structure and its corresponding capacity, and the second on the degradation of some structural property. For each
technique, one or more damage parameters can be computed, and from these a damage index can be estimated. Several
alternative choices for the damage parameters are identified. Procedures for computing values for these parameters, and
relating them to damage indices, are suggested.
INTRODUCTION
The prediction of the amount of seismic damage that a structure is likely to sustain during its lifetime is a
probabilistic problem. Nevertheless, deterministic analysis can be a valuable tool. This paper reviews concepts
and techniques for the use of deterministic analysis in seismic damage prediction.
Two types of procedure are considered, the first based on the balance between some demand o n the structure
and its corresponding capacity, and the second o n the degradation of some structural property. For each
procedure, one o r more dnmnge pnratneters can be computed, and from these a structural damage index can be
estimated. This index can then be related to a n economic damage index. This paper identifies a number of
alternative choices for the damage parameters, describes how the values of these parameters might be
computed, and suggests relationships between damage parameters and damage indices.
Although some damage parameters can be computed by linear structural analysis, the more rational
approach is to use inelastic analysis. A number of computer programs are able t o perform inelastic static and
dynamic structural analyses. While such programs have been widely used in research, they have not found wide
use in formal damage prediction. Also, such programs may need to be modified so that the structural response
quantities needed to compute damage parameters are included in their output. This paper identifies several
features which might need to be added to a structural analysis program to support damage computation.
It is important t o emphasize that there are many uncertainties in damage computation, particularly in the
modelling assumptions for structural analysis and in the choice of earthquake ground motions. The actual
strengths of structural materinls and members may be very different from those assumed for analysis, and in
fact will usually not be known. F o r example, Reference 7 reports in-situ tests showing concrete compressive
strengths of 8000 psi where the design value was 3000 psi, and steel yield strengths ranging from 56 to 62.5 ksi
where the nominal value was 50 ksi. It is also difficult to identify those characteristics of ground shaking which
describe the damaging power of an earthquake, yet ground motion is the biggest single source of uncertainty.'
C'hopra and Kan' and Powell and Row29 studied the effects of varying the ground motion a n d modelling
assumptions for analysis, and both concluded that uncertainties in the ground motion are more critical.
It is also difficult tocorrelatecomputed response with damage. Damage will typically be required in terms of
an economic damage index, which is the ratio of repair cost to replacement cost.4, l o , 1 6 * ' 9 3 0 , 3 2 , 3 3 It 1s
''9 '
difficult, however, to identify structural response parameters which can be correlated reliably with such a
damage index. Parameters such as base shear,". 39 d i ~ p l a c e m e n t26~ ~inter-storey
. drift,', 1 3 * 15, 1 6 , * O ductility
ratio2*'* 2 3 * 34 and energy dissipation 2 . 9 . 2 2 * 3 1 have been used, yet there are insufficient data available to
provide consistent correlation between these parameters and economic damage.
DAMAGE CONCEPTS
General
The problem of damage definition and computation is complex, and this paper does not claim to offer
definitive solutions. Rather, the emphasis is on reviewing and clarifying concepts and on identifying structural
response quantities which might be used to compute damage. These response quantities will be termed damage
parameters. Further, although an economic damage index will ultimately be required, the emphasis in this paper
is on the structural damage index, and on the relationship between this index and the damage parameters. The
establishment of a relationship between the structural and economic damage indices is then a separate (and
complex) problem, which is addressed only superficially.
There are two basic procedures which can be used to compute a structural damage index, the first based on
demand versus capacity and the second based on degradation. The concepts underlying these two procedures
are discussed in the following sections. Details for specific damage parameters are then considered.
Degradation procedure
The degradation procedure is based on estimation of a property for a structure, substructure or member in
its undamaged state, and a corresponding estimation in its damaged state. Possible choices for the structural
property include stiffness, strength and energy dissipation capacity. All three of these were used in
Reference 20, and an interesting correlation between energy dissipation and stiffness degradation has been
reported in Reference 18.
where 6 is the damage parameter, P, is the value of the structural property in the damaged state and Po is the
value in the undamaged state. The structural damage index must be expressed as a function of the damage
parameter (or as a function of more than one parameter).
