Professional Documents
Culture Documents
net/publication/259221590
CITATIONS
17
4 authors, including:
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Simplifying and improving design criteria for concentrically braced frames View project
All content following this page was uploaded by Pier Paolo Rossi on 15 June 2015.
ASYMMETRIC BUILDINGS
tel. +39-095-7382274
fax +39-095-7382249
email: emarino@dica.unict.it
Abstract
Traditional nonlinear static methods, e.g. the original version of the N2 method implemented in
Eurocode 8, are not always effective in the assessment of asymmetric structures because of the er-
rors committed in the evaluation of the torsional response. To overcome this shortcoming, two
methods have recently been suggested by Kreslin and Fajfar (2012) and Bosco et al. (2012). In par-
ticular, the method proposed by Kreslin and Fajfar adjusts the results of the nonlinear static analysis
by means of those of a standard modal response spectrum analysis. In the method proposed by
Bosco et al., the researchers suggested the use of two nonlinear static analyses, characterized by
In this paper, the two improved nonlinear static methods and the original N2 method are applied to
predict the maximum dynamic response of single- and multi-storey systems subjected to artificial
and recorded accelerograms. The results highlight that the improved nonlinear static methods pro-
vide estimates which are more accurate than those of the original N2 method. Further, the compari-
son of the results identifies the range of the structural properties within which the original N2 meth-
od is still reliable and the range within which one of the two improved methods should be preferred.
2
1 INTRODUCTION
It is widely recognised in the scientific community that a proper seismic assessment of existing
structures requires the comparison between the displacement capacity, i.e. the displacement corre-
sponding to the achievement of a given limit state, and the displacement demand of the structure,
i.e. the displacement caused by the earthquake ground motion. Nonlinear static methods developed
in the recent past (Antoniou and Pinho 2004; Chopra and Goel 2002; Fajfar 1999; Freeman 1998)
constitute a valid response to this need. In fact, they represent a powerful tool for the evaluation of
the inelastic response of structures and do not require the computational effort of nonlinear time-
history analyses. Most seismic codes allow the use of these methods. In particular, Eurocode 8
(CEN 2004) has adopted the N2 method, hereinafter called original N2 method, developed by Fajfar
and his research team (Fajfar and Fishinger 1988; Fajfar 1999; Fajfar and Gaspersic 1996).
The original N2 method considers that the structure vibrates predominantly in a single mode. For
this reason, it is suitable for predicting the response of regular planar frames, particularly if these
structures are low-rise. The response of asymmetric structures, instead, always sees important con-
tributions from more than one mode of vibration and is characterised by the simultaneous transla-
tion and rotation of the decks. While the in-plan distribution of the actual maximum displacements
of asymmetric structures is nonlinear (because these displacements are reached at different times),
the distribution provided by a single nonlinear static analysis is linear. Therefore, the single nonlin-
ear static analysis is not able to predict properly the in-plan distribution of the maximum displace-
ments of asymmetric buildings (Bosco et al. 2012; Chopra and Goel 2004). This observation partic-
ularly applies to torsionally flexible systems where the nonlinearity of the in-plan distribution of the
To make the nonlinear static methods reliable for the estimation of the displacement demand of
plan-asymmetric structures, several researchers have proposed improvements to the standard proce-
dure (Bosco et al. 2012; Bento et al. 2010; Chopra and Goel 2004; Fajfar et al. 2005; Kreslin and
Fajfar 2012; Fujii 2011). These new procedures are called hereinafter improved nonlinear static
3
methods. Among those recently proposed, two improved nonlinear static methods are analysed here:
the extended N2 method (Fajfar et al. 2005; Kreslin and Fajfar 2012) and the corrective eccentricity
method (Bosco et al. 2012). Both these two nonlinear static methods require a computational effort
much lower than that required by nonlinear time-history analysis and do not need modelling of the
The extended N2 method is based on the assumption that the results of the elastic analysis of an
asymmetric structure, properly normalised, represent the upper bound of the torsional amplification
of the inelastic response. Hence, the inelastic seismic response of asymmetric structures can be con-
servatively predicted by adjusting the results of the original N2 method by means of correction fac-
tors derived by a standard (elastic) modal response spectrum analysis. These correction factors can-
not be lower than unity because they consider that any favourable torsional effect arising from the
elastic analysis is reduced in the inelastic range of behaviour. It should be noted that the final ver-
sion of the method (Kreslin and Fajfar 2012) requires one nonlinear static analysis (pushover) for
each direction of the seismic action and considers the effects of higher modes in both plan and ele-
According to the corrective eccentricity method (Bosco et al. 2012), two nonlinear static analyses
have to be performed for each direction of the seismic action. The lateral forces are applied to two
points of the deck which are generally different from the centre of mass (CM). The distances be-
tween the points where the lateral force is applied and CM are named corrective eccentricities (ei)
and are calculated by means of analytical expressions (Bosco et al. 2012). The envelope of the re-
sults obtained by the two pushover analyses provides the prediction of the displacement demands.
The effect of higher modes along the height of the building is not explicitly considered by the cor-
rective eccentricity method. However, it could be incorporated in the pushover analysis through the
adoption of a proper load pattern (e.g. Gupta and Kunnath 2000; Marino et al. 2003; Mwafy and
4
In this paper, the benefits of the two improved nonlinear static methods are assessed by comparing
the error produced by these methods in the prediction of the displacement demands with that pro-
duced through the use of the original version of the N2 method. Further, the results obtained by
means of the extended N2 method and those provided by the corrective eccentricity method are
compared in order to identify the cases in which one of the two methods should be preferred. The
benchmark response is obtained by nonlinear dynamic analysis which is the most advanced numeri-
cal method of analysis. A comparison with results of shaking table tests would be even more relia-
ble (SeismoStruct Verification Report 2012). However, the limited number of lab tests on asymmet-
ric structures would not allow an extensive validation of the considered nonlinear static methods.
