Analysis and Design of Bridges

NATO ASI Series

Advanced Science Institutes Series
A Senes presenting the results of activities sponsored by the NATO SCience
Committee, which alms at the dissemination of advanced sCientific and technological
knowledge, With a view to strengthening links between sCientific commUnities
The Series IS published by an International board of publishers In conjunction With the
NATO SCientific Affairs DIVISion

A
B

Life Sciences
Physics

Plenum Publishing Corporation

London and New York

Mathematical and
Physical Sciences

D Reidel Publishing Company

Dordrecht and Boston

Behavioural and
Social Sciences
Applied Sciences

Martlnus NIJhoff Publishers

The Hag ue/Boston/Lancaster

Computer and
Systems Sciences
Ecological Sciences

Springer-Verlag
Berlin/Heidelberg/New York

E
F

Series E: Applied Sciences -

No. 74

Analysis and Design of Bridges

edited by

Cetin
Y,lmaz
,
Associate Professor
Civil Engineering Department
Middle East Technical University
Ankara, Turkey

s.

Tanvir Wasti

Professor
Civil Engineering Department
Middle East Technical University
Ankara, Turkey

1984

Martinus Nijhoff Publishers

The Hague / Boston / Lancaster

Published In cooperation with NATO SCientifiC Affairs DIvIsion

Proceedings of the NATO Advanced Study Institute on Analysis and Design of

Bridges, C;:e~me, Izmir, Turkey, June 28-July 9, 1982
Library of Congress Cataloging in Publication Data
Main entry under title:
Analysis and design of bridges.
(NATO advanced science institutes series. Series E,
Applied sciences ; 74)
Proceedings of the NATO Advanced Study Institute on
Analysis and Design of Bridges held at Cesme and Izmir,
Turkey, June 28-Ju1y 9, 1982.
"Published in cooperation with NATO Scientific Affairs
Division. II
1. Bridges--Design--Congresses. I. Y11maz, ~etin.
II. Wast1, Syed Tanvir, 1941 III. NATO Advanced
Study Institute on Analysis and Design of Bridges (1982
ge~me and izmir, Turkey) IV. North Atlantic Treaty
Organization. Scientific Affairs Division. V. Series.
TG300.A58 1984
624' .25
83-25461

ISBN-13: 978-94-009-6124-1
DO I: 10.1007/978-94-009-6122-7

e-ISBN-13: 978-94-009-6122-7

Distributors for the United States and Canada: K1uwer Boston, Inc., 190 Old Derby
Street, Hingham, MA 02043, USA
Distributors for all other countries: Kluwer Academic Publishers Group, Distribution
Center, P.O. Box 322, 3300 AH Dordrecht, The Netherlands

Copyright 1984 by Martinus Nijhoff Publishers, The Hague

Softcover reprint of the hardcover 1st edition 1984
All rights reserved. No part of this publication may be reproduced, stored in a
retrieval system, or transmitted, in any form or by any means, mechanical,
photocopying, recording, or otherwise, without the prior written permission of the
publishers,
Martinus Nijhoff Publishers, P.O. Box 566, 2501 CN The Hague, The Netherlands

v
PREFACE
The Proceedings of the NATO Advanced Study Institute on
Analysis and Design of Bridges held at ~eme, lzmir, Turkey from
28 June 1982 to 9 July 1982 are contained in the present volume.
The Advanced Study Institute was attended by 37 lecturers and
participants from 10 different countries. The Organizing Committee
consisted of Professors P. Gtilkan, A. C. Scordelis, S.T. Wasti and
9. Yl.lmaz.
The guidelines set by NATO for the Advanced Study Institute
require it to serve not only as an efficient forum for the
dissemination of available advanced knowledge to a selected group
of qualified people but also as a platform for the exploration of
future research possibilities in the scientific or engineering
areas concerned. The main topics covered by the present Advanced
Study Institute were the mathematical modelling of bridges for
better analysis and the scientific assessment of bridge behaviour
for the introduction of improved design procedures. It has been
our observation that as a result of the range and depth of the
lectures presented and the many informal discussions that took place,
ideas became fissile, the stimulus never flagged and many gaps in
the engineering knowledge of the participants were "bridged".
Here we particularly wish to mention that valuable informal presentations of research work were made during the course of the Institute
by Drs. Friedrich, Karaesmen, Lamas and Parker. Contributing
tangibly to the success of the Advanced Study Institute were the
several opportunities for social get-togethers in which the ladies
a Iso took part.
The editing of the proceedings has entailed a moderate amount
of work in the case of a few manuscripts and only very slight
interference with some others. In general the proceedings faithfully
represent the lectures as transmuted from the spoken to the written
word by the lecturers themselves. Unfortunately Professor B.T. Yenis
contribution was not received in time for publication.
Inspite of the inevitable overlapping, it has been found
suitable and convenient to group the lectures under the following
five main headings :
I.
II.
III.
IV.
V.

Design Considerations
Analysis Methods for Steel Bridges
Concrete Slab and Box Girder Bridges
Dynamics and Vibration
Earthquake and Wind Effects

The lectures in the present volume cover an impressive range

of subjects and it is our hope L1at they will of use to a large
number of engineers and researchers in civil, structural and bridge
engineering.

VI

While the manuscripts were being assembled

for the
publishers, we learnt with shock and dismay of the tragic death
(in December 1982) of Professor Conrad P. Heins, Jr. in an a~r
crash in China. Conrad iiConnie ii i1eins was Professor at the University
of Maryland, a well known researcher in Bridge Engineering and a
dynamic personality. Characteristically, he was one of the first
lecturers to send us his manuscripts. R.I.P.
We wish to take this opportunity to thank the many individuals
and organizations whose assistance contributed to the success of
the Advanced Study Institute.
1 The NATO Scientific Affairs Division, Brussels, Belgium
for substantially funding the Institute.
2. Middle East Technical University, Ankara, Turkey for
providing facilities and services.
3. AFA Prefabrike Beton Sanayi ve Ticaret A.., tzmir, Turkey
for generously hosting social and technical activities.
4. yUKSEL PROJE, A.., Ankara, Turkey for hosting social
events and funding services.
5. SETAN Engineering Consultants, Ankara, Turkey for funding
serv~ces .
6. BMB (United Engineering Consulting Group Co. Ltd.)
Ankara, Turkey for funding services.
Last, but certainly not least, the book in your hands has
been produced with care and skill by Martinus Nijhoff Publishers,
The Hague. The Netherlands. We would like to thank them for their
patience and understanding and, especially, for the prompt cooperation we received from Mrs. Henny Hoogervorst.
Ankara
April 1983

t;:ETtN YILMAZ
S. TANVIR WAS TI
Edi tors

VII

TABLE OF CONTENTS

Preface
I. DESIGN CONSIDERATIOr.rS
Christian Menn
Prestressing of Curved Bridges

Fritz Leonhardt
Certain Aspects of Bridge Design

23

II. ANALYSIS METHODS FOR STEEL BRIDGES

Ce lal Kos tern
Overloading and Inelastic Response of Steel Multigirder
Bridges

49

Con rad P. He ins

Steel Curved I and Box Girder Design and Analysis

57

H. Roy Evans
Simplified Methods for the Analysis and Design of
Bridges of Cellular Cross-section

95

III. CONCRETE SLAB AND BOX GIRDE R BRIDGES

Alex C. Scordelis
Berkeley Computer Programs for the Analysis of Concrete
Box Girder Bridges

119

S. Tanvir l,vasti and Alex C. Scordelis

Comparative Structural Behavior of Straight, Curved and
Skew Reinforced Concrete Box Girder Bridge Models

191

Celal Kostem
Lateral Live Load Distribution
Highway Bridges

213

Christ ian Menn

Skew Slab Bridges

~n

Prestressed Concrete

225

VIII
IV. DYNAMICS AND VIdRATION
Nuri Akka and ~etin YLlrnaz
Fluid-Structure Interaction in Analysis of Bridge piers.
Part I : Theoretical Preliminaries

243

YLlrnaz and Nuri Akka

Fluid-Structure Interaction in Analysis of Bridge Piers.
Part II : Nurrerical Approach

259

~etin

Etienne Zeller
Vibrations and Dynamic Behaviour of Actual Bridges

285

V. EARTH':lUAKE AND WIND EFFECTS

Con rad P. He ins
Seismic Design of Highway Bridges

343

Polat Giilkan
Analysis and Design of Bridges for Earthquake
Effects

375

Etienne Zeller
Wind Loads on Bridge Structures

415

List of Participants

449

I. DESIGN CONSIDERATIONS

PRESTRESSING OF CURVED BRIDGES

Christian Menn
Professor fur Baustatik und Konstruktion
Eidgenossische Technische Hochschule
Zurich

SUMMARY

A review of some structural concepts, prestressing schemes,

analysis methods, and fundamental behavior of curved prestressed
concrete bridges is presented.
The application of several structural schemes to curved
bridges is presented. Equilibrium requirements, the effect of
moment redistribution on equilibrium requirements, and the effects
of external constraints are reviewed.
An approximate method for the analysis of curved girder bridges
is developed. A comparison with an elasticity solution demonstrates
the applicability of the method.
Based on the structural requirements and approximate analysis
method presented,
optimal prestressing
requirements
for
the
equalization of torsion in simple- and continuous-span curved
prestressed box girder bridges are developed.

1.

INTRODUCTION

Twenty years ago both the design and construction of curved

bridges posed considerable problems. In comparison with straight
bridges they were also much more expensive to erect. Today, however,
the difference in cost between straight and curved bridges is much
less significant. The alignment of highways is no longer determined
by a desire for straight bridges. Thus in modern highway systems
most bridges are at least partially curved.
1.1

Structural Systems for Curved Bridges

The customary structural systems for straight bridges are

generally applicable for curved bridges, as long as the span and the
curvature of the latter do not exceed certain values.
In the case of ordinary beam bridges, the girders are entirely
adapted to the curvature of the roadway. The load capacity is
thereby not significantly reduced.
In the case of arch bridges, the inner edge of the arch in plan
is adapted to the curvature of the roadway and the outer edge of the
arch is symmetric with respect to the straight axis of the arch (Fig.
1). The column faces on the inner side are vertical, while on the
outer side they are oblique. This results in an aesthetically
pleasant form. The span of the arch and the curvature of the roadway
are, however, limited.
A cable stayed system can also be utilized in a curved bridge.
If the pylons have the form of a V (Fig. 2), the stay cables are
outside of the bridge deck throughout the bridge and do not encroach
on the road space. The span and curvature of the roadway are,
however, limited with the use of this scheme.
1.2

Load Carrying Behavior of Curved Girders

The load carrying behavior of a curved girder is characterized

by the fact that torsion moments are required for equilibrium. This
so called equilibrium torsion clearly differs from the constraint
torsion which very often occurs.
In the case of equilibrium torsion, the torsion moments are in
principle independent of the torsional stiffness of the girder (Fig.
3a). A large torsional stiffness yields small deformations; a
vanishing torsional stiffness yields a mechanism.
In the case of constraint torsion, the torsion moments are
directly dependent on the torsional stiffness (Fig. 3b). A large
torsional stiffness results in large torsion moments. A vanishing
torsional stiffness results in vanishing torsional moments; the

FIG. 1 J

FIG. 2:

CU1MID ARCH BlUtIGB

CURVED CABLE STAYED SYSTEM

~Q
Jr

J
1/2

1/2

f!

T = Q. ~
GK -

O-collapse!

Fig 30

FIG. 3A:

EQUILIBRIUM TORSION

~J~-l/-2-;Q-l-/2-t/.
Fig 3b

FIG. 3B:

CONSTRAINT TORSION

3 GK'I
T=Q 16.(3.EI+GK)

GK -

0 -

T -

equilibrium of the system, however, remains.

Torsion moments and bending moments in a curved system are
interrelated through equilibrium conditions. Therefore, a redistribution of the bending moments in a statically indeterminate
system results in a corresponding redistribution of the torsion
moments (Fig. 4). This effect must be considered in the case of
moment redistribution resulting from a change of the statical system
(e.g. cantilever construction bridges) or when checking the ultimate
load capacity of a system in accordance with the theory of
plasticity.
In most cases, the main beam section forces (N, V, M, T) in a
curved bridge may be calculated from the basic requirements of static
equilibrium. In T- and box- cross sections, however, transverse
internal stresses which are dependent on the beam section forces are
required in order to satisfy static equilibrium requirements. The
investigation of these additional stresses is better based on a
simple analysis than a finite element analysis, since the load
carrying behavior can be clearly detected and thereby the
reinforcement can be efficiently patterned.
In curved girders, prestressing can be used to mOdify not only
the bending moments in the concrete section but also the torsion
moments and the transverse stresses.
The cross section of a curved bridge is usually designed as a
hollow box. In principle, T cross sections are also applicable. They
are, however, weaker in torsion and exhibit larger transverse
stresses. The subsequent deformations are therefore larger for T
cross sections than they are for box cross sections.
1.3

Bearings and Constraints

Curved bridges should always be supported by torsion resistant

bearing arrangements at the abutments. Point bearings are often
used on the columns. Single columns with point bearings are suitable
especially for overpasses, since they provide optimal transparency
and roadway space. The rigidity of the system is greater, however,
when torsion resistant bearings are used at the intermediate supports
as well as at both the abutments (Fig. 5).
The displacements at the free end of a curved cantilever
girder, due to temperature changes and shrinkage on one hand and due
to prestressing and creep on the other, do not take the same
direction (Fig. 6). Bearings and expansion Joints are generally
designed to allow free displacements only along the tangent to the
bridge axis. Transverse restraint from the bearings results in
redundant forces due to temperature changes and shrinkage. These
forces are usually not significant. Bridges with a great degree of

K;7
, _-,
....

./

Moments occording to
th~ory

of elasticity

Continous girder

Fig 4

FIG. 4:

MOMENT REDISTRIBUTION

redistributed moments;
Theory of plasticity

Conti lever construction

a) Torsion resistant bearings at

b) Torsion resi stant bearings at abutments;

all supports

Point bearing at columne

Fig 5

FIG 5:

<p

ARRANGEMENT OF BEARINGS

'Po

=wLl.T+t. cs

Fig 6

FIG 6:

DISPLACEMENTS OF A CURVED GIRDER

(P)

10
curvature could conceivably be designed without any Joints.
design, although certainly feasible, is seldom used.

2.

Such a

CALCULATION OF SECTION FORCES

Calculation of beam section forces based on the theory of

elasticity is relatively tedious. Therefore, the computation is
usually performed with the assistance of a computer program. In the
theory of elasticity, both equilibrium and compatibility conditions
are satisfied. The determination of the deformations due to bending
and torsion, however, is based on the assumed rigidity of the
homogeneous concrete cross section and the assumed linear elastic
behavior of the ~~terials. At high stress levels these assumptions
are of only limited applicability, so even the results of such an
elastic computer analysis must be regarded as approximate.
A simple method for the calculation of the beam section forces
can therefore be of significant value. Such a method is presented in
the following.
This
method
always
satisfies
static equilibrium. Compatibility conditions are satisfied only for an infinite torsional
stiffness GK. Such a calculation of the beam section forces is very
simple and is hardly less reliable than the I exact' theory of
elasticity. Indeed, as long as static equilibrium is satisfied,
such an approximation of the compatibility conditions is acceptable.
This simplified method is therefore also suitable for the
determination of optimal prestressing.
The equations of equilibrium for a curved girder element (Fig.
7) are as follows:
~

+ qds

dr + M"d<j) + (eoq + t) ods

Todcp - Vods
dM

0
0
0

Thus,
d2 M
ds 2
dT
Os

1dI'
-(q - FOs)

M
-(F + eq + t)

=-~

The bending and torsion moments in the curved girder can in

principle be calculated directly from these relations in the same
manner as for a straight girder. An approximate method based on the
requirements of static equilibrium may be applied here to some
advantage.
For

curved

continuous

girder,

the

bending moments

are

Fig 7

FIG. 7:

GEOMETRY AND SECTION FORCES

simplified
method

M*
A -125.00

T*
0

-12550 -0.023
- 0.278

2.863

1.042

62.461

1.047

62.50

62.50 -2.344
0

2865

exact
method

-4.167

62.549 -2.339
0

-4.164

Fig8
FIG. 8:

EXAMPLE OF PROPOSED METHOD

12

calculated first as if it were a straight girder, satisfying the

compatibility conditions of the straight girder. The radius of the
straight girder is infinite, so the torsion term in the bending
moment relation is neglected. The torsion moments are subsequently
calculated from the bending moments resulting from the straight
girder analysis, making use of the compatibility conditions for
torsion. The method may be summarized as follows:
d2M

ds2
dT

as

-q

_(!i
+ e'q +
r

t)

This method guarantees satisfaction of static equilibrium in

the curved girder, but the conditions of compatibility according to
the theory of elasticity are not precisely met.
I f necessary, the bending moments may be recalculated in a
second step, taking into account the calculated torsion moments.
The proposed method converges, however, very rapidly in the case of
normal angles <PO
l/r. The first step of the calculation is
normally sufficient.
A comparison of an exact analysis with the one-step simplified
analysis is shown in Fig. 8.
Calculation of the transverse internal stresses requires
additional investigation. On a girder element, the deviation forces
from the normal stresses are in equilibrium with the differential
shear flow (Fig. g).
The shear flow consists of two parts: the shear flow resulting
from st. Venant Torsion and the shear flow resulting from the
longitudinal
bending
of
the
cross-sectional
panels. Exact
calculation of these two parts is relatively difficult. It is based
on assumptions concerning transverse and longitudinal stiffnesses
and is therefore not of significant importance in the design of the
reinforcement. In general an estimation of the two parts of the
shear flow is sufficient: the deformations of the cross section due
to transverse bending and longitudinal bending of the panels must be
about equal.

PRESTRESSING

An arbitrarily acting prestressing force P (Fig. 10) yields the

following section forces in the concrete cross section of a
statically determinate curved girder:

13

d+dv h+dT

----.. ----....T ~

M
o . d., =ro .ds =r:ti
. ds

dT =- M d.,. = dv

M
r
ds

=.....2L= ___M_ds
2bh

2bhr

rh

2rh

Fig 9
FIG. 9:

DEVIATION FORCES AND DIFFERENTIAL SHEAR FLOW

t
C = Shear centre
S = Centre of gravity
y

Fig 10

p
FIG. 10:

COMPONENTS OF PRESTRESSING FORCE P

14

v cy
da
da
-p x (ay .~
~.)
ux - a Z CI'X"

pxa y
daz
da
-p x '[(a-c
(a-c
...::::::yj
y y )'=-:-ux
Z
z ) ax:
AS can be seen, torsion in the concrete cross section may be
caused by tendons in the webs as well as by tendons in the flanges.
In the following, the required arrangement of the prestressing
tendons in order to ensure a specific distribution of the torsion
moments (e.g. Tcp = - Tcg) will be presented, presuming the girder
cross section is a symmetrical one cell hollow box.

3.1

Tendons in the Webs

Tendons located only in the webs of the curved girder result

the following section forces in the concrete cross section:

TCp '1 0

_a l
y

a yr

da

-b/2

Vcy

Vcz

Ncy

Mcz

-..

Mcz

ail

-..

Vcy

a zr

_a l
z

Mcy

da~

Vcz

dar
z
CI'X"

= CI'X"

As a result the torsion moments become

Tcp

da~
Px
- 2 ' b CI'X"

and the derivative provides the corresponding torque:

mtp

= - d:~p = ~

d2 a l

1 b

qpz

~n

15
The distribution of torsion moments in a simply supported
curved girder under dead load is shown in Fig. 11. The equalizing
prestress torsion Tcp = - Tcg requires the following geometry for the
tendons in the web on the outer side of the curve:
da;

2Tc

--ax

Pxb

azl

2.

fTc.dx

and the reverse geometry in the web on the inner side of the curve
(Fig. 12).
This tendon geometry for resistance of torsion may always be
superimposed on the tendon geometry designed for the resistance of
bending. Therefore, in simply supported girders, equalization of
torsion and resistance to dead load bending may be provided with the
same tendons (Fig. 12).
The distribution of torsion moments in a continuous curved
girder under dead load is shown in Fig. 13. This distribution is
clearly different from that in the simply supported girder under the
same load (Fig. 11). Redundant moments due to prestressing designed
to resist bending influence the torsional moments,
generally
resulting in the torsional moment distribution shown in Fig. 14.
In this case the equalizing prestress torsion TCl' = - Tcg
requires the following geometry for the tendons in the web on the
outer side of the curve:
da; _

--ax -

~_~

In contrast with the simply supported girder, this tendon

geometry for equalization of torsion can not always be superimposed
on the most effective tendon geometry for resistance of bending (Fig.
15)

3.2

Tendons in the Flanges

Tendons located only in the flanges of the curved girder result

in the following section forces in the concrete cross section:
pS

Tcp

t-

Mcy

pi

Px

_ps.a s
x z

pi.ai
x Z

....

as
z

-(h-a zi )

Vcy

Vcz

Mcy

Mcz

p~, pi
x

16

Fig 11

FIG 11:

TORSION IN A SIMPLY SUPPORTED CURVED GIRDER UNDER DEAD LOAD

a:

b.

!=

=-

~
c:

~.
~

a:
b:

Fig 12

FIG.

