4

Nonlinear Analysis of
Bridge Structures

4.1 Introduction .................................................................4-1

4.2 Analysis Classification and General Guidelines.........4-2
Classifications • General Guidelines
4.3 Geometric Nonlinearity Formulation ........................4-4
Two-Dimensional Members • Three-Dimensional
Members
4.4 Material Nonlinearity Formulations...........................4-9
Structural Concrete • Structural and
Reinforcement Steel
4.5 Nonlinear Section Analysis........................................4-17
Basic Assumptions and Formulations • Modeling and
Solution Procedures • Yield Surface Equations
4.6 Nonlinear Frame Analysis .........................................4-21
Elastic–Plastic Hinge Analysis • Refined Plastic Hinge
Analysis • Distributed Plasticity Analysis
Mohammed Akkari 4.7 Practical Applications ................................................4-22
California Department Displacement-Based Seismic Design • Static
of Transportation
Push-Over Analysis • Example 4.1 — Reinforced-
Lian Duan Concrete Multicolumn Bent Frame with
California Department P-D Effects • Example 4.2 — Steel Multicolumn Bent
of Transportation Frame Seismic Evaluation

4.1 Introduction
In recent years, nonlinear bridge analysis has gained a greater momentum because of the need to
assess inelastic structural behavior under seismic loads. Common seismic design philosophies for
ordinary bridges allow some degree of damage without collapse. To control and evaluate damage,
a postelastic nonlinear analysis is required. A nonlinear analysis is complex and involves many
simplifying assumptions. Engineers must be familiar with those complexities and assumptions to
design bridges that are safe and economical.
Many factors contribute to the nonlinear behavior of a bridge. These include factors such as
material inelasticity, geometric or second-order effects, nonlinear soil–foundation–structure inter-
action, gap opening and closing at hinges and abutment locations, time-dependent effects due to
concrete creep and shrinkage, etc. The subject of nonlinear analysis is extremely broad and cannot
be covered in detail in this single chapter. Only material and geometric nonlinearities as well as

4-1

© 2003 by CRC Press LLC

4-2 Bridge Engineering: Seismic Design

FIGURE 4.1 Lateral load–displacement curves of a frame.

some of the basic formulations of nonlinear static analysis with their practical applications to seismic
bridge design will be presented here. The reader is referred to the many excellent papers, reports,
and books [1–8] that cover this type of analysis in more detail.
In this chapter, some general guidelines for nonlinear static analysis are presented. These are
followed by discussion of the formulations of geometric and material nonlinearities for section and
frame analysis. Two examples are given to illustrate the applications of static nonlinear push-over
analysis in bridge seismic design.

4.2 Analysis Classification and General Guidelines

Engineers use structural analysis as a fundamental tool to make design decisions. It is important
that engineers have access to several different analysis tools and understand their development
assumptions and limitations. Such an understanding is essential to select the proper analysis tool
to achieve the design objectives.
Figure 4.1 shows lateral load vs. displacement curves of a frame using several structural analysis
methods. Table 4.1 summarizes basic assumptions of those methods. It can be seen from Figure 4.1
that the first-order elastic analysis gives a straight line and no failure load. A first-order inelastic
analysis predicts the maximum plastic load-carrying capacity on the basis of the undeformed
geometry. A second-order elastic analysis follows an elastic buckling process. A second-order inelas-
tic analysis traces load–deflection curves more accurately.

Classifications
Structural analysis methods can be classified on the basis of different formulations of equilibrium,
the constitutive and compatibility equations as discussed below.
Classification Based on Equilibrium and Compatibility Formulations
First-order analysis: An analysis in which equilibrium is formulated with respect to the unde-
formed (or original) geometry of the structure. It is based on small strain and small displace-
ment theory.

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-3

TABLE 4.1 Structural Analysis Methods

Features
Constitutive Equilibrium Geometric
Methods Relationship Formulation Compatibility

Elastic Elastic Original undeformed Small strain and small

Rigid–plastic Rigid plastic geometry displacement
First-order Elastic–plastic hinge Elastic perfectly plastic
Distributed plasticity Inelastic

Elastic Elastic Deformed structural Small strain and

Rigid–plastic Rigid–plastic geometry (P-D and P-d) moderate rotation
Second-order Elastic–plastic hinge Elastic perfectly plastic (displacement may be
Distributed plasticity Inelastic large)

True large Elastic Elastic Deformed structural Large strain and large
displacement Inelastic Inelastic geometry deformation

FIGURE 4.2 Second-order effects.

Second-order analysis: An analysis in which equilibrium is formulated with respect to the deformed
geometry of the structure. A second-order analysis usually accounts for the P-D effect (influ-
ence of axial force acting through displacement associated with member chord rotation) and
the P-d effect (influence of axial force acting through displacement associated with member
flexural curvature) (see Figure 4.2). It is based on small strain and small member deformation,
but moderate rotations and large displacement theory.
True large deformation analysis: An analysis for which large strain and large deformations are
taken into account.
Classification Based on Constitutive Formulation
Elastic analysis: An analysis in which elastic constitutive equations are formulated.
Inelastic analysis: An analysis in which inelastic constitutive equations are formulated.
Rigid–plastic analysis: An analysis in which elastic rigid–plastic constitutive equations are
formulated.
Elastic–plastic hinge analysis: An analysis in which material inelasticity is taken into account by
using concentrated “zero-length” plastic hinges.
Distributed plasticity analysis: An analysis in which the spread of plasticity through the cross
sections and along the length of the members are modeled explicitly.

© 2003 by CRC Press LLC

4-4 Bridge Engineering: Seismic Design

Classification Based on Mathematical Formulation

Linear analysis: An analysis in which equilibrium, compatibility, and constitutive equations are
linear.
Nonlinear analysis: An analysis in which some or all of the equilibrium, compatibility, and
constitutive equations are nonlinear.

General Guidelines
The following guidelines may be useful in analysis type selection:
• A first-order analysis may be adequate for short- to medium-span bridges. A second-order
analysis should always be encouraged for long-span, tall, and slender bridges. A true large
displacement analysis is generally unnecessary for bridge structures.
• An elastic analysis is sufficient for strength-based design. Inelastic analyses should be used
for displacement-based design.
• The bowing effect (effect of flexural bending on member’s axial deformation), the Wagner
effect (effect of bending moments and axial forces acting through displacements associated
with the member twisting), and shear effects on solid-webbed members can be ignored for
most bridge structures.
• For steel nonlinearity, yielding must be taken into account. Strain hardening and fracture
may be considered. For concrete nonlinearity, a complete strain–stress relationship (in com-
pression up to the ultimate strain) should be used. Concrete tension strength can be neglected.
• Other nonlinearities, most importantly soil–foundation–structural interaction, seismic
response modification devices (dampers and seismic isolations), connection flexibility, AND
gap close and opening, should be carefully considered.

4.3 Geometric Nonlinearity Formulation

Geometric nonlinearities can be considered in the formulation of member stiffness matrices. The
general force–displacement relationship for the prismatic member as shown in Figure 4.3 can be
expressed as follows:

{F } = [ K ]{D} (4.1)

where {F} and {D} are force and displacement vectors and [K] is stiffness matrix.
For a two-dimensional member as shown in Figure 4.3a:

{F } = {P1a , F2 a , M3a , P1b , F2 b , M3b }

T
(4.2)

{D} = {u1a , u2 a , q3a , u1b , u2 b , q3b }

T
(4.3)

For a three-dimensional member as shown in Figure 4.3b:

{ F } = { P 1a, F 2a, F 3a, M 1a, M 2a, M 3a, P 1b, F 2b, F 3b, M 1b, M 2b, M 3b }
T
(4.4)

{ D } = { u 1a, u 2a, u 3a, q 1a, q 2a, q 3a, u 1b, u 2b, u 3b, q 1b, q 2b, q 3b }
T
(4.5)

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-5

FIGURE 4.3 Degrees of freedom and nodal forces for a framed member. (a) Two-dimensional and (b) three-
dimensional members.

Two sets of formulations of stability function-based and finite-element-based stiffness matrices are
presented in the following section.

Two-Dimensional Members
For a two-dimensional prismatic member as shown in Figure 4.3a, the stability function-based
stiffness matrix [9] is as follows:

AE AE
-------- 0 0 – -------- 0 0
L L
12EI – 6EI – 12EI – 6EI
------------f
3 2
- f2
1 ------------ 0 - f ------------
--------------
3 2
- f2
L L L L
6EI
4f 3 0 - f2
-------- 2f 4
[K] = L
2 (4.6)
AE
-------- 0 0
L
12EI 6EI
3
-f
----------- - f2
--------
2
L L
4f 3

where A is the cross-section area; E is the material modulus of elasticity; L is the member length;
and f1, f2 , f3 , and f 4 can be expressed by stability equations and are listed in Table 4.2. Alternatively,
fi functions can also be expressed in the power series derived from the analytical solutions [10]
as listed in Table 4.3.
Assuming polynomial displacement functions, the finite-element-based stiffness matrix [11,12]
has the following form:

[K ] = [K ] + [K ]
e g (4.7)

where [Ke] is the first-order conventional linear elastic stiffness matrix and [Kg] is the geometric
stiffness matrix, which considers the effects of axial load on the bending stiffness of a member.

