Beruflich Dokumente
Kultur Dokumente
To cite this Article Papanikolaou, Vassilis K. and Elnashai, Amr S.(2005) 'EVALUATION OF CONVENTIONAL AND
ADAPTIVE PUSHOVER ANALYSIS I: METHODOLOGY', Journal of Earthquake Engineering, 9: 6, 923 941
To link to this Article: DOI: 10.1080/13632460509350572
URL: http://dx.doi.org/10.1080/13632460509350572
This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
Journal of Earthquake Engineering, Vol. 9, No. 6 (2005) 923-941
@ Imperial College Press @ Imperial College k
wWw.lcpress.co.uk
VASSILIS K. PAPANIKOLAOU
Labomtory of Reinfonced Concrete and Masonry Structures
Civil Engineering Department, Aristotle University of Thessdoniki
P. 0.Box 482, Thessalaniki, 54124, Greece
A M R S. ELNASHAI
Willet Professor, Director.Mid- America Earthquake Center
University of Illinois, 205 North Mathews, Urbana, Illinois, 61801, USA
Downloaded By: [University of Illinois] At: 18:56 8 July 2010
In this paper, a methodology is suggested and tested for evaluating the relative perfor-
mance of conventional and adaptive pushover methods for seismic response assessment.
The basis of the evduation procedure is a quantitative measure for the difference in
response between these methods and inelastic dynamic analysis which is deemed to be
the most accurate. Various structural leveb of evaluation and different incremental r e p
resentations for dynamic analysis are also suggested. This method is applied on a set of
eight different reinforced concrete structural systems subjected t o various strong motion
records. Sample results are presented and discussed while the full results are presented
alongside concIusions and recommendations, in a companion paper.
1. Introduction
Until recently, seismic assessment and design has relied on linear or equivalent lin-
ear (with reduced stiffness) analysis. In this approach, simple models are used for
various components of the structure, which are subjected to seismic forces evalu-
ated from elastic or design spectra, and reduced by force reduction (or behaviour)
factors. The ensuing displacements are amplified to account for the reduction of
applied forces. This procedure, though simple and easy to apply in the design office
environment, fails to fit within the principle of failure mode control which is part
of performance-based assessment and design [Elnashai, 20001.This in turn has led
to an increase in the use of inelastic analysis as a more realistic means of assessing
the deformationd state in structures subjected to strong p o u n d motion.
Static and dynamic analyses use the same material constitutive relationships
with the exception being static monotonic analysis that does not require unloading
and reloading models. Both use principles of equilibrium and compatibility with
the difference.being that dynamic equilibrium of forces includes damping and iner-
tial effects. Both make use of iterative procedures to arrive at convergent solutions.
Finally, whereas in static inelastic analysis the variable is the current level of dis-
placement or force, in dynamic analysis, it is time. Therefore, the four features
that may d e c t the difference in the level of complexity and computing resource
requirements are:
Static monotonic analysis requires only monotonic-constitutive models.
Dynamic analysis requires treatment of structural damping and' mass
distribution.
Static analysis to collapse is repeated as many times as the deformation causing
collapse divided by the displacement increment necessary for convergence; this is
likely to be in the tens of steps.
Dynamic analysis is a static analysis repeated as many times as the duration
Downloaded By: [University of Illinois] At: 18:56 8 July 2010
of the earthquake divided by the time step for response history analysis; this is
likely to be in the thousands of time increments.
The outcome of the above discussion leads to the conclusion that static analysis
requires simpler models, representation of stiffness and strength only, and a fraction
of the number of analyses, compared to dynamic analysis. This is the underlying
reason for the increased use of pushover analysis in the industry, and the inclusion of
static inelastic methods in assessment guidance notes (e.g. FEMA 2731274 [FEMA,
19971 and sequels) as well as modern design codes (e.g. the new draft of Eurocode 8).
