3

Dynamic Analysis

3.1 Introduction .................................................................3-1

Static vs. Dynamic Analysis • Characteristics of
Earthquake Ground Motions • Dynamic Analysis
Methods for Seismic Bridge Design
3.2 Single-Degree-of-Freedom System .............................3-3
Equation of Motion • Characteristics of Free Vibration
• Response to Earthquake Ground Motion • Response
Spectra • Example of an SDOF System
3.3 Multi-Degree-of-Freedom System ............................3-15
Equation of Motion • Free Vibration and Vibration
Modes • Modal Analysis and Modal Participation
Factor • Example of an MDOF System • Multiple-
Rambabu Bavirisetty
Support Excitation • Time History Analysis
California Department
of Transportation 3.4 Response Spectrum Analysis .....................................3-23
Single-Mode Spectral Analysis • Uniform-Load
Murugesu Vinayagamoorthy
Method • Multimode Spectral Analysis • Multiple-
California Department
of Transportation Support Response Spectrum Method
3.5 Inelastic Dynamic Analysis........................................3-31
Lian Duan
Equations of Motion • Modeling Considerations
California Department
of Transportation 3.6 Summary.....................................................................3-34

3.1 Introduction
The primary purpose of this chapter is to present dynamic methods for analyzing bridge structures
when subjected to earthquake loads. Basic concepts and assumptions used in typical dynamic
analysis are presented first. Various approaches to bridge dynamics are then discussed. A few
examples are presented to illustrate their practical applications.

Static vs. Dynamic Analysis

The main objectives of a structural analysis are to evaluate structural behavior under various loads
and to provide the information necessary for design, such as forces, moments, and deformations.
Structural analysis can be classified as static or dynamic: while statics deals with time-independent
loading, dynamics considers any load where the magnitude, direction, and position vary with time.
Typical dynamic loads for a bridge structure include vehicular motions and wave actions such as
winds, stream flow, and earthquakes.

3-1

© 2003 by CRC Press LLC

3-2 Bridge Engineering: Seismic Design

FIGURE 3.1 Ground motions recorded during recent earthquakes.

Characteristics of Earthquake Ground Motions

An earthquake is a natural ground movement caused by various phenomena, including global
tectonic processes, volcanism, landslides, rock-bursts, and explosions. The global tectonic processes
are continually producing mountain ranges and ocean trenches at the Earth’s surface and causing
earthquakes. This section briefly discusses the earthquake input for seismic bridge analysis. Detailed
discussions of ground motions are presented in Chapter 1.
Ground motion is represented by the time history or seismograph in terms of acceleration,
velocity, and displacement for a specific location during an earthquake. Time-history plots contain
complete information about the earthquake motions in the three orthogonal directions (two hor-
izontal and one vertical) at the strong-motion instrument location. Acceleration is usually recorded
by strong-motion accelerograph, and the velocities and displacements are determined by numerical
integration. The accelerations recorded at locations that are approximately the same distance away
from the epicenter may differ significantly in duration, frequency content, and amplitude due to
different local soil conditions. Figure 3.1 shows several time histories of recent earthquakes.
From a structural engineering view, the most important characteristics of an earthquake are the
peak ground acceleration (PGA), duration, and frequency content. The PGA is the maximum
acceleration and represents the intensity of a ground motion. Although the ground velocity may
be a more significant measure of intensity than the acceleration, it is not often measured directly,
but determined using supplementary calculations [1]. The duration is the length of time between
the first and the last peak exceeding a specified strong motion level. The longer the duration of a
strong motion, the more energy is imparted to a structure. Since the elastic strain energy absorbed
by a structure is very limited, a longer strong earthquake has a greater possibility to enforce a
structure into the inelastic range. The frequency content can be represented by the number of zero
crossings per second in the accelerogram. It is well understood that when the frequency of a regular
disturbing force is the same as the natural vibration frequency of a structure (resonance), the
oscillation of structure can be greatly magnified and effects of damping become minimal. Although

© 2003 by CRC Press LLC

Dynamic Analysis 3-3

earthquake motions are never as regular as a sinusoidal waveform, there is usually a period that
dominates the response.
Since it is impossible to measure detailed ground motions for all structure sites, the rock motions
or ground motions are estimated at a fault and then propagated to the Earth’s surface using a
computer program considering the local soil conditions. Two guidelines [2,3] recently developed
by the California Department of Transportation provide the methods to develop seismic ground
motions for bridges.

Dynamic Analysis Methods for Seismic Bridge Design

Depending on the seismic zone, geometry, and importance of the bridge, the following analysis
methods may be used for seismic bridge design:
• The single-mode method (single-mode spectral and uniform-load analysis) [4,5] assumes
that seismic load can be considered as an equivalent static horizontal force applied to an
individual frame in either the longitudinal or transverse direction. The equivalent static force
is based on the natural period of a single degree of freedom (SDOF) and code-specified
response spectra. Engineers should recognize that the single-mode method (sometimes
referred to as equivalent static analysis) is best suited for structures with well-balanced spans
with equally distributed stiffness.
• Multimode spectral analysis assumes that member forces, moments, and displacements due to
seismic load can be estimated by combining the responses of individual modes using methods
such as the complete quadratic combination (CQC) method and the square root of the sum of
the squares (SRSS) method. The CQC method is adequate for most bridge systems [6], and
the SRSS method is best suited for combining responses of well-separated modes.
• The multiple support response spectrum (MSRS) method provides response spectra and the
peak displacements at individual support degrees of freedom by accurately accounting for
the spatial variability of ground motions, including the effects of incoherence, wave passage,
and spatially varying site response. This method can be used for multiply supported long
structures [7].
• The time-history method is a numerical step-by-step integration of equations of motion. It
is usually required for critical/important or geometrically complex bridges. Inelastic analysis
provides a more realistic measure of structural behavior when compared with an elastic
analysis.
Selection of the analysis method for a specific bridge structure should not be purely based on
performing structural analysis, but be based on the effective design decisions [8]. Detailed discus-
sions of the above methods are presented in the following sections.

3.2 Single-Degree-of-Freedom System

The familiar spring–mass system represents the simplest dynamic model and is shown in Figure 3.2a.
When the idealized, undamped structures are excited either by moving the support or by displacing
the mass in one direction, the mass oscillates about the equilibrium state forever without coming
to rest. But real structures do come to rest after a period of time due to a phenomenon called
damping. To incorporate the effect of the damping, a massless viscous damper is always included
in the dynamic model, as shown in Figure 3.2b.
In a dynamic analysis, the number of displacements required to define the displaced positions
of all the masses relative to their original positions is called the number of degrees of freedom
(DOF). When a structural system can be idealized with a single mass concentrated at one location
and moved in only one direction, this dynamic system is called an SDOF system. Some structures,

© 2003 by CRC Press LLC

3-4 Bridge Engineering: Seismic Design

FIGURE 3.2 Idealized dynamic model. (a) Undamped SDOF system; (b) damped SDOF system.

FIGURE 3.3 Examples of SDOF structures. (a) Water tank supported by single column; (b) one-story frame
building; (c) two-span bridge supported by single column.

such as a water tank supported by a single-column, one-story frame structure and a two-span bridge
supported by a single column, could be idealized as SDOF models (Figure 3.3).
In the SDOF system shown in Figure 3.3c, the mass of the bridge superstructure is the mass of
the dynamic system. The stiffness of the dynamic system is the stiffness of the column against side
sway, and the viscous damper of the system is the internal energy absorption of the bridge structure.

Equation of Motion
The response of a structure depends on its mass, stiffness, damping, and applied load or
displacement. The structure could be excited by applying an external force p(t) on its mass or
by a ground motion u(t) at its supports. In this chapter, since the seismic loading is induced
by exciting the support, we focus mainly on the equations of motion of an SDOF system
subjected to ground excitation.

© 2003 by CRC Press LLC

Dynamic Analysis 3-5

FIGURE 3.4 Earthquake–induced motion of an SDOF system.