SEISMIC DAMAGE PREDICTION 72 1
For each damage parameter there will be a threshold value and a critical value. The structural damage index
will be zero if the calculated value of the parameter is less than the threshold value, implying no significant
damage. The damage index will be unity when the calculated value of the parameter is equal to its ultimate
value. implying failure. In between, the damage index must be related to the parameter by some damage
function. The simplest function is the linear relationship
where DI, is the structural damage index; and 6,, 6, and 6, are the calculated, threshold and ultimate values,
respectively, for the damage parameter. A rather more general form is
DI, = (-)
S,-b,
(3)
6, - 6,
The effect of the exponent, m, is shown in Figure 1. Other relationships could, of course, be used.
I .o
0.0
Figure I . Effect of exponent, m. on relationship between damage index and damage parameter
If damage is based on the maximum value or maximum range of the damage parameter, then 6, will be this
maximum. If a cumulative value is used, 6, may be based on a formal damage theory. One such theory is
reviewed in the following section.
where 6, is the cumulative value of the damage parameter, A and bare properties of the structure, and hi is the
value of the damage parameter for the ith half cycle. It can be shown that
1
A = - d- b - I (5)
U
The effect of the property b is as follows. For a given value of E d , , the amount of damage will usually be
larger for a small number of large cycles than for a large number of small cycles. In equation (6)this implies a
value of b > 1. If a structure or structural component were subjected to cyclic deformation at constant
amplitude, with an inelastic deformation equal to S per half cycle, then from equation (6)
h c = S u N ( ~ ~ (7)
where N is the number of half cycles. At failure, 6, = S , and N = N u = number of half cycles to failure. Hence
b
Nu(:;) = 1
or logNu+blog(:;) =0
This equation defines a line similar to the classical S-N line for high cycle fatigue of steel components.
However, the slope of the line is different from that for a typical S - N curve. For low cycle damage of structural
components a typical value of b is 1.6 or 1.8,”. 2 6 whereas for high cycle S-N curves the corresponding value is
2-5 or 3.
Damage parameters which can be calculated from response quantities for the complete structure, or which
require only a minimum amount of information from individual elements, are considered in this section.
Parameters which require more detailed element information are considered later.
Strength demand
If the strength demand on a structure, calculated assuming elastic behaviour, does not exceed the yield
strength, damage to the structure will be small. On the other hand, if the calculated demand approaches, or
exceeds the ultimate strength, damage may be high. Hence, strength demand can be used as a damage
parameter. Hasselman et al.“ suggest that strength demand, as measured by base shear, is an appropriate
damage parameter for low-rise buildings (less than 5 storeys).
Lew and Takahashi” have developed a ‘Rapid Seismic Analysis Procedure’ to assess damage, in which the
elastic base shear demand is used as a damage parameter, an estimate of the yield base shear is taken as the
threshold value and an estimate of the ultimate base shear is taken as the ultimate value. The damage index is
calculated by equation ( 2 ) .The strength demand is calculated using an elastic response spectrum analysis, in a
way which accounts approximately for inelastic behaviour. Procedures to estimate the yield base shear and
ultimate base shear are also outlined in Reference 21.
There are two major problems with this type of approach. First, the procedures used to estimate the yield
and ultimate base shears are very approximate. Second, elastic base shear demands greater than the ultimate
base shear do not necessarily imply complete failure. Nevertheless, the procedure is simple, and can be
appropriate for preliminary damage estimates.
Displacernentldeforination demand
Ductility is the ability of a structure or structural component to deform inelastically without fracture, and
preferably without substantial loss of strength. It has been common to express displacement and deformation
demands in terms of ductility ratios, calculated as the ratio of maximum deformation to deformation at first
yield.
At the structure level, displacement demand is typically expressed as a displacement ductility ratio. For a
single degree of freedom (SDOF) system, this ratio, p, is defined as
p=F Y
(9)
1) Estimate yield values of storey drift for each storey. Let r y be the corresponding displacement vector.
