Advantages and Limitations of Adaptive and Non-Adaptive Force-Based Pushover Procedures

May 7, 2004 14:47 WSPC/124-JEE 00151
Journal of Earthquake Engineering, Vol. 8, No. 4 (2004) 497–522

c Imperial College Press
by UNIVERSIDAD DE CHILE SISTEMA DE SERVICIOS DE INFORMACION Y BIBLIOTECAS (SISIB) on 09/05/17. For personal use only.
ADVANTAGES AND LIMITATIONS OF ADAPTIVE AND

NON-ADAPTIVE FORCE-BASED PUSHOVER PROCEDURES
S. ANTONIOU
Department of Civil Engineering, Imperial College London
Imperial College Road, London SW7 2BU, UK
J. Earth. Eng. 2004.08:497-522. Downloaded from www.worldscientific.com
R. PINHO
European School for Advanced Studies in Reduction of Seismic Risk (ROSE School)
Collegio Alessandro Volta, Via Ferrata 17, 27100 Pavia, Italy
Received 25 November 2002

Revised 30 October 2003
Accepted 30 October 2003
The recent drive for use of performance-based methodologies in design and assessment
of structures in seismic areas has significantly increased the demand for the development
of reliable nonlinear inelastic static pushover analysis tools. As a result, the recent years
have witnessed the introduction of the so-called adaptive pushover methods, which,
unlike their conventional pushover counterparts, feature the ability to account for the
effect that higher modes of vibration and progressive stiffness degradation might have
on the distribution of seismic storey forces. In this paper, the accuracy of these force-
based adaptive pushover methods in predicting the horizontal capacity of reinforced
concrete buildings is explored, through comparison with results from a large number
of nonlinear time-history dynamic analyses. It is concluded that, despite its apparent
conceptual superiority, current force-based adaptive pushover features a relatively minor
advantage over its traditional non-adaptive equivalent, particularly in what concerns the
estimation of deformation patterns of buildings, which are poorly predicted by both types
of analysis.
Keywords: Force-based adaptive pushover; incremental dynamic analysis; FAP.
1. Introduction
Under the pressure of recent developments, seismic codes have begun to explicitly
require the identification of sources of inelasticity in structural response, together
with the quantification of their energy absorption capacity. Ideally, such perfor-
mance evaluation of structural systems subjected to earthquake loading should be
based on nonlinear time history analysis. However, the intrinsic complexity and the
additional computational effort required by the latter (especially if a fibre-based
distributed inelasticity modelling philosophy is adopted) do not justify its use in
ordinary engineering applications.
As a result of the above, nonlinear static, as opposed to dynamic, pushover
analysis has been gaining significance over recent years as a tool for assessment and
497
May 7, 2004 14:47 WSPC/124-JEE 00151
498 S. Antoniou & R. Pinho
design verification. Indeed, and despite its relative simplicity and ease of use, this
numerical tool can provide information on many important response characteristics
that cannot be obtained from an elastic static or dynamic analysis. Conventional

pushover consists in the application and monotonic increase of a predefined lateral
storey force pattern, kept constant throughout the analysis. However, such a proce-
dure exhibits a number of limitations, mainly related to its inability to account for
the progressive stiffness degradation, the change of modal characteristics and the
period elongation of a structure subjected to monotonic loading. As a result, recent
years have witnessed the introduction of the so-called adaptive pushover methods,
which overcome, at least from a conceptual viewpoint, such limitations.
In this paper, and following a brief review of latest developments in the field,
the accuracy of force-based adaptive pushover methods in predicting the horizontal

capacity of reinforced concrete buildings is explored, through comparison with re-
sults from a large number of nonlinear time-history dynamic analyses. An extended
version of the fully adaptive pushover algorithm proposed by Elnashai [2001], which
manages to encompass all advanced features that have been recently proposed and
developed by a number of researchers and can thus be deemed as a satisfactory
representative of the current state-of-the-art in the field, is employed.
The results of the extensive parametric study carried out, summarised in the
body of the current presentation, indicate that, despite its apparent conceptual
superiority, current force-based adaptive pushover features a relatively minor ad-
vantage over its traditional non-adaptive counterpart, particularly in what concerns
the estimation of deformation patterns of buildings, which are poorly predicted by
both types of analysis. Therefore, possible areas of improvement of the adaptive
scheme are also identified and suggested as potential solutions to the seemingly
inherent shortcomings of current force-based pushover analysis methods.
2. Nonlinear Static Pushover Analysis in Earthquake Engineering

Pushover analysis has served well as an efficient and easy-to-use alternative to
dynamic time-history analysis, since, despite its simplicity, it is capable of providing
important structural response information. Indeed, pushover can be employed to
identify critical regions, where inelastic deformations are expected to be high, and
strength irregularities in plan or elevation that might cause important changes in
the inelastic dynamic response characteristics [e.g. Krawinkler and Seneviratna,
1998]. In addition, pushover analysis can also provide realistic estimations of force
demands in potentially brittle elements, such as shear-dominated members, or the
consequences of strength deterioration of given members on the overall structural
stability. Finally, this type of analysis is also capable of predicting the sequence
of yielding and/or failure of structural components and the progress of the overall
capacity curve of the structure, thus verifying the adequacy of the seismic load path
(accounting for both structural and non-structural elements of the system).
May 7, 2004 14:47 WSPC/124-JEE 00151
Force-based Pushover Procedures 499

14 Table 1Each of the three groups should be sort of subdivided in subgroups of 2 lines. Hence, can you
14 Table 1
Each ofplease try to replicate the row spacing shown ininTable
the three groups should be sort of subdivided 1 included
subgroups afterHence,
of 2 lines. this list?
can you
15 However, pleaseReplace
Fig. 6 conventional try to replicate
image the row
with the spacing
one shownafter
included in Table
this 1list
included after this list?
15
pushover analysis exhibits also shortcomings and limita-
16 Fig. 6Table 2 Replace Can image with the one
the headings included
of Table 2 beafter this listcentred?
vertically
16 tions
16
that
Table 2 confine
line 2 Can its
the range
headings ofof application
Table
Replace “masses” with “mass”2 be and
vertically raise
centred? doubts about its effectiveness
16 to line 2
accurately Replace “masses”
estimate structural with “mass”
seismic demand, as demonstrated by a number
16 line 3 Replace “the response acceleration” with “seismic forces”
16 line 3 Replace “the response acceleration” with “seismic forces”
16 line
of researchers 4 [Lawson Replace “heights”
et al.,with 1994; with “height”
Krawinkler and Seneviratna, 1998; Kim and
16 line 4 Replace “heights” “height”
17 line 23 Replace “applications” with “application”
17 D’Amore,
17
line 23 1999; Replace
line 30
Naeim and Lobo,
“applications” 1999].
with For instance, the deformation predictions
“application”
Replace “0.5better” with “0.5% plots described the elastic (or better”
17 line 30 Replace “0.5better” with “0.5% plots described the elastic (or better”
can
18 be highly
line 8 inaccurateReplace if“(0.5”
higher with modes
“±0.5” are important and/or if the structure is
18 line 8 Replace “(0.5” with “±0.5”
18 18 line 12
pushed line 12 into
highly Replace
its “versus.”
nonlinear
Replace “versus.” with “versus” with “versus”
post-yield range. The latter is characterised by
20 20 line 5line 5 ReplaceReplace “at with
“at softening
four” four”“at
with “at
thethe the four”
four”
gradual
20 line 13
degradation
line 13 Replace
and
Replace “reasonable”
of structural
with “reasonably”
system during the analysis pro-
20 “reasonable” with “reasonably”
20 cedure,
20 line 15which
line 15 inReplace
turnReplace
leads to significant
“clear “clear
’global ’global
winner’, the”elongation
winner’, the”
with of the
with“global
“clear “clear periods
“global
winner”, the” (see
winner”, the” Fig. 1(a))
21 21 change
and line 6line of
6 modalReplaceReplace
shape “that” withcorresponding”
“that” characteristics
with “the “the corresponding”
(see Fig. 1(b)), as a result of the tendency
21 21 line 12 line 12 Replace Replace
“load”“load” with “loading”
with “loading”
for deformations
21 line 14
to concentrate
line 14 Replace Replace
at the locations
“configurations. Contrary”
that sustain more
with “configurations
damage.
and, contrary”
21 “configurations. Contrary” with “configurations and, contrary”

21 Moreover,
21 line 19 being
line 19 Replace a static
Replace method,
“achievement”
“achievement” pushover analysis
with performance”
with “good “good performance” reproduces material strain-
21 21 line 20 Remove “such”
ing only, neglecting other sources of energy dissipation that are associated with
line 20 Remove “such”
21 21 line 39 line 39 Replace Replace “instance,
“instance, inertia”inertia” with “instance,
with “instance, dynamicdynamic
inertia” inertia”
23
dynamic
23 line 16 response,
line 16 Replace such
Replace as“adaptive
“adaptive
kinetic energy
provides”
provides”
and“adaptive
with
with “adaptive
viscous
pushover
damping,
pushover
provides”provides” as well as dura-
23 tion effects.
23 line line 16
16 Finally,
Replace three-dimensional
Replace
“than”“than” with “over”
with “over” effects are difficult to incorporate, whereas
23 23 effects
the lineof17cyclic
line 17 ReplaceReplace “counterparts
“counterparts
earthquake with”
with” with
loading with “counterparts,
“counterparts,
cannot with” with”In summary, pushover
be modelled.
24 24 - - Reference
Reference “Bommer “Bommer et al. [1999]”
et al. [1999]” should should come“Bracci
come before before “Bracci et al. [1997]”
et al. [1997]”
24 analysis
24 - lacks
- many important
Reference
Reference features
“Campos-Costa
“Campos-Costa of&dynamic
& Pinto Pinto [1999]”
[1999]” nonlinear
should shouldbefore
come analysis
come and thus
before[1995]”
“CEN “CEN [1995]” can-
24 not
24 really linesubstitute
line 30 30 Replace the
Replace latter in the
“Brussels.”
“Brussels.” with withrole of theBelgium.”
“Brussels,
“Brussels, Belgium.” most accurate tool for structural
25 25 -
analysis -and assessment. Reference
Reference “Kim & “Kim & D’Amore
D’Amore [1999]”[1999]” shouldbefore
should come come“Krawinkler
before “Krawinkler and Seneviratna
and Seneviratna [1998]” [1998]”
25 25 - - Reference
Reference “Papanikolaou
“Papanikolaou [2000]”[2000]”
should should come“Paret
come before beforeet.
“Paret et. al. [1996]”
al. [1996]”
25 25 However,
- - andReplace
nonetheless,
Replace several possible developments can considerably im-
“Priestley,
prove the efficiency “Priestley, M.J.N.
of the method, [2003] Myths
M.J.N. [2003] and
bringing
Myths Fallacies in Earthquake
it Fallacies
and some steps Engineering,
to Revisited,
closer Engineering,
in Earthquake The Mallet
the realistic
Revisited, The Mallet
MilneMilne
Lecture, IUSS Press,
Lecture, IUSS Pavia,
Press, Italy.”
Pavia, Italy.”
modelling of nonlinear with with time-history analysis. Indeed, as shown further ahead, some
of the aforementioned “Priestley, limitations [2003]can
M.J.N.M.J.N.
“Priestley, Mythsbe
[2003] and overcome
Myths Fallacies within Earthquake
in Earthquake
and Fallacies theEngineering,
derivation of aRevisited
Revisited
Engineering, fully Press,
(IUSS (IUSS Press,
Pavia, Italy).”
adaptive procedure, Pavia,
which Italy).”
accounts for both higher mode contributions as well
25 - Replace
25 - Replace
as alteration of the local
“SeismoStruct resistance
[2003] Computer andprogram
modalforcharacteristics
static and dynamic of the structure,
nonlinear
“SeismoStruct [2003] Computer program for static and dynamic nonlinear analysis of framed
as
analysis of framed
structures [online]. Available from URL: http://www.seismosoft.com.”
induced by the progressive
with
structures accumulation
[online]. Availableof damage.
from URL: In this way, the
http://www.seismosoft.com.”stiffness degra-
with [2004] SeismoStruct - A computer program for static and dynamic nonlinear analysis
“SeismoSoft
“SeismoSoft
of framed structures[2004] SeismoStruct
[online]. Available from - A computer program for static and dynamic nonlinear analysis
URL: http://www.seismosoft.com.”
of framed structures [online]. Available from URL: http://www.seismosoft.com.”
___ before loading at peak capacity
___ before loading at peak capacity
6
period (sec)
6 first mode
second first
mode mode
period (sec)
5 third mode
Mode 1
5
second mode
third mode
Mode 1
4
4
3
3 ___ before loading at peak capacity
2 ___ before loading at peak capacity
2
1
Mode 2
1
0
Mode 2
0.0% 0.5% 1.0% 1.5% 2.0%
0 total drift
0.0% 0.5% 1.0% 1.5% 2.0%
total drift
(a) (a) (b)

