Вы находитесь на странице: 1из 20

Engineering Structures 23 (2001) 407–424

www.elsevier.com/locate/engstruct

Static pushover versus dynamic collapse analysis of RC buildings


A.M. Mwafy, A.S. Elnashai *
Department of Civil and Environmental Engineering, Imperial College, Imperial College Road, London SW7 2BU, UK

Received 2 February 2000; received in revised form 23 May 2000; accepted 26 May 2000

Abstract

Owing to the simplicity of inelastic static pushover analysis compared to inelastic dynamic analysis, the study of this technique
has been the subject of many investigations in recent years. In this paper, the validity and the applicability of this technique are
assessed by comparison with ‘dynamic pushover’ idealised envelopes obtained from incremental dynamic collapse analysis. This is
undertaken using natural and artificial earthquake records imposed on 12 RC buildings of different characteristics. This involves
successive scaling and application of each accelerogram followed by assessment of the maximum response, up to the achievement
of the structural collapse. The results of over one hundred inelastic dynamic analyses using a detailed 2D modelling approach for
each of the twelve RC buildings have been utilised to develop the dynamic pushover envelopes and compare these with the static
pushover results with different load patterns. Good correlation is obtained between the calculated idealised envelopes of the
dynamic analyses and static pushover results for a defined class of structure. Where discrepancies were observed, extensive
investigations based on Fourier amplitude analysis of the response were undertaken and conservative assumptions were
recommended. 2001 Elsevier Science Ltd. All rights reserved.

Keywords: Pushover analysis; Time–history collapse analysis; RC buildings; Fourier amplitude analysis
408 A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424

1. Introduction applying a predefined lateral load pattern which is


distributed along the building height. The lateral forces
Inelastic time–history analysis is a powerful tool for are then monotonically increased in constant proportion
the study of structural seismic response. A set of with a displacement control at the top of the building
carefully selected ground motion records can give an until a certain level of deformation is reached. The
accurate evaluation of the anticipated seismic target top displacement may be the deformation
performance of structures. Despite the fact that the expected in the design earthquake in case of designing a
accuracy and efficiency of the computational tools have new structure, or the drift corresponding to structural
increased substantially, there are still some reservations collapse for assessment purposes. The method allows
about the dynamic inelastic analysis, which are mainly tracing the sequence of yielding and failure on the
related to its complexity and suitability for practical member and the structure levels as well as the progress
design applications. Moreover, the calculated inelastic of the overall capacity curve of the structure.
dynamic response is quite sensitive to the characteristics The static pushover procedure has been presented and
of the input motions, thus the selection of a suite of developed over the past twenty years by Saiidi and
representative acceleration time–histories is mandatory. Sozen [1], Fajfar and Gaspersic [2] and Bracci et al. [3] ,
This increases the computational effort significantly.
among others. The method is also described and
The inelastic static pushover analysis is a simple option
for estimating the strength capacity in the post-elastic recommended as a tool for design and assessment
range. The technique may be also used to highlight purposes by the National Earthquake Hazard Reduction
potential weak areas in the structure. This procedure Program ‘NEHRP’ (FEMA 273) [4] guidelines for the
involves seismic rehabilitation of existing buildings. Moreover,
the technique is accepted by the Structural Engineers
Association of California ‘SEAOC’ (Vision 2000) [5]
* Corresponding author. Fax: +44 207 594 6053.
among other analysis procedures with various level of
E-mail address: a.elnashai@ic.ac.uk (A.S. Elnashai).
complexity. This analysis procedure is selected for its

-0296/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved. PII:
S0141-0296(00)00068-7
applicablity to performance-based seismic design and level of ductility. The dynamic pushover envelopes
approaches and can be used at different design levels to are then compared with the force– deformation curves
verify the performance targets. Finally, it is clear from obtained from inelastic static pushover analysis
recent discussions in code-drafting committees in considering different lateral loading patterns. The
Europe that this approach is likely to be recommended in procedure offers an opportunity for full comparisons
future codes. between the two methods of analysis up to ultimate
The technique has been evaluated in several previous collapse.
studies [6–10], to some extent, with different emphasis.
In most of the previous work, only comparative studies
between dynamic and static pushover analysis have been 2. Description of the buildings
assessed at certain loading levels, i.e. design level, or at
equal top displacement (roof displacement from In order to achieve the aforementioned objectives,
pushover equal to the maximum dynamic roof twelve RC buildings are considered, split into three
displacement). The results have been presented mainly groups: sets of four 8-storey irregular frame, four
in terms of global quantities, i.e. deformations, 12storey regular frame and four 8-storey dual frame-wall
calculated hysteretic energy and structural damage structures. Within each group, combination of two
indices. The main aim of this paper is to develop design ground accelerations (0.15 and 0.30 g) and three
complete pushover-like load– displacement curves from design ductility classes (High, Medium and Low) lead to
incremental dynamic analysis up to collapse for a range the four cases mentioned above. The selection of four
of structural configurations representing the most cases for each configuration is motivated by the desire to
common types of RC building, including different compare the performance of structures design according
structural systems, building heights, design acceleration to a ductility class set of rules but for different ground
A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424 409

acceleration and for the same ground acceleration but the member length. It is widely accepted that this
different ductility class rules. The value of the force technique is more accurate than the point–hinge models
reduction factor (behaviour factor q in EC8 and response mainly used in many other programs, especially when
modification factor R in UBC) increases and rigorous large axial force variations exist. The program has been
standards on member detailing requirements are verified elsewhere [14–16].
imposed for higher ductility classes. Table 1 shows the To accurately predict the inelastic seismic response of
definition of the set of structures under consideration the structure with sufficient accuracy, due care has been
where the elastic force reduction factors used in the given to create detailed and efficient models of the
design as well as the observed elastic fundamental structures, taking into account all necessary geometric
period, obtained from elastic free vibration analyses, are and strength characteristics of columns, beams and
also given. beam– column connections. Towards minimising the
Each building has been designed and detailed in computational requirements and the volume of input and
accordance with Eurocode 8 [11], Parts 1-1 to 1-3, as a output data to be handled, an effort was made to select
representative of a seismic design code applicable to powerful two-dimensional models that can provide, with
more than one country. While the second and third appropri-
groups are regular in plan and in elevation, the first
group exhibits two sources of irregularity in elevation.
The first storey has a greater height than the remaining
ones and severance at the first storey of some
intermediate columns, which are supported by long span
beams. The geometric characteristics of the structures
are illustrated in Fig. 1.
The overall plan dimensions of the configurations
considered are 15 m×20 m. The total heights are 25.5,
36 and 24 m for groups 1, 2 and 3, respectively, with
equal storey heights of 3 m except the first storey of
group 1 , which is 4.5 m high. While the lateral force
resisting system for groups 1 and 2 is moment frames,
group 3 possesses both a central core extending over the
full height and moment frames on the perimeter. The
floor system is solid slab in groups 1 and 2, and a waffle
slab in group 3. Live loads and loading from floor
finishes and partitions are both assumed to be 2.0 kN/m2.
All buildings are assumed to be founded on medium soil
type ‘B’ of EC8 (firm). The cross section capacities have
been computed by considering a characteristic cylinder
strength of 25 N/mm2 for concrete and a characteristic
yield strength of 500 N/mm2 for both longitudinal and
transverse steel. More details regarding member cross
section sizes and reinforcements are given in Fardis [12].