12:

-~~

=:24

equalizing torsion moments

equalizing bending moments
c:
superposition

TENDON GEOMETRY FOR A SIMPLY SUPPORTED CURVED GIRDER

17

Fig 13
FIG 13:

TORSION IN A CONTINUOUS CURVED GIRDER UNDER DEAD LOAD

Fig 14
FIG. 14:

TORSION MOMENT INCLUDING REDUNDANT MOMENT

DUE TO PRESTRESSING

a;

Fig 15
FIG. 15:

a:
b:

equalizing torsion moments

equalizing bending moments
c: superposition

TENDON GEOMETRY FOR A CONTINUOUS CURVED GIRDER

18

da z

---ax

Vcz

Mcz

Vcy = 0

s
-p~. y

pi.ai
x y

da s
-pstfx
x
x

P~tfx

-+

da i

-+

a;

as
y'

da s
da i
tfx, tfx

As a result the torsion moments become

.
da i
s da i
-p x .a z .tfx
Tcp = pl..htfx
x
x
x
d s

pS.h.~
x
x

Tcp

. da s
-p .al."tfx
x z
x

and the derivative provides the corresponding torque:

d2 a i

dT

~p

px

a~.~y

2 s
l.. d a v i s
pxaz~
azqpy

mtp
3.3

~~

Transverse Stresses due to Prestressing

The deviation forces of the tendons consist in general of three

components:
K
Deviation forces qpy
due to the curvature of the girder in the
qK
= _ K
x-y plane
py
r
2

Deviation forces due to qpy

PY

Deviation forces due to qpz :

As a result the deviation forces
following torque about the shear center:

~
da

dx

d2 a

__
z

dx2
cause

the

- qpy.(az-c z }
qpz .(ay - c y )
The deviation forces q; of the tendons are in equilibrium
with the deviation forces
the normal stresses in the concrete.
y

cf6 o'l

qpy

qcy

Based on equilibrium and the distribution of normal stresses

19

0cx the corresponding transverse stresses due to ~y can be readily

determined. The deviation forces qpy and qpz (Fig. 16) as well as
the torque mtp(~ ) and mtp(qpz) are in equilibrium with the
differential shear ~low.

qpy +

~~y
dV c

-?xqpz +

dV cz

dT(V C:i)
ax
dA c

Ac

----a:x-

dV cz

J dT(Vdxcz ) .dA c

----a:x-

Ac
dTc(qE:i)
mtp(qpy) +
ax

dTc(qEZ)
ax

mtp(qpz) +

dT(T )

c
---ax :; f --axdTc

dA c

The transverse stresses due to these deviation forces and the

applied torque are calculated according to the distribution of the
differential shear flow. In the case of a constant cross section the
differential shear flow exhibits the same distribution as the shear
flow due to the shear forces Vcy and Vcz and the torsion moment Tc
(taking into account only St. Venant Torsion) (Fig. 17).
The deviation forces qpy and qpz can be separated into two
independent loading cases:
1: The resultant of the deviation forces passing through the shear
center (no torque component).
2: The
deviation
forces
causing
only
torques
(the
torque
resultant)
For each loading case the transverse stresses can be calculated
taking into consideration the equilibrium with the differential
shear flow (Figs. 18,19).
Superposition of the different transverse stresses shows that,
supposing the geometry of the tendons such that Tc :; - Tc ' and
respectively Vcy :; Vcz :; 0 (Section 3.1), for tendonsP~n the webs the
transverse stresses become twice as large, but for tendons in the
flanges the transverse stresses can be compensated for.

20

Fig 16
FIG. 16: DEVIATION FORCES qpy' qpz
foTo(-Vcyl

y-

Fig 17
FIG. 17:

SHEAR FLOWS DUE TO -Vcy ' -V cz ' -Tc

21
qpy in C

I~

. 11

Il,~=~ll

dVCY )
dx

S (qPY)

S --

qpz in C

I~

It

11

dVCZ)
S (-

S (qpz)

dx

m,p (qpy)

It

S (m,p)

(~~

--

~I

If" "JI

~I

~I//=/'It

-----.-

m,p (qpz)

If

S (m,p)

Fig 18

It
S(

~~

22

S(mtp)

m,p(qpz)

=-

m,

It

Fig 19

FIG. 19:

S(m,p)

TRANSVERSE STRESSES DUE TO TORSION AND PRESTRESSING

23

CERTAIN ASPECTS OF BRIDGE DESIGN

Prof.Dr. Fritz Leonhardt

Stuttgart, West Germany

ABSTRACT. During tllese lectures at the NATO Advanced Study Institute

it is my intention to treat certain aspects of bridge design that
stem from my experience of several decades as a researcher and
bridge designer. From the viewpoint of general interest as well as
applicability, the following topic has been chosen for the first
part of the presentation : the importance of partial prestressing.
For the second session of lectures, as per the request of the
organizers, it is my intention to treat the design and construction
of cable-stayed bridges.
1.

PARTIAL

PRESTRESSIL~G

The behaviour of fully prestressed concrete structures,

especially bridges, during the last 30 years has not been satisfactory
in all cases. Structures with I-beams for high ratios of live load
to dead load often developed camber due to elastic and creep
deformations of the precompressed tension chord member. The large
number of tendons necessary for full prestressing reduce the
concrete area in the critical section and do often not allow
sufficient cover and spacing.
In quite a number of bridges, mainly continuous box girder
bridges, cracks appeared in zones which should have had only
compressive stresses according to design calculations. Such cracks
had a rather large width, in one case 5 rnrn, proving that the small
amount of reinforcing bars was stressed above yield at the moment
of cracking. It was found that these cracks had mainly been caused

24

by internal restraint forces due to temperature differences or

differences of shrinkage and creep, especially if thin slabs are
connected to thick portions of the structure. Some of these bridges
almost failed due to these temperature cracks because the amount of
reinforcing bars was too small. The possible range of these
temperature differences was measured in several countries and was
found to be as large as about 25 0 K. If the stresses caused by such
temperature differences are calculated for the homogeneous concrete
structure, then the engineer gets worried because these stresses
can eas ily reach values fa r above the tens ile strength of conc re te
or close to about half the amount of live load stresses. Therefore,
if tensile stresses due to these temperature effects would have to
be avoided by additional precompression, the amount of prestressing
force would have to be increased considerably, causing difficulties
in structural detailing, mainly for finding the necessary place for
these additional tendons and their anchorages. In addition transverse
vertical and horizontal prestressing would be necessary, which would
make construction work far too complicated.
However, it would, indeed, be wrong to increase the degree of
prestressing for avoiding such tensile stresses, because these
restraint forces decrease and in many cases almost disappear as
soon as a few cracks in the concrete relieve the degree of restraint
by decreasing the bending or tensile rigidity (Fig. 1). Therefore,
these restraint forces can practically be neglected as far as the
safety against failure for the ultimate limit state is concerned.
This ultimate limit state for full load actions governs the
dimensioning of the necessary steel area (prestressing steel +
rebars) for the tension chord members in girders under bending and
no additional steel is needed against temperature restraint forces,
if a sufficient portion of this steel is used for crack control and
securing ductility.
It is well known and verified by many tes ts that fine hair
cracks in concrete structures have no adverse influence on the
durability or on safety against corrosion. The task is only to keep
the crack width limited to harmless values.
Experience with testing of large scale prestressed concrete
girders shows that the cracks, which must come if the load goes up
above service load in order to find the ultimate capacity, will open
quickly if the anxJUnt of reinforcing bars is ve ry small. This unfavourable behaviour is due to the fact that the bond strength of
grouted tendons, especially of large tendons, is only between 20
and 40% of the bond strength of normal reinforcing bars (1).
Therefore, the bond gets destroyed over a considerable length if
the rebars are stressed above yielding at cracking and are,
therefore, not able to prevent this loss of bond.

25

Service load State

Ultimate Limit State

brittle failure

~ ductile

I~red

by AT

--I ~ deformation
eAT

Fig. I

f.i. curvature

failure

by 1.5 AT

Restrained forces, like M~T = Moment due to ~T caused

by restrained deformation, decrease under load effects
above service load towards ul timate load; high degree
of prestressing leads to brittle failure.

We know from tests that, in such cases, we even do not obtain

the full capacity which we calculate under the assumption of a
perfect bond corresponding to a straight line strain diagram. This
means that not only for crack control but also for ultimate safety
a larger amount of rebars is needed if we use the big size grouted
tendons for post tensioned structures.
Of course, this additional steel for rebars contributes to the
ultimate load capacity and allows a reduction in the amount of
prestressing steel. Simultaneously the post cracking behaviour of
the structure becomes more ductile and less sensitive against
unforeseen actions or against temperature stresses.
All these facts lead to the conclusion that we get a better
behaviour of our prestressed concrete structures by reducing the
degree of prestressing and increasing the amount of unstressed
reinforcing bars, or in other words: partial prestressing improves
the serviceability performance and simultaneously improves ultimate

26
safety, if the structural details are well designed.
1.1

Performance requirements for serviceability

Besides the ultimate safety against load actions there are

performance requirements for serviceability to be observed, which
relate to admissible deformations and crack width, to durability
and safety against corrosion (2).
For most structures, however, fine hair cracks can be allowed
and the prescribed limit of the crack width can be secured with a
low degree of prestressing or even with no prestressing, if the
reinforcing bars are designed in order to fulfil the conditions
for crack width limitation as described below. Even a small amount
of prestressing helps to limit the crack width, because the jump
of steel stress at cracking gets reduced considerably by a rather
small amount of prestressing (Fig. 2). There is a wide range of
freedom to choose the necessary degree of prestressing, if the
area of prestressing steel plus the area of reinforcing bars
satisfy the ultimate limit condition. There are cases in which the
upward deflection due to creep of girders must be limited or kept
zero in order to fulfil serviceability requirements; the degree of
prestressing can then be governed by this condition, which cannot
be fulfilled with a high degree of prestressing.
The quality of the structure and also the durability are
practically independent of the degree of prestressing, if the
serviceability performance requirements and those for the safety
towards ultimate capacity are fulfilled.
1.2

Choice of the degree of prestressing

Practical experience suggests the following definition of the

prestressing degree :
(= moment of decompression after losses
moment for maximum service load

max ~L+LL

This definition has the advantage that the engineer can directly
see if the degree which he intends to choose will, for instance,
cover the full dead load moment, ~L' or will cover dead load plus
30% of live load or will cover only 80% of dead load.
Prof. H. Bachmannof Zurich has published charts giving the
necessary amounts of prestressing and reinforcing steel for
rectangular cross sections depending on ttle degree of pres tress ing
(Fig. 3). He further gives the increase of steel stress in the
prestressing steel and the steel stress in reinforcing bars due to
full service load, again depending on the degree of prestressing
(Fig. 4). For a prestressed concrete railroad bridge with I-beams
such charts were calculated in my office (Fig. 5 and 6). In these

27

300
~= res M

Ph

=-0.6}
=-0.4 prestressed with P
=-0.2

I for h =25 N/mm2

0.2%

Fig. 2

0.5%

1.0%

Jump of steel stress at cracking

e = related eccentricity of resulting normal force
Ap c~n only be added to As for well bonded pretens ~oned s tee 1.

charts a sufficiently large =n~mum percentage of reinforcement was

assumed to prevent opening of cracks at the cracking load. The
min Pr = As/Ac,ef' related to the affected concrete area, was
assumed to be about 0.5%.
The charts give the minimum of total steel quantity ~ + As at
a prestressing degree of f= 0.6 for the rectangular cross section
and of ~ = 0.82 for the I-beam. For higher prestressing degrees
additional steel is needed to fulfil
the condition of 0c = 0
(full prestress) or 0c < admissible Oct (limited prestress). This
additional steel is not needed for ultimate limit design. In the
case of the rectangular cross section, the steel needed for full
prestressingGe=l.O gives a global factor of safety of y = 2.7 where

28

1.8 (Swiss code) would be sufficient.

If the unit price for the prestressing steel compared to the

unit price for reinforcing steel is assumed to be 3:1 (dotted lines
in Fig. 3 and 5) one finds the minimum of cost for the necessary
amount of steel to carry the bending moments. This minimum is given
for a degree of prestressing between 0.45 and 0.6 depending on the
type of cross section. One has to check yet if the necessary amount
of reinforcing bars can eas ily be placed in the available concrete
section. The largest space in between such bars depends on the
allowed crack width. No crack width verification is needed, if
these spacings in the tension chord zone do not exceed the following
values :
Crack width limit mm
max.bar spacing mm

0.1

0.2

0.3

80

120

150

The charts of the steel stresses show that the stress amplitudes
remain far below the fatigue limits if for rectangular cross sections
the degree of prestressing is chosen above 0.4 and for I-beams above
0.5 and if the frequently repeated portion of the live load remains
below about 7070 of the dead load. This means that choosing partial
prestressing in most practical cases does not cause any fatigue
problems if the degree of prestressing and the ratio LL:DL are
within these limits. Therefore, even for railroad bridges partial
prestressing can be chosen.
1.3

Partial prestressing and shear or torsion

In some countries vertical or inclined prestressing bars in

webs in box or I-girders can be found lately in order to avoid
principal tensile stresses due to shear or torsion. In my experience
this type of prestressing is in most cases not necessary, because
inclined shear cracks have practically never been found in
prestressed concrete bridges. Such cracks would appear only with
loads causing bending moments well above the MD (moment of decompression), normally above full service load. This is partially due to
the fact that a part of the shear forces is carried inside the
chord members and does not cause tension in the web. Of course, the
spacing of the stirrups should be rather close, so that, if a shear
crack would develop, the crack width would remain very small.
One of the biggest prestressed concrete bridges in Germany is
the Kochertalbrlicke, Geislingen, which has a box girder with 2 webs
only 45 cm thick which have to carry a 6 lane freeway with a total
width of 30 m over spans of 138 m. TIle maximum shear stress under
service load in these webs runs up to 5.8 N/mm 2 and the principal

29

tensile stresses were calculated as 3.8 N/mm2 No web prestressing

was chosen for shear or torsion in spite of these high shear and
tensile stresses, but the stirrups for the shear reinforcement were
spaced down to 8 cm near the intermediate supports. Even additional
inclined stirrups were placed. But no shear crack could be found
so far. It is therefore my hope that those authorities, who still
stick to the request of full prestressing, may soon introduce codes
allowing partial prestressing in order to get the economical and
technical advantages of this design concept.

<~<'

~~!;J

Ap

A. = S 460 '550

A.

min. cost .,../'

/'

/'

=S 1500 '1700

.,/

/'

Up ,in = 0.75 fp ,0.2

/'
(Swiss Cod.)

3Ap

A
min. ~ =-'-~ 0.5'.
r Ac ,.'