With the proliferation in the use of static inelastic (pushover) methods for seis-
mic assessment and design comes controversy. Many researchers have contributed
developments to enhance the performance of the pushover technique [Freemanet al.,
1975; Bracci et al., 1997; Chopra and Goel, 2002 among others]. They, including
the writers of this paper, advocate the use of pushover in its various forms in lieu
of dynamic analysis. On the other hand, resistance to the use of pushover analysis
comes from two opposing ends of the complexity spectrum. Advocates of simplicity
legitimately state that the single load distribution pushover fails to capture the
actual behaviour, and that more advanced versions are too complex for practical
applications. Others state, also legitimately, that inelastic dynamic effects cannot be
captured.by any static method. Hence, full dynamic analysis is necessary. Previous
research, with very few exceptions, has focused on developing pushover techniques
without ~ssessingtheir performance comprehensively. This would require applics
tion of static pushover to a wide range of structures ranging from low to high rise,
regular to highly irregular, subjected to a large number of earthquake records cov- ,
ering a wide, range of magnitudes, distance, site conditions and source mechanism,
monitored on various structural levels. Towards this challenge, the objective of this
paper is to first suggest a general methodology for evaluation of pushover methods,
including conventional and adaptive as compared to inelastic response history anal-
ysis. Then, apply this methodology to a series of different structural systems under
various strong motion records. The above are deployed to answer, at least in part,
Evaluation of Conventional and Adaptive Pushwet Analysis I: Methodology 925
1994; Faella, 1996; Krawinkler and Seneviratna, 19981. Pushover analysis per se is
not a recent development. However, this study is concerned with multi-degree of
freedom inelastic analysis of complex structures, which is relatively recent.
Conventional pushover analysis is the nonlinear incremental-iterative solution
of the equilibrium equation K U = P in a finite element formulation, where K is
the nonlinear stiffness matrix, U is the displacement vector and P is a predefined
load vector applied laterally over the height of the structure in relatively small load
increments (Fig. 1). This lateral load can be a set of forces or displacements that
have a necessarily constant ratio throughout the analysis (fixed pattern). At the
end of each iteration, the reaction vector (P,) of the structure is assembled from
all finite element contributions. The out of balance forces are iteratively reapplied
until convergence to a specified tolerance is reached [Bathe, 1982):
where
B is the strain displacement matrix of each element
CTNLis the element nonlinear stress vector as determined by its material
constitutive law
Downloaded By: [University of Illinois] At: 18:56 8 July 2010
too simplistic, particularly for nonductile structures that exhibit severely pinched
hysteretic behaviour.
Static pushover analysis neglects dynamic effects and therefore, during strong
mot ion, nonlinear structural behaviour can be described by balancing the
dynamic equilibrium at every time step. As pushover analysis focuses only on
the strain energy of the structure during monotonic loading, it neglects other
sources of energy mainly associated with dynamic components of forces such as
kinetic and viscous damping energy.
The conventional pushover analysis procedure does not account for the progres-
sive changes in the modal properties during nonlinear yielding and cracking in
the structure which leads to period elongation and different spectral amplifica-
tions. This is due to the constant lateral load pattern used, which ignores the
potential redistribution of inertia forces and higher mode effects, as yielding and
cracking governs the inelastic structural behaviour.
It is clear that conventional pushover analysis, due to its static nature, lacks
many features of its dynamic counterpart that may be critical in certain cases.
However, it provides the engineer with an efficient alternative, not only to expensive
inelastic dynamic malysisi but also to standard seismic code practice. Consequently,
some possible developments to enhance this are suggested below:
response is then calculated by combining the individual peak responses using the
SRSS rule. The aforementioned methods reported varying degrees of success but
still did not account for damage accumulation and resulting modification of the .
modal parameters. This could significantly affect the structural response.
Therefore, adaptive methods accounting for the variation of the modal proper-
ties during the inelastic process are considered to be a promising alternative. Bracci
et al. [I9971and Lefort [2000]used the inelastic storey forces of the previously equi-
librated load step to update the lateral load pattern. Storey force distribution was
obtained by either adding an increment of force to the equilibrated force vector
from the previous step or by a new set of forces accounting for the current state of
resistance distribution. Gupta and Kunnath [2000] proposed a constantly updated
load pattern depending on the instantaneous dynamic characteristics of the struc-
ture and a site specific spectrum. Before each load increment, eigenvalue analysis
was carried out and the storey forces for each mode were calculated by multiplying
the storey weight with the modal participation factor, mode shape and spectral
amplification. A static analysis was then performed for each mode of interest and
calculated responses were first combined with SRSS and then added to the previous
load step. An alternative approach was suggested by Albanesi et al. [2002], where
the loading took into account both the inertial properties and the kinetic energy of
the structure generated under strong motion.
Work undertaken by the writers and their cwworkers has developed a robust
procedure for adaptive pushover analysis that is shown to be superior to, or at worst,
as good as conventional pushover. Formulations given by Papanikolaou (20001,pub- .
(KT).
These ingredients can form the new lateral load pattern a s described below
(Fig. 3). The modal load for mode i, applied on the j degree of heedom is defined:
Q'.M.~
ui is the modal participation factor of mode i, vi = @~T,M.@'
After defining the lateral load profiles for all different modes, a modal combina-
tion (ABS/SRSS/CQC) produces the updated load vector. Before the application
of the updated lateral load vector, normalisation and multiplication by the current
loading level is performed so 'that the magnitude of the pushover load is still applied
incrementally, as done in the conventional pushover approach.