The displacement of the ground motion ug , the total displacement of the single mass ut , and
the relative displacement between the mass and ground u (Figure 3.4) are related by

ut = u + ug (3.1)

By applying Newton’s law and D’Alembert’s principle of dynamic equilibrium, it can be

shown that

fI + fD + fS = 0 (3.2)

where fI is the inertial force of the single mass and is related to the acceleration of the mass by
fI = m uúút ; fD is the damping force on the mass and related to the velocity across the viscous damper
by fD = c uú ; fS is the elastic force exerted on the mass and related to the relative displacement
between the mass and the ground by fS = k u , where k is the spring constant; c is the damping
ratio; and m is the mass of the dynamic system.
Substituting these expressions for fI , fD , and fS into Eq. (3.2) gives

m uúút + c uú + k u = 0 (3.3)

The equation of motion for an SDOF system subjected to a ground motion can then be obtained
by substituting Eq. (3.1) into Eq. (3.3), and is given by

m uúú + c uú + k u = - m uúúg (3.4)

Characteristics of Free Vibration

To determine the characteristics of the oscillations such as the time to complete one cycle of
oscillation ( Tn ) and number of oscillation cycles per second ( w n ), we first look at the free vibration
of a dynamic system. Free vibration is typically initiated by disturbing the structure from its
equilibrium state by an external force or displacement. Once the system is disturbed, the system
vibrates without any external input. Thus, the equation of motion for free vibration can be obtained
by setting uúúg to zero in Eq. (3.4) and is given by

m uúú + c uú + k u = 0 (3.5)

© 2003 by CRC Press LLC

3-6 Bridge Engineering: Seismic Design

FIGURE 3.5 Typical response of an SDOF system. (a) Undamped; (b) damped.

Dividing Eq. (3.5) by its mass m will result in

uúú + Ê ˆ uú + Ê ˆ u = 0
c k
(3.6)
Ë m¯ Ë m¯

uúú + 2x w n + w 2nu = 0 (3.7)

where w n = k / m the natural circular frequency of vibration or the undamped frequency;

x = c ccr the damping ratio; ccr = 2 m w n = 2 k m = 2 k w n the critical damping coefficient.
Figure 3.5a shows the response of a typical idealized, undamped SDOF system. The time required
for the SDOF system to complete one cycle of vibration is called the natural period of vibration
(Tn ) of the system and is given by

2p m
Tn = = 2p (3.8)
wn k

Furthermore, the natural cyclic frequency of vibration fn is given by

© 2003 by CRC Press LLC

Dynamic Analysis 3-7

FIGURE 3.6 Response of an SDOF system for various damping ratios.

w 1 k
f n = -----n- = ------ ---- (3.9)
2p 2p m

Figure 3.5b shows the response of a typical damped SDOF structure. The circular frequency of
the vibration or damped vibration frequency of the SDOF structure, w d , is given by
w d = w n 1 - x2 .
The damped period of vibration ( Td ) of the system is given by

2p 2p m
Td = = (3.10)
wd 1 - x2 k

When x = 1 or c = ccr , the structure returns to its equilibrium position without oscillating and
is referred to as a critically damped structure. When x > 1 or c > ccr , the structure is overdamped
and comes to rest without oscillating, but at a slower rate. When x < 1 or c < ccr , the structure is
underdamped and oscillates about its equilibrium state with progressively decreasing amplitude.
Figure 3.6 shows the response of SDOF structures with different damping ratios.
For structures such as buildings, bridges, dams, and offshore structures, the damping ratio is less
than 0.15 and thus can be categorized as underdamped structures. The basic dynamic properties
estimated using damped or undamped assumptions are approximately the same. For example, when
x = 0.10 , w d = 0.995w n , and Td = 1.01Tn .
Damping dissipates the energy out of a structure in opening and closing of microcracks in
concrete, stressing of nonstructural elements, and friction at the connection of steel members. Thus,
the damping coefficient accounts for all energy-dissipating mechanisms of the structure and can be
estimated only by experimental methods. Two seemingly identical structures may have slightly
different material properties and may dissipate energy at different rates. Since damping does not
play an important quantitative role except for resonant responses in structural responses, it is
common to use average damping ratios based on the types of construction materials. Relative
damping ratios for common types of structures, such as welded metal of 2 to 4%, bolted metal
structures of 4 to 7%, prestressed concrete structures of 2 to 5%, reinforced concrete structures of
4 to 7%, and wooden structures of 5 to 10%, are recommended by Chmielewski et al. [9].

© 2003 by CRC Press LLC

3-8 Bridge Engineering: Seismic Design

FIGURE 3.7 Induced earthquake force vs. time on an SDOF system.

Response to Earthquake Ground Motion

A typical excitation of an earth movement is shown in Figure 3.7. The basic equation of motion of
an SDOF system is expressed in Eq. (3.4). Since the excitation force m uúúg cannot be described by
simple mathematical expression, closed-form solutions for Eq. (3.4) are not available. Thus, the
entire ground excitation needs to be treated as a superposition of short-duration impulses to evaluate
the response of the structure to the ground excitation. An impulse is defined as the product of the
force times duration. For example, the impulse of the force at time t during the time interval dt
equals -m uúúg ( t)dt and is represented by the shaded area in Figure 3.7. The total response of the
structure for the earthquake motion can then be obtained by integrating all responses of the
increment impulses. This approach is sometimes referred to as “time history analysis.” Various
solution techniques are available in the technical literature on structural dynamics [1,10].
In seismic structural design, designers are interested in the maximum or extreme values of the
response of a structure as discussed in the following sections. Once the dynamic characteristics
( Tn and w n ) of the structure are determined, the maximum displacement, moment, and shear on
the SDOF system can easily be estimated using basic principles of mechanics.

Response Spectra
The response spectrum is a relationship of the peak values of a response quantity (acceleration, velocity, or
displacement) with a structural dynamic characteristic (natural period or frequency). Its core concept in
earthquake engineering provides a much more convenient and meaningful measure of earthquake effects than
any other quantity. It represents the peak response of all possible SDOF systems to a particular ground motion.
Elastic Response Spectrum
This, the response spectrum of an elastic structural system, can be obtained by the following steps [10]:
1. Define the ground acceleration time history (typically at a 0.02-second interval).
2. Select the natural period Tn and damping ratio x of an elastic SDOF system.
3. Compute the deformation response u(t ) using any numerical method.
4. Determine uo , the peak value of u(t ) .
Calculate the spectral ordinates by D = uo , V = 2p D / Tn , and A = (2 p / Tn ) D .
2
5.
6. Repeat Steps 2 and 5 for a range of Tn and x values for all possible cases.
7. Construct results graphically to produce three separate spectra as shown in Figure 3.8 or a
combined tripartite plot as shown in Figure 3.9.
It is noted that although three spectra (displacement, velocity, and acceleration) for a specific
ground motion contain the same information, each provides a physically meaningful quantity. The
displacement spectrum presents the peak displacement. The velocity spectrum is related directly to
the peak strain energy stored in the system. The acceleration spectrum is related directly to the peak
value of the equivalent static force and base shear.

© 2003 by CRC Press LLC

Dynamic Analysis 3-9

FIGURE 3.8 Example of response spectra (5% critical damping) for Loma Prieta 1989 motion.

A response spectrum (Figure 3.9) can be divided into three ranges of periods [10]:
• Acceleration-sensitive region (very short period region): A structure with a very short period
is extremely stiff and expected to deform very little. Its mass moves rigidly with the ground,
and its peak acceleration approximately equals the ground acceleration.
• Velocity-sensitive region (intermediate-period region): A structure with an intermediate
period responds greatly to the ground velocity than other ground motion parameters.
• Displacement-sensitive region (very long period region): A structure with a very long period
is extremely flexible and expected to remain stationary while the ground moves. Its peak
deformation is closer to the ground displacement. The structural response is most directly
related to ground displacement.
Elastic Design Spectrum
Since seismic bridge design is intended to resist future earthquakes, use of a response spectrum obtained
from a particular past earthquake motion is inappropriate. In addition, jagged spectrum values over
small ranges would require an unreasonable accuracy in the determination of the structure period [11].
It is also impossible to predict a jagged response spectrum in all its details for a ground motion that
may occur in the future. To overcome these shortcomings, the elastic design spectrum, a smoothened
idealized response spectrum, is usually developed to represent the envelopes of ground motions recorded
at the site during past earthquakes. The development of an elastic design spectrum is based on statistical
analysis of the response spectra for the ensemble of ground motions. Figure 3.10 shows a set of elastic
design spectra in Caltrans Bridge Design Specifications [12]. Figure 3.11 shows project-specific accel-
eration response spectra for the California Sonoma Creek Bridge.
Engineers should recognize the conceptual differences between a response spectrum and a
design spectrum [10]. A response spectrum is only the peak response of all possible SDOF systems
due to a particular ground motion, whereas a design spectrum is a specified level of seismic

© 2003 by CRC Press LLC

3-10 Bridge Engineering: Seismic Design

FIGURE 3.9 Tripartite plot–response spectra (1994 Northridge Earthquake, Arleta–Rordhoff Ave. Fire Station).

design forces or deformations and is the envelope of two different elastic design spectra. The
elastic design spectrum provides a basis for determining the design force and deformation for
elastic SDOF systems.
Inelastic Response Spectrum
A bridge structure may experience inelastic behavior during a major earthquake. The typical elastic
and elastic–plastic responses of an idealized SDOF to severe earthquake motions are shown in
Figure 3.12. The input seismic energy received by a bridge structure is dissipated by both viscous
damping and yielding (localized inelastic deformation converting into heat and other irrecoverable
forms of energy). Both viscous damping and yielding reduce the response of inelastic structures
compared with elastic structures. Viscous damping represents the internal friction loss of a structure
when deformed and is approximately a constant because it depends mainly on structural materials.
Yielding, on the other hand, varies depending on structural materials, structural configurations,
and loading patterns and histories. Damping has negligible effects on the response of structures for
the long-period and short-period systems and is most effective in reducing response of structures
for intermediate-period systems.