The investigation is first conducted on single-storey asymmetric systems because these systems
allow an extensive parametric analysis. This part of the study investigates a large set of single-
storey systems and considers the variation of many structural parameters. These systems are repre-
sentative of the majority of existing buildings. However, it is well known that simplified single-
storey systems describe only the main aspects of the torsional behaviour of real asymmetric build-
ings and do not cover important peculiarities related to the heightwise distribution of some structur-
al properties (Anagnostopoulos et al. 2010; De Stefano et al. 2006; Ghersi et al. 2007). Hence, in a
second part of the study the investigated nonlinear static methods are applied to a set of multi-storey
structures.
The procedure for the application of the extended N2 method can be summarised in three steps.
First, the original N2 method is applied to evaluate the displacement demand of the building. Seis-
mic forces are applied to the mass centres CM and considered as acting separately along two hori-
zontal directions. According to Kreslin and Fajfar, the pushover analyses can be performed on ei-
ther two planar models (one for each direction of the seismic forces) or one 3-D model. In this pa-
5
per, the former option is adopted. The single pushover analysis is carried out until the displacement
of the centre of mass of the roof reaches the displacement demanded by the assigned peak ground
Second, two modal response spectrum analyses of the 3-D model, one for each of the two horizontal
directions of the seismic action, are performed. While the effects of the modes of vibration of the
single modal response spectrum analysis are combined according to the CQC rule, those of the two
modal response spectrum analyses are combined by the SRSS rule. Further, the results of the modal
response spectrum analysis are scaled so that the roof displacement at CM equals that provided by
the pushover analysis. The two scaled modal response spectrum analyses are used to define two sets
of correction factors to be applied to the results of the pushover analyses. One of the sets of correc-
tion factors refers to the in-plan distribution of the displacements of the floors and the other to the
heightwise distribution of the storey drifts. In particular, the correction factor regarding the in-plan
distribution of the displacements is calculated at the roof level as the ratio of the displacement at an
arbitrary location to the displacement at the centre of mass CM. The correction factor referring to the
storey drifts is calculated at the centre of mass of each storey as the ratio of the storey drift obtained
by modal response spectrum analysis to the storey drift obtained by pushover analysis. If a correc-
tion factor (in plan or in elevation) is smaller than 1.0, the value 1.0 is used instead. This is because
Third, the results of the pushover analysis are adjusted by means of the two sets of correction fac-
tors. The correction factor determined for displacements also applies to storey drifts. According to
the Authors, the in-plan correction factors should be calculated dividing the modal roof displace-
ments of the asymmetric system by the modal roof displacement of the system obtained by restrain-
ing the deck rotation of the 3-D scheme, hereinafter named corresponding planar system. However,
the displacements provided at the centre of mass CM by the planar and 3-D models are generally
close each other and thus these approaches provide similar values of the correction factors.
6
The effectiveness of the extended N2 method is based on the assumption that the torsional amplifi-
with respect to that achieved with nonlinear dynamic analysis (Marusic and Fajfar 2005; Perus and
Fajfar 2005). This is true in most cases; however, a recent study by De Stefano and Pintucchi
(2010) has shown that at the flexible side of structures characterized by very high torsional stiffness
the inelastic torsional response determines an amplification of the translational displacements which
is often greater than that corresponding to an elastic behaviour. The very limited number of applica-
tions present in the literature on multi-storey buildings (Bhatt and Bento 2011) confirms the poten-
tiality of the extended N2 method in predicting the torsional response of buildings. Nonetheless, the
Authors underline the importance of investigating this aspect further in order to consolidate this
The method predicts the displacement demand of asymmetric buildings through the envelope of the
results of two nonlinear static analyses for each direction of the seismic action. The lateral forces
are applied to two points of the deck which are generally different from the centre of mass CM. The
difference between the abscissa of the point where the force F is applied and the abscissa of CM is
named corrective eccentricity ei. The corrective eccentricities are expressed as a function of four
parameters which affect the torsional response of asymmetric structures: the rigidity eccentricity er
(distance between the centre of rigidity CR and CM), the ratio Ωθ of the torsional to lateral frequen-
cies of the corresponding torsionally balanced system (i.e. of the system obtained by shifting CM
into CR), the strength eccentricity es (distance between the centre of strength CS and CM) and the
ratio Rμ of the elastic strength demand to the actual strength of the corresponding planar system.
While er and Ωθ influence the torsional response in both the elastic and inelastic range of behaviour
(Hejal and Chopra 1987; Goel and Chopra 1990; Palermo et al. 2013), es and Rμ influence only the
inelastic torsional response (Goel and Chopra 1990). The corrective eccentricities are linear func-
7
ei ai es bi er i = 1 or 2 (1)
where the coefficients ai and bi are functions of Ωθ and Rμ as reported by Bosco et al. (2012).
The rigidity eccentricity and the ratio of the uncoupled torsional to lateral frequencies can be evalu-
ated rigorously for single-storey systems and regularly asymmetric multi-storey systems (Hejal and
Chopra 1987). However, several approximated methods are available in the literature for the deter-
mination of er and of non-regularly asymmetric systems (Makarios and Anastassiadis 1998a and
1998b; Bosco et al. 2013; Calderoni et al. 2002; Doudomis and Athanatopoulou 2008; Habibullah
1998; Makarios 2008; Marino and Rossi 2004; Moghadam and Tso 2000).