© 2003 by CRC Press LLC

4-6 Bridge Engineering: Seismic Design

TABLE 4.2 Stability Function-Based fi Equations for Two-Dimensional Member

Axial Load P
f Compression Zero Tension

(kL)3 sin kL (kL)3 sinh kL

f1 1
12f c 12f t

(kL)2 (1 - cos kL) (kL)2 (cosh kL - 1)

f2 1
6f c 6f t
(kL)(sin kL - kL cos kL) (kL)(kL cosh kL - sinh kL)
f3 1
4f c 4f t
(kL)(kL - sin kL) (kL)(sin kL - kL)
f4 1
2f c 2f t

Note: f c = 2 - 2 cos kL - kL sin kL ; f t = 2 - 2 cosh kL - kL sinh kL ; k = P / EI .

TABLE 4.3 Power Series

Expression of fi Equations

 (2n + 1)! [m(kL) ]

1 2 n
1+
f1
n =1
12f
•

 (2n + 2)! [m(kL) ]

1 1 2 n
+
f2 2 n =1
6f
•

 (2(n + 3)!) [m(kL) ]

1 2 n +1 2 n
+
f3 3 n =1
4f
•

 (2n + 3)! [m(kL) ]

1 1 2 n
+
f4 6 n =1
2f
•

 (2n + 4)! [m(kL) ]

1 2(n + 1) 2 n
f +
12 n =1

Note: minus sign = compression;

plus sign = tension.

AE AE
-------- 0 0 – -------- 0 0
L L
12EI – 6EI – 12EI – 6EI
------------
3
------------
2
- 0 --------------
3
- ------------
2
-
L L L L
6EI
4 0 --------
- 2
[K] = L
2 (4.8)
AE
-------- 0 0
L
12EI 6EI
sym -----------
3
- --------
2
-
L L
4

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-7

Í ú
Í 6 -L -6 -L ú
Í 0 ú
Í 5 10 5 10 ú
Í 2 L2 L L2 ú
Í 0 - ú
15 10 30 ú
[K ]
g
PÍ
=m Í
L 0 0 0 ú
ú (4.9)
Í
Í ú
Í 6 L ú
Í sym.
5 10 ú
Í ú
Í 2 L2 ú
ÍÎ 15 úû

It is noted [13] that Eqs. (4.8) and (4.9) exactly coincide with the stability function-based stiffness
matrix when taken with only the first two terms of the Taylor series expansion in Eq. (4.6).

Three-Dimensional Members
For a three-dimensional frame member as shown in Figure 4.3b, the stability function-based stiffness
matrix has the following form [14]:

Èf s1 0 0 0 0 0 - f s1 0 0 0 0 0 ù
Í fs 7 0 0 0 fs6 0 -fs 7 0 0 0 fs6 ú
Í ú
Í fs 9 0 -fs8 0 0 0 -fs 9 0 -fs8 0 ú
Í GJ GJ ú
Í 0 0 0 0 0 - 0 0 ú
Í L L
Í fs 4 0 0 0 fs8 0 fs5 0 úú (4.10)
Í fs 2 0 -fs6 0 0 0 fs3 ú
[K ] = Í f s1 0 0 0 0 0 ú
ú
Í
Í fs 7 0 0 0 -fs6 ú
Í ú
Í Sym. fs 9 0 fs8 0 ú
Í GJ ú
Í 0 0 ú
L
Í fs 4 0 ú
Í ú
ÍÎ f s 2 úû

where G is shear modulus of elasticity; J is torsional constant; and f s1 to f s 9 are expressed by

stability equations and listed in Table 4.4.
Finite-element-based stiffness matrix has the form [15]:

Èfe1 0 0 0 0 0 - fe1 0 0 0 0 0 ù
Í fe 7 0 0 0 fe 6 0 - fe 7 0 0 0 fe 6 ú
Í ú
Í fe 9 0 - fe8 0 0 0 - fe 9 0 - fe8 0 ú
Í GJ GJ ú
Í 0 0 0 0 0 - 0 0 ú
Í L L
Í fe 4 0 0 0 - fe8 0 fe 5 0 úú (4.11)
Í fe 2 0 - fe 6 0 0 0 fe3 ú
[ Ke ] = Í fe1 0 0 0 0 0 ú
ú
Í
Í fe 7 0 0 0 - fe 6 ú
Í ú
Í Sym. fe 9 0 fe8 0 ú
Í GJ ú
Í 0 0 ú
L
Í fe 4 0 ú
Í ú
ÍÎ fe 2 úû

© 2003 by CRC Press LLC

4-8 Bridge Engineering: Seismic Design

TABLE 4.4 Stability Function-Based fsi for Three-Dimensional Member

Stability Functions Si
f si Compression Tension

1 1
EA
f s1 = S1
L
S1
1-
EA
4 P 3 L2 y
[
H + Hz ] 1-
EA
[
4 P 3 L2 y
H ¢ + Hz¢ ]
( 4 + f y ) EI Z (aL)(sin aL - aL cos aL) (aL)(aL cosh aL - sinh aL)
f s 2 = S2 S2
(1 + f y ) L 4f a 4f a
(2 - f y ) EI Z (aL)(aL - sin aL) (aL)(sinh aL - aL)
f s 3 = S2 S3
(1 + f y ) L 2f a 2f a
( 4 + f z ) EI y (bL)(sin bL - bL cos bL) (bL)(bL cosh bL - sinh bL)
f s 4 = S4 S4
(1 + f z ) L 4f b 4f b
(2 - f z ) EI y (bL)(bL - sin bL) (bL)(sinh bL - bL)
f s 5 = S2 S5
(1 + f z ) L 2fb 2fb

f s 6 = S6
6 EI Z (aL)2 (1 - cos aL) (aL)2 (cosh aL - 1)
S6
(1 + f y ) L2 6f a 6f a

f s 7 = S7
12 EI Z (aL)3 sin aL (aL)3 sinh aL
S7
(1 + f y ) L3 12f a 12f a
6 EI y (bL)2 (1 - cos bL) (bL)2 (cosh bL - 1)
f s8 = S8 S8
(1 + f z ) L 2
6f b 6f b

12 EI y (bL)3 sin bL (bL)3 sinh bL

f s 9 = S9 S9
(1 + f z ) L 3
12fb 12fb

a = P / EI Z fa 2 - 2 cos aL - aL sin aL 2 - 2 cosh aL + aL sinh aL

b = P / EI y fb 2 - 2 cos bL - bL sin bL 2 - 2 cosh bL + bL sinh bL

H y = bL( M ya
2
+ M yb
2
)(cot bL + bL cos ec 2bL) - 2( M ya + M yb )2 + 2bLM ya M yb (cos ecbL)(1 + bL cot bL)

Hz = aL( Mza2 + Mzb2 )(cot aL + aL cos ec 2 aL) - 2( Mza + Mzb )2 + 2aLMza Mzb (cos ecaL)(1 + aL cot aL)

H y¢ = bL( M ya
2
+ M yb
2
)(coth bL + bL cos ech 2bL) - 2( M ya + M yb )2 + 2bLM ya M yb (cos echbL)(1 + bL coth bL)

Hz¢ = aL( Mza2 + Mzb2 )(coth aL + aL cos ech 2 aL) - 2( Mza + Mzb )2 + 2aLMza Mzbb (cos echaL)(1 + aL coth aL)

Èf g1 f g10 - f g11 0 0 0 0 - f g10 f g11 0 0 0 ù

Í fg7 0 f g12 f g13 fg6 - f g10 -fg 7 0 f g14 - f g13 fg6 ú
Í ú
Í fg9 f g15 -fg6 f g13 f g11 0 -fg9 f g16 -fg6 - f g13 ú
Í ú
Í f g17 f g18 f g19 - f g12 - f g15 - f g17 - f g 20 f g 21 ú
Í fg 4 0 0 - f g13 fg6 - f g 20 -fg5 f g13 ú (4.12)
Í ú
fg2 -fg6 - f g13 f g 21 - f g13 -fg3 ú
[ ] Í 0
Kg = Í
f g1 f g10 - f g11 0 0 0 ú
Í ú
Í fg7 0 - f g14 f g13 -fg6 ú
Í Sym. fg9 - f g16 fg6 f g13 ú
Í ú
Í f g17 f g18 f g19 ú
Í ú
Í fg 4 0 ú
Í f g 2 úû
Î

where fei and fgi are given in Table 4.5.