© 2003 by CRC Press LLC

Dynamic Analysis 3-11

FIGURE 3.10 Typical Caltrans elastic design response spectra.

FIGURE 3.11 Acceleration response spectra for Sonoma Creek Bridge.

© 2003 by CRC Press LLC

3-12 Bridge Engineering: Seismic Design

FIGURE 3.12 Response of an SDOF to earthquake ground motions. (a) Elastic system; (b) inelastic system.

In seismic bridge design, a main objective is to ensure that a structure is capable of deforming
in a ductile manner when subjected to a larger earthquake loading. It is desirable to consider the
inelastic response of a bridge system to a major earthquake. Although a nonlinear inelastic dynamic
analysis is not difficult in concept, it requires careful structural modeling and intensive computing
effort [8]. To consider inelastic seismic behavior of a structure without performing a true nonlinear
inelastic analysis, the ductility-factor method can be used to obtain the inelastic response spectra
from the elastic response spectra. The ductility of a structure is usually referred to as the displace-
ment ductility factor m defined by Figure 3.13:

Du
m= (3.11)
Dy

where Du is ultimate displacement capacity and D y is yield displacement.

The simplest approach to developing the inelastic design spectrum is to scale the elastic design
spectrum down by some function of the available ductility of a structural system:

ARSelastic
ARSinelastic = (3.12)
f (m )

Ï1 for T n £ 0.03 sec

Ô
f ( m ) = Ì 2m – 1 for 0.03 sec < T n £ 0.5 sec (3.13)
Ô
Óm for T n ≥ 0.5 sec

© 2003 by CRC Press LLC

Dynamic Analysis 3-13

FIGURE 3.13 Lateral load–displacement relations.

For the very short period ( Tn £ 0.03 sec) in the acceleration-sensitive region, the elastic displacement
demand D ed is less than displacement capacity D u (see Figure 3.13). The reduction factor f (m ) = 1
implies that the structure should be designed and remain at elastic to avoid excessive inelastic defor-
mation. For the intermediate period (0.03 sec < Tn £ 0.5 sec) in the velocity-sensitive region, elastic
displacement demand D ed may be greater or less than displacement capacity D u and the reduction
factor is based on the equal-energy concept. For the very long period ( Tn > 0.5 sec) in the displace-
ment-sensitive region, the reduction factor is based on the equal-displacement concept.

Example of an SDOF System

Given
An SDOF bridge structure is shown in Figure 3.14. To simplify the problem, the bridge is assumed
to move only in the longitudinal direction. The total resistance against the longitudinal motion
comes in the form of friction at bearings, and this could be considered a damper. Assume the
following properties for the structure: damping ratio x = 0.05, area of superstructure A = 3.57 m2,
moment of column Ic = 0.1036 m4, Ec of column = 20,700 MPa, material density r = 2400 kg/m3,
length of column Lc = 9.14 m, and length of the superstructure Ls = 36.6 m. The acceleration response
curve of the structure is given in Figure 3.11. Determine (1) natural period of the structure, (2)
damped period of the structure, (3) maximum displacement of the superstructure, and (4) maxi-
mum moment in the column.
Solution

12 Ec Ic 12(20700 ¥ 10 6 )(0.1036)
Stiffness: k = = = 33690301 N/m
L3c 9.143

Mass: m = A Ls r = (3.57)(36.6)(2400) = 313, 588.8 kg

k 33, 690, 301

Natural circular frequency: w n = = = 10.36 rad/s
m 313, 588.8

© 2003 by CRC Press LLC

3-14 Bridge Engineering: Seismic Design

FIGURE 3.14 SDOF bridge example. (a) Two-span bridge schematic diagram; (b) single -column bent; (c) idealized
equivalent model for longitudinal response.

w n 10.36
Natural cyclic frequency: fn = = = 1.65 cycles/s
2p 2p

1 1
Natural period of the structure: Tn = = = 0.606 s
fn 1.65

The damped circular frequency is given by

w d = w n 1 - x 2 = 10.36 1 - 0.052 = 10.33 rad/s

The damped period of the structure is given by

2p 2p
Td = = = 0.608 s
w d 10.33

From the ARS curve, for a period of 0.606 s, the maximum acceleration of the structure will be
0.9 g = 1.13 ¥ 9.82 = 11.10 m/s. Then,

The force acting on the mass = m ¥ 11.10 = 313588.8 ¥ 11.10 = 3.48 MN

FL3c 3.48 ¥ 9.143

The maximum displacement = = = 0.103 m
12 EIc 12 ¥ 20700 ¥ 0.1036

FLc 3.48 ¥ 9.14

The maximum moment in the column = = = 15.90 MN-m
2 2

© 2003 by CRC Press LLC

Dynamic Analysis 3-15

3.3 Multi-Degree-of-Freedom System

The SDOF approach may not be applicable for complex structures such as multilevel frame struc-
tures and bridges with several supports. To predict the response of a complex structure, the structure
is discretized with several members of lumped masses. As the number of lumped masses increases,
the number of displacements required to define the displaced positions of all masses increases. The
response of a multi-degree-of-freedom (MDOF) system is discussed in this section.

Equation of Motion
The equation of motion of an MDOF system is similar to the SDOF system, but the stiffness k,
mass m, and damping c are matrices. The equation of motion to an MDOF system under ground
motion can be written as

[M]{ uúú} + [C]{ uú} + [K]{ u} = -[M] {B} uúúg (3.14)

The stiffness matrix [K] can be obtained from standard static displacement-based analysis
models and may have off-diagonal terms. The mass matrix [M] due to the negligible effect of mass
coupling can best be expressed in the form of tributary lumped masses to the corresponding
displacement degrees of freedom, resulting in a diagonal or uncoupled mass matrix. The damping
matrix [C] accounts for all the energy-dissipating mechanisms in the structure and may have off-
diagonal terms. The vector {B} is a displacement transformation vector that has values 0 and 1
to define degrees of freedom to which the earthquake loads are applied.

Free Vibration and Vibration Modes

To understand the response of MDOF systems better, we look at the undamped, free vibration of
an N degrees-of-freedom (N-DOF) system first.
Undamped Free Vibration
By setting [C] and uúúg to zero in Eq. (3.14), the equation of motion of undamped, free vibration
of an N-DOF system can be shown as:

[M]{ uúú} + [K]{ u} = 0 (3.15)

where [M] and [K] are n ¥ n square matrices.

