Response Spectrum Analysis of Structures Subjected To Seismic Actions

Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

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

net/publication/278661406

Response Spectrum Analysis of Structures Subjected to Seismic Actions

Chapter · January 2013


DOI: 10.1007/978-3-642-36197-5_133-1

CITATION READS

1 6,499

1 author:

Michalis Fragiadakis
National Technical University of Athens
121 PUBLICATIONS   1,496 CITATIONS   

SEE PROFILE

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

WG 13: SEISMIC ASSESSMENT, DESIGN AND RESILIENCE OF INDUSTRIAL FACILITIES - EUROPEAN ASSOCIATION OF EARTHQUAKE ENGINEERING View project

RASOR: Risk Assessment for the Seismic protection Of industRial facilities View project

All content following this page was uploaded by Michalis Fragiadakis on 31 August 2015.

The user has requested enhancement of the downloaded file.


Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

Response Spectrum Analysis of Structures Subjected to Seismic


Actions
Michalis Fragiadakis*
Laboratory for Earthquake Engineering, National Technical University of Athens (N.T.U.A.), Athens, Greece

Synonyms
Lateral force method; Modal analysis; Modal response spectrum analysis; Modal superposition;
Response spectrum; Seismic performance assessment

Introduction
Response spectrum analysis (RSA) is a method widely used for the design of buildings. Concep-
tually the method is a simplification of modal analysis, i.e., response history (or time history)
analysis (RHA) using modal decomposition, that benefits from the properties of the response
spectrum concept. The purpose of the method is to provide quick estimates of the peak response
without the need to carry out response history analysis. This is very important because response
spectrum analysis (RSA) is based on a series of quick and simple calculations, while time history
analysis requires the solution of the differential equation of motion over time. Despite its approx-
imate nature, the method is very useful since it allows the use of response spectrum, a very
convenient way to describe seismic hazard.
RSA is very appealing to practicing engineers because seismic loading is defined by means of
a response spectrum. All design codes worldwide define seismic input (or hazard) by means of
a code-compliant, typically smooth, response spectrum that can be easily adjusted according to the
site seismic hazard. Such spectra are able to implicitly take into consideration the fact that structures
are designed to resist seismic actions by deforming inelastically. Two variations of RSA are offered
in almost every seismic design code. Following the notation of Eurocode 8 (EC8 2004), these
variations are the “lateral force method” and the “modal response spectrum analysis.” This entry
discusses the underlying concepts of both methods in a comprehensive manner. The interested
reader is also advised to consult the classic textbooks of Chopra (2000) and Clough and
Penzien (1975).

Modal Analysis
Before proceeding to the discussion of response spectrum analysis, it is essential to have a good
understanding of modal analysis, also known as modal superposition method. Modal analysis aims
at transforming a fully coupled problem of N degrees of freedom to N uncoupled single-degree-of-
freedom (SDOF) problems that can be solved individually, and, in the end, the individual solutions
can be superimposed to obtain the solution of the initially coupled problem. The advantage is that
simpler methods, including closed-form methods, can be used to solve the single-degree-of-freedom

*Email: mfrag@mail.ntua.gr

Page 1 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

problem. From a computing standpoint, the number of operations required is substantially reduced
when modal decomposition is preferred (Bathe 1996). A second reason is that only a small number
modes need to be taken into consideration, while the rest can be neglected with a minimum loss of
accuracy. This considerably speeds up the whole process, while the number of modes required
depends on the type of loading and the structural properties (i.e., stiffness, damping).
The equation of motion of a system with N degrees of freedom is written as

mü þ cu_ þ ku ¼ qðt Þ (1)

where m, c, and k are the mass, damping, and stiffness matrices, respectively, while u €, u_ , and
u denote the vectors of acceleration, velocity, and displacement and are always functions of time.
q(t) is the vector of applied loads, again defined as function of time. If the dynamic load is a ground
motion history (acceleration time history), then q(t) is obtained as

ug ðt Þ
qðtÞ ¼ mB€ (2)

where B is a vector of order N whose entries are equal to 1 for translational degrees of freedom in the
direction of the ground motion and zero otherwise. It is also reminded that the number of modes is N,
equal to the number of the degrees of freedom of the system.
By definition, the mode shape vector Fn describes the shape of the nth mode, and the vector of
displacements is equal to the sum:

X
N
uðtÞ ¼ Fn yn ðt Þ (3)
n¼1

The coupled equation of motion of Eq. 1 can be uncoupled after substituting u(t) using Eq. 3 and left-
multiplying it with FT. This is achieved with the aid of the orthogonality condition, where the
products FTmmFn, FTmcFn and FTmkFn are equal to zero if n 6¼ m. The following generalized
quantities can be defined when n ¼ m:

