You are on page 1of 28

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/330523315

Response spectrum analysis of frame structures: reliability-based comparison


between complete quadratic combination and damping-adjusted combination

Article  in  Bulletin of Earthquake Engineering · January 2019


DOI: 10.1007/s10518-019-00559-7

CITATION READS

1 117

4 authors:

Francesco Basone Paolo Castaldo


Kore University of Enna Politecnico di Torino
8 PUBLICATIONS   7 CITATIONS    102 PUBLICATIONS   682 CITATIONS   

SEE PROFILE SEE PROFILE

L. Cavaleri Fabio Di Trapani


Università degli Studi di Palermo Politecnico di Torino
109 PUBLICATIONS   1,083 CITATIONS    54 PUBLICATIONS   621 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Strength and Stiffness Predictions of Masonry Infilled Frames View project

STRIT (Strumenti e Tecnologie per la gestione del Rischio delle Infrastrutture) View project

All content following this page was uploaded by Fabio Di Trapani on 23 January 2019.

The user has requested enhancement of the downloaded file.


Bulletin of Earthquake Engineering
https://doi.org/10.1007/s10518-019-00559-7

ORIGINAL RESEARCH

Response spectrum analysis of frame structures:


reliability‑based comparison between complete quadratic
combination and damping‑adjusted combination

Francesco Basone1 · Paolo Castaldo2 · Liborio Cavaleri3 · Fabio Di Trapani2

Received: 29 April 2018 / Accepted: 9 January 2019


© Springer Nature B.V. 2019

Abstract
In the framework of seismic design of structures, response spectrum analysis (RSA) is the
most commonly used approach in practice. The most popular combination rule is the com-
plete quadratic combination (CQC) which is also prescribed by the most of seismic design
codes and is based on the assumptions that the seismic acceleration is a white noise process
and the peak factor ratios associated to the total and modal responses are unitary. Recently,
the damping adjusted combination (DAC) rule has been developed for base-isolated struc-
tures to overcome the aforementioned simplified assumptions. Although it has been proved
that the simplifications about peak factors lead to noticeable errors in the case of base-
isolated structures, the accuracy gain of DAC with respect to CQC in the case of fixed-base
structures is still unknown. Therefore, the paper presents an in-depth study on the RSA of
three-dimensional frame structures, aimed to evaluate the accuracy of the above methods.
Two reference classes of frame structures having different degree of complexity are con-
sidered. Average interstorey drift and floor torsion responses, obtained from a set of Time
History Analyses are compared with those of the modal combination rules. Lognormal
joint probability density functions of the predictive errors from CQC and DAC are finally
evaluated for a reliability assessment of the two combination rules under bidirectional seis-
mic excitations.

Keywords  Complete quadratic combination (CQC) · Damping-adjusted combination


(DAC) · Response spectrum analysis (RSA) · Time history analysis (THA) · Reliability
assessment · Joint PDF

* Francesco Basone
francesco.basone@unikore.it
1
Facoltà di Ingegneria ed Architettura, Università degli Studi di Enna “Kore”, Enna, Italy
2
Dipartimento di Ingegneria Strutturale, Edile e Geotecnica, Politecnico di Torino, Turin, Italy
3
Dipartimento di Ingegneria Civile, Ambientale, Aerospaziale, dei Materiali, University
of Palermo, Palermo, Italy

13
Vol.:(0123456789)
Bulletin of Earthquake Engineering

1 Introduction

Response Spectrum Analysis (RSA) is currently the most popular approach for seismic
design of building structures. Although more sophisticated performance-based techniques
are available in literature (Fajfar 2000; Carvalho et al. 2013; Birely et al. 2012; Cavaleri
et al. 2012; Castaldo et al. 2016; Basone et al. 2017; Cavaleri et al. 2017), the simplicity
of RSA, combined with the use of proper reduction factors and capacity design rules, still
provides a good compromise between safety and practical efficiency. RSA was originally
developed by Rosenblueth (1951) and subsequently revised and modified by a number of
researchers to account the correlation between modal responses (Der Kiureghian 1981;
Rosenblueth and Elorduy 1961) and high-frequency responses (Der Kiureghian and Naka-
mura 1993). RSA has also been used to assess seismic response of buildings equipped with
viscoelastic dampers (e.g. Palmeri 2006).
The simplest combination rule is the Square Root of Sum of Squares (SRSS) (Rosen-
blueth 1951), which neglects the statistical correlation between the different modes of
vibration. Although such a combination is simple, remarkable inaccuracy in assessing
structural response is obtained when modal frequencies are close one to each other. To
overcome this drawback, the Complete Quadratic Combination (CQC) rule (Wilson et al.
1981) was developed. The latter implicitly assumes that: (1) the median of the extreme
values of a generic state variable is almost coincident with its mean value; (2) seismic
acceleration is a white noise process; (3) the peak factor associated with a state variable is
equal to the peak factors of the modal contributes. CQC rule combines the ordinates of a
given response spectrum by using the cross-correlation coefficients of the modal oscillators
(Der Kiureghian 1980, 1981). Acceptable accuracy generally observed for conventional
constructions (Penzien 1997; Cacciola et  al. 2004), along with its simplicity, stimulated
its diffusion. To date, CQC rule is still adopted in seismic codes (Eurocode 8 and NTC
08) as conventional method to perform a response spectrum analysis. However, it has been
recognized that conventional CQC rule becomes ineffective for non-classically damped
structures, and alternative techniques have been proposed (Villaverde 1988; Falsone and
Muscolino 1999) but the larger computational effort limited their application.
In order to overcome this deficiency, Damping Adjusted Combination (DAC) rule
(Muscolino et  al. 2013) has been developed and validated for base-isolated buildings.
Authors have shown that such a combination rule is more versatile and accurate than tradi-
tional ones for structural systems having modal damping ratios different each other and dif-
ferent with respect to the reference one. Furthermore, simplified assumptions of the CQC
rule are removed. Indeed, seismic excitation is modeled as zero-mean stationary Gauss-
ian processes having Power Spectral Density (PSD) function consistent with the target
response spectrum. Specific expressions are given to evaluate modal peak factors as well
as peak factors of all quantities. A strong reduction of the computational demand of the
DAC rule was obtained by the same authors proposing a simplified combination named
DAC-WN assuming the hypothesis that the seismic acceleration is a white noise (WN). In
fact, the authors showed a good agreement between DAC-WN and DAC rules recommend-
ing their use when nonclassical damping in base-isolated building is limited. Although the
effectiveness of such combination was demonstrated, no considerations were done for clas-
sically damped and fixed-base structures.
Recently, Menun et  al. (2015) studied the influence of the peak factor assumptions and
quantified their impact on the structural response of a nine-story symmetric and unsymmet-
ric-plan building computed through the conventional CQC rule. The authors showed that

