Professional Documents
Culture Documents
A Consecutive Modal Pushover Procedure For Estimating The Seismic Demands of Tall Buildings - Mehdi Poursha, Et Al, 2009
A Consecutive Modal Pushover Procedure For Estimating The Seismic Demands of Tall Buildings - Mehdi Poursha, Et Al, 2009
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
article info a b s t r a c t
Article history: The nonlinear static procedure (NSP), based on pushover analysis, has become a favourite tool for use
Received 6 August 2007 in practical applications for building evaluation and design verification. The NSP is, however, restricted
Received in revised form to single-mode response. It is therefore valid for low-rise buildings where the behaviour is dominated
19 October 2008
by the fundamental vibration mode. It is well recognized that the seismic demands derived from the
Accepted 20 October 2008
Available online 10 December 2008
conventional NSP are greatly underestimated in the upper storeys of tall buildings, in which higher-
mode contributions to the response are important. This paper presents a new pushover procedure
Keywords:
which can take into account higher-mode effects. The procedure, which has been named the consecutive
Consecutive modal pushover (CMP) modal pushover (CMP) procedure, utilizes multi-stage and single-stage pushover analyses. The final
procedure structural responses are determined by enveloping the results of multi-stage and single-stage pushover
Seismic demands analyses. The procedure is applied to four special steel moment-resisting frames with different heights.
Higher-mode effects A comparison between estimates from the CMP procedure and the exact values obtained by nonlinear
Tall buildings response history analysis (NL-RHA), as well as predictions from modal pushover analysis (MPA), has been
carried out. It is demonstrated that the CMP procedure is able to effectively overcome the limitations of
traditional pushover analysis, and to accurately predict the seismic demands of tall buildings.
© 2008 Elsevier Ltd. All rights reserved.
ζn damping ratio of the nth mode in which Γn is the modal participating factor of the nth mode and
ωn natural frequency of the nth mode φn is the corresponding mode-shape. The displacement of an N
Γn modal participating factor of the nth mode degree-of-freedom (NDOF) system, u, can be defined by the sum
αn effective modal participating mass ratio of the nth of the modal contributions:
mode N
δt target displacement of the roof
u( t ) =
X
φn qn (t ) (4)
n =1
upper-bound pushover analysis [18] was developed in which the where the modal co-ordinate, qn (t ), is governed by
contributions of the first two elastic modes to the invariant load
q̈n + 2ζn ωn q̇n + ωn2 qn = −Γn üg (t ) (5)
pattern were combined through the absolute sum rule. A single
pushover analysis was then implemented with the resulting load in which ωn and ζn are the natural vibration frequency and
pattern. Other researchers proposed a two-phase load pattern [19], damping ratio of the nth mode, respectively. Γn is obtained as
and found it suitable for approximating the seismic capacity curve follows:
only for low- to mid-rise buildings. This procedure was not verified φnT mi
with respect to the estimation of the seismic demands of buildings. Γn = . (6)
In another study, an ‘‘incremental response spectrum analysis’’ φnT mφn
(IRSA) [3] was developed. In this procedure, whenever a new The solution to Eq. (5) is
plastic hinge occurs, elastic modal spectrum analysis is executed in
qn (t ) = Γn Dn (t ) (7)
order to take into account the changes in the dynamic properties
of the structure. More recently, an adaptive modal combination where Dn (t ) is governed by the equation of motion for a single-
(AMC) [20] procedure was proposed, in which the applied lateral degree-of-freedom (SDOF) system subjected to üg (t ):
forces are updated in accordance with the changes in the dynamic
characteristics during inelastic analysis, for each mode. The total D̈n + 2ζn ωn Ḋn + ωn2 Dn = −üg (t ). (8)
response is obtained by combining the peak modal responses, The floor displacements, u(t ), can be expressed by substituting Eq.