M n ¼ FTn mFn
C n ¼ FTn cFn (4)
K n ¼ FTn kFn

It is also assumed that

Qn ðt Þ ¼ FTn qðt Þ ¼ FTn mB€


ug ðt Þ: (5)

The coupled equation of motion (Eq. 1) is equivalent to the following set of equations:

M 1 y€1 þ C 1 y_1 þ K 1 y1 ¼ Q1 ðt Þ
M 2 y€2 þ C 2 y_2 þ K 2 y2 ¼ Q2 ðtÞ
(6)
... ...
M n y€n þ C n y_n þ K n yn ¼ Qn ðtÞ

If we divide by Mn and assume Cn ¼ 2xnon and Kn ¼ Mno2n, the equations are further simplified to

Page 2 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

Ln
y€n þ 2xn on y_n þ o2n yn ¼  u€g ðtÞ (7)
Mn

where

X
N
Ln ¼ FTn m ¼ mnj Fnj (8)
j¼1

For brevity, only the last equation of Eq. 7 is shown. Equations 6 or 7 are not coupled and therefore
can be solved separately to obtain yn. Equation 7 can be seen as the equation of motion of a SDOF
with frequency on subjected to a strong ground motion whose amplitude € ug(t) has been multiplied
with Ln/Mn. Otherwise, the solution of Eq. 7 is the solution of a SDOF with frequency on subjected

ug(t) and then multiplied with Ln/Mn.
The ratio Ln/Mn is the participation factor of mode n, which always sums to one and may take both
positive and negative values. This quantity depends on how the modes have been normalized and is
not equal to the contribution of a mode to a response quantity.
The vector of displacements u is back-calculated using Eq. 3, and every other response quantity
can be subsequently determined once u is known. For example, the vector of elastic forces will be

f ðt Þ ¼ kuðt Þ ¼ . . .
(9)
. . . ¼ kF1 y1 ðt Þ þ kF2 y2 ðt Þ þ    þ kFn yn ðtÞ:

For damped systems, the above decoupling is feasible only if the damping matrix c is orthogonal,
otherwise Eq. 6 does not apply. Usually c is derived as a combination of k and m, i.e., c ¼ a0k + a1m,
where a0, a1 are constants. This form of damping is known as Rayleigh damping. Another approach
would be to form the damping matrix so that every mode is damped with Cn ¼ 2xnon. This is
achieved with the aid of the formula (Chopra 2000):
!
XN
2xn on T
c¼m Fn Fn m (10)
n¼1
Mn

A fundamental property of modal analysis is that the response can be accurately captured if a,
relatively, small number of modes are considered. Therefore, the question is how many modes
should be included in modal analysis. The answer depends on the applied load and on the response
parameter examined. For example, more modes are required to accurately capture the base shear
than the roof displacement. Since the exact solution is not a priory known, one has to identify the
response parameters that are likely to be sensitive to high modes and then decide the number of
modes to include. In lieu of the above, some building codes suggest that the number of modes
considered should be chosen so that the effective modal masses for the modes taken into account
amounts to at least the 90 % of the total mass of the structure. Due to the relationship of the effective
modal mass and the base shear, this rule implies that the error in the base shear estimate should be
less than 10 %. Moreover, according to Eurocode 8 (EC8 2004), the analysis should include all
modes with effective modal masses greater than 5 % of the total.

Page 3 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

a
b 2

1.5

acceleration (g)
0.5

−0.5

−1

−1.5

−2
0 5 10 15 20 25 30 35 40
time (sec)

Fig. 1 (a) Plane frame geometry, (b) the Imperial Valley record (15 Oct. 1979)

Example 1: Modal Analysis of a Plane Frame Determine the response of the three-storey pane
frame of Fig. 1a when subjected to the Imperial Valley (1979) ground motion record (Fig. 1b).
Assume k ¼ 10,000 kN/m, m ¼ 20 t, and 5 % damping of the critical for every mode.
The mass and the stiffness matrices are obtained as follows:
2 3 2 3
1:5k k 0 15, 000 10, 000 0
6 7 6 7
k ¼ 4 k 2k k 5 ¼ 4 10, 000 20, 000 10, 000 5 kN=m
0 k k 0 10, 000 10, 000

2 3 2 3
2m 0 0 40 0 0
m¼4 0 2m 05¼4 0 40 0 5 t:
0 0 m 0 0 20

The eigenvalue problem, |km2| ¼ 0, is first solved to obtain the eigenmodes (or just “modes”) and
the eigenperiods (or just “periods”) of the frame. The eigenperiods are T ¼ [0.98, 0.31, 0.21]s and
the mode vector is
2 3
0:69 0:95 0:38
F ¼ ½F1 F2 F3  ¼ 4 0:92 0:16 0:83 5
1:00 1:00 1:00