13
Bulletin of Earthquake Engineering

the errors introduced by this simplified assumption are higher when two or more separated
modal frequencies give significant contributions to the overall response. An overestimation of
the contribution of the higher modes and an underestimation of the contribution of the lower
modes have been highlighted. Two examples of conventional frame structures demonstrated
that the inclusion of the peak factor adjustments can increase conventional SRSS and CQC
predictions of about 17%. Although peak factor adjustment is simplified, it can be used to
evaluate the response of structures whose significant modes of vibration are close to the domi-
nant frequency of the ground motion.
Based on the aforementioned considerations, the accuracy in predicting structural response
of conventional typologies of RC (reinforced concrete) framed buildings computed by the
CQC rule and the most recent DAC-WN rule is focused in this paper, particularly addressing
the different propagation of predictive errors under uniaxial and biaxial seismic excitations in
reliability terms. Indeed, DAC-WN removes the simplified peak factor assumption providing
closed-form expressions to evaluate peak factors of the response, but the gain in terms of accu-
racy in predicting seismic response remains unknown. The influence of structural regularity
on the predictive accuracy of the two methods is also investigated. Similarly to the study car-
ried out by Menun et al. (2015), two different 3D nine-story structural models, having differ-
ent degree of complexity, are analyzed. Furthermore, for each structural configuration, mass
to stiffness ratios are varied to explore a quite wide frequencies range of interest for reinforced
concrete frame structures. For the comparison, the benchmark solution is obtained by aver-
aging fundamental demand parameters (interstorey drifts and relative floor torsions) result-
ing from time history responses with artificial spectrum-compatible accelerograms. Seismic
response under bidirectional seismic excitation is finally investigated in terms of reliability by
defining lognormal bivariate probability density function (Castaldo et al. 2015, 2016) of the
predictive errors on the interstorey drifts between the two main directions.

2 Reference formulation of the dynamic problem

The classic modal formulation of the dynamic problem is here briefly recalled for sake of clar-
ity. With reference to a linear n-degree-of-freedom (n-DOF) classically damped structural sys-
tem, subjected to the horizontal ground acceleration Ü g (t), its dynamic behavior is described
by the following equation of motion:
̇ + Ku(t) = −M𝜏 Ü g (t)
̈ + Cu(t)
Mu(t) (1)
where M, C and K are the mass, damping and stiffness matrices; 𝝉 is the n-dimensional
array listing the influence coefficients of the ground shaking; 𝐮(t),𝐮(t)
̇ and 𝐮̈ (t) are the dis-
placement, velocity and acceleration vectors, respectively.
The linear Rayleigh model is herein adopted for the damping matrix:
C = c 0 M + c1 K (2)
where c0 and c1 are the Rayleigh damping constants having units ­s−1 and s, respectively.
By introducing the modal coordinate transformation, the displacement vector takes the
expression:


m
𝐮(t) = 𝜱q(t) = 𝛟j qj (t) (3)
j=1

13
Bulletin of Earthquake Engineering

in which q(t) is the modal displacement vector collecting the j-th modal displacement
qj(t), Φ is the modal matrix, having order n × m, collecting the m eigenvectors ϕj, nor-
malized with respect to the mass matrix M and resulting as a solution of the following
eigenproblem:

K −1 M𝜱 = 𝜱𝜴−2
(4)
𝜱T M𝜱 = Im

where Ω = diag{ω1…. ω2} is the spectral matrix, that is a diagonal matrix listing the
undamped natural circular frequencies ωi, Im is the identity matrix of order m and the
superscript T denotes the transpose operator. By applying the coordinate transformations to
Eq. (1), pre-multiplying both members by 𝜱T  , the following set of decoupled second order
differential equations is obtained:
̇ + 𝜴2 q(t) = 𝜞 Ü g (t)
q̈ (t) + 𝜦q(t) (5)
̇ and q̈ (t) are the modal velocity and acceleration vectors, 𝜞 = 𝜱 M𝝉 is the
in which q(t) T

vector collecting the modal participation factors and 𝜦 is the generalized damping matrix
given by:

𝜦 = 𝜱T C𝜱 (6)
For classically damped structures, the generalized damping matrix is a diagonal matrix list-
ing the quantities 2 ζi ωi, being ζi the i-th modal damping ratio.

3 Complete quadratic combination (CQC)

Conventional RSA with CQC rule requires the following steps: (1) computing the modal
matrix and the associated spectral matrix by solving the real-values eigenproblem; (2) select-
ing the number m of the vibrational modes to be retained for the structure; (3) defining the
generic structural response of interest y(t) as a combination of the m modal responses, so that:


m
y(t) = ei 𝜃i (t) (7)
i=1

where θi(t)= qi(t)/Γi, ei is the influence coefficient that is the contribute of the i-th mode
shape to the response quantity of interest affected by the i-th participation factor [for exam-
ple ei= ϕri Γi when the r-th structural displacement has to be evaluated, that is y(t) = ur(t)].
The term θi(t) represents the solution of the following differential equation, obtained by
manipulating the i - th of Eq. (5):
𝜃̈i (t) + 2𝜁i 𝜔i 𝜃̇ i (t) + 𝜔2i 𝜃i (t) = Ü g (t) (8)
(4) evaluating the design value of y(t) according to CQC rule:
√ ( ) ( )
√m m
√∑ ∑ Ae Ti , 𝜁i Ae Tk , 𝜁k
Y CQC
=√ 𝜌ik ei ek (9)
i=1 k=1 𝜔2i 𝜔2k

where ρik is the correlation coefficient between i-th and k-th modes of vibration, usually
computed under the assumption that the seismic acceleration is a zero-mean stationary

13
Bulletin of Earthquake Engineering

Gaussian white noise (Wilson et al. 1981) and Ae(Ti,ζi) is the i -th ordinate of the response
spectrum in terms of pseudo-acceleration. This quantity depends in turn on the periods of
vibration Ti= 2π/ωi and on the viscous damping ratios ζi of the m modes of vibration.