It is noteworthy that additional complexity required to perform adaptive
pushover analysis is considerable, in terms of accessing a robust eigenvalue solver,
efficiently updating the applied force vector, and switching if necessary to a fixed-
distribution displacement control past the peak point on the load-displacement
curve. However, onus of these complications is on the software, not the user. The
software package used throughout this study fuliils all the above requirements and
the only complexity left to the user is the reasonable and sufficient mass
. represen-
.
tation of the structure.
2001; Varnvatsikos and Cornell, 20021. According to this approach, the structural
model under consideration is excited by the same strong motion input, scaled to
different PGA values. For every scaling factor, the absolute maximum dynamic
response parameters (load displacement, moment curvature etc.) are plotted on a
two dimensional chart, similarly to pushover analysis (Fig. 4). The difference is
that each point now represents a Full-run inelastic dynamic analysis, while each
point on the static pushover curve is the response at one load step. The obvious
drawback of this procedure is the extremely high cost in editing time, and computer
resources. For this reason, it was implemented in the used software package to run
automatically without any user intervention.
The selection of the dynamic response absolute maxima is an issue which may
require further discussion regarding the fact that they may occur in different time
instants. Nevertheless, it is assumed that this approach reflects the corresponding
monotonic pushover nature where response maxima occur simultaneously. Another
absolute
maxima
/
I
Xmax
40.
. ** 0 0 s-
35 - * r a m
-
30 -
25 -
Pushover
20
IDAMax-Max
15 -
A IDA Max-Corresponding
10 -
x IDA Corresponding-Max
5
advantage i s t h a t the use of absolute maxima sets a safety upper bound in the
structural response.
For reasons of completeness, an alternative approach capturing the correspond-
ing values Y,,,, X,,, of both response maxima X, ,Y respectively, within the
same time instant has also been implemented (Fig. 5). As a result, two more sets
of IDA points are introduced (Fig. 6).
d , = a b s (Ypi-YDi)/YDi,
i = 1 t 0 4 (D5 ignored). (4)
Averaging all difference values suggests that equal weight has been assigned to
each IDA point. This is considered realistic when the scaling step of the record
remains constant throughout the analysis, which is analogous to the constant load,
increment used in pushover methods. The formula (Ypi- YDi)/YDi was selected
because the reference value (denominator YDi)is the dynamic response (in other
words comparing static to dynamic, not the opposite). For easy and fast application,
the CCDF method was implemented as & additional tool in the used software
package. An application of the CCDF method is shown in Fig. 8.
Evalwtion of Conventaond and Adaptive Pwhover Analysis I: MeUIodology 933
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.4 0.5 Q.55
Global drift (m)
6. Levels of Evaluation
Apart from the usual practice of monitoring base shear versus global drift, storey
shear versus interstorey drift and element moment versus section curvature are
also suggested for the evaluation of conventional and adaptive pushover methods
in order to capture possible hidden features of the inelastic response that may not
be explicitly reflected on the global scale.
Global level: The h s t monitored quantity was the base shear (V) versus top
displacement (d) (Fig. 9). Horizontal forces (V,) of the support nodes were added
and plotted against the horizontal displacement of the top floor. In dynamic anal-
ysis though, the support displacement is subtracted from the top displacement
in order to establish the global drift of the structure.
Storey level: The second monitor is the storey shear (VS) versus intentorey
drift ( d 3 ) (Fig. 10). The storey force (v)
is derived by adding all individual
element shear forces (V,S) which are equal to (M' - ~ ~ ) where/ t , M' and
are the top and bottom element moments and t is the storey height.
Section level: The ha1 monitor used in this study accounts for the local
behaviour (section level) during inelastic static and dynamic analysis through
plotting the element moment (M-)Venus curvature ( K ) (Fig. 11). The latter is
derived from (et - ~ ~ )where / h Q and E* are the top and bottom layer strains
respectively and h is the height of the section.
Downloaded By: [University of Illinois] At: 18:56 8 July 2010
c! If.
Fig. 11. Monitoring element moment versus section curvatuie.
Evaluation of Conventional and Adaptive Pwhover Analysis I: Methodoiugy 935
7. Analysis Setup
The components of a large scale analytical study targeting the evaluation of con-
ventional and adaptive pushover procedures are presented in subsequent sections.
These include analysis software and methods, structural models under considera-
tion and strong motion input. The analysis results are presented in a companion
paper.
regularity, these structures were selected for their long na&ral periods and the
possibility to attract higher mode effects d k n g the inelastic process.