using an appropriate combination scheme at the end. The accuracy (7) into Eq. (4), so that
of this procedure was not demonstrated for the estimation of
N
plastic rotations of the hinges. X
The main objective of the present paper is therefore to propose a u( t ) = Γn φn Dn (t ) . (9)
n=1
new procedure which can take into account higher-mode effects in
the pushover analysis of tall buildings, and can improve estimates Making use of Eq. (6), the effective modal mass, Mn∗ , and the
of seismic demands, especially of the plastic rotations of the hinges. effective modal participating mass ratio for the nth mode, αn , can
The new procedure, called the consecutive modal pushover (CMP) be defined as
procedure, benefits from the concepts of structural dynamics
Mn∗ = Ln Γn (10)
and uses multi-stage and single-stage pushover analyses that are
∗
elaborated in detail in this paper. To demonstrate its applicability Mn
αn = (11)
and effectiveness, the procedure has been applied to four special M∗
steel moment-resisting frames. Predictions based on the MPA in which
procedure are also presented for the sake of comparison to
those obtained by the CMP procedure. Seismic demands obtained Ln = φnT mi (12)
by approximate pushover procedures are compared with exact N
X
solutions derived from nonlinear response history analysis. The M∗ = mj (13)
results of the comprehensive analyses, which are summarized j =1
M. Poursha et al. / Engineering Structures 31 (2009) 591–599 593
where M ∗ is the total mass of the structure obtained by summation a triangular or uniform load distribution. Finally, the seismic
of the lumped masses, mj , over all floor levels. An important result demands are obtained by enveloping the peak responses derived
can be derived by pre-multiplying Eq. (3) by iT : from the multi-stage and single-stage pushover analyses.
The details of the CMP procedure are expressed as a sequence
N N
X X of the following steps:
Mn∗ = mj . (14)
n=1 j =1 1. Calculate the natural frequencies, ωn , and the mode-shapes,
This result shows that the sum of the effective modal masses over
φn . These properties are determined by eigen-analysis of the
linearly elastic structure for the first three modes. The mode-
all modes (N) is equal to the total mass of the building [8]. As a
shapes are normalized so that the roof component of φn equals
result from Eqs. (11), (13) and (14), the summation of effective
unity (φrn = 1).
modal participating mass ratios over all modes is equal to unity.
2. Compute s∗n = mφn , where s∗n shows the distribution of
incremental lateral forces over the height of the structure for
3. Consecutive modal pushover procedure
the nth stage of multi-stage pushover analysis.
3. Compute the total target displacement of the structure at the
The consecutive modal pushover (CMP) procedure is presented,
roof, δt .
which can be used to estimate the peak response of inelastic struc-
tures subjected to earthquake excitation. Some pushover analyses 4. The CMP procedure consists of single-stage and multi-stage
are employed in the CMP procedure since it is possible to use dif- pushover analyses. First apply the gravity loads and then per-
ferent pushover analyses and to envelope the results [16]. The pro- form these displacement-control pushover analyses according
cedure uses multi-stage and single-stage pushover analyses. The to the following sub-steps:
multi-stage pushover analysis benefits from consecutive imple- 4.1 Perform the single-stage pushover analysis using an
mentation of modal pushover analyses, including a limited number inverted triangular load pattern for medium-rise buildings
of modes, such that when one stage (one modal pushover analysis) and a uniform force distribution for high-rise buildings,
has been completely performed, the next stage (the next modal until the control node at the roof sways to the predefined
pushover analysis) begins with an initial structural state (stress total target displacement, δt .
and deformation) which is the same as the condition at the end 4.2 The second pushover analysis is a two-stage pushover
of the previous stage. Consecutive modal pushover analyses are analysis. In the first stage, perform the nonlinear static
carried out with force distributions using mode-shapes obtained analysis, using the incremental lateral forces s∗1 = mφ1 ,
from eigen-analysis of the linearly elastic structure. Changes in until the displacement increment at the roof reaches ur1 =
the modal properties of the structure are ignored when the struc- β1 δt (Eq. (15); i = 1), where β1 = α1 (Eq. (16); i = 1).