4 Damping‑adjusted combination (DAC)

Conventional CQC rule tends to lose in accuracy when viscous damping ratios are different
from the reference one (ζ0 = 0.05) assumed in response spectra of the seismic codes (Muscolino
et al. 2013). For this reason a Damping Correction Factor (DCF) is suggested for different lev-
els of energy dissipation (Muscolino et al. 2013). However, recent studies have also shown the
unreliability of such semi-empirical adjustment (Lin and Chang 2003; Lin et al. 2005).
Damping adjusted combination (Muscolino et al. 2013) was proposed to overcome this
drawback. The design value of y(t) defined by means of DAC rule applies as:
√ ( ) ( )
√m m
√∑ ∑ Di Ae Ti , 𝜁0 Dk Ae Tk , 𝜁0
Y DAC = √ 𝜌ik ei ek 𝜒i(0) 𝜒i(0) (10)
i=1 k=1 𝜔2i 𝜔2k

in which Di is the damping adjustment factor for the i-th mode defined as follows:

𝜆0 ⟨𝜃i (t)⟩
Di = � � � (11)
𝜆0 𝜃i(0) (t)

and 𝜆0 ⟨⋅⟩ is the zero-th spectral moment of the response of the modal oscillator 𝜃i (t) purged
by the participation factors and defined as:

∫ � i � Ug
𝜆0 ⟨𝜃i (t)⟩ = �H (𝜔)�2 G ̈ (𝜔)d𝜔 (12)
0

where GÜ g (𝜔) is the one-sided PSD function of the seismic input and Hi (𝜔) is the complex-
valued Frequency Response Function (FRF) of the i -th modal oscillator evaluated by:
[( 2 ) ( )]−1
Hi (𝜔) = 𝜔i − 𝜔2 + j 2𝜁i 𝜔i 𝜔 (13)

The denominator of Eq. (11) is the Standard Deviation (SD) of the stationary seismic
response of a dummy SDOF oscillator associated with the i-th mode of vibration. The
dummy oscillator has a unit mass, a reference value of the viscous damping ratio ζ0 and an
undamped period of vibration Ti= 2π/ωi. Therefore, its dynamic response is ruled by:

𝜃̈i(0) (t) + 2𝜁0 𝜔i 𝜃̇ i(0) (t) + 𝜔2i 𝜃i(0) (t) = Ü g (t) (14)

The term χ(0)


i in Eq. (10) is a dimensionless correction coefficient defined as:

PF⟨y(t)⟩
𝜒i(0) = � �
(15)
PF 𝜃i(0) (t)

13
Bulletin of Earthquake Engineering

in which PF⟨⋅⟩ is the dimensionless peak factor (PF) of the random process.
Although the formal similarities between the DAC rule in Eq. (10) and the classical CQC
rule in Eq. (9), the main difference consists of the introduction of both the damping adjustment
factors Di and the correction term χ(0)
i . Specifically, the latter coefficient takes into account the
effects of different values in the PFs of the structural response and the damping ratio which
could be different for each dummy oscillator, overcoming one of the main limitations of the
traditional CQC rule. It can be observed that the DAC degenerates into CQC by selecting the
damping adjustment factors and correlation coefficients equal to 1.

4.1 Closed form expressions of damping‑adjusted combination rule (DAC‑WN)

The traditional CQC rule is usually applied under the simplifying assumption that the ground
acceleration can be modeled as a stationary white noise, with energy uniformly distributed
over the frequencies. The same assumption is carried out with the purpose of deriving sim-
ple closed-form expressions for the coefficients introduced in the DAC rule. Therefore, the
quantities Di, ρik and χ(0)
i introduced in Eq. (10) are particularized for Gg(ω)= 1 (unitary white
noise) at each ω and ζ0 (typical value of viscous damping ratio for the response spectra given
in seismic codes). The correlation coefficient between i-th and k-th modes of vibration can be
obtained as for the CQC rule, namely:
8 √ √ � �
𝜌ik =
Ci,k
𝜉i 𝜉k 𝜔i 𝜔k 𝜔i 𝜔k 𝜉i 𝜔i + 𝜉k 𝜔k (16)

where
Ai,k = 𝜔2i + 𝜔2k
Bi,k = 𝜔2i − 𝜔2k ( ) (17)
Ci,k = B2i,k + 4Ai,k 𝜉i 𝜉k 𝜔i 𝜔k + 4 𝜉i2 + 𝜉k2 𝜔2i 𝜔2k

In Eqs. (16) and (17) ζi and ζk are the viscous damping ratios of the i-th and k-th mode,
respectively.
The damping adjustment factor becomes:

𝜉0
Di = (18)
𝜉i

The evaluation of the correction coefficients can be simplified. The denominator in Eq. (15)
is the PF proposed by Vanmarcke (1975) for the i-th dummy oscillator under white noise exci-
tation evaluated as:
⟨ ⟩ √ { [ ( √ )]}
PF 𝜃i(0) (t) = 2 ln 0.04601𝜔i Tg 1 − exp −0.3283 ln(0.4601𝜔i Tg ) (19)

It depends only on the viscous damping ratio ζ0 along with the dimensionless quantity
ωiTg, being Tg the duration of the strong phase of the ground motion (Muscolino et al. 2013).
The numerator in Eq. (15) is the Vanmarcke’s PF for the structural response y(t), evaluated as:
� � � ��
PF⟨y(t)⟩ = 2 ln 2.89N + ⟨y(t)⟩ 1 − exp (−Q⟨y(t)⟩) (20)

13
Bulletin of Earthquake Engineering

where the quantities N + ⟨y(t)⟩ and Q⟨y(t)⟩ are so defined:



1 𝜆2 ⟨y(t)⟩
N ⟨y(t)⟩ =
+
T
2𝜋 𝜆0 ⟨y(t)⟩ g (21)

Q⟨y(t)⟩ = 1.77q⟨y(t)⟩1.20 ln (2.89N + ⟨y(t)⟩)

in which q⟨y(t)⟩ is the dimensionless bandwidth parameter, given by:



𝜆1 ⟨y(t)⟩2
q⟨y(t)⟩ = 1 − (22)
𝜆0 ⟨y(t)⟩𝜆2 ⟨y(t)⟩

The first three spectral moments of the i-th dummy oscillator introduced in Eqs.  (21)
and (22) are given respectively as (Der Kiureghian 1980):

� m m
� ei e k � �
𝜆0 ⟨y(t)⟩ = 2𝜋 𝜁i 𝜔i + 𝜁k 𝜔k
i=1 k=1
Cik
� � ��
�m m
� ei ek � � � � 𝜔i
𝜆 1 ⟨y(t)⟩ = 𝜁i Aik + 2𝜁k 𝜔i 𝜔k Zi + 𝜁k Aik + 2𝜁i 𝜔i 𝜔k Zk − Bik ln
i=1 k=1
C ik 𝜔k

� � ei e k
m m
� �
𝜆2 ⟨y(t)⟩ = 2𝜋 𝜔i 𝜔k 𝜁i 𝜔k + 𝜁k 𝜔i
i=1 k=1
Cik
(23)

where

⎛ ⎞
1 − 𝜁i2 ⎟
1 −1 ⎜
Zi = � tan ⎜ ⎟ (24)
⎜ 𝜁i ⎟
1 − 𝜁i2
⎝ ⎠

Equations (17)–(24) provide closed expressions of all quantities which significantly reduce


the computational effort required.