./

10

pr.str.ssing

d.gr

~~~~--0~.4-----0~.6-----0~.8----~1D~x

Fig. 3 and 4

--

-..""

100

"

dUp

..........

Influence of the degree

of prestressing on the
necessary quantities of
p-steel Ap and rebar-steel
As and on the steel stress
due to full service load.

"

above G'p due to

prestreSSing

",
.................

L--------L~----~L-------~=_----~~~~--~--~~~~.x

0.2

0.4

prestressing

0.6

degree

0.8

1.0

30

Ap = S 1500 /1700
As

= S 450 /500

200

,~g
It>
(")

0.90

prestressing degree

Fig. 5 and 6

.l

Influence of the degree of

prestressing on the necessary
quantities of p-steel An and
reb ar s tee 1 As and on the s tee 1
stress due to full service load
for I-beam.

200

100

~----~~------~------~--------L------~--~--~~X

0.6

prestressing

degree

0.8

to

31
2.

CABLE - STAYED BRIDGES

2.1

The development to the multi-stay-cable system

The idea to support a beam by stays from a tower is very old.

The rebirth of the system must be. credited to F. Dischinger (l) who
pointed out the advantages of high stresses in stays of high
strength steel. The first modern cable-stayed bridges were beam
bridges with only 2 to 6 stay supports in the main span. The spans
between the stay supports were between 30 and 60 m and needed large
bending stiffness of the beam with a depth of 3 to 4 m. The stay
forces were large so that several ropes were needed to build up the
cables. The anchorages of these cables were rather complicated. A
considerable amount of auxiliary structures were needed to erect
such bridges (Fig. 7).

multi stay cable bridge

Fig. 7

mainly tension and compression

Development from stayed beam to the proper

multi-stay-cable system

In the further development it was found that most difficulties

disappeared if a large number of stays is used with spacings at the
deck anchorage of only 8 to 15 m, so that free cantilevering erection
is possible without any auxiliary supports (Fig. 7). This development
to the multi-stay-cable bridges practically led to a new type of
main girder system which could no more be defined as a beam girder.
It has its own proper qualities. What was formerly a beam girder,

32

has become mainly the compressive chord member of a cantilever

structure hung up to towers by inclined stay cables. Similar to
a chord of a truss, this compressive chord member does not need
much bending stiffness, because the triangle tower-stay-chord
gives abundant stiffness for getting deflection lines with
curvatures which fulfil the performance requirements for highway
and railroad traffic. Therefore, the depth of the longitudinal
girders or stringers in the deck structure is almost independent
of the main span and should be chosen small, in order to avoid
unnecessarily large longitudinal bending moments in the deck
structure.
This multi-stay-cable system is in fact a new system, different
from our classical systems of beam girders, arches, or suspension
bridges with stiffening girders. Comparative calculations showed
that this system deflects less under highway traffic loadings than
slender continuous beams or suspension bridges. It has also superior
qualities as far as the dynamic behavior is concerned, mainly due
to its large amount of system damping.
These superior qualities are obtained if highly stressed cables
of sufficient inclination are used so that vertical deflections
will be small. The cable stiffness is the paramount parameter in
this system,it is usually described by the formula
A
A

s Eeff

y2
1

1/,2

(1)

c E0

l2a'

where :
A
s

area of cable steel

modulus of elasticity of straight vertical cable
weight of cable (incl. corrosion protection) related to
A .1/,

I/,c

s c
horizontal span of cable

tensile stress of cable, influencing the sag of cable

The evaluation for Eeff is shown in Fig. 8. The stiffness of

such cables increases with the third power of the steel stress and
decreases with the second power of the horizontal span length due
to the sag effect (change of sag by change of stress). The relation
between"live and dead load has influence on the dimensioning of the
cables and therefore on the stiffness under dead load conditions.
Depehding on this p:g ratio the cable stiffness might become
insufficient for I/,c > 250 m or for main span lengths of 500 m.

33
However, is is easy to reduce the sag effect for longer cables by
introducing so-called stiffening ropes as shown in Fig. 9. In this
way the high stiffness of the cable-stayed system can be maintained
for very long spans, up to main spans of 1500 m for railroad and
highway bridges.

180
140

60

O+---~----~--~--~~lc

100

200

300

Fig. 8
Effective modulus of
elasticity gives influence
of the sag of the cable
on its stiffne s s

400

T=Ac G'

stiffening ropes

Fig. 9

Stiffening ropes to reduce the sag effect on the stiffness

of the cables

34

2.2

The arrangement of the stay cables

There are many possibilities for the arrangement of the

cables. There is the fan-shaped configuration (Fig. 10) in which
all cables join at the head of the tower. In the harp-shaped
arrangement (Fig. 11) all cables are parallel and their anchorages
at the tower are distributed over the height of the tower. This
arrangement needs mo~ steel for the cables, results in higher
compressive normal forces in the deck, and causes bending moments
in the tower. From a technical and economical view it is, therefore,
interior to the fan-shaped arrangement, but for the appearance of
the bridge it might be superior, because all cables look parallel
also in view under a skew angle. This aesthetic advantage was
decisive for the choice of the harp-system in the famous Dusseldorf
bridge family crossing the River Rhine (4).
If it is requested that cables should be easily replaceable
in cases of accidents, then it is difficult to realize the pure fan
arrangement in which all cables join theoretically at one point

Fig.lO Fan-shaped configuration of stay cables

Fig. 11 Harp-shaped configuration of stay cables

35

Fig. 12 Fan-shaped arrangement. anchorages of cables at tower distributed

above the tower top. For replaceable cables it proved to be simpler

to have the anchorages of the cables distributed vertically over a
certain length of the tower. This results in a cable configuration
as shown in Fig. 12. The length of this distribution of the
anchorages can also be larger but then bending moments in the tower
leg are caused by differences in the horizontal components of the
cable forces due to certain live load positions.
Of course, other configurations of the cables are possible
mainly depending on local conditions for the ratios between main
and side spans. A harmonic arrangement of the cables is important
for the aesthetic quality of such bridges and, therefore, the
choice should be made with care and diligence.
2.3

The ratio between main and side spans

The ratio between side span 1 and main span has influence
on the stress changes mainly of the back stay cables, which hold
the tower head back to the anchor pier. Live load in the main span
increases these stresses, live load in the side span decreases them,
and in long side spans with 1/ > 0.4, the cables could become
more or less slack. The back stay cables get the largest stress
amplitudes of all cables and these amplitudes must be kept safely
below the fatigue strength of the cables for that part of the live
load which may occur a million times (for instance at highway
bridges 40% of maximal live load).
The ratio 1/ has further influence on the amount of vertical
anchoring forces at the anchor pier. This anchor force decreases
with increasing 1 / . Towers on both sides of the main span are
assumed here. If there is only one tower on one side (unsymmetrical
case) then the main span acts similar as if the span length is
about l.8.

36

A good choice of the ratio between side and main span is

important for a good and economical design. As .an aid for this
choice a chart has been calculated in which we find the parameter
p:g plotted vertically, and horizontally the length of the main
span in meters (5).
Of course, the choice be tween the length of s ide and main span
depends also on local conditions of water depth, of foundation data
or on the wish to have a tower on one side of the main span only.
2.4

The optimal height and stiffness of towers

The height of the towers has influence on the necessary amount

of cable steel and on the longitudinal compressive forces in the
bridge deck. The higher the tower, the smaller will be the quantities
of the cable steel and the compressive forces. The curves in Fig.13
show that it is of no use to make the towers higher than about 0.2
up to 0.25, because one has also to consider the quantities needed
for the tower. For bridges with the tower on one side the h/ must
be related to 1.8.
In the longitudinal direction the towers should be slender and
have a small bending stiffness, so that live loads in the main span
do not cause large bending moments in the tower but get the back
stay cables acting. Transmitting the unbalanced horizontal components
from the top of the towers to the ground by back stay cables is much
more economical than by bending resistance of the towers.

Vl

fan shape

harp shape

1i~:~ ~:~=th

J.--/--J.

.:;-3
c
c

a;

Fig. 13
Quantity of cable steel as
a functIOn of relative height

iii

of tuwers

:l

CT21----\l---'<--,--~--_.--

QJ

L.

2
u
.E

37
Longitudinal bending stiffness of towers, which is characteris tic for Morand i' s early stayed beam bridges, ge t ve ry large
moments, which must be carried by the foundation and can easily
double their cost. In order to avoid this, some towers of cablestayed bridges have been built with foot hinges so that the foundations are concentrically loaded (Mannheim and Oberkasse 1 bridges
across the River Rhine) (4, 6).

0)

C3[~-1
-

__ ...J

section 1-1

cl

b)r -, I

i
low

Flg. 14
Two cable planes along
the edges of the deck
with different solutions
for the towers

2.5

The arrangement of the cables transversely and cross sections

of the deck structures

In normal cases the bridge deck should be hung up along its

edges, resulting in two planes of cables and two towers, standing
just outside of the railing of the bridge deck. The towers should
get an unsymmetrical cross section with most of the load carrying
area and consequently the center of gravity close to the bridge
deck, so that the cable planes can be vertical or must be only
slightly inclined (Fig. 14a). The wind loads acting on the tower
by the cables are small so that no horizontal bracing between the
two towers is necessary, if the tower legs are tapered and fixed
in the foundation.

38

If the height of the towers is considerably larger than the

width of the bridge, then a horizontal beam connecting the towers
at the top may be useful, allowing a small inclination of the
cable plane or of the tower legs, so that the cable planes can be
kept vertical (Fig. l4b).
For long spans, A-shaped towers add to the good appearance of
such bridges, because all cables join at the one tower top (Fig.14c).
A good example is the Pont de St.Nazaire with a main span of 404 m,
crossing the River Loire near its mouth (7). Joining the cables of
the two planes at tne top increases also the torsional stiffness
of the brid ge deck.
For high level bridges it is desirable to join the tower legs
under the bridge deck in order to narrow the necessary width of
the foundation, as was done in the design for the Faroe Bridge in
Denmark (Fig. l4d).
The cross section of the deck structure can be very simple,
if the bridge is hung up along its edges. No torsional rigidity is
necessary because the cables give a stiff support along each edge
and the deflection is small, so that unsymmetrical loading gives
only a very small transverse inclination of the deck. Therefore,
for a width of the br~dge up to around 15 m, a simple massive or
hollow concrete slab with ribs along the edges is sufficient (Fig.
l5a). The edge rib allows us to anchor the cables at any point and
ensures the buckling safety.
For wider bridges, cross girders are necessary which should
be arranged with a spacing of only 3 to 5 m, so that the concrete
slab or the orthotropic steel plate can easily span longitudinally.
Here most of the reinforcing bars of the steel stiffening ribs run
longitudinally and help to carry the compressive normal forces
(Fig. l5b and c). The concrete deck slab is advantageous for main
spans up to about 6QO m, even if steel girders are used transversely
for constructional reasons. Composite action should be provided. If
the deck slab is of concrete, then all longitudinal girders should
also be of concrete so that no creep problems arise due to the high
longitudinal compressive stresses. The all steel bridge with an
orthotropic plate deck becomes mandatory for the very big spans in
order to reduce the dead loads. Even for very large highway bridges
longitudinal edge girders with a depth between 1.0 and -2.5 mare
sufficient and can be used for anchoring the cables.
2.6

Situation at the ends of the cable stayed bridges

Near the ends of the side spans or of a main span hung up

to one tower only, there are the only regions with large bending
moments. As a consequence the angular changes of the deflection
line at such free ends of the longitudinal girder are large. This

39

a}

b)

f . ,,""" "1""" f"" T

>15m

'''CO M"" co

c)

~2t"

:I~
\

~~

ii?~~

Fig. 15 a

Cross section of a concrete bridge with a wIdth b

<..

Fig. 15 b c

Cross section of concrete or steel bridge with b

> 12

15 m
m

is acceptable for normal highway bridges but is causes difficulties

for railroad bridges which can easily be avoided, if tne girder
continues with an increased depth into a small approach span. This
solution can generally be recommended, if approach bridges follow
behind the side or main spans. By such continuity the uplift forces
of the back stay cables can be counteracted by the weight of this
adjoining span and by ballast concrete within the depth of this
girder extending to both sides of the anchor pier (Fig. 16). It
might be sufficient to have only a cantilever, which carries a
hinged bearing of the approach bridge. The continuity allows us
also to distribute the anchorages for the back stay cables over a
certain length behind the axis of the anchor pier.

40

side span
appnooch span

Fig. Hi

2.7

'

anchored

Continuity to the approadl bridge allows lengthwise

distribution of the back stay cables

Arrangement of bearing and joints

Vertical bearings should only be arranged at the end of the

side spans but not at the tower, where it is better to continue the
elastically deformable support condition which is given by the
stay cables. If stiff vertical bearings at the towers would be
chosen, then large lon~itudinal bending IOOments would cause trouble
and require more bending stiffness than necessary for the rest of
the bridge.
The bearings for horizon tal transverse loads, like wind loads,
mus t be arranged in a way to allow angular changes of the wind
girder in a horizontal plane. At the towers these wind bearings can
be simple rubber pads which act directly against the tower legs and
which should have an open gap of about 3 mm for the unloaded
condition to allow vertical and aorizontal movements of the deck
easily. At the end of the side span the wind bearing can be placed
in the center-line of the bridge. It must allow angular changes
in horizontal and vertical planes and longitudinal movements of
the bridge.
For horizontal loads in the longitudinal direction, which are
mainly caused by braking forces, different arrangements are possible;
for symmetrical bridges no fixed bearing is needed, if at the end
of each side span speciay fluid bearings are used, which allow for
the slow temperature elonge.tions without much resistance but which
are sufficiently stiff to react to breaking forces without much
deformation. In this case two expansion joints are needed.
In seismic regions it is easily possible to design the bearings
with bolted steel angles limiting the movement for normal service

41

conditions, which, however, break away if extreme seismic amplitudes

occur. Shock absorbers have to damp further excessive movement, so
that the big mass of the deck structure would not cause damaging
forces to the towers and piers. In this way cable-stayed bridges
can be made very safe against earthquake attacks.
2.8

Maximum span length

The experience gained by designing and constructing a large

number of cable-stayed bridges during the last 25 years allow us
to state that rnulti-stay-cable bridges as described can be built
for highway traffic with spans up to 700 m, for railroads up to
about 500 m with prestressed concrete. Designs for steel bridges
have been made with main spans of about I 300 and I 500 m (6 lanes
highway and 2 tracks railroad) for the Messina Straits Crossing.
No structural difficulties were found (5).
I

Fig.17
Anchorage of HiAm-cables
in the tower and in the concrete edge girder of the
deck

42
2.9

The towers

The shape of the towers is sometimes a subject for special

architectural treatment. As engineers we should try to keep the
forms as simple as possible, making use of tapering, good proportions and suitable profiles of the cross sections to obtain a
pleasing appearance.
Experience proves that concrete towers are cheaper than steel
towers, and the cost difference increases with the size of the
bridge. Box sections are preferable to solid sections, so that
access to the cable anchorages at the top can be provided inside
the tower shafts.
Over the years, many different types of saddles or anchorages
for the cables have been designed. However, there is now
wide
agreement that the cables should be anchored in a way which allows
easy replacing of each cable. This led to a rather simple solution
as shown in Fig. 17. The cable socket is pulled through a steel
pipe embedded in the concrete and a ring nut is turned on the
thread, to hold the socket in place. The anchor chamber in the
tower head must be sufficiently large to handle the equipment which
is needed for this work. The horizontal tensile force, which results
from two opposite anchor forces, is counteracted by horizontal
prestressing tendons just outside the anchor chamber, which keep
the concrete of the tower head under horizontal compression.
2.10 The cables and their anchorage
The cables are the most important members of this system,
they must, therefore, be safe against fatigue and corrosion. The
best quality cables must be chosen. A large amount of testing
results and practical experience is available for the judgement,
which type of cable would be the best (8, 9, 10, 11). In this
lecture there is not sufficient opportunity to present this knowledge in detail. The following recommendations can be given, based
on this knowledge :
1. Steel ropes, protected only by painting should not be used for
large cable-stayed bridges as such ropes have failed at several
large bridges and have had to be replaced.
2. Parallel wire or parallel strand cables are superior to ropes by
their fully elastic behaviour and well defined modulus of
elasticity.
3. The corrosion protection should be absolutely secured by placing
the steel of the cables inside a tube, which is tightly connected
to the anchorages. The tube can be of black polyethylene (PE),
which has been proven to keep its qualities for at least 20 years
and will probably need no maintenance for 40 or 50 years, if

43

correctly handled during the transport, erection and injection.

4. The fatigue strength of the cables depends mainly on the fatigue
strength of the anchorage. The normal zinc filled sockets of
ropes give only a rather low fatigue strength, because the high
strength of the wires gets damaged by the high temperature,
which must be used for pouring the metal. Fatigue tests proved
that the stress changes, which can safely be carried by such
anchorages, are as 100 to 120 N/mm2 for large ropes with a
diameter bigger than 80 mm.
Tlick walled steel pipes around the cables can also be used to
help carrying the live load. As a result the stress amplitudes
of the wires are reduced. This was done at the Main Bridge in
Hochst (12) and at the Brotonne Bridge (13).
5. Special anchorages have been developed, for instance the BBR
HiAmranchorage (9) with a cold filling material in the conical
cone or other types like the one of Freysinnet International,
which was used for the Vi go Bridge in Spain (14).
2.11 Dynamic Behaviour
If stiff cables are chosen with Eeff > 180 000 N/mm 2 then
very favourable dynamic behaviour of the cable-stayed bridges can
be expected. This is especially true for the aerodynamic behaviour
and can be traced to the following phenomena :
1. The strain behaviour of the cables is non-linear due to the
influence of the sag effect.
2. Each of the many inclined cables with the mass of the deck
belonging to it has a different frequency. Whenever forces act,
which excite oscillations of the bridge in a certain mode, then
the development of the amplitudes is immediately interrupted by
the interference of cables with different frequencies. The
mu1ti-stay-cable system develops hereby a system damping of normal
concrete or steel structures like beams , arches or suspension
bridges. Due to this system damping resonance oscillations with
large amplitudes are impossible and only such reasonance oscillations can become dangerous for long span bridges.
This favourable dynamic behaviour was first proved by a few
tests at the 90 m long cable-stayed pedestrian bridge across the
Schillerstrasse in Stuttgart, a light weight steel bridge (only
150 kg/m 2 dead weight), which can easily be excited but develops
amplitudes of not more that about 5 mm. The second proof was
given by dynamic model tests at the ISMES Institute of Bergamo,
which were conducted for the railroad bridges over the Parana
(15).Fig. 18 shows a typical oscillation diagram from these tests.

44

~l~
..J...--.330 m --,j.

[4i)

kN/m

6,0

L35m 1 98.3m

10000n"g

train

at

V: 108.8 km/h

system damping
(mm) 0 102030405060

t(s)

-2~

-400

-6:'"
-80.,

measured values related to 1= 330 m

Fig. 18
Oscillation diagram
cable-stayed bridge

of a

The good behaviour was confirmed by the experience during the

free cantilevering construction of these and other bridges and
also by the behaviour under railroad traffic.
This system damping prevents any resonance oscillations and
secures hereby the aerodynamic safety even for very long spans,
if a large number of cables and sufficiently stiff cables are
involved and if the span/width ratio is not larger than about
40. No aerodynamic shaping is necessary. Theories, which have
been developed to verify aerodynamic safety of suspension bridges
have, therefore, only limited validity for these multi-stay-cable
bridges; the same is true for sectional model tests in wind
tunnels, as long as the system damping cannot be imitated with
sufficient similarity in such tests.
The system damping cannot prevent torsional oscillations of
bridges which are hung up by cables in one plane along the
center-line; furthermore, the frequencies of bending and torsional
oscillations are independent of each other. There may be a
danger of getting torsional resonance if the box girder has a
shape which would give a large cM factor (shape factor related
to pitching moments), so that the wind forces could start
torsional deformations.

45

Our knowledge in this respect is still limited and, therefore,

a diligent investigation must be recommended. Another dynamic
problem is the oscillation of the cables themselves, as was
observed at seve ral cab le-s tayed bridges, espec ially at the
Brotonne Bridge, where car dampers had been installed at the
roadway level. The authors use neoprene dampers at the end of
the steel pipe, in which the end of the cable is enclosed over
a certain length near the anchorages (Fig. 17). This neoprene
damper has to prevent also angular changes of the cab le at the
anchorage which would cause additional bending stresses. So far
these dampers have actually prevented amplitudes larger than
about 20 mID in cables with lengths up to about 150 m, if the
cables have cement injection inside the PE-tubes, which give a
high damping to the cab les themselves.
2.12 Cable-stayed bridges for railroad
The strong damping of multi-stay-cable bridges makes tnis
system very suitable for railroad traffic, especially for modern
high speed railroads. Of course, the dead load masses should be
large so that dead load stresses in the cables are sufficiently