Equation (3.15) could then be rearranged to

È ù
Í[K] - w n [M]ú { fn} = 0
2
(3.16)
Î û

where {fn } is the deflected-shape matrix. Solution to this equation can be obtained by setting

[K ] - w n 2 [ M] =0 (3.17)

The roots or eigenvalues of Eq. (3.17) will be the N natural frequencies of the dynamic system.
Once the natural frequencies ( w n ) are estimated, Eq. (3.16) can be solved for the corresponding
N independent, deflected-shape matrices (or eigenvectors), {fn } . In other words, a vibrating system
with N-DOFs will have N natural frequencies (usually arranged in sequence from smallest to largest),

© 2003 by CRC Press LLC

3-16 Bridge Engineering: Seismic Design

corresponding N natural periods Tn, and N natural mode shapes {fn } . These eigenvectors are
sometimes referred to as natural modes of vibration or natural mode shapes of vibration. It is
important to recognize that the eigenvectors or mode shapes represent only the deflected shape
corresponding to the natural frequency, not the actual deflection magnitude.
The N eigenvectors can be assembled in a single n ¥ n square matrix [F] , modal matrix, where
each column represents the coefficients associated with the natural mode. One of the important
aspects of these mode shapes is that they are orthogonal to each other. Stated mathematically,

{f } [ K ]{f } = 0 {f } [ M ]{f } = 0
T T
If w n π w r , n r
and n r
(3.18)

[K ] *
= [F ] [ K ] [F ]
T
(3.19)

[M ] *
= [F ] [ M ] [F ]
T
(3.20)

where [K] and [M] have off-diagonal elements, whereas K* [ ] and M*[ ] are diagonal matrices.
Damped Free Vibration
When damping of the MDOF system is included, the free-vibration response of the damped system
will be given by

[M]{uúú} + [C]{uú} + [K]{u} = 0 (3.21)

The displacements are first expressed in terms of natural mode shapes, and later they are multi-
plied by the transformed natural mode matrix to obtain the following expression:

[M ]{Yúú} + [C ]{Yú} + [K ]{Y} = 0

* * *
(3.22)

where M* [ ] [ ]
and K* are diagonal matrices given by Eqs. (3.19) and (3.20) and

[C ]
*
= [F] [C] [F]
T
(3.23)

[ ] [ ] [ ]
While M* and K* are diagonal matrices, C* may have off-diagonal terms. When C* [ ]
has off-diagonal terms, the damping matrix is referred to as a nonclassical or nonproportional
[ ]
damping matrix. When C* is diagonal, it is referred to as a classical or proportional damping
matrix. Classical damping is an appropriate idealization when similar damping mechanisms are
distributed throughout the structure. Nonclassical damping idealization is appropriate for the
analysis when the damping mechanisms differ considerably within a structural system.
Since most bridge structures have predominantly one type of construction material, bridge
structures could be idealized as a classical damping structural system. Thus, the damping matrix
of Eq. (3.22) will be a diagonal matrix for most bridge structures. And the equation of nth mode
shape or generalized nth modal equation is given by

Yúún + 2x n w n Yún + w 2 Yn = 0 (3.24)

Equation (3.24) is similar to Eq. (3.7) of an SDOF system. Also, the vibration properties of each
mode can be determined by solving Eq. (3.24).

© 2003 by CRC Press LLC

Dynamic Analysis 3-17

FIGURE 3.15 Rayleigh damping variation with natural frequency.

Rayleigh Damping
The damping of a structure is related to the amount of energy dissipated during its motion. It could
be assumed that a portion of the energy is lost due to the deformations, and thus damping could
be idealized as proportional to the stiffness of the structure. Another mechanism of energy dissi-
pation could be attributed to the mass of the structure, and thus damping idealized as proportional
to the mass of the structure. In Rayleigh damping, it is assumed that the damping is proportional
to the mass and stiffness of the structure.

[C] = ao [M] + a1 [K] (3.25)

The generalized damping of the nth mode is then given by

Cn = ao Mn + a1Kn (3.26)

Cn = ao Mn + a1 w n 2 Mn (3.27)

Cn
xn = (3.28)
2 Mnw n

ao 1 a
xn = + 1 wn (3.29)
2 wn 2

Figure 3.15 shows the Rayleigh damping variation with natural frequency. The coefficients ao and
a1 can be determined from specified damping ratios at two independent dominant modes (say, ith
and jth modes). Expressing Eq. (3.29) for these two modes will lead to the following equations:

ao 1 a1
xi = + w (3.30)
2 wi 2 i

ao 1 a
xj = + 1w (3.31)
2 wj 2 j

© 2003 by CRC Press LLC

3-18 Bridge Engineering: Seismic Design

When the damping ratio at both the ith and jth modes is the same and equals x , it can be shown that
2w i w j 2
ao = x a1 = x (3.32)
wi + w j wi + w j

It is important to note that the damping ratio at a mode between the ith and jth mode is less than
x . And in practical problems, the specified damping ratios should be chosen to ensure reasonable
values in all the mode shapes that lie between the ith and jth mode shapes.

Modal Analysis and Modal Participation Factor

In previous sections, we have discussed the basic vibration properties of an MDOF system.
Now, we will look at the response of an MDOF system to earthquake ground motion. The basic
equation of motion of the MDOF for an earthquake ground motion given by Eq. (3.14) is
repeated here:

[M]{uúú} + [C]{uú} + [K]{u} = -[M] {B} uúúg

The displacement is first expressed in terms of natural mode shapes, and later it is multiplied by
the transformed natural mode matrix to obtain the following expression:

[M ]{Yúú} + [C ]{Yú} + [K ]{Y} = - [F] [M]{B} uúú

* * * T
g
(3.33)

And the equation of the nth mode shape is given by

Mn* Yúún + 2x n w n Mn*Yún + w 2 Mn* Yn = Ln uúúg (3.34)

Mn* = {fn } [M]{fn }

T
where (3.35)

Ln = - {fn } [M] [B]

T
(3.36)

The Ln is referred to as the modal participation factor of the nth mode.

By dividing Eq. (3.34) by Mn* , the generalized modal equation of the nth mode becomes

Ê L ˆ
Yúún + 2x n w n Yún + w 2 Yn = Á n* ˜ uúúg (3.37)
Ë Mn ¯

Equation (3.34) is similar to the equation motion of an SDOF system, and thus Yn can be
determined by using methods similar to those described for SDOF systems. Once Yn is established,
the displacement due to the nth mode will be given by un (t ) = fn Yn (t ) . The total displacement due
to combination of all mode shapes can then be determined by summing up all displacements for
each mode and is given by

u( t ) = Â f Y (t )n n
(3.38)

© 2003 by CRC Press LLC

Dynamic Analysis 3-19

FIGURE 3.16 Three-span continuous-frame bridge structure of MDOF example. (a) Schematic diagram; (b) lon-
gitudinal degree of freedom; (c) transverse degree of freedom; (d) rotational degree of freedom; (e) mode shape 1;
(f) mode shape 2; (g) mode shape 3.

This approach is sometimes referred to as the classical mode superposition method. Similar to
the estimation of the total displacement, the element forces can also be estimated by adding the
element forces for each mode shape.

Example of an MDOF System

Given
The bridge shown in Figure 3.16 is a three-span continuous-frame structure. Details of the
bridge are as follows: span lengths are 18.3, 24.5, and 18.3 m.; column length is 9.5 m; area of
superstructure is 5.58 m2; moment of inertia of superstructure is 70.77 m4; moment of inertia
of column is 0.218 m4; modulus of elasticity of concrete is 20,700 MPa. Determine the vibration
modes and frequencies of the bridge.

Solution
As shown in Figures 3.16b, c, and d, five degrees of freedom are available for this structure. Stiffness
and mass matrices are estimated separately, and the results are given here.

© 2003 by CRC Press LLC

3-20 Bridge Engineering: Seismic Design

È126318588 0 0 0 0 ù
Í ú
Í 0 1975642681 -1194370500 -1520122814 -14643288630 ú
Í ú
[K ] = Í 0 -1194370500 1975642681 14643288630 1520122814 ú
Í ú
Í 0 -1520122814 14643288630 479327648712 119586857143 ú
Í ú
ÍÎ 0 -14643288630 1520122814 119586857143 479327648712 úû

È81872 0 0 0 0ù
Í ú
Í 0 286827 0 0 0ú
Í ú
[ M] = Í 0 0 286827 0 0ú
Í ú
Í 0 0 0 0 0ú
Í ú
ÍÎ 0 0 0 0 0 úû

The condensation procedure will eliminate the rotational degrees of freedom and will result in
three degrees of freedom. (The condensation procedure is performed separately, and the result is
given here.) The equation of motion of free vibration of the structure is

[ M ]{uúú} + [ K ]{u} = {0}

Substituting condensed stiffness and mass matrices into the above equation gives

È81872 0 0 ù Ïuúú1 ¸ È126318588 0 0 ù Ïu1 ¸ Ï0 ¸

Í 0 Ô Ô Ô Ô Ô Ô
Í 286827 0 úú Ìuúú2 ý + ÍÍ 0 1975642681 -1194370500 úú Ìu2 ý = Ì0 ý
Ô Ô 1975642681 ûú ÔÓu3 Ôþ ÔÓ0 Ôþ
ÎÍ 0 0 286827úû Óuúú3 þ ÍÎ 0 -1194370500

The above equation can be rearranged in the following form:

1
[ M ]-1[ K ] { f} = { f}
w2

Substitution of appropriate values in the above expression gives the following:

È 1 0 0 ù
Í 818172 ú È126318588 0 0 ù Ï f1n ¸ Ï f1n ¸
1 Í ú Í Ô Ô Ô Ô
-1215625977úú Ìf2 n ý = Ìf2 n ý
1
Í 0 0 ú Í 0 1518171572
w 2n Í 286827
Í 0 1 úú ÍÎ 0 -1215625977 1518171572 úû ÔÓf3n Ôþ ÔÓf3n Ôþ
0
Î 286827 û

È154.39 0 0 ù Ï f1n ¸ Ï f1n ¸

1 Í Ô Ô Ô Ô
0 5292.9 -4238.2 úú Ìf2 n ý = Ìf2 n ý
w 2n Í
ÍÎ 0 -4238.2 5292.9 úû ÔÓf3n Ôþ ÔÓf3n Ôþ

By assuming different vibration modes, natural frequencies of the structure can be estimated.