7 24 G . H. POWELL AND R. ALLAHABADI
Lo I
‘Y
DISPLA~EMENT
Lo
(a) Based On First Yield
4
Collapse Load
r,, DISPLACEMENT
J
(b) Based On Collapse Load
LoA I
‘Y
Equal Areas
ry r e = rm DISPLACEMENT
(a) Elastic Displacement = Inelastic Displacement
LOAD
al Areas
Re
rY re rm DISPLACEMENT
(b) Elastic Energy = Inelastic Energy
rY rc m
‘ DISPLACEMENT
(c) Elastic Force = Inelastic Force
= J(2)
For a SDOF building, this corresponds to the first of the above assumptions.
An approach based on the second of the assumptions has been proposed by Blume et aL4 By equating the
<
shaded areas in Figure 3(b), it can be shown, for # 0, that
1
Guidelines for obtaining the yield base shear (R,)and the base shear demand (R,)are described in Reference 4.
The ductility ratio was used as the damage parameter, and a damage index was calculated using equation (3).
The threshold value, 6,, was taken as one, and ultimate values, 6,, were recommended for high-rise, low-rise
industrial and low-rise residential buildings.
A further simple method has been proposed by Hasselman et al." as follows.
1) Estimate an interstorey drift ratio, A, using a procedure described in Reference 16. This procedure
assumes that the first mode response is dominant, that the mode shape is a straight line and that the mass
of the building is uniform over its height.
2) Compute a drift ductility, p = A,'Ay, where A y = 0.0044 for concrete buildings and 0.0077 for steel
buildings.
3) If p is greater than one, assume that the response is inelastic. A and p are then adjusted by a procedure
outlined in Reference 16.
4) Calculate an economic damage index using a formula given in Reference 16..
All of the above methods use simplified analyses to compute the displacement ductility. More sophisticated
approaches, making use of inelastic static or dynamic analyses, could be devised. However, if such analyses
were carried out it would probably be inconsistent to attempt to reduce the results to a single ductility ratio
based on equation (9). Instead, it would be more appropriate to consider separate substructures or members.
DISPLACEMENT
considered at the structure level, although no specific procedure appears to have been proposed. A possible
procedure is as follows.
1) Estimate the energy dissipation capacity of the structure. One way of doing this is to apply some fixed
gravity load, select a reasonable lateral load pattern (e.g. one based on design code requirements3*), then,
using inelastic static analysis, load the structure laterally to progressively increasing load or displacement
levels. At each level, calculate the dissipated energy (areas under the load-displacement curves, probably
for the lateral loads only), and hence obtain a relationship between dissipated energy and load (or
displacement). From this relationship estimate an ultimate capacity.
2 ) Perform either (a) an inelastic dynamic analysis or (b) a cyclic inelastic static analysis (following a
predetermined load or displacement sequence), and compute the total dissipated energy. For a static
analysis this can be done as in Step 1. For a dynamic analysis the external work can not be used directly,
because it includes elastic work, damping work and kinetic energy as well as hysteretic work. In general it
will be necessary to sum the hysteretic work from individual members.
3) Select as the damage parameter the ratio of hysteretic work to energy capacity. The threshold value, S,,
will presumably be zero. The ultimate value, S, could exceed one, since a structure may have a larger
dissipation capacity when loaded cyclically than when loaded monotonically.
There are likely to be several difficulties in applying this procedure, particularly in estimation of the
dissipation capacity and the choice of 6,. Also, if an inelastic analysis is to be carried out, it might be more
appropriate to compute damage at the substructure or member level rather than at the structure level.
where b, and b, are arbitrary vectors (with the same units), R is a load vector and r is the displacement vector
obtained by analysing the structure for R. One choice is bl = b, = R. This gives
R ~ R
KR=-
RTr
The ratio of this scalar stiffness in the damaged state to its value in the undamaged state, calculated using
SEISMIC DAMAGE PREDICTION 727
tangent stiffnesses,has been termed the current st@ness parameter by berg at^.^ From equation (1 5 ) the current
stiffness parameter, S,, is defined as
(RTR)/(RTrd)--RTr,
S, = -
(R TR)/(R Tro) R Trd
where subscripts d and 0 indicate the damaged and undamaged states, respectively. Since rd = K d ' R and
r, = KU'R, equation (16) can be written as
and
s, = K,d
-
K40
If 9 is mass normalized, then since 4 T K 4 = r0'9~M+, it follows that
QZd.9&Po
s+=z
where (od and roo = circular frequencies in the damaged and undamaged states, respectively.