high to make them very stiff. Sufficient dead load is easily
obtained, if the ballast for the tracks continues over the bridge
as is requested now by most railroad companies.
Further, the deck should be a prestressed concrete structure
for main spans Q,<400 m. The depth of the longitudinal girders must
be slightly larger than for highway bridges in order to keep the
curvature and the change of the gradient of the deflection line
sufficiently small. Short side spans and a combination with highway
traffic help to reduce the stress amplitudes. Several long span
cable-stayed railroad bridges have been built or are under
construction.

REFERENCES

1.

Leonhardt, F. : ItVorlesungen Uber Massivbau-Spannbeton lt

Volume 5, Stuttgart, 19.

2.

CEB-FIP Model Code, Table 15.1, Paris, 1978.

3.

Dischinger, F. : HangebrUcken fUr schwerste Verkehrslasten

Der Bauingenieur, 3/1949, p .65 and following and p.107 and
following.

4.

Tamms, F., Beyer, E. : KniebrUcke DUsseldorf Book edited by

Beton Verlag DUsseldorf 1969.

46

5.

Leonhardt, F. and Zellner, W. : "Cable Stayed Bridges",

IABSE Surveys S-13/80, ZUrich, May 1980.

6.

Volke, E. ; Rademacher, C.-H. : NordbrUcke Mannheim Ludwigshafen (Kurt-S chumache r-BrUcke) De r S tah lb au 1973, Heft
4, p. 97, Heft 5, p. l38, Heft 6, p. 161.

7.

Vanbourdolle, M, Ciolina, J., Bacarrere, J.: Le pont de Saint

Nazaire-Saint Brevin , Annales de l'Institute Technique du
Batiment et des Travaux Publics 347 (1977), pp. 13-43.

8.

Andra, W., Zellner, W. : Zugglieder aus ParalleldrahtbUndeln

und ihre Verankerung bei hoher Dauerschwellbelastung, Die
Bautechnik, 1969, Hefte 8 and 9, pp. 263-268, 309-315.

9.

Andr~, W., Saul, R. : Versuche mit Blindeln auz parallelen

Drahten und Litzen fUr die ~ordbrlicke Mannheim-Ludwigshafen
und das Zeltdach in Mlinchen, Die Bautechnik, 1974, Hefte 9,
10, 11, pp. 289-298, 332-340, 371-373.

10.

Andr.1, W., Saul, R. : Die Festigkeit, insbesondere Dauerfestigkeit

1anger Paralleldrahtblindel, Die Bautechnik, 1979, Heft 4,
pp. 128-130.

11.

Freyssinet Cable Stays, Technical Outline, Brochure by

Freyssinet International, Ref. F1. 246A/08.79.

12.

Schambeck, H. : Bau der zweiten MainbrUcke der Farbwerke

Hoechst AG - Konstruktion und Ausflihrung Vortrage Betontag
1973, pp. 359-172, Wiesbaden, Deutscher Beton-Verein e.V.

13.

Mathivat, J. : The Brotonne Bridge, Proceedings of the Eight

Congress of the Federation Internationa1e de 1a Precontrainte,
May 1978, Part 2, pp. 164.

14.

De Miranda F., Leone, A., Passaro, A. : II ponte stra11ato

sullo stretto di Rande presso Vigo (Spagna) delle Autopistas
del Atlantico - Costruzioni Metal1iche 2/1979.

15.

Leonhardt, F., Zellner, W., Saul, R. : Zwei Schragkabelbrlicken

flin Eisenbahn - und Stra enverkehr User den Rio Parana
(Argentinien) - Der Stah1bau, 1979, Hefte 8 and 9.

II. ANALYSIS METHODS FOR STEEL BRIDGES

49

OVERLOADING AND INELASTIC RESPONSE OF STEEL MULTIGIRDER BRIDGES

Celal N. Kostem, Professor of Civil Engineering

Fritz Engineering Laboratory, 13, Lehigh University
Bethlehem, Pennsylvania 18015 USA

INTRODUCTION
Highway bridges may be designed in accordance with the allowable stress, load factor, or load and resistance factor design
approach. When these bridges are loaded with the regular design
vehicles the response of the structure will be linearly elastic, or
very close to it. The collapse load levels of these bridges can be
computed by anyone of the accepted methods available in the literature. However, the overloading of the bridge may propel its structural response to a level that is above the proportional limit but
far below the collapse load. In the bridge engineering literature
there exists a major gap in the prediction of the structural response of the bridge at post-proportional limit, and also the determination of the extent of damage to the superstructure. This paper
provides an overview of the inelastic response of highway bridges
composed of steel girders, rolled, built-up or plate girders, and
reinforced concrete deck slab.
The various research programs that are summarized herein have
all employed the finite element displacement method, Derivations
and the development of the mathematical models have been extremely
laborious. Presentation of any formulation would not have increased the engineering contributions of this paper. It is recommended that readers interested in the mathematical derivations of
the employed models, and the verification of these models, should
refer to the references cited. The paper is aimed at bridge
engineers.
1.1

The Overloading of Bridges

50

The extensive ~esearch carried out on the overloading behavior

of simple span pres~ressed concrete I-beam bridges, which were designed in accordance with the AASHTO Standard Specifications for
Highway Bridges (9), indicated that prior to the occurrence of any
distress in the beams the reinforced concrete deck slab will undergo substantial damage in the form of cracking of the concrete for
increased overload levels (5,7). These cracks occur approximately
at the mid-spacing between the beams at the bottom of the slab,
that is positive bending of the deck. The cracks are essentially
parallel to the girders. Regardless of the extent of the damage
to the deck the prestressed concrete I-beams did not have any distress up until extremely high load levels, at which level the deck
slab was essentially "destroyed."
Bridge engineers speculated that since both steel girder and
prestressed concrete I-beam bridges are designed under the same
specifications (9), which reflects the same structural response
expectations, similar structural response should be expected from
steel multigirder highway bridges. The research have showed that
there exists substantial differences between the response characteristics of steel and prestressed concrete bridges at load levels
that are above the proportional limit (2,3,7).
2

THE METHOD OF ANALYSIS

The "routine" methods of analysis of beam-slab highway bridgesthe structures which inherently have a very high degree of structural indeterminancy-employ the oversimplified "distribution factor"
approach (9,10). The studies have indicated that even for the
elastic regime of the structure 'this approach is far from being
satisfactory (2,7). Thus, a more detailed scheme needs to be
employed which will account for the structural indeterminancy, as
well as the gradual initiation and spread of material nonlinearities.
The finite element modeling with plate bending elements, with membrane stiffness properties, and the eccentric beam elements would
provide a realistic approach to the elastic simulation of the superstructure. To account for the material nonlinearities, it is essential that both the beam and the plate bending elements be subdivided
into layers, through the depth of the structural member (Fig. 1).
Throughout the analysis with the gradually increased load levels,
using the tangent stiffness formulation, stress and strain, as well
as the damage, history of each layer of each element need to be
monitored. A complicated analysis scheme as such requires the
development of a detailed software package. Program BOVAS (Bridge
Overload Analysis-Steel) has been developed to analyze simple or
continuous steel multigirder highway bridges (4). A number of
benchmark studies have been conducted to verify the accuracy of the
program (2,3,4). It has been found that the program can predict
the full load versus deformation (or damage) history of a given

51

bridge for a given loading condition, up to the collapse of the

bridge superstructure, if an acceptable finite element mesh is
employed. On the average the maximum discrepancy noted between the
analytical simulation via program BOVAS versus the available full
scale or the model studies was found to be about 5%. Thus, the
observations and conclusions presented herein have a high degree of
reliability as compared to any other method of "approximation" of
the overload and inelastic response of bridge superstructures.

NONLINEARITIES ASSOCIATED WITH OVERLOADING

In the simulation of the post-linear elastic response of steel

multigirder bridges the following possible sources of nonlinearities
need to be initially considered: nonlinearity of stress-strain curve
for concrete and steel, cracking and crushing of the concrete, yielding of steel, local and global stability of the steel girders and
their components, post-buckling behavior of steel components, and
slip between the deck and girders.
In addition, the question regarding the small versus large deflection formulation needs to be
looked into. The research have illustrated that formulation of the
problem on undeformed geometry does not cause any inherent inaccuracies in the analysis scheme (3).
3.1

Interface Slip

The slip between the slab and the girders could theoretically
be a major problem, especially at increased overload levels. This
problem could be further compounded due to the recent trend in the

8y

---~
~
z

'w
Fig. 1

Slab and Girder Layering

............ 8

~.

52
increased use of "partial shear connectors."
Detailed analytical research, and its correlation with the
available test data, have demonstrated that for fully composite
construction the slip is not a progressive phenomenon for increased
overload levels (8).
In the case of the use of partial shear connectors slip does not occur under service loads, not even at load
levels in excess of service loads. For higher load levels, a slip
of a finite amount can take place, which causes a slight "jump" in
the load-deformation curve, corresponding to a slight incremental
loss of stiffness (8). The load deformation curve continues to
climb, with the previous slope, without any further disturbance,
as far as the slip contribution is concerned, until the spread of
extensive damage to the superstructure. Additional finite slip can
also take place. It has been noted that for fully composite or
partial shear connection cases the neglect of the slip will not
result in any major inaccuracy in the prediction of the post-linear
elastic response of the bridge superstructure. The result of this
simplification will be the slight overestimation of the stiffness
of the bridge.
In the case of fully noncomposite construction it is noted that
there still exists a substantial initial composite interaction
between the deck and the girders at design load levels.
For
increased overload levels there will be intermittent slip of undefinable magnitude. The slip, in this case, again is not progressive. For increased live load levels the frictional forces
between the deck and the girders increase, thus arresting the
slip in part.
In any engineering analysis and research the contribution of the slip should preferably not be overemphasized. The
complications arising from its inclusion do not warrant the increase
in accuracy.
3.2

Material Nonlinearities

The research have demonstrated that the inclusion of material

nonlinearities is essential in the correct assessment of the inelastic response of steel multigirder highway bridges (2,3,7). This
is especially the case if the study requires the simulation of the
full-range response, i.e. from zero load level up to the collapse
of the bridge.
If an investigation is aimed at the determination of the rated
capacity of a given bridge, and if serviceability criteria will
have to be considered, as has been the case with the common practice in the USA (10), then the allowable stress levels for the
girder steel and for the reinforcing bars are far below the levels
at which material nonlinearities will be pronounced.
In these types
of analysis it is noted that serious considerations that might be
given to the representation of the nonlinearity of the steel stress-

53

strain curve may not be critical (3). However, the nonlinearity of

the concrete stress-strain curve can not be neglected, especially
if the rating provisions do not specify the stress levels in concrete (10).
The cracking of the concrete must be simulated as realistically
as possible (7). Even though the serviceability considerations
require the determination of the crack width in reinforced concrete
structures, the experience of the researchers with the use of any
given crack width formula for any given structure-carefully monitored throughout the loading-has not been encouraging (7). Due to
various factors, which are beyond the control of the bridge engineer,
the crack width formulae do not give a reliable estimate for an
existing structure, for which very little may be known as far as
the exact placement of the reinforcing bars. Instead, it was found
that the use of the crack depth, which can be reliably predicted,
is a better measure of the serviceability 1imit in the definition
of the allowable load levels in the overload analysis (5). The
critical parameter, especially in the negative moment region of
continuous bridges, where the deck cracking will initiate at the
top of the slab, is the determination of the crack depth for infrequent overloadings. Due to this infrequency, the cracks will be
"working cracks," such that after the traverse of the vehicle the
cracks will close due to the rebound of the bridge superstructure (6).
3.3

Stability Considerations

The research have indicated that for rolled or built up sections the web buckling, with the exception of at the supports,
could easily be retarded by vertical stiffeners.
In the case of
plate girder construction, especially those with deep and slender
webs, the web buckling takes place almost at the same load level
that some relatively deep cracks, for example, one-fourth the slab
thickness, form in the deck. Regardless of the negative connotation of the web buckling, this is a local constrained phenomenon.
After the web panel buckling, the "truss action" takes over (3).
The post buckling strength of the web is usually more than sufficient
to permit the traverse of infrequent overloads (2,3).
In the case of continuous construction lateral torsional buckling is a possibility in the negative moment region. However, even
though this buckling mode is theoretically possible, due to the
contribution of the wind bracing, and X-bracing or diaphragms, the
lateral torsional buckling becomes an unlikely phenomenon (2).
3.4

Serviceability Limits

The main thrust of the research summarized herein is the

ability to predict the post-linear elastic behavior of the bridge.
Such a response is not uncommon for many special permit vehicles.

54
The key issue for the bridge engineer is to define the extent of
permissible "recoverable damage" to the bridge. In the US bridge
engineering practice, i.e. AASHTO Bridge Maintenance Manual, the
allowable stresses for the steel girders have been set low enough
to prevent the damage to the steel superstructure. It should be
noted that these low stresses are set forth to account for the
adverse effects of the residual stresses and the stress concentrtion at the connection details. The current serviceability limits
require extensive in depth studies in order to relate the possible
post-linear elastic behavior of the bridge components to the bridge
life expectancy as well as the permissible load levels.

EXAMPLE BRIDGE

To illustrate the capabilities of the research summarized

herein one example is presented. A four span continuous bridge with
span lengths 21.34 m + 27.43 m + 27.43 m + 21.34 m (70 ft. + 90 ft.
+ 90 ft. + 70 ft.) was field tested by the researchers at the
University of Tennessee. The cross section of the bridge is shown
in Fig. 2. Figure 3 shows the load-deflection curve for the field
test and the finite element analysis.
It can be seen that there
exists an excellent agreement between the actual behavior and the
analytical prediction at elastic as well as at the transition into
highly nonlinear "knee," and at the ultimate load level. The
analysis indicated that slab cracking initiates at 1154 kN (259.5
kip) load, which is far in excess of the service loading. At
2475 kN (556.4 kips) first yielding begins in the bottom flange of
the girder immediately under the loading. Careful inspection of
the load-deflection curve indicates that for these load levels the
bridge response is almost linear. Thus the initiation and spread
of the damage have no perceptible effects on the overall structural
response of the bridge. Additional observations on the behavior of

~W36lt170
(Typ)

Ilj48m.m\ .2540mm
(57')
(100")

48m,!!' I
I .2540mm.
I .2540mm.
I11(57')
000")
(100")

Fig. 2

Cross Section

55
DEFLECTION AT SECTION A-A (In.)

10

20

r---------~r---------~r-~====~~!;:;::~~::l===~1200
5000

4000

-;;;

Co

0
<I

-'

800 :;;;
0

f-

-'

o Unlv. Tenn. Tell

-'

<I

af-

3000
2000

o PrOf>lrom BOVAS

-'

f-

400 a

f-

o
DEFLECTION AT SECTION A-A (mm)

Fig. 3

Load versus Deflection Curve

this bridge testing and analysis can be found in reference 2.

The load levels given above are above the practical limits.
If the analysis includes the residual stresses (results of which
will not be presented herein), there may be a 20-30% decrease in
the load levels that will cause yielding in the girders.
It should
be recalled that there are substantial stress concentrations in
the steel connection details, causing grave concern for the initiation of the fatigue crack growth.

GENERAL OBSERVATIONS

Extensive observations can be listed with applicability to

certain types of construction, and also dependent on the type of
loading (6)_ However, it is possible to arrive at general observations and conclusions that are applicable to all steel multigirder bridges, regardless of loading and the construction practice:
(1)

(2)

In continuous beam-slab bridge superstructures the

first failure is the cracking of the slab concrete
in the negative moment region.
Regardless of the type of loading, the negative
moment regions suffer the greatest amount of damage_

56

(3)

(4)

(5)

1.

2.

3.

4.

5.

6.

7.

8.

9.
10.

The effects of cross-bracing in more uniform lateral

live load distribution are negligible.
In deep plate girders the web panel buckling frequently
occurs at load levels higher than the design loads.
This buckling does not lead to any progressive damage.
Exclusion of the web panel buckling in any analytical
simulation can lead to the incorrect assessment of
the stresses in the superstructure.
In cover-plated girders, the plastification of the
flange initiates near the tip of the cover-plate,
a most undesirable location in view of fatigue
considerations.

Fisher, T.A., J.W. Fisher, C.N. Kostem and D.R. Mertz. Design
and Retrofit for Fatigue Damage in Web Gap, Fatigue of Steel
and Concrete Structures, IABSE Colloquium Proceedings (Lausanne,
Switzerland, 1982) 535-543.
Hall, J.C. and C.N. Kostem.
Inelastic Analysis of Steel Multigirder Highway Bridges, Fritz Engineering Laboratory Report
No. 435.1 (Lehigh University, Bethlehem, Pennsylvania, USA,
1980).
Hall, J.C. and C.N. Kostem.
Inelastic Overload Analysis of
Continuous Steel Multigirder Highway Bridges by the Finite
Element Method," Fritz Engineering Laboratory Report 432.6
(Lehigh University, Bethlehem, Pennsylvania, USA, 1981).
Heishman, C.A. and C.N. Kostem. Users Manual for Program BOVAS Detailed Version, Fritz Engineering Laboratory Report No. 435.2
(Lehigh University, Bethlehem, Pennsylvania, USA, 1982).
Kostem, C.N. Further Parametric Studies on the Overloading of
Highway Bridges, Fritz Engineering Laboratory Report No. 434.2
(Lehigh University, Bethlehem, Pennsylvania, USA, 1980).
Kostem, C.N. A Parametric Study on the Overloading of Steel
Multigirder Bridges, Fritz Engineering Laboratory Report No.
435.3 (Lehigh University, Bethlehem, Pennsylvania, USA, 1983).
Peterson, W.S. and C.N. Kostem. The Inelastic Analysis of
Beam-Slab Highway Bridge Superstructures, Fritz Engineering
Laboratory Report No. 378B.5 (Lehigh University, Bethlehem,
Pennsylvania, USA, 1975).
Tumminelli, S.C. and C.N. Kostem. Finite Elements for the
Elastic Analysis of Composite Beams and Bridges, Fritz
Engineering Laboratory Report No. 432.3 (Lehigh University,
Bethlehem, Pennsylvania, USA, 1978).
Standard Specifications for Highway Bridges. The American
Association of State Highway and Transportation Officials,
Washington, D.C., USA, 1977.
Manual for Maintenance Inspection of Bridges. The American
Association of State Highway and Transportation Officials,
Washington, D.C., USA, 1978.

57

STEEL CURVED I AND BOX GIRDER DESIGN AND ANALYSIS

C. P. Heins, Professor
Institute for Physical Science and Technology and
Civil Engineering Department
University of Maryland, College Park, MD 20742
INTRODUCTION
During the past ten years extensive research studies have been
undertaken in the area of the design and analysis of curved bridge
structures.

Such studies were necessitated by increased demands for

curved highway alignment and aesthetic considerations.

It is therefore

the purpose of this paper to present the results of some of these

studies and the resulting design specifications.

ANALYSIS TECHNIQUES
The determination of the induced forces in a horizontally curved
bridge system requires consideration of the interaction of the entire
bridge structure.

The influences of torsion (St. Venant and warping)

must be included in such analysis.

As indicated in the AASHTO code (2), the effects of curvature

58

may be neglected in evaluating the bending moments in the longitudinal

members, according to the following;
Number of Girders

Angle for 1 span

Angle for 2 or more SEar

20

30

3,4

30

40

40

50

or more

However, the influence of such curvature on the induced torsional

forces should be examined.

As a guide for such examination the following

has been found (26);

For

2 0.4

evaluation of stresses due to pure torsion may

be neglected, but warping stresses must be included.

For

~ ~

10.

evaluation of stresses due to warping may

be neglected.

where:

L[G~/EIW]1/2

and is plotted in Fig. 1

Analysis techniques. - There are various analytical methods

available for determining the response of curved bridges
In general these methods can be categorized as follows;
1.

Static Method-U.S. Steel

2.

Computer Matrix Grid Method

a)

Three DOF

b)

Five DOF

3.

Space Frame Grid Method

4.

Finite Element Method

(29)

(23)
(23)

(23,24,28)

(23,24,28~33).

59

5.

Finite Differences Method

(8,9,10,15)

The first method "Static Method-U.S. Steel" (27,29) allm~s for

direct dtermination of the forces in the longitudinal girders and
diaphragms by making certain assumptions relative to load or force
distribution, and generally does not require a computer.

The method

can be explained by examining a free body of a curved flange, Fig.

2 and 3.

Noting that the resultant forces (F) in the flange are not

in equilibrium in the vertical direction, therefore an artifical

force q is developed, and equals;
M
R

1
h

(1)

where; q = horizontal load, along each flange, required to balance

the flange forces

The effective concentrated torque T at a given diaphragm, can now

be computed using this q force; which gives;

however:

H h

therefore;

T
T

therefore;
where:

!i.

R
H
R

and

(2)

distance between diaphragms

The total torque applied to the entire bridge, shown in Fig. 4,

is therefore;

60

EM

(3)

The resistance to this total torque is in the form of shears

V, which will be assumed linear with respect to the CG of the girders;
therefore;
V

3D

11 6

v1 - 1.3

and

The shear on the outside girders can now be determined from

equilibrium;

3D

VI

x 2

"6

VI - D + VI
VI

_ 10 D
9

VI

E~

V2

Vl/3

!:! ~
R

D
~
"9= E !:!
R

- 1-11I D

(4)
(5)

forces can be written as;

E M~

where

= E !:! ~
R

R ~

In general the

D
+ V2 - "6
- E

RD

(6)

is given as follows for the various number of girders;

Girders

1.0

1.0

1.11

5
1.25

6
1.40

1.56

10

1.72

1.88

2.04

61

The shear forces

as additional loads.

can now be applied to the individual girders

The diaphragms are designed for the shear

forces and end moments, shown in Fig. 4.

The second method to be utilized consists of a computer oriented
matrix method, which may consist of three degrees of freedom (28)
or five degrees of freedom (24).
effects, but excluded warping.

The 3 DOF grid, accounts- for torsional

The 5DOF grid, contains the warping

in the top and bottom flanges, thus increasing the

degreees of

freedom by two.
The space frame grid method, consists of 6 DOF, and considers
the girder flanges modeled as beams and the webs as a series of
cross and vertical elements.

The modeling of the girders in this manner