© 2003 by CRC Press LLC

Dynamic Analysis 3-21

Substitution of vibration mode {1 0 0} will result in the first natural frequency.

T

È154.39 0 0 ù Ï1 ¸ È154.39ù Ï1 ¸
1 Í Ô Ô 1 Í Ô Ô
0 5292.9 -4238.2 ú Ì0 ý = 2 Í 0 úú = Ì0 ý
ú
w 2n Í w
ÍÎ 0 -4238.2 5292.9 ûú ÔÓ0 Ôþ n
ÍÎ 0 úû ÔÓ0 Ôþ

Thus, w 2n = 154.39 and w n = 12.43 rad / s

By substituting the vibration modes of {0 1 1} and {0 1 – 1} in the above expression, the
T T

other two natural frequencies are estimated as 32.48 and 97.63 rad/s.

Multiple-Support Excitation
So far we have assumed that all supports of a structural system undergo the same ground motion.
This assumption is valid for structures with foundation supports close to each other. However, for
long-span bridge structures, supports may be widely spaced. As described previously, earth motion
at a location depends on the localized soil layer and the distance from the epicenter. Thus, bridge
structures with supports that lie far from each other may experience different earth excitation. For
example, Figure 3.17c, d, and e shows the predicted earthquake motions at Pier W3 and Pier W6
of the San Francisco–Oakland Bay Bridge (SFOBB) in California. The distance between Pier W3
and Pier W6 of the SFOBB is approximately 1411 m. These excitations are predicted by the California
Department of Transportation by considering the soil and rock properties in the vicinity of the
SFOBB and expected earth movements at the San Andreas and Hayward faults. Note that the earth
motions at Pier W3 and Pier W6 are very different. Furthermore, Figure 3.17c, d, and e indicates
that the earth motion not only varies with the location, but also varies with direction. Thus, to
evaluate the response of long, multiply supported, and complicated bridge structures, use of the
actual earthquake excitation at each support is recommended.
The equation of motion of a multisupport excitation would be similar to Eq. (3.14), but the only
difference is now that {B} uúúg is replaced by a displacement array { uúúg } . And the equation of motion
for the multisupport system becomes

[M]{ uúú} + [C]{ uú} + [K]{ u} = -[M] { uúúg } (3.39)

where { uúúg } has the acceleration at each support location and has zero value at nonsupport
locations. By using the uncoupling procedure described in the previous sections, the modal equation
of the nth mode can be written as

Ng

ÂM
Ln
Yúún + 2x n w n Yún + w 2 Yn = - *
uúúg (3.40)
l =1 n

where Ng is the total number of externally excited supports.

The deformation response of the nth mode can then be determined as described in previous
sections. Once the displacement responses of the structure for all the mode shapes are estimated,
the total dynamic response can be obtained by combining the displacements.

Time-History Analysis
When the structure enters the nonlinear range, or has nonclassical damping properties, modal
analysis cannot be used. A numerical integration method, sometimes referred to as time-history
analysis, is required to get more accurate responses of the structure.

© 2003 by CRC Press LLC

3-22 Bridge Engineering: Seismic Design

FIGURE 3.17 San Francisco–Oakland Bay Bridge. (a) Vicinity map; (b) general plan elevation; (c) longitudinal
motion at rock level; (d) transverse motion at rock level; (e) vertical motion at rock level; (f) displacement response
at top of Pier W3.

In a time-history analysis, the timescale is divided into a series of smaller steps, dt. Let us say the
response at ith time interval has already been determined and is denoted by ui , uúi , uúúi . Then, the
response of the system at ith time interval will satisfy the equation of motion (Eq. [3.39]).

[M]{ uúúi } + [C]{ uúi } + [K]{ ui } = -[M]{ uúúgi } (3.41)

© 2003 by CRC Press LLC

Dynamic Analysis 3-23

The time-stepping method enables us to step ahead and determine the responses ui +1, uúi +1, uúúi +1 at
the i + 1th time interval by satisfying Eq. (3.39). Thus, the equation of motion at i + 1th time interval
will be

[M]{ uúúi+1} + [C]{ uúi+1} + [K]{ ui+1} = -[M]{ uúúgi+1} (3.42)

Equation (3.42) needs to be solved prior to proceeding to the next time step. By stepping through
all the time steps, the actual response of the structure can be determined at all time instants.
Example of Time-History Analysis
Pier W3 of the SFOBB was modeled using the ADINA [13] program, and nonlinear analysis was
performed using the displacement time histories. The displacement time histories in three directions
are applied at the bottom of Pier W3, and the response of the Pier W3 was studied to estimate the
demand on Pier W3. One of the results, the displacement response at the top of Pier W3, is shown
in Figure 3.17f.

3.4 Response Spectrum Analysis

Response spectrum analysis is an approximate method of dynamic analysis that gives the maximum
response (acceleration, velocity, or displacement) of an SDOF system with the same damping ratio,
but with different natural frequencies, respond to a specified seismic excitation. Structural models
with n degrees of freedom can be transformed to n single-degree systems, and response spectra
principles can be applied to systems with many degrees of freedom. For most ordinary bridges, a
complete time history is not required. Because the design is generally based on the maximum
earthquake response, response spectrum analysis is probably the most common method used in
design offices to determine the maximum structural response due to transient loading. In this
section, we will discuss basic procedures of response spectrum analysis for bridge structures.

Single-Mode Spectral Analysis

Single-mode spectral analysis is based on the assumption that earthquake design forces for structures
respond predominantly in the first mode of vibration. This method is most suitable to regular linear
elastic bridges to compute the forces and deformations, but is not applicable to irregular bridges
(unbalanced spans, unequal stiffness in the columns, etc.) because higher modes of vibration
significantly affect the distribution of the forces and resulting displacements. This method can be
applied to both continuous and noncontinuous bridge superstructures in either the longitudinal
or transverse direction. Foundation flexibility at the abutments can be included in the analysis.
Single-mode analysis is based on Rayleigh’s energy method — an approximate method that
assumes a vibration shape for a structure. The natural period of the structure is then calculated by
equating the maximum potential and kinetic energies associated with the assumed shape. The
inertial forces pe ( x ) are calculated using the natural period, and the design forces and displacements
are then computed using static analysis. The detailed procedure can be described in the following
steps:
1. Apply uniform loading po over the length of the structure and compute the corresponding
static displacements us ( x ) . The structure deflection under earthquake loading, us ( x, t ) , is
then approximated by the shape function, us ( x ), multiplied by the generalized amplitude
function, u(t ) , which satisfies the geometric boundary conditions of the structural system.
This dynamic deflection is shown as

u( x, t ) = us ( x ) u(t ) (3.43)

© 2003 by CRC Press LLC

3-24 Bridge Engineering: Seismic Design

2. Calculate the generalized parameters a, b, and g using the following equations:

Ú
a = us ( x ) dx (3.44)

b=
Ú w( x) u ( x) dxs
(3.45)

g=
Ú w( x) [u ( x)] dx s
2
(3.46)

where w( x ) is the weight of the dead load of the bridge superstructure and tributary
substructure.
3. Calculate the period Tn

g
Tn = 2p (3.47)
po ga

where g is acceleration of gravity (mm/s2).