The first column gives the first mode and every row corresponds to the modal displacement of the
respective storey. The system of Eq. 7 now becomes

84:22
y€1 þ 0:64y_1 þ 40:8y1 ¼  u€g ðtÞ
72:62
11:32
y€2 þ 2:04y_2 þ 418:2y2 ¼ u€g ðtÞ
56:90
2:09
y€3 þ 3:03y_3 þ 916:0y3 ¼  u€g ðtÞ
53:60

Page 4 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

0.1

0.05 0.0506 m

y1(t)
0

−0.05

−0.1
0 5 10 15 20 25 30 35 40

x 10−3
2
0.0016 m
1
y2(t)

−1

−2
0 5 10 15 20 25 30 35 40

x 10−4
2
0.0002 m
1
y3(t)

−1

−2
0 5 10 15 20 25 30 35 40
time (sec)

Fig. 2 Response histories of modal displacements yn (t) (Eq. 3)

The modal displacements yn(t) are obtained solving the above equations either using Duhamel’s
integral or, preferably, numerically using the Newmark’s method (Bathe 1996; Chopra 2000). The
modal response histories of yn(t) are shown in Fig. 2. According to the three plots, the modal
displacement of the first mode is considerably larger than that of modes two and three (note the
different scale of the vertical axes). Moreover, the response histories of higher modes have a more
rich frequency content compared to that of lower modes. Figure 3 shows the modal displacements of
the third storey, obtained using Eq. 3. The first three plots show the displacement response history of
every mode and the fourth plot shows the actual/total displacement of the third storey. Regarding the
amplitude and the frequency content, the observations of Fig. 2 still hold. Moreover, the maxima of
every response history occur at different time instants. The figure also implies that we can have
a good approximation of the final solution using the first mode only and discarding modes two and
three with a minor loss of accuracy.

Page 5 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

0.0506 m
0.04
0.02

u1(t)
0
−0.02
−0.04
−0.06
0 5 10 15 20 25 30 35 40

x 10−3
2
0.0016 m
1
u2(t)

0
−1
−2
0 5 10 15 20 25 30 35 40

x 10−4
2
0.0002 m
1
u3(t)

0
−1
−2
0 5 10 15 20 25 30 35 40

0.1
0.05 0.052 m
utot(t)

0
−0.05
−0.1
0 5 10 15 20 25 30 35 40
time (sec)

Fig. 3 Response history of the modal displacements of the third storey. The plot in the bottom shows the actual/total
response history of the third storey

2 3
71:5068 28:9232 5:0115
c¼4 75:9453 26:4174 5
sym 40:4784

The final solution of the last plot in the bottom of Fig. 3 has been verified also by solving the equation
of motion (Eq. 1) with the aid of Newmark’s method. For this purpose, we use Eq. 10 to obtain the
damping matrix c shown above. Direct integration algorithms are usually simpler and their addi-
tional computational effort is easily handled by modern computers. Modal analysis may be preferred
because it (i) helps decompose the problem and allows drawing qualitative conclusions, (ii) is faster
when only the most significant modes are considered, and (iii) is the basis of response spectrum
modal analysis.

Page 6 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

Fig. 4 Acceleration response spectra for three strong ground motion records

Response Spectrum
In seismic design problems, in most occasions, it is sufficient to know only the peak values of the
response, defined as the maximum of the absolute value of a response quantity r(t) that varies in time
(Chopra 2005):

r0 ¼ maxjrðt Þj (11)

where r0 is the absolute value of r(t). The response spectrum of a quantity is the plot of the maximum
(or peak) value of a response quantity (e.g., displacement, velocity, acceleration) against the full
range of interest of the natural vibration period values Tn.
The word spectrum (plural spectra or spectrums) was first used in science within the field of optics
to describe the rainbow of colors in visible light when separated using a prism. In other words,
“spectrum” implies a broad range of conditions or behaviors grouped and studied together. If a given
period value Tn refers to a different SDOF structure, a response spectrum of a quantity by definition
provides the maximum value of the response quantity of interest under a given ground motion for the
whole range of structures of interest. Therefore, such plots give the maximum response for the
whole range (or spectrum) of SDOF structures.

Properties of Response Spectra


Figure 4 shows the response spectra of three recorded ground motion records. Some very interesting
observations regarding the properties of the response spectra can be made. All three spectra intersect
the vertical axis at an acceleration value equal to the peak ground acceleration, i.e., the maximum
acceleration value of the recorded ground motion. This point corresponds to a very rigid SDOF
system with period equal to zero. Moreover, the Manjil (1990) and the Assisi-Stallone (1997) ground
motions are clearly stronger compared to the Parkfield (1966) record, while the Manjil record has
a richer frequency content compared to the Assisi-Stallone record. This stems from the fact that the
Manjil record maintains high acceleration values for a wide range of periods, while the Assisi-
Stallone spectrum decays quickly. A rich frequency content is also observed in the Parkfield record.