include warping influences by summing the effects of axial loads and

vertical bending moments on the flanges.
The finite element method, idealizes the bridge as a series of intersect
ing plates with specified boundary conditions.

Although, the method,

in

general gives an excellent representation of the bridge, excessive data and

computer time is required.
The finite difference procedure permits direct solution of the different
curved girder equation.

This equation, as developed by Vlasov (5), represent

the load-deflection response of curved bearns, including warping influences.

The details of this technique will now be given.

The load deformation

response of a single curved girder is given by the following expressions:

1ItI

"

EI

''''

w <P

EI
"
+~<P
R

(7)

62

EI

-~Tl"" + (
R

EIX+G~

) n" - EI

9'"

EI
- ~ + m
R2
z

+ GK "
T

(8)

where:

Vertical Deflection

Transverse Rotations

E1x

Bending Stiffeners

G~

Torsional Stiffeners

E1w

Warping Stiffness

= Radius

Equations (7) and (8) can be written in central difference form,

resulting in the following Equations (9) and (10);
EI
~ + EI
R2
x
EI

r2GK~
T

[~'

2t:,2G~

("+EI )+
. [ EI} .2( El +0"'1 ~'7 2.2(EI2+~l

4-f-H1 x

4\R

it-2

n-l

EI

~ + EI

R2

EI

w
R

n+2

R2

R +

r~Iw

-[,t:t.

-R2

n+l

n
=~t:,

} .2~J
2(EI; .o",l]

- 4 -+EI
R2

(9a)

63

El

w
R

-[4~~l + .2(EIX:eR,16~~~ +

w
n - 2

114El
x
6(El) + 2112GK.r + _ _
R2
n

-[4 (E1w) + 112G~

El

2112(E1w + GK.r)

n - 1

-f~~~ + "(EYeR,
-[4( E1W) + 112GK.rJ
n + 1

E1

w
R

m 114

(lOa)

El
w
n + 2

or
[Kl]{n} + [K 2 ]{<P}

114
qy

(9b)

[K3 ]{n} + [K4]{}

m 114

(lOb)

where the

node~points

are prescribed as shown in Fig. 5.

If the bridge now consists of a series of such curved girders, attached

by a series of diaphragms, as shown in Fig. 6, then Eqs. (9) and (10) can be
appropriately modified.

Consider first the response of a single diaghragm

#1, where the end actions developed are given by;

~a

a
na

QDa

~n
QDn
and

[K]

is a

diaphragms.

4x4

[K5 ]

(11)

1h

matrix and can represent either beam or truss type

64

Similarly the end actions developed in diaphragm

~n

4>n
nn

QDn

~b

[K 6 ]

(12)
4>b

QDb
The total diaphragm

#2, are given by:

4>b

induced forces acting at point

"n"

can be

obtained by adding Eqs. (11) and (12), giving;

4>a
(13)

~J [::J
=

where:

(14)
(15)
and

K ..
~J

= force

at points due to a deflection at point j, or points 2 and

4 represent vertical deflections and shears and points 1 and 3 represent

rotations and moments.
The interaction of the curved girders and diaphragms can be Obtained
by adding Eq. (13) to Eqs. (9b) and (lOb).

It should be noted that Eqs. (9b)

and (lOb) represent the general response of curved girder #1, subjected to a
uniform loads qy and mz '

At node

"n"

an additional load per length will

65

be imposed at the diaphragm and is equal to

Mnn/n and

QDn/n.

Therefore

the final equations are;

(16)

(17)

where:

[K 7 ]

[K3 ] + [0, 0, K33 + Kll , 0, 0]

(K8 ]

[K 4 ] + [0. O. K34 + K12 , 0, 0]

[K9 ]

[K3l , K13 ]

(K lO ]

[K32 , K14 ]

(Kn]

[Kl ] + [0, 0, K43 + K21 , 0, 0]

(K 12 ]

(K 2 ] + (0, 0, K44 + K22 , 0, 0]

[K13 ]

[K4l , K23 ]

[K14 ]

(K42 , K24 ]

{Ij>}

{n}

G:J

,
=

[~: J

If the node at which Eqs. (16) and (17) are being applied does not
correspond to a diaphragm connection, then the

K ..

1J

terms are set equal

to zero.
The above referenced equations have been incorporated into a computer
program (31) and utilized in developing design aids, as will now be described.

66
DESIGN AIDS

As mentioned previously, the analysis of curved girder systems

requires the girder section properties

Ix'~'

and Iw.

The designer,

therefore, is required to estimate the size of the girders in order to

compute these stiffnesses.

This size estimation can be achieved if the

primary internal girder forces, bending moment

and bimoment

which induce normal stresses, were known with some degree of confidence.
Then the basic bending equation,

x
I

(18)

and warping equation,

BiW
o

Bi

(I

(19)

Iw )

w n

could be applied by assuming a proportion of the design stress for

and

and computing the required (I

Ie)

0b

and (I Iw ) properties.
w n

Estimation of these stresses and/or forces has been obtained by performing

a thorough system analysis of single and multispan curved bridges (31,32).
The resulting force equations and graphs, in addition to impact factors
and girder property equations, for I-girder bridges, will now be presented.
Impact Factors.

The dynamic response of single span curved

67

girder bridges subjected to a sprung mass vehicle has been predicted

by a Fourier Series and Lump Mass Techniques (35).

These techniques

were then applied in determining the response of typical highway

bridges, and thus, the evaluation of impact factors.

The impact factor

for the bending, Fig. 7, and Fig. 8 lists the impact factor for the
torsional effects.

The curves are listed for various girder span

lengths and central angles

e=

(L/R).

The curves are for a vehicle

speed of V = 60 mph; other curves are available for velocities of 20

and 40 mph (35).
The impact factors are applied as in conventional pactice;
M

Bi

MSTATIC(l + IF)

(20)

BiSTATIC(l + IT)

(21)

where
M
Bi

Induced Bending Moment

Induced Bimoment

Live Load Design. -

The design of any bridge element requires the

establishment of forces in that element.

If the bridge has a straight

alignment, the forces can be determined readily through use of distribution

factors and simple beam and torsional forces creates an indeterminate
situation.

The following equations were, therefore, developed (34),

utilizing relationship between single curved responses (33) and the

system (30,32), to permit evaluation of the live load forces developed

68

in a curved composite bridge I-girder system.

These forces can then

be used to establish the induced stresses and thus, proper girder section.
Amplification Factor - K1

All the internal forces and deformations for a single curved and
straight girder have been evaluated using the various computer programs
(33,36).

The ratio of the reactions for these two girders gives the

following,

f(S.C.G.)
f(S.S.G.)

(22)

This factor describes the immediate effect of curvature relative

to a straight member.

The graphical representation of these data and

its analysis gives the following general equations (23) and (24),

Kmoment
K
-bimoment=

0.15(L/R) + 1

(23)

[(35n) (L/R) 2 - 15(L/R)] x 103

(24)

where
n

R/100,

R > 100'

Radius - Ft.

Span Length - Ft.

Distribution Factor - K2
The evaluation of the true distribution of load to each girder,
and this realistic values of internal forces, can be considered by
analyzing the curved girder as a system.

The number of tracks used in

69

the analysis would be dependent on the number of lanes.

The ratio of

these resulting maximum forces to those in a single curved girder gives

f(Sy.C.G.)
f(S.C.G.)
A plot of this ratio vs.

Rand

general equations (26) and (27).

Rand L

(25)

L will yield the following

In all instances, the parameters

refer to the midlength and corresponding radius of the system.

These equations are valid for girder systems which contain four, six,
and eight girders spacings of 7', 8', 9', and 10'.

Kmoment

(n

+ 3) (OR4L) + 0.6

K
- .11(R/L)
-bimoment- M

L < 70'

K
- (M - l)(R/L)
-bimoment6

L > 70'

(26)

(27a)
(27b)

where
R > 100'

R/lOO,

L/50,

Radius - Ft.

Span Length - Ft.

L > 50'

Reduction Factor K3
Because many bridge structures are continuous, it is desirable to
obtain some factors which can be applied to the simple span data to give

70

preliminary forces in continuous spans.

This factor can be written as

f(Sy.C.G.)N
f(Sy.C.G.)

(28)

where
N

Number of spans (2 or 3)

Utilizing a computer program (36). the maximum forces in a two

or three span curved bridge system of four. six. and eight girders
were evaluated under various critical loadings.
and the resulting
Table 1.

K3

A study of all the data

values give the following values listed in

The data is described relative to number of spans and is

independent of number of girders.

It should be emphasized that the two

and three span girder systems must all contain equal span lengths with
a maximum given span length of 100.0'.

For example. for a three span

system. the total maximum bridge length would then be 3 xL

= 300.0'.

Evaluation of Girder Forces and Deformation

With the various factor equations available. it is now possible
to evaluate preliminary forces in a curved girder bridge. relative
to the forces in a striaght girder.
1.

The procedure is as follows:

Evaluate maximum function "F" for a single straight girder of

length L subjected to a line of AASHO wheel loads.
function would be Fbending.

This

The function for Fbimoment is

assumed to be equal to one.

2.

Evaluate amplification factors

equations (23) and (24)

for the midspan length L and radius R of the bridge system.

71

3.

Evaluate distribution factors

K2

equations (26) and (27)

for the given midspan length L, number of girders in system,

and radius R.

4.

Select reduction factor

K3

from Table 1 if system is continuous

span.
5.

Determine maximum function F of curved girder system, i.e.

Max. Moment Static

(2

and
Max. Bimoment Static
6.

(30)

Multiply Max. Moment and Max. Bimoment by (1 + I) Equations (20)

and (21), respectively.

In order to account for girder spacing S, the resulting action of single
straight girser subjected to a line of truck wheels should be multiplied
by a Factor

S,

where

1.29

for

7' or 8'

1.57

for

9' or 10'

Also, all of the parameters

Rand

are referenced to the center-

line of the bridge system.

Dead Load Design.

The dead load response of curved bridge

systems has been predicted by Murphy (31), which more realistically

represents the response of the bare steel frame system.

This technique

has been computerized and then applied in the development of design

72

curves.

The curves represent the maximum bending stress (Ob)

warping stress
S.

(0 )
w

and

as a function of R, L, I , and diaphragm spacing

x

The curves were based on typical girder properties obtained from

a survey of bridge design as a function of girder length, as given in

Figure 15.

If the actual design properties are different from those

referenced values given in Figure 15, a correction factor is required

to modify the chart values.
It has been found that for bending stress,
(31)

where
0b

Actual or Corrected Bending Stress

0b

Bending Stress form Referenced Properties

Cl

sIs

Referenced Section Modulus

Design Section Modulus

The warping stress is related by the following,

ow
a

ow

(32)

Actual or Corrected Warping Stress

Warping Stress from Referenced Properties
(W /1 ){W /1 0
n wnw

Referenced Value
Design Value
The design charts, Figures 9 through 14, list the induced stresses
and

normalized relative to the applied dead load (W) per

73

length (K/in.).

Thus, any variation in dead load may be considered

The normalized stress vs. the (R/L) values are

in an actual design.

plotted for span lengths of L

are dependent on stiffness

= 100',

125', and 150'.

These plots

(I ) , number of girders, and diaphragm

x

spacing.
The plots of

vs.

(R/L) for L

100',125', and 150' (Figures

4,6,8) are independent of diaphragm spacing and

in the development of the curves.
The plots of

vs.

(R/L)

(I ), as determined
x

(~).

are dependent on diaphragm spacing

and stiffness (I ), as given in Figures 10, 12, and 14.

x

These curves

are limited to a four-girder system and are for specified stiffnesses,

however, for other stiffnessed, the following equation may be applied,
I

x -rx
I

(33)

where

Referenced Modified Factor

I
I

rx

Actual Stiffness
Reference Chart Value given in Figures 10, 12, and 1.

Warping Stress Adjustment

The curves that are presented herein are the results relative to
the four-girder system.
and .",ht-girder

.yot_.

In order_to account for the behavior of six

the (

": )

chart value _ t b.

These chart values are modified by the following equation,

~difi.d.

74

()
a

/j

chart

(- )
cr

(34)

where

"C':)=

(": y

(35)

Change in the value of the chart due to an

increase in the number of girders (six
or eight) beyond the referenced four girders,
as given in Figure 10

I
I

w
w

Reference Property given in Figure 16

Design Property
Design Diaphragm Spacing

Girder Spacing
The curves have been developed for girder spacing of 8.0', however
spacings up to 10' are acceptable.
Torsional Properties.

In order to determine design stresses

or distortions, the girder section property is required.

If the girder

is subjected to torsion and bending, the torsional properties, in

addition to bending properties, will be required.

The exact solution

of the torsional properties of composite sections has been demonstrated

(37).

However, by idealizing the composite section, a series of simplified

equations can be developed and used for design (11).

Figure 17 describes a typical composite girder and pertinent
dimensions.

Neglecting the top girder flange and modifying the concrete

75
slab thickness, as shown in Figure 12, with the dimensions defined
as follows:

G /G.
c

Using these dimensions, the

resulting torsional properties are determined as follows,

Shear Center
(36)

Normalized Warping Functions

Slab:

Beam:

W
n

(37)

c
(38)

Warping Stiffness
I

(39)

Torsional Constant
1

(40)

"3
where
m

G /G

With the evaluation of these torsional parameters, the resulting

normal stresses in the composite section due ti bimoment can be evaluated
as follows:

76

Warping Nominal Stress

Slab:

Steel:

BiW
n
(J

(41)

BiW
n

___
5

(J

(42)

DESIGN SPECIFICATIONS
The AASHTO design criteria (1,2), as applied to steel I and Box
girder bridges, is given in Table 2.
straight and curved elements.

These criteria are listed for both

77

~------------------------0

100

Mono bo.

"2:30

0
\()

30

.,-:

1.0

-----------------------------~

Multiple I

ITT

'.0

It-O.S-3

~~

D5~

_J ___ l ___ l __ L __L_:.l __ -L __ L __ l

o

0.1

0.2

0.1

0.'

O~

Cent-,I angle.'

0.'

OJ

0.1

0.9

Figure 1

78

Fig. 2

lq , lDl
~

0/3
Ie
.I'"

1\

0/6

,.'t:

0/6

I
.,

1\

0/3

Fig. 4

n-l

n+1

~2

,/ \.-I

Fig. 5

~Ol

_l
~
n-

nH

n+2

~""O<"
Fig. 6

79

.t
I

(.tIl UOJ.lU .Llv.r.!I 'IVliOIS'JOJ.

80

".
200.

I"'.

I!I C.S.

JO.

o.
I.

..

1.

zo.

10.

la.

Figure 9
IOt1OUZI!D IWU'IHl mESS

L-lOOO'
12,000. ~It

Ia -

"".
400.

lOO

200

1/L-2,S

100
IlL -

---.--24

-r-I-22.

20.

18

16

l~

.
I

figure 10

81

mas

lQIW.IZID IEHJDC

L .. I25.D'

lID.

UO.

200

uo.

1\;"10

.ao.

...

..

z.

1.

...

zo.
Figure 11

.................... mas
L-WO'

x.." 10,000

m4

5<10

...
lID

IlL" lot

'00.

lO'

'4.

22

20.

II

Ii.

14.

DW'lwat srlCIlG S - rr.

Figure 12

82

1GIIIiIl.mq,~_

L-UDO'

~.U5.OIIl ~Io

F~f:ute

14

Y[SE<:T1O.

IC)DIJUIS

.,.. sp_

...-l

83

m)

La:ilth

l.

2.

1.

10.

50

no.

90.

130.

ISO

10

...

...
I~

.or.

...
figure 15
100.

60
60.
t,..Spm (fcJ

100.

JOQ

250.

x..' ",000

100

..4

1. ... 125

Z50,

lO.

200

;...

lb~

O.

1lO.

.. l2,ooo

~!1..50.

lO.

O.

5.

10.

--60l<don

--IGUdc.

.0

. 0

..'

't

..

OIL

5.

10.

I
I

In'

L .. lSQ'
\

"

I}

... .' III "

f/--ol
bt

x..' W."'"

100
SO.

O.

l3O.

L 100'

;...

100.

140.

;... 200.

I~P~.
JOQ.

1Jll.

OIL
1

Figure 16

..

.,
Figure 17

Three
Span

Span

Two
.70

.70

.70

.60

.75

.65

.35

.35

.45
.40

.75

.75

.90

1.00

Number
of
~ending Moment ~eflection ~otation KSt . Venant ~arping Torsion ~imoment KShear
Spans

K3 - REDUCTION FACTOR FOR MAXIMUM FUNCTION IN TWO AND THREE SPAN BRIDGES

TABLE 1

-1:0-

00

moment)

(pos! ti ve

it

Flange

Compression

Item

< 3250 < 24

t~~~

Straight

TABLE 2

Working Stress Design

< 4400

--to
(b)

1.

0.6 ( r )
b

[30+8000(0.1-R )
fw

and Pw

Pw1 or Pw2' where;

(+) use smaller Pw1 or Pw2

0.95

-(-r) [1"""]51
b

--f

fw (_) use Pw1

fb

/
b

Pw2 =

iff

or

Pw1 =

+ (!:.)(!:.)
R b

(!,.) 2F
0.55 F [1- ~1p P
Y
41T2E
B w

PB

where

Fb

and

t~~

Curved

VI

00

bJ

moment)

Compression
Flange
(negative

Item

I fy

6140

k
Y

< 60 or 13,300

< 60

t~

fb :'. 57.6 x 106

IF

13300 <

(~)2

fb :'. 0.55Fy
_ .
0 224F [1 -sin _11
Y
2

~~t~

6140 <

fb ~ 0.55 Fy

Straight

TABLE 2

- b/ t IF
( 1330071Ln
y
)

=
0.55 F

IF
y
y

6140 . X

I
f

< 13,300 or 60

Y 0.224F { sin
0.326Fy +
y

t~1F

if
t

IF

13300

11

"2

t 2 x 106 _

57 6(.!.)Z1066
. b

Fb - 57.6(1;)

F
b

k~JJ

113.4(-~)2

f 2

x 106

13,300 - t
(13,300 6140X )

Fb is smaller of the following;

k~

6 = 11_9.0(;V)2
y

where r---~-

FbY-

IF

6140 <

IFy

1-9.2[/1
y

Curved

fv
where X = 1 +4/3(- - 0.15) ~1

Fb

k:'.

Working Stress Design (Continued)

0-

00

~:~

10

Stiffener

with

moment)

Compression
Flange
(negative

Item

iK

~ 0.S5F y -

t~

>

B=

~t3w

0.125K
for n=l

where
3 4
$ ={O.07K: for 0>1

with longitudinal Stiffener

Stiffener requirement

fb :'; 14.4 x 106 K(;)2

IF
y

3580/K

~]
t
y

Xl

Xl

where

Fb

<

Curved

<

(n

1)2

5.34 + 2.84 (l/wt 3)1/3

K
fv
0.93 + (1.6 - ({)(y)

1-912(/)

1 (n>1)

0.5SF

IF

~ < 3070/K Xl

Working Stress Design (Continued)

0.224F y [1 - Bin!
2 66S0/K -

6650/K < ~ < 60

t ~

fb

IF

3070 /K < ~ < 60 or 66S0/K

fb ~ O.SSF y

IFy

~ ~ 3070

TABLE 2

Straight

5.34

1 (n=l)

--.J

00

hJ

with
stiffener

moment)

Compress ion
Flange
(negative

Item
Straight

TABLE 2

t~

O.l{~s)t~

or

Fb

14.4K(;)26 106

= 14.4K(.!.)2 x 10 6 -

Fb is smaller value of

IF"y

w
~

106

66S01i< X2 -30701i< Xl

66S01i< X
2 -

v K
2 t 2
14.4(K.) (~)

f 2

(f)

in

ylf i

S.34J

66S01i< X2
< ."': < 60

+ 0.224F

f 2
1-9.0(f)
y

3Z6F

rLO. y
I

IF

Curved

<."': < 66S01i< X or 60

X2 ~ 1 - 2.13(F~)
y

Where

'b

IF

30701i< X

Working Stl'eSB Design (Continued)

}] 6

00
00

Stiffener

With
Long1 tud Ina1
and Transverse

Item

Working Stress Design(Continued)

> 8t 3 w

sAf
a

spacing of transverse stiffener

~
E

Af : Area of flange including longitudinal stiffener

fs: maximum longitudinal bending stress

a:

where

I t "- o.10(nH)3w3

K = '[1 +(~)212 + 87.3

1
(1\+1) 2 (~) 2 [l+() .1(n H) I

Use same formula, but use Kl instead of K

Straight

TABLE 2
Curved

'"'

00

ll

Transverse
Sti ffener

With

L.lb= =

Stiffener

r:n:.

Without

Web

Item

< 5.625 x 10 7 ::. ~

v (d I t) 2
3

fv -

do

0.B7(1-C)

stiffener spacing

F (d /t )2

2.2 x lOB [1 +(dl do) 2 J

<J[C+~
2
3
( old)

d0 < 1. 5d

t~

d < 23,000. ::. 170

t -

~ < 150

Straight

TABLE 2

::'1.0

0.02

F (d /t )2
Y

+ (dId )2 J

0.B7(1-C) J
,d---:L
11 +\-o/d)

C = 2.2 x lOB [1

[C

1. 5d

Fb

- J

<

< 1.0

23000 {1.19 - 10(/.) + 34 (:O)2}::'170

fv - 3

%::.

Curved

/R'-:' 0.02 use straight Girder criteria

If d O/R

IIf

Same

Working Stress Design (Continued)

)g

with
Transverse
Stiffener

Web

Item

d
25 (--%)2 _ 20 > 5.0

rl t 3
> 0
- 10.92 J

Stiffener Criteria

Straight

TABLE 2

< 2600

!?

n;

d2
0.95flt

(/

~d

= 1.0+

~)2

Curved

0.78

- 0.78) 4
)
1775
Z

2.

- 20) X::' 5.0

d
1.0 for dO

[25(

t3 J

> d

Stiffener Criteria

Working Stress Design (Continued)

0.78~-f~1.0

92
REFERENCES
1.

"Standard Specifications for Highway Bridges,"

Washington, D.C., 1977.

12th Edition, AASHTO,

2.

"Guide Specifications for Horizontally Curved Highway Bridges,"

AASHTO, Washington, D.C., 1980.

3.

T.V. Galambos, "Tentative Load Factor Design Criteria for Curved

Steel Bridges," Rept. No. 50, Washington, Univ., St. Louis, MO,
May 1978.

4.

R. Wolchuck, R. Mayrbaurl, :Proposed Design Specifications for Steel

Box Girder Bridges," FHWA Rept. TS-80-205, January 1980.

5.

C.P. Heins, "Box Girder Design," AISC Engineering Journal, Vol. IS,
No.4, December 1978.

6.

C.P. Heins, L.J. Hua, "Proportioning Box Girder Bridges," ASCE St. Div.
Jour., Vol. 106, No. STll, November 1980.

7.

C.P. Heins, D.H. Hall, "Designers Guide to Steel Box Girder Bridges,"
Bethlehem Steel Corp., Bethlehe~, PA, 1981.

8.

C.P. Heins, J.C. Olenick, "Curved Box Beam Bridges Analysis,"

Computers and Structures Journal, Vol. 6, pp. 65-73, Pergamon Press,
London, 1976.

9.

C.P. Heins, F.H. Sheu, "Computer Analysis of Steel Box Girder

Bridges," Civil Engineering Dept. Report, Univ. Of Maryland, June 1981.

10.

C.P. Heins, J.C. Olenick, "Diaphragms for Curved Box Beam Bridges,"
ASCE St. Div. Jour., Vol. 101, No. STlO, October 1975.

11.

C.P. Heins, A, Sahin, "Natural Frequency of Curved Box Girder

Bridges," ASCE St. Dvi. Jour., Vol. lOS, No. ST12, December 1979.

12.

C.P. Heins, R. Humphrey, "Bending and Torsion Interaction of Box

Girders," ASCE St. Div. Jour., Vol. IDS, No. ST5, May 1979.

13.

C.P. Heins, et al., "Curved Steel Box Girder Bridges: A Survey,"

ASCE St. Div. Jour., Vol. 104, No. STIl, November 1978.

14.

C.P. Heins,et aI., "Curved Steel Box Girder Bridges: State of Art,"
ASCE St. Div. Jour., Vol. 104, No. STll, November 1978.

15.

C.P. Heins, W.H. Lee, "Curved Box Girder Field Test, " ASCE St.
Div. Jour. , Vol. 107, No. ST2, February 1981.

93

16.

C.H. Yoo, J. Buchanan, C.P. Heins, W.L. Armst:rong, "Loading Response

of a Continuous Box Girder Bridge," Proceeedings ASCE Specialty
Conference of Metal Bridges, St. Louis, MO, November 1974.

17.

C.P. Heins, J.Y. Shyu, ''Moment Capacity of Box Girders," Institute

for Physical Science and Technology Report, University of Maryland,
June 1981.

18.

Heins, C.P., Huo, C., "Truck and Wind Load Effects on Girder Slab
Bridge With/Without Wind Bracing," AISI Project 308, Effects of
Bracing on I Girder Bridges, Inst. for Phy. Sci. and Tech.,
Tech. Note BN-988-A, University of Maryland, College Park, MD,
April 1980.

19.

Heins, C.P., Wind Load Effects on Tangent and Skew I Girder

Bridges With/Without Lateral Wind Bracing," Addendum, AISI Project 308,
Inst. for Phys. Sci. and Tech., University of Maryland, College Park,
MD, July 1980.

20.

Heins, C.P., Jin, J., "Load Distribution of Braced Curved I Girder

Bridges," AISI Project 308, Effects of Bracing I Girder Bridges,
Inst. for Phys. Sci. and Tech., University of Maryland, College Park,
MD, June 1982.

21.

Heins, C.P., Kato, H., "Bottom Lateral Bracing Effects of Curved I

Girder Bridges," AISI Project 308, Effects on I Girder Bridges,
Inst. for Phys. Sci. and- Tech., University of Maryland, College Park,
MD, February 1981.

22.

Heins, C.P., Kuo, T.C., "Live Load Distribution of Simple Span Steel
I-Beam Composite Highway Bridge at Ultimate Load," Department of
Civil Engineering. University of Maryland, April 1973.

23.

Wilson, L., Bathe. K.J., Peterson, F.E., Dovey, H., "User's Guide
to Program SAP," University of California. Department of Civil
Engineering, November 1972.

24.

Schelling, D.R Heins. C.P., Sikes, G.H., "State of the Art of

Curved Girder Bridge Programs." Computer and Structures. Vol. 9,
pp. 27-37, Pergamon Press, London, 1978.

25.

Heins, C.P., Sheu, F.H., "Design/Analysis of Curved Box Girder

Bridges." Computers and Structures. Vol. 15, pp. 241-258,
Pergamon Press. London, 1978.

26.

Nakai. H., Heins. C.P., "Analysis Criteria for Curved Bridges,"

ASCE St. Div. Jour., Vol. 103, No. ST7, July 1977.

94

27.

Merrl.tt, F.S., "Structural Steel Designers Handbook," McGraw Hill

Book Co., New York, N.Y. , 1972.
----~~~~

28.

Lavelle, F.H.m Boick, L.S. "A Program to Analyze Curved Girder

Bridges", Engr. Bullentin No.8, Univ. of Rhode Island, Kingston,
Rhode Island, 1965

29.

United States Steel Corporation, Highway Structures Design Handbook,

Vol. I, 1967.

30.

Bell, L.C., Heins, C.P., "The Solution of Curved Bridge Systems

Using the Slope-Deflection Fourier Series Method", Civil Engineering
Rept. No. 30, Univ. of Maryland, College Park, MD, June 1968.

31.

Murphy, E.L., Heins, C.P., "Dead Load Analysis of Single Span

Curved Bridges", Civil Engineering Rept. No. 52, Univ. of Maryland,
College Park, MD, June 1973.

32.

Heins, C.P., "The Presentation of the Slope-Deflection Method for

the Analysis of Curved Orthotropic Highway Bridges", Civil Engineering
Rept. No. 15, June 1967.

33.

Spates, K.R., Heins, C.P., "The Analysis of Single Curved Girders

with Various Loadings and Boundary Condition", Civil Engineering
Rept. No. 20, June 1968.

34.

Siminou, J., Heins, C.P., "Proposed Equations for Preliminary Design

of Curved Girder Bridge Systems", Civil Engineering Rept. No. 25,
June 1969.

35.

Vashi, K.M., Heins, C.P., "Impact Factors for Curved Highway Bridges".
Civil Engineering Rept. No. 32, Spetember 1969.

36.

Bell, L.C., Heins, C.P., "Curved Girder Computer Manual", Civil