5 Comparison between CQC and DAC‑WN rules in estimating seismic


demand of RC structures under unidirectional seismic excitations

5.1 Reference structural models

Seismic response of two reference nine-story RC structures has been investigated. The two
structural models selected are characterized by different geometry in plan and elevation.
The reference structural models, named M1 and M2 respectively, have been selected in
order to assess the reliability of CQC and DAC-WN rules in estimating seismic response
of buildings characterized by a different rate of distribution of dynamic properties over the
modes due to the different structural regularity. A 3D view of the both buildings is shown
in Fig. 1, while geometric details of the floor plans are reported in Table 1.

13
Bulletin of Earthquake Engineering

Fig. 1  3D view of the buildings and position of the center of mass of each floor: a M1 model; b M2 model

For each structural model, stiffness of columns is varied in order to consider a suffi-
ciently large range of dominant frequencies (3–30 rad/s) of interest for reinforced concrete
frame structures. Hence, 2 × 11 databases of structural models are defined. The damping
ratio of the buildings is assumed constant for each mode and equal to 5%. The variation
of modal frequencies within each database results in a modification of the contribution
of each mode to the overall response without a modification in the correlation between
the modes. However, the modal peak factor ratios, evaluated when DAC-WN is used,
change, allowing a comparison of predictions by CQC and DAC-WN rules with respect to
the benchmark solution. An overview of the modal properties of each model is shown in
Table 2 for the first five modes.

5.2 Definition of unidirectional RSA with CQC, of unidirectional RSA with DAC‑WN


and of time history analyses

Response spectrum analysis is carried out for the models using the original formulation of
CQC rule, and the simplified DAC rule under white noise input (DAC-WN) as defined in
the previous section. Direct integration time history analysis (THA) is also carried out as
benchmark solution, using 30 sampled ground motion signals, which are generated to be
spectrum compatible with the reference response spectrum and to represent the record-
to-record variability (Castaldo et al. 2017). The structures are supposed to have rigid floor
diaphragms, hence 3 × 9 (27) degrees of freedom are considered.
Artificial accelerograms used for THA have been generated using the expression
by Shinozuka and Sato (1967) based on a spectrum-compatible PSD function. The
procedure proposed by Cacciola (2010) has been used to evaluate the PSD function,

13
Bulletin of Earthquake Engineering

Table 1  Geometric details of floor plans of the reference structures


Reference structural
Plan configurations and dimensions [m]
model
All floor plans

M1

Floor-plan #1 Floor-plan #2 Floor-plan #3

Floor-plan #4 Floor-plan #5 Floor-plan #6

M2

Floor-plan #7 Floor-plan #8 Floor-plan #9

verifying that generated accelerograms complied with EC8 rules, which provide that
mean response spectrum ordinates are not less than 90% of the corresponding values
of the 5% damping elastic response spectrum and that the mean zero period spectral
ordinate is not less than agS, where ag is the peak ground acceleration and S the soil fac-
tor. The target spectrum used for the generation of the sets of signals has been defined
assuming a return period ­TR = 475 years with ag = 0.237 g. The spectra of the 30 gener-
ated ground motions are shown in Fig. 2, together with mean and target spectra.
Figure 3 shows the PSD function of the target spectrum and the dominant frequen-
cies of each structure herein selected. It can be clearly observed that dominant mode
frequencies significantly change for each considered structure, in such a way to cover a
sufficiently wide frequency range of interest for reinforced concrete structures.

13
Bulletin of Earthquake Engineering

Table 2  Modal participating mass ratios and frequencies of structures for the first five modes
Structures Mode M1 models M2 models
mi,x (%) mi,y (%) ωi (rad/s) mi,x (%) mi,y (%) ωi (rad/s)

S1 1 0.003 80.26 3.18 9.98 61.76 3.30


2 79.58 0.001 3.21 61.45 10.01 3.34
3 0.648 0.001 3.43 0.17 0.050 3.94
4 0.005 11.53 9.01 5.82 7.853 8.38
5 11.49 0.001 9.03 8.05 5.822 8.39
S2 1 0.004 80.32 3.92 9.97 61.78 3.96
2 79.55 0.001 3.96 61.48 9.99 4.01
3 0.629 0.001 4.23 0.16 0.05 4.73
4 0.003 11.56 11.10 7.32 6.36 10.07
5 11.53 0.001 11.14 6.54 7.30 10.08
S3 1 0.004 80.20 5.54 9.98 61.75 5.54
2 79.55 0.001 5.60 61.45 10.00 5.61
3 0.629 0.001 5.98 0.17 0.051 6.61
4 0.003 11.56 15.70 5.43 8.24 14.08
5 11.53 0.001 15.75 8.45 5.43 14.09
S4 1 0.004 0.802 7.84 9.99 61.73 8.90
2 79.55 0.000 7.92 61.42 10.01 9.00
3 0.629 0.000 8.46 0.17 0.051 10.62
4 0.003 0.116 22.21 4.17 9.49 22.62
5 11.53 0.000 22.27 9.71 4.19 22.64
S5 1 0.004 80.18 10.59 9.97 61.78 10.44
2 79.53 0.001 10.71 61.47 9.99 10.56
3 0.621 0.001 11.42 0.16 0.051 12.47
4 0.003 11.58 30.00 7.07 6.61 26.54
5 11.55 0.001 30.10 6.80 7.06 26.56
S6 1 0.002 80.51 13.24 9.98 61.76 12.97
2 79.77 0.001 13.37 61.46 10.00 13.12
3 0.730 0.001 14.39 0.17 0.051 15.49
4 11.32 0.001 37.69 6.05 7.63 32.97
5 0.013 11.41 37.71 7.82 6.05 33.00
S7 1 0.004 80.10 15.23 9.99 61.73 15.08
2 79.50 0.001 15.39 61.42 10.01 15.25
3 0.604 0.001 16.40 0.18 0.051 17.99
4 0.003 11.63 43.09 4.01 9.65 38.32
5 11.58 0.001 43.25 9.87 4.03 38.35
S8 1 0.004 80.08 17.29 9.99 61.70 17.15
2 79.47 0.001 17.48 61.38 10.02 17.35
3 0.586 0.001 18.60 0.18 0.052 20.45
4 0.002 11.64 48.90 3.00 10.66 43.59
5 11.61 0.001 49.10 10.89 3.02 43.63