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-9

TABLE 4.5 Elements of Finite-Element-Based Stiffness Matrix

Linear Elastic Matrix Geometric Nonlinear Matrix

AE 4 EI Z 2 Fxb L Fxb L
f e1 = ; fe2 = f g1 = 0 ; f g2 = f g 4 = ; f g3 = f g5 =
L L 15 30

2 EIz 4 EI y 6 Fxb Fxb Mza + Mzb

f e3 = ; fe4 = f g7 = f g9 = ; f g 6 = f g8 = ; f g10 =
L L 5L 10 L2

2 EI y 6 EI Z M ya + M yb M ya M xb
fe5 = ; fe6 = f g11 = ; f g12 = ; f g13 =
L L2 L2 L L

12 EI Z 6 EI M yb Mza Mzb
fe7 = ; f e8 = 2 y f g14 = ; f g15 = ; f g16 =
L3 L L L L

12 EI y Fxb I p Mzb Mza M ya M yb

fe9 = f g17 = ; f g18 = - ; f g19 = -
L3 AL 6 3 3 6

Mza + Mzyb M ya + M yb
f g 20 = ; f g 21 =
6 6

Iz and Iy are moments of inertia about the z–z and y–y axis, respectively; Ip is the polar moment of
inertia.

Stiffness matrices considering warping degree of freedom and finite rotations for a thin-walled
member were derived by Yang and McGuire [16,17].
In conclusion, both sets of the stiffness matrices have been used successfully when considering
geometric nonlinearities (P-D and P-d effects). The stability function-based formulation gives an
accurate solution using fewer degrees of freedom when compared with the finite-element method.
Its power series expansion (Table 4.3) can be implemented easily without truncation to avoid
numerical difficulty.
The finite-element-based formulation produces an approximate solution. It has a simpler form
and may require dividing the member into a large number of elements in order to keep the (P/L)
term a small quantity to obtain accurate results.

4.4 Material Nonlinearity Formulations

Structural Concrete
Concrete material nonlinearity is incorporated into analysis using a nonlinear stress–strain rela-
tionship. Figure 4.4 shows idealized stress–strain curves for unconfined and confined concrete in
uniaxial compression. Tests have shown that the confinement provided by closely spaced transverse
reinforcement can substantially increase the ultimate concrete compressive stress and strain. The
confining steel prevents premature buckling of the longitudinal compression reinforcement and
increases the concrete ductility. Extensive research has been made to develop concrete stress–strain
relationships [18–25].

Compression Stress–Strain Relationship

Unconfined Concrete
A general stress–strain relationship proposed by Hognestad [18] is widely used for plain concrete
or reinforced concrete with a small amount of transverse reinforcement. The relation has the
following simple form:

© 2003 by CRC Press LLC

4-10 Bridge Engineering: Seismic Design

FIGURE 4.4 Idealized stress–strain curves for concrete in uniaxial compression.

Ï È 2e Ê e ˆ 2 ù
Ô fco¢ Í c - Á c ˜ ú e c £ e co
Ô Í e co Ë e co ¯ ú
Ô Î û
fc = Ì (4.13)
Ô È Ê e - e ˆù
Ô fco¢ Í1 - bÁ c o ˜ ú e co < e c £ e u
ÔÓ ÍÎ Ë e u - e co ¯ úû

2 fco¢
e co = (4.14)
Ec

where fc and ec are the concrete stress and strain; fco¢ is the peak stress for unconfined concrete
usually taken as the cylindrical compression strength fc¢ ; eco is strain at peak stress for unconfined
concrete usually taken as 0.002; eu is the ultimate compression strain for unconfined concrete taken
as 0.003; Ec is the modulus of elasticity of concrete; b is a reduction factor for the descending branch
usually taken as 0.15. Note that the format of Eq. (4.13) can be also used for confined concrete if
the concrete-confined peak stress fcc¢ and strain ecu are known or assumed and substituted for fco¢
and eu, respectively.

Confined Concrete — Mander’s Model

Analytical models describing the stress–strain relationship for confined concrete depend on the
confining transverse reinforcement type (such as hoops, spiral, or ties) and shape (such as circular,
square, or rectangular). Some of those analytical models are more general than others in their
applicability to various confinement types and shapes. A general stress–strain model (Figure 4.5)
for confined concrete applicable (in theory) to a wide range of cross sections and confinements was
proposed by Mander et al. [23,24] and has the following form:

fcc¢ (e c / e cc ) r
fc = (4.15)
r - 1 + (e c / e cc )
r

È Ê f¢ ˆù
e cc = e co Í1 + 5 Á cc - 1˜ ú (4.16)
Î Ë f ¢
co ¯û

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-11

FIGURE 4.5 Stress–strain curves — Mander model.

Ec
r= (4.17)
Ec - Esec

fcc¢
Esec = (4.18)
e cc

where fcc¢ and ecc are peak compressive stress and corresponding strain for confined concrete. fcc¢
and ecu, which depend on the confinement type and shape, are calculated as follows:
Confined Peak Stress
1. For concrete circular section confined by circular hoops or spiral (Figure 4.6a):

Ê 7.94 fl ¢ 2 fl ¢ ˆ
fcc¢ = fco¢ Á 2.254 1 + - - 1.254˜ (4.19)
Ë f ¢
co f ¢
co ¯

1
fl ¢ = Kr f (4.20)
2 e s yh

Ï (1 - s¢ / 2 d )2 / (1 - r ) for circular hoops

Ô s cc
Ke = Ì (4.21)
Ô(1 - s¢ / 2 ds ) / (1 - rcc ) for circular spirals
Ó

4 Asp
rs = (4.22)
ds s

© 2003 by CRC Press LLC

4-12 Bridge Engineering: Seismic Design

FIGURE 4.6 Confined core for hoop reinforcement. (a) Circular hoop and (b) rectangular hoop reinforcement.

where fl ¢ is the effective lateral confining pressure; Ke is confinement effectiveness coefficient, fyh
is the yield stress of the transverse reinforcement, s¢ is the clear vertical spacing between hoops or
spiral; s is the center-to-center spacing of the spiral or circular hoops; ds is the centerline diameter
of the spiral or hoops circle; rcc is the ratio of the longitudinal reinforcement area to section core
area; rs is the ratio of the transverse confining steel volume to the confined concrete core volume;
and Asp is the bar area of transverse reinforcement.
2. For rectangular concrete section confined by rectangular hoops (Figure 4.6b)
The rectangular hoops may produce two unequal effective confining pressures flx¢ and fly¢ in
the principal x and y direction defined as follows:

flx¢ = Ke r x fyh (4.23)

fly¢ = Ke ry fyh (4.24)

È n
( wi¢ )2 ù Ê s¢ ˆ Ê s¢ ˆ
Í1 -
Í
 1-
6bc dc úú ÁË 2b ˜¯ Á1 - 2 d ˜
Ë c¯
Ke = Î û
i =1 c
(4.25)
(1 - rcc )
Asx
rx = (4.26)
s dc

Asy
ry = (4.27)
s bc

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-13

FIGURE 4.7 Peak stress of confined concrete.

where fyh is the yield strength of transverse reinforcement; wi¢ is the ith clear distance between
adjacent longitudinal bars; bc and dc are core dimensions to centerlines of hoop in x and y direction
(where b ≥ d), respectively; and Asx and Asy are the total area of transverse bars in x and y direction,
respectively.
Once flx¢ and fly¢ are determined, the confined concrete strength fcc¢ can be found using the
chart shown in Figure 4.7 with flx¢ being greater than or equal to fly¢ . The chart depicts the general
solution of the “five-parameter” multiaxial failure surface described by William and Warnke [26].
As an alternative to the chart, the authors derived the following equations for estimating fcc¢ :

Ï Aflx2 + Bflx + C fly < flx and fly £ 0.15

Ô
Ôf -f
Ô lx
fcc = Ì D+C fly < flx and fly > 0.15
ly
(4.28)
Ô 0. 3 - f ly
Ô
ÔÓC fly = flx

A = 196.5 fly2 + 29.1 fly - 4 (4.29)

B = -69.5 fly2 + 8.9 fly + 2.2 (4.30)

C = -6.83 fly2 + 6.38 fly + 1 (4.31)

D = -1.5 fly2 - 0.55 fly + 0.3 (4.32)

fcc¢ f¢ fly¢
where fcc = ; flx = lx ; fly =
fco¢ fco¢ fco¢

© 2003 by CRC Press LLC

4-14 Bridge Engineering: Seismic Design

Note that by setting fl ¢ = 0.0 in Eq. (4.19), Eqs. (4.16) and (4.15) will produce to Mander’s
expression for unconfined concrete. In this case and for concrete strain ec > 2 eco, a straight line that
reaches zero stress at the spalling strain esp is assumed.
Confined Concrete Ultimate Compressive Strain
Experiments have shown that a sudden drop in the confined concrete stress–strain curve takes place
when the confining transverse steel first fractures. Defining the ultimate compressive strain as the
longitudinal strain at which the first confining hoop fracture occurs, and using the energy balance
approach, Mander et al. [27] produced an expression for predicting the ultimate compressive strain
that can be solved numerically.
A conservative and simple equation for estimating the confined concrete ultimate strain is given
by Priestley et al. [7]:

1.4rs fyh e su
e cu = 0.004 + (4.33)
fcc¢

where esu is the steel strain at maximum tensile stress. For rectangular section rs = rx + ry as defined
previously. Typical values for ecu range from 0.012 to 0.05.
Equation (4.33) is formulated for confined sections subjected to axial compression. It is noted
that when Eq. (4.33) is used for a section in bending or in combined bending and axial compression,
then it tends to be conservative by a least 50%.
Chai et al. [28] used an energy balance approach to derive the following expression for calculating
the concrete ultimate confined strain as

g 2 f yh
Ï r s e su -------------
-
Ô g 1 f cc ¢ confined by reinforcement
e cu = e sp + Ì (4.34)
g 2 f yj
Ô r e ------------- confined by circular steel jackets
Ó sj suj g 1 f cc -
¢

where e sp is the spalling strain of the unconfined concrete (usually = 0.003 to 0.005), g1 is an
integration coefficient of the area between the confined and unconfined stress–strain curves; and
g2 is an integration coefficient of the area under the transverse steel stress–strain curve. The confining
ratio for steel jackets rsj = 4tj/(Dj – 2tj); Di and tj is outside diameter and thickness of the jacket,
respectively; fyj is yield stress of the steel jacket. For high- and mild-strength steels and concrete
compressive strengths of 4 to 6 ksi (27.58 to 41.37 MPa), Chai et al. [28] proposed the following
expressions:

Ï 2000rs
Ô 1 + (1428 )4 0.25 for Grade 40 Steel
g2 Ô ( rs )
=Ì (4.35)
g1 Ô 2000rs
for Grade 60 Steel
(
Ô 1 + (1480r )0.25 0.4
Ó s )
Confined Concrete — Hoshikuma’s Model
In additional to Mander’s model, Table 4.6 lists a stress–strain relationship for confined concrete
proposed by Hoshikuma et al. [25]. The Hoshikuma model was based on the results of a series of
experimental tests covering circular, square, and wall-type cross sections with various transverse
reinforcement arrangement in bridge piers design practice in Japan.

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-15

TABLE 4.6 Hoshikuma et al. [25] Stress–Strain

Relationship of Confined Concrete

Ï È Ê ˆ ù
n -1
Ô E e Í1- 1 e c ú e c £ e cc
Ô c c Á ˜
Í n Ë e cc ¯ ú
fc = Ì Î û
Ô
Ô f ¢ - E (e - e ) e cc < e c £ e cu
Ó cc des c cc

Ec e cc f¢ f ¢2
n = ; e cu = e cc + cc ; Edes = 11.2 co
Ec e cc - fcc¢ 2 Edes rs f yh

Ï rs f yh
Ô 1.0 + 3.8 f for circular section
fcc¢ Ô ¢
co
= Ì
fco¢ Ô rs f yh
Ô1.0 + 0.76 f ¢ for square section
Ó co

Ï rs fsh
Ô 0.002 + 0.033 f ¢ for circular section
Ô co
e cc = Ì
Ô r f
Ô0.002 + 0.013 f ¢
s sh
for square section
Ó co

FIGURE 4.8 Idealized stress–strain curve of concrete in uniaxial tension.

Tension Stress–Strain Relationship

Two idealized stress–strain curves for concrete in tension is shown in Figure 4.8. For plain concrete,
the curve is linear up to cracking stress fr. For reinforced concrete, there is a descending branch
because of bond characteristics of reinforcement. A trilinear expression proposed by Vebe et al. [29]
is as follows:

Ï 0.5E c e c e c £ e c1 = 2 f r § E c
Ô
f c = Ì f r [ 1 – 0.8E c ( e c – 2 f r § E c ) ] e c1 < e c £ e c2 = 2.625 f r § E c (4.36)
Ô
Ó f r [ 0.5 – 0.075E c ( e c – 2.625 f r § 4E c ) ] e c < e c3 = 9.292 f r § E c

where fr is modulus of rupture of concrete.

© 2003 by CRC Press LLC

4-16 Bridge Engineering: Seismic Design

FIGURE 4.9 Idealized stress–strain curve of structural steel and reinforcement.

Structural and Reinforcement Steel

For structural steel and nonprestressed steel reinforcement, its stress–strain relationship can be
idealized as four parts: elastic, plastic, strain hardening, and softening, as shown in Figure 4.9. The
relationship is, commonly expressed as follows:

Ï E s es 0 £ es £ e y
Ôf e sy < e s £ e sh
Ô y
Ô e s – e sh
- ( f su – f y )
f s = Ì f y + ----------------- e sh < e s £ e su (4.37)
Ô e su – e sh
Ô
e s – e su
Ô f u – ------------------ ( f su – f sb ) e cu < e s £ e sb
Ó e sb – e su

where fs and es are stress of strain in steel; Es is the modulus of elasticity of steel; fy and ey are yield
stress and strain; esh is hardening strain; fsu and esu are maximum stress and corresponding strain;
and fsb and esb are rupture stress and corresponding strain.

Ï 14 e y for Grade 40
Ô
e sh = Ì (4.38)
ÓÔ 5 e y for Grade 60

ÏÔ 0.14 + e sh for Grade 40

e su = Ì (4.39)
ÓÔ 0.12 for Grade 60

For the reinforcing steel, the following nonlinear form can also be used for the strain-hardening
portion [28]:

È m(e s - e sh ) + 2 (e s - e sh )(60 - m) ù
fs = f y Í + ú for e sh < e s £ e su (4.40)
ÍÎ 60(e s - e sh ) + 2 2(30r + 1)
2
úû

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-17

m =
(f su )
/ fy (30r + 1) - 60r - 1
2

(4.41)
15r 2

r = e su - e sh (4.42)

fsu = 1.5 fy (4.43)

For both strain-hardening and -softening portions, Holzer et al. [30] proposed the following
expression:

e s – e sh Ê f su ˆ e s – e sh ˆ
- ------- – 1 exp Ê 1 – -----------------
f s = f y 1 + ----------------- - for e sh < e s £ e sb (4.44)
e su – e sh Ë f y ¯ Ë e su – e sh¯

For prestressing steel, its stress–strain behavior is different from that of nonprestressed steel.
There is no obvious yield flow plateau in its response.

4.5 Nonlinear Section Analysis

Basic Assumptions and Formulations

The main purpose of section analysis is to study the moment–thrust–curvature behavior. In a
nonlinear section analysis, the following assumptions are usually made:
• Plane sections before bending remain plane after bending;
• Shear and torsional deformation is negligible;
• Stress–strain relationships for concrete and steel are given;
• For reinforced concrete, a prefect bond between concrete and steel rebar exists.
The mathematical formulas used in the section analysis are (Figure 4.10):
Compatibility equations
fx = e / y (4.45)

fy = e / x (4.46)
Equilibrium equations

P =
Ú
A
s dA = Âs A
i =1
i i (4.47)

Mx =
ÚA
s y dA = Âs y A
i =1
i i i (4.48)

My =
Ú s x dA = Â s x A
A i =1
i i i (4.49)

© 2003 by CRC Press LLC

4-18 Bridge Engineering: Seismic Design

FIGURE 4.10 Moment–curvature–strain of cross section.

Modeling and Solution Procedures

For a reinforced-concrete member, the cross section is divided into a proper number of concrete and
steel filaments representing the concrete and reinforcing steel as shown in Figure 4.10d. Each concrete
and steel filament is assigned its corresponding stress–strain relationships. Confined and unconfined
stress–strain relationships are used for the core concrete and for the cover concrete, respectively.
For a structural steel member, the section is divided into steel filaments and a typical steel
stress–strain relationship is used for tension and compact compression elements, and an equivalent
stress–strain relationship with reduced yield stress and strain can be used for a noncompact com-
pression element.
The analysis process starts by selecting a strain for the extreme concrete (or steel) fiber. By using
this selected strain and assuming a section neutral axis (NA) location, a linear strain profile is con-
structed and the corresponding section stresses and forces are computed. Section force equilibrium is
then checked for the given axial load. By changing the location of the NA, the process is repeated until
equilibrium is satisfied. Once equilibrium is satisfied, for the assumed strain and the given axial load,
the corresponding section moment and curvature are computed by Eqs. (4.48) and (4.49).
A moment–curvature (M–F) diagram for a given axial load is constructed by incrementing the
extreme fiber strain and finding the corresponding moment and the associated curvature. An
interaction diagram (M–P) relating axial load and the ultimate moment is constructed by incre-
menting the axial load and finding the corresponding ultimate moment using the above procedure.