Engineering Rept. No. 30, September 1968.

37.

Kuo, J.T.C., Heins, C.P., "Torsional Properties of Composite Steel

Bridge Members", Civil Engineering Rept. No. 37, June 1970.

38.

Kuo, J.T.C., Heins, C.P., "Behavior of Composite Beams Subjected

to Torsion", Civil Engineering Rept. No. 39, February 1971.

39.

Heins, C.P., "Curved I Girder Design", Highway Research Record

No. 547, Highway Research Board, January 1975.

95

S]}1PLIFIED ME1HlDS FOR THE ANALYSIS AND DESIGN OF BRIDGES OF

CELLULAR CROSS-SECTION

H.R. Evans

Department of Civil and Structural Engineering, University College,

Cardiff.

SYNOPSIS
The naner describes a number of simplified methods that have
been developed for the analysis of cellular bridge decks; the
behaviour in the linear elastic and in the non-linear nost-buckling
range is considered. Each method assumes an idealized structural
behaviour to provide a simplified solution that may be carried out
on a microcomputer. The methods have been develoned specifically
for use during the nreliminary stages of design when reneated
analyses may be required to establish ontimum nroportions. The
paper includes some typical comparisons of the results given by
the proposed methods with the results of other established methods.
1

INTRODUCTION

In recent years there has been a great increase in the computing power that is available to the bridge designer and this has
made it ~ossible for powerful techniques, requiring extensive
computations, to be applied to the analysis of bridge structures.
Of major importance in this context is the finite element method (1)
which, by virtue of its adantability, enables a wide range of
bridge loading, geometry and sunnort conditions to be taken into
account in analysis.
Despite the increase in available computing Dower, the finite
element method must still be regarded as a rather exnensive method
of analysis. Certainly, at the initial design stage, when the
designer is endeavouring to ontimize the prODortions of his structure by repeated analysis, the costs involved in using the method
are unaccentable. Furthermore, since most modern design codes are
based on limit state considerations, it is essential to be able to

96

determine the ultimate loads of bridge components. In such analyses,

since due account must be taken of both feometric and material nonlineari ty, an incremental procedure must be adopted; the consequent
demands upon computing time made by a finite element solution are
exceptionally high.
There is, therefore, a real need for simolified alternatives
to the finite element method for use at the preliminary design
stage. Ideally, once the designer has established a reasonably
efficient structure by repeated analysis employing these approximate
techniques, a final check analysis by the finite element method
should be carried out.
The advent of more powerful conmuters in no way lessens the
value of such simolified alternative methods. In parallel with the
development in main frame' conmuting power in recent years there
has been an even more significant development in microconmuters
and programmable calculators. The sinmlified techniques to be
discussed in this paper may be programmed for solution on microcomputers of the types that are becoming increasingly available to
the design engineer.
The development and application of several simnlified, anproximate methods will be outlined in this paper. The first part of
the paper will concentrate upon the elaStic, global analysis of
cellular girders, whereas the second part will consider the calculation of the collapse load of complete multi-cellular girders.
Where possible, comparisons will be made between the approximate
results calculated by the simplified methods and those obtained,
either from a full finite element analysis, or from experiments.
2

GWBAL ELASTIC ANALYSIS OF CELLULAR STRUCTURES

Three techniques associated with the global analysis of

cellular girders in the elastic range will be discussed in some
detail. The methods are :
(a)
the nodal section method
the bar simulation method
(b)
(c)
the grillage method
2. I The fudal Section Method (2,3)
The nodal section method has been developed to provide a
relatively simple method of predicting the moments, stresses and
deflections in single-span and continuous box girders. It may be
applied to girders where the geometry of cross-section varies
along the span and to girders that are curved in ulan. In the case
of a straight girder of uniform cross-section the solution becomes
extremely simple and can be carried out conveniently on a microcomputer.
The method is based on the adaptation of the "ordinary" method
of analysis for folded plate roofs to the analysis of closed box
sections.

97

In general, any component plate of a cellular girder bends

both in its plane and perpendicular to its plane. It is assumed in
the nodal section method that the bending action of each plate
perpendicular to its plane can be represented by considering a
series of transverse one-way slab strips and that the in-plane
bending action of each plate can be idealized as that of a beam
spanning in the longitudinal direction between the supports.
On the basis of these assumntions, the simplified structural
behaviour shown in Fig.l, may be considered. The structural action
is assumed to consist of the action of a series of one-way,
transverse frames, elastically supported at the joints by a series
of inter-connected longitudinal plate beams. The frames are assumed
to transmit shears and moments in the transverse direction only,
this action being termed the "transverse frame action". The plate
beams, on the other hand, only transmit those force components
within their planes and this action is termed the "longitudinal
plate action"-.
The diagram gives an outline of the procedure followed in the
nodal section solution. The structure is first idealized by taking
a number of transverse nodal sections and a transverse frame is
considered to be positioned at each nodal section; in the case of
a girder of varying cross-sectional geometry, the dimensions of
each frame will be different. The transverse frame action under
the external loading is analysed by assuming that each frame is
rigidly supported at the joints and the transverse bending moments
and joint reactions are then determined simply for each frame in
turn.
Since the frame and plate systems are considered to be interconnected at the joints, the reactions calculated for the frame
system are applied as joint loads on to the plate system.
The longitudinal plate system is then analysed by considering
each individual plate to act as a beam spanning between the supports.
It is ensured that equilibrium of longitudinal stresses and compatibility of longitudinal strains is maintained at the edges between
plates and the longitudinal membrane stresses in the plates and
the joint deflections are calculated. Since the transverse frames
were analysed assuming the joints to be rigidly supported and
since these joints were allowed to deflect in the longitudinal
plate analysis, the displacements of the plate and frame systems
are not compatible at this stage of the solution. These incompatibilities can be conveniently expressed in terms of the sway
displacement of each component plate, see Fig. 2, and removed by
an iterative technique that is termed the "sway correction procedure".
In this procedure, the sway displacements are applied to the
transverse frame and the resulting moments evaluated by carrying
out another transverse frame analysis. The associated reaction
forces are then applied to the plate system and new joint

98

nodal sections

transverse
frame action

~
H [

interaction
between frame ;
and
:
'I
plate systemsl=t:
:
;:
I

,I
,

longitudinal
plate action

FIG. 1

'"

IDEALIZATION OF S1RUC1URAL BFHAVIOUR IN

NJDAL SECTION MEfOOD

M = 6 EIlI

12

FIG. 2

SWAY DISPLACEMENT OF TYPICAL PLA1E

deflections are calculated; a new set of incompatibilities will

arise from these deflections. The whole process is then repeated
and, by using a suitable accelerator, rapid convergence of the
iterative solution is guaranteed for virtually all types of cellular cross-sections.
The accuracy of the nodal section approach has been established
by several comparisons (2) with other methods. An interesting study
was reported by Anderson and Waddell (4) in 1975, wherein the
application of the method to the analysis of an actual bridge
structure was described. The bridge was divided into 3 parts, each
of different cross-sectional form, as shown in Fig. 3. The parts
comprised of :
(a) a girder of varying web depth, cantilevering from the supports;
this girder was of single-cell construction with edge cantilevers as shown in Fig. 3a.
(b) a twin-cell section with edge cantilevers, as in Fig. 3b,
forming the suspended centre span.
(c) a twin-cell section, as in Fig. 3c, forming the 3-span continuous approach viaduct.
Figure 3 also shows the various loading conditions that were
considered. Some results, typical of those obtained by Anderson
and Waddell, are shown in Fig. 4 for each of the three parts of
the bridge. Those quantities that are of most importance in bridge
design, viz. transverse bending moments, longitudinal direct
stresses and web shears are plotted. The nodal section values
obtained in each case are compared to those predicted by other
methods (5,6,7) and very good agreement is observed.
The application of the nodal section method to box girders
that are curved in plan has been discussed in detail by Al Rifaie
and Evans (3) and these authors reported several comparisons of
nodal section values with experimental and finite element results
for curved girders (8). Figure 5 shows typical results obtained
by these authors illustrating the effects of curvature on a singlecell concrete girder. This girder was subjected to a uniformly
distributed loading over the complete area of the top flange and
the curves show the variation of the transverse moments, longitudinal stresses and vertical deflections at mid-span, with the degree
of curvature. Obviously, as the radius of curvature becomes large,
the solutions tend towards the symmetrical, straight box condition.
The finite element and nodal section curves illustrate the same
trends and show a satisfactory degree of agreement.
On the basis of the many comparisons that have been carried
out with the results obtained from other methods and from experiments, it may be concluded that the nodal section method yields
results of good accuracy. It provides a convenient and economic
design tool for analysing the linear elastic behaviour of box
girder bridges.

100

II

Cal

IDDlce

FIG. 3 CROSS-SECTIONS ANALYSED BY ANDERSON AND WADDELL (4)

(a)

In-plane web shears (cantilever structure)

(b)

Transverse bending moments (suspended span)

(c)

Longitudinal stresses (continuous viaduct)

FIG. 4

______other methods
G nodal section

TYPICAL RESULTS OBTAINED BY ANDERSON AND WADDELL (4)

101

'?

ie

10

~...

-- ---

1000

so

.-.- _

040

"

III

>

20

10

!II

.=

_ _ - -'>- -

It"

,-

;'

v'

2'0

1ii

radius
span

- -I!r- Finite: de:.mcnt _1""'11

1'8

Nodal 5C"cticn wlulf

I'"

--

III

....A

.c.

e 1.4~
III

-I/,- _

inner edge

"
2

1
...

outer edge

"JO

"2

,0

g 0'&

CII..Itu edge:.

:;; 0'"

:l

"'=

0'4

WO.2

'l.

....J

radius
span

radius
-span

l'

.e

V60

...u

.2

..

<;:::
til

'0

-0u

'"

'50

'40
.~

20

outer edge:

'10

>

FIG. 5

RFSULTS FOR A CURVED BOX GIRDER

OBTAINED BY EVANS AND AL-RIFAIE (8)

102

2.2

The Bar Simulation Method (9,10,11)

When a box girder bends under an applied vertical loading, the
longitudinal direct stresses set up within the flanges are assumed
to be uniformly distributed across the flange width in normal
engineering beam theory. However, in most cases, particularly in a
wide flange, the shear deformations developed within the plane of
the flange plate lead to a non-uniform distribution of the stresses,
as shown in Fig. 6. This phenomenon is known as the "shear lag"
effect and its neglect can lead to an underestimation of the
stresses in the flange at positions close to the webs and can result
in an unsafe design.
The bar simulation method has been developed (9,10,11) to
predict this effect in a relatively simple way. The method has been
adapted from a technique originally used in the aircraft engineering
field (12) and it assumes that a flange plate may be idealized as
an assemblage of axial load-carrying "bar" members and shearcarrying "sheet" members.
Figure 7 shows the idealization of the flanges of three
typical box girders where the original, unstiffened flange is artificially divided into axial load-carrying and shear-carrying components. The axial load-carrying capacity of the flange is assumed to
be concentrated at a number of discrete bar members which are interconnected by a sheet that is assumed to be capable of carrying
shears only. The accuracy of the solution will increase with the
number of bars taken.
The idealized flange is then analysed by considering the
equilibrium of the forces developed in the bars and sheets and the
compatibility of bar and sheet displacements. A number of simultaneous differential equations are thus established, the total number
of equations being equal to the number of bars assumed. These
equations can be solved for any loading and support conditions and
explicit solutions have been presented by Evans and Taherian (11)
for the conditions of most relevance to bridge analysis.
Typical results obtained by the bar simulation method are
compared to finite element values in Figs. 8,9 and 10. Figure 8
shows the influence of the span/flange width ratio of the girder
upon the shear lag effect. The curves plotted in Fig. 8 refer to
the mid-span stresses in a simply supported girder under a unifOrmly
distributed load and show that the shear lag effect is very dependent upon the span/width ratio. The ratio is plotted on the horizontal axis in the diagram and it is seen that for short, wide
girders, the effects of shear lag are very large, whereas they
almost disappear for long, narrow girders. The curves plotted in
Fig. 8 show that for the single-span girders considered, the
results predicted by the bar simulation method lie very close to the
finite element values.

103

stress distribution
due to shear lag

FIG. 6

EFFECT OF SHEAR LAG ON FLANGE STRESSES

(a)

single-cell box

spine-beam girder

double-cell box

FIG. 7

TYPICAL IDEALIZATIONS OF FLANGE PLATES

104

112

Finite element
Bar Simulation

span/width
FIG. 8

VARIATION OF SHEAR LAG EFFECT WITH SPAN/WIDTH RATIO

f-!

015

hoi

Interm_cHace

suppo"

48

Two-span girder

FIG. 9

SHEAR LAG EFFECT IN CONTINUOUS GIRDER

II

105

[5t...=:',
II'

II'

d:::::J~

==",,==",==m

Spu _ 12m

Span_<tOm

(a) spine-beam concrete box

(b) four-cell steel box

FIG. 10

SHEAR LAG EFFECT ON STRESSES IN MULTI-CELL GIRDERS

FIG. 11

GRILLAGE IDEALIZATION

106

Results for continuous single-cell box girders are plotted in

Fig. 9. The edge and mid-flange stress variations for a two-span
girder, symmetrically supported, but with only one span loaded,
are plotted in Fig. 9a; corresponding values for a four-span girder
are plotted in Fig. 9b. In both cases, the bar simulation and finite
element stress curves show very good agreement at virtually all
positions on the span.
Stress distributions across the flange width of two girders of
very different cross-sections are plotted in Fig. 10. Stresses for
a typical spine beam concrete box are shown in Fig. loa and corresponding values for a four-cell steel box are given in Fig. lOb. In
each case the stress distributions at the mid- and quarter-span
cross-sections are plotted and the bar simulation and finite element
predictions are again observed to be in good agreement.
These results show that the bar simulation method can be used
to determine the shear lag effect in single- and multi-span girders
of various cross-sectional types. By adopting the explicit solutions
of the differential equations that have been presented (11), the
shear lag effect can be calculated very easily at the design stage.
2.3 The Grillage Method (13,14)
A multi-cellular structure may be idealized as a grillage,
wherein the real continuous plated structure is represented by a
series of discrete, orthogonally intersecting I-bearns, as shown in
Fig. 11. Once such an idealization of the structure has been established the analysis of the resulting grillage becomes very easy and
may be carried out conveniently on a micro-computer.
There are, however, two difficulties associated with the representation of a continuous plated structure by a series of discrete
beam members. The first of these is how to deal with the effects of
shear lag, see earlier in Fig. 6, in the flanges. It has been
shown (13) that shear lag can be taken into account conveniently by
adopting an effective width concept where the actual flange width
is replaced by a reduced "effective" width, over which the stress
is considered to be uniformly distributed. The extent to which the
width must be reduced can be determined from the bar simulation
method described above or from empirical tables (15). Once the
effective width of flange acting with each web is known, the
flexural properties of the idealized grillage I-beams can be
determined.
The second difficulty that arises from the idealization is to
ensure that the torsional stiffness of the closed cells is adequately
represented. It has been established (13) that a satisfactory
representation can be achieved by assuming that each I-beam
possesses a torsional stiffness equal to that of a single closed
cell.

107

The accuracy of such a grillage idealization can be illustrated

by considering the typical l6-cell structure shown in Fig. l2a.
The structure is square in plan and spans between simple end
supports. Figure l2b shows the longitudinal stresses developed at
the mid-span cross-section for three very different loading
conditions and Fig. l2c shows the corresponding mid-span deflections.
Grillage points are compared to finite element curves, showing
good agreement in each case.
Cellular structures frequently contain openings in the web
and these openings can cause a significant reduction in the shear
stiffness of the web plates and in the torsional stiffness of the
closed cells. Detailed experimental and finite element studies of
these effects have been carried out by Evans and Shanmugam (14,16)
and these studies have enabled the effects of openings to be
represented by the introduction of two empirical coefficients :
(a) an effective shear area coefficient (Ks) by which the full
web area should be multiplied to give an effective shear area.
(b) an effective torsional coefficient (KT) by which the full
torsional stiffness of a cell should be mUltiplied to give
an effective stiffness.
The results of the parametric studies carried out by Evans
and Shanmugam (16) were finally reduced to the form of the three
curves shown in Fig. 13. The shear area coefficient KS was found
to be largely independent of both the girder width/depth ratio
(b/d) and the width/thickness ratio (b/t). Its variation with the
diaphragm spacing/depth ratio (a/d) and the web onening ratio is
shown in Fig. l3a. The effective torsion coefficient KT was also
found to be independent of the width/thickness ratio (b/t). Its
variation with the other two parameters is shown in Fig. l3b.
Once the appropriate values of the reduction coefficients
have been determined from these curves, the grillage analysis of
a girder with openings may be carried out in the usual way. Results
for a typical l6-cell structure, simply supported on all edges are
shown in Fig. 14. The girder was subjected to equal loads at all
internal nodes, as indicated, and three different extents of web
openings were considered. In the first case the webs were solid,
then openings representing 12.5% of the original intact web area
were introduced and, finally, very large openings of 31.25% were
considered.
The grillage values (program GRILEJ) are shown as points for
comparison with the finite element curves and very good agreement
is observed in Fig. 14 for both deflection and stress values. The
plotted deflection values show the very significant loss in
stiffness occasioned by the introduction of openings.
Figure l5a shows the variation of the central deflection of
the 16-cell girder with web opening ratio for two different end
support conditions. For this diagram, because of the computing

108

"'-.,,"--r<'-::..-

~_~+-1"--(,

/!:::~===:?'

FIG. 12(a)

W
v b_
, ,'l,.<:9

ij
,I,

~n
! .
I

1"-

1-2.em

,,-1.~ em

I'

*\.2~1'2em

,_

300 em b _ 300 em b - 300 em b - 300 em

TYPICAL 16-CELL STRUCTURE

25P

60

I.

2'5P

40
20
25P

o
3

5
0

10

2
- - FIn;le elemenl melhod
o
Grillage me1hod

FIG. 12(b)

LONGITUDINAL
STRESSES (N/mlh

3
4

---Finne elemenl melhod

Grillage melhod
0

FIG. 12 (c)

VERTICAL DEFLECTIONS
Crrun)

109

~O~------~------~~'-----~~------7.40
Web opening: 'III

(a)

Effective shear area coefficient

CKS)

~O~------~~------~20'-------~~------~40
Web_nlng:'III

Cb)

Effective torsional coefficient

FIG. 13

CKr)

DESIGN CURVES TO ALLOW FOR EFFECTS

OF OPENINGS

110

"

~ .......
-

1'()

~,

2-0

,----,

1IJ

-,-i

4-0

-\..

-" "-

........

C---

(l)

lH

S-O

(l)

--- --------j

x --_

3-0

-,-

................

---

Flnl'e element resunl for Itructure

- - lolld weill
- - - - '2-5'1ftwebopenlngl
- - - 31'25'1ftweb openings
GRllEJ 'Huh.
.tructure
o IOlidwelll
x
12-5% web openings

-, _

I.

'--l

lor

&0

,----,

200

-Z

FIG. 14 RESULTS FOR 16-CELL

GIRDER

- - - __

_
J
...

. "~-~~

'-'

,S-o

1IJ

<l)

1IJ
1IJ

<l)

/'

;...

10-0

C/J

/'

--

/ .x- ./"~~ _ _ '

/-/

,.

../

5-0 -- --

()o()~~-----------------~~~------------------~~~
6-0

SINc!u,e limply ,uported on lou, lides

_ _ GRILEJ ,esuUI
')(
,,,nile element re.ults
5 -0
SlNetu,e wllh SImply IUpported edges
and clamped ends
__ - GRllEJ ,esults

"-0

finite element result.

3-0

tl ------

2-0

Jf",.,.""'''

'"

... ..-

"

"

x
Ala

,"

_-8

80-0
- - GRllEJ re8ulll
o
Finite element reautll

60-0
400
2()O

'-a

(b)

10

Web opening: '1ft

24-cell girder

--------,,.,,,O,...--------.iiOI<-------"3""h---

O-O O
!:-

(a)

OpenIng: '1ft

16-cell girder
FIG. IS

VARIATION OF DEFLECTIONS (mm) WITH OPENING SIZE

III

time involved, only three finite element solutions were obtained

for each case so that these values are shown as points. Several
grillage solutions were carried out to enable the continuous
curves to be drawn. A similar comparison is shown in Fig. 15b for
a 24-cell structure and, in both cases, the finite element points
lie close to the grillage curves, indicating good agreement.
These comparisons, and many others, have confirmed the
accuracy of the grillage approach for the analysis of multicellular girders. Moreover, the grillage solution time of 18 seconds
for the 24-cell girder (see Fig. l5b) , compared to the corresponding equivalent time of 3 hours for a finite element solution,
indicates the usefulness of the grillage approach as a design tool.
3

ANALYSIS OF NON-LINEAR BEHAVIOUR AND COLLAPSE

With the current changes in many design codes from the
allowable stress approach to limit state concepts, it is important
to be able to calculate the collapse loads of structures. The
availability of simplified methods is of even greater importance
in this context, since the incremental procedures that must be
adopted to allow for the effects of material and geometrical nonlinearity make the computing time required for a finite element
solution unacceptable for the analysis of anything other than an
isolated element. When an individual panel is analysed in this way,
it is difficult to prescribe the correct boundary conditions to
represent the restraints imposed by adjacent panels.
The extension of the grillage approach to the analysis of the
non-linear and collapse behaviour of a multi-cellular structure
will now be considered.
3.1 The Prediction of the Non-linear and Colla se Behaviour
o a Multi-Cellular Girder (17
The grillage method discussed in section 2.3 can be extended
to analyse the non-linear and collapse behaviour of cellular
structures. The high bending moments arising from lateral loading
usually cause buckling in the slender compression flanges of these
girders. In the post-buckling range, the non-uniform distribution
of flange stress, caused by the shedding of load away from the
buckled central region of the flange to the edge regions, may be