4. Calculate the static loading pe ( x ) which approximates the inertial effects associated with the
displacement us ( x ) using the ARS curve or the following equation [4]:

b Csm
pe ( x ) = w( x ) us ( x ) (3.48)
g

1.2 AS
Csm = (3.49)
Tm 2 / 3

where Csm is the dimensionless elastic seismic response coefficient; A is the acceleration
coefficient from the acceleration coefficient map; S is the dimensionless soil coefficient based
on the soil profile type; Tn is the period of the structure as determined above; pe ( x ) is the
intensity of the equivalent static seismic loading applied to represent the primary mode of
vibration (N/mm).
5. Apply the calculated loading pe ( x ) to the structure as shown in Figure 3.18 and compute
the structure deflections and member forces.
This method is an iterative procedure, and the previous calculations are used as input parameters
for the new iteration leading to a new period and deflected shape. The process is continued until
the assumed shape matches the fundamental mode shape.

Uniform-Load Method
The uniform-load method is essentially an equivalent static method that uses the uniform
lateral load to compute the effect of seismic loads. For simple bridge structures with relatively
straight alignment, small skew, balanced stiffness, relatively light substructure, and no hinges,
the uniform-load method may be applied to analyze the structure for seismic loads. This
method is not suitable for bridges with stiff substructures such as pier walls. This method
assumes continuity of the structure and distributes earthquake force to all elements of the
bridge and is based on the fundamental mode of vibration in either a longitudinal or transverse

© 2003 by CRC Press LLC

Dynamic Analysis 3-25

FIGURE 3.18 Single-mode spectral analysis method. (a) Plan view of a bridge subjected to transverse earthquake
motion; (b) displacement function describing the transverse position of the bridge deck; (c) deflected shape due to
uniform static loading; (d) transverse free vibration of the bridge in assumed mode shape; (e) transverse loading (f)
longitudinal loading.

direction [5]. The period of vibration is taken as that of an equivalent single mass–spring
oscillator. The maximum displacement that occurs under the arbitrary uniform load is used
to calculate the stiffness of the equivalent spring. The seismic elastic response coefficient Csm
or the ARS curve is then used to calculate the equivalent uniform seismic load, using which
the displacements and forces are calculated. The following steps outline the uniform-load
method:

© 2003 by CRC Press LLC

3-26 Bridge Engineering: Seismic Design

FIGURE 3.19 Structure idealization and deflected shape for uniform load method. (a) Structure idealization; (b)
deflected shape with maximum displacement of 1 mm.

1. Idealize the structure into a simplified model and apply a uniform horizontal load ( po ) over
the length of the bridge as shown in Figure 3.19. It has units of force/unit length and may
be arbitrarily set equal to 1 N/mm.
2. Calculate the static displacements us ( x ) under the uniform load po using static analysis.
3. Calculate the maximum displacement us,max and adjust it to 1 mm by adjusting the uniform
load po .
4. Calculate bridge lateral stiffness K using the following equation:

po L
K = (3.50)
us,max

where L is total length of the bridge (mm) and us,max is maximum displacement (mm).
5. Calculate the total weight W of the structure, including structural elements and other relevant
loads such as pier walls, abutments, columns, and footings, by

W=
Ú w( x)dx (3.51)

where w(x) is the nominal, unfactored dead load of the bridge superstructure and tributary
substructure.
6. Calculate the period of the structure Tn using the following equation:

2p W
Tn = (3.52)
31.623 gK

where g is acceleration of gravity (m/s2).

© 2003 by CRC Press LLC

Dynamic Analysis 3-27

7. Calculate the equivalent static earthquake force pe using the ARS curve or using the following
equation:

CsmW
pe = (3.53)
L

8. Calculate the structure deflections and member forces by applying pe to the structure.

Multimode Spectral Analysis

The multimode spectral analysis method is more sophisticated than single-mode spectral analysis
and is very effective in analyzing the response of more complex linear elastic structures to an
earthquake excitation. This method is appropriate for structures with irregular geometry, mass, or
stiffness. These irregularities induce coupling in three orthogonal directions within each mode of
vibration. Also, for these bridges, several modes of vibration contribute to the complete response
of the structure. A multimode spectral analysis is usually done by modeling the bridge structure
consisting of three-dimensional frame elements with structural mass lumped at various locations
to represent the vibration modes of the components. Usually, five elements per span are sufficient
to represent the first three modes of vibration. To capture the i th mode of vibration, a general rule
of thumb is that the span should have at least (2i - 1) elements. For long-span structures, many
more elements should be used to capture all the contributing modes of vibration. To obtain a
reasonable response, the number of modes should be equal to at least three times the number of
spans. This analysis is usually performed with a dynamic analysis computer program such as
ADINA [13], GTSTRUDL [14], SAP2000 [15], ANSYS [16], or NASTRAN [17]. For bridges with
outrigger bents, C-bents, and single-column bents, rotational moment of inertia of the superstruc-
ture should be included. Discontinuities at the hinges and abutments should be included in the
model. The columns and piers should have intermediate nodes at quarter points in addition to the
nodes at the ends of the columns.
By using the programs mentioned above, frequencies, mode shapes, member forces, and joint
displacements can be computed. The following steps summarize the equations used in the multi-
mode spectral analysis [5].
1. Calculate the dimensionless mode shapes {f } and corresponding frequencies
i w i by

È ù
Í[K] – w [M]ú{ u} = 0
2
(3.54)
Î û

where

ui = Âf y = F y
j =1
j j i
(3.55)

y j = modal amplitude of jth mode; f j = shape factor of jth mode; F = mode-shape matrix.
The periods for ith mode can then be calculated by

2p
Ti = (i = 1, 2,º, n) (3.56)
wi

2. Determine the maximum absolute mode amplitude for the entire time history is given by

© 2003 by CRC Press LLC

3-28 Bridge Engineering: Seismic Design

{f } [M]{B} uúú
T
T 2 S (x , T )
Yi (t )max = i a 2i i i g
(3.57)
4p {f } [M]{f }
T
i i

where Sa (xi , Ti ) = gCsm is the acceleration response spectral value; Csm is the elastic seismic
response coefficient for mode m = 1.2 AS Tn2 / 3 ; A is the acceleration coefficient from the
acceleration coefficient map; S is the dimensionless soil coefficient based on the soil profile
type; Tn is the period of the nth mode of vibration.
3. Calculate the value of any response quantity Z(t) (shear, moment, displacement) using the
following equation:

Z (t ) = Â A Y (t )
i =1
i i
(3.58)

where coefficients Ai are functions of mode shape matrix (F) and force-displacement
relation-ships.
4. Compute the maximum value of Z(t) during an earthquake using the mode-combination
methods described in the next section.
Modal Combination Rules
The mode combination method is a very useful tool for analyzing bridges with a large number of
degrees of freedom. In a linear structural system, maximum response can be estimated by mode
combination after calculating natural frequencies and mode shapes of the structure using free-
vibration analysis. The maximum response cannot be computed by adding the maximum response
of each mode because different modes attain their maximum values at different times. The absolute
sum of the individual modal contributions provides an upper bound that is generally very conser-
vative and not recommended for design. There are several different empirical or statistical methods
available to estimate the maximum response of a structure by combining the contributions of
different modes of vibrations in a spectral analysis. Two commonly used methods are the square
root of sum of squares (SRSS) and the complete quadratic combination (CQC).
For an undamped structure, the results computed using the CQC method are identical to those
using the SRSS method. For structures with closely spaced dominant mode shapes, the CQC method
is precise whereas SRSS estimates inaccurate results. Closely spaced modes are those within 10% of
each other in terms of natural frequency. The SRSS method is suitable for estimating the total
maximum response for structures with well-spaced modes. Theoretically, all mode shapes must be
included to calculate the response, but fewer mode shapes can be used when the corresponding
mass participation is over 85% of the total structure mass. In general, the factors considered to
determine the number of modes required for the mode combination are dependent on the structural
characteristics of the bridge, the spatial distribution, and the frequency content of the earthquake
loading. The following list [14] summarizes several commonly used mode-combination methods
to compute the maximum total response. The variable Z represents the maximum value of some
response quantity (displacement, shear, etc.), Zi is the peak value of that quantity in the i th mode,
and N is the total number of contributing modes.
1. Absolute Sum: The absolute sum is the sum of the modal contributions:

Z= ÂZ
i =1
i
(3.59)

© 2003 by CRC Press LLC

Dynamic Analysis 3-29

2. SRSS or Root Mean Square (RMS) Method: This method computes the maximum by taking
the square root of the sum of squares of the modal contributions:

1/ 2
È N
ù
Z=Í
ÍÎ
Âi =1
Z ú
úû
i
2
(3.60)

3. Peak Root Mean Square (PRMS): Absolute value of the largest modal contribution is added
to the root mean square of the remaining modal contributions:

Z j = max Z i (3.61)

1/ 2
È N
ù
Z=Í
ÍÎ
Âi =1
Z úi
úû
2
+ Zj with i π j (3.62)

4. CQC: Cross-correlations between all modes are considered:

N N 1§2

Z = ÂÂZ
i=1 j=1
i r ij Z j (3.63)

rij =
(
8 xi x j xi + rx j r 3/ 2 ) (3.64)
(1 - r )2 2
+ 4xi x j r 1 + r ( 2
) + 4 (x 2
i )
+ x 2j r 2

where

wj
r = (3.65)
wi

5. Nuclear Regulatory Commission Grouping Method: This method is similar to the RMS method
with additional accounting for groups of modes whose frequencies are within 10%.