In multi-storey systems the heightwise distribution of the centres of strength can be obtained by a
pushover analysis of the building which is restrained to translate along one direction. At each step
of the pushover analysis, the centre of strength of the first storey is evaluated as the point in which
the base shear is applied. The position of CS changes as the pushover analysis progresses and the
roof displacement increases because of the gradual yielding of the resisting elements. Hence, the
strength eccentricity es is conventionally calculated as the distance between the centre of strength
corresponding to the target roof displacement and the centre of mass. The Rμ factor is evaluated as
the ratio of the required elastic base shear to the lateral strength of the structure. The elastic base
shear is calculated by modal response spectrum analysis of the corresponding planar system. The
lateral strength is calculated as the total base shear provided by the pushover analysis of the corre-
sponding planar system when the roof displacement equals the target value. A numerical example
SYSTEMS
In this section the effectiveness of the nonlinear static methods in estimating properly the maximum
dynamic displacements of single-storey asymmetric systems is verified. To this end, a large set of
single-storey systems is subjected to 10 pairs of ground motions and their seismic response is de-
8
termined by nonlinear dynamic analysis. The average of the displacement demands obtained for the
10 bidirectional ground motions is assumed as the benchmark for the nonlinear static methods. The
number of the accelerograms is a compromise between the minimum value (seven) suggested by
Eurocode 8 and larger values (twenty or more) suggested by other Authors (Bosco and Marino
2013; Kreslin and Fajfar 2012; Marino and Nakashima 2006) in order to have a better evaluation of
the average response of structures subjected to seismic actions which are represented by the as-
Then, the errors committed by the nonlinear static methods (original N2 method, extended N2
method and corrective eccentricity method) in the estimate of the distribution of the maximum dy-
namic displacements are determined and analysed. Two approximations may undermine the accura-
cy of the nonlinear static methods: one is related to the evaluation of the target displacement and the
other to the in-plan distribution of the displacement demand (Kreslin and Fajfar 2012). These two
approximations are independent and can be separately considered. As the object of this paper is to
test and compare the effectiveness of the analysed nonlinear static methods in predicting the in-plan
distribution of the displacement demand, the errors in the estimate of the target displacement of the
centre of mass of the roof (Bosco et al. 2009) are eliminated. To achieve this goal, the target dis-
placement of the centre of mass of the asymmetric system is set equal to the average of the 10 max-
The analysed single-storey systems are monosymmetric and can be categorised into 9 groups (see
Table 1). The systems belonging to the reference group (Sys01) are those adopted for the calibration
of the analytical equations of the corrective eccentricities (Bosco et al. 2012). These systems are
constituted by a rigid deck which is rectangular in plan. The deck has dimensions, denoted as B and
L in Figure 1a, equal to 12.5 m and 29.5 m, respectively. The mass m of the deck is equal to 1416 t,
while the mass radius of gyration rm about CM is equal to 0.312 L. The deck is supported by 4 resist-
ing elements arranged along the x-axis of the assumed reference system (Fig. 1) and by 8 resisting
9
elements arranged along the y-axis. All the resisting elements have lateral stiffness and strength in
their plane only. The force-displacement relationship of each vertical resisting element is assumed
to be bilinear with no strain hardening. The contribution γx of the resisting elements arranged along
the x-axis to the total torsional stiffness about CR is equal to 20%. The values adopted for rm and γx
are proper for systems with a rectangular plan. The systems are mass eccentric (MES) and the un-
coupled periods in the x- and y-directions, Tx and Ty, (i.e. the periods of the corresponding torsional-
Systems belongings to groups Sys02 and Sys03 differ from the reference systems in the value of the
strain hardening ratio h of the resisting elements (ratio of the post-yield lateral stiffness to the elastic
one). In fact, the ratio h is set equal to 0.05 and 0.10 for systems belonging to groups Sys02 and
Sys03, respectively. Systems of groups Sys04-Sys07, instead, differ from the reference systems in
the values of the uncoupled periods, which are in the range from 0.3 s to 1.8 s. Unlike the reference
systems, those belonging to group Sys08 (Fig. 1b) are stiffness eccentric (SES). Finally, systems
Sys09 are different from the others because they have a deck which is squared in plan (Fig. 1c),
sustained by 8 resisting elements along the x-axis and by 8 resisting elements along the y-axis. Fur-
ther, in these latter systems the mass radius of gyration rm about CM is equal to 0.406 L and the con-
tribution γx of the resisting elements arranged along the x-axis to the total torsional stiffness about
CR is equal to 50%.
For each group of systems, different structural schemes are generated by modifying the position of
CM, the lateral stiffness and the lateral strength of the resisting elements according to the procedures
described in Ghersi and Rossi (2000) and Bosco et al. (2012), respectively. These procedures are
applied so as to obtain asymmetry with respect to the y-axis and to maintain symmetry with respect
to the x-axis. To analyse single-storey systems which are representative of the majority of real
buildings, the four parameters (Ωθ, er, Rμ and es) that mostly influence the torsional response have
been varied in a large range of values. In particular, the torsional to lateral uncoupled frequency
ratio Ωθ is assumed as ranging from 0.6 to 1.4 in steps of 0.05 so as to analyse both torsionally flex-
10
ible (Ωθ < 1) and torsionally rigid systems (Ωθ > 1). Values of er ranging from -0.10 L to 0 in steps
of 0.025 L are instead considered to include both systems with large to small rigidity eccentricity.
The total lateral strength of the systems is derived by assigning values of the parameter Rμ ranging
from 1.0 to 6.0 in steps of 1.0 so as to consider systems that may experience moderate to large plas-
tic deformations during the seismic event. As the value Rμ = 1.0 only ensures that the mean re-
sponse the planar systems is elastic, the value Rμ = 0.5 is also considered to analyse asymmetric
systems which are expected to behave elastically. Finally, the total lateral strength is distributed
among the resisting elements so as to return values of es ranging from -0.10 L to 0.10 L in steps of
0.025 L; 10 random distributions of the lateral strength have been considered for each value of es.
The Newmark method is used to evaluate the dynamic response of the analysed systems. The pa-
rameters and are set equal to 0.25 and 0.5, respectively. The Rayleigh formulation is considered
for damping. In particular, the viscous damping factor is set equal to 5% for the first and the third
modes of vibration of the system. The seismic input consists of a set of 10 bidirectional ground mo-
tions. The x- and y-components of each ground motion are generated by the SIMQKE computer
program (SIMQKE 1976). Each accelerometric component is scaled by means of a trapezoidal en-
velope function with initial, central (stationary part) and final parts of 3, 22.5 and 5 s, respectively
(Bosco and Rossi 2009, 2013; Ghersi and Rossi 2001; Rossi and Lombardo 2007). The mean re-
sponse spectrum of the 10 x- and y-accelerometric components is compatible with the elastic spec-
trum proposed by EC8 for soil type C and characterized by 5% damping ratio and peak ground ac-
celeration ag equal to 0.35 g (Marino et al. 2005). Further, the mean of the zero period spectral re-
sponse acceleration values is not lower than the value stipulated by EC8.