3. Modelling approach and assumptions

The inelastic analyses have been performed using the


adaptive static and dynamic structural analysis program
ADAPTIC, a program developed at Imperial College
[13] for the nonlinear analysis of steel, reinforced
concrete and composite structures under static and
dynamic loading. The program utilises the layered
‘fibre’ approach for inelastic RC frame analysis and has
the capability of predicting the large displacement
response of elastic and inelastic plane and space frames.
It has also the feature of representing the spread of
inelasticity within the member cross section and along
Table 1
Definition of the structural systems under analysis
Reference No. of storeys and structural Ductility
Group A.M.name
410 class 407–424Design acc. (g)
system / Engineering Structures 23 (2001)
Mwafy, A.S. Elnashai Force red. factor Elas. fund. period ( s )

1 IF-H030 8-storey irregular frame High 0.30 4.00 0.674


IF-M030 Medium 3.00 0.654
IF-M015 Medium 0.15 3.00 0.719
IF-L015 Low 2.00 0.723
2 RF-H030 12-storey regular frame High 0.30 5.00 0.857
RF-M030 Medium 3.75 0.893
RF-M015 Medium 0.15 3.75 0.920
RF-L015 Low 2.50 0.913
3 FW-H030 8-storey regular frame-wall High 0.30 3.50 0.538
FW-M030 Medium 2.625 0.533
FW-M015 Medium 0.15 2.625 0.592
FW-L015 Low 1.75 0.588

Fig. 1. Plane and cross sectional elevation of the buildings: (a) 8-storey irregular frame buildings; (b) 12-storey regular frame buildings; (c) 8-
storey regular frame-wall buildings. supported by the fact that conservative response
parameters will be obtained as a result of the domination
of gravity loads in long beam spans in the frame
ate selection of parameter values, acceptable
structures. On the other hand, the critical strain and shear
representation of the cyclic inelastic behaviour on
criteria are expected to occur in the coupling beams in
member and structure levels, while guaranteeing
the dual structural systems. Taking advantage of
numerical stability. The choice of two-dimensional
symmetry, only the interaction of two distinct frames
modelling may be also justified in the light of satisfying
(one internal and another external) is considered. Both
basic code requirements for such type of modelling.
lateral load-resisting frames are assembled using an
Two-dimensional analyses are undertaken in one
overlay approach, which is illustrated in Fig. 2 for the
direction only (global X- direction of frame structures
irregular frame structure. The two frames are coupled
and global Z-direction of frame-wall ones). This is
appropriately with regard to translational and rotational
A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424 411

degrees of freedom by 2D joint elements to meet the Modelling of the core is achieved by making use of
assumption of infinite in-plane stiffness of the slab in the two flexural elements, for each wall at each storey, in
normal direction. order to account for splicing of bars at mid-storey
The structural mesh utilises three elements for beam– height. The elements are located at the centroid of the
column members, the lengths of which are determined core Ushaped cross section and connected with beams at
on the basis of the critical member lengths. These each storey level using two rigid links. In addition, five
lengths are determined according to EC8 provisions for elements are used to represent each coupling beam, with
different ductility classes. The ends of horizontal bidiagonal reinforcement represented by vectorial
elements within the beam–column joints are considered resolution of the inclined reinforcement area along the
rigid. Consequently, two elements are added to each longitudinal and transverse directions. The same method
beam at its extremities. Furthermore, shear spring is utilised to represent the bidiagonal shear
connection elements are introduced to represent the reinforcement in some other beams and in the lower two
shear stiffness of the beam–column connection. To storeys of the core of the FW-H030 and FW-M030
simplify calculations of the shear stiffness of the joint, building.
the force– deformation relationship for both concrete Reinforced concrete column-section and T-section are
and steel reinforcement within the joint is assumed to be utilised for modelling of columns and beams,
linear elastic. Despite the simplicity of the joint respectively. Both sections, taken from ADAPTIC

modelling, global structural response obtained have been library, allow the geometrical definition of the section as
extensively compared and checked with analyses well as that of the confined concrete region within it.
performed by Salvitti and Elnashai [17] and Taking into account the available cross sections in
Panagiotakos and Fardis [18]. These show a good ADAPTIC library, a reasonable approximation is made
conceptual agreement with the current modelling results to replace the original U-shaped section of the core of
since the drift values are on the whole higher than the the framewall structures by a T-section, with the same
values by Salvitti and Elnashai [17] where no provision stiffness properties. The approximation may be justified
for beam–column connection behaviour was made. For in the light of the two-dimensional modelling which
the sake of brevity, only some results of the comparison neglects torsion and the regularity of the structure both
are shown in Fig. 3. The results of the current study, for in terms of stiffness and strength. Reinforcement
two of the 12-storey frame buildings, are between results patterns are varied for each section as a function of
of the rigid beam–column joint modelling of the former stirrup spacing in accordance with those specified in the
and the flexible, one member lumped plasticity design. Confinement factors are evaluated as described
modelling, of the latter where bar slip effects within the in Eurocode 8, and varied along the member length
joints and member shear deformations are considered. according to the arrangement of transverse
reinforcements. The effective slab width participating in
beam deformation is taken as
Fig. 2. The overlay technique considered and description of the beam–column joint modelling.
412 A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424