13
Bulletin of Earthquake Engineering

Table 2  (continued)
Structures Mode M1 models M2 models
mi,x (%) mi,y (%) ωi (rad/s) mi,x (%) mi,y (%) ωi (rad/s)

S9 1 0.004 80.13 21.04 9.98 61.76 20.51


2 79.52 0.001 21.26 61.46 10.00 20.75
3 0.614 0.001 22.67 0.17 0.052 24.49
4 0.003 11.61 59.55 6.05 7.63 52.13
5 11.56 0.001 59.75 7.82 6.05 52.17
S10 1 0.004 80.21 25.99 9.98 61.76 25.23
2 79.55 0.001 26.26 61.45 10.00 25.52
3 0.631 0.001 28.04 0.17 0.052 30.12
4 0.004 11.56 73.64 5.66 8.02 64.12
5 11.53 0.001 73.85 8.22 5.66 64.17
S11 1 0.005 80.26 30.38 9.97 61.77 30.51
2 79.40 0.001 30.71 61.47 10.00 30.87
3 0.525 0.001 32.56 0.16 0.053 36.44
4 0.002 11.53 85.70 6.86 6.83 77.54
5 11.71 0.001 86.17 7.01 6.84 77.60

Fig. 2  Target spectrum and gen-


erated ground motion spectra

Structural responses of buildings defined above were obtained considering the earth-
quake acting along X-direction. Seismic demand monitored parameters were interstorey
drifts at every floor in X direction and relative floor torsions resulting by CQC, DAC-
WN and THA. Interstorey drifts were evaluated referring to the centers of mass at each
floor. Torsional effects were evaluated normalizing the relative torsion angle (ΔΘ) of
floors by the interstorey height (h). The latter are important because high values of tor-
sion produce non-negligible increases of stress on the structure. The benchmark values
for each structural model were obtained by averaging THA results obtained using as
input the before mentioned 30 ground motions.

13
Bulletin of Earthquake Engineering

Fig. 3  Target spectrums PSD and fundamental natural frequencies of the considered structures

5.3 Interstorey drifts along X direction of M1 and M2 models

Interstorey drifts were evaluated along X direction by THA, CQC and DAC-WN. The drift
profiles obtained for the regular structural configurations (M1) at the different fundamental
frequencies are shown together in Fig.  4. Percentage variations of RSA predictions with
respect to THA average values can be also observed in Fig. 5. Both CQC and DAC-WN
provide a good estimation of the translational response with an average error of about 10%
and the maximum predictive errors do not exceed 15%. The larger errors are generally rec-
ognized at the top floors and for the structural configurations presenting the lowest funda-
mental frequencies.
Figures 6 and 7 show the same diagrams for the irregular models (M2) highlighting
that the predictions by DAC-WN present the same level of accuracy as for the regular

Fig. 4  Interstorey drifts computed by THA, RSA with CQC and RSA with DAC-WN for M1 model along
X-direction

13
Bulletin of Earthquake Engineering

Fig. 5  Percentage variations of RSAs drift predictions with respect to THA average values for M1 model
along X-direction

Fig. 6  Interstorey drifts computed by THA, RSA with CQC and RSA with DAC-WN for M2 model along
X-direction

building. On the contrary, higher inaccuracy is recognized observing results by CQC


response spectrum analyses, which show underestimations even larger than − 30%. Also
in this case, the predictive error grows when the deformability of structures increases,
while adequate accuracy of CQC is recognized for stiffer buildings.

13
Bulletin of Earthquake Engineering

Fig. 7  Percentage variations of RSAs drift predictions with respect to THA average values for M2 model
along X-direction

5.4 Torsional response of M1 and M2 models

Relative floor torsion profiles evaluated by THA and RSA for M1 models are shown in
Fig. 8. It can be generally observed that results by CQC significantly overestimate the
average response obtained by time history analyses. The average percentage overestima-
tion (Fig. 9) is about 50%. Results by DAC-WN show instead an overall underestimation

Fig. 8  Relative floor torsions computed by THA, RSA with CQC and RSA with DAC-WN for M1 model

13
Bulletin of Earthquake Engineering

Fig. 9  Percentage variations of RSAs relative floor torsion predictions with respect to THA average values
for M1 model

of the torsional response. However, the magnitude of the average percentage error
(− 25%) is significantly lower.
Torsional responses of irregular models (Figs.  10, 11) present significant errors by
both CQC and DAC-WN only in correspondence of the last two floors, where buildings
present a relevant change of torsional stiffness. Even in this case, predictive errors by
CQC are larger than those recorded by DAC-WN of about two times.

Fig. 10  Relative floor torsions computed by THA, RSA with CQC and RSA with DAC-WN for M2 model

13
Bulletin of Earthquake Engineering

Fig. 11  Percentage variations of RSAs relative floor torsion predictions with respect to THA average values
for M2 model

5.5 Overall considerations

Percentage errors with respect to THA averages are shown in Figs. 12 and 13 for the most
significant demand parameters. In detail, first, top and maximum interstorey drifts and
interstorey relative torsions are considered. Percentage errors are reported as a function of
the fundamental frequencies of structures. The predictive accuracy of the two combination
rules has not shown a clear dependence on fundamental frequencies. However, estimation

Fig. 12  Percentage errors of RSA predictions with respect to THA averages at the different fundamental
frequencies of M1 model

13
Bulletin of Earthquake Engineering

Fig. 13  Percentage errors of RSA predictions with respect to THA averages at the different fundamental
frequencies of M2 model

errors have generally reduced for stiffer structures. On the whole, results show that the
translational response of regular buildings subjected to unidirectional seismic excitation
is estimated with sufficient accuracy by both CQC and DAC-WN rules. CQC resulted less
accurate in predicting interstorey drifts of irregular buildings when these are character-
ized by large deformability. Even torsional response is more accurately estimated by using
DAC-WN, however for the regular buildings, absolute values of relative torsions have low
magnitude and may result not relevant to the overall response. On the contrary, relative tor-
sions values are not negligible for the irregular buildings, hence significant overestimations
or underestimations can be expected by adopting CQC rule.