The effectiveness of the nonlinear static methods is assessed by comparing the maximum displace-
ments provided by the nonlinear dynamic analysis along the y-axis to those provided by the nonlin-
11
ear static methods. In particular, the percentage difference between the displacement evaluated by
the nonlinear static method and the corresponding value obtained by the dynamic analysis is calcu-
lated. Positive percentage differences indicate a conservative estimate provided by the nonlinear
Three types of errors are calculated with regard to the whole system: (i) the maximum unconserva-
tive error (indicated as max UCo error in figures), (ii) the maximum conservative error (max Co
error in figures), and (iii) the average absolute error, i.e. the average of the absolute values of the
errors committed for all the resisting elements (AAv error in figures).
To summarise the results obtained for the large number of structures analysed, each of the above
errors is processed and reported in a compact form. As an example, Figure 2 illustrates the main
steps of the process for the graphical representation of the maximum unconservative error. In par-
ticular, the error committed in the prediction of the dynamic displacement is first calculated in the
assigned system (identified by fixed values of Ωθ, Rμ es, er and by an assigned distribution of
strength among the resisting elements) with regard to each resisting element arranged along the
y-direction, as shown in Figure 2a. This error is conservative for some resisting elements and un-
conservative for some other elements. As the values of the errors corresponding to the 10 strength
distributions are always very close to one another, the average of the 10 values is considered as rep-
resentative of the results corresponding to the different strength distributions. This error is calculat-
ed for systems with fixed values of Ωθ and Rμ and different values of es and er and the results are
plotted as in Figure 2b. Then, the average value of the errors committed for the systems character-
ized by different values of es and er is determined (Fig. 2c). The procedure is repeated for all the
considered values of Ωθ and Rμ so as to obtain the graph plotted in Figure 2d which is representative
of a whole group of systems. The same procedure is adopted to summarise the results in terms of
12
3.4 Discussion of the results
The errors committed by the considered nonlinear static methods are shown and compared in Fig-
ures 3, 4, 5, 6 and 7. Grey, dark and white surfaces represent the errors committed by the original
N2 method, the extended N2 method and the corrective eccentricity method, respectively.
Figure 3 shows the results obtained for the reference systems (Sys01) and highlights that the origi-
nal N2 method is able to predict accurately the displacements of torsionally rigid structures. In fact,
for systems with Ωθ larger than 1.2 the dynamic displacements are only slightly underestimated or
overestimated (up to 10%). Further, the average errors are modest because the in-plan distribution
of the displacements of torsionally rigid systems is almost linear and thus it can be well approxi-
mated by the results of a single pushover analysis. Conversely, significant errors are committed by
the original N2 method in the prediction of the displacements of very torsionally flexible structures.
50%. Figure 3 also shows that the use of the improved nonlinear static methods enhances signifi-
cantly the prediction of the displacement demand. In fact, both the improved methods lead to small
unconservative and average errors. In the worst case, the dynamic displacements are underestimated
by the extended N2 method up to -21% and by the corrective eccentricity method up to -16%. Fur-
ther, the extended N2 method is generally more conservative for torsionally rigid structures (due to
the no-reduction rule) and for systems which are well excited into the inelastic range of behaviour
Similar results are obtained for single-storey systems sustained by resisting elements with strain
hardening (Sys02 and Sys03 systems). As an example, Figure 4 shows that the maximum uncon-
servative errors committed by the improved nonlinear static methods for the Sys03 systems are
slightly lower than those reported for the reference systems while the maximum conservative and
the average errors are virtually equal to those of the reference systems. Further, Figure 4 confirms
that the use of the original N2 method leads to significant errors in the prediction of the displace-
ment demand of asymmetric structures except for torsionally rigid systems (Ωθ > 1.2).
13
Figures 5 and 6 show the results for systems characterised by Tx=Ty=0.3 s (Sys04) and Tx=Ty=1.8 s
(Sys07), respectively. It is evident that the effectiveness of the extended N2 method in the predic-
tion of the in-plan distribution of displacements is almost independent of the period of vibration.
The effectiveness of the corrective eccentricity method is similar for systems with Tx=Ty ranging
from 0.6 s to 1.8 s. Figure 6 shows that the errors committed for Sys07 systems (Tx=Ty=1.8 s) are
generally close to those obtained for the reference systems (Tx=Ty=1.0 s) (compare with Fig. 3). The
corrective eccentricity method is instead less effective in very short period systems (Sys04) (Fig. 5).
For these systems, an increase of all the considered errors is observed. For instance, the unconserva-
tive errors increase up to -32% while they are, at most, equal to -20% for the reference systems.
However, also for Sys04 systems the corrective eccentricity method is generally more reliable than
Results obtained for SES systems (not reported in any figure) and systems characterised by a square
deck (Fig. 7) confirm the previously described findings. The benefit obtained by the improved non-
linear static methods is, however, less evident for systems characterised by a square deck.
SYSTEMS
In this section the nonlinear static methods are applied to predict the seismic response of 12 r.c.
buildings and the results are compared to those of nonlinear dynamic analysis. The pushover analy-
sis is carried out assuming an adaptive load pattern which is able to match the maximum inelastic
displacements of the corresponding planar system. This choice is consistent with that made for sin-
gle-storey systems and is adopted for all the three nonlinear static methods analysed in the paper.
Beams and columns are modelled by elastic members with flexural plastic hinges at the two ends.
An elastic perfectly plastic moment-rotation relation is defined for each member based on a mo-
ment-curvature analysis. The moment curvature analysis of columns is performed assuming an axial
14
force equal to that resulting from the gravity loads of the seismic design situation. Both the pusho-
ver and nonlinear dynamic analyses are performed by the SAP computer program (CSI 2009).
The nonlinear dynamic analyses are performed assuming a Rayleigh formulation for damping. In
the planar structures, the parameters and are calculated assuming a viscous damping factor
equal to 5% for the first and second translational modes in the direction under examination. The
A set of 22 natural bidirectional ground motions is used for the nonlinear dynamic analyses. This
set is the “Far Field” record set used in the FEMA-P695 project (FEMA 2009). The accelerograms
are scaled as described in the study by Kreslin and Fajfar (2012). The benchmark seismic response
to be predicted by the nonlinear static methods is the average of the seismic responses resulting
from 44 nonlinear dynamic analyses (the horizontal components of each ground motion are first
assigned to the x- and y-directions and then the analysis is repeated swapping the directions of the
two components).