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-19

FIGURE 4.11 Moment–thrust–curvature curve. (a) Reinforced concrete section; (b) steel I-section.

For a reinforced-concrete section, the yield moment is usually defined as the section moment at
onset of yielding of the tension reinforcing steel. The ultimate moment is defined as the moment
at peak moment capacity. The ultimate curvature is usually defined as the curvature when the
extreme concrete fiber strain reaches ultimate strain or when the reinforcing rebar reaches its
ultimate (rupture) strain (whichever takes place first). Figure 4.11a shows typical M–P–F curves
for a reinforced-concrete section.

© 2003 by CRC Press LLC

4-20 Bridge Engineering: Seismic Design

FIGURE 4.12 General yield surfaces. (a) Reinforced concrete section; (b) steel section.

For a simple steel section, such as rectangular, circular-solid, and thin-walled circular section,
a closed-form of M–P–F can be obtained using the elastic-perfectly plastic stress–strain rela-
tions [4, 31]. For all other commonly used steel section, numerical iteration techniques are
used to obtain M–P–F curves. Figure 4.11b shows typical M–P–F curves for a wide-flange
section.

Yield Surface Equations

The yield or failure surface concept has been conveniently used in inelastic analysis to describe the
full plastification of steel and concrete sections under the action of axial force combined with biaxial
bending. This section will present several yield surface expressions for steel and concrete sections
suitable for use in a nonlinear analysis.
Yield Surface Equations for Concrete Sections
The general interaction failure surface for a reinforced-concrete section with biaxial bending, as
shown in Figure 4.12a, can be approximated by a nondimensional interaction equation [32]:

n
Ê My ˆ
m
Ê Mx ˆ
Á M ˜ + Á M ˜ = 1.0 (4.50)
Ë xo ¯ Ë yo ¯

where Mx and My are bending moments about x–x and y–y principal axes, respectively; Mxo and
Myo are the uniaxial bending capacity about the x–x and y–y axes under axial load P; and the
exponents m and n depend on the reinforced-concrete section properties and axial force. They can
be determined by a numerical analysis or experiments. In general, the values of m and n usually
range from 1.1 to 1.4 for low and moderate axial compression.
Yield Surface Equation for Doubly Symmetrical Steel Sections
The general shape of yield surface for a doubly symmetrical steel section as shown in Figure 4.12b
can be described approximately by the following general equation [33]:

ax ay
Ê M ˆ Ê M ˆ
x
Á M ˜ + Á y ˜ = 1.0 (4.51)
Ë pcx ¯ Ë M pcy ¯

where Mpcx and Mpcy are the moment capacities about respective axes, reduced for the presence of
axial load; they can be obtained by the following formulas:

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-21

TABLE 4.7 Parameters for Doubly Symmetrical Steel Sections

Section Types ax ay bx by

Solid rectangular 1.7 + 1.3 (P/Py) 1.7 + 1.3 (P/Py) 2.0 2.0
Solid circular 2.0 2.0 2.1 2.1
I-shape 2.0 1.2 + 2 (P/Py) 1.3 2 + 1.2 (Aw/Af )
Thin-walled box 1.7 + 1.5 (P/Py) 1.7 + 1.5 (P/Py) 2 - 0.5 B ≥ 1.3 2 - 0.5 B ≥ 1.3
Thin-walled circular 2.0 2.0 1.75 1.75

Where B is the ratio of width to depth of the box section with respect to the bending axis.

È Ê Pˆ xù
b

M pcx = M px - Á ˜ ú
Í 1 (4.52)
Í Ë Py ¯ ú
Î û

È Ê Pˆ yù
b

M pcy = M py 1 - Á ˜ ú
Í (4.53)
Í Ë Py ¯ ú
Î û

where P is the axial load; Mpx and Mpy are the plastic moments about x–x and y–y principal axes,
respectively; and ax, ay, bx, and by are parameters that depend on cross-sectional shapes and area
distribution and are listed in Table 4.7.
Equation (4.51) represents a smooth and convex surface in the three-dimensional stress-resultant
space. It is easy to implement in a computer-based structural analysis.
Orbison [15] developed the following equation for a wide-flange section by trial and error and
curve fitting:

2 2 4 2
Ê Pˆ Ê M ˆ Ê M ˆ Ê Pˆ Ê M ˆ
1.15 Á ˜ + Á x ˜ + Á y ˜ + 3.67 Á ˜ Á x ˜
Ë Py ¯ Ë M px ¯ Ë M py ¯ Ë Py ¯ Ë M px ¯
(4.54)
2 2 4 2
Ê Pˆ Ê M ˆ Ê M ˆ Ê M ˆ
+ 3.0 Á ˜ Á y ˜ + 4.65 Á x ˜ Á y ˜ = 1.0
Ë Py ¯ Ë M py ¯ Ë M px ¯ Ë M py ¯

4.6 Nonlinear Frame Analysis

Both the first-order and second-order inelastic frame analyses can be categorized into three types
of analysis: (1) elastic–plastic hinge, (2) refined plastic hinge, and (3) distributed plasticity. This
section will discuss the basic assumptions and applications of those analyses.

Elastic–Plastic Hinge Analysis

In an elastic–plastic hinge (lumped plasticity) analysis, material inelasticity is taken into account
using concentrated “zero-length” plastic hinges. The traditional plastic hinge is defined as a zero-
length point along the structure member that can maintain plastic moment capacity and rotate
freely. When the section reaches its plastic capacity (for example, the yield surface as shown in
Figure 4.12), a plastic hinge is formed and the element stiffness is adjusted [34,35] to reflect the
hinge formation. For regions in a framed member away from the plastic hinge, elastic behavior is
assumed.

© 2003 by CRC Press LLC

4-22 Bridge Engineering: Seismic Design

For a framed member subjected to end forces only, the elastic–plastic hinge method usually
requires only one element per member, making the method computationally efficient. It does not,
however, accurately represent the distributed plasticity and associated P-d effects. This analysis
predicts an upper-bound solution (see Figure 4.1).

Refined Plastic Hinge Analysis

In the refined plastic hinge analysis [36], a two-surface yield model considers the reduction of plastic
moment capacity at the plastic hinge (due to the presence of axial force) and an effective tangent
modulus accounts for the stiffness degradation (due to distributed plasticity along a frame member).
This analysis is similar to the elastic–plastic hinge analysis in efficiency and simplicity and, to some
extent, also accounts for distributed plasticity. The approach has been developed for advanced design
of steel frames, but detailed considerations for concrete structures still need to be developed.

Distributed Plasticity Analysis

Distributed plasticity analysis models the spread of inelasticity through the cross sections and along
the length of the members. This is also referred to as plastic zone analysis, spread-of-plasticity
analysis, and elastoplastic analysis by various researchers. In this analysis, a member needs to be
subdivided into several elements along its length to model the inelastic behavior more accurately.
Two main approaches have been successfully used to model plastification of members in a second-
order distributed plasticity analysis:
1. Cross-sectional behavior is described as an input for the analysis by means of
moment–thrust–curvature (M–P–F) and moment–thrust–axial strain (M–P–e) relations
which may be obtained separately from section analysis as discussed in Section 4.5 or approx-
imated by closed-form expressions [31].
2. Cross sections are subdivided into elemental areas and the state of stresses and strains is
explicitly, traced using the proper stress–strain relations for all elements during the analysis.
In summary, the elastic–plastic hinge analysis is the simplest one, but provides an upper-bound
solution. Distributed plasticity analysis is considered the most accurate and is generally computa-
tionally intensive for larger and complex structures. Refined plastic hinge analysis seems to be an
alternative that can reasonably achieve both computational efficiency and accuracy.

4.7 Practical Applications

In this section, the concept and procedures of displacement-based design and the bases of the static
push-over analysis are discussed briefly. Two real bridges are analyzed as examples to illustrate
practical application of the nonlinear static push-over analysis approach for bridge seismic design.
Additional examples and detailed discussions of nonlinear bridge analysis can be found in the
literature [7,37].