(b)
Fig. 1. Changes in (a)
(a)periods of vibration and (b) mode shapes with deformation
(b) level.
Fig. 1. Changes in (a) periods of vibration and (b) modal shapes with deformation level.
Fig. 1. Changes in (a) periods of vibration and (b) mode shapes with deformation level.
May 7, 2004 14:47 WSPC/124-JEE 00151
dation, the period elongation and the influence of all significant modes of vibration
can be explicitly considered. In addition, finding a way to somehow incorporate
the expected ground motion in the analysis might also allow the attainment of
site-specific results, as required in some applications. This can be achieved with
the utilisation of response spectra representative of such areas, derived from actual
accelerograms or code provisions.
In recent years, a number of research endeavours have been carried out with the
objective of introducing the aforementioned developments. The majority of such
are briefly summarised in the subsections that follow.
2.1. Pushover procedures considering higher-mode effects

One of the first attempts to consider higher mode effects was made by Paret et al.
[1996] and Sasaki et al. [1998], who suggested the simple, yet efficient, multi-modal
pushover procedure (MMP). This comprises several pushover analyses under forc-
ing vectors representing the various modes deemed to be excited in the dynamic
response. The individual pushover curves are then converted to the Acceleration-
Displacement Response Spectrum (ADRS) format, after which the Capacity Spec-
trum Method [Freeman, 1975] is utilised to compare the structural capacity with
the earthquake demand. In this manner, it becomes evident which mode is more
critical and where damage is likely to occur. The procedure is intuitive and does
indeed provide qualitative information on higher mode effects, which conventional
single mode pushover analysis fails to highlight. However, the effects of these higher
modes cannot be easily quantified, since the method does not provide an estimation
of the response.
A refinement of the multi-modal pushover procedure is the PRC method
(Pushover Results Combination), which has been recently proposed by Moghadam
and Tso [2002]. According to this method, the maximum seismic response is again
estimated by combining the results of several pushover analyses, which are car-
ried out using load patterns that match the modal shapes of a predefined number
of vibration modes. The final structural response is obtained as a weighed (using
the respective modal participation factors) summation of the pushover results from
each analysis. Usually, the first 3 or 4 modes are considered.
A similar procedure, also based on MMP, is the MPA (Modal Pushover Analy-
sis) method suggested by Chopra and Goel [2002]. According to MPA, the pushover
curves corresponding to forcing vectors representing the various modes of vibra-
tion are idealised and transformed into bilinear curves of single-degree-of-freedom
(SDOF) equivalent systems, so as to calculate the target deformation and the corre-
sponding response parameters for each mode separately. The total demand is then
determined by combining the peak modal demands using the SRSS rule. Typically,
two or three modes are enough to achieve accurate results. When compared to
Response Spectrum Analysis, the method gave rather good estimates for response
parameters such as interstorey drifts or floor displacements, and the results pre-
May 7, 2004 14:47 WSPC/124-JEE 00151
sented were superior to the predictions of the pushover analyses with the fixed
force distributions suggested by FEMA [ATC, 1997], which greatly underestimated
drift demands.
Overall, the above multi-modal procedures constitute a significant improvement
of conventional pushover analysis. The former are theoretically more robust and
conceptually more attractive, since they explicitly consider the response of more
than one vibration mode and the influence of the expected ground motion, thus
yielding results that are closer to rigorous inelastic time-history analysis. However,
these methods do not account for damage accumulation and resulting modification
of the modal parameters, which might considerably affect the response character-
istics of a given structure. Therefore, the fully adaptive methods presented in the
following subsection seem to constitute a better alternative, due to their ability

to adapt the load applied to the structure at different deformation levels (i.e. at
different analysis steps), leading to conceptually more correct results.
2.2. Adaptive pushover procedures

Bracci et al. [1997] were the first to introduce a procedure that utilises fully adaptive
patterns. The analysis starts by assuming an initial lateral load distribution, usu-
ally triangular, whereas the additional loads imposed in subsequent increments are
calculated from the instantaneous base shear and storey resistances of the previous
load step. The procedure was implemented in the dynamic analysis package IDARC
[Valles et al., 1996] leading to the attainment of apparently promising results. Lefort
[2000] implemented an extended version of this method, which employed an addi-
tional force scaling equation to account for higher mode contributions, obtaining,
however, limitedly-accurate response predictions.
A different adaptive methodology was proposed by Gupta and Kunnath [2000],
in which the applied load is constantly updated, depending on the instantaneous
dynamic characteristics of the structure, and a site-specific spectrum can be used to
define the loading pattern. According to the method, eigenvalue analysis is carried
out before each load increment, accounting for the current structural stiffness state.
The number of modes of interest that are considered is predefined and the storey
forces for each mode are estimated as the product between the modal participation
factor, mass-normalised mode shape, weight of the storey and spectral amplification
of the mode being considered. Then, a static analysis is carried out for each mode
independently and the calculated action effects for each mode are combined with
SRSS and added to the corresponding values from the previous step. At the end of
the step, the structural stiffness state is assessed so as to be used in the eigenvalue
analysis of the next step. The estimates of interstorey drifts and the sequence of
the formation of local or global collapse mechanisms presented in the paper were
satisfactory. However, the employment of SRSS to combine the different pushover
responses, derived for each mode, implies that structural equilibrium is not satisfied
at the end of each step.
May 7, 2004 14:47 WSPC/124-JEE 00151
Another method proposed by Requena and Ayala [2000], who discussed two vari-
ations of adaptive pushover (referred to as approaches 2-A and 2-B) and compared
them with a modal fixed-pattern scheme. Whilst in one of the proposed variants
the storey loads were derived through SRSS combination of modal forces, in the
second option an “equivalent fundamental mode” is first determined (through an
SRSS combination of the vibration modes shapes) and then used to compute the
lateral loads. This procedure is repeated whenever the structural stiffness changes,
as new plastic hinges form and develop, hence the lateral load distribution reflects
the current state of inelasticity. The two alternative adaptive methods are indeed
appealing since they are theoretically rigorous and they explicitly account for higher
modes and spectral contributions. In addition, the procedures suggested in the pa-
per for the determination of the target displacement are also of particular interest,
albeit beyond the scope of the present work. However, and unfortunately, the an-
alytical results presented by Requena and Ayala [2000] were only limited and the
accuracy of the procedures could not be effectively assessed or judged.
A wholly alternative adaptive pushover methodology has been recently proposed
by Albanesi et al. [2002], who suggested an Adaptive Energy-based Pushover Anal-
ysis (AEPOA) whereby the imposed lateral force/displacement profiles at each
step supposedly take into account not only the inertial properties of the struc-
ture but also the kinetic energy that the latter is expected to mobilise when sub-
jected to earthquake loading. The results obtained, however, did not seem partic-
ularly exceptional, since, compared to conventional pushovers, the curves derived
by the AEPOA method were very unstable and tended to stop at very low defor-
mation levels (less than 1% in some cases), whereas the predicted values of the
base shear strength of the building were unreliable. In addition, no clear proce-
dure for applying and updating the lateral loads (forces or displacements) was
described.
Finally, Elnashai [2001] proposed an adaptive pushover scheme that seemed
to encompass, within a single-analysis pushover algorithm, all advanced features
described above. This single-run procedure is fully adaptive and multi-modal and
accounts for system degradation and period elongation by updating the force distri-
bution at every step (or at predefined steps) of the analysis. The dynamic properties
of the structure are determined by means of eigenvalue analyses that consider the
instantaneous structural stiffness state, at each analysis step. Site- or record-specific
spectral shapes can also be explicitly considered in the scaling of forces, so as to
account for the dynamic amplification that expected ground motion might have on
the different vibration modes of the structure. This force-based adaptive pushover
algorithm, deemed as a satisfactory representative of the current state-of-the-art
in the field, was further developed and tested by the authors of the present work,
for improved stability and accuracy, and is thus presented and discussed in greater
detail in the remaining parts of this presentation.
May 7, 2004 14:47 WSPC/124-JEE 00151
3. The Force-Based Adaptive Pushover Algorithm (FAP)

The force-based adaptive pushover algorithm adopted and developed within the
current presentation has been implemented in SeismoStruct [SeismoSoft, 2004], a

fibre-modelling Finite Element program for seismic analysis of framed structures,
which can be freely downloaded from the Internet. Full details on this computer
package can be found in its accompanying manual. The implementation of the pro-
posed algorithm can be structured in four main stages; (i) definition of nominal
load vector and inertia mass, (ii) computation of load factor, (iii) calculation of
normalised scaling vector and (iv) update of loading force vector. Whilst the first
step is carried out only once, at the start of the analysis, its three remaining coun-
terparts are repeated at every equilibrium stage of the nonlinear static analysis
procedure, as described in the following subsections.
3.1. Nominal load vector and inertia mass