Fig. 3. Verification of modelling assumptions (a) RF-H030 building and (b) RF-L015 building (average for four artificial accelerograms).
the beam width plus 7% of the clear span of the represent the distribution of inertia forces imposed on
structural member on either side of the web. Horizontal the building. The first shape is calculated as SRSS
and vertical structural members including core walls are combinations (for the first three modes) of the load
modelled using two-dimensional cubic elasto-plastic distributions obtained from modal analyses of the
beam–column elements, where a cubic shape function is buildings. The choice of this load shape is made to take
used for the transverse displacement [13]. This into consideration the anticipated effect of higher modes
formulation is intended to represent short lengths of RC of vibrations for moderate long period and irregular
elements, consequently, axial strain is assumed to be structures (the 12-storey and the 8-storey frame
constant along the element length. The numerical buildings), as well as for buildings with hybrid lateral
integration of the governing equations for this element is resistance systems (the 8-storey frame-core structures).
performed over two Gauss sections, which have a fixed The design code lateral load pattern and a uniform load
position within the element length. The inelastic distribution shape have been also utilised. The latter
response of the cross section is assembled from represents the lateral forces that are proportional to the
contributions of individual layers for which inelastic vertical distribution of the mass at various levels. On the
cyclic material constitutive relationships are applied. other hand, the code lateral load shape represents the
The cubic elasto-plastic elements are combined with forces obtained from the predominant mode of vibration.
material models for concrete, which account for active The use of the uniform load shape may be justified in the
confinement and reinforcing steel with nonlinear light of a possible soft storey mechanism of the 8-storey
hardening. On the concrete side, the uniaxial constant irregular buildings. If this mechanism occurs the
confinement concrete model, Martinez-Rueda and response will be controlled by a large drift in the first
Elnashai [19], has been chosen. For steel, the advanced storey. Therefore, this load distribution may give better
multisurface steel model for cyclic plasticity, which predictions of the overall response. The inverted
defines the stress–strain response of steel in terms of a triangular (code) and the rectangular (uniform) load
series of cubic polynomials, Elnashai and Izzuddin [20], shapes also represent the extreme cases from the linear
is utilised. The parameters used in the material models distribution point of view. The shape of the lateral load
are the mean values. should be selected on the light of anticipated changes in
inertia forces as the structure moves from the elastic to
the plastic phases. Ideally, this shape should be modified
4. Load pattern and seismic action with the changes in inertia forces during the actual
earthquake. These changes mainly depend on the
According to the data used in the design [12], a live characteristics of both the record and the structure.
Several trials [2,3] have been made to permit of changes
load Q=2.0 kN/m2 is considered to calculate the total
in inertia forces with the level of inelasticity through the
gravity loads on the frames, which is applied as point
use of adaptive load patterns. The underlying approach
loads at nodes. Using the appropriate coefficients from
of this technique is to redistribute the lateral load shape
EC8, the vertical loads are combined with seismic
with the extent of inelastic deformations. The load shape
actions in a combination of 1.0G+0.15Q+E L for all is suggested to be redistributed according to the global
stories except the top floor, where it was taken equal to displacement shape, the level of storey shear demands or
1.0G+0.30Q+E L . To account for inertia effects during a combination of mode shapes obtained from secant
dynamic analysis, masses are calculated in a manner stiffnesses. This redistribution is performed at each time
consistent with the gravity loading combinations and are step, which leads to a substantial increase in the
represented by lumped 2D mass elements. computational effort. Moreover, the pushover analysis
Due to the fact that the lateral force profiles in static has not been widely established as yet in the design
pushover analyses will influence the structural response, office environment. Therefore, for common types of
three different load patterns have been utilised to building the need for more taxing approaches is by no
A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424 413

mean fully established. The variable load distribution given in Table 2, while their acceleration response
option may be appropriate for special and long period spectra for 5 % damping are shown in Fig. 5.
structures, despite that eminence of this technique has In order to apply the outlined procedure for the
not been confirmed yet [9,10]. On this basis, the evaluation of dynamic collapse envelopes, scaling of the
aforementioned fixed load distribution shapes have been records utilised is frequently required. The technique of
utilised for the current study. It is also worth mentioning scaling earthquake records to possess equal values of
that the NEHRP (FEMA 273) guidelines recommend spectrum intensity was based on a proposal by Housner
utilising fixed load patterns with at least two load [23]. The spectrum intensity is defined as the area under
profiles. The first shape should be the uniform load the pseudo-velocity spectrum between certain period
distribution and the other is the code profile or the load limits. It is suggested in the current study to modify the
shape obtained from multimodal analyses. The code limits employed in the original method (between 0.1 and
lateral load distribution is allowed if more than 75% of 2.5 s) to be between 0.8 T y and 1.2 T 2D , where T y and T 2 D
the total mass participates in the predominant mode. are the inelastic periods of the structure at global
Time–history analyses employ four artificially yielding and at twice the design ground acceleration,
generated 10-s duration acceleration records, referred to respectively. This follows the proposal of
as Art-rec1 to Art-rec4, as well as two natural records. MartinezRueda [24,25], modified for the range used
The artificial accelerograms were generated to fit the here. Therefore, the normalisation factor for an
Eurocode 8 elastic response spectrum for medium soil accelerogram (n) is equal to the ratio SI c /SI n . Where, SI c
class as shown in Fig. 4 for a PGA=0.3 g. The use of the and SI n are the areas under the code velocity spectrum
artificial accelerograms is in order to allow effective and the velocity spectrum of the scaled accelerogram,
comparisons and calibrations with the design code. respectively. SI c and SI n are calculated between periods
Moreover, the effect of the vertical component of the of 0.8 T y and 1.2
seismic excitation is worthy of consideration [21], T 2D , as explained above.
particularly for the irregular frame structures where the It is also worth mentioning that there is no need to use
the aforementioned scaling method with the artificial
accelerograms since they are already spectrum-
compatible. Hence, the four artificial records are scaled
according to their PGA. The buildings are analysed first
under the artificial records at different PGA levels and
the recorded top response time history is utilised to
obtain the inelastic periods T y and T 2D of each building
from Fourier analyses (average for four artificial
records). The scaling factors are then calculated for the
longitudinal component of natural records and used for
scaling the accelerograms up to collapse. The factors
used to scale the longitudinal earthquake component are
also used to scale the vertical component of the motions,
when employed, to keep the V/H ratio constant. Table 3
shows the average normalisation factors to ground
acceleration 0.30 g for each of the three groups of
Fig. 4. Acceleration spectra for the artificial accelerograms (5% buildings. Finally, it should be noted that the quoted
damping). values of PGA are not of the natural or scaled records
planted columns are supported by long span beams. but rather multiples of the design ground acceleration.
Towards this end, two natural ground motions have been
selected in terms of the V/H ratio (peak vertical-to-
horizontal acceleration). The Kobe (Hyogo-ken Nanbu
at Kobe University, Japan, 1995) and the Loma Prieta
(Northern California at Saratoga ‘Aloha Ave’, USA,
1989) earthquakes are employed and applied with and
without the vertical components, giving two analyses for
each record. However, for the sake of brevity, results of
the effect of vertical ground motion on the seismic
response are not presented herein. Comprehensive
results of this study are given elsewhere [22].
Characteristics of the records that have been used are
Table 2
Characteristics of records used in analysis

Earthquake
414 A.M. Mwafy, A.S. Elnashai / Engineering Date Ms 407–424
Structures 23 (2001) Station PGA (g) V/H No. of input runs

Horiz. Vert.