Page 7 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

Table 1 Spectral amplification factors for 84-percentile confidence (Newmark and Hall 1982)
Damping (%) Acceleration Velocity Displacement
50 percentile (median) 3.21–0.68x 2.31–0.41x 1.82–0.27x
84 percentile 4.38–1.04x 3.38–0.67x 2.73–0.45x

Apart from the observations of the above paragraph, the most important property of a response
spectrum is that it provides the maximum acceleration of a single-degree-of-freedom (SDOF)
system without the need of performing a dynamic analysis, since all necessary calculations
have been already performed in order to draw the spectrum. If Sa (Tn, x) is the x%-damped spectral
acceleration (ordinate of the acceleration spectrum), the peak deformation of a SDOF with period Tn
and damping x will be
 2
Tn
umax ¼ o2n S a ðT n , xÞ ¼ S a ð T n , xÞ (12)
2p

Similarly the peak value of the equivalent static force fs,max will be

S a ðT n , xÞ
fs, max ¼ kumax ¼ mS a ðT n , xÞ ¼ W (13)
g

where k, m, and W are the stiffness, mass, and weight of the SDOF, respectively.
Integrating the acceleration response history would give the velocity, while the integral of the
velocity history provides the displacement response history. This means that acceleration, velocity,
and displacement are proportional (by a factor o), except for a phase shift. The phase shift does not
influence the maximum response values and therefore the following approximations can be made:

S d ðT n , xÞ ¼ jumax j
S v ðT n , xÞ ¼ ju_ max j  oS a ðT n , xÞ ¼ PSV (14)
S d ðT n , xÞ ¼ j€
umax j  o2 S d ðT n , xÞ ¼ oPSV ¼ PSA

where PSV and PSA are the pseudospectral velocity and the pseudospectral acceleration, respec-
tively. They are not the exact maxima of velocity and acceleration, but they are very close to the
actual values and are usually used in engineering practice.

Design or Code-Compliant Spectra


The response spectra of Fig. 4 refer to the response of all possible SDOF systems of specific ground
motion records. On the other hand, a design spectrum recognizes that ground motions at a site may
originate from different seismic sources. Therefore, a design spectrum is representative of an
ensemble of ground motions and, contrary to the single-record spectra of Fig. 4, its shape is always
smooth. Design spectra have been derived from statistical analysis from an ensemble of response
spectra and correspond either to a mean or to a mean plus one standard deviation spectrum. They are
also known as “uniform hazard spectra,” since they refer to the same probability of exceedance for
the whole range of periods.
A procedure for constructing a design spectrum from peak values of ground acceleration, velocity,
and displacement has been presented by Newmark and Hall (1982). The procedure uses the
amplification factors of Table 1 and is applied in a logarithmic tripartite log paper where

Page 8 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

the horizontal axis denotes frequency (Hz). The steps required are quite simple (Fig. 5): (i) first, draw
the ground parameters (peak acceleration, velocity, and displacement) each parallel to the
corresponding axis. (ii) Then draw lines parallel to them at distances obtained from Table 1. For
example, for damping 5 %, the amplification factor of the median acceleration is 3.210.68ln(5) ¼
2.11 and 1.65 and 1.39 for the velocity and the displacement spectrum, respectively. (iii) Draw a line
connecting the response spectrum at 8 Hz with the ground motion at 33 Hz so that the response of
very stiff structures is equal to that of the ground. (iv) Finally, draw another line connecting the
response spectrum at 1/10 Hz with the ground motion curve at 1/33 Hz so that the peak response of
very flexible systems coincides with the peak ground deformation. Other approaches for creating
both elastic and inelastic spectra can be found in the literature.
Design codes give smooth elastic or design spectra. The distinction between elastic and design
spectrum is that the design spectrum is derived from the elastic by dividing the ordinates of the latter
by a response modification factor R (according to the terminology used in the USA) or a behavior
factor called q in the European practice. Some codes, such as Eurocode 8 (European Committee for
Standardization 2004), anchor the spectrum to the peak ground acceleration, but this is not always
the case. In seismic codes that do not specify explicitly the PGA value, it can be inferred either from
the acceleration value at period equal to zero or by dividing the flat segment of the spectrum (plateau)
with the acceleration amplification factors of Table 1.
The Eurocode 8 elastic spectrum is given by the expression below and is plotted in Fig. 6:

Fig. 5 Construction of design spectra according to Newmark and Hall (1982)

Page 9 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

Fig. 6 The elastic design spectrum of European Committee for Standardisation (2004)

