Você está na página 1de 5

224

4.6.1.1 Modal and Spectral Analyses

Fundamentals of Earthquake Engineering

(4.11)

The response of MDOF systems to a transient signal may be calculated by decomposing the system
into series of SDOF systems, calculating the response of each in the time domain and then algebraically
combining the response history to obtain the response of the MDOF system. This is modal analysis. If
the analysis is only focused on the maximum response quantities, then the various modal maxima are
calculated under the effect of a response spectrum representing the transient signal, and the maxima
are combined to give an upper bound of the maximum response of the MDOF. This is modal spectral
analysis, or spectral analysis for short. Both the above methods are applicable only to linear elastic
systems, since they employ superposition. Modal analysis may be considered a time-domain solution,
whereas it can be argued that modal-spectral analysis is a frequency-domain solution.
Two concepts are needed for the development of modal analysis. These are the principle of superposition and the convolution integral. Selection of earthquake spectra (input) and adequate combinations
of modes are essential to perform modal spectral analysis. For a SDOF system, it can be shown that
the displacement at time t is given by the solution of equation (3.14) in Chapter 3.
The coupled equation of motion for MDOF structures given in matrix form in equation (4.9.2) can
be rewritten for linearly elastic systems as follows:
M x + C x + K x = M I xg

(4.12)

By a change of basis, equation (4.11) yields a set of uncoupled equations of motion, each of which
represents a SDOF system. The procedure is summarized below:
(a) Assume that the displacement vector can be expressed in the following form:
x = F Y (t )

(4.13)

where is the modal matrix and Y(t) is the vector of modal (or normal) coordinates. The modal
matrix is non-singular positive and hence can be inverted. Note that the columns of the matrix
, i.e. the modes of vibrations i, are not known at this stage;
(b) Formulate the eigenvalue problem for the MDOF system as follows:
K Fi = i2 M Fi

(4.14)

(c) Compute the N eigenvalues (or frequencies), and eigenvectors (or modes of vibration) from equation (4.13). This is a conventional eigenvalue analysis. Alternatively, Ritz vectors can also be
employed, especially for complex structural systems, they provide more accurate results for the
same number of modes computed through the eigenvectors. The mode with the lowest frequency
is the fundamental mode and the corresponding frequency is the fundamental frequency of vibration. Once the frequencies are known, they can be substituted one at a time into the following
equation:

(K 2 M ) x = 0

(4.15)

which can be solved for the relative amplitudes of motion for each of the displacement components in the particular mode of vibration. The key characteristic of the mode shapes is that they
are orthogonal with respect to the mass M and stiffness K matrices.
(d) Assume mode-proportional damping (i.e. total damping is the sum of the modal damping contributions), given by:
FiT C F j = 2 i i ij

Response Evaluation

225

(4.16.1)

In most FE codes, the mass and stiffness proportional damping is used as an efficient technique
of assembling a damping matrix without reference to the element contribution. If two modes only
are involved, this is termed Rayleigh damping and is given by the following expression:
C= M+ K

(4.16.2)

The parameters and can be evaluated if the damping ratio i is known for any two modes.
Using the following relationship:

+ i2 = 2 i i

+ 2j
2j
(4.16.3)

two simultaneous equations in and are derived for two known values of i. Consequently,
the damping ratio i in any mode can be calculated as below:

i =

(4.17.4)

(4.17.3)

(4.17.2)

(4.17.1)

The above assumption is essential to retain the option of solving decoupled equations of motion.
Since the mode shapes are orthogonal to M and K, they are also orthogonal to the Rayleigh
damping matrix.
(e) Formulate the equations of motion in terms of normal (or generalized) coordinates Yi:
Yi + 2 i i Yi + i2 Yi = i x g

K i
i
M

where the angular frequency i for the ith mode is:

i =
in which M i is the generalized mass given as follows:
i = FiT M Fi
M
and K i represents the generalized stiffness expressed by:
K i = FiT K Fi

i =

Li
i
M

(4.17.6)