Kobe (Japan) 17/01/95 7.20 KBU 0.276 0.431 1.56 2


Loma Prieta (USA) 18/10/89 7.17 SAR 0.319 0.349 1.09 2
Artificial Records Art-rec1, Art-rec2, Art-rec3,
4
and Art-rec4

is both inelastic and geometrically nonlinear. The large


displacement formulation is an updated Lagrangian
form, where convected member axes are used to derive
member deformations.
The criteria used for defining collapse are classified
into two groups; local and global criteria. Two failure
criteria on the member-level are applied: the ultimate
curvature, which is normally controlled by the maximum
compression strain at the extreme fibre of the confined
concrete and shear failure in any structural member. An
empirical axial load-sensitive shear model capable of
providing an experimentally verifiable estimate of shear
supply in RC members was proposed by Priestley et al.
[26] and has been utilised in this study [27]. The code
shear supply model has also been employed after elimin
Fig. 5. Elastic spectra for the long. component of the natural records chosen: a limit corresponds to a maximum inter-storey
(5% damping). drift of 3% of the storey height, formation of a sidesway
mechanism and reduction in lateral resistance by
Table 3 considering the load–displacement curve of the
Normalisation factors for ground acceleration 0.30 g structure. Additionally, the criterion used to define
global yield threshold, which is essential for the
Earthquake IF-buildings RF-buildings FW-buildings proposed scaling method of the records, is selected as
Average the yield displacement of the equivalent elasto-plastic
system with reduced stiffness evaluated as the secant
Artificial 0.30 0.30 0.30 0.30 stiffness at 75% of the ultimate load of the real system.
Kobe (KBU) 0.54 0.61 0.56 0.57
The utilised shear models are implemented with other
Loma Prieta 1.15 1.25 1.32 1.24 collapse and yield criteria in a post-processing program
(SAR) connected to ADAPTIC [28]. This post-processsor
traces the shear supply–demand situation at each time
step at both ends of all members. It also performs the
appropriate calculations to evaluate the local and global
5. Collapse criteria and incremental dynamic collapse response parameters of the structure and directly apply
‘dynamic pushover’ results the selected criteria.

Three types of analyses have been performed using


the structural models described earlier. Eigenvalue
analyses are conducted to determine the elastic periods
and the mode shapes of the buildings needed for
calculating the first lateral load profile of the static
pushover analysis (combination of loads from modal
shapes). Inelastic static pushover and dynamic analyses
are then performed using the calculated lateral load
shapes and the seismic actions with increasing severity.
The analyses are progressed until all the predefined
collapse limits are exceeded. In both static and dynamic
analyses, permanent loads are first applied and iteration
to equilibrium is performed. This is followed by
applying the horizontal action (loads or ground
acceleration). The analysis ating the design safety
factors. On the structure level, three collapse criteria are
A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424 415

The results of more than 1300 inelastic time–history for the two buildings designed for the higher design
analyses were employed to perform regression analyses ground acceleration gives the impression of being higher
to obtain the dynamic pushover (ideal) envelope for each than the other pair of results. The low correlation of the
of the twelve examined buildings. Figs. 6–8 depict former and the high correlation of the latter are reflected
dynamic response points and the fitted regression in the correlation values which are equal to 0.69 and
equations of the response of the buildings subjected to 0.66 for the first pair and 0.93 and 0.88 for the second
the eight seismic actions considered for all limit states. one. It should be noted that the main difference between
The fitted envelopes for the upper and lower response the two pairs is in the longitudinal and transverse
points, the number of analyses carried out, the design reinforcement of the structural members, while the
base shear and the correlation coefficient for each case dimensions of the cross-sections for this group are the

Fig. 6. Dynamic collapse analysis results for the irregular frame structures.
are also shown. The actual response of the 8-storey same except a slight changes, mainly in the beams
irregular frame structures illustrated in Fig. 6 show how crosssection width (from 0.35 m for the first pair to 0.30
the results of the eight seismic actions follow the same m for the second one). In spite of the aforementioned
trend and shape the pushover envelopes without the need observations, the difference in scatter between the higher
to apply curve fitting. This is clear from the correlation and the lower design ground acceleration pair diminishes
coefficient values, which are almost invariably above when calculating the difference between the lower and
0.9. It is worth mentioning that the number of time– the upper response envelope for each case (quotient of
history analyses shown on each graph varies according minimum and maximum strength for the eight records).
to the number of trials needed to identify the collapse This value is equal to 0.69 and 0.72 for the 0.30 g design
and the yield limits, as discussed above. ground motion pair and 0.76 and 0.74 for the other pair.
Concerning the 12-storey regular frame structures, This is more consistent since the calculated inelastic
Fig. 7 shows a higher scatter in the dynamic analysis periods, which are the main cause of the different
results of different ground motions than the results of the response, are very close for the four buildings, as
8-storey irregular frame buildings. Moreover, the scatter subsequently discussed. Finally, the difference in the
416 A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424

correlation values between the two pair of buildings can pair is 0.35 m, compared to 0.25 m for the other pair.
only be justified in the light of the lower number of runs This causes an increase in the mass at each storey level
( or response points) needed to achieve yield and for the former, hence higher amplification of base shear
collapse for each pair. This number is equal to 97 and demand. Alongside the high initial stiffness of this pair,
105 for the first pair and 60 and 73 for the second one. the difference in response between the two pairs of
The high sensitivity to changes in the input motion building shown in Fig. 8 can be explained.
observed in the 12-storey frame buildings are also 6. Contribution of the elongated period to the seismic

Fig. 7. Dynamic collapse analysis results for the regular frame structures.
reflected, to a lesser extent, in the 8-storey frame-wall response
group. Fig. 8 illustrates the results of the time–history
analyses for this group. At collapse limit state, a scatter The scatter observed for some buildings is mainly in
is observed for values of V Min /V Max shown in Table 4. A the post-elastic range, and is associated with the spread
higher hardening stiffness is also observed for the 0.30 g of yielding and member failure throughout the structure.
design ground acceleration pair compared to the other Subsequently, the stiffness of the structure decreases, the
pair. This is not observed in the other two groups of fundamental period elongates and the distribution of the
building. Previous analytical investigations [29,30] have inertia forces along the building undergoes continuous
indicated that base shear demands of wall structures are change. To provide insight into the response of the
sensitive to higher mode effects. Once a plastic hinge investigated buildings, extensive analyses in the
has formed at the base of the wall, higher mode effects frequency domain (Fourier analyses) of the acceleration
can considerably amplify the base shear as well as the response at the top have been conducted to identify the
shear at each storey level. The results shown in Fig. 8 predominant inelastic period of each building under
confirm that such amplification may occur and could be consideration. Fig. 9 illustrates the calculated periods
large. It is also worth mentioning that the thickness of (average for the eight seismic actions) at the design and
the core-walls for the higher design ground acceleration twice the design ground acceleration, along with the
A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424 417

Fig. 8. Dynamic collapse analysis results for the frame-wall structures.