8
>
> ag S ½1 þ ð2:5  1ÞðT =T B Þ , if 0  T  TB
< a S  2:5
g , if TB  T  TC
S a ðT , xÞ ¼ (15)
>
> a g S  2:5 
ðT C =T Þ  , if TC  T  TD
:
ag S  2:5 T C T D =T 2 , if T D  T  4sec

where S is the soil factor, and TB, TC, and TD are period values that change the shape of spectrum
according to the soil conditions. Eurocode 8 recommends the above spectrum for earthquakes of
magnitude above 5.5, while another expression is recommended for smaller earthquakes.  is the
damping correction factor equal to 1 when x ¼ 5 % (typical for reinforced concrete structures) and
obtained as
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ 10=ð5 þ xÞ  0:55 (16)

ag is the design acceleration which is equal to the importance factor gI times the design peak ground
acceleration (PGA). Figure 6 shows two response spectra obtained with the formula of EC8 (Eq. 15).
The spectrum with the solid line refers to ground type A (S ¼ 1.2, TB ¼ 0.15, TC ¼ 0.5, TD ¼ 2.0) and
will be used for the design of a reinforced concrete structure (x ¼ 5 %). The importance factor (gI) is
considered equal to 1.0 and the design PGA is 0.40 g. The spectrum with the dashed line refers to
ground type B (S ¼ 1.35, TB ¼ 0.2, TC ¼ 0.8, TD ¼ 2.0), PGA equal to 0.30 g, and gI ¼ 1.0. Note that
both spectra intercept the vertical axis at a value equal to agS, while the horizontal plateau (for
x ¼ 5 %) is at agS2.5. 2.5 is the acceleration spectral amplification factor approximately equal to that
given by the expression of Table 1. Both PGAs adopted for the site under consideration have a 10 %
probability of being exceeded in 50 years (return period Tm ¼ 475 years), and therefore the seismic
hazard of both response spectra is uniform and equal to a 10 % probability of being exceeded in
50 years.

Page 10 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

Response Spectrum Modal Analysis

Peak Modal Responses


Modal analysis provides the entire response history for a given ground motion record. For design
purposes, its application requires a design ground motion record that is representative of the seismic
hazard at the site. The use of such records is rare and requires special skills to select them. Hence, in
engineering practice, the seismic hazard is defined preferably with the aid of regional response
spectra as discussed the “Properties of Response Spectra” section.
Moreover, for design purposes, we usually use the maximum value of a response parameter and
not the entire response history. Since every mode can be treated as an independent SDOF system, the
maximum response values of a mode can be easily obtained from the corresponding response
spectrum (Eqs. 12 and 13).
If Sd(Tn, x), Sv(Tn, x), and Sa(Tn, x) denote the spectral displacement, velocity and acceleration,
respectively, the maximum modal displacements are obtained from a response spectrum as

Ln Ln T 2n
yn,max ¼ S d ð T n , xÞ ¼ S a ðT n , xÞ (17)
Mn M n 4p2

The maximum displacement and the equivalent lateral force of the jth storey will be

Ln Ln T 2n
ujn,max ¼ S d ðT n , xÞFnj ¼ S a ðT n , xÞFnj (18)
Mn M n 4p2

Ln
sjn,max ¼ mj S a ðT n , xÞFnj (19)
Mn

Alternatively, the maximum values of the member forces due to the nth mode can be obtained by
static analysis by loading the structure with the maximum equivalent lateral forces of Eq. 19. The
calculation would be
 
Ln
fn ¼ mFn S a ðT n , xÞ (20)
Mn

and the modal displacements are obtained as sn,max ¼ k1fn.

Modal Combination Rules


RSA provides only the maximum response parameter values of every mode. Once the maxima of
every mode are known, we need to calculate the exact maximum value of a response quantity. This
calculation is not straightforward. If the maximum value of every mode had occurred exactly at the
same time instant, then combining Eqs. 3 and 17, we would get

XN  2


 Ln T 
uo ¼ M 4p2 Fn S a ðT n , xÞ  maxðuðt ÞÞ
n
(21)
n¼1 n

The equation above provides a conservative estimation of umax, since, as also shown in Fig. 2, the
maxima of every mode do not occur concurrently. This upper bound is usually too conservative and

Page 11 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

rarely used for design. In practice, modal combination rules, as discussed in the next section, are
preferred.