A scalar secant, rather than tangent, stiffness can also be calculated, as follows.
If this calculation is performed in both the damaged and undamaged states, a secant stiffness ratio can be
calculated. For damage estimation this will tend to be more useful than a stiffness ratio based on tangent
stiffnesses, since it accounts for a larger part of the load-displacement relationship.
Once calculated, a stiffness ratio could be used to assess damage as follows.
1) Calculate the scalar stiffness in the undamaged state.
2 ) Perform an inelastic dynamic analysis or a cyclic static analysis to damage the structure.
3) Recalculate the scalar stiffness in the damaged state.
Hence obtain the damage index using equations (1)and (2). The threshold value, a,, would presumably be zero
(i.e. no stiffness degradation), and an ultimate value, 6,, presumably less than one, would have to be chosen.
It is important to note that this approach can be applied only ifstigness degradation can be modelled
appropriately in the analysis. The structural elements used in computer programs for seismic analysis often do
not include stiffness degradation.
728 G . H. POWELL A N D R. ALLAHABADI
Strength degradation
Although it does not appear to have been used, strength degradation could be an appropriate damage
parameter. The procedure might be as follows:
I ) Calculate a measure of the strength of the structure, Po, in the undamaged state, using static inelastic
analysis. A suitable measure might be ultimate base shear, calculated for some distribution of lateral load
and some (fixed) magnitude of gravity load.
2) Perform either an inelastic dynamic analysis or a cyclic static analysis to damage the structure.
3 ) Recalculate the damaged strength, P,, using a second static analysis.
Hence obtain the damage parameter using equation (2). The threshold value, 6,, would presumably be zero (i.e.
no strength loss). An ultimate value, 6,, presumably less than one, would have to be chosen.
This method can be applied only if strength degradation can be modelled appropriately. Also, as noted
before, if a detailed inelastic analysis is to be carried out, damage computation based on substructures, or
members might be more appropriate.
1) Load the undamaged structure statically to a specified displacement, either monotonically or through a
load cycle. Calculate the work done (in the case of monotonic loading, the energy absorbed; in the case of
a load cycle, the energy dissipated).
2) Perform an inelastic dynamic analysis, or a cyclic static analysis, to damage the structure.
3) Reload the structure statically to the same displacement as in Step 1, and recalculate the work done.
4) Base the damage parameter on the ratio of the work done on the damaged structure in Step 3 to that done
on the undamaged structure in Step 1.
This procedure will require threshold and ultimate values for the damage parameter. Other problems to be
faced are the selection of the load distribution and displacement magnitude, and the relationship between the
damage parameter and the work quantities. Again, if inelastic dynamic analysis is to be carried out, more
sophisticated methods might be more appropriate.
1) If the damage parameter were the drift of a storey substructure, this drift could be obtained from the
nodal displacements, which are response quantities for the complete structure.
SEISMIC DAMAGE PREDICTION 729
2) If the damage parameter were the energy dissipated in the girders and columns of a storey, this could be
obtained from the storey shear vs storey drift relationship. The storey shear would generally have to be
obtained by summing the shears in the column elements.
3) If the damage parameter were the energy dissipated in the columns alone, it would be necessary to
compute the energies for the column elements, and add them to get the storey value.
Some computer programs deal with substructures at intermediate levels in the assembly of the complete
structure properties. if the substructures used for such an analysis were to correspond to the substructures of
interest for damage prediction, it would be possible to use the substructure results directly. In general, however,
the analysis substructures are likely to differ from those required for damage prediction. Hence, substructure
quantities for damage prediction must generally be obtained as indicated in the above examples.