Table 4
Observed response at global collapse for the eight records

Group Reference name Roof disp. (mm) D Min /D Max Base shear (kN) VMin /VMax Storeya
D Max D Min Mean VMax V Min Mean

1 IF-H030 613 503 542 0.82 11,614 9918 10,567 0.85 1, 2, 3, 5


IF-M030 635 500 570 0.79 13,930 12,713 13,146 0.91 1, 2, 4, 5
IF-M015 492 381 449 0.77 7699 6663 7123 0.87 1, 4, 5
IF-L015 590 380 465 0.64 9229 8102 8685 0.88 1, 2, 4, 5
2 RF-H030 690 580 625 0.84 15,647 11,568 13,689 0.74 2, 4, 8, 9
RF-M030 796 611 684 0.77 16,278 12,076 13,990 0.74 4, 5, 8, 9
RF-M015 735 630 681 0.86 9743 9234 9453 0.95 5, 6, 8
RF-L015 785 607 694 0.77 12,735 11,009 11,972 0.86 4, 5, 9
3 FW-H030 643 599 631 0.93 20,821 15,520 17,849 0.75 2, 3, 6, 7
FW-M030 660 576 625 0.87 23,300 18,123 20,738 0.78 2, 5, 7
FW-M015 643 590 621 0.92 12,724 8769 10,642 0.69 2, 3
FW-L015 652 598 626 0.92 16,153 11,604 13,425 0.72 2, 3, 7
a
Location where interstorey drift collapse criterion is observed for the eight ground motions.
elastic period for each building calculated from respectively. It is observed that the average percentage
eigenvalue analyses. of elongation in the period is (100%), (90%), and (60%).
It is clear to what extent the fundamental periods of The percentage increase is clearly related to the overall
the buildings are elongated as a result of the spread of stiffness of the structural system of the building. The
cracks and yielding. The average elastic periods for the maximum calculated elongation is recorded in the most
three groups of building are 0.69, 0.90, and 0.56 s,
respectively. On the other hand, the calculated inelastic
periods at the design and twice the design ground
acceleration are (1.30–1.46), (1.65–1.80), and (0.81–
1.00) s,
418 A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424

Fig. 9. Elastic and inelastic (at the design and twice the design ground acceleration) predominant response periods of the buildings — average
for the eight seismic actions. average inelastic fundamental period is about 1.75 s,
which corresponds to high amplification in the Loma
Prieta (SAR) record only. This also accounts for the high
flexible system, where the first storey can be considered response of the 12-storey frame structures when
as a soft storey; whereas the minimum elongation is subjected to the latter record. The same applies, to a
observed in the stiff frame-wall structural system. The lesser extent, to the artificial and Loma Prieta (SAR)
results point towards an important conclusion, records when imposed on the frame-wall structures
employment of elastic periods of vibration in estimating (inelastic period 0.91 s), compared to the Kobe (KBU)
design forces leads to high levels of overstrength (ratio record. For this reason the observed scatter for this
of actual-to-required strength). Moreover, they lead to group is less than the 12-storey buildings. On the other
nonuniform safety margins for different structural hand, the ordinates of the spectra correspond to the
systems. inelastic period of the 8-storey irregular buildings are
To facilitate the comparison between the input equivalent, hence the high correlation for this group.
accelerograms utilised in this study in terms of the
frequency content, the records are scaled to ground
acceleration equal to 0.30 g and used to obtain the 7. Inelastic static-to-collapse ‘static pushover’
Fourier spectrum for each record. The normalisation analyses
factors used are the average of the scaling factors
utilised to perform the collapse analysis for the three Following the success in obtaining the incremental
groups of buildings, as shown in Table 3. The Fourier dynamic response envelopes for the twelve buildings
amplitude spectra for the acceleration records (one of the under investigation, inelastic static pushover analyses
utilised artificial records, Artrec1, and the horizontal are performed to assess the applicability of the technique
component of the two natural records) are shown in Fig. (for different load distributions) in predicting the overall
10. The average inelastic period for each group of dynamic response of structures. Figs. 11–13 illustrate the
building is also shown on the graphs. It is clear that the base shear vs top displacement plots for the three lateral
input motions, with the exception of the Kobe (KBU) load profiles utilised along with the incremental dynamic
record, have high amplitude that may amplify the effect envelopes for the twelve buildings. The dynamic
of the second mode of vibration for structures with pushover curve for each case is shown in the form of the
period between 0.35 and 0.50 s. The amplitude in this upper and lower response envelope as well as the best fit
period range is higher than the amplitude corresponding of the time–history analysis results. The global yield and
to the fundamental period of the three groups of collapse thresholds are also shown. It should be pointed
structures. This is one of the reasons for the scatter in the out that it was decided to choose only one global yield
results of the 12-storey regular frame and the 8-storey limit from the limits obtained from the four pushover
frame-wall structures. It is verified that the high envelopes (the three static and the dynamic one). This is
response points in Figs. 7 and 8 are for the artificial and due to the need to unify and simplify obtaining this limit,
Loma Prieta (SAR) records, while the low response is which is necessary for the suggested method of scaling
for the Kobe (KBU) record. This may also justify the input seismic actions explained earlier. The yield
obtaining a higher maximum base shear corresponding limit state obtained from the static pushover analysis
to almost an identical top deflection when applying the using the code lateral load shape is selected for this
same ground motion with higher PGA. Furthermore, the purpose. For collapse, the observed upper and lower
Fourier spectral ordinate corresponding to the inelastic global collapse limits from the eight earthquake records
fundamental period of the buildings can also be utilised
to justify the scatter in the results of the second and the
third group of buildings. For the 12-storey buildings, the
A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424 419

Fig. 10. Fourier amplitude spectra for the input accelerograms (scaled to 0.30 g) and the average inelastic period of the buildings.