In conventional pushover, a nominal load vector P0 is defined at the start of the
analysis. The magnitude of such vector is of reduced relevance since it does not affect
the attained results, due to the fact that pushover solution algorithms automatically
scale the load vector in order to meet the analysis target. Its real usefulness resides
instead on the fact that (i) it defines the structural nodes where the loads are
applied and (ii) it characterises the load distribution shape (triangular, uniform,
etc.) that is to be used throughout the entirety of the analysis.
In adaptive pushover, however, the loading vector shape is automatically defined
and updated by the solution algorithm at each analysis step, for which reason
the nominal vector P0 must always feature a uniform (rectangular) distribution
shape, in height, so as not to distort the load vector configuration determined in
correspondence to the dynamic response characteristics of the structure at each
analysis step (see Sec. 3.3). In other words, nominal loads, as defined at the start of
the analysis, must be equal at all storeys, whilst their magnitude may be arbitrarily
chosen (it is usually convenient to define an initial load vector with such magnitude
so that the analysis load factor λ can be made to range between zero and unity,
the latter corresponding to the attainment of full horizontal structural capacity, or
thereabouts).
In addition, and still within the realms of the definition of the start-up conditions
for the analysis, it is noteworthy that adaptive pushover requires the inertia mass
M of the structure to be modelled, so that eigenvalue analysis, employed in the
updating of the load vector shape (see Sec. 3.3), may be carried out. In the proposed
adaptive pushover algorithm, both lumped and distributed mass elements may be
employed, and freely spread throughout the structure.
The above highlights that, from a usability point-of-view, the implemented adap-
tive pushover algorithm effectively presents no additional effort and/or requirements
with respect to its conventional non-adaptive counterpart, since the only element
of novelty, in terms of analysis input, is the introduction of the building’s inertia
May 7, 2004 14:47 WSPC/124-JEE 00151
mass, which, however, can usually be obtained directly from the vertical gravity
loads, already included in any type of pushover analysis.
3.2. Computation of load factor

In the proposed algorithm, the magnitude of the loading vector P at any given
analysis step is obtained, in general terms, by the product of its nominal coun-
terpart P0 , defined above, and the load factor λ at that step (see Eq. (1)). The
latter is automatically increased, by means of a load control or response control
incrementation strategy, until a predefined analysis target, or numerical failure, is
reached.
P = λ · P0 . (1)
Load control refers to the case where the load factor is directly incre-
mented/controlled by the pushover algorithm whilst the structural response
(e.g. nodal displacements and rotations) corresponding to such loading level (force
level in this case) is subsequently determined. In other words, and in the case of
FAP, a force-controlled pushover is carried out, with the load factor being used to
scale directly the applied force vector until the point of peak capacity.
In the case of response control, on the other hand, it is the response of the
structure (e.g. a given nodal displacement or rotation) that is first directly incre-
mented/controlled, after which the load factor corresponding to such deformation
level can be computed. The load factor variation, therefore, is not prescribed by
the user, but is instead automatically calculated by the program so that the ap-
plied loading vector (force vector in this case) at a particular increment corresponds
to the attainment of the target response displacement/rotation at the controlled
node. It is noted that this type of load factor incremental scheme differs signifi-
cantly from a displacement-controlled analysis since it is the response deformation
of the structure, as opposed to a loading displacement vector, that is controlled by
the program.
A more detailed description of the workings of such loading/solution schemes
is beyond the scope of the current presentation, interested readers being instead
referred to the manual of the FE program, indicated earlier, where the proposed
adaptive pushover algorithm has been implemented. Here, it suffices to note that for
the purpose of the current parametric study, and in general for all force-based adap-
tive pushover applications, the response control algorithm scheme is the preferred
option since it (i) fully captures irregular response features such as soft-storey fail-
ures, (ii) models the softening post-peak loading/response branch of the structure,
(iii) provides an even distribution of force-displacement curve points throughout
both pre- and post-peak loading/response ranges and (iv) provides direct control
of structural deformations, which, as widely acknowledged, provide a better insight
into the damage incurred by the structure.
May 7, 2004 14:47 WSPC/124-JEE 00151
3.3. Calculation of normalised scaling vector

The normalised modal scaling vector, F̄ , used to determine the shape of the load
vector (or load increment vector) at each step, is computed at the start of each load
increment. In order for such scaling vector to reflect the actual stiffness state of the
structure, as computed at the end of the previous load increment, an eigenvalue
analysis is firstly carried out. To this end, the Lanczos algorithm [Hughes, 1987] is
employed to determine the modal shapes and participations factors of any given pre-
defined number of modes, after which the modal storey forces Fij can be determined
as:
Fij = Γj φij Mi , (2a)

where i is the storey number and j is the mode number, Γj is the modal participation
factor for the jth mode, φij is the mass normalised mode shape atvalue for the ith
before
beforeloading
loading atpeak
peakcapacity
capacity ___
___
storey
66
and the jth mode, and M i is the mass of the ith storey.
Alternatively, the modal storey forces can be computed using Eq. (2b), given
period (sec)
period (sec)
first
firstmode
mode
second
secondmode
mode Mode
Mode 11
below,
55
where
third
an additional parameter Sa,j that represents the acceleration response
thirdmode
mode
spectrum
44 ordinate corresponding to the period of vibration of the jth mode, here-
after referred to as spectral amplification, is entered into the computation of F ij . In
33
other words, the modal storey forces are weighted by beforethe S value at
loadinga
beforeloading
thecapacity
instanta-
peakcapacity
atatpeak ___
___
neous period of that mode, so as to take into account the effects that the frequency
22
content
11 of a particular input time-history, or spectrum, have in the response of the
Mode
Mode22
structure being analysed:
00
0.0%
0.0% 0.5%
0.5% 1.0%
1.0% 1.5%
1.5% 2.0%
2.0%
=drift
Fijtotal
total Γj φij Mi Sa,j .
drift (2b)
Previous research (a)

(a)[Mwafy and Elnashai, 2000; Antoniou, 2001] (b) indicated that
(b)
Fig.
consideringFig.1.
1.Changes
the inin(a)
spectral
Changes periods
periodsof
ofvibration
amplification
(a) and
and(b)
of each
vibration mode
modeshapes
particular
(b) with
withdeformation
mode
shapes in level
the computation
deformation level of
the Fij , that is, the use of Eq. (2b) in favour of Eq. (2a), could contribute to an im-
provement in the similitude between static pushover and dynamic inelastic analysis
5000
5000
base shear (kN)
base shear (kN)

base shear (kN)
base shear (kN)
6000
6000
4000
4000
5000
5000
3000
3000 4000
4000
3000
3000
2000
2000
2000
2000
dynamic
dynamic dynamic
dynamic
1000
1000 with
withspectral
spectralamplification
amplification with
withspectral
amplification
without
withoutspectral
amplification 1000
1000 without
withoutspectral
amplification
00 00
0%0% 1%
1% 2%
2% 3%
3% 0%0% 1%
1% 2%
2% 3%
3%
total
totaldrift
drift total
totaldrift
drift
(a) Model
(a)
(a) ModelWM15-AR
Model WM15-AR
WM15-AR (b) Model
(b)
(b) ModelRM15-AR
Model RM15-AR
RM15-AR
Fig.2.2.Effect
Fig. Effectof
ofspectral
amplificationin
inresults
resultsobtained
obtained
Fig. 2. Effect of spectral amplification in adaptive pushover results.
normalised
normalised nominal
nominal
shapeatatstep
shape stept t loadvector
load vector newforces
new forcesapplied
appliedat
atstep
steptt
PPt t == [[λλt t]] ×× ×× =

=
normalised = 1
normalised
∑∑ shape = 1
Fig. ( ( representation
Fig.3.3. Graphical
Graphical ))
shape
representation ofloading
of loadingforce
forcevector
vectorcalculation
calculationwith
withtotal
totalupdating
updating
May 7, 2004 14:47 WSPC/124-JEE 00151
results. These observations were confirmed in the present study, where the inclu-
sion of ground motion characteristics in the adaptive pushover analysis (through
the use of response spectra) provided an equally closer fit to the dynamic results,
as shown in Fig. 2. Although in several cases the predictions with and without
spectral amplification coincided, in none did the pushover with spectral amplifica-
tion (i.e. different Sa for each mode) perform worse. Therefore, the utilisation of
a spectral shape, either code-defined or derived from actual records, to determine
modal storey forces is, in general, recommended.
Ideally, in order to assure full consistency between demand and supply duc-
tilities, multiple response spectra, derived for varying values of equivalent viscous
damping, should be employed so as to reflect the actual energy dissipation charac-
teristics of the structure at each deformation level (i.e. at each analysis step). The
implementation of such refinement, however, was beyond the scope of the current
work, for which reason a single constant response spectrum derived for an equiva-
lent viscous damping value of 5%, which the authors acknowledge may contribute
to an overestimation of the first mode contribution, was used throughout the en-
tirety of each analysis. Future work will target the assessment of the effects that
such apparent limitation has on attained results.
The lateral load profiles of each vibration mode are then combined using ei-
ther the Square Root of the Sum of Squares (SRSS, Eq. 3(a)), if the modes can
be assumed as fully uncoupled, or the Complete Quadratic Combination (CQC,
Eq. 3(b)) method, if cross-coupling of modes and respective viscous damping is to
be considered. It is noted that in Eq. (3b), n stands for the number of modes whilst
ρjk is the cross-model correlation coefficient, which can be approximated by Eq. (4)
if one assumes all modes to feature the same damping coefficient ξ:
v
u n 2
uX
Fi = t Fij , (3a)
j=1
v
u n X
n
uX
Fi = t (Fij · ρjk · Fik ) , (3b)
j=1 k=1
8 · ξ 2 · (1 + r) · r1.5 ωk
ρjk = , r= . (4)
(1 − r2 )2 + 4 · ξ 2 · r · (1 + r)2 ωj
Further, since only the relative values of storey forces (Fi ) are of interest in the
determination of the normalised modal scaling vector F̄ , which defines the shape,
not the magnitude, of the load or load increment vector (see Sec. 3.4), the forces
obtained by Eqs. (3a) or (3b) are normalised with respect to the total value, as
follows:
Fi
F̄i = . (5)
ΣFi
May 7, 2004 14:47 WSPC/124-JEE 00151
It is noted that, as highlighted by Priestley [2003], the use of SRSS or CQC rules
to combine modal forces may not always lead to the attainment of an adequate load
vector shape that accurately represents the dynamic response characteristics of the
structure at a particular deformation level. In fact, if true reproduction of dynamic
analysis response is the objective of the pushover analysis, then the normalised
scaling vector should ideally be obtained through a weighted vectorial addition
of the contribution of each mode, to avoid the not necessarily realistic storey force
increase when the force vector from one mode is summed, through SRSS or CQC, to
the fundamental one. The employment of such alternative modal force combination
procedure, however, calls for additional in-depth studies that are conspicuously
beyond the scope of the present endeavour. Hence, and taking also account of the
fact that in the present work observed modes of vibration at every deformation
stage were always sufficiently apart, the SRSS modal combination procedure was
consistently employed throughout the analyses.
It is equally noteworthy that when the structural response reaches its post-
peak range, the eigen-solver can no longer output real eigen-solutions, due to the
presence of negative values in the diagonal of the stiffness matrix which in turn lead
to imaginary periods of vibration, corresponding to wholly unfeasible modal shapes.
In such cases, the load vector shape is no longer changed (only its magnitude is
updated), effectively meaning that a conventional non-adaptive pushover analysis
is employed thereafter. This is acceptable since, in the majority of applications,
entrance in the post-peak response range corresponds to the formation of a given
failure mechanism, which then remains qualitatively unchanged until collapse is
reached. Furthermore, and as described in Sec. 4.3, when a given storey starts failing
due to damage accumulation, force-based adaptive algorithms tend to generate
unrealistically high load concentrations at that given storey, leading to equally
unfeasible high drift values. Hence, and for this reason also, the employment of
a constant load vector shape in the post-peak softening structural response range
seems justified.
Finally, in cases where a very large number of analysis steps are employed and/or
when the structural model is very large, it might prove advantageous for the load
vector shape to be updated with a frequency lower than the number of analysis
steps, so as to reduce the computation effort. Such feature is explicitly available
in the algorithm developed by the authors. Within the framework of the current
parametric study, however, for reasons of accuracy and analysis stability, the load
vector shape was updated at every analysis step, with up to ten modes of vibration
being considered in its computation.
3.4. Update of loading force vector