6 Reliability analysis of CQC and DAC‑WN in estimating seismic


demand of RC structures under bidirectional seismic excitation

Previous results have shown that translational and torsional responses of buildings sub-
jected to unidirectional seismic excitation may be affected by non-negligible errors when
estimated by RSA. For bidirectional ground motions, errors computed in both directions
unavoidably cumulate and, therefore the accuracy of CQC and DAC-WN in predicting
bidirectional response building structures is here tested by means of a probabilistic inves-
tigation, aimed at assessing the probability that predictive error falls within a specified
tolerance interval. Time history analyses are carried out using the same ground motion
set before described, considering each earthquake acting 100% along X direction and 30%
along Y direction for both M1 and M2 structures. Interstorey drifts along X and Y direc-
tions are evaluated at each floor in correspondence of two monitoring points, namely the
center of mass (CM) and the corner column (CC) at the top right (Fig. 14). The same drifts
are then estimated with the response spectrum analysis by CQC and DAC-WN. The choice
to evaluate the response at CM and CC depends on the different magnitude by torsional
demand affecting the displacements of these points.

13
Bulletin of Earthquake Engineering

Fig. 14  Scheme of application of
bidirectional seismic excitation
and monitoring points

For each structural model at every fundamental frequency and for each monitoring
point, the ratios between RSA drift predictions by CQC and DAC-WN along X and Y
THA THA
(ΔCQC
X
, ΔCQC
Y
, ΔDAC−WN
X
, ΔDAC−WN
Y
) and time history averages (ΔX , ΔY ) are evaluated
as:

ΔCQC ΔCQC
εCQC
X
= X
THA
; εCQC
Y
= Y
THA (25)
ΔX ΔY

DAC−WN
ΔDAC−WN
X DAC−WN
ΔDAC−WN
Y
εX = THA
; εY = THA (26)
ΔX ΔY

The ratios above defined provide a more effective representation of predictive errors,
intrinsically indicating overestimations and underestimation whether the ratio is greater
than or less than 1.
Predictive errors so defined are shown in Figs. 15 and 16 as a function of the domi-
nant frequencies of structures. For the regular structural models it can be observed that
the accuracy of estimations has a certain dependence with the position of the monitor-
ing point. Estimations at the corner column are, in fact, less accurate because of the
more strict dependence on the torsional response. This trend cannot be recognized for
the irregular configurations, for which larger errors with respect to M1 models are rec-
ognized independently from the monitoring position.
By considering predictive errors εX and εY of models M1 and M2 as random vari-
ables, it is possible to evaluate lognormal joint probability density functions (Castaldo
et  al. 2015, 2016) of the errors for CQC and DAW-WN at the different monitoring
points. Such functions allow a better understanding of the reliability of the structural
response obtained with both RSAs respect to the one obtained by THAs. In particular,
the reliability of both combination rules defined as the probability of error and as prob-
ability of overestimating or underestimating of the structural response was evaluated,

13
Bulletin of Earthquake Engineering

(a) (b)

(c) (d)
Fig. 15  Interstorey drifts predictive errors of for M1 Model at: a center of mass (X dir.); b center of mass
(Y dir.); c corner column (X dir.); d corner column (Y dir.)

for fixed tolerance intervals. Each lognormal bivariate joint probability density function
(j-PDF) is evaluated by means of the following expression:

1
f(εX , εY ) = √
2πεX εY σln(εX ) σln(εY ) 1 − ρ2
⎧ � �
⎡ ln(εX )−μln(ε ) 2 ⎤⎫
⎪ ⎢ X
+ ⎥⎪ (27)
⎪ 1 ⎢ σln(εX )
⎥⎪
× exp ⎨− � �� � � � ⎬
⎪ 2(1 − ρ) ⎢⎢ −2ρ ln(εX )−μln(εX ) ⎥⎪
2 2
ln(εY )−μln(εY ) ln(εY )−μln(εY )

+ ⎥⎪
⎩ ⎣ σln(εX ) σ ln(εY ) σ ln(εY ) ⎦⎭

in which, μln(εX ) , μln(εY ),σln(εX ) and σln(εY ) are the means and the standard deviations of the
natural logarithms of the random variables εX and εY , while ρ is the correlation coefficient.
The jPDF functions obtained for M1 and M2 models are shown in Figs. 17, 18, 19, 20
and the corresponding contour lines with the different tolerance intervals are illustrated in
Figs. 21, 22.
Generally, the regular structural model shows similar results for both CQC and DAC-
WN rules. Furthermore, it can be observed that all jPDFs do not exceed 25% error in
both X and Y directions. Different considerations can be expressed for the irregular

13
Bulletin of Earthquake Engineering

(a) (b)

(c) (d)
Fig. 16  Interstorey drifts predictive errors of for M2 Model at: a center of mass (X dir.); b center of mass
(Y dir.); c corner column (X dir.); d corner column (Y dir.)

Fig. 17  jPDF of interstorey drifts predictive errors for M1 model at the center of mass: a DAC-WN; b CQC

13
Bulletin of Earthquake Engineering

Fig. 18  jPDF of interstorey drifts predictive errors for M1 model at the center corner column: a DAC-WN;
b CQC

Fig. 19  jPDF of interstorey drifts predictive errors for M2 model at the center of mass: a DAC-WN; b CQC

structural models. jPDFs obtained with the CQC rule show a greater dispersion of the
errors respect to one obtained with the DAC-WN rule. Figure  22 clearly shows that
CQC rule leads errors up to 40% unlike the DAC-WN rule whose errors do not exceed
20%. Such trend highlighted that high inaccuracies can apply with the CQC rule for
structures dominated by high vibration modes. On the contrary, DAC-WN rule shows
good results for both regular and irregular structures.
The accuracy of CQC and DAC-WN rules in estimating bidirectional seismic
response is then assessed in probabilistic terms. By considering a number of pris-
matic percentage tolerance intervals (τ) with increasing amplitude centered in plan at
the point of coordinate [1,1] (THA average solution) (Figs. 21, 22), the probability that
the predictive error (expressed as percentage) falls within a specified tolerance interval