Fig. 11. Static and dynamic pushover analysis results for the irregular frame structures.
420 A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424

Fig. 12. Static and dynamic pushover analysis results for the regular frame structures.
as well as from the three pushover analyses are the code and the multimodal load patterns to the uniform
presented. It is important, however, to note that the static load shape. It is also noticeable that the difference
pushover and the incremental dynamic collapse analyses between load shape A (the design code load pattern,
are continued beyond all the predefined collapse which is almost an inverted triangle) and load shape B
thresholds. This is to ensure that all collapse states are (load shape from multimodal analysis) is very small.
bounded by the dynamic analysis. Although higher mode effects are confirmed in the
The static pushover method is rarely used to predict response of the second and the third group of buildings,
seismic demands when a particular ground motion is as explained earlier, the multimodal analysis load
imposed on a structure [9,10]. If this is needed, the top pattern did not show an enhanced capability to predict
‘target’ displacement expected when this ground motion these effects. This is due to the fact that this load shape
is imposed on the building should be estimated. It is represents the distribution of inertia forces in the elastic
beyond the scope of this study to address the approaches range only, while the amplification of higher mode
of estimating the target displacement. A review of these effects are observed in the post-elastic phase. Table 5
methods was given in the latter two references. Since the presents the results at global collapse limit state for the
main application of the static pushover analysis is to three load shapes. In terms of the predicted ultimate
estimate the seismic capacity of structures, the following strength and drift at collapse, the differences between
observations are driven by this requirement. In this load A and B are less than 4%, for the twelve buildings.
application of the procedure, the analysis is usually As a general trend, the collapse is observed earlier
continued until any of the predefined collapse criteria is when applying the uniform load than the triangular load.
exceeded. Collapse is observed slightly earlier than the triangular
In general, it is clear in all cases that the response of distribution when imposing the multimodal load. In Figs.
the buildings is sensitive to the shape of the lateral load 11–13 the lower collapse limits from static analyses are
distribution. This is particularly true when moving from always from the uniform load and the upper limits are
A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424 421

Fig. 13. Static and dynamic pushover analysis results for the frame-wall structures.

Table 5
Results at global collapse limit state for the three load patterns
Group Reference name Roof disp. (mm) Base shear (kN) Storeyd

Aa Aa Aa
Bb Cc Bb Cc Bb Cc

1 IF-H030 534 528 508 10,091 10,446 11,592 4th 3rd 3 rd


IF-M030 552 534 480 12,690 13,056 14,219 3rd 3rd 2 nd
IF-M015 474 462 432 6652 6914 7620 3rd 3rd 2 nd
IF-L015 516 498 450 8253 8508 9147 3rd 2nd 2 nd
2 RF-H030 648 624 552 12,135 12,499 14,650 5th 4th 3 rd
RF-M030 712 688 568 13,083 13,444 15,748 5th 5th 3 rd
RF-M015 656 640 600 7332 7554 9235 5th 5th 4 th
RF-L015 688 664 592 9817 10,136 12,175 5th 5th 4 th
3 FW-H030 570 560 535 13,243 13,796 16,425 3rd 3rd 3 rd
FW-M030 580 570 545 16,671 17,241 20,754 3rd 3rd 3 rd
FW-M015 575 565 530 7880 7988 9843 3rd 3rd 3 rd
FW-L015 590 580 545 10,001 10,119 12,490 3rd 3rd 2 nd
a b
Triangular load. Multimodal
load. c Uniform load. d Storey at which
collapse is observed.
from the triangular load. Moreover, the maximum uniform load pattern is observed in lower storeys than
interstorey drift collapse limit when employing the the recorded collapse from the triangular or the
422 A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424

multimodal loads. This observation is also recorded in Higher modes effect and concentration of inelastic
dynamic analyses, where collapse is observed in lower deformations are expected to be significant in the first
storeys for records that impose higher base shear. Fig. 14 group of structures where the buildings exhibit two
depicts the relationship between collapse limit states of sources of irregularity and weak storey.
the three load shapes for one of the investigated Notwithstanding, four static pushover analyses using the
buildings (building RF-L015). It should be emphasised simple triangular load pattern have succeeded in
that the load–deformation envelope is for global predicting the average results of more than 600 inelastic
response, which is a function of the point of application time–history analyses. It is also important to note that
of resultant force. The uniformly distributed load gives the good design of these buildings and the high
the lowest point; hence the maximum strength and overstrength associated with structural elements,
earlier global yield and collapse limit states. On the particularly the columns, prevented any undesirable
other hand, resultant in the triangular load case is mode of failure. The results show that utilising the
applied at a higher point; consequently lower strength triangular load shape only to predict the global response
and delayed global yield and collapse are observed in all of low rise frames as well as well designed irregular
cases. frame structures is adequate.
Despite the fact that all load shapes do not represent In contrast to the first group of buildings, the results of
the actual distribution of relative inertia forces during the the static pushover of the 12-storey group, illustrated in
dynamic analysis, almost an identical response is Fig. 12, show discrepancies with the dynamic response
observed in the first group of buildings between the envelope in the post-elastic range. While the static
dynamic analysis best-fit envelopes and the static pushover results of the triangular and the multimodal
response obtained from the triangular and the load pattern show a good agreement with the dynamic
multimodal distributions. On the other hand, the uniform results best fit in the elastic range, both give a
load overestimates the initial stiffness and the maximum conservative prediction of the maximum lateral strength,
base shear in the four buildings. Table 6 illustrates as also shown in Table 6 for the triangular load.
graphically the differences between the results of the However, in the four buildings the triangular load
static pushover analysis for the triangular and the response is higher than the lower limit envelopes
uniform load patterns on one side, and the incremental obtained from dynamic analyses employing natural and
dynamic analysis (average for eight records) on the artificial records. On the other hand, the capacity curve
other, at global collapse limit state. Since the triangular obtained from the uniformly distributed load
load shape is simple and show very close results with the overestimates the response in the elastic range.
multimodal load pattern, it was decided to exclude the However, it gives better prediction of the ultimate
latter from this comparison. It is clear that the uniformly strength. It is also clear from Fig. 12 that the triangular
distributed load is unconservative in predicting collapse load shape gives good prediction of the deformation at
limit states (underestimates the drift and overestimates collapse, while the uniform load underestimates the
the strength). The overall prediction of collapse using collapse limit state in the four buildings. It is concluded
the triangular load is significantly better. Although it for this group of buildings that the triangular distribution
slightly underestimates the average drift of dynamic is again the most suitable load pattern given that the
analysis in two buildings, it is between the upper and uniform load, which is rec-
lower drift limits obtained from the eight records, as
shown in Fig. 11.
A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424 423

Fig. 14. Differences between the three lateral load patterns.