The Square Root of the Sum of Squares (SRSS) Rule


The most common rule for modal combination is the Square Root of Sum of Squares (SRSS) rule.
According to this rule, the peak response of every mode is squared and then the squares are summed.
The estimation of the maximum response quantity of interest is the square of the sum. If we denote
r as an arbitrary response quantity (e.g., displacement, stress resultants) and rn is its peak modal
response, their SRSS combination is written as
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ro  r21 þ r22 þ    þ r2n (22)

In general, this combination rule gives good response estimates, especially for two-dimensional
problems. The major limitation is that in order to produce satisfying estimates, the modes should be
well separated, i.e., the eigenfrequencies should not have close values. If this condition is not met,
the CQC method should be used instead. A criterion to determine if two modes are well separated is

1 pffiffiffiffiffiffiffiffiffiffi
 1 þ 0:1 xn xm (23)
bnm

where bnm ¼ om/on ¼ Tn/Tm and xn, xm is the damping ratio of modes n and m, respectively.

The Complete Quadratic Combination (CQC) Rule


The complete quadratic combination (CQC) rule overcomes the limitations of the SRSS rule and
should be adopted if the condition of Eq. 23 is not satisfied. The rule is expressed as follows:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u N N
uX X
ro  t ϵnm rn rm (24)
n¼1 m¼1

where ϵnm is a correlation coefficient that takes values in the 0,1 range and is equal to 1 when n ¼ m.
If bnm is that of Eq. 23, the correlation term is calculated as
pffiffiffiffiffiffiffiffiffiffi
8 xn xm ðxn þ bnm xm Þb3=2
ϵnm ¼   
nm
  (25)
2 2
1  bnm þ 4xn xm bnm 1 þ bnm þ 4 x2n þ x2m b2nm
2

If the same modal damping is used for modes n and m (xn ¼ xm ¼ x), the equation reduces to

8x2 ð1 þ bnm Þb3=2


ϵnm ¼  
nm
(26)
2 2
1  bnm þ 4x bnm ð1 þ bnm Þ2
2

Note that according to Eq. 26 when bnm ¼ 1, i.e., for two modes with equal frequencies, their
correlation coefficient ϵnm is equal to 1. When the modes are well separated, the n 6¼ m terms of ϵnm
become very small and the CQC rule gives estimates very close to those of the SRSS rule. ϵnm can
take both positive and negative values and the CQC estimate may be above or below that of

Page 12 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

a 0.2 b
0.18 0.6
Spectral displacement (m)

Spectral acceleration (g)


0.16
0.5
0.14
0.12 0.4
0.1
0.3
0.08
0.06 0.2 0.384
0.04 0.370
0.1 0.203
0.0486
0.02
0.0040
0 0.0090 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
period T (sec) period T (sec)

Fig. 7 (a) Displacement response spectrum, (b) acceleration response spectrum

SRSS. Further discussion on the CQC rule can be found in Der Kiureghian (1981) and Wilson
et al. (1981).
Both the SRSS and the CQC method should be applied directly on the response quantity of
interest according to its modal value. Obtaining the value of a response quantity by combining the
peak values of other quantities must be avoided as it may result to gross errors.

Example 2: Response Spectrum Modal Analysis of a Plane Frame For the building of Fig. 1a,
estimate the storey displacements using response spectrum analysis. The structure is again subjected
to the Imperial Valley ground motion record of Fig. 1b.
The spectral displacements and the spectral accelerations of the Imperial Valley (1979) ground
motion record are shown in Fig. 7. We show both displacement and acceleration spectra for
comparison (although the acceleration spectrum is always adopted in practice). The modal displace-
ments yn,max are calculated with either of the expressions of Eq. 17. Thus, for the first mode, we
obtain:

Ln 84:22
y1,max ¼ S d ðT n , xÞ ¼  0:0486 ¼ 0:056m, or
Mn 72:62
Ln T 2n 84:22 0:982
y1,max ¼ S a ðT n , x Þ ¼  0:203  9:81 ¼ 0:056m
M n 4p2 72:62 4p2

Similarly, we obtain y2,max ¼ 0.0018m and y3,max ¼ 0.0002m for the second and the third mode,
respectively. Both methods yield practically the same value; the minor differences are due to round-
off errors. These values should coincide with the maxima of Fig. 2. The differences observed are
again due to round-off errors. Note that we have dropped the negative signs using the absolute values
of all response quantities.
The modal displacements of the first mode are (Eq. 3)
2 3 2 3
0:69 0:0388
u1,max ¼ F1 y1,max ¼ 4 0:92 5  0:056 ¼ 4 0:0519 5m:
1:0000 0:0565

The un,max values are summarized in Table 2.