Once the normalised scaling vector and load factor or load factor increment have
been determined, and knowing also the value of the initial nominal load vector,
the loading force vector at a given analysis step can be updated using one of two
alternatives; total or incremental updating.
May 7, 2004 14:47 WSPC/124-JEE 00151
normalised nominal
shape at step t load vector new forces applied at step t
Pt = λt × × =
∑=1
Fig. 3. Graphical representation of loading force vector calculation with total updating.
Fig. 3. Graphical representation of loading force vector calculation with total updating.
3.4.1. Total Updating

With total updating, the load vector Pt at a given analysis step t is obtained through
a full substitution of the existing balanced loads (load vector at previous step) by
a newly derived load vector, computed as the product between the current total
load
at factor λt ,drift
0% total the current
at normalised modal scaling
0.5% total drift at 1.0%vector F̄t and the
total drift nominal
at 1.5% total load
drift
vector P0 , as schematically represented in Fig. 3 and numerically translated into
Fig. 4. Total Updated loading force vector at different deformation levels (model WH30-NR1).
Eq. (6a):
normalised = λt · F̄t · P0 .new increment

Ptnominal (6a)
shape at step t load vector of forces
This load updating scheme does seem to be the most adequate if one is looking
∆Pt = ∆λt ×in full the dynamic
to reproduce × response characteristics
= of a structure subjected
to earthquake loading, since the load vector is fully updated at every single analysis
∑=1
step to reflect the stiffness state, hence the vibration properties, of the structure at
that particular deformation level. This effectively means that an adaptive pushover
new increment
existing forces of forces new forces applied at step t
that has been started with a, say, first mode of vibration triangular force distribu-
tion, may then be concluded with an applied force distribution that reflects a second
Pt=Pt-1+∆Pt= + =
or third mode of vibration, which might be controlling the response of the structure
at that stage of the analysis. This behaviour was clearly observed in some of the
case studies considered in the parametric study described in Sec. 4, as schematically
Fig. 5. Graphical representation of loading force vector calculation with incremental updating.
illustrated in Fig. 4 below.
Unfortunately, in addition to its auspicious full reproduction of inertia force dis-
tribution patterns observed in nonlinear dynamic analysis, Total Updating features
normalised nominal
also a conspicuous
Table 1.lack
shape of numerical
Definition
at step t stability,
ofload
the due
structural
vector to the
systems
new forces difficulty
applied at step t in introducing
considered.
the abrupt type of storey force changes depicted in Fig. 4 below. This is partic-
ularly evident ×
λt those
Pt = in cases where × the applied = force at one storey level reduces
Structure Structural No. of Ductility Design Behaviour T
Group with an increasing load factor, ∑=1 whereas the forces at the adjacent levels augment. uncracked
In 3.
Fig. reference
hisGraphical
work, Antoniou system
[2002] does
representation storeys
presentforce
of loading levelcalculation
a solution
vector to thisPGA (g) total
numerical
with factor (q) if (sec)
obstacle;
updating.
at one or more storeys there is reduction of the applied forces, all storey loads are
RF-H030 / RH30 High 5.00 0.697
0.30
RF-M030 / RM30 Regular Medium 3.75 0.719
1 12
frame Medium 3.75 0.745
RF-M015 / RM15 0.15
RF-L015 / RL15 Low 2.50 0.740
at 0% total drift at 0.5% total drift at 1.0% total drift at 1.5% total drift
IF-H030
Fig. 4.Fig.
Total / IH30
4. Updated loading
Total-Updated force
loading vector
force at atdifferent
vector
High
different deformation levels
deformation levels (model
(model
4.00 0.565
WH30-NR1).
WH30-NR1).
0.30
IF-M030 / IM30 Irregular Medium 3.00 0.536
2 normalised nominal 8 new increment
IF-M015 / IM15 frame
shape at step t load vector Medium
of forces 3.00 0.613
0.15
IF-L015
∆Pt = ∆λ/t IL15
× × = Low 2.00 0.614
∑=1
FW-H030 / WH30 High 3.50 0.569
new increment 0.30
FW-M030 / WM30 existing forces
Regular Medium
of forces 2.625
new forces applied at step t 0.557
3 8
Pt=Pt-1+∆P t=
frame-wall + Medium
= 2.625 0.601
FW-M015 / WM15
May 7, 2004 14:47 WSPC/124-JEE 00151
proportionally increased so that at any given storey the loading force is at least
equal to that of the previous step. This loading vector corrective technique, how-
ever, cannot, for obvious reasons, be employed with a response control algorithm,
the preferred choice for force-based adaptive pushover (Sec. 3.2, above). There-
fore, Total Updating has not been adopted in the current parametric study, being
instead replaced by the stable, and comparably accurate, Incremental Updating
scheme, described below.
3.4.2. Incremental Updating

With Incremental Updating, the load vector Pt at a given analysis step t is obtained
by adding to the load vector of the previous step Pt−1 (existing balanced loads) a
newly derived load vector increment, computed as the product between the current
load factor increment ∆λt , the current modal scaling vector F̄t and the nominal
load vector P0 , as schematically represented in Fig. 5 and numerically translated
into Eq. (6b).
normalised nominal
shape at step t new forces applied at step t
Pt = Pt−1load vector
+ ∆λ t · F̄t · P0 . (6b)
Pt = λt × that the results
It is noteworthy × obtained with = Incremental Updating were con-
sistently very similar to those ∑=1 obtained with Total Updating, in those cases where
Fig. 3.theGraphical
latter couldrepresentation of loading
indeed be applied. force to
This seems vector
comecalculation
as a result ofwith
thetotal
fact updating.
that
in both schemes the trends of the load distributions are similar, as shown by Anto-
niou [2001, 2002], even though the absolute values of the forces differ significantly
in some cases. In addition, even in those limited cases where differences between the
two load updating modalities were not negligible, it was by no means clear which
variant provided better estimates. Finally, it is also noted that Bracci et al. [1997],
Gupta and Kunnath [2000] and Requena and Ayala [2000], whose work has been
briefly
at 0%reviewed
total driftin Sec.at2 0.5%
of this paper,
total drift have also adopted
at 1.0% total drift the use of incremental
at 1.5% total drift
load vector updating.
Fig. 4. Total Updated loading force vector at different deformation levels (model WH30-NR1).
normalised nominal new increment

shape at step t load vector of forces
∆Pt = ∆λt × × =
∑=1
new increment
existing forces of forces new forces applied at step t
Pt=Pt-1+∆Pt= + =
Fig. 5. Graphical
Fig. 5. Graphical representation
representation of loadingforce
of loading force vector
vectorcalculation with incremental
calculation updating. updating.
with incremental
Table 1. Definition of the structural systems considered.
Structure Structural No. of Ductility Design Behaviour Tuncracked

Group
reference system storeys level PGA (g) factor (q) (sec)
RF-H030 / RH30 High 5.00 0.697

May 7, 2004 14:47 WSPC/124-JEE 00151
4. Parametric Study
In what follows, sample results of an extensive parametric study that aimed at
assessing the effectiveness of the proposed force-based adaptive pushover algorithm

(FAP), are presented. To this end, results obtained with the latter are compared
to well-established static and dynamic analysis methods, all of which have been
applied to a large number of case studies, as discussed below.
4.1. Introduction
Adaptive pushover is intended to be a method for general use in design and as-
sessment. It is therefore imperative to verify its efficiency for different structural
configurations subjected to equally diverse input ground motions. Extensive pre-

liminary testing, comprising 1200 dynamic and pushover analysis on 12 different
“stick-models”, has already been carried out by Antoniou [2001], for which reason
the current study focused instead in the application of the proposed algorithm to
more realistic structural models. A series of frame systems were thus considered and
analysed, using conventional pushover procedures as well as the suggested adaptive
analysis scheme. Furthermore, and in order to fully assess the method’s accuracy,
or lack of, a considerable number of dynamic analyses were conducted, with the
corresponding results being then compared with those from force-based adaptive
pushover.
Three different structural configurations were employed in this study; a 12-
storey regular frame, an eight-storey irregular frame and a dual (wall-frame) system.
Moreover, different ductility classes and design ground accelerations were consid-
ered, resulting in a total of 12 structural models. The latter represent common
reinforced concrete structures and are based on buildings designed and detailed at
the University of Patras [Fardis, 1994], seemingly according to the 1995 version of
Table 1. Definition of the structural systems considered.
Structure Structural No. of Ductility Design Behaviour Tuncracked