Table 6
Differences between static and dynamic pushover analysis at global collapse limit state

Group Difference in drift Difference in strength

Triangular Uniform Triangular Uniform

ommended by the NEHRP guidelines (FEMA 273), is The comparison between static and dynamic pushover
unconservative in predicting the response and the drift at analysis for this group of buildings shows more
collapse. discrepancies than the second group, especially for the
Concerning the frame-wall group of buildings, the 0.30 g design ground motion pair. As explained earlier,
differences between the static pushover and the dynamic these differences are mainly due to higher mode effects,
analysis results are larger than for the other two groups. which amplify the base shear following formation of
In terms of the predicted elastic response and initial first plastic hinge at the base of the wall. In pushover
stiffness, the triangular and the multimodal load shapes analysis, once the wall attained its ultimate lateral
show good correlation with the dynamic analysis best fit strength, it will deform by plastic hinging at the base
for three buildings and a conservative prediction for the [31]. Clearly, for this type of structure (shear frame
fourth (FW-M030). In the post-elastic range, the two response plus cantilever wall response) the amplification
load shapes underestimate the lateral strength obtained of the base shear during the dynamic analysis is difficult
from the time–history analyses. Table 6 shows that the to predict by pushover analysis. However, the triangular
triangular load prediction of strength at global collapse load profile shows good correlation in the elastic range,
is between 20 and 26% less than the average results of conservative predictions of the ultimate strength and
dynamic analyses. Similar to the 12-storey buildings, the reasonable estimations of the collapse limit state
uniform load overestimates the elastic response but gives (underestimates the drift by about 8%). Hence, it may be
better prediction of the lateral strength at collapse for employed for estimating the seismic capacity and
this group of buildings. Moreover, none of the collapse limit state.
investigated load patterns give reasonable prediction of
the high hardening stiffness obtained from dynamic
analysis for the higher design ground motion pair. With
regard to predicting the drift at collapse, both the
triangular and the uniform load patterns are
unconservative. This is clear from Fig. 13 and Table 6.
424 A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424

Finally, if the static pushover analysis is utilised as a static pushover analysis using three lateral load
tool for predicting seismic demands instead of distributions and local and global limit state criteria are
estimating capacities, the analysis is usually performed utilised. Based on the large amount of information
until the roof drift corresponding to the design ground obtained, which is nonetheless far from comprehensive,
acceleration is attained. Table 7 presents the average for the following conclusions are drawn:
eight ground motions of the maximum top displacement
observed from time–history analyses at the design I Subject to adequate modelling of the structure, careful
ground acceleration. Clearly, the target displacement is selection of the lateral load distribution and articulate
almost always below the global yield limit state. The interpretation of the results, pushover analysis can
comparison between the static pushover and the dynamic provide insight into the elastic as well as the inelastic
analysis discussed above show that the triangular load response of buildings when subjected to earthquake
gives better estimation of the response in the elastic ground motions.
range. In few buildings, however, it underestimates the I Static pushover analysis is more appropriate for low
initial stiffness. In contrast, in the same range the
rise and short period frame structures. For
uniform load shape overestimates the stiffness and the
welldesigned buildings but with structural
base shear in all cases. From the design point of view,
irregularities, the results of the procedure also show
the uniformly distributed load is conservative for the
good correlation with the dynamic analysis. In this
twelve buildings investigated. It is concluded that the
study, response obtained for a group of four 8-storey
use of two load distributions is needed for estimating the
irregular frame buildings using an inverted triangular
seismic demand. The simple triangular or the
lateral load distribution is identical to inelastic time–
multimodal shape, which correlate well with dynamic
history analysis.
analysis results and the uniform load pattern which
shows a conservative prediction of demands in almost I The experience gained from previous studies can help
all cases considered. This conclusion is supported by the to eliminate the discrepancies between static and
observations obtained from the results of the frame-wall dynamic analysis results for special and long period
group of buildings. The uniform load pattern can provide buildings. These differences are mainly due to the
limited capability of the fixed load distribution to
Table 7
Observed maximum roof displacement at the design ground acceleration (average for eight ground motions)
Group 1 Roof disp. (mm) Group 2 Roof disp. (mm) Group 3 Roof disp. ( mm )

IF-H030 285 RF-H030 290 FW-H030 115


IF-M030 202 RF-M030 304 FW-M030 120
IF-M015 138 RF-M015 146 FW-M015 56
IF-L015 136 RF-L015 163 FW-L015 67
a conservative estimation of shear demand below predict higher mode effects in the post-elastic range.
collapse limit states; hence it can support preventing To overcome this problem, more than one load
undesirable shear failure. Moreover, it shows reasonable pattern should be selected to guarantee providing an
estimation of shear at global collapse limit state (at 3% accurate or slightly conservative prediction of
interstorey drift). It is worth mentioning that, according capacities and demands.
to the SEAOC (Vision 2000), complete collapse is I The investigation carried out on two sets of four
considered once the interstorey drift exceeds 2.5%. 12storey frame buildings and four 8-storey frame-
Utilising this definition of collapse leads to obtaining wall structures show that a conservative prediction of
conservative prediction of shear demand in all buildings capacity and a reasonable estimation of deformation
investigated when employing the uniformly distributed is obtained using the simple triangular or the
load. multimodal load distribution. The same load patterns
slightly underestimate the demand of some buildings
in the elastic range. On the other hand, the uniform
8. Conclusions load provides a conservative prediction of seismic
demands in the range before first collapse. It also
The applicability and accuracy of inelastic static yields an acceptable estimation of shear demands at
pushover analysis in predicting the seismic response of the collapse limit state.
RC buildings are investigated. Twelve RC buildings I Comparison between the triangular and the multimodal
with various characteristics, incremental dynamic distribution results show differences less than 4 %,
analysis employing eight natural and artificial records, for the twelve buildings, since the former captures the
A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424 425