ture experiences nonlinear yielding under increasing lateral loads Then implement the second stage of analysis using the
during pushover analysis. The number of modes in the consec- incremental lateral forces s∗2 = mφ2 until the displacement
utive modal pushover analyses (i.e. the number of stages in the increment at the roof equals ur2 = β2 δt (Eq. (15); i = 2),
multi-stage pushover analysis) depends on the fundamental pe- where β2 = 1 − α1 (Eq. (17); i = 2). It is noted that
riod, T , of the building structure. When the fundamental period of the initial condition in the second stage of the two-stage
the building (with a moment-resisting frame system) is less than pushover analysis is the same as the state at the last step of
2.2 s, the multi-stage pushover analysis is carried out in two stages. analysis in the first stage.
For buildings with fundamental periods of 2.2 s or more, both two- 4.3 The third analysis is a three-stage pushover analysis. It is
and three-stage pushover analyses are used. The displacement only performed for buildings having a fundamental period
increment at the roof is obtained, in each stage of multi-stage anal- of 2.2 s or more. The first stage is exactly the same as the
ysis, as the product of a factor and the total target displacement of first stage of the two-stage pushover analysis described
the roof. This factor is determined from the initial modal proper- above. After the first stage, continue the nonlinear static
ties of the structure. The displacement increment, uri , at the roof in analysis using the incremental lateral forces s∗2 = mφ2
the ith stage of multi-stage pushover analysis, is therefore calcu- until the displacement increment at the roof reaches ur2 =
lated as β2 δt (Eq. (15); i = 2) where β2 = α2 (Eq. (16); i = 2).
uri = βi δt (15) Thereafter, perform the third (last) stage of the three-stage
pushover analysis using the incremental lateral forces s∗3 =
in which mφ3 . The displacement increment at the roof at this stage
βi = αi for the stages before the last stage (16) is equal to ur3 = β3 δt (Eq. (15); i = 3) where β3 =
1 − α1 − α2 (Eq. (17); i = 3). It is again noted that the initial
and condition at each stage of the analysis is the same as the
Ns −1
X state at the end of the analysis in the previous stage.
βi = 1 − αj for the last stage (17) 5. Calculate the peak values of the desired responses, such as
j=1 displacements, storey drifts, and hinge plastic rotations, for the
pushover analyses described above. The peak values resulting
where δt is the total target displacement at the roof, and Ns is the
from the one-, two-, and three-stage pushover analyses are
number of stages included in the multi-stage pushover analysis.
denoted by r1 , r2 , and r3 , respectively.
Also, αi is the effective modal mass ratio for the ith mode, which
is derived from Eq. (11). Several different approaches can be used 6. Calculate the envelope, r, of the peak responses as follows:
to establish the total target displacement at the roof level. This r = Max {r1 , r2 } T < 2.2 s (18)
displacement can be determined by using the capacity spectrum
r = Max {r1 , r2 , r3 } T ≥ 2.2 s (19)
method [2], the displacement coefficient approach [4,5], the N2
method [15,16], or dynamic analysis of the structure [14,23,24,28]. showing that the seismic demand of the inelastic structure
As demonstrated previously, in addition to multi-stage pushover in the CMP procedure is obtained by enveloping the peak
analysis, single-stage pushover analysis is used in the procedure. responses resulting from the single- and multi-stage pushover
The single-stage pushover analysis is performed separately with analyses.
594 M. Poursha et al. / Engineering Structures 31 (2009) 591–599
Table 1
Characteristics of the analyzed frames.
No. No. of storeys h (m) b (m) Seismic mass of floors (kg s2 /m) Periods
T1 (s) T2 (s) T3 (s)
Fig. 3. (a) Pseudo-acceleration spectra and (b) displacement spectra of the set of far-field records of ground motions, damping ratio = 5%.
Table 2
List of the ground motions used.
No. Earthquake name Date Magnitude Station name Station number Component (deg) PGA (g)
Fig. 4. Peak values of storey drift ratios derived from pushover analyses used in the
CMP procedure and from NL-RHA for the 10-storey frame. Fig. 5. Peak values of storey drift ratios derived from pushover analyses used in the
CMP procedure and from NL-RHA for the 20-storey frame.