1/ 2
È N G e e ù
Z = Í
Í
Î
ÂZ
i =1
i
2
+ Â Â Â
g =1 n=s m=s
Z ¥ Z ú
g
n
ú
û
g
m nπm (3.66)

where G is number of groups; s is mode shape number where the g th group starts; e is
mode shape number where the g th group ends; and Zig is the i th modal contribution in
the g th group.
6. Nuclear Regulatory Commission Ten Percent Method: This method is similar to the RMS
method with additional accounting for all modes whose frequencies are within 10%.

1/ 2
È N
ù
Z=Í
ÍÎ
ÂZ i =1
i
2
+2 Â Zn Zm ú
úû
(3.67)

© 2003 by CRC Press LLC

3-30 Bridge Engineering: Seismic Design

The additional terms must satisfy

wn - wm
£ 0 .1 for 0 .1 £ m £ n £ N (3.68)
wm

7. Nuclear Regulatory Commission Double-Sum Method: This method is similar to the CQC
method.
1/ 2
È N N ù
Z = Í
ÍÎ Â Â ZZ
i =1 j =1
i j e ij ú
úû
(3.69)

( )
-1
È ÏÔ w ¢ - w ¢ ¸Ôù
e ij = Í1 + Ì ýú
i j

( )
(3.70)
Í ÔÓ xi¢w i + x¢j w j Ôþúû
Î

[ ]
1/ 2
w ¢i = w 1 - xi2 (3.71)

2
xi¢ = xi + (3.72)
td w i

where td is the duration of support motion.

Combination Effects
Effects of ground motions in two orthogonal horizontal directions should be combined while
designing bridges with simple geometric configurations. For bridges with long spans, outrigger
bents, and cantilever spans, or where effects due to vertical input are significant, vertical input
should be included in the design along with two orthogonal horizontal inputs. When bridge
structures are analyzed independently along each direction using response spectra analysis, then
responses are combined either using methods, such as the SRSS combination rule as mentioned in
the previous section, or using the alternative method described below. For structures designed using
equivalent static analysis or modal analysis, seismic effects should be determined using the following
alternative method for the following load cases:
1. Seismic load case 1: 100% Transverse + 30% Longitudinal + 30% Vertical
2. Seismic load case 2: 30% Transverse + 100% Longitudinal + 30% Vertical
3. Seismic load case 3: 30% Transverse + 30% Longitudinal + 100% Vertical
For structures designed using time-history analysis, the structure response is calculated using the
input motions applied in orthogonal directions simultaneously. Where this is not feasible, the above
alternative procedure can be used to combine the independent responses.

Multiple-Support Response Spectrum Method

Records from recent earthquakes indicate that seismic ground motions can significantly vary at
different support locations for multiply supported long structures. When different ground motions
are applied at various support points of a bridge structure, the total response can be calculated by
superposition of responses due to independent support input. This analysis involves combination
of dynamic response from single-input and pseudo-static response resulting from the motion of
the supports relative to each other. The combination effects of dynamic and pseudo-static forces

© 2003 by CRC Press LLC

Dynamic Analysis 3-31

due to multiple support excitation on a bridge depend on the structural configuration of the bridge
and the ground motion characteristics. Recently, Kiureghian et al. [7] presented a comprehensive
study on the multiple-support response spectrum (MSRS) method based on fundamental principles
of stationary random vibration theory for seismic analysis of multiply supported structures that
accounts for the effects of variability between the support motions. Using the MSRS combination
rule, the response of a linear structural system subjected to multiple-support excitation can be computed
directly in terms of conventional response spectra at the support degrees of freedom and a coherency
function describing the spatial variability of the ground motion. This method accounts for the three
important effects of ground motion spatial variability, namely, the incoherence effects, the wave-passage
effect, and the site-response effect. These three components of ground motion spatial variability can
strongly influence the response of multiply supported bridges and may amplify or deamplify the
response by one order of magnitude. Two important limitations of this method are nonlinearities in
the bridge structural components and/or connections and the effects of soil–structure interaction. This
method is an efficient, accurate, and versatile solution and requires less computational time than a true
time history analysis. Following are the steps that describe the MSRS analysis procedure.
1. Determine the necessity of variable-support motion analysis: Three factors that influence the
response of the structure under multiple-support excitation are the distance between the
supports of the structure, the rate of variability of the local soil conditions, and the stiffness
of the structure. The first factor, the distance between the supports, influences the incoherence
and wave-passage effects. The second factor, the rate of variability of the local conditions,
influences the site response. The third factor, the stiffness of the superstructure, plays an
important role in determining the necessity of variable-support motion analysis. Stiff struc-
tures such as box-girder bridges may generate large internal forces under variable-support
motion, whereas flexible structures such as suspension bridges easily conform to the variable-
support motion.
2. Determine the frequency response function for each support location: Programs such as SHAKE
[18] can be used to develop these functions using borehole data and time-domain site
response analysis. Response spectra plots, peak ground displacements in three orthogonal
directions for each support location, and a coherency function for each pair of degrees of
freedom are required to perform the MSRS analysis. The comprehensive report by Kiureghian
[7] provides all the formulas required to account for the effect of nonlinearity in the soil
behavior and the site frequency involving the depth of the bedrock.
3. Calculate the structural properties: Properties such as effective modal frequencies, damping
ratios, influence coefficients, and effective modal participation factors ( w i , xi , ak , and bki )
are to be computed externally and provided as input.
4. Determine the response spectra plots, peak ground displacements in three directions, and a
coherency function for each pair of support degrees of freedom required to perform MSRS analysis:
Three components of the coherency function are incoherence, wave-passage effect, and site-
response effect. Analysis by an array of recordings is used to determine the incoherence
component. The models for this empirical method are widely available [19]. Parameters such
as shear wave velocity, the direction of propagation of seismic waves, and the angle of
incidence are used to calculate the wave-passage effect. The frequency-response function
determined in the previous steps is used to calculate the site-response component.

3.5 Inelastic Dynamic Analysis

Equations of Motion
Inelastic dynamic analysis is usually performed for the safety evaluation of important bridges to
determine the inelastic response of bridges when subjected to design earthquake ground motions.

© 2003 by CRC Press LLC

3-32 Bridge Engineering: Seismic Design

Inelastic dynamic analysis provides a realistic measure of response because the inelastic model
accounts for the redistribution of internal actions due to the nonlinear force displacement behav-
ior of the components [20–25]. Inelastic dynamic analysis considers nonlinear damping, stiffness,
load deformation behavior of members including soil, and mass properties. A step-by-step inte-
gration procedure is the most powerful method used for nonlinear dynamic analysis. One impor-
tant assumption of this procedure is that acceleration varies linearly while the properties of the
system such as damping and stiffness remain constant during the time interval. By using this
procedure, a nonlinear system is approximated as a series of linear systems and the response is
calculated for a series of small equal intervals of time Dt and equilibrium is established at the
beginning and end of each interval.
The accuracy of this procedure depends on the length of the time increment Dt . This time
increment should be small enough to consider the rate of change of loading p(t ) , nonlinear
damping and stiffness properties, and the natural period of the vibration. An SDOF system and its
characteristics are shown in Figure 3.20. The characteristics include spring and damping forces,

FIGURE 3.20 Definition of a nonlinear dynamic system. (a) Basic SDOF structure; (b) force equilibrium; (c)
nonlinear damping; (d) nonlinear stiffness; (e) applied load.