Shear wall structures: Two models designed with high (H) and low (L) ductil-
ity recommendat ions with a design acceleration of 0.30 g and 0.15 g , respectively
(Fig. 13).These structures were selected because they exhibited a cantilever (sin-
gle degree of freedom) instead of frame behaviour and hence, be dominated by
the fundamental mode.
Irregular structures (in elevation): Two models designed with high (H) and
low (L) ductility recommendations with a design acceleration of 0.30g and 0.15 g,
respectively (Fig. 14). These structures were selected for their irregularity in
elevation and the possibility to produce soft storey mechanisms and attract higher
mode effects.
ICONS frame: The following model was based on the full scale structure
(ICONS)tested in Ispra, Italy on 1999 (Fig. 15). It was selected for its strong
irregularity in plan; representing design and construction practice in many South-
ern European and Mediterranean countries in the 50s and 60s [Pinho and
Elnashai, 2001].
The SPEAR frame: The final model is a 3D frame based on a full scale struc-
ture tested in 2003 within the European network on Seismic Performance Assess-
ment and Rehabilitation (SPEAR). It features irregularities both in plan and
elevation (Fig. 16). Similar to the ICONS Erames, this structure was designed
according to outdated design codes and with no seismic design provisions. It
will be constructed from weak concrete and smooth bars. It has heavily imbal-
anced stifihess in two orthogonal directions as well as large eccentricity in plan
Euduatim of Conventional and Adaptive Pwhouer Analysis I: Methodology 937
S:==ys : 8
with a shear wall corc. Total height : 24.00 m
Typical storey height : 3.00 m
Smctural ductility : High / Low Total length : 15.00 rn
Design acceleration : 0.30s/ 0.15s
Concrete strength : f,= 30 MPa Number of nodes : 238
Downloaded By: [University of Illinois] At: 18:56 8 July 2010
Steel yield strength : fy = 585 MPa Number of elements : 248 (conv. pushover)
Actual structure
=,-
FE modelling and mass distribution
- ----- . -- -- -- - - ----, I :
-
7.3. Strong motion chamcteristics
*
Four different strong motion iecords wereimposed on the above eight structural
systems, selected in terms of origin and frequency content: Two artificial records,
one representing the European earthquake of a 975 year return period (E975)
938 V. K. Papanikolaou tY A. S. Elhashai
Storeys : 4
Total height : 10.80 m
2D reinforced concrete frame Typical storey height : 2.70 m
with irregularity in plan. Total length : 12.50 rn
Storeys : 3
Total height : 8.75 m
Typical storey height : 3.00 m (first : 2.75 m)
Total length (x) : 9.00 m
3D reinforced concrete frame Total length Q : 10.375 m
with irregularity in plan and elevation.
Concrete strength : f,= 20 MPa
Number of nodes : 366 Longitudinal steel strength : fy= 400 MPa
Number of elements : 384 (conv. pushover) Transverse steel strength : fy= 220 MPa
and another derived from the Eurocode 8 design code response spectrum (EC8), and
two natural records, one with liow (Emeryville, Loma Priet a earthquake, 1989) and
another with high (Sibnta Monica, Northridge earthquake, 1994) frequency content,
providing high ampliifications in the long and short period range respectively. All
four records were normalised to a PGA of 0.3g in order to be comparable during
Evaluation of Conventional and Adaptive Pmhover Andysis I: Methodology 939
-03J 15 sec o 1 2 3 4
Downloaded By: [University of Illinois] At: 18:56 8 July 2010
Fig. 17. Scaled accelerograrns and elastic response spectra of the four records (E9.75,EC8,
Emeryvilie, St. Monica).
the variable scaling of the incremental dynamic analysis approach. Twenty-five runs
for each record were performed with a scaling factor from 0.2 to 5.0 (up to 1.5 g
of PGA) with a step of 0.2. The accelerograrns and elastic response spectra are
depicted in Fig. 17.
8. Conclusions
A methodology for evaluating the performance of conventional and adaptive static
pushover methods is presented in this paper. The cornparat ive assessment is based
on the comparison of the structural response between the two pushover methods and
incremental represent at ion of inelastic dynamic analysis (IDA), using a simple and
efficient quantitative measure, the Capacity Curve Discrepancy Factor (CCDF).
The structural response is monitored on several levels, global, storey, and section.
All of the above techniques are implemented in the software package Zeus-hi and
applied to a number of different structural systems under the effect of various strong
motion acceleration records. The analysis results are presented and discussed in a
companion paper. It is believed that the proposed genera methodology provides
the researcher with a versatile tool for evaluating the performance of existing and
new advanced static pushover analysis techniques.