(a) Storey drift ratios of the 10-storey frame. (b) Storey drift ratios of the 15-storey frame.
Fig. 6. Height-wise variation of the storey drifts for the 10- and 15-storey frames.
(a) Storey drift ratios of the 10-storey frame. (b) Storey drift ratios of the 15-storey frame.
Fig. 7. Errors in the storey drifts for the 10- and 15-storey frames.
than those from the CMP at some lower storeys. For example, in in general, provides substantially better estimates of hinge plastic
the 15-storey frame, the storey drift ratios are underestimated by rotations than the MPA procedure for high-rise frames. It is
up to 11.7% and 25% at the upper eight storeys by the CMP and MPA noteworthy that a key aspect of the CMP procedure is the fact
procedures, respectively. In the lower seven storeys, the errors that modal pushover analyses are carried out continuously. The
from the MPA and CMP procedures reach 5% and 14%, respectively. consecutive implementation of modal pushover analyses means
More detailed discussion about the results of the MPA procedure that rotations of the plastic hinges are continuously accumulated
is available in [25]. at the mid and upper floor levels during the modes of interest in
The height-wise variation of hinge plastic rotations and the the multi-stage pushover analysis, whereas the MPA procedure
errors from pushover procedures for the medium-rise frames are attempts to estimate the total response quantities by combining
shown in Figs. 8 and 9, respectively. The figures show that the the individual peak responses obtained separately for each
MPA procedure fails to accurately predict hinge plastic rotations mode. The trend of the higher-mode influences and incremental
at the upper floor levels of the 10- and 15-storey frames, whereas application of lateral forces in the multi-stage pushover analysis
noticeable improvement was achieved in the estimates of these of the CMP procedure indicate a significant improvement in
rotations by the CMP procedure. The figures illustrate that the CMP
comparison with the MPA procedure in predicting hinge plastic
procedure is in closer agreement with the NL-RHA. As a result, the
rotations at the mid and upper floor levels. However, the pushover
hinge plastic rotations obtained by the CMP procedure are more
procedure suffers from the limitation that it is unable to take into
accurate than those obtained by the MPA procedure, especially at
account the cumulative rotation of hinges due to cyclic hysteretic
the mid and upper floors. It is noted that the 100% errors resulting
behaviour [21]. It is noted that the results derived from the CMP
from the CMP procedure at the last floor of the 10-storey frame and
procedure may be occasionally inaccurate at some mid or lower
at the first floor of the 15-storey frame are ignored since the plastic
rotations predicted by the CMP procedure are zero, whereas the floors of high-rise frames in which the plastic rotations obtained by
rotations obtained by NL-RHA are very small. At some lower floor NL-RHA are small. Some large errors in hinge plastic rotations (such
levels, the CMP procedure occasionally provides better estimates of as in the case of the last floor of frame S3) are also not important
plastic rotations than the MPA procedure, and vice versa. Also, the since the rotations computed by NL-RHA are very small.
CMP procedure tends to slightly overestimate the plastic rotation As can be seen from Figs. 6, 8, 10 and 12, the height-wise
of the hinges at some lower floor levels. distribution of storey drifts and hinge plastic rotations derived
From Figs. 10–13, consequences similar to those for the from the CMP procedure is more similar to the distribution
medium-rise frames can be almost deduced for the high-rise (20- obtained by the benchmark solution (NL-RHA) than to that
and 30-storey) frames. Figs. 10 and 11 provide evidence that the obtained by the MPA procedure. This achievement by the CMP
MPA and CMP procedures are accurate enough in predicting storey procedure is more remarkable for the plastic rotations of the hinges
drift ratios. As can be seen from Figs. 12 and 13, the CMP procedure, in comparison with the MPA procedure.
M. Poursha et al. / Engineering Structures 31 (2009) 591–599 597
(a) Hinge plastic rotations of the 10-storey frame. (b) Hinge plastic rotations of the 15-storey frame.