© 2003 by CRC Press LLC

Dynamic Analysis 3-33

forces acting on mass of the system, and arbitrary applied loading. The force equilibrium can be
shown as

fi (t ) + fd (t ) + fs (t ) = p(t ) (3.73)

and the incremental equations of motion for time t can be shown as

m Duúú(t ) + c(t ) Duú(t ) + k (t ) Du(t ) = Dp(t ) (3.74)

Current damping fd (t ) , elastic forces fs (t ) are then computed using the initial velocity uú(t ) ,
displacement values u(t ) , nonlinear properties of the system, damping c(t ) , and stiffness k (t ) for
that interval. New structural properties are calculated at the beginning of each time increment based
on the current deformed state. The complete response is then calculated by using the displacement
and velocity values computed at the end of each time step as the initial conditions for the next time
interval and repeating until the desired time.

Modeling Considerations
A bridge structural model should have sufficient degrees of freedom and proper selection of lin-
ear/nonlinear elements such that a realistic response can be obtained. Nonlinear analysis is usually
preceded by a linear analysis as a part of a complete analysis procedure to capture the physical and
mechanical interactions of seismic input and structure response. Output from the linear response
solution is then used to predict which nonlinearities will significantly affect the response and to
model them appropriately. In other words, engineers can justify the effect of each nonlinear element
introduced at the appropriate locations and establish the confidence in the nonlinear analysis. While
discretizing the model, engineers should be aware of the trade-offs between the accuracy, compu-
tational time, and use of the information such as the regions of significant geometric and material
nonlinearities. Nonlinear elements should have material behavior to simulate the hysteresis relations
under reverse cyclic loading observed in the experiments.
The general issues in modeling of bridge structures include geometry, stiffness, mass distribution,
and boundary conditions. In general, abutments, superstructure, bent caps, columns and pier walls,
expansion joints, and foundation springs are the elements included in the structural model. The
mass distribution in a structural model depends on the number of elements used to represent the
bridge components. The model must be able to simulate the vibration modes of all components
contributing to the seismic response of the structure.
Superstructure: Superstructure and bent caps are usually modeled using linear elastic three-
dimensional beam elements. Detailed models may require nonlinear beam elements.
Columns and pier walls: Columns and pier walls are usually modeled using nonlinear beam
elements having response properties with a yield surface described by the axial load and biaxial
bending. Some characteristics of the column behavior include initial stiffness degradation due to
concrete cracking, flexural yielding at the fixed end of the column, strain hardening, and pinching
at the point of load reversal. Shear actions can be modeled using either linear or nonlinear load
deformation relationships for columns. For both columns and pier walls, torsion can be modeled
with linear elastic properties. For out-of-plane loading, flexural response of a pier wall is similar to
that of columns, whereas for in-plane loading the nonlinear behavior is usually shear action.
Expansion joints: Expansion joints can be modeled using gap elements that simulate the nonlinear
behavior of the joint. The variables include initial gap, shear capacity of the joint, and nonlinear
load deformation characteristics of the gap.
Foundations and abutments: Foundations are typically modeled using nonlinear spring elements
to represent the translational and rotational stiffness of the foundations to represent the expected

© 2003 by CRC Press LLC

3-34 Bridge Engineering: Seismic Design

behavior during a design earthquake. Abutments are modeled using nonlinear spring and gap
elements to represent the soil action, stiffness of the pile groups, and gaps at the seat.

3.6 Summary
This chapter has presented the basic principles and methods of dynamic analysis for the seismic
design of bridges. Response spectrum analysis — the SDOF or equivalent SDOF-based equivalent
static analysis — is efficient, convenient, and most frequently used for ordinary bridges with simple
configurations. Elastic dynamic analysis is required for bridges with complex configurations. A
multisupport response spectrum analysis recently developed by Kiureghian et al. [7] using a lumped-
mass beam element mode may be used in lieu of an elastic time-history analysis.
Inelastic response spectrum analysis is a useful concept, but the current approaches apply only
to SDOF structures. An actual nonlinear dynamic time-history analysis may be necessary for some
important and complex bridges, but linearized dynamic analysis (dynamic secant stiffness analysis)
and inelastic static analysis (static push-over analysis) (Chapter 4) are the best possible alternatives
[8] for the most bridges.

References
1. Clough, R. W. and Penzien, J., Dynamics of Structures, 2nd ed., McGraw-Hill, New York, 1993.
2. Caltrans, Guidelines for Generation of Response — Spectrum-Compatible Rock Motion Time
History for Application to Caltrans Toll Bridge Seismic Retrofit Projects, the Caltrans Seismic
Advisor Board ad hoc Committee on Soil–Foundation-Structure Interaction, California Depart-
ment of Transportation, Sacramento, 1996.
3. Caltrans, Guidelines for Performing Site Response Analysis to Develop Seismic Ground Motions
for Application to Caltrans Toll Bridge Seismic Retrofit Projects, The Caltrans Seismic Advisor
Board ad hoc Committee on Soil–Foundation-Structure Interaction, California Department of
Transportation, Sacramento, 1996.
4. AASHTO, LRFD Bridge Design Specifications, American Association of State Highway and Trans-
portation Officials, Washington, D.C., 1994.
5. AASHTO, LRFD Bridge Design Specifications, 1996 Interim Version, American Association of State
Highway and Transportation Officials, Washington, D.C., 1996.
6. Wilson, E. L., der Kiureghian, A., and Bayom, E. P., A replacement for SSRS method in seismic
analysis, J. Earthquake Eng. Struct. Dyn., 9, 187, 1981.
7. Kiureghian, A. E., Keshishian, P., and Hakobian, A., Multiple Support Response Spectrum Analysis
of Bridges Including the Site-Response Effect and the MSRS Code, Report No. UCB/EERC-97/02,
University of California, Berkeley, 1997.
8. Powell, G. H., Concepts and Principles for the Application of Nonlinear Structural Analysis in
Bridge Design, Report No. UCB/SEMM-97/08, Department of Civil Engineering, University of
California, Berkeley, 1997.
9. Chmielewski, T., Kratzig, W. B., Link, M., Meskouris, K., and Wunderlich, W., Phenomena and
evaluation of dynamic structural responses, in Dynamics of Civil Engineering Structures, W. B.
Kratzig and H.-J. Niemann, Eds., A.A. Balkema, Rotterdams, 1996.
10. Chopra, A. K., Dynamics of Structures, Prentice-Hall, Englewood Cliffs, NJ, 1995.
11. Lindeburg, M., Seismic Design of Building Structures: A Professional’s Introduction to Earthquake
Forces and Design Details, Professional Publications, Belmont, CA, 1998.
12. Caltrans, Bridge Design Specifications, California Department of Transportation, Sacramento, CA,
1991.
13. ADINA, User’s Guide, Adina R&D, Inc., Watertown, MA, 1995.
14. GTSTRUDL, User’s Manual, Georgia Institute of Technology, Atlanta, 1996.
15. SAP2000, User’s Manual, Computers and Structures Inc., Berkeley, CA, 1998.

© 2003 by CRC Press LLC

Dynamic Analysis 3-35

16. ANSYS, User’s Manual, Vols. 1 and 2, Version 4.4, Swanson Analysis Systems, Inc., Houston, TX,
1989.
17. NASTRAN, User’s Manual, MacNeil Schwindler Corporation, Los Angeles, CA.
18. Idriss, I. M., Sun J. I., and Schnabel, P. B., User’s manual for SHAKE91: a computer program for
conducting equivalent linear seismic response analyses of horizontally layered soil deposits, Report
of Center for Geotechnical Modeling, Department of Civil and Environmental Engineering, Univer-
sity of California at Davis, 1991.
19. Abrahamson, N. A., Schneider, J. F., and Stepp, J. C., Empirical spatial coherency functions for
application to soil-structure interaction analysis, Earthquake Spectra, 7, 1991.
20. Imbsen & Associates, Seismic Design of Highway Bridges, Sacramento, CA, 1992.
21. Priestly, M. J. N., Seible, F., and Calvi, G. M., Seismic Design and Retrofit of Bridges, John Wiley &
Sons, New York, 1996.
22. Bathe, K.-J., Finite Element Procedures in Engineering Analysis, 2nd ed., Prentice-Hall, Englewood
Cliffs, NJ, 1996.
23. ATC 32, Improved Seismic Design Criteria for California Bridges: Provisional Recommendations,
Applied Technology Council, 1996.
24. Buchholdt, H. A., Structural Dynamics for Engineers, Thomas Telford, London, 1997.
25. Paz, M., Structural Dynamics — Theory and Computation, 3rd ed., Van Nostrand Reinhold, New
York, 1991.

© 2003 by CRC Press LLC