Group Reference System Storeys Level PGA (g) Factor (q) (sec)
RF-H030/RH30 High 5.00 0.697
0.30
RF-M030/RM30 Regular Medium 3.75 0.719
1 12
RF-M015/RM15 Frame Medium 3.75 0.745
0.15
RF-L015/RL15 Low 2.50 0.740
IF-H030/IH30 High 4.00 0.565
0.30
IF-M030/IM30 Irregular Medium 3.00 0.536
2 8
IF-M015/IM15 Frame Medium 3.00 0.613
0.15
IF-L015/IL15 Low 2.00 0.614
FW-H030/WH30 High 3.50 0.569
0.30
FW-M030/WM30 Regular Medium 2.625 0.557
3 8
FW-M015/WM15 Frame-Wall Medium 2.625 0.601
FW-L015/WL15 Low 0.15 1.75 0.588
May 7, 2004 14:47 WSPC/124-JEE 00151

coupling beam
12 × 3.0 = 36.0 m
structural
4.5+7 × 3.0 = 25.5 m

wall
8 × 3.0 = 24.0 m
REGULAR SYSTEM IRREGULAR SYSTEM DUAL SYSTEM
Fig. 6. Geometric characteristics of the regular, irregular and dual systems.
Eurocode 8 [CEN, 1995]. Subsequently, they were modelled by Mwafy [2001] under
the framework of a different project, and were then adapted by Rovithakis [2001]
for the purpose of the current research endeavour. Their general characteristics are
defined in Table 1 and schematically illustrated in Fig. 6.
It is also worth noting that many of the modelled buildings feature a column-
sway, rather than beam-sway, type of failure mechanism, and are thus likely to
experience significant modifications of the applied force vector shape at different
deformation levels. This serves well the purpose of the current study, since it is in
such non-beam-sway cases that conventional non-adaptive pushovers struggle the
most to accurately reproduce the dynamic response characteristics of a building
subjected to a given input motion.
Four input time-histories, consisting of one artificially-generated accelerogram
[Campos-Costa and Pinto, 1999] and three natural records (Loma Prieta earth-
quake, USA, 1989), were employed for the dynamic analyses of the study. The
selection of these four records aimed at guaranteeing that the twelve buildings de-
scribed above would be subjected to a wide-ranging type of earthquake action, in
terms of frequency content, peak ground acceleration, duration and number of high
amplitude cycles. Indeed, the original PGA of the records varies between 0.12g and
0.93g, the spectral shapes are markedly distinct and provide high amplifications at
different periods (see Fig. 7), and the ratio between the significant duration (defined
as the interval between the build up of 5% and 95% of the total Arias Intensity
[Bommer and Martinez-Pereira, 1999]) and the total duration ranges from 22% to
72%. The characteristics of the records are summarised in Table 2, whereas their
elastic response spectra for an equivalent viscous damping of 5% are shown in Fig. 7.
The twelve building configurations subjected to four records resulted in a total
number of 12 × 4 = 48 test cases, deemed adequate to provide a representative sam-
ple for the verification of the proposed algorithm. Two conventional pushover pro-
cedures were carried out for each of the 12 buildings, for which the two distributions
May 7, 2004 14:47 WSPC/124-JEE 00151
Table 2. Characteristics of the records employed in this study.
Peak Peak
DE CHILE SISTEMA DE SERVICIOS DE INFORMACION Y BIBLIOTECAS (SISIB) on 09/05/17. For personal use only.
Record Ground Response 5% AI 95% AI Total Significant teff /ttotal

Acceleration Acceleration Threshold Threshold Duration ttot Duration teff
AR 0.30g 1.28g 2.32 sec 12.75 sec 15.0 sec 10.43 sec 69.5%
NR1 0.25g 0.90g 11.23 sec 20.16 sec 40.0 sec 8.93 sec 22.3%
1.0 1.0 1.0 1.0

1.0 1.0 1.0 1.0
0.8 0.8 0.8 0.8

0.8 0.8 0.8 0.8
(g)(g)
(g) (g)
(g)(g)
(g) (g)
0.6 0.6
acceleration
0.6 0.6
acceleration
acceleration
acceleration
0.6 0.6
acceleration
0.6 0.6
acceleration
acceleration
acceleration
0.4 0.4 0.4 0.4

0.4 0.4 0.4 0.4
0.2 0.2 0.2 0.2

0.2 0.2 0.2 0.2
0.0 0.0 0.0 0.0

0 0 1 1 2 2 3 3 0 0 1 1 2 2 3 3
0.0 0.0 0.0 0.0
0 0 1 1 2 2 3 3 period (sec)
period (sec) 0 0 1 1 2 2 3 3 period (sec)
period (sec)
period (sec)period (sec) period (sec) period (sec)
(a) Record
(a) AR
(a)Record
RecordAR
AR (b)Record
(b) Record
(b)NR1
Record NR1
NR1
(a)AR
(a) Record Record AR (b)NR1
(b) Record Record NR1
0.5 0.5 3.0 3.0
0.5 0.5 3.0 3.0
0.4 0.4 2.4 2.4

0.4 0.4 2.4 2.4
(g) (g)
(g) (g)
(g)(g)
(g)(g)
0.3 0.3 1.8 1.8

acceleration
acceleration
acceleration
acceleration
0.3 0.3 1.8 1.8

acceleration
acceleration
acceleration
acceleration
0.2 0.2 1.2 1.2

0.2 0.2 1.2 1.2
0.1 0.1 0.6 0.6

0.1 0.1 0.6 0.6
0.0 0.0 0.0 0.0
0 0 1 1 2 2 3 3 0 0 1 1 2 2 3 3
0.0 0.0 period (sec) 0.0 0.0 period (sec)
0 0 1 1 2 2 3 3 period (sec) 0 0 1 1 2 2 3 3 period (sec)
(c) Record period (sec)period (sec) period (sec) period (sec)
(c) NR2
Record NR2 (d) Record
(d)NR3
Record NR3
(c) Record
(c)NR2
Record NR2 (d) Record
(d)NR3
Record NR3
Fig. 7. (c)
Fig. 7. Elastic Record
response
Elastic NR2 spectra
spectra
response of the four records
of the (d)equivalent
Record
(5% equivalent
four records (5% NR3
viscous damping)
viscous damping)
Fig. 7. Elastic
Fig. 7. response spectra of
Elastic response the four
spectra records
of the (5% equivalent
four records viscous viscous
(5% equivalent damping)damping)
Fig. 7. Elastic response spectra of the four records (5% equivalent viscous damping).
proposed in the NEHRP Guidelines [ATC, 1997] were employed; the uniform dis-
UNIVERSIDAD
tribution, whereby lateral forces are proportional to the total mass at each floor
8000 8000
level, and the triangular distribution, in which seismic forces are proportional to
(kN)
(kN)
(kN)
(kN)
8000 8000
12000
(kN)
the product of floor mass and storey height. Furthermore, in excess of ten adaptive
(kN)
(kN)
12000
(kN)
7000 7000
by shear
shear
shear
12000
11000 12000
shear
7000 7000 11000

pushovers were run per building, employing 10000 the 11000
different variants of FAP discussed
shear
shear
shear
shear
6000 11000
6000 10000
base
base
base
base
6000 6000
in Sec. 3. The results of the 144 pushover10000 analyses
9000
9000carried out were then compared
10000
base
base
base
5000
base
5000 9000
8000 9000
5000to the envelopes derived with the recently proposed Incremental Dynamic Analysis
8000
5000
8000
7000 8000
4000 7000
4000
4000 4000
(IDA) procedure [Hamburger et al., 2000; 7000Vamvatsikos
6000 7000
6000
and Cornell, 2002; Mwafy
3000
3000and Elnashai, 2000; Papanikolaou, 2000], whereby a structure is subjected to a
6000
5000 6000
5000
3000 3000 5000
4000 5000
2000
2000series of nonlinear time-history
dynamic
adaptive
analysis
dynamic of
3000
4000
increasing
4000
4000
intensity (e.g. peak ground
dynamic
adaptive
dynamic
dynamic adaptive 3000 dynamic adaptive
2000 2000 uniform dynamic uniform dynamic
adaptive uniform 3000
2000 3000 adaptive uniform
1000 triangular adaptive 2000 triangular adaptive
1000 uniform triangular
uniform uniform triangular
1000
2000 2000 uniform
1000 1000 triangular triangular 1000 triangular triangular
0 10000 10000
0% 0 1% 2% 3% 0% 1% 2% 3%
0 0% 1% 2% 3% 0 0% 1% 2% 3%
0 total drift 3% 0 total drift
0% 0% 1% 1% 2% 2% total drift 3% 0% 0% 1% 1% 2% 2% 3% drift 3%
total
total drift total drift total drift total drift
(a) Model
(a)RL15-NR1
Model RL15-NR1 (b) Model
(b)WM30-NR3
Model WM30-NR3
Fig.(a) Model
(a)RL15-NR1
8. Adaptive Model
and conventional
Fig. 8. Adaptive
(b) Model
RL15-NR1 pushover curves vs. dynamic (b)WM30-NR3
Model
analysis
and conventional pushover curves vs. dynamic
WM30-NR3
envelopes
analysis envelopes
Fig. 8. Adaptive and conventional pushover curves vs. dynamic analysis envelopes
May 7, 2004 14:47 WSPC/124-JEE 00151
acceleration is incrementally scaled from a low elastic response value up to the

attainment of a pre-defined post-yield target limit state). To this end, about 20
inelastic dynamic analyses with variable scaling factors were performed per case,
leading to a total of approximately 1000 nonlinear time-history runs.
All the analyses, dynamic or static, were carried out up to the point of 3%
global drift, employing the previously introduced FE package, capable of predicting
the large displacement behaviour of space frames under static or dynamic loading,
taking into account both local (beam-column effect) and global (large displace-
ments/rotations effects) geometric nonlinearities as well as material inelasticity. The
spread of the latter along the member length and across the section area is explicitly
represented through the employment of a fibre modelling approach, implicit in the
formulation of the inelastic beam-column frame elements employed in the analyses.