Page 13 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

Table 2 Response spectrum modal displacements


Storey 1 Storey 3 Storey 3
u1,max 0.0388 0.0519 0.0565
u2,max 0.0017 0.0003 0.0018
u3,max 0.0001 0.0001 0.0002
Sum (Eq. 21) 0.0403 0.0519 0.0580
SRSS (Eq. 22) 0.0389 0.0519 0.0565
CQC (Eq. 24) 0.0389 0.0519 0.0565
RHA 0.0333 0.0468 0.0520

Table 3 bnm values used for the CQC method


Mode 1 2 3
1 1.0000 0.3123 0.2110
2 3.2021 1.0000 0.6757
3 4.7396 1.4799 1.0000

If the maxima of un(t) occurred at the same time instant, then we could apply Eq. 21 to get the
maximum storey displacements. Although this is not correct, it can be used to give a conservative
upper bound. Therefore
 
X N 
 
umax <  Fn yn,max  ¼ jF1 y1,max j þ jF2 y2,max j þ jF3 y3,max j
 n¼1 

2 3 2 3 2 3 2 3
0:69 0:95 0:38 0:0403
¼ 4 0:92 5  0:0565 þ 4 0:16 5  0:0018 þ 4 0:83 5  0:0002 ¼ 4 0:0519 5m: (27)
1:0 1:0 1:0 0:0580

The above values clearly overestimate the actual values obtained with response history analysis
which are shown in the last row of Table 2. For a more accurate and less conservative estimation of
the maximum displacements, we use the SRSS or the CQC rule. A sample calculation of the top
storey displacement using the SRSS rule will be
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u3,o ¼ 0:05652 þ 0:00182 þ 0:00022 ¼ 0:0565m:

For storeys 1 and 2, the maximum displacements are u1,o ¼ 0.0389 m and u2,o ¼ 0.0519 m (Table 2).
The CQC method requires first to calculate bnm and ϵnm. The values of bnm are shown in Table 3,
and Table 4 shows ϵnm. A sample calculation of b12 would give b12 ¼ T2/T1 ¼ 0.31/0.984 ¼ 0.3123.
ϵ12 is obtained as

8  0:052  ð1 þ 0:3123Þ  0:31233=2


ϵ12 ¼  2 ¼ 0:005587 (28)
1  0:31232 þ 4  0:052  0:3123  ð1 þ 0:3123Þ2

Page 14 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

Table 4 ϵnm values used for the CQC method


Mode 1 2 3
1 1.0000 0.0056 0.0026
2 0.0056 1.0000 0.0592
3 0.0056 1.0000 0.0592

Moreover, when N ¼ 3, Eq. 24 becomes

r2o ¼ r21 þ r21 þ r22 þ r23 þ ðϵ12 þ ϵ21 Þr1 r2 þ


þðϵ23 þ ϵ32 Þr2 r3 þ ðϵ13 þ ϵ31 Þr1 r3 :

With the aid of Table 4, the maximum displacement of the third storey becomes

u23,o ¼ 0:05652 þ 0:00182 þ 0:00022


þ 2  0:0056  0:0565  0:0018
þ 2  0:0592  0:0018  0:0002
(29)
þ 2  0:0026  0:0565  0:0002
6 8 8
pffiffiffiffiffiffiffiffiffiffiffiþ 1:4  10 þ 4:2  10 þ 5:9  10 ¼ 0:0032
¼ 0:0032
u3,o ¼ 0:032 ¼ 0:0565

Note that for the problem studied, the correlation terms are very small (less than 106), and thus
the final prediction is similar to that of the SRSS method. Also, cautiousness is required due to the
fact that in the CQC calculation, the r terms maintain their sign (here, everything is positive). This
means that the CQC estimates could be greater or smaller than that of SRSS.
Another approach to obtain the modal displacements is through applying the equivalent lateral
forces of Eq. 20. For the first mode, the calculation will be
 
L1
f1 ¼ mF1 S a ðT 1 , xÞ
M12 32 3 2 3
40 0 0 0:6870 63:3358
84:22 4
¼ 0 40 0 5 4 0:9184 50:202g ¼ 4 84:6704 5kN:
72:62
0 0 20 1:0000 46:0954

The equivalent lateral forces are applied on the structure to the yield the modal displacements:
2 3 2 3
63:3358 0:0388
u1 ¼ k1 f 1 ¼ k1 4 84:6704 5 ¼ 4 0:0519 5m
46:0954 0:0565

The reader should appreciate that the base shear of the first mode, Vs1, is equal to the sum of the
entries of f1, which yields Vs1 ¼ 194.10kN.