13
Bulletin of Earthquake Engineering

Fig. 20  jPDF of interstorey drifts predictive errors for M2 model at the center corner column: a DAC-WN;
b CQC

Fig. 21  jPDF of interstorey drifts predictive errors for M1 model by DAC-WN and CQC at: a center of
mass; b corner column

( P(ε < τ) ) can be evaluated by calculating the volume below of the jPDF and delimited
by the tolerance prism. Such a probability obviously grows with enlarging the tolerance
interval. Therefore, a greater reliability of the structural response is obtained when a
higher probability with a small tolerance is achieved.
A clear comparison is shown in Fig. 23a–d, where probability curves are reported for
DAC-WN and CQC.
From Fig. 21 and the associated diagrams in Fig. 23a, b, it can be observed that for
the regular structural model (M1), DAC-WN and CQC rules have approximately the
same reliability. The 90% probability that predictive error is lower than the tolerance is
on average achieved with similar tolerance values at the center of mass (10% for CQC

13
Bulletin of Earthquake Engineering

Fig. 22  jPDF of interstorey drifts predictive errors for M2 model by DAC-WN and CQC at: a center of
mass; b corner column

(a) (b)

(c) (d)
Fig. 23  Probability that predictive errors fall within the tolerance interval by DAC-WN and CQC: a M1
model at center of mass; b M1model at corner column; c M2 model at center of mass; d M2 model at cor-
ner column

13
Bulletin of Earthquake Engineering

and 11% for DAC-WN) and at the external node (17.5% for CQC and 18.5% for DAC-
WN), confirming a good accuracy of both RSA methods.
For the irregular model (M2) (Figs. 22, 23b, c), similar reliability is shown by DAC-
WN with respect to the regular case, while a significantly larger dispersion of the results
is observed from CQC jPDFs. In fact, the 90% probability that predictive error is lower
than the tolerance is achieved with a tolerance of 12.5% by DAC-WN for both CM
and CC points. The same tolerance interval provided a probability of about 45% when
CQC rule is adopted with a 90% probability achieved for a tolerance of 30%. Although
DAC-WN rule is based on the simplified assumption of white noise input processes, the
stability shown with respect to the complexity of structural models allows suggesting
the method gives more reliable predictions for the cases of irregular structures, with
respect to those obtainable by CQC. The estimation of the modal peak factors of each
considered vibration mode makes it possible to increase the accuracy of the seismic
response of buildings whose response is also governed by lower vibration modes. On
the contrary, the simplified assumptions of the CQC rule lead to larger inaccuracies in
the response of irregular structures compromising its reliability.
The previous analyses provided information on the magnitude of the predictive error
in absolute value, confirming the stability of DAC-WN rule accuracy level even for irreg-
ular structural configurations. A further investigation has been carried out to assess the
probability of overestimating and underestimating more than a specific percentage value
by DAC-WN and CQC rules. Since the RSA represents a less onerous alternative to the
THA for the design and verification of the structures, it is important that the RSA does not
excessively underestimate structural responses. By defining the generic percentage overes-
timation and underestimation limits with the symbols ω+ and ω− respectively, it is possible
to consider a number of prisms centered at [0,0], having side 1 + ω+/100 or 1 − ω−/100.
The probability that the overestimation error (ε+) is larger than ω+ [ P(ε+ > ω+ ) ] is calcu-
lated as the volume below jPDF function external to the prism of side 1 + ω+/100. At the
same time, the probability that the underestimation error (ε−) is larger (in absolute value)
than ω− is calculated as the volume below the jPDF function internal to the prism of side
1 − ω−/100. Such a probability is expressed as P(ε− > ω− ) to formally consider the negative
sign of the errors. A summary of results is illustrated in Fig. 24a–d. Seismic response is
more frequently underestimated by both DAC-WN and CQC. However, results by DAC-
WN are more balanced (60% of underestimation to 40% of overestimation) with respect to
CQC, which has shown probabilities of underestimation even achieving 80%.
For the regular structural model (M1), underestimations and overestimations by both
DAC-WN and CQC did not exceed ± 20%. This trend was also confirmed for M2 model
predictions by DAC-WN, while results by CQC have shown underestimations even larger
than 30%, although associated with low probabilities.
General trend shows that DAC-WN leads to underestimate the seismic response less
than CQC rule. Such consideration is more evident for the irregular structural model
respect to the regular one. As in the previous case, the reliability of DAC-WN is proved to
be better than CQC rule for irregular models. Indeed, the simplified assumption of CQC
rule does not excessively affect the seismic response of regular models but it leads to larger
inaccuracies when the contributions of lower vibration modes increase.

13
Bulletin of Earthquake Engineering

(a) (b)

(c) (d)
Fig. 24  Probabilities of overestimation and underestimation more than a specified limit (ω) by DAC-WN
and CQC: a M1 model at center of mass; b M1 model at corner columns; c M2 model at center of mass; d
M2 model at corner column

7 Conclusions

Response spectrum analysis is commonly employed as a practice method to assess


design seismic response of structures. CQC rule currently represents the standard RSA
method. The latter is based on the main simplified assumption that peak factors associ-
ated with the response quantities of interest are equal to the peak factors of the con-
tributing modal responses. The DAC rule and the simplified DAC-WN remove this
limitation with a good computational efficiency. The paper has presented an in-depth
comparison of the two RSA approaches (CQC and DAC-WN) in reliability terms aimed
at assessing their accuracy in estimating seismic response of buildings with different
degree of complexity. Two main structural configurations have been defined: M1 (regu-
lar) and M2 (irregular). Mass/stiffness ratios of models were varied within the same
structural configurations. The most significant demand parameters have been evaluated
by DAC-WN and CQC under unidirectional and bidirectional seismic excitations. Out-
comes have been compared with average values by time history analyses carried out
with 30 spectrum compatible ground motions. Based on the obtained results the follow-
ing conclusions can be drawn:

13
Bulletin of Earthquake Engineering

• CQC and DAC-WN provided a similar good reliability in estimating translational


response of M1 model (regular). Average error was about 10%. DAC-WN presented the
same accuracy even for M2 model (irregular), while higher inaccuracy was highlighted
by CQC results for the most deformable building configurations, where underestima-
tions were even larger than − 30%;
• torsional response was estimated more accurately by DAC-WN for both M1 and M2
models, although the absolute values of relative torsions obtained by the regular build-
ings have low magnitude and may result not relevant to the overall response. Significant
overestimations or underestimations can be instead expected by CQC rule applied to
irregular configurations;
• for the case of bidirectional seismic excitation of M1 model, lognormal bivariate prob-
ability density functions of predictive errors have confirmed that DAC-WN and CQC
rules have equivalent reliability for regular structural models with a 90% probability
to have percentage errors lower than 11% for the displacements at the center of mass
(CM) and 18% at the external node (CC). M2 model response is estimated with the
same level of accuracy as M1 model by DAC-WN, while noticeable inaccuracy is
obtained by CQC with a 90% probability of not exciding a tolerance of 30%;
• seismic demand is generally underestimated by both RSA methods. Acceptable under-
estimation percentages were observed from results by M1 model. M2 model response
was instead significantly underestimated by CQC.