Structural members were subdivided into 4–5 elements, with smaller elements at
member ends so as to ensure that inelasticity could be accurately modelled. Beams
and columns were modelled as extending from the centre of one beam-column joint
to the centre of the next, in order to take account, albeit in an indirect and empiri-
cal manner, of joint flexibility that could be induced by joint shear distortion, yield
penetration and/or bar slip. Inertia mass was taken as permanent vertical load plus
30% of its variable counterpart. No viscous damping was considered in the dynamic
analysis, since energy dissipation through hysteresis is already implicitly included
within the nonlinear fibre model formulation of the inelastic frame elements, and
non-hysteretic type of damping was assumed to be negligible within the scope of
the present application.
In order to appraise the applicability and the effectiveness of force-based adap-
tive pushover, a series of top displacement versus base shear plots (capacity curves)
has been created, comparing the adaptive and conventional pushover results against
the dynamic analysis envelopes obtained for each of the four accelerograms de-
scribed above. The interstorey drift and storey shear profiles, more representative of
local response characteristics, at four different deformation levels (0.5%, 1.0%, 1.5%
and 2.5% total drift) have also been plotted. Typically, the 0.5% plots described
the elastic (or better, pre-yield) behaviour of the buildings, those of 1.0–1.5% sig-
nalled the point of global yielding, where the stiffness changes significantly and the
load distributions are rapidly updated, whilst 2.5% global drift is deep within the
inelastic range.
Finally, it is noted that by comparing pushover results with IDA output obtained
for each single accelerogram, as opposed to the statistical average of all dynamic
cases, a much more demanding and precise assessment of the static procedures is
effectively carried out, since structural response peculiarities introduced by indi-
vidual input motions are not smoothed out through results averaging. Within this
non-statistical verification framework, and in order to facilitate interpretation of
the most important observations and exemplification of the significant conclusions,
only representative plots are given henceforth.
May 7, 2004 14:47 WSPC/124-JEE 00151
4.2. Capacity curves

As noted in Sec. 3, the most advantageous force-based pushover variant, when both
numerical stability and results accuracy are considered, is that featuring response
control, spectral amplification and incremental updating. Consequently, in all plots
provided hereafter, only such FAP configuration will be compared with conventional
pushover (uniform and triangular distributions), as well as IDA results.
It is also noted that in the current work the dynamic analysis envelopes consist
of the
1.0
locus of maximum total drift versus 1.0corresponding base shear (i.e. peak
base shear1.0
within a time-window (±0.5 seconds 1.0
of the instant of maximum drift
0.8 0.8
occurrence).
0.8 This reflects a pragmatic, rather than
0.8 judicious, choice, since it has
acceleration (g)
acceleration (g)
not been possible to unequivocally demonstrate that any of the three available
acceleration (g)
0.6
0.6
acceleration (g)
0.6 0.6
choices
0.4 for plotting dynamic analyses envelopes 0.4 (max drift versus max shear, max
0.4 0.4
drift
0.2
versus corresponding shear, corresponding 0.2
drift versus max shear) is more
meaningful0.2
than the other, for all structural types 0.2
and analysis cases. Future work
0.0 0.0
will hopefully
0 0.0
0
1 clarify
1
2 this
2
matter,
3
period
3
with
(sec)
the authors 0.0 not excluding
0
0
1
1
2 the possibility
2
3
3period (sec)
that
the use of only (a) one
Recordof ARthe three possible dynamic analysis
period (sec)
results
(b) Record NR1 interpretation
period (sec)
(a) Record AR (b) Record NR1

to 0.5
match 0.5
an equally unique force-based adaptive 3.0
3.0
pushover might be deemed as
inappropriate.
0.4 2.4
0.4 2.4
In Fig. 8, representative cases of comparative plots for models with different
acceleration (g)
acceleration (g)
acceleration (g)
acceleration (g)
0.3 1.8
structural0.3 characteristics, and for different records, 1.8 are shown. It is observed that
the IDA
0.2
0.2 points lay between the uniform and the 1.2
1.2 triangular distributions, at least
within
0.1
0.1
the pre-yield range (some noticeable 0.6 exceptions
0.6
where these two distribu-
tions
0.0
did not provide upper and lower bounds, 0.0
even at pre-yield range, have been
0 0.0 1 2 3 00.0 1 2 3
recorded, 0 as discussed
1 2later inperiod
3this section).
(sec)
period (sec)
The adaptive
0 1 pushover 2 curves are
3period (sec)typi-
period (sec)
cally located between (c) Record NR2
the two conventional curves as well, providing (d) Record NR3 a slightly closer
(c) Record NR2 (d) Record NR3
Fig.Fig.
7. Elastic response spectra of of
thethe
four records (5% equivalent viscous
viscousdamping)
fit to the IDA 7. points.
Elastic response spectra four records (5% equivalent damping)
In addition to the above, however, there have also been several examples where
the pushover methods (adaptive or not) failed to correctly predict the dynamic
8000
8000
base shear (kN)
base shear (kN)

base shear (kN)
base shear (kN)
12000
12000
7000 7000
11000
11000
6000 6000 10000
10000
90009000
5000 5000
80008000
70007000
4000 4000
60006000
3000 3000 50005000
40004000
2000 2000 dynamic
dynamic dynamic
dynamic
adaptive
adaptive 30003000 adaptive
adaptive
uniform
uniform 20002000 uniform
uniform
1000 1000 triangular
triangular triangular
triangular
10001000
0 0 0 0
0% 0% 1% 1% 2% 2% 3% 3% 0% 0% 1% 1% 2%2% 3%3%
total driftdrift
total total drift
total drift
(a)(a)
(a) Model
Model
Model RL15-NR1
RL15-NR1
RL15-NR1 (b)
(b)
(b) Model
Model
Model WM30-NR3
WM30-NR3
WM30-NR3
Fig.Fig. 8. Adaptive
8. Adaptive andand conventional
conventional pushover
pushover curves
curves vs.vs. dynamic
dynamic analysisenvelopes
analysis envelopes
Fig. 8. Adaptive and conventional pushover curves versus dynamic analysis envelopes.
44
May 7, 2004 14:47 WSPC/124-JEE 00151
structural behaviour throughout the entire deformation range, even in the pre-yield
region. The most serious complications have been caused by record NR2, which fea-
tures an acceleration response spectrum with large amplifications for periods up to

0.4 sec followed by a very steep descending branch (see Fig. 7), which, associated to
the softening of the building’s response and consequential period elongation, lead
to scenarios where higher modes dominate the structural behaviour of the stud-
ied structures. This resulted in highly irregular interstorey drift and storey shear
profiles, shown in Sec. 4.3, and accordingly peculiar IDA envelopes. Consequently,
none of the force-based pushover procedures employed here, including the adap-
tive algorithm, managed to capture the characteristics of the dynamic response of
structural systems such as those illustrated in Fig. 9.
As noted in Fig. 9, in some cases, the dynamic base shear was considerably
higher than the static shear estimates provided by the uniform distribution (or, less
frequently, lower than the triangular). These are induced by the highly irregular
shapes of the inertia forces at the time step of maximum displacement, which cannot
be adequately reproduced, throughout the entire deformation range, by any of the
force-based static procedures employed in this work. Moreover, on many of such
cases, the slope of the IDA envelope was seen to change abruptly at a certain
point of the curve (e.g. Fig. 9(a), drift of 1.7%), due to the fact that, for some of
these structures, when subjected to a given accelerogram with a certain frequency
content, the increase of scaling factors, thus deformation demand and hence periods
of vibration, would at a certain point introduce a sudden change in the dominant
base shear (kN)
9000
response mode of vibration, which would then lead to a markedly diverse response
8000
trend, clearly identified in the dynamic analysis
7000
envelopes. Similar observations
have been made by Vamvatsikos and Cornell 6000 [2002] and Antoniou [2001].
5000
4000
4.3. Interstorey drifts and shear profiles
3000
dynamic
2000 adaptive
Interstorey drifts are crucial parameters in terms of structural response,uniform
especially
triangular
1000
in view of the recent developments in the field
0
of performance-based engineering,
0% 1% 2% 3%
total drift
base shear (kN)
base shear (kN)
9000 3000
8000
2500
7000
6000 2000
5000
1500
4000
3000 1000
dynamic dynamic
2000 adaptive adaptive
uniform 500 uniform
1000 triangular triangular
0 0
0% 1% 2% 3% 0% 1% 2% 3%
total drift total drift
(a) Model RH30-NR2 (a) Model RH30-NR2 (b) Model IM15-NR2 (b) Model IM15-NR2
base shear (kN)
3000
Fig. 9. Adaptive and conventional pushover curves versus dynamic analysis envelopes.
2500
2000
1500
1000
dynamic
adaptive
500 uniform
triangular
0
0% 1% 2% 3%
total drift
May 7, 2004 14:47 WSPC/124-JEE 00151

12
dynamic
11 adaptive
since they are closely 10
related to the damage sustained uniform by a building. Therefore, it
triangular
is essential for static 9 methods to provide not only correct capacity curves, but also
8
drift values as close7as possible to the predictions of more rigorous dynamic analyses.
For this reason,6 the drift profiles obtained in each of the pushover and dynamic
5
analyses were examined 4
at the four levels of total displacement introduced earlier;
0.5%, 1.0%, 1.5% and 3 2.5%. It is noted that the drift profiles from the dynamic anal-
ysis represent in fact 2
1
envelopes of peak response at the specified total displacement
levels, not actual profiles
0.0%
at a given instant
0.5%
of time. In other 1.0%
words, they represent
interstorey drift
the drift maxima at each storey as obtained in the one time-history analysis (20
were carried out for 12 each pair of building/record scenarios) where their sum
12
dynamic
comes
dynamic
12 11 adaptive
dynamic
11 the closest adaptive
to each10of the aforementioned total
11
10
adaptive
displacement
uniform limits. uniform
triangular
uniform triangular
10
In Fig. triangular
10, the sequence
9 of the storey deformations
9
for one representative case
9 8
8 is shown. In general, 8
the predictions were reasonably
7
accurate in the pre-yield and
7
7
reasonably early inelastic
6 range for both the6 adaptive and conventional procedures
6 5
5
(Figs. 10(a) and 10(b)).5
Although there was 4
no clear “global winner”, the force-
4
4 based adaptive pushover 3 predictions were 3 found to be slightly more accurate in
3 2
2
the majority of cases, 2 in consistency with the
1
observations on the capacity curves,
1
1
0.0% 0.5% 1.0%
0% 1% 2%
0.0% 0.5% 1.0% interstorey drift interstorey drift
interstorey drift
12 (a) total drift 120.5% (b) total drift 1.0%

dynamic dynamic
12 11 adaptive
11 dynamic adaptive
11 adaptive uniform 12 uniform
10 10 dynamic
triangular
uniform triangular
10 11 adaptive
9 triangular 9
uniform
9 10 triangular
8 8
8 9
7 7
7 8
6 6
6 7
5 5
5 6
4 4
4 5
3 3
3 4
2 2
2 3
1 1
1 2
0.0% 0.5% 1.0% 0% 1% 2%
0% 1% 2% interstorey drift 1 interstorey drift
interstorey drift
0% 1% 2% 3% 4%
12
(a) total drift 0.5% (b)
(a)total
total drift
drift 1.0%
0.5%
interstorey drift (b) total drift 1.
(a) total drift 0.5% dynamic(b) total drift 1.0%
11 adaptive
uniform 12
10 12 dynamic
12 triangular
dynamic 11 dynamic adaptive
9 adaptive adaptive
11 11 uniform
uniform 10 uniform triangular
108 triangular
10 triangular
9
7
9 9
8
6
8 8
7
5
7 7
6
4
6 6
5
3
5 5
2 4
4 4
1 3
3 3
2
20% 1% 2% 2
interstorey drift 1
1 1
0% 1% 2% 3% 4%
0% 1% (a) total2%
drift 0.5% 3% 4% 0% 1% 2% (b)
3% total
4% drift
5% 1.0%
6% 7% 8% interstorey drift
interstorey drift interstorey drift
12
(c) total drift 1.5% (c) total drift 1.5% (d) total drift 122.5%
dynamic (d) total drift 2.5%
11 adaptive
12 dynamic
11
10 Fig. Interstorey
Fig. 10. Model RM30-NR3.
10. Model
dynamic RM30-NR3. Interstorey
uniform
adaptive drift profiles at triangular
drift profiles
11
different deformation
atlevels.
different deformation levels adaptive
uniform
9 uniform 10 triangular
10 triangular
8 9
9
8
7 8 6
6 7
7
5 6
6
4 5
5
3 4
4
2 3
3
1 2
2
0% 1% 2% 3% 4% 1
1
interstorey drift
0% 1% 2% 3% 4% 5% 6% 7% 8%
0% 1% 2% 3% 4% 5% 6% 7% 8% interstorey drift
interstorey drift
12 (c) total drift 1.5%