(4.17.5)

The factor i is called the modal participation factor and provides a measure of the degree to
which the ith mode participates to the global dynamic response. This factor is as below:

where:
Li = FiT M I

(f) Compute the solutions of the system of N uncoupled equations in normal coordinates given in
equation (4.17.1). The response of the ith mode of vibration at any time t can be expressed by
the convolution (Duhamel) integral in the form:

226
Yi(t ) =

Li
A (t )
i i i
M

N
i =1

M
i

Li

(4.18)

Fundamentals of Earthquake Engineering

Ai (t ) M Fi
(4.19)

where Ai(t) is given by the solution of equation (3.14) in Chapter 3. Alternatively, the equation
of motion can be solved numerically in the time or frequency domain. These approaches are
known as direct integration method and fast Fourier transform, respectively.
(g) Compute the total elastic restoring force as follows:
R = K F Y (t ) =

N
i

L2i

M
i =1

Ai(t )
(4.20)

(h) Compute the total seismic base shear VB. It can be obtained by summing the effective earthquake
forces over the height of the structure:
VB =

Li
A (t ) Fi
i i
M

(4.21)

(i) Compute the relative displacement with respect to the base of the structure corresponding to the
ith mode of vibration:
x i = F Yi (t ) =

L2i

i =1

N
i

(4.22)

Equation (4.16.3) makes damping frequency-dependent. The procedure illustrated in (d) to compute
i will usually over-damp the higher modes of vibration, thus affecting the reliability of results for
high-rise structures or systems subjected to near-field earthquake ground motions. Proportional damping
can be visualized as immersion of the structure in a non-physical fluid whose viscosity becomes infinite
for rigid-body motion of the structure ( = 0). For higher frequency modes, viscosity acts to damp relative motion of the MDOF, with increasing effect as increases. Non-physical high-frequency vibrations, also known as noise, generated by numerical response simulation can be damped by the term
K.
i in equation (4.20) is defined as the effective modal mass. This quantity generally
The term L2i M
diminishes inversely with the order of modes. For example, in regular shear frame buildings, the fundamental mode accounts for up to 8590% of the total mass. Therefore, summing the response for the
first two to three modes will represent the MDOF system. On the other hand, slender long-span bridges
usually respond in tens or even hundreds of modes, all of which will be required to achieve adequate
representation of the MDOF. The sum of the modal masses is the total mass of the structure; i.e.:

i =1

M = M

Equations (4.19) and (4.21) express the entire history of actions and deformations of MDOF structures. Lumped systems with N degrees of freedom possess N independent mode shapes. It is thus possible to express the deformed shape of the structure in terms of amplitudes of these shapes by treating
them as generalized coordinates Y(t) as shown in equation (4.18).
In seismic analysis, the evaluation of maximum values of displacements and internal forces rather
than their whole time history, is often the primary purpose, especially in design. Peak responses obtained
for individual modes can be combined using statistical methods. The modal spectral (or spectral, or

Response Evaluation

227

response spectrum) analysis estimates peak values of structural response by combining maximum modal
contributions. These maxima are determined from earthquake response spectra for elastic SDOFs. The
spectral analysis procedure is summarized in the following steps:

Li
Sai
FiT M Fi
(4.23)

(a) Compute modes and frequencies of the MDOF by following steps (a) to (d) of the procedure for
modal analysis given above.
(b) Compute for each mode the generalized mass M i and the modal participation factor i from
equations (4.17.3) and (4.17.5), respectively.
(c) Select an acceleration spectrum (e.g. as in Section 3.4.2).
(d) Compute the spectral accelerations Sai corresponding to the periods Ti determined for each mode
of vibration.
(e) Compute the maximum inertia forces for each mode. The vector of earthquake forces Fmax,i (t)
for the ith mode is as follows:
Fmax, i(t ) = M Fi
(f) Compute the maximum values of response parameters, e.g. actions (moments, shears, axial loads
and torsion, if any) and deformations (displacements and rotations) discussed in Section 4.8. The
response quantities can be determined from static analysis.
(g) Combine the quantities determined in step (f) for each mode to determine the total response
parameters.