characteristics of the most important mode of [1] Saiidi M, Sozen MA. Simple nonlinear seismic analysis of R/C
vibration. The load distribution from multimodal structures. Journal of the Structural Division, ASCE
analysis only represents the distribution of inertia 1981;107(ST5):937–51.
forces in the elastic range; hence higher mode effects [2] Fajfar P, Gaspersic P. The N2 method for the seismic damage
analysis of RC buildings. Earthquake Engineering and
are not entirely accounted for in the post-elastic Structural Dynamics 1996;25:31–46.
domain.
[3] Bracci JM, Kunnath SK, Reinhorn AM. Seismic performance
I The elongation in the fundamental period of structures and retrofit evaluation of reinforced concrete structures. Journal
due to extensive yielding and cracking during of Structural Engineering, ASCE 1997;123(1):3–10.
earthquakes depends on the overall stiffness of the [4] FEMA. NEHRP guidelines for the seismic rehabilitation of
structural system of the building. In the current study, buildings. FEMA 273, Federal Emergency Management
Agency,
the observed elongation ranges between 100% for the
1996.
most flexible irregular frame system and 60% for the
stiff frame-wall structural system. Employment of [5] SEAOC. Performance based seismic engineering of buildings.
Vision 2000 Committee, Structural Engineers Association of
elastic periods in seismic code does not therefore California, Sacramento, CA, 1995.
provide uniform levels of safety for different [6] Lawson RS, Vance V, Krawinkler H. Nonlinear static push-over
structural systems. analysis — why, when, and how? In: Proceedings 5th US
I The results of the dynamic collapse analysis show NCEE, vol. 1. IL, USA: Chicago, 1994:283–92.
clearly that each earthquake record exhibits its own [7] Mitchell D, Paultre P. Ductility and overstrength in seismic
peculiarities, dictated by frequency content, duration, design of reinforced concrete structures. Canadian Journal of
Civil Engineering 1994;21:1049–60.
sequence of peaks and their amplitude. The dispersion
[8] Faella G, Kilar V. Asymmetric multistorey R/C frame
in the results of different ground motions depends on structures: push-over versus nonlinear dynamic analysis. In:
the characteristics of both the structure and the Proceedings 11th ECEE, Paris, 1996, CD-Rom.
record. The Fourier spectral analysis is an important [9] Krawinkler H, Seneviratna GDPK. Pros and cons of a pushover
tool to investigate the observed variability of the analysis of seismic performance evaluation. Engineering
results and to identify the elongated inelastic periods Structures 1998;20(4-6):452–64.
of the structure. [10] Tso WK, Moghadam AS. Pushover procedure for seismic
analysis of buildings. Progress in Structural Engineering and
I The importance of pushover analysis as an assessment Materials 1998;1(3):337–44.
and design tool warrants much needed further [11] Eurocode 8. Design provisions for earthquake resistance of
developments. These may be classified as ‘tools’ and structures. Part 1-1, 1-2 and 1-3, Comite Europeen de
‘behaviour’. There is considerable scope for Normalisation, European Pre-standard ENV 1998-1-1, 1-2, and
development of tools for more efficient and versatile 1-3, Bruxelles, 1994.
pushover analysis techniques. One such development [12] Fardis MN. Analysis and design of reinforced concrete
would be the continuous assessment of the effect of buildings according to Eurocodes 2 and 8. Configuration 3, 5
and 6, Reports on Prenormative Research in Support of
inelasticity on the load distribution used, taking into Eurocode 8, 1994.
account the shape of the spectrum. This would enable
[13] Izzuddin BA, Elnashai AS. ADAPTIC — a program for
the accurate and realistic analysis of highly irregular adaptive large displacement elastoplastic dynamic analysis of
structures. With regard to ‘behaviour’, analysis of a steel, concrete and composite frames. ESEE Report No. 89/7,
larger sample of buildings that includes high-rise Imperial College, London, 1989.
structures and structures with heavily irregular [14] Elnashai AS, Elghazouli AY. Performance of composite
strength distribution is needed. steel/concrete members under earthquake loading, Part I:
Analytical model. Earthquake Engineering and Structural
Dynamics 1993;22(4):315–45.
To close, it is emphasised that, notwithstanding the [15] Broderick BM, Elnashai AS. Seismic resistance of composite
range of structures analysed, the number of records beam–columns in multi-storey structures, Part 2: Analytical
employed and the rigour of the limit state criteria model and discussion of results. Journal of Construction Steel
Research 1994;30(3):231–58.
monitored, the conclusions are, strictly speaking,
applicable to the range investigated. However, some [16] Elnashai AS, Pinho R. Repair and retrofitting of RC walls using
selective techniques. Journal of Earthquake Engineering
generality may be claimed by noting that every effort 1998;2(4):525–68.
has been made to select distinct structural systems, [17] Salvitti LM, Elnashai AS. Evaluation of behaviour factors for
comprehensive limit states and verified investigation RC buildings by non-linear dynamic analysis. In: Proceedings
tools. 11 th WCEE, Acapulco, Mexico, 1996, CD-Rom, Paper No.
1820.
[18] Panagiotakos TB, Fardis MN. Effect of column capacity design
on earthquake response of reinforced concrete buildings. Journal
References of Earthquake Engineering 1998;2(1):113–45.
426 A.M. Mwafy, A.S. Elnashai / Engineering Structures 23 (2001) 407–424

[19] Martinez-Rueda JE, Elnashai AS. Confined concrete model


under cyclic load. Materials and Structures 1997;30(197):139–
47.
[20] Elnashai AS, Izzuddin BA. Modelling of material nonlinearities
in steel structures subjected to transient dynamic loading.
Earthquake Engineering and Structural Dynamics 1993;22:509–
32.
[21] Papazoglou AJ, Elnashai AS. Analytical and field evidence of
the damaging effect of vertical earthquake ground motion.
Earthquake Engineering and Structural Dynamics
1996;25:1109–37.
[22] Mwafy AM. Seismic performance of code-designed RC
buildings. PhD thesis, Imperial College, University of London,
2000. [23] Housner G. Spectrum intensities of strong-motion
earthquakes. In: Proceedings of the Symposium on Earthquake
and Blast Effects on Structures. CA, USA: Los Angeles,
1952:20–36.
[24] Martinez-Rueda JE. Energy dissipation devices for seismic
upgrading of RC structures. PhD thesis, Imperial College,
University of London, 1997.
[25] Martinez-Rueda JE. Scaling procedure for natural
accelerograms based on a system of spectrum intensity scales.
Earthquake Spectra 1998;14(1):135–52.
[26] Priestley MJN, Verma R, Xiao Y. Seismic shear strength of
reinforced concrete columns. Journal of Structural Engineering,
ASCE 1994;120(8):2310–29.
[27] Elnashai AS, Mwafy AM, Lee D. Collapse analysis of RC
structures including shear. ASCE Structures Congress, New
Orleans, 1999.
[28] Mwafy AM. Seismic performance of RC buildings under
multiaxial earthquake loading. PhD transfer report, Imperial
College, London, 1998.
[29] Ghosh SK. Required shear strength of earthquake resistant
reinforced concrete shear walls. In: Krawinkler H, Fajar P,
editors. Nonlinear Seismic Analysis and Design of Reinforced
Concrete Buildings. Oxford: Elsevier Science, 1992.
[30] Seneviratna GDPK. Evaluation of inelastic MDOF effects for
seismic design. PhD thesis, Stanford University, 1995.
[31] Krawinkler H. New trends in seismic design methodology. In:
Proceedings 10th ECEE. The Netherlands: Rotterdam,
1995:821–30.