I t may also be noted that there will not always be a one-to-one correspondence between analysis elements
and structure members, since a member may be modelled using several elements. In general, therefore, member
quantities for damage prediction may involve the results for more than one element.
The demand-vs-capacity procedure can be applied directly to substructures and members provided (a) that
values of the chosen damage parameter can be obtained from either the structure or element responses, and
(b) that suitable threshold and ultimate values can be specified. The degradation procedure is more difficult to
apply, because it depends on calculating a structure property, such as strength, for the substructure or member
in both its undamaged and damaged states. In some cases it may be possible to infer a value for the property
from the structure or element responses. In general, however, it will be necessary to analyse the substructure or
member as a separate structure, isolated from the complete structure. For example, to calculate the degraded
strength of a girder member it will generally be necessary to model the member as a separate structure, and load
it to determine its strength. While this can be done relatively easily for an undamaged member, computer
programs generally do not include options for separating out members in the damaged state. This is because
there is no simple way to isolate one member from the damaged structure, since the structural elements will
have undergone inelastic deformations and may have complex initial states, with residual forces and
deformations.
At the structure level, the following damage parameters were considered in the preceding section:
(1) strength demand (calculated by elastic analysis), ( 2 ) displacement demand, (3) deformation demand
(average storey drift), (4) energy dissipation demand, (5) tangent stiffness degradation, (6) secant stiffness
degradation, (7) strength degradation, and (8) degradation in energy dissipation capacity. In principle, all of
these parameters can also be considered at the substructure and member levels. In practice, however, some of
the parameters may be difficult to calculate. In the following sections, each of the parameters is considered, for
the particular examples of (a) a substructure consisting of one storey of a frame, and (b) a girder member in
which plastic hinges may form at the member ends.
Strength demand
For a storey substructure the most likely strength parameter is the storey shear, and for a girder it is the
member end moment. The demand will be calculated by elastic analysis, and the capacity must be estimated.
The main problem with this approach is that it may not be rational to base damage estimates on elastic
demands.
Displacement~deforma~ion deinand
The deformation demand on a storey will most likely be the storey drift.**1 3 . 15. 1 6 . 4 0 The deformation
demand on a girder could be total member end rotation or (more likely) plastic hinge rotation. Estimation of
the capacities will require judgement, but otherwise the procedures should be straightforward. Note, however,
that the two chosen examples are simple ones. If a column rather than a girder, were chosen, it might be
necessary to account for the effect of axial force on the hinge rotation capacity. This would be a substantial
complication.
730 G. H. POWELL A N D R. ALLAHABADI
relationships for substructures or members. The difference is in whether the relationship is processed to obtain
information on stiffness, strength or energy dissipation.
Summary
Damage indices computed at the substructure or member levels can provide more detailed damage
information than those computed directly at the structure level. In particular, if damage is localized rather than
distributed more or less uniformly, damage indices computed at the structure level may be unreliable.
However, whereas there are several well defined techniques for computing damage indices at the structure
level, the techniques at the substructure and member levels are less well defined. A major reason for this is that
damage parameters at the structure level can be computed by performing relatively simple operations on the
analysis results for the complete structure. At the substructure and member levels, however, it is necessary to
perform more complex operations, which are generally not supported by the available computer programs.
Even if damage parameters of many types could be computed routinely, major difficulties remain in
choosing parameters which correlate closely with damage, in selecting threshold and ultimate values for the
parameters, and in correlating damage parameters with damage indices.
Of all the possible damage parameters, those based on deformation demand appear to be the best choice, in
terms of both simplicity and rationality. Those based on degradation are attractive intuitively, but they depend
greatly on the element modelling assumptions, and they can be difficult to compute.
i=I
i w,
where wi is the weight assigned to substructure or member i. The weighting factor can reflect the replacement
cost and/or the relative importance of the substructure in maintaining the integrity of the structure. For
example, the lower storeys of a building might be assigned more importance than the upper storeys. The
weighting factor for any storey could also depend on the magnitude of the damage index for that storey, so that
severely damaged storeys are weighted more heavily. One approach, used, for example, by Park et ~ l . , ~is’ to
make wi proportional to DI,. The equation for the overall index is then
n
i=I
Member damage indices can similarly be combined directly to obtain an overall structure index.