By summarizing results, it can be concluded that seismic demand of simple and regular
structural configurations can be effectively estimated by CQC. For these cases, in fact, the
adoption of DAC or DAC-WN would increase computational demand without a significant
improvement of the accuracy. On the contrary, CQC rule applied to irregular configura-
tions characterized by noticeable torsional demand may lead to significant inaccuracies and
underestimations and the use of DAC or DAC-WN can be recommended as a valid oppor-
tunity. Results obtained are extrapolated from the investigated database. Further reliability
tests on different structural configurations and structural types would be desirable.

References
Basone F, Cavaleri L, Muscolino G, Di Trapani F (2017) Incremental dynamic based fragility assessment
of reinforced concrete structures: stationary versus non-stationary artificial ground motions. Soil Dyn
Earthq Eng 103:105–117
Birely AC, Lowes L, Lehman D (2012) A model for the practical nonlinear analysis of reinforced-concrete
frames including joint flexibility. Eng Struct 130:455–465
Cacciola P (2010) A stochastic approach for generating spectrum compatible fully nonstationary earth-
quakes. Comput Struct 88(15–16):889–901
Cacciola P, Colajanni P, Muscolino G (2004) Combination of modal responses consistent with seismic input
representation. J Struct Eng (ASCE) 130(1):47–55
Carvalho G, Bento R, Bhatt C (2013) Nonlinear static and dynamics analyses of reinforced concrete build-
ings-comparison of different modelling approaches. Earthq Struct 4(5):451–470
Castaldo P, Palazzo B, Della Vecchia P (2015) Seismic reliability of base-isolated structures with friction
pendulum bearings. Eng Struct 95:80–93
Castaldo P, Palazzo B, Della Vecchia P (2016) Life cycle-cost and seismic reliability analysis of 3D systems
equipped with FPS for different isolation degrees. Eng Struct 125:349–363
Castaldo P, Ferrentino T, Palazzo B (2017) Seismic reliability-based ductility demand evaluation for inelas-
tic base-isolated structures with friction pendulum devices. Earthq Eng Struct Dyn 46(8):1245–1266

13
Bulletin of Earthquake Engineering

Cavaleri L, Di Trapani F, Macaluso G, Papia M (2012) Reliability of code proposed models for assessment
of masonry elastic moduli. Ing Sismica 29(1):38–59
Cavaleri L, Di Trapani F, Asteris PG, Sarhosis V (2017) Influence of column shear failure on pushover
based assessment of masonry infilled reinforced concrete framed structures: a case study. Soil Dyn
Earthq Eng 100:98–112
CEN, European Committee for Standardisation. Eurocode 8: design provisions for earthquake resistance of
structures, part 1.1: general rules, seismic actions and rules for buildings, prEN 1998-1; 2003
CS.LL.PP. DM 14 Gennaio, Norme tecniche per le costruzioni. Gazzetta Ufficiale della Repubblica Italiana;
2008
Der Kiureghian A (1980) Structural response to stationary excitation. J Eng Mech Div 106(6):1195–1213
Der Kiureghian A (1981) A response spectrum method for random vibration analysis of MDF systems.
Earthq Eng Struct Dyn 9(5):419–435
Der Kiureghian A, Nakamura Y (1993) CQC modal combination rule for high-frequency modes. Earthq
Eng Struct Dyn 22(11):943–956
Fajfar P (2000) A nonlinear analysis method for performance-based seismic design. Earthq Spectra
16:573–592
Falsone G, Muscolino G (1999) Cross-correlation coefficients and modal combination rules for non-classi-
cally damped systems. Earthq Eng Struct Dyn 28(12):1669–1684
Lin YY, Chang KC (2003) A study on damping reduction factor for buildings under earthquake ground
motions. J Struct Eng (ASCE) 129(2):206–214
Lin YY, Miranda E, Chang KC (2005) Evaluation of damping reduction factors for estimating elastic
response of structures with high damping. Earthq Eng Struct Dyn 34(11):1427–1443
Menun C, Reyes JC, Chopra AK (2015) Errors caused by peak factor assumptions in response spectrum
based analyses. Earthq Eng Struct Dyn 44(11):1729–1746
Muscolino G, Palmeri A, Versaci C (2013) Damping-adjusted combination rule for the response spectrum
analysis of base-isolated buildings. Earthq Eng Struct Dyn 42(2):163–182
Palmeri A (2006) Correlation coefficients for structures with viscoelastic dampers. Eng Struct
28(8):1197–1208
Penzien J (1997) Evaluation of building separation distance required to prevent pounding during strong
earthquakes. Earthq Eng Struct Dyn 26(8):849–858
Rosenblueth E (1951) A basis for aseismic design. Ph.D. thesis. University of Illinois: Urbana, IL
Rosenblueth E, Elorduy J (1961) Responses of linear systems to certain transient disturbances. In: Proceed-
ings of the 4th world conference on earthquake engineering, Santiago, Chile
Shinozuka M, Sato Y (1967) Simulation of nonstationary random process. J Eng Mech 93(1):11–40
Vanmarcke EH (1975) On the distribution of the first-passage time for normal stationary random processes.
J Appl Mech 42(1):215–220
Villaverde R (1988) Rosenblueth’s modal combination rule for systems with non-classical damping. Earthq
Eng Struct Dyn 16(3):931–942
Wilson EL, Der Kiureghian A, Bayo EP (1981) A replacement for the SRSS method for seismic analysis.
Earthq Eng Struct Dyn 9(2):187–192

Publisher’s Note  Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.

13
View publication stats

You might also like