In order to eliminate the errors in the estimate of the seismic response of planar structures, an adap-
tive pushover analysis (PO) is considered. As the seismic response is investigated in terms of the in-
plan distribution of displacements and storey drifts, two adaptive load patterns are used (one for
The first load pattern is defined in such a way as to obtain, at all the storeys of the corresponding
planar structure, horizontal displacements equal to those of the nonlinear dynamic analysis. To this
end, a displacement-controlled pushover analysis (named PO-D) of the corresponding planar struc-
ture is performed using a displacement load vector proportional to the maximum dynamic dis-
placements of this system, increased with a constant step . The lateral force distribution corre-
sponding to the incremental displacement distribution is then evaluated at each step of the PO-D.
15
The obtained force distributions are used in the successive adaptive pushover analysis PO of the
asymmetric structure. In the PO the single force distribution is applied until the incremental dis-
The second adaptive load pattern, used for the prediction of storey drifts, is obtained by the de-
scribed procedure using the sum of the maximum storey drifts of the dynamic analysis as displace-
The analysed buildings are obtained through slight modifications of the building analysed by Kres-
lin and Fajfar (2012) to test the extended N2 method. This is a r.c. mass eccentric 8-storey building
with a square deck (L × L) sustained by eight frames arranged along two orthogonal directions
(Fig. 8). Each frame has three spans with length equal to 8.0 m. The interstorey height is equal to
5.00 m at the first two storeys and is equal to 3.10 m at all the other storeys. All the columns have
dimensions 60 × 60 cm and longitudinal reinforcement equal to 8 Ø25. The beams have dimensions
40 × 60 cm and longitudinal reinforcement equal to 4 Ø20 in the lower part of the section and
4 Ø20 plus 3 Ø12 in the upper part. Concrete C25/30 and steel reinforcement B500 are used. The
floor masses are equal to 481 t for the two lower storeys and 453 t for the other storeys; the radii of
gyration of mass are equal to 9.8 m (0.408 L). The centres of mass are lined up along a vertical axis
and present an eccentricity with respect to the geometrical centre of the deck equal to 5% of the
plan dimensions in each of the horizontal directions. Because of the symmetry of the structure the
centres of rigidity are lined up along a vertical axis passing through the geometrical centre of the
deck. Similar consideration applies to the centres of strength CS because the steel reinforcement is
symmetric with respect to the geometrical centre of the deck. Details on the design of this building
This building has been modified to obtain asymmetric structural systems characterised by different
values of rigidity and strength eccentricities and uncoupled torsional to lateral frequency ratio.
Since the centres of rigidity and strength have equal position in plan, no distinction is made herein-
16
after between rigidity and strength eccentricities. Thus, the symbols ex and ey are adopted to repre-
sent the eccentricities of the rigidity and strength centres along the x- and y-axes, respectively.
The analysed buildings may be subdivided into two sets. The first set consists of 6 buildings which
are obtained by varying the coordinates of the centre of mass. In particular, three values of eccen-
tricity e (–0.05 L, –0.10 L and –0.15 L) are considered in order to analyse buildings with low (-L)
medium (-M) or high eccentricity (-H). Further, for each value of the eccentricity, both mono-
(ex=0, ey = e) and bi-eccentric (ex= ey = e) buildings are analysed. The bi-eccentric buildings with
low eccentricity (e = –0.05 L) is equal to that analysed by Kreslin and Fajfar. As allowed in EC8,
the cracked concrete elements are modelled reducing the elastic stiffness to one-half of the corre-
The second set of buildings is obtained assuming that columns identified by symbol A in Figure 8
and beams AB, BB and BA are steel members. Further, the above-mentioned beams are linked to
the columns by pinned connections. The stiffness of the concrete members is assumed equal to the
stiffness of the uncracked elements. This choice allows the Authors to ensure that the lateral stiff-
ness of the buildings belonging to the second set be virtually the same as that of the buildings of the
first set. The 6 positions of the centre of mass considered for the first set of buildings are also con-
sidered for the buildings of the second set. The uncoupled torsional to lateral frequency ratio Ωθ of
the analysed buildings is calculated as the ratio of the average value (in elevation) of the radii of
gyration of the lateral stiffness rk to rm. The radii of gyration of the lateral stiffness are calculated at
each storey as suggested in Bosco et al. (2013) and the average value is equal to 12.85 m and
7.45 m for buildings belonging to the first and second sets, respectively. The buildings of the first
set are torsionally rigid (Ωθ = 1.31) while those of the second set are torsionally flexible (Ωθ = 0.76).
The Rμ factor calculated by the procedure described in Section 2.2 is equal to 2.38 and 3.49 for the
17
4.4 Discussion of the results
The comparison between the seismic responses obtained by the three considered nonlinear static
methods and the nonlinear dynamic analysis are expressed in terms of (i) in-plan distribution of the
floor displacements u at each story and (ii) heightwise distribution of the storey drift at the stiff (dr1)
and flexible sides (dr2) of the structures. When the corrective eccentricity method is used, the seis-
mic forces are applied to the considered buildings with the eccentricities listed in Table 2. It is
worth noting that the value of the corrective eccentricities is independent of the rigidity and strength
eccentricities in the orthogonal direction. When the extended N2 method is used, instead, the in-
plan correction factors listed in Table 3 are applied to adjust the results of the pushover analysis of
the corresponding planar structure. Note that no correction factors are considered along the height
of the building because an adaptive load pattern is used in the pushover analysis.
Figures 9 and 10 show the in-plan distribution of the displacements at the 8th storey of the torsion-
ally rigid structure with low eccentricity and that of the displacements of the torsionally flexible
structure with high eccentricity. Attention is focused on these structures because their seismic be-
haviour bounds that of the other analysed structures. In fact, the two structures are representative of
buildings with moderate and significant torsional response, respectively. The results of the mono-
eccentric buildings are plotted in Figures 9a and 10a, while those of the bi-eccentric buildings are
Figure 9 shows that all the considered methods are able to predict the in-plan distribution of the
displacements of the structures with Ωθ = 1.31, Rμ = 2.38 and e = –0.05 L. In fact, at each storey the
dynamic displacements are never underestimated or overestimated more than ±10%. The worst un-
conservative estimate is determined by the application of the original N2 method at the stiff side of
the bi-eccentric building. The corrective eccentricity method provides results which are close to the
actual values. These results are conservative for the mono-eccentric building and slightly underes-
timated at the stiff side of the bi-eccentric building. The extended N2 method is always conserva-
tive.