Displacement-Based Seismic Design

Basic Concept
In recent years, displacement-based design has been used in the bridge seismic design practice as a
viable alternative approach to strength-based design. Using displacements rather than forces as a
measurement of earthquake damage allows a structure to fulfill the required function (damage-
control limit state) under specified earthquake loads.
In a common design procedure, one starts by proportioning the structure for strength and
stiffness, performs the appropriate analysis, and then checks the displacement ductility demand

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-23

FIGURE 4.13 Load–deformation curves.

against available capacity. This procedure has been widely used in bridge seismic design in California
since 1994. Alternatively, one could start with the selection of a target displacement, perform the
analysis, and then determine strength and stiffness to achieve the design level displacement. Strength
and stiffness do not enter this process as variables; they are the end results [38,39].
In displacement-based design, the designer needs to define a criterion clearly for acceptable
structural deformation (damage) based on postearthquake performance requirements and the
available deformation capacity. Such criteria are based on many factors, including structural type
and importance.
Available Ultimate Deformation Capacity
Because structural survival without collapse is commonly adopted as a seismic design criterion for
ordinary bridges, inelastic structural response and some degradation in strength can be expected
under seismic loads. Figure 4.13 shows a typical load–deformation curve. A gradual degrading
response as shown in Figure 4.13 can be due to factors such as P-D effects and/or plastic hinge
formulation. The available ultimate deformation capacity should be based on how great a reduction
(degradation) in structure load-carrying capacity response can be tolerated [21].
In general, the available ultimate deformation capacity can be referred to as the deformation that
a structure can undergo without losing significant load-carrying capacity [40]. It is therefore rea-
sonable to define available ultimate deformation as that deformation when the load-carrying capac-
ity has been reduced by an acceptable amount after the peak load, say, 20%, as shown in Figure 4.13.
This acceptable reduction amount may vary depending on required performance criteria of the
particular case.
The available deformation capacity based on the design criteria requirements need not correspond
to the ultimate member or system deformation capacity. For a particular member cross section, the
ultimate deformation in terms of the curvature depends on the shape, material properties, and loading
conditions of the section (i.e., axial load, biaxial bending) and corresponds to the condition when the
section extreme fiber reaches its ultimate strain (ecu for concrete and esp for steel). The available ultimate
curvature capacity fu can be chosen as the curvature that corresponds to the condition when section
moment capacity response reduces by, say, 20% from the peak moment.
For a framed structure system, the ultimate deformation in terms of the lateral displacement depends
on structural configurations, section behavior, and loading conditions and corresponds to a failure state
of the frame system when a collapse mechanism forms. The available lateral displacement capacity Du
can be chosen as the displacement that corresponds to the condition when lateral load-carrying
capacity reduces by some amount, say, 20% from its peak load. In current seismic design practice

© 2003 by CRC Press LLC

4-24 Bridge Engineering: Seismic Design

in California, the available frame lateral displacement capacity commonly corresponds to the first
plastic hinge reaching its ultimate rotational capacity.
Analysis Procedures
Seismic analysis procedures used in displacement-based design can be divided into three groups:
Group I: Seismic displacement and force demands are estimated from an elastic dynamic time
history or a response spectrum analysis with effective section properties. For concrete struc-
tures, cracked section properties are usually used to determine displacement demands, and
gross section properties are used to determine force demands. Strength capacity is evaluated
from nonlinear section analysis or other code-specified methods, and displacement capacity
is obtained from a static nonlinear push-over analysis.
Group II: Seismic displacement demand is obtained from a specified response spectrum and
initial effective stiffness or a substitute structural model [38] considering both the effective
stiffness and the effective damping. Effective stiffness and displacement capacity are estimated
from a nonlinear static push-over analysis.
Group III: A nonlinear inelastic dynamic time-history analysis is performed. Bridge assessment
is based on displacement (damage) comparisons between analysis results and the given
acceptance criteria. This group of analyses is complex and time-consuming and used only
for important structures.

Static Push-Over Analysis

In lieu of a nonlinear time-history dynamic analysis, bridge engineers in recent years have used
static push-over analyses as an effective and simple alternative when assessing the performance of
existing or new bridge structures under seismic loads. Given the proper conditions, this approximate
alternative can be as reliable as the more accurate and complex ones. The primary goal of such an
analysis is to determine the displacement or ductility capacity, which is then compared with dis-
placement or ductility demand obtained for most cases from linear dynamic analysis with effective
section properties. However, under certain conditions, the analysis can also be used in the assessment
of the displacement demand, as will be illustrated in the examples to follow.
In this analysis, a stand-alone portion from a bridge structure (such as bent-frame with single
or multiple columns) is isolated and statically analyzed, taking into account whatever nonlinear
behavior is deemed necessary (most importantly and commonly, material and geometric nonlinear
behavior). The analysis can utilize any of the modeling methods discussed in Section 4.6, but plastic
hinges or distributed plasticity models are commonly used. The analytical frame model is first subjected
to the applied tributary gravity load and then is pushed laterally in several load (or displacement)
increments until a collapse mechanism or a given failure criterion is reached. Figure 4.14 shows a
flowchart outlining a procedure using static push-over analysis in seismic design and retrofit evaluation.
When applying static push-over analysis in seismic design, it is assumed that such analysis can
predict with reasonable accuracy the dynamic lateral load–displacement behavior envelope, and
that an elastic acceleration response spectrum can provide the best means for establishing required
structural performance.

Example 4.1 — Reinforced-Concrete Multicolumn Bent Frame

with P-D Effects
Problem Statement
The as-built details of a reinforced-concrete bridge bent frame consisting of a bent cap beam and two
circular columns supported on pile foundations are shown in Figure 4.15. An as-built unconfined
concrete strength of 5 ksi (34.5 MPa) and steel strength of 40 ksi (275.8 MPa) are assumed. Due to lack
of adequate column transverse reinforcement, the columns are retrofitted with 0.5-in. (12.7-mm)-thick

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-25

FIGURE 4.14 An alternative procedure for bridge seismic evaluation.

steel jacket. The bottom of the column is assumed to be fixed; however, since the footing lacks top mat
and shear reinforcement, the bottom with a pinned connection is also to be considered. The frame is
supported on a stiff pile–foundation and the soil–foundation–structure interaction is to be ignored.
Use static nonlinear push-over analysis to study the extent of the P-D effect on the lateral response
of the bent frame when the columns are assumed fixed at the base in one case and pinned in another
case. Assume that the columns are retrofitted with steel jacket in both cases and determine whether
the footing retrofit is also required. Use 0.7 g ground acceleration and the ARS spectrum with 5%
damping shown in Figure 4.16.
Analysis Procedure
The idealized bent frame, consisting of the cap beam and the two retrofitted column members, is
discretized into a finite number of beam elements connected at joints, as shown in Figure 4.17. The
idealized column and cap beam cross sections are divided into several concrete layers and reinforcing
steel layers as shown. Two different concrete material properties are used for the column and cap

© 2003 by CRC Press LLC

4-26 Bridge Engineering: Seismic Design

FIGURE 4.15 As-built plan — Example 4.1.

FIGURE 4.16 Specific ARS curve — Example 4.1.

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-27

FIGURE 4.17 Analytical model — Example 4.1. (a) Local layered cap beam section: 12 concrete and 2 steel layers;
(b) layered column section: 8 concrete and 8 steel layers; (c) discretized frame model.

beam cross sections. The column concrete properties incorporated the increase in concrete ultimate
stress and strain due to the confinement provided by the steel jacket. In this study the column
confined ultimate concrete compressive stress and strain of 7.5 ksi (51.7 MPa) and 0.085 are used,
respectively. The total tributary superstructure dead load of 1160 kips (5160 kN) is applied uniformly
along the length of the cap beam. The frame is pushed laterally in several load increments until
failure is reached.
For this study, failure is defined as the limit state when one of the following conditions first takes place:
1. A concrete layer strain reaches the ultimate compressive strain at any member section;
2. A steel layer strain reaches the rupture strain at any member section;
3. A 20% reduction from peak lateral load of the lateral load response curve (this condition is
particularly useful when considering P-D).
The lateral displacement corresponding to this limit state at the top of the column defines the
frame failure (available) displacement capacity.
A nonlinear analysis computer program NTFrame [41,42] is used for the push-over analysis. The
program is based on the distributed plasticity model, and the P-D effect is incorporated in the model
second-order member stiffness formulation.
Discussion of the Results
The resulting frame lateral load vs. displacement responses are shown in Figure 4.18 for the cases
when the bottom of the column is fixed and pinned. Both cases will be discussed next, followed by
concluding remarks.
Column Fixed at Bottom Case
In this case the column base is modeled with a fixed connection. The lateral response with and
without the P-D effect is shown in Figure 4.18a. The sharp drop in the response curve is due to
several extreme concrete layers reaching their ultimate compressive strain at the top of the column.
The effect of P-D at failure can be seen to be considerable but not as severe as shown in Figure 4.18b
with the pinned connection. Comparing Figures 4.18a and b, one can observe that fixing the bottom
of the column resulted in stiffer structural response.

© 2003 by CRC Press LLC

4-28 Bridge Engineering: Seismic Design

FIGURE 4.18 Lateral load vs. displacement responses — Example 4.1. Lateral response (a) fixed column (b) with
penned column.