taken into account by adopting an effective width approach.
The reduced effective width in the post-buckling range can be
determined from well-established empirical formulae. It will be
dependent upon the magnitude of the stress developed in the
compression flange and upon the flange buckling stress. Since the
effective flange width to be assumed depends upon the magnitude of
the flange stresses and since these stresses, in turn, depend upon
the effective width assumed, a closed form solution is not possible.

112

Thus, in the non-linear, post-buckling range the grillage

solution has to be carried out incrementally. Starting from the
linear elastic range before buckling, the applied loading is
increased in steps. For each load, section properties corresponding
to the flange stresses calculated for the previous load level are
assumed for the members and an elastic grillage analysis is carried
out. The procedure is approximate since the assumed member properties
are always one step behind the applied loading, so that the
magnitude of the chosen loading increment affects the accuracy of
the solution. However, it has been found that the solution converges rapidly with the increment size.
At each loading level, every grillage member is considered
individually; the average shear stress in the web is combined with
the flange stress and the Von Mises yield criterion is applied.
If yield is observed to have occurred in any of the members, they
are not allowed to carry any further loading and are effectively
removed from the girder. Complete failure of the structure occurs
when a sufficient number of individual members have failed so
that a collapse mechanism can develop.
Typical results are shown in Fig. 16 where the central deflection of a 24-cell girder is plotted agains the applied load; midspan stress values are also plotted. The non-linear relationship
between load and deflection is apparent and the girder is observed
to fail at a load of 4900 kN. This value predicted by the grillage
method lies within 7% of the load of 5260 kN obtained from a yield
line analysis, thus showing satisfactory agreement.
Further results for a typical 28-cell structure are presented
in Fig. 17, where the mode of failure of the girder is shown on
the schematic plan views on the left, and the table on the right
of the diagram indicates the magnitudes of the bending and shearing
stresses in the various members as they reach yield.
First yield is observed to have occurred in the transverse
members at the centre of the span at a load of 75 kN. These
members are close to the longitudinal edges so that shear is the
predominant action, as shown by the tabulated values. Before increasing the load to 80 kN, the four yielded members were removed
and during this increment, yield is observed to have occurred in
four further transverse members, with shear again predominating.
Upon removing these four members and applying a further increment
of load to 85 kN, widespread yield is observed allowing the
formation of a collapse mechanism and consequent failure of the
girder.
Unfortunately, little theoretical or experimental evidence
exists at present to enable the accuracy of the proposed grillage
approach in the non-linear:range to be proved conclusively. The
complexity of a multi-cellular structure also makes it impossible
to obtain finite element results for comparison. The limited

113
Flange edgeltr...: N/mm2
8~0~____~4~0____~80~____~1~20~__~1~80~__~2~OO~__~2~40T-___

:li6000

FIG. 16 RESULTS FOR 24-CELL

GIRDER
2000

Can".1 CSellexlon: mm
Cen"al dolitxlon 01 tllo ,truclU," with thlckor wOb,
.round tho porlphe'Y
Flange Idoe Itrell a' the centre 01 the Itructure with
thlckor wobi around the poripho'Y

-I

-I

1-

sr.

:t

2-

load per node

bendinG

Shed I

~t

strcs~

..ess

rIItio
u/o v

...

0.~2

0.86

0.84

1.00

0.57

1.00

0.78

0.67

load per node

85 kN

1.00

l/T v

1.00

0.~8

1.00

1.00

0.78

1.00

0.00

0.34

1.00

10

0.60

0.99

stage 3
80 kN

ratIo

.:;

2-

stage 2

load pe.- node - 75 kN

~

Ii

1.00

0.~7

12

1.00

0.00

FIG. 17 TYPICAL RESULTS INDICATING

COLLAPSE OF 28-CELL GIRDER

114

comparisons that have been carried out to date have shown encouraging agreement but the author is currently conducting a series of
tests on large-scale models to provide further data.
The extension of the grillage approach has been carried out
without adding greatly to the complexity of the solution process.
Consequently, even in its non-linear form, it may still be regarded
as a convenient design aid.
4

CONCLUSION
A number of different methods of analysis for cellular bridge
structures have been described in this paper. In each method, an
idealized structural behaviour is assumed so as to provide a simplified solution procedure, making each method suitable for use during
the preliminary stages of design when repeated analyses may be
required to establish optimum proportions. Some typical comparisons
with results obtained from other methods have been presented to
illustrate the accuracy that can be achieved by the simplified
approaches; such comparisons are presented in more detail in the
listed references. All the methods presented are suitable for
solution on the types of microcomputers that are becoming increasingly available to the design engineer.
REFERENCES
1.
2.
3.
4.
5.

6.
7.
8.

Zienkiewicz, O.C. The Finite Element Method, McGraw Hill Ltd,

London, 1977.
Rockey, K.C. and Evans, H.R. The Nodal Section Method for the
Analysis of Box Girders. Proc.Int.Assoc.Bridge and Struct.Eng.
Vol. 35, 1975, pp. 185-216.
AI Rifaie, W.N. and Evans, H.R. An Approximate Method for the
Analysis of Box Girder Bridges that are Curved in Plan. Proc.
Int.Assoc.Bridge and Struct.Eng., Feb. 1979, pp. 1-15.
Anderson, R.G. and Waddell, J.W. Analysis of Concrete Box
Girder Structures by the Nodal Section Method, PTRC Conference,
London, July 1975.
Highway Engineering Computer Branch of the Department of the
Environment. Program for the Analysis of Box Girder Bridges
using Quadrilateral Finite Elements, Report HECB/B/14 (QUEST),
May 1975.
Cherng, Y.K. STRIP Programs for the Analysis of Straight and
Curved Box Girder Bridges. Report and User Manual, R. Travers
Morgan, Consulting Engineers, London, March 1972.
Scordelis, A.C. Box Girder Bridges - A Folded Plate Analysis.
Department of the Environment Bridge Design Computer
Application Programs (MUPDI) Report HECBjB/ll, Nov. 1970.
Evans, H.R. and AI Rifaie, W.N. An Experimental and Theoretical
Investigation of the Behaviour of Box Girders Curved in Plan.
Proc.Instn.Civ.Engrs., Part 2, 1975, 59, Jrne, pp. 323-352.

115

9.

Evans, H.R. and Taherian, A.R. The Prediction of the Shear

Lag Effect in Box Girders. Proc.Instn.Civ.Engrs., Part 2,
1977, 63, March, pp. 69-92.
10. Taherian, A.R. and Evans, H.R. The Bar Simulation Method for
the Calculation of Shear Lag in Multi-cell and Continuous
Box Girders. Proc.Instn.Civ.Engrs., Part 2, 1977, 63, Dec.,
pp. 881-897.
11. Evans, H.R. and Taherian, A.R. A Design Aid for Shear Lag
Calculations. Proc.Instn.Civ.Engrs., Part 2, 1980, 69, June,
pp. 403-424.
12. Kuhn, P. Stresses in Aircraft and Shell Structures. McGraw
Hill, New York, 1956.
13. Evans, H.R. and Shanmugam, N.E. An Approximate Grillage
Approach to the Analysis of Cellular Structures. Proc.Instn.
Civ.Engrs., Part 2, 1979, 67, March, pp. 133-154.
14. Evans, H.R. and Shanmugam, N.E. The Elastic Analysis of
Cellular Structures Containing Web Openings. Proc.Instn.Civ.
Engrs., Part 2, 1979, 67, Dec, pp. 1035-1063.
15. British Standards Institution. Steel Concrete and Composite
Bridges; Part 3 - Code of Practice for Design of Steel
Bridges. BS5400, 1982.
16. Shanmugam, N.E. and Evans, H.R. An Experimental and Theoretical Study of the Effects of Web Openings on the Elastic
Behaviour of Cellular Structures. Proc.Instn.Civ.Engrs.,
Part 2, 1979, 67, Sept., pp. 653-676.
17. Sh anmugam , N.E. and Evans, H.R. A Grillage Analysis of the
Non-linear and Ultimate Load Behaviour of Cellular Structures
under Bending Loads. Proc.Instn.Civ.Engrs., Part 2, 1981, 71,
Sept., pp. 705-719.

III. CONCRETE SLAB AND BOX GIRDER BRIDGES

119

BERKELEY COMPUTER PROGRAMS FOR THE ANALYSIS OF CONCRETE

BOX GIRDER BRIDGES

Alex C. Scordelis
Professor of civil Engineering
University of California
Berkeley, California USA

SUMMARY
A review of the analytical methods and computer programs, which
have been developed at the University of California at Berkeley for
the linear and nonlinear analysis of reinforced and prestressed
concrete box girder bridges, is presented. Short descriptions of
15 different computer programs based on one of the following methods
of analysis are given:
(1) folded plate elasticity method; (2)
finite strip method; (3)
finite element method; (4)
finite segment
method; and (5)
nonlinear analysis method. These various programs
provide the capability for the detailed analysis of single and
multicell box girder bridges of one or more spans, which. are straight,
curved, skew or of arbitrary geometry in plan. General loadings and
arbitrary boundary conditions may be treated.
The linear analysis programs assume the analytical model to be
a linear uncracked homogeneous concrete system. The nonlinear
analysis programs utilize analytical models which include the
concrete, reinforcing steel and prestressing steel and can predict
the structural response of bridges, under initial and time dependent
load histories, through their elastic, cracking, inelastic and
ultimate ranges.
Five representative numerical examples are presented to
illustrate some of the solutions which can be obtained using the
Berkeley computer programs. Results obtained are compared with
experimental results and those from other analytical solutions.

120
1.

INTRODUCTION

The continuing expansion of the highway network throughout the

world is largely the result of the great increase in traffic,
population, and the extensive growth of metropolitan urban areas.
This expansion has led to many changes in the use and development
of various kinds of bridges.
In recent years, single or multicell
reinforced and prestressed concrete box girder bridges have been
proposed and widely used as economic and aesthetic solutions for
the overcrossings, undercrossings, separation structures and viaducts
found in today's modern highway systems. These structures can be
constructed to follow any desired alignment in plan. The very large
torsional rigidity of the box girder's closed cellular section
provides structural efficiency, while its broad, unbroken soffit,
viewed from beneath is more esthetically pleasing than open-web
type systems.
In the span ranges of 60 to 100 ft (18 to 31 m), cast-inplace multicell reinforced concrete box girder bridges which are
straight, skew, curved or of arbitrary geometry in plan are widely
used for overcrossings, undercrossings and interchange structures.
For longer spans up to about 200 ft (61 m), post-tensioned, prestressed, cast-in-place multicell box girder bridges are often
used. For spans between 200 and 800 ft (61 to 244 m), commonly
used in viaducts or water crossings, segmentally erected prestressed concrete box girder bridges of one, two or more cells may be
used. The typical box girder bridge is a folded plate structure
consisting of top and bottom concrete slabs monolithically connected by a series of vertical, sloping or curved webs (Fig. 1).
Present and past designs of box girder bridges usually have
been based on linear elastic analyses of simplified models of
uncracked homogeneous systems. Empirical formulas are often used
to determine the transverse distribution of longitudinal girder
moments and vertical shear forces in the webs due to design truck
vehicle wheel loads. Empirical formulas are also used to determine
local transverse and longitudinal slab design moments.
In many
recent bridge specifications, an alternate load factor method of
design based on ultimate strength calculations is permitted, but
little guidance is given on how this can be applied to complex box
girder bridge geometries. Basic knowledge on the nonlinear
behavior and ultimate strength of concrete box girder bridges is
lacking at the present time.
Because of their extensive use in California, a continuing
program of research on box girder bridges has been conducted at the
university of California at Berkeley since 1965. A systematic
plan was developed to study successively straight, simple and
continuous bridges, curved bridges, skew bridges and bridges of
arbitrary plan geometry. For each of these configurations the

121

FIG. 1

MULTI-CELL BOX GIRDER BRIDGE CROSS SECTIONS

Ny
FIG. 2

Nxy

Nx

MXY~MXY

DISPLACEMENTS AND INTERNAL FORCES ON A DIFFERENTIAL ELEMENT

FIG. 3

, Rtr

~
I

FIG. 4

PRISMATIC BOX GIRDER BRIDGES

(JOINT

.... \'j"j

(i=-ijj

\~J

JOINT AND PLATE EDGE FORCES AND DISPLACEMENTS

IN GLOBAL AND LOCAL COORDINATE SYSTEMS

122

approach has been to: (1) study the available literature; (2) develop
analytical methods and general computer programs; (3) perform
experimental studies on elastic models to verify the analytical
methods developed if deemed necessary; (4) make analytical parameter
studies; (5) test large scale reinforced concrete models or prototypes; and (6) develop recommended design procedures.
Initially,
reinforced concrete bridges and linear methods of analysis were
studied, but in recent years since 1977 the emphasis in this
research has shifted to prestressed concrete bridges and to nonlinear
methods of analysis. The results of this continuing research program
at the University of California have been described in a series of
research reports and technical papers which are listed as references
in Appendix 2, under A and B respectively.
The purpose of the present paper is to review the analytical
methods and computer programs, which have been developed at Berkeley
for box girder bridges, and to present numerical results from
several examples. Both linear and nonlinear analytical models and
methods will be described in the body of the paper, while in
Appendix 1 short descriptions of 15 different computer programs for
box girder bridges, which are available, will be described. Detailed information on these computer programs and their theoretical
basis can be found in the published research reports and technical
papers listed as references in Appendix 2.

2.

LINEAR ELASTIC ANALYSIS

An analytical solution of the true response of a concrete

box girder bridge under load is complicated by the usual factors
common to other reinforced or prestressed concrete structural
systems. It is a highly indeterminate structure; it is made of two
materials, concrete and steel; under increasing load it experiences
cracking and material nonlinearities and thus some redistribution
of internal forces; and also the internal forces are time dependent
because of creep and shrinkage in the concrete. Nevertheless, as
for other reinforced or prestressed concrete systems, such as frames,
slabs, and shells, it has generally been accepted that for design
purposes, the distribution of internal forces, moments and displacements in a box girder bridge due to applied loads can be
based on linear elastic analyses of an uncracked homogeneous
concrete system.
In a complete analysis of a multi-cell box girder bridge, all
of the internal forces and displacements shown on a typical element
in Fig. 2 taken from a deck or web plate of the bridge should be
determined. The internal forces Nx , Ny and Nxy are termed membrane
forces while Mx, My, Mxy' Qx and Qy are internal forces due to
plate bending.
In many approximate analyses certain internal
forces are assumed to be negligible and are thus taken as zero.

123

Of prime interest from a design standpoint are those internal

forces which determine the reinforcing or prestressing steel
requirements for the bridge. These are Nx for -the main longitudinal
tension steel; Nxy for the diagonal tension steel; and My and Ny
for the transverse steel.
Many analytical models and methods have been developed for the
linear elastic analysis of box girder bridges. Among these are
approximate methods based on simplified structural behavior such as
the use of an elementary or extended beam method; an equivalent
beam grillage or anisotropic slab to represent the system; exact
and approximate methods based on folded plate theory; and numerical
solutions based on finite strip, finite element, finite segment
or finite difference methods. No attempt will be made to review
all of these methods here.
Instead, only solutions and associated
computer programs which have been developed at the University of
California at Berkeley will be described. These linear elastic
analyses of an uncracked homogeneous structure give complete
solutions for all of the internal forces, moments and displacements
shown in Fig. 2 for any type or position of loading on the bridge.
The linear elastic solutions and computer programs to be described
can be classified under the following methods of analysis; (1)
folded plate elasticity method; (2) finite strip method; (3) finite
element method; and (4) finite segment method. Each of these
methods has its advantages and disadvantages and the appropriate
method should be carefully selected based on its advantages and
applicability for the particular problem to be solved.

3.

FOLDED PLATE ELASTICITY METHOD

A multi-cell prismatic box girder bridge (Fig. 3) may be

modelled analytically as a folded plate system consisting of an
assembly of longitudinal plate elements interconnected at joints
along their longitudinal edges and simply supported at the two ends.
The known quantities input into the problem include the geometry,
dimensions and material properties of the plate elements, the
surface and joint loadings and the boundary conditions along the
longitudinal joints. Each plate element selected is assumed to
extend longitudinally over the entire span and transversely
between designated joints on the cross-section (Fig. 4).
An analysis for applied loads with any arbitrary longitudinal
distribution for straight bridges or any arbitrary circumferential
distribution for curved circular bridges may be performed using a
direct stiffness harmonic analysis. The applied loads are first
resolved into Fourier series components. An analysis is made for the
components of each particular harmonic and then the final results are
obtained by summing the results for all harmonics used to represent
the load. Once the solution technique, which involves extensive

124
computations, has been devel~ped for a single harmonic, it can be
reused for any harmonic, and thus the method is ideally suited to
the application of a digital computer.
The analysis for each harmonic load has the advantage that for
straight or curved circular bridges such loads will produce displacements of the same variation and vice versa and thus a single
characteristic value may be used to describe any force or displacement pattern. For example, the displacement pattern for the nth
harmonic:

mrx

rex)

ro sin

reel

ro sin

e-; for

mr8

(1)

for straight bridges

curved circular bridges

(2)

eo

may be described by a single value rD. L and

define the span
for the straight and circular bridges respectively (Fig. 3). This
makes it possible to treat an entire joint as a single nodal point
and to operate with single forces and displacements instead of
functions.
If the condition of static equilibrium and geometric
compatibility are maintained at a nodal point they will automatically
be satisfied along the entire longitudinal joint. Thus the threedimensional prismatic bridge problem may be treated as a twodimensional problem in the transverse direction. A direct stiffness method applied to such a system results in a structure stiffness matrix which is extremely well conditioned for solution since
the non-zero coefficients are all grouped in a narrow band about
the main diagonal.
Each longitudinal joint has four characteristic degrees of
freedom; it can displace vertically and horizontally in the plane
of the cross-section; it can move longitudinally tangent to the
joint; and it can rotate about an axis tangent to the joint. These
directions define a global coordinate system for displacements or
forces at the joint (Fig. 4a).
The direct stiffness method is described in detail in many
textbooks and other pUblications. Thus it need only be briefly
outlined here by the following steps for loads applied at the joints.
1.

Determine the 8 x 8 element stiffness matrix k

in S = kv for each plate element, relating the edge
forces S to the edge displacements V in the local
coordinate system (Fig. 4b).

2.

Transform the element stiffnesses from a local k

(Fig. 4b) to a global k (Fig. 4c) coordinate
system. Assemble these 8 x 8 matrices into the
structure stiffness matrix K, which for the
structure in Fig. 4a would be a 24 x 24 matrix.

125

3.

Solve the equilibrium equations, R = Kr, where R

represents the applied loads, for the unknown
joint displacements r.

4.

Determine the plate element internal forces and

displacements (Fig. 2) by expressions relating
these quantities to the joint displacements.

The basic logic of a general computer program for the direct

stiffness harmonic analysis is independent of the method chosen for
determining the element stiffness matrices.
3.1

Straight Single Span Bridges

For straight bridges, formulas based on elasticity theory for

the stiffness matrix coefficients defining membrane and plate
bending action for isotropic, linearly elastic plates were first
utilized to develop the computer program MULTPL (1965) [AI, Bl, B2,
B6] for straight single span bridges, simply supported at the two
ends. The elasticity equations defining the plane stress problem
for membrane loads in the plane of the plate and classical thin
plate theory for loads normal to the plate are used to determine
the necessary expressions for the fixed edge-forces and moments due
to applied loads between the joints and also for the element stiffness matrix k in the local coordinate system needed in the direct
stiffness harmonic solution. Results obtained by this elasticity
method represent the most exact available solution for a linear
elastic analysis of box girder bridges and they are often used to
measure the accuracy of results by other methods.
3.2

Straight Continuous Span Bridges

The above solution for a single span bridge without interior

supports was extended to continuous span bridges with interior
rigid diaphragms in the computer program MUPDI (1966) [AI, B3, B6]
and to bridges with interior flexible diaphragms and supporting
planar frame bents in the computer program MUPDI3 (1971) [A13, B6,
B16]. A force method of analysis is used in combination with the
direct stiffness harmonic analysis described above.
For structures supported by flexible planar frame bents the
redundants are taken as the interaction forces between the folded
plate system and the bents (Fig. 5). The interaction forces are
represented by a set of three joint forces at each longitudinal
joint (Fig. 5d) consisting of vertical, horizontal and rotational
components in the plane of the transverse cross-section. The
analysis is carried out in the following sequence of steps:
1.

With the redundants X set equal to zero (Fig. 5b),

126

(a)

Elevation of the Structure

----1
-(b)

(c)

Primary Structure

Under Unit Redundant Force

~OLDED
~LATES

RIGIO LINKS

~+F------'~4r ~
GIRDER AXIS

SUPPORT
BENT

(d)

Joint Redundant Forces

FIG. S

(e)

Idealized Frame Bent

ANALYSIS OF A BOX GIRDER BRIDGE ON A FLEXIBLE BENT

127
the folded plate structure is analyzed for the
given external applied load using a direct stiffness
harmonic analysis as described above. A displacement vector 00 (24 x 1 in Fig. 5) is found for this
case, which defines the displacements at the points
where the redundants are to act.

2.

The folded plate structure is next analyzed for

unit values of each of the redundant forces X
(Fig. 5c) and the corresponding flexibility
matrix FI (24 x 24 in Fig. 5) for the folded plate
system is formed.

3.

The idealized planar frame bent (Fig. 5e), composed

of one dimensional elements, is then analyzed by
the direct stiffness method. The total structure
stiffness matrix for the frame bent is formed and
a static condensation is carried out to eliminate
the degrees of freedom which do not correspond
to the redundant forces. Finally, the flexibility
matrix F2 (24 x 24 in Fig. 5) corresponding to the
unit redundant forces is found by inverting the
frame stiffness matrix.

4.