dynamic (d) total drift 2.
May 7, 2004 14:47 WSPC/124-JEE 00151
shown above. However, in the highly inelastic region, the FAP method failed to
provide correct response estimates, by exaggerating the inelastic deformations in the
locations of damage and underestimating them elsewhere (Figs. 10(c) and 10(d)).
Such erroneous behaviour seems to be related to the fact that at onset of damage
on a given storey, the resulting reduction in stiffness and consequent changes in
the modal shapes derived by eigenvalue analysis at the corresponding analysis step
induces a concentration at that storey of the increment of forces for the next step,
thus increasing even more its damage and hence resulting in a sort of “vicious
circle” that leads to exaggeratedly large deformation levels at that such damaged
locations.
It is also noteworthy that, whereas the non-adaptive pushover with uniform
distribution yielded equally flawed results in that range, its triangular counterpart,
despite its fixed loading pattern, managed to provide better drift estimates at high
deformation levels. This can be explained by the fact that, typically, the damage
occurred at the lower floor levels for all structural configurations and, contrary
to the uniform and the adaptive patterns, with the triangular distribution the
forces applied to the failed storeys were smaller than those applied to the upper
storeys, resulting in more reasonable drift predictions. However, and although the
above does indeed constitute an interesting observation, it should be stressed that
such good performance of the fixed triangular patterns comes as a result of the
location of damage in the considered case studies. Had damage occurred at the
upper floors, the triangular distribution would have failed to predict accurately
the drifts profiles. Such behaviour was indeed observed by Antoniou [2002], where
non-adaptive pushovers with triangular distributions failed to provide adequate
predictions for those structural models that featured a soft-storey located at mid-
height.
Similar observations were made also for the storey shear profiles obtained from
each of the pushover and dynamic analyses at the aforementioned four levels of
structural deformation (0.5%, 1.0%, 1.5% and 2.5% total drift). Although differ-
ences were not significant, FAP shear distribution predictions were found to be typ-
ically between the results from the uniform and triangular non-adaptive pushover
analyses, and usually closer to the dynamic envelopes, at least for total deformations
levels of up to 1.5%. Thereafter, it was conventional pushover with triangular force
distribution that presented the best predictions, for the reasons discussed above.
There have also been cases where the interstorey drift and shear distribution
predictions of the adaptive and the fixed distributions were totally inadequate, even
at the early stages of deformation. Typical examples of such flawed behaviour are
depicted in the drift profiles of Fig. 11 and shear distributions of Fig. 12, the latter
clearly confirming that for a full reproduction of dynamic response characteristics it
is required that static analysis methods feature the capability of reversing the sign
of applied storey forces (in Fig. 12(a), for instance, dynamic inertia forces applied
at storeys eight and nine are negative). This is a feature that, as highlighted in
Sec. 3.3, current force-based pushover schemes, adaptive or not, seem to be lacking.
May 7, 2004 14:47 WSPC/124-JEE 00151
1212 12
12 dynamic
dynamic
dynamicdynamic 88 8
8
dynamic
dynamic
dynamic
dynamic
1111 11 adaptive
adaptive
adaptive adaptive
adaptive
adaptive
11 adaptive adaptive
uniform
uniform
uniform uniform
uniform
uniform
uniform 77 7 uniform
1010 10
10 triangular
triangular
triangular 7 triangular
triangular
triangular
99 9
9 66 6
6
88 8
8
77 7 55 5
5
7
66 6
6 44 4
4
55 5
5
33 3
44 4
4
3
33 3
3 22 2
2
22 2
2
11 1
11 1
1
1
0.0% 0.0%
0.0% 0.5% 0.5%
0.5% 1.0% 1.0% 0.0%
1.0% 0.0% 0.0% 0.5% 0.5%
0.5% 1.0%
1.0%1.0%
0.0% 0.5% interstorey
interstorey
drift drift 1.0% 0.0% 0.5% interstorey
interstorey
drift 1.0%
interstorey drift
interstorey drift interstorey drift drift
interstorey drift
(a) Model
(a) (a)
Model RL15-NR2
Model
RL15-NR2
RL15-NR2 (b)
(b) (b)Model
Model
(b) Model
IH30-NR2
IH30-NR2
IH30-NR2
(a) Model
(a) RL15-NR2
Model RL15-NR2 (b) Model IH30-NR2
Model IH30-NR2
Fig.Fig.
Fig. 11.Interstorey
11.
Fig. 11.
Interstorey
11. Interstorey
driftdrift
drift
Interstorey profiles
profiles
profiles
drift at 0.5%
at
profiles0.5%
at 0.5%totaltotal
total driftdrift
drift for different
for for
different
different
model
model
model configurations
configurations
configurations
Fig. 11. Interstorey drift profiles atat0.5%
0.5% total
total drift
drift for
fordifferent
differentmodel
modelconfigurations
configurations.
12 12
12 12 dynamic
dynamic 8 8 dynamic
dynamic
dynamicdynamic 8 8 dynamic
dynamic
11 11 adaptive
adaptive adaptive
adaptive
11 11 adaptive
adaptive adaptive
adaptive
uniform
uniform uniform
uniform
10 10 uniformuniform 7 7 uniform
uniform
10 10 triangular
triangular 7 7 triangular
triangular
triangular
triangular
9 9
9 9 6 6
6 6
8 8
8 8
7 7 5 5
7 7 5 5
6 6
6 6 4 4
4 4
5 5
5 5
4 4 3 3
4 3 3
4
3 3 2 2
3 3 2 2
2 2
2 2 1 1
1 1 1 1
1 1
0 0 1000 1000 2000 2000 3000 3000 4000 4000 0 0 1000 1000 2000 2000 3000 3000
0 0 1000 1000 2000 2000 3000 3000 4000 4000 0 0 1000 1000 2000 2000 3000 3000
storey
storey
shearshear
(kN)(kN) storey
storey
shearshear
(kN)(kN)
storeystorey
shear shear
(kN) (kN) storey shearshear
storey (kN) (kN)
(a) (a)
Model
Model
RM15-NR2
RM15-NR2 (b) (b)
Model
Model
IL15-AR
IL15-AR
(a) (a) Model
(a)Model RM15-NR2
RM15-NR2
Model RL15-NR2 (b) Model
(b) Model
(b) Model IL15-AR
IL15-AR
IL15-AR
Fig.Fig.
12. 12.
Storey
Storey
shear
shear
profiles
profiles
at 0.5%
at
at 0.5%
totaltotal
driftdrift
for for
different
for different
model
model
configurations
configurations
Fig. Fig.
12. 12. Storey
Storey shearshear profiles
profiles at 0.5% 0.5%
totaltotal
driftdrift different
for different model
model configurations
configurations
Fig. 12. Storey shear profiles at 0.5% total drift for different model configurations.
It is interesting to note that the above deficiencies were not always reflected
on the capacity pushover curves, discussed previously, which, on occasions, were
seen to be correct even for those cases where storey drift and/or shear profiles were
incorrect. This highlights the importance of considering local parameters, such as
storey drift and/or storey shear profiles, when assessing the efficiency of pushover
methods, adaptive or not.
5. Conclusions
In this work, the accuracy of adaptive and non-adaptive force-based pushover meth-
ods in predicting the seismic response characteristics of reinforced concrete build-
ings has been explored. To this end, a fully adaptive pushover algorithm, which
explicitly accounts for the effects that progressive stiffness degradation and higher-
mode effects have on structural response, has been developed and employed in the 7 7
7 7
verification study. The latter comprised comparisons between the capacity curves
May 7, 2004 14:47 WSPC/124-JEE 00151
and drift/shear profiles obtained through static and dynamic nonlinear fibre-based
analysis on a number of real structural models of different configurations and duc-
tility classes, subjected to equally diverse input ground motions; each pushover was
compared with the IDA output obtained for each individual accelerogram, as op-
posed to the average of the latter, to avoid smoothening out of important structural
response peculiarities introduced by individual records.
A summary of the main observations and general conclusions of the present
study is presented below:
• Effective assessment of the accuracy of pushover algorithms, through comparison