18
Different considerations apply to the prediction of the in-plan distribution of the displacements of
the structures with Ωθ = 0.76, Rμ = 3.49 and e = –0.15 L (Fig. 10). In fact, the application of the
original N2 method provides large unconservative errors, up to -50%, in the case of both the mono-
and bi-eccentric buildings. The extended N2 method is generally conservative, but the displace-
ments can be seriously overestimated (up to 64%). Some slight unconservative errors can be found
at the flexible side of the mono-eccentric building. The corrective eccentricity method provides re-
sults which are closer to the actual values and determines only slight unconservative errors (-5%) at
To discuss the results obtained for all the considered multi-storey buildings, a compact representa-
tion of the errors is reported. To this end, the percentage error committed in the estimate of the dis-
placements at the stiff or flexible side is first calculated at each storey. Then, the average value in
elevation is calculated and plotted as a function of the normalised eccentricity of the building e/L. In
each plot, a different type of line (dotted, dashed or dashed dotted line) is used to identify the adopt-
ed nonlinear static method; different symbols are adopted to pinpoint results obtained for mono-
eccentric (square symbols) or bi-eccentric (circle symbols) buildings. Figure 11a shows that, if the
original N2 method is applied (dotted lines) to the buildings of the first set (Ωθ = 1.31 and
Rμ = 2.38), the larger the rigidity eccentricity, the greater the underestimation of the displacements
at the stiff side. In particular, for bi-eccentric buildings with e equal to –0.15 L, the average error in
elevation is equal to –30%. The extended N2 method (dashed dotted line) matches almost exactly
the displacement at the stiff side of bi-eccentric buildings, independently of the eccentricity e; in-
stead, it is conservative for mono-eccentric buildings (average errors up to 17%). The corrective
eccentricity method (dashed line) is slightly unconservative for bi-eccentric buildings (errors up to –
14%). Figure 11b confirms the effectiveness of all the considered nonlinear static methods in the
prediction of the displacements at the flexible side: the original N2 and the corrective eccentricity
methods provide almost correct results, while the extended N2 method is slightly conservative (er-
rors up to 8%).
19
If buildings belonging to the second set are considered (Ωθ = 0.76 and Rμ = 3.49), the original N2
method appears to be ineffective in predicting the deck rotations (Fig. 12a, average unconservative
error up to –42% at the stiff side). Both the improved nonlinear static methods enhance the predic-
tion of the torsional response reducing the unconservative error significantly. Specifically, the ex-
tended N2 and the corrective eccentricity methods provide conservative results at the stiff side (con-
servative errors up to 62% and 20%, respectively); independently of the adopted method of analy-
sis, the unconservative and conservative errors committed in the estimate of the flexible side
The above-mentioned results are confirmed when the seismic response is expressed in terms of sto-
rey drift. In this regard, Figure 13 shows that the heightwise distribution of the storey drift at the
stiff side of the torsionally rigid structures with low eccentricity is well predicted by all the nonline-
ar static methods. Instead, the application of the original N2 method to torsionally flexible struc-
tures with high eccentricity (Fig. 14) leads to significant unconservative errors at all the storeys. In
these cases, the extended N2 method is conservative and the corrective eccentricity method pro-
vides results which are closer to the dynamic values even if slightly unconservative at the upper
The average percentage errors committed in elevation in the evaluation of the storey drift at the stiff
and flexible sides of the buildings are also calculated and plotted in Figures 15 and 16 as a function
of the normalised eccentricity e/L. The above-mentioned figures confirm the findings obtained by
5 CONCLUSIONS
In this paper the effectiveness of three nonlinear static methods (the original N2 method, the ex-
tended N2 method and the corrective eccentricity method) in the prediction of the seismic response
of asymmetric systems is compared. Specifically, the selected methods are applied to predict the
seismic response of a large set of asymmetric single-storey systems and that of 12 multi-storey
20
buildings. The prediction of the storey displacements and, in the case of the multi-storey buildings,
also that of the storey drifts is compared to the response obtained by nonlinear dynamic analyses.
The results show that the original N2 method is able to predict only the response of very torsionally
rigid systems. The use of the two improved nonlinear static methods always enhances the prediction
of the displacement demand of asymmetric structures and the obtained enhancement is remarkable
especially in the case of torsionally flexible structures. Based on the obtained results, the extended
N2 method distinguishes because it is always conservative and very easy to apply. However, it is
sometimes overly conservative. The corrective eccentricity method, instead, provides results which
are closer to those obtained by nonlinear dynamic analyses. This latter method requires a greater
computational effort than the extended N2 method, because three pushover analyses are needed
Finally, it is worth noting that the results obtained in this study should be further validated against
results of shaking table tests or models which are more refined than those used in this paper (e.g.,
fiber models could be used in order to take account of interaction phenomena arising in the nonline-
ar range of behaviour).
ACKNOWLEDGEMENTS
The Authors wish to thank Dr. Annalisa Licandro for her contribution to the numerical analyses
REFERENCES
Anagnostopoulos S.A., Alexopoulou C., Stathopoulos K.G. (2010) An answer to an important con-
troversy and the need for caution when using simple models to predict inelastic earthquake re-
sponse of buildings with torsion. Earthquake Engineering and Structural Dynamics 39: 521–
540.
21
Antoniou S., Pinho R. (2004). Development and verification of a displacement-based adaptive
Bento R., Bhatt C., Pinho R. (2010). Verification of nonlinear static procedures for a 3D irregular
Bhatt C., Bento R. (2011) Assessing the seismic response of existing RC buildings using the ex-
Bosco M., Ghersi A., Marino E.M. (2009) On the Evaluation of Seismic Response of Structures by
Nonlinear Static Methods. Earthquake Engineering and Structural Dynamics 38: 1465–1482.