Alternatively, member indices can first be combined to obtain a storey index, and the storey indices can then be
combined to obtain an overall index. A procedure of this type was used by Czarnecki’.
usual to make a distinction between them, and they are typically assumed to be equal. This is a questionable
assumption. However, the issue is a complex one, and is not addressed further herein,
NON-STRUCTURAL DAMAGE
The preceding sections have considered structural damage. Prediction of non-structural damage is equally
important, and can generally be based on simpler criteria. There are two main types of non-structural
component in a building, the first being sensitive primarily to floor motions, and the second being sensitive
primarily to storey drift.
Components of the first type include equipment, appendages, electrical systems and elevators. For these
components, Culver et aL8 selected peak floor acceleration and peak floor velocity as damage parameters.
Similar procedures have been described by Hasselman et and Bouhafs.' Relationships between damage
and floor acceleration have also been proposed by Ferrite.'
Components of the second type include window glass, partition walls, and exterior cladding. There is
reasonable agreement on the relationships between drift and damage.I3. Is. l 6 For example, for a good quality
component it is typically assumed that there is no damage for drift ratios less than 0.001, and 100 per cent
damage for drift ratios greater than 0.02.
CONCLUSION
This paper has attempted to establish a framework for the estimation of damage. A damage theory can be
formulated in terms of a damage index. This index is a function of one or more damage parameters, which are
structural response quantities that can be computed and correlated with damage. The damage parameter may
be a single maximum value, a maximum range or a cumulative value. Damage parameters can be defined at the
structure, substructure or member level, and corresponding damage indices can be computed. Two basic
procedures for calculating damage indices have been identified, based on demand vs capacity and on
degradation. A procedure for obtaining an overall damage index from the substructure and member indices
has been outlined.
A distinction has been made between structural damage index and economic damage index, to emphasize
that they are different concepts. In practice, however, the two are likely to be assumed to be equal.
Damage parameters defined at the structure level make use essentially of structure results (e.g. node
displacements, work done by applied loads). The calculation of these parameters tends to be relatively simple,
and several specific procedures are available. However, damage parameters at the structure level provide little
or no information on the type and location of the damage. Damage parameters defined at the substructure and
member levels must generally make use of element response results (e.g. plastic hinge rotations, strength
degradation). These parameters can provide more detailed damage information. However, they tend to be
more difficult to calculate, and relatively few specific procedures are available.
In the demand vs capacity approach it tends to be relatively easy to compute the demand, but more difficult
to estimate the capacity. In the degradation approach the reverse tends to be the case. The degradation
approach is intuitively attractive, but the amount of degradation can depend greatly on the modelling
assumptions.
Finally, it must be kept in mind that there are many uncertainties, and that accuracy in a deterministic
analysis may be more illusory than real. The illusion of accuracy can be particularly dangerous when detailed
damage computations are performed at the member level. Ideally, the modelling and analysis assumptions
which control the uncertainty should be identified, and those assumptions should be varied to assess their
effects on the result. Formal Monte Carlo techniques could even be applied. With the present state of the art
this would be extremely expensive, but if the modelling, analysis and damage computation tasks were
automated, it could become feasible.
ACKNOWLEDGEMENT
The research described in this paper was performed under National Science Foundation Grant No. CEE
8414905. The support of the Foundation is gratefully acknowledged.
SEISMIC DAMAGE PREDICTION 733
REFERENCES
I. M. A. Austin, K. S. Pister and S. A. Mahin,’A methodology for computer-aideddesign ofearthquake-resistant steel structures’, Report
N o . U C B / E E R C - 8 5 /13. Earthquake Engineering Research Center, University of California. Berkeley, CA, 1985.
2. H. Banon, J. M. Biggs and H. M. Irvine, ‘Seismic damage in reinforced concrete frames’, J . struct. diu. ASCE, 107, 1713-1729 (1981).