Using the curve shown in Figure 4.18a, the displacement demand for the fixed column case with
P-D effect is calculated as follows:
Step 1: Calculate the Initial Effective Stiffness Keff
The computer results showed that the first column extreme longitudinal rebar reached yield
at lateral force of 928 kips (4128 kN) at a corresponding lateral yield displacement of 17 in.
(431.8 mm); therefore:

Keff = 928/17 = 55 kips/in. (9.63 kN/mm)

Step 2: Calculate an Approximate Fundamental Period Tf

W 1160
Tf = 0.32 = 0.32 = 1.5 s
Keff 55

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-29

Step 3: Determine the Damped Elastic Acceleration Response Spectrum (ARS) at the Site in g’s
By using the given site spectrum shown in Figure 4.16 and the above calculated period, the
corresponding ARS for 5% damping is 0.8.
Step 4: Calculate the Displacement Demand Dd

ARS ( W ) 0.8 ( 1160 )

D d = ---------------------- = ------------------------ = 16.9 in. ( 429.3 m )
K eff 55

(in this case the yield and demand displacements are found to be practically equal).
In much of the seismic design practice in California, the effect of P-D is usually ignored if the P-
D moment is less than 20% of the design maximum moment capacity. Adopting this practice and
assuming that the reduction in the moment is directly proportional to the reduction in the lateral
force, one may conclude that at displacement demand of 16.9 in. (429.3 mm), the reduction in
strength (lateral force) is less than 20%, and as a result the effects of P-D are negligible.
The displacement demand of 16.9 in. (429.3 mm) is less than the failure state displacement
capacity of about 40 in. (1016 mm) (based on a 20% lateral load reduction from the peak). Note
that for the fixed bottom case with P-D, the displacement when the extreme concrete layer at the
top of the column reaches its ultimate compressive strain is about 90 in. (2286 mm).
Column Pinned at Bottom Case
In this case the column bottom is modeled with a pinned connection. Note that the pinned condition
assumption is based on the belief that in the event of a maximum credible earthquake, the col-
umn–footing connection would quickly degenerate (degrade) and behave like a pinned connection.
The resulting lateral responses with and without the P-D effect are shown in Figure 4.18b. In this
case the effects of P-D are shown to be quite substantial.
When considering the response without the P-D, one obtains a displacement demand of 38 in.
(965.2 mm) (based on a calculated initial stiffness of 18.5 kips/in. (3.24 kN/mm) and a correspond-
ing structure period of 2.5 s). This displacement demand is well below the ultimate (at failure)
displacement capacity of about 115 in. (2921 mm). As a result, one would conclude that the retrofit
measure of placing a steel jacket around the column with no footing retrofit would be adequate.
The actual response, however, is the one that includes the P-D effect. In this case the effect of P-
D resulted in a slight change in initial stiffness and frame period — 15.8 kips/in. (2.77 kN/mm)
and 2.7 s, respectively. However, beyond the initial stages, the effects are quite severe on the load–dis-
placement response. The failure mode in this case will most likely be controlled by dynamic
instability of the frame. MacRae et al. [43] performed analytical studies of the effect of P-D on
single-degree-of-freedom bilinear oscillators (i.e., single-column frame) and proposed some pro-
cedures to obtain a limiting value at which the structure becomes dynamically unstable. The process
requires the generation of the proper hysteresis loops and the determination of what is termed the
effective bilinear stiffness factor. Setting aside the frame dynamic instability issue, the calculated initial
stiffness displacement demand is about 38 in. (965.2 mm) and the displacement capacity at 20%
reduction from peak load is 24 in. (609.6 mm).
Referring to the curves with P-D in Figure 4.18, it is of interest to mention, as pointed out by
Mahin and Boroschek [44], that continued pushing of the frame will eventually lead to a stage when
the frame structure becomes statically unstable. At that stage the forces induced by the P-D effect
overcome the mechanical resistance of the structure. Note that the point when the curve with P-D
effect intersects the displacement-axis (as shown in Figure 4.18b) will determine the lateral displace-
ment at which the structure becomes statically unstable. Dynamic instability limits can be 20 to
70% less than the static instability, depending on the ground motion and structural characteristics
[44]. Note that dynamic instability is assumed not to be a controlling factor in the previous case
with fixed column.

© 2003 by CRC Press LLC

4-30 Bridge Engineering: Seismic Design

In conclusion, if the as-built column–footing connection can support the expected column
moment obtained from the fixed condition case (which is unlikely), then retrofitting the column
with steel jacket without footing retrofit is adequate. Otherwise, the footing should also be retrofitted
to reduce (limit) the effect of P-D.
It should be pointed out that in this example the analysis is terminated at the completion of the
first plastic hinge (conservative), whereas in other types of push-over analysis, such as event-to-
event analysis, the engineer may choose to push the frame farther until it forms a collapse mecha-
nism. Also, unlike the substitute structure procedure described by Priestley et al. [7] in which both
the effective system stiffness and damping ratio are adjusted (iterated) several times before final
displacement demand is calculated, here only the initial effective stiffness and a constant specified
structure damping are used.
As a final remark, the P-D effect in bridge analysis is normally assumed small and is usually
ignored. This assumption is justified in most cases under normal loading conditions. However, as
this example illustrated, under seismic loading, the P-D effect should be incorporated in the analysis,
when large lateral displacements are expected before the structure reaches its assumed failure state.
In the design of a new bridge, the lateral displacement and the effect of P-D can be controlled. When
assessing an existing bridge for possible seismic retrofit, accurate prediction of the lateral displace-
ment with P-D effects can be an essential factor in determining the retrofit measures required.

Example 4.2 — Steel Multicolumn Bent Frame Seismic Evaluation

Problem Statement
The as-built details of a steel bridge bent frame consisting of a bent cap plate girder and two built-up
columns supported on a stiff pile–foundation are shown in Figure 4.19. Steel is Grade 36. Site-specific
displacement response spectra are given in Figure 4.20. For simplicity and illustration purposes, fixed
bases of columns are assumed and the soil–foundation–structure interaction is ignored.

FIGURE 4.19 As-built plane — Example 4.2.

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-31

FIGURE 4.20 Displacement response spectra — Example 4.2.

Evaluate lateral displacement capacity by using static nonlinear push-over analysis. Estimate
seismic lateral displacement demands by using the substitute structure approach, considering both
the effective stiffness and the effective damping. The effective damping x can be calculated by
Takeda’s formula [45]:

Ê 0.95 ˆ
Á1 - m - 0.05 m Dd ˜
Ë Dd ¯
x = 0.05 + (4.55)
p
D ud
m Dd = (4.56)
Dy

where µDd is displacement ductility demand; and Dud and Dy are displacement demand and yield
displacement, respectively.
Analysis Modeling
The bent frame members are divided into several beam elements as shown in Figure 4.21. The
properties of beam elements are defined by two sets of relationships for moment–curvature, axial
force–strain, and torsion–twist for the cap beam and columns, respectively. The available ultimate
curvature is assumed as 20 times yield curvature. The total tributary superstructure dead load of
800 kips (3558 kN) is applied at longitudinal girder locations. A lateral displacement is applied
incrementally at the top of the bent column until a collapse mechanism of the bent frame is formed.
Displacement Capacity Evaluation
The displacement capacity evaluation is performed by push-over analysis using the ADINA [46]
analysis program. Large displacements are considered in the analysis. The resulting lateral load
vs. displacement response at the top of columns is shown in Figure 4.22. The sudden drops in
the response curve are due to the several beam elements reaching their available ultimate curva-
tures. The yield displacement Dy = 1.25 in. (31.8 mm) and the available ultimate displacement
capacity (corresponding to a 20% reduction from the peak lateral load) Du = 2.61 in. (66.3 mm)
are obtained.

© 2003 by CRC Press LLC

4-32 Bridge Engineering: Seismic Design

FIGURE 4.21 Analytical model — Example 4.2.

FIGURE 4.22 Lateral load vs. displacement — Example 4.2.

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-33

Displacement Demand Estimation

A substitute structure approach with the effective stiffness and effective damping will be used to
evaluate displacement demand.
1. Try Dud = 3 in. (76.2 mm); from Figure 4.20, Eqs. (4.55) and (4.56), we obtain

600
K eff = --------- = 200 kips/in. ( 35.04 kN/mm )
3
W ( kips ) 800
T eff = 0.32 ------------------------------ = 0.32 --------- = 0.64 (s)
k eff ( kips/in. ) 200
D ud 3
m Dd = ------- = ---------- = 2.4
Dy 1.25
Ê 1 – ----------
0.95
- – 0.05 2.4ˆ
Ë 2.4 ¯
x = 0.05 + ------------------------------------------------------- = 0.15
p

From Figure 4.20, find Dd = 2.5 in. < Dud = 3 in. (76.2 mm).
2. Try Dud = 2.5 in. (63.5 mm); from Figures 4.20 and 4.22, Eq. (4.55), and (4.56), we obtain

830
K eff = --------- = 332 kips/in. ( 58.14 kN/mm )
2.5
W ( kips ) 800
T eff = 0.32 -------------------------------- = 0.32 --------- = 0.50
K eff ( kips/in. ) 332
D ud 2.5
m Dd = ------- = ---------- = 2
Dy 1.25
Ê 1 – 0.95 ---------- – 0.05 2ˆ
Ë 2 ¯
x = 0.05 + ------------------------------------------------- = 0.13
p

From Figure 4.20, find Dd = 2.45 in. (62.2 mm) close to Dud = 2.5 in. (63.5 mm). OK
Displacement demand Dd = 2.45 in. (62.2 mm).
Discussion
It can be seen that the displacement demand Dd of 2.45 in. (62.2 mm) is less than the available
ultimate displacement capacity of Du = 2.61 in. (66.3 mm). It should be pointed out that in the
actual seismic evaluation of this frame, the flexibility of the steel column to the footing bolted
connection should be considered.