Bosco M., Ghersi A., Marino E.M. (2012) Corrective Eccentricities for assessment by the nonlinear
Bosco M., Marino E.M. (2013) Design method and behavior factor for steel frames with buckling
Bosco M., Marino E.M., Rossi P.P. (2013) An analytical method for the evaluation of the in-plan
10.1007/s10518-013-9438-3.
Bosco M., Rossi P.P. (2009) Seismic behaviour of eccentrically braced frames. Engineering Struc-
Bosco M., Rossi P.P. (2013) A design procedure for dual eccentrically braced systems: numerical
Calderoni B., D’Aveni A., Ghersi A. Rinaldi, Z. (2002) Static vs. modal analysis of asymmetric
231.
Causevic M., Mitrovic S. (2011) Comparison between non-linear dynamic and static seismic analy-
9: 467–489.
22
CEN. EN 1998-1. (2004) EuroCode 8: Design of structures for earthquake resistance – Part 1:
General rules, seismic actions and rules for buildings. European Committee for Standardization,
Bruxelles.
Chopra A.K., Goel R.K. (2002) A modal pushover analysis procedure for estimating seismic de-
mands for buildings. Earthquake Engineering and Structural Dynamics 31: 561–582.
Chopra A.K., Goel R.K. (2004) A modal pushover analysis procedure to estimate seismic demands
for unsymmetric-plan buildings. Earthquake Engineering and Structural Dynamics 33: 903–927.
CSI Analysis Reference Manual For SAP2000 (2009), ETABS, and SAFE, ISO# GEN062708M1
De Stefano M., Marino E.M., Rossi P.P. (2006) Effect of Overstrength on the seismic behaviour of
Dynamics 39:1369–1394.
Doudomis I.N., Athanatopoulou A.M. (2008) Invariant torsion properties of multistorey asymmetric
buildings. The Structural Design of Tall and Special Buildings 17(1): 79-97.
Fajfar P., Fishinger M. (1988) N2 – A method for non-linear seismic analysis of regular buildings.
Fajfar P., Gaspersic P. (1996) The N2 method for the seismic damage analysis of rc buildings.
Fajfar P. (1999) Capacity spectrum method based on inelastic demand spectra. Earthquake Engi-
Fajfar P., Marušic D., Peruš I. (2005) Torsional effects in the pushover-based seismic analysis of
FEMA (2009) Quantification of building seismic performance factors. FEMA—P695 report Federal
23
Freeman S.A. (1998) The capacity spectrum method as a tool for seismic design. Proc. of the 11th
Fujii K. (2011) Nonlinear static procedure for multi-story asymmetric frame buildings considering
Ghersi A., Marino E.M., Rossi P.P. (2007) Static versus modal analysis: influence on inelastic re-
Ghersi A., Rossi P.P. (2000) Formulation of design eccentricity to reduce ductility demand in
Ghersi A, Rossi P.P. (2001). Influence of bi-directional ground motions on the inelastic response of
Goel R.K., Chopra A.K. (1990) Inelastic seismic response of one-story asymmetric-plan systems,
USA.
Gupta B, Kunnath S.K. (2000) Adaptive spectra-based pushover procedure for seismic evaluation
Habibullah A. (1998) Center of rigidity – a redefinition. Computers and Structures, Inc 3(1).
Hejal R., Chopra A.K. (1987) Earthquake response of torsionally-coupled buildings. Earthquake
of California at Berkeley.
Kreslin M., Fajfar P. (2012) The extended N2 method considering higher mode effects in both plan
Makarios T., Anastassiadis A. (1998a) Real and fictitious elastic axes of multi-storey buildings:
theory. The Structural Design of Tall and Special Buildings 7(1): 33-55.
Makarios T., Anastassiadis A. (1998b) Real and fictitious elastic axes of multi-storey buildings:
applications. The Structural Design of Tall and Special Buildings 7(1): 57-71.
24
Makarios T. (2008) Practical calculation of the torsional stiffness radius of multi-storey tall build-
ings. The Structural Design of Tall and Special Buildings 17(1): 39-65.
Marino E.M., Muratore M., Rossi P.P. (2003) Pushover analysis in the evaluation of the seismic
response of steel frames. Proc. of 3rd International Workshop and Seminar on “Behaviour of
Marino E.M., Nakashima M., Mosalam K.M. (2005) Comparison of European and Japanese seismic
Marino E.M., Nakashima M. (2006) Seismic performance and new design procedure for chevron
Marino E.M., Rossi P.P. (2004) Exact evaluation of the location of the optimum torsion axis. The
Marusic D., Fajfar P. (2005) On the inelastic seismic response of asymmetric buildings under bi-
Moghadam A.S., Tso W.K. (2000) Extension of Eurocode 8 torsional provisions to multi-storey
Mwafy A.M., Elnashai A.S. (2001) Static pushover versus dynamic collapse analysis of RC build-
Palermo M., Silvestri S., Gasparini G., Trombetti T. (2013) Physically-based prediction of the max-
Perus I., Fajfar P. (2005) On the inelastic torsional response of single-storey structures under bi-
Requena M., Ayala A.G. (2000) Evaluation of a simplified method for the determination of the non-
linear seismic response of RC frames. Proc. of the 12th World Conference on Earthquake Engi-
25
Rossi P.P., Lombardo A. (2007) Influence of the link overstrength factor on the seismic behaviour
Sasaki K.K., Freeman S.A., 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 6th U.S. National
_News=42.
SIMQKE (1976) A program for artificial motion generation, User’s manual and documentation,
26
Table 1. Main properties of the analysed single-storey systems.
Type of Uncoupled periods Strain hardening Deck dimensions Contribution
Label
Asymmetry (Tx, Ty) ratio h (B x L) x
Sys 01 MES 1.0 s 0.00 12.5 m x 29.5 m 20%
Sys 02 MES 1.0 s 0.05 12.5 m x 29.5 m 20%
Sys 03 MES 1.0 s 0.10 12.5 m x 29.5 m 20%
Sys 04 MES 0.3 s 0.00 12.5 m x 29.5 m 20%
Sys 05 MES 0.6 s 0.00 12.5 m x 29.5 m 20%
Sys 06 MES 1.4 s 0.00 12.5 m x 29.5 m 20%
Sys 07 MES 1.8 s 0.00 12.5 m x 29.5 m 20%
Sys 08 SES 1.0 s 0.00 12.5 m x 29.5 m 20%
Sys 09 MES 1.0 s 0.00 29.5 m x 29.5 m 50%
Table 2. Adopted corrective eccentricities.