3. P. Bergan, ‘Solution algorithms for nonlinear structural problems’, Proc. int. conf. eng. appl. finite element method Hovik, Norway
(1979).
4. J. A. Blume, R. E. Scholl and P. K. Lum, ‘Damage factors of predictingearthquake dollar loss probabilities’, Prepared Under Contract
No. 14-08-0001-1588 for the U.S. Geological Survey, URS/John A. Blume & Associates, Engineers, San Francisco, 1977.
5. M. Bouhafs, ‘Evaluation of the seismic performance of buildings’, Techniques rapid assessment seismic vulnerability, ASCE struct. cony.
New Orleans (1986).
6. A. K. Chopra and C. Kan, ‘Effectsof stiffness degradation of earthquake ductility requirements for multistory buildings’, Earthquake
eng. struct. dyn. 2, 3-5 (1973).
7. Computech Engineering Services, Correlutions of Inelastic Analysis and Destructiue Tests on a Reinforced Concrete Building, Berkeley,
CA, (1982).
8. C. G . Culver, H. S. Lew, G. C. Hart and C. W. Pinkham, ’Natural hazard evaluation of existing buildings’, Building Science Series N o .
61, US. National Science Bureau of Standards, Washington, D.C., Jan. 1975.
9. R. M. Czarnecki, ‘Earthquakedamage to tall buildings’, Report No. R73-8, Department of Civil Engineering, M.I.T., Cambridge, MA,
1973.
10. R. M. Czarnecki, R. E. Scholl and L. E. Malik, ‘Techniques for estimating earthquake probable loss for buildings and contents’,
Techniques rapid assessment seismic vulnerability, ASCE struct. cong. New Orleans (1986).
1 I . N. E. Dowling, ‘Fatigue failure prediction for complicated stress-strain histories’, T . & A. M. Report N o . 337. Department of
Theoretical and Applied Mechanics, University of Illinois, Urbana-Champaign, 1971.
12. M. Bassam El-Hafez and G . H. Powell, ‘Computer aided ultimate load design of unbraced multistory steel frames’, Report No.
UCBIEERC 73-3, Earthquake Engineering Research Center, University of California, Berkeley, CA 1973.
13. J. M. Ferrito, ‘Economics of seismic design for new buildings’, J . struct. diu. ASCE, 110, 2925-2938 (1984).
14. M. Fintel and S. K. Ghosh, ‘Explicit inelastic dynamic design procedure for aseismic structures’, ACI J . , Title No. 79-12, Mar.-Apr.
(1982).
15. H. Guangqian, et a/.,‘A new procedure for aseismic design of multi-storeyed shear type buildings based upon deformation checking’,
Proc. 8 t h world conJ earthquake eng. San Francisco, 5, 4 8 3 4 9 0 (1984).
16. T. K. Hasselman, R. T. Eguchi and J. H. Wiggins. ‘Assessment ofdamageability ofexisting buildings in a natural hazardsenvironment’.
Technical Report N o . 80-1332-1, J. H. Wiggins Company, 1980.
17. G . W. Housner and P. C. Jennings, ‘Earthquake design criteria for structures’, C. 1. T. Report EERC 76-06, Pasadena, CA, 1977.
18. H. Iemura, ’Hybrid experiments on earthquake failure criteria of reinforced concrete structures’, Proc. 8th world conf earthquake eng.
San Francisco 6, 103-110 (1984).
19. H. Krawinkler, ‘Selection of loading histories for seismic testing of structural components’, Proc. S E S A - J S M E int. conf: exper. mech..
Hawaii (1982).
20. H. Krawinkler and M. Zoheri, ‘Cumulative damage in steel structures subjected to earthquake ground motions’, Comput. struct. 16,
531-541 (1983).
2 I , T. K. Lew and S. K. Takahashi, ‘Rapid seismic analysis procedure’, Report N o . 5 1-78-02, Naval Facilities Engineering Command, Port
Hueneme, CA, 1978.