References
1. Chen, W. F., Plasticity in Reinforced Concrete, McGraw-Hill, New York, NY, 1982.
2. Clough, R. W. and Penzien, J., Dynamics of Structures, 2nd ed., McGraw-Hill, New York, 1993.
3. Fung, Y. C., First Course in Continuum Mechanics, 3rd ed., Prentice-Hall Engineering, Science &
Math, Englewood Cliffs, NJ, 1994.

© 2003 by CRC Press LLC

4-34 Bridge Engineering: Seismic Design

4. Chen, W. F. and Han, D. J., Plasticity for Structural Engineers, Gau Lih Book Co., Ltd., Taipei,
Taiwan, 1995.
5. Chopra, A. K., Dynamics of Structures: Theory and Applications to Earthquake Engineering, Prentice-
Hall, Englewood Cliffs, NJ, 1995.
6. Bathe, K. J., Finite Element Procedures, Prentice-Hall Engineering, Science & Math, Englewood
Cliffs, NJ, 1996.
7. Priestley, M. J. N., Seible, F., and Calvi, G. M., Seismic Design and Retrofit of Bridges, John Wiley &
Sons, New York, 1996.
8. Powell, G. H., Concepts and Principles for the Applications of Nonlinear Structural Analysis in
Bridge Design, Report No. UCB/SEMM-97/08, Department of Civil Engineering, University of
California, Berkeley, 1997.
9. Chen, W. F. and Lui, E. M., Structural Stability: Theory and Implementation, Elsevier, New York, 1987.
10. Goto, Y. and Chen, W. F., Second-order elastic analysis for frame design, J. Struct. Eng. ASCE,
113(7), 1501, 1987.
11. Allen, H. G. and Bulson, P. S., Background of Buckling, McGraw-Hill, London, 1980.
12. White, D. W. and McGuire, W., Method of Analysis in LRFD, Reprints of ASCE Structure Engi-
neering Congress ’85, Chicago, 1985.
13. Schilling, C. G., Buckling of one story frames, AISC Eng. J., 2, 49, 1983.
14. Ekhande, S. G., Selvappalam, M., and Madugula, M. K. S., Stability functions for three-dimensional
beam-column, J. Struct. Eng. ASCE, 115(2), 467, 1989.
15. Orbison, J. G., Nonlinear Static Analysis of Three-dimensional Steel Frames, Department of Struc-
tural Engineering, Cornell University, Ithaca, NY, 1982.
16. Yang, Y. B. and McGuire, W., Stiffness matrix for geometric nonlinear analysis, J. Struct. Eng. ASCE,
112(4), 853, 1986.
17. Yang, Y. B. and McGuire, W., Joint rotation and geometric nonlinear analysis, J. Struct. Eng. ASCE,
112(4), 879, 1986.
18. Hognestad, E., A Study of Combined Bending and Axial Load in Reinforced Concrete Members,
University of Illinois Engineering Experimental Station, Bulletin Series No. 399, Urbana, IL, Nov.
1951.
19. Kent, D. C. and Park, R., Flexural members with confined concrete, J. Struct. Div. ASCE, 97(ST7),
1969, 1971.
20. Popovics, S. A., Review of stress–strain relationship for concrete, J. ACI, 67(3), 234, 1970.
21. Park, R. and Paulay, T., Reinforced Concrete Structures, John Wiley & Sons, New York, 1975.
22. Wang, W. C. and Duan, L., The stress–strain relationship for concrete, J. Taiyuan Inst. Technol., 1,
125, 1981.
23. Mander, J. B., Priestley, M. J. N., and Park, R., Theoretical stress–strain model for confined concrete,
J. Struct. Eng. ASCE, 114(8), 1804, 1988.
24. Mander, J. B., Priestley, M. J. N., and Park, R., Observed stress–strain behavior of confined concrete,
J. Struct. Eng. ASCE, 114(8), 1827, 1988.
25. Hoshikuma, J. et al., Stress–strain model for confined reinforced concrete in bridge piers, J. Struct.
Eng. ASCE, 123(5), 624, 1997.
26. William, K. J. and Warnke, E. P., Constitutive model for triaxial behavior of concrete, Proc. IABSE,
19, 1, 1975.
27. Mander, J. B., Priestley, M. J. N., and Park, R., Seismic Design of Bridge Piers, Research Report No.
84-2, University of Canterbury, New Zealand, 1984.
28. Chai, Y. H., Priestley, M. J. N., and Seible, F., Flexural Retrofit of Circular Reinforced Bridge Col-
umns by Steel Jacketing, Report No. SSRP-91/05, University of California, San Diego, 1990.
29. Vebe, A. et al., Moment-curvature relations of reinforced concrete slab, J. Struct. Div. ASCE,
103(ST3), 515, 1977.

© 2003 by CRC Press LLC

Nonlinear Analysis of Bridge Structures 4-35

30. Holzer, S. M. et al., SINDER, A Computer Code for General Analysis of Two-Dimensional Rein-
forced Concrete Structures, AFWL-TR-74-228 Vol. 1, Air Force Weapons Laboratory, Kirtland AFB,
NM, 1975.
31. Chen, W. F. and Atsuta, T., Theory of Beam-Columns, Vols. 1 and 2, McGraw-Hill, New York, 1977.
32. Bresler, B., Design criteria for reinforced concrete columns under axial load and biaxial bending,
J. ACI, 32(5), 481, 1960.
33. Duan, L. and Chen, W. F., A yield surface equation for doubly symmetrical section, Struct. Eng.,
12(2), 114, 1990.
34. King, W. S., White, D. W., and Chen, W. F., Second-order inelastic analysis methods for steel-frame
design, J. Struct. Eng. ASCE, 118(2), 408, 1992.
35. Levy, R., Joseph, F., and Spillers, W. R., Member stiffness with offset hinges, J. Struct. Eng. ASCE,
123(4), 527, 1997.
36. Chen, W. F. and S. Toma, Advanced Analysis of Steel Frames, CRC Press, Boca Raton, FL, 1994.
37. Aschheim, M., Moehle, J. P., and Mahin, S. A., Design and Evaluation of Reinforced Concrete
Bridges for Seismic Resistance, Report No. UCB/EERC-97/04, University of California, Berkeley,
1997.
38. Priestley, N., Myths and Fallacies in Earthquake Engineering — Conflicts between Design and
Reality, in Proceedings of Tom Paulay Symposium — Recent Development in Lateral Force Transfer
in Buildings, University of California, San Diego, 1993.
39. Kowalsky, M. J., Priestley, M. J. N., and MacRae, G. A., Displacement-Based Design, Report No.
SSRP-94/16, University of California, San Diego, 1994.
40. Duan, L. and Cooper, T. R., Displacement ductility capacity of reinforced concrete columns, ACI
Concrete Int., 17(11), 61, 1995.
41. Akkari, M. M., Nonlinear push-over analysis of reinforced and prestressed concrete frames, Struc-
ture Notes, State of California, Department of Transportation, Sacramento, July 1993.
42. Akkari, M. M., Nonlinear push-over analysis with p-delta effects, Structure Notes, State of Califor-
nia, Department of Transportation, Sacramento, Nov. 1993.
43. MacRae, G. A., Priestley, M. J. N., and Tao, J., P-delta design in seismic regions, Structure System
Research Project Report No. SSRP-93/05, University of California, San Diego, 1993.
44. Mahin, S. and Boroschek, R., Influence of geometric nonlinearities on the seismic response and
design of bridge structures, Background Report, California Department of Transportation, Division
of Structures, Sacramento, 1991.
45. Takeda, T., Sozen, M. A., and Nielsen, N. N., Reinforced concrete response to simulated earth-
quakes, J. Struct. Div. ASCE, 96(ST12), 2557, 1970.
46. ADINA, ADINA-IN for ADINA User’s Manual, ADINA R & D, Inc., Watertown, MA, 1994.

© 2003 by CRC Press LLC