with time-history analysis results, should not be based on the comparison of
capacity curves alone, but rather include an evaluation and comparison of local
response quantities, such as storey drift and storey shear profiles.
• Non-adaptive pushover analyses, with the triangular and uniform distributions,
do not always provide curves that constitute a lower and an upper bound to
the IDA response points. Moreover, they fail to reproduce accurately the local
dynamic response characteristics of buildings, particularly within the post-peak
range.
• Force-based adaptive pushover provides slightly improved capacity curves over its
conventional counterparts, with a closer fit to the IDA points. Although the differ-
ences were not always impressive, they were clear in the vast majority of the cases.
• Accordingly, up to global yield, FAP derives interstorey drift and storey shear
profiles closer to the dynamic envelopes. However, in the highly inelastic range
it tends to introduce excessive forces at the locations of plastic hinges and, sub-
sequently, to overestimate the drift values at such storeys.
• In some of the studied cases, none of the force-based static analysis methods em-
ployed in this study could correctly predict the dynamic response characteristics
of buildings, even at pre-yield deformation stages, due to the difficulties faced by
current force-based pushover algorithms in modelling a reversal of applied storey
force sign.
• Overall, it can be stated that force-based adaptive pushover, in its present state
of development, features a relatively minor advantage over its traditional non-
adaptive counterparts, all of which present limitations in the accurate prediction
of dynamic response characteristics of RC buildings.
The above clearly indicates that further research work is required before force-
based adaptive pushover can be considered as a valid alternative to nonlinear time-
history analysis. As highlighted in the body of the document, such developments
might include (i) the implementation of multiple response spectra to determine
the spectral amplification of different modes at varying deformation stages and/or
(ii) the employment of weighted vectorial addition, as opposed to SRSS and CQC
combination rules, in the computation of normalised scaling vector, as suggested
by Priestley [2003].
May 7, 2004 14:47 WSPC/124-JEE 00151
Acknowledgements
The authors would like to acknowledge the important contribution of Professor A.
S. Elnashai, with whom the current research endeavours were initiated at Imperial
College of London. The first author is also grateful to the financial support provided
by the European Commission at different stages of the current research project,
through the research networks ICONS (Innovative Concepts for Seismic Design of
New and Existing Structures) and SAFERR (Safety Assessment For Seismic Risk
Reduction), and also by means of a Marie Curie Fellowship at the ROSE School
(Pavia, Italy). Finally, the authors thank Professor Nigel Priestley for his valuable
suggestions during the final stages of this work.
References
Albanesi, T., Biondi, S. and Petrangeli, M. [2002] “Pushover analysis: An energy based
approach,” Proc. of the Twelfth European Conference on Earthquake Engineering,
London, United Kingdom, Paper No. 605.
Antoniou, S. [2001] “Pushover analysis for seismic assessment of RC structures,” Transfer
Report, Department of Civil Engineering, Imperial College London, United Kingdom.
Antoniou, S. [2002] “Advanced inelastic static analysis for seismic assessment of struc-
tures,” PhD Thesis, Imperial College London, University of London, United King-
dom.
ATC [1997] “NEHRP Guidelines for the seismic rehabilitation of buildings,” FEMA Re-
port No. 273, Federal Emergency Management Agency, Applied Technology Council,
Washington, D.C.
Bommer, J. and Martinez-Pereira, A. [1999] “The effective duration of earthquake strong
motion,” Journal of Earthquake Engineering 3(2), 127–172.
Bracci, J. M., Kunnath, S. K. and Reinhorn, A. M. [1997] “Seismic performance and
retrofit evaluation of reinforced concrete structures,” Journal of Structural Engineer-
ing 123(1), 3–10.
Campos-Costa, A. and Pinto, A. V. [1999] “European seismic hazard scenarios — An
approach to the definition of input motion for testing and reliability assessment of
Civil Engineering structures,” JRC Special Publication No. X.99.XX, Joint Research
Centre, Ispra, Italy.
CEN [1995] European Prestandard ENV 1998-1-1: Eurocode 8 — Earthquake resistant
design of structures, Part 1: General rules and rules for buildings. Comite Europeen
de Normalisation, Brussels, Belgium.
Chopra, A. K. and Goel, R. K. [2002] “A modal pushover analysis procedure for estimating
seismic demands for buildings,” Earthquake Engineering and Structural Dynamics 31,
561–582.
Elnashai, A. S. [2001] “Advanced inelastic static (pushover) analysis for earthquake ap-
plications,” Structural Engineering and Mechanics 12(1), 51–69.
Fardis, M. N. [1994] “Analysis and design of reinforced concrete buildings according to
Eurocodes 2 and 8. Configurations 3,5 and 6,” Reports on Prenormative Research in
Support of Eurocode 8.
Freeman, S. A., Nicoletti, J. P. and Tyrell, J. V. [1975] “Evaluations of existing build-
ings for seismic risk — A case study of Puget Sound Naval Shipyard, Bremerton,
Washington,” Proc. of the First U.S. National Conference on Earthquake Engineer-
ing, Oakland, California, pp. 113–122.
May 7, 2004 14:47 WSPC/124-JEE 00151
Gupta, B. and Kunnath, S. K. [2000] “Adaptive spectra-based pushover procedure for

seismic evaluation of structures,” Earthquake Spectra 16(2), 367–391.
Hamburger, R. O., Foutch, D. A. and Cornell, C. A. [2000] “Performance basis of guidelines
for evaluation, upgrade and design of moment-resisting steel frames,” Proc. of the
Twelfth World Conference on Earthquake Engineering, Auckland, New Zealand, Paper
No. 2543.
Hughes, T. J. R. [1987] The Finite Element Method, Linear Static and Dynamic Finite
Element Analysis (Prentice-Hall, Inc.).
Kim, S. and D’Amore, E. [1999] “Push-over analysis procedure in earthquake engineering,”
Earthquake Spectra 15(3), 417–434.
Krawinkler, H. and Seneviratna, G. D. P. K. [1998] “Pros and cons of a pushover analysis
of seismic performance evaluation,” Engineering Structures 20(4–6), 452–464.
Lawson, R. S., Vance, V. and Krawinkler, H. [1994] “Nonlinear static push-over analysis
— Why, when, and how?” Proc. of the Fifth US National Conference on Earthquake
Engineering, Earthquake Engineering Research Inst., Oakland, California, 1, 283–292.
Lefort, T. [2000] “Advanced pushover analysis of RC multi-storey buildings,” MSc Disser-
tation, Department of Civil and Environmental Engineering, Imperial College London,
United Kingdom.
Moghadam, A. S. and Tso, W. K. [2002] “A pushover procedure for tall buildings,” Proc. of
the Twelfth European Conference on Earthquake Engineering, London, United King-
dom, Paper No. 395.
Mwafy, A. M. and Elnashai, S. A. [2000] “Static pushover versus dynamic-to-collapse
analysis of RC buildings,” ESEE Research Report No. 00/1, Department of Civil and
Environmental Engineering, Imperial College London, United Kingdom.
Mwafy, A. M. [2001] “Seismic performance of code-designed RC buildings,” PhD Thesis,
Imperial College London, University of London, United Kingdom.
Naeim, F. and Lobo, R. [1999] “Avoiding common pitfalls in pushover analysis,” Proc. of
the Eighth Canadian Conference on Earthquake Engineering, Canadian Association
for Earthquake Engineering, Vancouver, British Columbia, Canada, pp. 269–274.
Papanikolaou, V. [2000] “Development and verification of adaptive pushover analysis pro-
cedures,” MSc Dissertation, Department of Civil and Environmental Engineering,
Imperial College London, United Kingdom.
Paret, T. F., Sasaki, K. K., Eilbeck, D. H. and Freeman, S. A. [1996] “Approximate
inelastic procedures to identify failure mechanisms from higher mode effects,” Proc.
of the Eleventh World Conference on Earthquake Engineering, Disc 2, Paper No. 966.
Priestley, M. J. N. [2003] Myths and Fallacies in Earthquake Engineering, Revisited (IUSS
Press, Pavia, Italy).
Requena, M. and Ayala, G. [2000] “Evaluation of a simplified method for the determi-
nation of the nonlinear seismic response of RC frames,” Proc. of the Twelfth World
Conference on Earthquake Engineering, Auckland, New Zealand, Paper No. 2109.
Rovithakis, A. [2001] “Verification of adaptive pushover analysis procedures,” MSc Disser-
tation, Department of Civil and Environmental Engineering, Imperial College London,
United Kingdom.
Sasaki, K. K., Freeman, S. A. and Paret, T. F. [1998] “Multi-mode pushover procedure
(MMP) — A method to identify the effects of higher modes in a pushover analysis,”
Proc. of the Sixth US National Conference on Earthquake Engineering, Earthquake
Engineering Research Inst., Oakland, California, 12 pp.
SeismoSoft [2004] SeismoStruct — A Computer Program for Static and Dynamic Non-
linear Analysis of Framed Structures [online]. Available from URL: http://www.
seismosoft.com.
May 7, 2004 14:47 WSPC/124-JEE 00151
Valles, R. E., Reinhorn, A. M., Kunnath, S. K., Li, C. and Madan, A. [1996] IDARC2D
Version 4.0: A Computer Program for the Inelastic Analysis of Buildings, Technical
Report NCEER 96-0010, State University of New York at Buffalo, Buffalo, NY.
Vamvatsikos, D. and Cornell, C. A. [2002] “Incremental dynamic analysis,” Earthquake

Engineering & Structural Dynamics 31(3), 491–514.

Advantages and Limitations of Adaptive and Non-Adaptive Force-Based Pushover Procedures

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Advantages and Limitations of Adaptive and Non-Adaptive Force-Based Pushover Procedures

Caricato da

Copyright:

Formati disponibili

May 7, 2004 14:47 WSPC/124-JEE 00151

Journal of Earthquake Engineering, Vol. 8, No. 4 (2004) 497–522

ADVANTAGES AND LIMITATIONS OF ADAPTIVE AND

Received 25 November 2002

Keywords: Force-based adaptive pushover; incremental dynamic analysis; FAP.

498 S. Antoniou & R. Pinho

that cannot be obtained from an elastic static or dynamic analysis. Conventional

the accuracy of force-based adaptive pushover methods in predicting the horizontal

2. Nonlinear Static Pushover Analysis in Earthquake Engineering

Force-based Pushover Procedures 499

21 “configurations. Contrary” with “configurations and, contrary”

(a) (a) (b)

500 S. Antoniou & R. Pinho

2.1. Pushover procedures considering higher-mode effects

Force-based Pushover Procedures 501

following subsection seem to constitute a better alternative, due to their ability

2.2. Adaptive pushover procedures

502 S. Antoniou & R. Pinho

Force-based Pushover Procedures 503

3. The Force-Based Adaptive Pushover Algorithm (FAP)

current presentation has been implemented in SeismoStruct [SeismoSoft, 2004], a

procedure, as described in the following subsections.

3.1. Nominal load vector and inertia mass

504 S. Antoniou & R. Pinho

3.2. Computation of load factor

Force-based Pushover Procedures 505

3.3. Calculation of normalised scaling vector

Fij = Γj φij Mi , (2a)

Previous research (a)

base shear (kN)

base shear (kN)

PPt t == [[λλt t]] ×× ×× =

506 S. Antoniou & R. Pinho

Force-based Pushover Procedures 507

3.4. Update of loading force vector

508 S. Antoniou & R. Pinho

3.4.1. Total Updating

normalised = λt · F̄t · P0 .new increment

Force-based Pushover Procedures 509

3.4.2. Incremental Updating

normalised nominal new increment

Table 1. Definition of the structural systems considered.

Structure Structural No. of Ductility Design Behaviour Tuncracked

RF-H030 / RH30 High 5.00 0.697

510 S. Antoniou & R. Pinho

assessing the effectiveness of the proposed force-based adaptive pushover algorithm

configurations subjected to equally diverse input ground motions. Extensive pre-

Table 1. Definition of the structural systems considered.

Structure Structural No. of Ductility Design Behaviour Tuncracked

Force-based Pushover Procedures 511

4.5+7 × 3.0 = 25.5 m

REGULAR SYSTEM IRREGULAR SYSTEM DUAL SYSTEM

Fig. 6. Geometric characteristics of the regular, irregular and dual systems.

512 S. Antoniou & R. Pinho

Table 2. Characteristics of the records employed in this study.

Record Ground Response 5% AI 95% AI Total Significant teff /ttotal

1.0 1.0 1.0 1.0

0.8 0.8 0.8 0.8

0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0

0.4 0.4 2.4 2.4

0.3 0.3 1.8 1.8