Acknowledgement - - -
The work described in the paper was funded by the Mid-America Earthquake Center
at the University of Illinois at Urbana-Champaign.
References
Albanesi, T., Biondi, S. and Petrangeli, M. [20021 "Pushover analysis: An energy based
approach," P d i n g s of the Twelfth eumpean C o n f m c e on Earthquake Engineer-
ing, London, United Kingdom, Paper No. 605.
Antoniou, S. (20031 Pushover Analysis for Seismic Design and Assessment of RC Struc-
tures, PhD Thesis, Engineering Seismology and Earthquake Engineering Section,
Imperial College, London, UK.
Bathe, K. J. [I9821 Finite Element Pmcedures in Engineering Andysis, Prentice Hall.
Bracci, J. M., Kunnath, S. K. and Reinhorn, A. .M. [I9971 "Seismic performance and
retrofit evaluation of RC structures, ASCE, ST Division 123(1), 3-10.
Chopra, A. K. and Goel, R K. 120021 "A modal pushover analysis procedure for estimating
seismic demands for buildings," Earthquake Engineering and Stmcturd Dynamics
3 1(3), 561-582.
Elnashai; A. S. [2000] "Advancedinelastic static (pushover) analysis for seismic design and .
assessment," G.Penelis International Synposiurn on Concrete and Masonry Struc-
tu~es,Aristotle University of Thessaloniki, Greece.
Downloaded By: [University of Illinois] At: 18:56 8 July 2010
Elnashai, A. S. [ZOO21 "Do we really need inelastic dynamic analysis?" Journal of Earth-
p u k e Engineering 6 (Special Issue I), 123-130.
Elnashai, A. S., Papanikolaou, V. and Lee, D. H. [2002-20051 Zeus-NL - A Pmgram for
Inelustac Dynamec Analysis of Stmctunes - User Manual, Mid-America Earthquake
Center, University of Illinois at U rbana- Champaign.
Elnashai, A. S., Pinho, R. and Antoniou, S. [2000] INDYAS - A Pmgmm for Inelustic
Dynamic Analysis of Structures, ESEE Research Report, Imperial College, London. '
Fajfar, P. and Fischinger, M. [I9881 "N2 - Method for nonlinear seismic analysis of regular
structures," Proceedings of the Ninth World Conference on Earthquake Engineering
(TokyeKyoto, Japan, 1988) 5, 111-116.
Fardis, M. N. (19941 Analysis and Design of Reinfomed Concrete Buildings According to
Evmwdes 2 and 8, Configurations 3, 5 and 6,Reports on Prenormative Research in
Support of Eurocode 8.
FEMA [I9971 NEHRP Guidelines for the Seismic Rehabilitation of Buildings Developed by
the Building Seismic Safety Council for the Fedeml Emergency Management Agency,
F E M A Report, No. 273, Washington, D.C.
Reeman, S. A., Nicoletti, J. P. and Tyrell, J. V. 119751 "Evaluation of existing building for
seismic risk -A case study of Puget Sound Naval Shipyard Bremerton, Washington,"
Proceedings of the United States National Confeence on Earthquake Engineering,
Berkeley, pp. 113-122.
Gulkan, P. and Sozen, M. A. [I9741 "Inelastic response of reinforced concrete structures
to earthquake'rnotions," ACI J o u m d 71, 604410.
Gupta, B. and Kunnath, S. K. [2000] UAdaptive spectra-based pushover procedure for
seismic evaluation of structures," Earthquake S p e c h l6(2), 367-39 1.
Izzudin, B. A. and Ehashai, A. S. [I9891ADAPTIC - A P m p m for Adaptive Large Dis-
placement Elaptoplai@ac Dynamic Analysis of Steel, C o n m t e and Composite h m e s ,
ESEE Research Report, No. 89-7, Imperial College, London.
Krawinkler, H.and Seneviratna, G. D. P. K. [I9981 "Pros and cons of a pushover analysis
of seismic performance evaluation," "Engineering Structures," 20 4-6, 452-464.
L a w n , R. S., Vance, V. and Krawinkler, H. [1996]Nonlinear static pushover analysis:
* .
Why, how, when and when not to use it, LLPmceedtngsof the 65th Annual Convention
of S t r u c t u d Engineers Associotaorr of Cdifomiu," Structural Engineers Association
of California (Sacramento, 1996), pp. 17-36.
Evaluation of Conventional and Adaptive Pushover Analysis I: Methodology 941