Fig. 8. Height-wise variation of the hinge plastic rotations for the 10- and 15-storey frames.
(a) Hinge plastic rotations of the 10-storey frame. (b) Hinge plastic rotations of the 15-storey frame
Fig. 9. Errors in the hinge plastic rotations for the 10- and 15-storey frames.
(a) Storey drift ratios of the 20-storey frame. (b) Storey drift ratios of the 30-storey frame.
Fig. 10. Height-wise variation of the storey drifts for the 20- and 30-storey frames.
The locations of the plastic hinges for the 15- and 20-storey been demonstrated that the seismic demands are considerably
frames are shown in Figs. 14 and 15, respectively. They were underestimated in the upper storeys of tall frames through the
obtained by the MPA and CMP procedures, together with NL- FEMA force distributions. It can therefore be concluded that, in
RHA. The results of the MPA procedure, shown in the figures, the case of tall building structures, the CMP procedure provides
were obtained by using the mean spectrum of the selected ground substantially better estimates of seismic response than the FEMA
motions. As shown in the figures, the MPA and CMP procedures force distributions.
are almost able to identify the yielding of beams in the upper
floor levels. The CMP procedure is nearly able to recognize yielding 8. Conclusions
at most of the floors, but it fails to predict yielding at some
locations. It was shown in the previous investigation [25] that To take into account higher-mode effects in pushover analysis
FEMA force distributions [4] are unable to identify the yielding of for estimating the seismic demands of tall building structures, the
beams at the upper floor levels of tall frames, in which higher- consecutive modal pushover (CMP) procedure has been proposed.
mode contributions to the response are important. It has also The procedure employs multi-stage and single-stage pushover
598 M. Poursha et al. / Engineering Structures 31 (2009) 591–599
(a) Storey drift ratios of the 20-storey frame. (b) Storey drift ratios of the 30-storey frame.
Fig. 11. Errors in the storey drifts for the 20- and 30-storey frames.
(a) Hinge plastic rotations of the 20-storey frame. (b) Hinge plastic rotations of the 30-storey frame.
Fig. 12. Height-wise variation of the hinge plastic rotations for the 20- and 30-storey frames.
(a) Hinge plastic rotations of the 20-storey frame. (b) Hinge plastic rotations of the 30-storey frame.
Fig. 13. Errors in the hinge plastic rotations for the 20- and 30-storey frames.
analyses. In the multi-stage pushover analyses, modal pushover significant improvement has been achieved in estimating the hinge
analyses are conducted consecutively with force distributions, plastic rotations through the CMP procedure. The plastic rotations
using mode-shapes derived from the eigen-analysis of the linearly produced by the CMP procedure are substantially better than those
elastic structure. The single-stage pushover analysis is performed obtained by the MPA procedure, especially at the mid and upper
with a triangular or uniform load distribution. The seismic floor levels, when compared to NL-RHA. The improvement in the
demands are then determined by enveloping the peak responses CMP procedure results from the consecutive implementation of
resulting from the multi-stage and single-stage pushover analyses. modal pushover analyses, so that rotations of the plastic hinges
The former analyses control the seismic demands in the mid and are continuously accumulated at the mid and upper floor levels
upper storeys, whereas the latter analysis controls the responses during the modes of interest in the multi-stage pushover analysis.
in the lower storeys of tall buildings. Also, the height-wise distribution of the hinge plastic rotations
Using the CMP procedure, storey drifts can be estimated with produced by the CMP procedure is, in general, more similar to that
acceptable accuracy. At some (upper) storeys, the storey drifts obtained by the benchmark solution (NL-RHA) than that produced
are more accurately estimated by the CMP than by the MPA, by the MPA procedure. The CMP procedure is more accurate than
whereas the MPA yields better estimates of storey drift than the the FEMA load distributions in predicting the seismic demands of
CMP at some other (lower) storeys of the analyzed tall frames. A tall buildings.
M. Poursha et al. / Engineering Structures 31 (2009) 591–599 599