22. J. Lin and S. A. Mahin, ‘Effect of inelastic behavior on the analysis and design of earthquake resistant structures’, Report N o .
U C B / E E R C - 8 5 / 0 8 ,Earthquake Engineering Research Center, University of California, Berkeley, CA, 1985.
23. S. A. Mahin and V. V. Bertero, ‘Problems in establishing and predicting ductility in aseismic design’, Proc. int. con$ earthquake struct.
eny. St. Louis, (1976).
24. S. A. Mahin and J. Lin, ‘Construction of inelastic response spectra for single degree of freedom systems’, Report N o . UCBIEERC-
83117, Earthquake Engineering Research Center, University of California, Berkeley, CA, 1983.
25. N. M . Newmark and W. J. Hall, ‘Procedures and criteria for earthquake resistant design’, National Bureau of Standards Building
Science Series 46, Building Practices for Disaster Mitigation, Proc. workshop sponsored by the National Science Foundation and the
National Bureau of Standards, Boulder, Colorado (1972).
26. Y. J. Park and A. H. S. Ang,‘Mechanistic seismic damage model for reinforced concrete’,.!. struct. div. ASCE Eng. I 1 1,722-739 (1985).
27. Y. J. Park, A. H . S. Ang and Y.W. Wen, ‘Seismic damage analysis of reinforced concrete buildings, J . struct. diu. ASCE, 111,740-757
(1985).
28. G . H. Powell and M. Bassam El-Hafez, ‘Subassemblage concept for tall frame analysis’, J . struct. diu. ASCE. 100, 1611-1625 (1974).
29. G . H. Powell and D. G . Row, ‘Influence of design and analysis assumptions on computed inelastic response of moderate tall frames’,
Report N o . EERC 76- 1 1, Earthquake Engineering Research Center, University of California, Berkeley, CA. 1976.
30. C. Rojahn, ’ATC methodology and data for rapid assessment of seismic vulnerability’, Techniques rapid assessment seismic
uulnerubility, ASCE struct. cong. New Orleans (1986).
31. P. Ruiz and J. Penzien, ‘Probabilistic study of the behavior of structures during earthquakes’, Report N o . EERC 69-3, Earthquake
Engineering Research Center, University of California, Berkeley, CA, 1969.
32. T. A. Sabol and G . C. Hart, ‘Seismic risk analysis of a multi-site portfolio of buildings’, Techniques rapid assessmenf seismic
vulnerability, ASCE sfruct. cong. New Orleans (1986).
33. Charles Scawthorn, ‘Rapid assessment of seismic vulnerability’, Techniques rapid assessment seismic vulnerability, A S C E strucf.cony.
New Orleans (1986).
34. R. E. Scholl, ‘Seismic damage assessment for high rise buildings’, Prepared Under Contract No. 14-08-001-16814 for the U. S.
Geological Survey, URS,/John A. Blume & Associates, San Francisco, 1979.
35. R. Shepherd and P. C. Jennings. ‘Experimental investigation-correlation with analysis’, Workshop earthquake resistant reinforced
concrete build. constr. University of California, Berkeley, CA (1977).
36. J. Simons and G. H. Powell, ‘Solution strategies for statically loaded nonlinear structures’, Report No. U C B / E E R C - 8 2 / 2 2 ,Earthquake
Engineering Research Center, University of California, Berkeley, CA, 1982.
734 G. H. POWELL A N D R. ALLAHABADI
37. M. T. Suidan and R. A. Eubanks, 'Cumulative fatigue damage in seismic structures', J . struct. d i n A X E , 99,923-943 (1973).
38. Uniform Building Code, International Conference of Building Officials, Whittier, CA, 1985.
39. R. V. Whitman, et a/., 'Seismic design analysis; methodology and pilot application', Report No. MIT-CE-R74-16, Department of Civil
Engineering, M.I.T., Cambridge, MA, 1974.
40. E. H. Wong and R. V. Whitman, 'Correlations between earthquake damage and strong ground motion', Report No. R75-23, M.I.T.
Department of Civil Engineering, Cambridge, MA, 1975.