i =1

2
i

(4.24)

Decisions are needed for the number of modes to be combined and the combination method. The
choice of number of modes to be combined has implications on both accuracy and economy of the
procedure. In most cases of structural applications, two to three modes are sufficient, as mentioned
earlier. The objective is to account for at least 8590% of the total mass, which is achieved in regular
structures with relative ease. In special structures, e.g. slender long-span bridges, reaching the minimum
8590% limit may require combining tens or even hundreds of modes.
Various approximate formulae for superposition may be used in spectral analysis. The most commonly used methods are the square root of the sum of the squares (SRSS) and the complete quadratic
combination (CQC). A reasonable safe upper bound on the overall response parameters is obtained by
assuming that the response measures in the different modes are uncorrelated. For three-dimensional
structures with a large number of almost similar periods of vibration, this assumption is not
applicable.
In the SRSS, the total value of the response parameter E is given by:

E=

If the difference between two modal frequencies is less than 10%, the SRSS may lead to underestimating the structural response. Notwithstanding, the SRSS combination approach secures a safe
upper bound on global response quantities in most cases, as mentioned above. In some cases, local
response parameters may not be a safe upper bound, due to the effect of higher modes on local
quantities. The simplified code approach presented in Section 4.6.3 is indeed a simplification of the
SRSS method, by just replacing the modal mass of the fundamental, or predominant, mode by the
total mass.
When modes are closely spaced, a combination approach that includes cross-modal contributions is
required, since the closely spaced modes are at least partially correlated. This procedure may be used

228

i =1 j =1

E E

N
j

Fundamentals of Earthquake Engineering

ij
(4.25.1)

for all structures; where cross-correlations are low or non-existent, the cross-coupling terms will be
small or zero. The CQC is expressed as follows:

E=

(1 r 2 )

2
2

+ 4 2r (1 + r )

8 2(1 + r ) r 3 2
(4.25.2)

where ij is a cross-modal coefficient. This coefficient is generally expressed as a function of the modal
frequencies and damping characteristics and, for equal modal damping, i.e. i = j = , is as follows
(Der Kiureghian, 1980):

ij =
where r = j/i; the coefficient ij varies between 0 and 1 for i = j. If the modal frequencies of the
MDOF are well separated, the off-diagonal terms tend to zero and the CQC method approaches the
SRSS.
Estimates of the total value of the response parameter E obtained by CQC rule may be larger or
smaller than the estimates provided by the SRSS rule (Chopra, 2002). Figure 4.33 shows the bending
moment diagrams computed from response spectral analysis for a plane frame extracted from the sample
SPEAR building in Figure 4.2. The modal combination rules discussed above, i.e. SRSS and CQC, are
utilized. The damping value used for analyses is 5%. SRSS and CQC provide values that are in good
agreement. The modal analysis of the 3D frame shows that the frequencies of the system are not closely
spaced; the minimum difference between two frequencies is greater than 10%. However, the sample
SPEAR structure is a multi-storey building with asymmetric plan and hence the SRSS leads to reasonable estimates of response. The differences between the values computed through the CQC and SRSS
are lower than 10%.

4.6.1.2 Response History Analysis

105.00

175.43

160.35

78.28

87.80

157.05

143.87

142.20

129.41

178.55

192.54

114.72

158.75

142.97

86.32

In contrast to the frequency-domain solutions presented in Section 4.6.1.1 (notwithstanding the special
nature of modal analysis), the response of MDOF systems to a transient signal may be calculated by
80.50

131.10

140.73

Figure 4.33 Bending moments (in kNm) computed through response spectral analyses using two different modal
combinations for three-storey frame: square root of the sum of the squares (left) and complete quadratic combination
(right)

Você também pode gostar