Page 15 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

Multistorey Buildings with Arbitrary Plan Configuration


Response spectrum analysis has been so far discussed for plane frames. The extension to three-
dimensional buildings, although straightforward, requires some further attention. When three-
dimensional multistorey buildings are studied, we assume that each storey behaves as a rigid
diaphragm with three inplane degrees of freedom, two translational, and one rotational. Therefore,
the total number of degrees of freedom is 3N, where N is the number of storeys. Moreover, the
direction of seismic loading is arbitrary but can always be analyzed to two components parallel to the
x and y axes and denoted € ug,x, €ug,y, respectively. The equation of motion (Eq. 1) is now written:
2 3 2 3 2 3
u€x  ux u€g , x
m ~ 4 u€g,y 5
~ 4 u€y 5 þ c~ u_ x u_ y u_ y þ k~4 uy 5 ¼ m (30)

uy uy 0

where the mass matrix is


2 3
mx 0 0
m 4
~¼ 0 my 0 5 (31)
0 0 IO

and c~and k~ are the corresponding damping and stiffness matrices, respectively. IO is a diagonal
matrix of order N whose entries are the moments of inertia of every storey about the vertical axis that
passes through the center of mass. The modal vector consists of three subvectors, Fn ¼ hfxnfynfyni,
and the participation factor Lhn/Mn is now calculated from the quantities:

Lhxn ¼ fTxn m, for u€gx


(32)
Lhyn ¼ fTyn m, for u€gy

and

~ n
M n ¼ FTn mF (33)

Kn and Cn are defined as before using Eq. 4. Equation 32 depends on the component of ground
motion considered u€g,x or u€g,y . For every mode, we solve again Eq. 7 to obtain the modal
displacements yn and the lateral displacements and the torsional rotation of the diaphragm:

D E
un ¼ uxn uyn uyn ¼ fxn fyn fyn yn ðt Þ (34)

The modal displacements can be obtained from the design response spectrum using Eq. 17
or performing static analysis of the building applying the equivalent static forces associated with
the nth mode peak response (Eq. 20). In this case, assuming u€g ¼ u€gx , the lateral forces fn will be

Page 16 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

Lnx
fn ¼ mFn S a ðT n , xÞ (35)
Mn

and
1
un ¼ k~ f n : (36)

Response Spectrum Modal Analysis in Seismic Design Codes


Various design codes and guidelines, including European and US codes, suggest the use of both
linear and nonlinear methods of analysis. Nonlinear methods are beyond the scope of this entry,
while linear methods are based on response spectrum modal analysis and on a simplified method that
considers only the first mode of response. According to the terminology of Eurocode 8, linear elastic
analysis can be performed either with (a) the “lateral force method of analysis” for buildings meeting
specific conditions or (b) with the “modal response spectrum analysis,” applicable to all types of
buildings and being the reference method for obtaining seismic response estimates.
According to the “lateral force method of analysis,” a lateral load pattern that follows the first
mode is applied (e.g., Eq. 35). In Eurocode 8, the method is applicable only if the fundamental
period of vibration is less than 2s or 4TC, where TC is the corner period of the design spectrum.
A second condition that also has to be satisfied is that the structure should meet the criteria for
regularity in plan. Different criteria may be found in other codes. The “modal response spectrum
analysis” is recommended for all other cases, with the only exception of buildings with seismic
isolation provided by highly nonlinear devices. Practically every design code in the world uses these
two methods for linear elastic analysis and recommends criteria to determine if their applicable.

Summary
Response spectrum modal analysis has been presented, discussing that it is a simplified version of
modal analysis and appropriate for structural design. The method allows the use of smooth design
spectra for the assessment and the design of structures. All concepts discussed are presented in
a numerical example, while the adaptation of the method by modern design codes has been
conceptually explained.

Cross-References
▶ Assessment of Existing Structures Using Inelastic Static Analysis
▶ Assessment of Existing Structures Using Linear Analysis
▶ Assessment of Existing Structures Using History Analysis
▶ Modal Analysis
▶ Nonlinear Dynamic Seismic Analysis
▶ Time History Seismic Analysis

Page 17 of 18
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_133-1
# Springer-Verlag Berlin Heidelberg 2013

References
Bathe KJ (1996) Finite element procedures. Prentice Hall, New York
Chopra AK (2000) Dynamics of structures: theory and applications to earthquake engineering,
2nd edn. Prentice Hall, Englewood Cliffs
Chopra AK (2005) Earthquake dynamics of structures a primer, 2nd edn. Earthquake Engineering
Research Institute, Oakland
Clough RW, Penzien J (1975) Dynamics of structures. McGraw-Hill, New York, 634 pp
Der Kiureghian A (1981) A response spectrum method for random vibration analysis of MDF
systems. Earthq Eng Struct Dyn 9:419–435
European Committee for Standardisation (2004) Eurocode 8 (EC8) Design of structures for earth-
quake resistance – part 1: general rules for buildings. Brussels
Newmark NM, Hall WJ (1982) Earthquake spectra and design. Earthquake Engineering Research
Institute, Berkeley
Wilson EL, Der Kiureghian A, Bayo EP (1981) Replacement for the SRSS method in seismic
analysis. Earthq Eng Struct Dyn 9(2):187–192

Page 18 of 18

View publication stats

You might also like