System er= es=e Ωθ Rμ e1 e2
Mono-01, Bi-01 -0.05 L 1.312 2.384 -0.025 L -0.005 L
Mono-02, Bi-02 -0.10 L 1.312 2.384 -0.050 L -0.010 L
Mono-03, Bi-03 -0.15 L 1.312 2.384 -0.074 L -0.015 L
Mono-04, Bi-04 -0.05 L 0.760 3.489 -0.063 L -0.036 L
Mono-05, Bi-05 -0.10 L 0.760 3.489 -0.125 L -0.072 L
Mono-06, Bi-06 -0.15 L 0.760 3.489 -0.226 L -0.130 L
Side 2 (flexible)
Side 1 (stiff)
B CR=G CS CM
Side 2 (flexible)
x
Side 1 (stiff)
es er
L B CR=G CS CM x
(b) es er
y
Side 2 (flexible)
Side 1 (stiff)
CR CS CM
B
x
es
er L
Figure 1. Plan layout of the analysed systems: (a) Sys01-07; (b) Sys08; (c) Sys09.
(a) (b)
CR CS CM x
(c) (d)
Figure 3. Maximum unconservative, conservative and average errors for Sys01 systems.
Original N2 method Extended N2 method Corrective eccentricity method
Figure 4. Maximum unconservative, conservative and average errors for Sys03 systems (h = 10%).
Original N2 method Extended N2 method Corrective eccentricity method
Figure 5. Maximum unconservative, conservative and average errors for Sys04 systems (Tx = Ty =0.3 s).
Original N2 method Extended N2 method Corrective eccentricity method
Figure 6. Maximum unconservative, conservative and average errors for Sys07 systems (Tx = Ty =1.8 s).
Original N2 method Extended N2 method Corrective eccentricity method
Figure 7. Maximum unconservative, conservative and average errors for Sys09 systems (x =50%).
Y
3.1 m
A B B A
3.1 m
ex
3.1 m
B C CM C B 3.1 m
ey
L = 24 m 3.1 m
CR=CS X
3.1 m
B C C B
5.0 m
A B B A
5.0 m
8m 8m 8m
L = 24 m
(a) (b)
u u
(cm) 8th storey extended N2 (cm) 8th storey
method
30 30
nonlinear
dynamic
20 analyses 20
corrective
eccentricities
10 10
N2 method
0 0
- L /2 CR CM L /2 - L /2 CR CM L /2
30 30
nonlinear
dynamic
20 corrective analyses 20
eccentricities
10 10
N2 method 8th storey 8th storey
0 0
- L /2 CR CM L /2 - L /2 CR CM L /2
Figure 10. Storey displacement of the buildings with Ωθ = 0.76, Rμ = 3.49, e = – 0.15 L:
(a) mono-eccentric building; (b) bi-eccentric building.
Mono-eccentric buildings Bi-eccentric buildings
(a) (b)
Err (%) Err (%)
50% 50%
25% 25%
0% 0%
-25% -25%
-50% -50%
Stiff Side Flexible Side
-75% -75%
-0.05 -0.10 e /L -0.15 -0.05 -0.10 e /L -0.15
Figure 11. Average error along the height in the prediction of the displacement of the buildings
with Ωθ = 1.31 and Rμ = 2.38 at the (a) stiff and (b) flexible side.
Mono-eccentric buildings Bi-eccentric buildings
(a) (b)
Err (%) Err (%)
50% 50%
25% 25%
0% 0%
-25% -25%
-50% -50%
Stiff Side Flexible Side
-75% -75%
-0.05 -0.10 e /L -0.15 -0.05 -0.10 e /L -0.15
Figure 12. Average error along the height in the prediction of the displacement of the buildings
with Ωθ = 0.76 and Rμ = 3.49 at the (a) stiff and (b) flexible side.
Original N2 method Extended N2 method Corrective eccentricity method
(a) (b)
ns ns
extended N2 method
7 7
6 corrective 6
eccentricities
5 5
4 nonlinear 4
dynamic
3 3
analyses
2 2
N2 method
1 1
0 0
0.0% 0.5% 1.0% 1.5% 2.0% d r1 0.0% 0.5% 1.0% 1.5% 2.0% d r1
Figure 13. Storey drift at the stiff side of the buildings with Ωθ = 1.31, Rμ = 2.38, e = – 0.05 L:
(a) mono-eccentric building; (b) bi-eccentric building.
Original N2 method Extended N2 method Corrective eccentricity method
ns corrective
ns
7 eccentricities 7
6 extended 6
N2 method
5 5
4 N2 4
method
3 3
2 nonlinear 2
dynamic
1 analyses 1
0 0
0.0% 0.5% 1.0% 1.5% 2.0% d r1 0.0% 0.5% 1.0% 1.5% 2.0% d r1
Figure 14. Storey drift at the stiff side of the buildings with Ωθ = 0.76, Rμ = 3.49 e = –0.15 L:
(a) mono-eccentric building; (b) bi-eccentric building.
Mono-eccentric buildings Bi-eccentric buildings
(a) (b)
Err (%) Err (%)
50% 50%
25% 25%
0% 0%
-25% -25%
-50% -50%
Stiff Side Flexible Side
-75% -75%
-0.05 -0.10 e /L -0.15 -0.05 -0.10 e /L -0.15
Figure 15. Average error along the height in the prediction of the storey drift of the buildings with
Ωθ = 1.31 and Rμ = 2.38 at the (a) stiff and (b) flexible side.
Mono-eccentric buildings Bi-eccentric buildings
(a) (b)
Err (%) Err (%)
50% 50%
25% 25%
0% 0%
-25% -25%
-50% -50%
Stiff Side Flexible Side
-75% -75%
-0.05 -0.10 e /L -0.15 -0.05 -0.10 e /L -0.15
Figure 16. Average error along the height in the prediction of the storey drift of the buildings with
Ωθ = 0.76 and Rμ = 3.49 at the (a) stiff and (b) flexible side.