Geometric compatibility requires that

(3a)

(3b)

or
00

FX

(3c)

where

(4)

and thus the redundants

x
5.

F -I

may be found from Eg.

(3c)

00

(5)

The simply supported folded plate structure and the

planar frame bent can now be analyzed, subjected to
the known applied external loading and the known
redundant forces, to determine the final stresses
and displacements in the structure.

For interior diaphragms which are not externally supported, a

similar force method is used to determine the redundants. However,

128

in this case, because the diaphragms can undergo three degrees of

rigid body motion in their own plane in addition to the deformation
of the diaphragms themselves, additional transformation and condensation of the flexibility matrices FI and F2 are required. These
are done automatically in the computer programs and are described
in detail in Ref. A13.

4.

FINITE STRIP METHOD

When using the direct stiffness harmonic analysis, the direct

application of the elasticity theory to determine the stiffness
matrix of plate elements which are not isotropic becomes exceedingly
complex and resort must be made to simpler approaches. A theory
known as the finite strip method is used in these cases. The finite
strip method may be thought of as a special form of the finite
element method.
It approximates the behavior of each plate by an
assemblage of narrower longitudinal finite strips for which
selected displacement patterns varying as harmonics longitudinally
and as polynomials in the transverse direction are assumed to
represent the behavior of the strip in the total structure. Local
coordinate systems and displacement interpolation functions for a
typical finite strip having a longitudinal span L = 2a and a
transverse width B = 2b are shown in Figs. 6a and 6b. with these
assumptions the displacement at any point in the strip can be
expressed in terms of the eight nodal point displacements shown in
Fig. 4b. Using successively strain-displacement relationships, a
stress-strain law and thence either the principle of virtual displacements or the principle of minimum potential energy, the element
stiffness matrix and also the generalized or consistent nodal
forces for surface loads can be derived for the finite strip.
4.1

Straight Single Span Bridges

The above method has been used to develop a computer program

MULSTR (1970) [A4, B5, B6, B16] for straight single span box girder
bridges made up of orthotropic plates with closely spaced eccentric
ribs or stiffeners in the longitudinal and transverse directions.
The properties of such an orthotropic plate are assumed to be
uniform over each longitudinal finite strip, but can vary from
strip to strip if desired. The two basic types of eccentric ribs
used are designated as torsionally soft ribs and torsionally stiff
ribs (Figs. 7a and 7b). The former consists of open slender
sections that have little torsional resistance, whereas the latter
includes open or closed box sections with considerable torsional
resistance. Detailed derivations for the element stiffness matrix
and the consistent nodal forces for such finite strips using the
basic theory for orthotropic plates with either torsionally soft
or stiff eccentric ribs are presented in Ref. A4.

129

L 20

82

.. TwVzz

Uz

-I

"

FIG. 6&

EDGE I

UI

y..... y

"'1

X,U,X

III

w,

81

F
&l

I,

Z,W

= K 10
Y= y/b

J(

POSITIVE NODAL POINT DISPLACEMENT AND LOCAL COORDINATE SYSTEM

I
I

.-~IIIIIIIII~

'

COS nTTi!

[-"-1-...:"
....
~

FIG. 6b

cca::a::crm 1" III

DISPLACEMENT INTERPOLATION FUNCTIONS FOR A FINITE STRIP

130

FIG. 7a

FIG. 7b

TYPICAL ELEMENT OF TOIISIONALLY SOFT ORTHOTROPIC PLATE

TYPICAL ELEMENT OF TOIISIONALLY STIFF OR'l'HOTROPIC PLATE

131
4.2

Circularly Curved Bridges

The finite strip method and direct stiffness harmonic analysis

have also been used to develop the computer programs CURSTR (1970)
[A6, B6, B8, B14, B16], CURDI (1974) [A20], and CURDIP (1980) [A26]
for box girder bridges which are circularly curved in plan (Fig.
8a) and made up of orthotropic plate elements without eccentric
stiffeners. Each plate element, which may be a segment from a
circular ring plate, a conical shell or a cylindrical shell (Figs.
8b, c, d), is idealized by a number of circumferential finite
strips. Displacement patterns in each strip are assumed to vary as
harmonics in the circumferential direction and as shown in Fig. 6b
in the transverse direction.
CURSTR can be used to analyze curved single span bridges and
is similar in concept to MULTPL for straight bridges. CURDI can be
used to analyze curved continuous span bridges with interior radial
diaphragms or supporting planar frame bents and is similar in
concept to MUPDI3 for straight bridges. By setting the radius to a
very large value in CURSTR or CURD I a close approximation to a
straight bridge is obtained and thus, if desired, results from a
finite strip method can be compared with those from the more exact
folded plate elasticity method used in MULTPL and MUPDI3.
CURDIP extends CURDI to take into account automatically the
effect of post-tensioned prestressing tendons. By defining as
input the tendon profiles, jack forces, jacking procedure and
friction properties, the corresponding equivalent loads are
computed automatically by the program and applied to the structure.
This important addition simplifies considerably the analysis of
prestressed concrete box girder bridges subjected to general dead
and live loadings.
In order to describe a general prestressing tendon profile in
CURDIP the real tendon is divided into piecewise linear segments
(Fig. 9). To minimize the necessary amount of input data each
tendon is first divided into portions which are thence subdivided
into various numbers of segments of equal length in each portion.
The tendons are assumed to be in the mid-thickness plane of the
finite strips. Cylindrical coordinates defining the tendon points
at each end of a segment are input.
These coordinates are transformed to a global Cartesian
coordinate system so that vector algebra may be used for the
necessary calculations to determine the equivalent loads due to
prestressing.
This process involves the following steps (Fig. lOa):
(1) the
length and direction of each tendon segment is determined; (2) the
angle 8i between segments at each tendon point is calculated;

132

rve d Br idg e
Ci rc ul ar ly Cu

(a)

,/

,.

"

(h)

.......

_----

Ri ng Pl ate

El~ent

,
,r,. (C)

Co nic al El~ent

FIG . B

(d)

'-;.~-.,

me nt
Cy lin dr ica l Ele

E ELEMENTS
GIR DE R BR IDG
CURVED BOX

133
(3) the standard friction formula is used to calculate the prestressing tendon force Fi existing in each tendon segment.
F.
l

(6)

where Fo = tendon force at jacking end; W = coefficient of friction;

a = total cummulative angle change from jacking end; k = wobble
friction coefficient; and S = cummulative distance along tendon
from jacking end; (4) with the force in each tendon segment known,
the concentrated equivalent load vector at each tendon segment
point can be found (Fig. lOb) in a global coordinate system by
vector algebra; (5) this load vector R is transformed back to a
cylindrical coordinate system and applied as statically equivalent
nodal joint loads Ri and Rj to the finite strip in which the tendon
point occurs (Fig. 11).

5.

ADVANTAGES AND DISADVANTAGES OF DIRECT STIFFNESS

HARMONIC ANALYSIS

Both the folded plate elasticity method and the finite strip
method make use of the direct stiffness harmonic analysis procedure described earlier. Advantages of this procedure are:
1.

It is well suited for computer programming and can

yield a complete and accurate solution in a
reasonable amount of computer time.

2.

Any desired theory can be used to determine the

response of the individual plate or strip elements.

3.

Both surface and joint loadings of arbitrary

longitudinal variation as well as prestressing
can be treated.

4.

Any combination of displacement and force

boundary conditions along the longitudinal joints
can be used.

Disadvantages of this procedure are:

1.

It is restricted to prismatic structures which

may have interior supports; but must be simply
supported at the extreme ends.

2.

The material and geometric properties of each

plate or strip element making up the cross-section
must be constant in the longitudinal direction.

134

SECTIONS
TENDON POINT

.......

PORTION (J)
10
FIG. 9

SEG~'ENTS

PORTION (1+ 1)
2 SEGMENTS

PRESTRESSING TENDON SUBDIVISION INTO PORTIONS AND SEGMENTS

SEGMENT:

,,,
,

CD

TENDON POINT:
TENDON ANGLE CHANGE:

0.

FIG.

lOa

FIG. lOb

LENGTHS, ANGLE CHANGES, AND TENDON SEGMENT FORCES

CON:ENTRATED EQUIVALENT LOAD AT TENDON POINT

135

i,j

FIG. 11a

FIG. llb

= STRIP EDGES
= TENDON POINT

TENDON PROFILE ON SIX FINITE STRIPS

STATICALLY EQUIVALENT JOINT LOADS

136

6.

FINITE ELEMENT METHOD

The finite element method is described extensively in textbooks

and in the technical literature.
In the finite element method the
actual continuum is replaced by an assembly of finite elements
interconnected at nodal points (Fig. 12). For a general box girder
bridge system, the finite elements may consist of two-dimensional
shell or plate elements and transverse or longitudinal one dimensional frame type elements.
Stiffness matrices which approximate
the behavior in the continuum, are developed for the finite elements
based on assumed displacement or stress patterns, after which an
analysis based on the direct stiffness method may be performed to
determine nodal point displacements and thence the internal stresses
in the finite elements.
The accuracy obtained is dependent on the
assumptions used in deriving the stiffness matrices and on the
fineness of mesh used in subdividing the structure. The results often
satisfy compatibility, but not necessarily equilibrium in the
continuum until a sufficiently fine mesh is used.
A number of investigators have developed general shell programs
which could be used for analyzing box girder bridges. However, if
available, it is better to use special purpose programs which take
advantage of the repetitive and special nature of these structures.
These should provide the required accuracy in the results with a
minimum of the following:
(1) required amount of input preparation;
(2) execution time and core storage in the computer; and (3) amount
of 9utput data reduction necessary for meaningful interpretation.
Three such special purpose programs developed at the University of
California for the linear elastic analysis of box girder bridges are
CELL (1970) [A7, B6, B7, BlO, B16), FINPLA (1967) [A2, A5, A6] and
FINPLA2 (1971) [A12, B6, B16].
6.1

Bridges with Arbitrary Plan Geometry and Constant Depth

Many multicell box girder bridges used in separation and interchange structures fall into this category. The computer program
CELL was written especially for this type of bridge in which the
bridge is made up of top and bottom horizontal deck slabs and
vertical longitudinal webs and transverse diaphragms (Fig. 14).
The deck slabs are idealized by quadrilateral elements having
a total of 5 degrees of freedom (DOF) per node, 3 translations and
2 rotations (Fig. 13a, b). The in-plane action of the quadrilateral elements is represented by the plane stress mixed model
Q8Dll having 2 translational DOF at each external corner node and
3 internal DOF (Fig. l3a). The mixed model is constructed using
separate expansions for the displacement and strain fields.
The
variations of the u and V components of the displacement field are
approximated by the standard bi-linear expansion for the 8 corner
node DOF and by bi-quadratic expansions for 2 of the internal DOF.

137

FIG. 12

J,Y

BOX GIRDER BRIDGE OF GENERAL GEOMETRY

YI.k+

8Yll~

~Y.Y=CONST.

L -_ _ _....;Z"--__

la)

J,V

Plane Stress Q8D11

-t=i-

8z 1

u"-tv

8., w,
Id)

1,8x

Plate Bend1ng ONE\-;

~I""

u~
u = a [ 1-)
a

1.

I,U

Plane Stress Ql2Rl2

FIG. 13

r:

.,U

"~~
&

.,8x

Plate Bend1ng Q19

::I

f"2

J,Y

IG)

Ib)

Plane Stress QUSPl2

8z 1

'-----~_~

Ie)

U,

WI~

.,u

x2

21-)
a

~~ .L
I f)

I~)J
a

21~)3
a

Beam Functlons for 0

Zl

FINITE ELEMENTS USED IN CELL AND FINPLA2 PROGRAMS

31~) 2 + 1
a

138

PLAN VIEW

E!

ELEVATION

TOP DECK

II

II

,I

F=RTICAL WEB

BOTTON DECK/

SECTION

FIG.

14

A- A

BOX GIRDER BRIDGE OF CONSTANT DEPTH WITH VERTICAL WEBS

139
The third internal DOF is used to enforce a constant shear-strain
variation over the entire element, which produces a more flexible
and better element. After the element stiffness is formed the 3
internal DOF are eliminated by an internal static condensation
process. The quadrilateral plate bending element Q19 (Fig. 13b)
used for the deck slabs is a compatible element made up of four
subtriangles, each of which has 11 DOF associated with full cubic
expansions of the w-displacement field and an enforced linear
variation of the normal slope along one edge. In combining the
four sub-elements, a quadrilateral with 19 DOF is obtained. However,
the 7 internal DOF are eliminated by static condensation leaving the
essential 3 DOF at each corner node, 2 rotations and a translation
(Fig.13b).
The veritcal webs of the bridge are idealized by special
rectangular spar elements having a total of 5 DOF at each corner
node, 3 translations and 2 rotations. A single element over the
entire depth of the bridge can be used to capture the essential
behavior of the web. The in-plane action of these elements is
represented by the model QUSP12 (Fig. 13c). A bilinear expansion
for u and v is associated with the 2 translational DOF at each
node and a cubic variation in the x-direction of V is defined by
the rotation 6 z i = dV/dX at each node. The plate bending is
represented by a simple one-way bending element ONEW having 2 DOF
at each node, a rotation and a translation (Fig. 13d). Orthotropic
plate properties and arbitrary loadings and boundary conditions
can be treated in CELL. Automatic element and coordinate generation options which minimize the required input data, have made
CELL a widely used program.
6.2

Bridges of General Geometry with Arbitrary Integrated ThreeDimensional Frame

A structure of this type (Fig. 12) is made up of quadrilateral

two dimensional elements and one dimensional frame elements, each
of which is assumed to have 6 DOF at each node, 3 translations and
3 rotations. Although 5 DOF have often been used for general shell
programs, special techniques are required to account for the missing DOF, the rotation normal to the shell or plate surface. The
use of 6 DOF for the two dimensional elements eliminates this
complication and also makes the integration of the standard 6 DOF
per node frame elements into the system a simple matter.
The quadrilateral elements for both the decks and the web may
have an arbitrary orientation in space. They are taken as flat
plate elements which give the best least squares fit through the
actual location of the corner nodes. For box girder bridges of
general plan and elevation geometry this assumption should be quite
good. The special feature of the plane stress element, Q12R12, is
that in addition to the commonly used 2 translational DOF per node,

140
each node is also assigned a rotational DOF (Fig. 13e). This is
defined as the average rotation about the element z-axis. The
internal element displacements u and v are assumed to vary linearly
with the 2 translational DOF and as beam functions (Fig. 13f) with
the rotational DOF. The assumed nodal rotations introduce small
angular discontinuities so that the element is not fully compatible.
The plate bending model used
to that described in the preceding
program. The stiffness matrix for
frame elements can be found in any

for the quadrilaterals is identical

section (Fig. 13b) for the CELL
the standard one dimensional
textbook.

The above elements were first incorporated in 1967 into the

program FINPLA to analyze straight box girder bridges made up of
rectangular plate finite elements, reinforced by one dimensional
transverse or longitudinal beam elements, and supported by single
or mUltiple column rigid frame bent supports. The transverse centerline dimensions of the box girder system must be constant along the
span in FINPLA, however, the thickness and orthotropic material
properties of the rectangular plate elements may vary along the
width and length of the bridge. The program has been used extensively for box girder bridges and requires a minimum of input data to
define the structure and the loading.
In 1970 FINPLA was extended to analyze box girder bridges
of general geometry (Fig. 12) in a computer program of much greater
complexity called FINPLA 2. FINPLA 2 can analyze bridges of varying
width and depth made up of quadrilateral planar finite elements
(Fig. 13b, f) and one dimensional frame elements, so that almost
any three dimensional bridge system can be studied. The form of the
input data was developed primarily to simplify its use for box
girder bridges. The structure is discretized by dividing it
longitudinally into a number of segments by vertical sections and
by subdividing each segment into finite elements. The structure
alignment is described by a longitudinal reference line, which may
be a straight line, a circular curve or an arbitrary planar string
polygon and cross-sections are defined with respect to this line.
Because of its complexity FINPLA2 has not been used extensively and
thus complete checking, documentation and elimination of errors in
all options and applications of the program cannot be assured.

7.

ADVANTAGES AND DISADVANTAGES OF FINITE ELEMENT METHOD

Advantages of this method are:
1.

It is the most general method available and can

treat arbitrary loadings, boundary conditions,
varying material and dimensional properties and
cutouts.

141
3.

One dimensional frame type elements can be readily

incorporated as an integral part of the structural
system.

Disadvantages of this method are:

8.

1.

It requires a greater amount of computer time

than a direct stiffness harmonic analysis to
obtain a solution of comparable accuracy.

2.

A refined mesh size must be used to achieve

accurate results in the vicinity of steep stress
gradients.

3.

Static equilibrium is not automatically satisfied

for the displacement models normally used, but is
approached as the mesh size is refined. Judgement
must be used in selecting an appropriate mesh layout
and in interpreting the results.

4.

Automatic mesh and load generation schemes need to

be incorporated into the computer program used to
avoid the large manual input of data otherwise
required.

FINITE SEGMENT METHOD

The name finite segment method is often used to designate

solutions in which larger elements than those used in the classical
finite element method are employed. Typically, these larger
elements capture the dominant structural behavior and satisfy
equilibrium. A computer program SIMPLA (1967) [A2, B4J based on
this method has given excellent solutions for straight box girder
bridges with constant transverse centerline dimensions along the
span and arbitrary boundary and support conditions. The basic
structural element used in this method is formed by dividing each
longitudinal plate into a number of finite segments along the span.
These finite segments each have a width equal to the transverse
distance between the longitudinal joints of the plate. Thickness
and material properties of each finite segment may be differenL.
In the analysis, the finite segments are first interconnected
transversely at one end of the bridge to form a full transverse
segment of the entire bridge cross-section. The solution proceeds
by a segment progression method along the span to connect one
transverse segment to the next until the far end of the bridge is
reached. The boundary conditions at the two ends of the bridge
provide sufficient equations to determine all the unknowns needed
in the solution of the problem.

142
Each plate segment is assumed to obey the ordinary theory of
folded plates. Thus slab action is determined by the behavior of
one way slab strips spanning between longitudinal joints, so that
torsional and longitudinal slab moments are neglected.
In-plane
membrane action is defined by elementary beam theory applied to
individual finite segments which results in a linear valuation of
longitudinal stresses over the width of each segment between longitudinal joints.
Advantages of the finite segment method are:
1.

A reduced number of elements is needed to define

the structure and thus the storage and computer
time required for solution is reduced.

2.

The internal forces found satisfy equilibrium.

Disadvantages of this method are:

9.

1.

The segment progression solution used may be

numerically sensitive requiring so-called "stopovers" along the span in the solution.

2.

Because of the larger elements used, highly

localized stresses or steep stress gradients
cannot be captured.

NONLINEAR ANALYSIS

Present design procedures for reinforced and prestressed

concrete box girder bridges by either allowable stress or load
factor design methods are generally based on internal forces
determined from linear elastic analyses of simplified models of
uncracked homogeneous concrete systems. There is a need to develop
rational methods of nonlinear analysis from a fundamental viewpoint
which can accurately predict the response of these complex bridqe
systems under initial and time-dependent load histories, expecially
through their cracking, inelastic and ultimate ranges, so as to
determine their nonlinear behavior, ultimate strength and mechanisms
of failure.
An important problem has arisen with respect to the many
hundreds of existing box girder bridges. Many bridge agencies are
attemptin~ to determine the overload capacities of these bridges
because of the demand for permission to use heavier and heavier
loads on the highway system. These bridges were originally
designed using linear elastic analyses of simplified analytical
models. Their true ability to carry the heavier loads being
proposed can only accurately be determined by means of analyses

143

which take into account the dominant nonlinearities under overloads

up to failure.
In this presentation, the most recent research at the University
of California in the nonlinear analysis of reinforced and prestressed
concrete box girder bridges, which was completed in three separate
Ph.D. dissertations under the supervision of the writer, will be
reviewed. Each dealt with seperate bridge types as follows:
(1) Straight, Cast-In-Place, Reinforced or Prestressed
Concrete Box Girder Bridges.
(2) Curved, Segmentally Erected, Single Cell,
Prestressed Concrete Box Girder Bridges.
(3) Cast-In-Place, Multicell, Reinforced Concrete
Box Girder Bridges of Arbitrary Plan Geometry.
The analytical models, solution techniques and computer
programs developed as part of each of these extensive research
studies will be described briefly. Detailed information on these
studies may be found in the research reports and technical papers
given as references in Appendix 2.
9.1

Straight Cast-In-Place, Reinforced or Prestressed Concrete

Box Girder Bridges

The nonlinear analysis of bridges, which can be modeled as

planar frames composed of one dimensional elements, can be performed
using the computer program PCFRAME (1977) [A29, B22, B23, B26].
This program was developed as part of the Ph.D. dissertation
completed by Y.J. Kang [A29] in 1977.
The analytical model consists of a series of interconnected
one dimensional elements between joints lying in a plane (Fig. 15).
Each joint has three degrees of displacement freedom, two translational and one rotational. Each element is assumed to have a
prismatic cross section, which has an axis of symmetry, but the
shape of the cross section may differ from element to element. The
element cross section is divided into a descrete number of concrete
and steel layers, (Fig. 15). These layers are defined by their
cross-sectional area and distance from the element reference plane,
which contains the longitudinal axis of the member. Different
material properties in the layers due to cracking of the concrete
and nonlinear material stress-strain relationships are monitored
during the analysis. Element stiffnesses and properties are
evaluated by a layer integration procedure.
Pre-tensioned, post-tensioned bonded and unhonded prestressing
tendons each having a given profile, initial tensioning force and

144

(a)

Actual and Idealized Prestressed Concrete Frame

(b)

Element

REINFORCING STEEL LAYER

CorKRETE LAYER

---11lZlZl'8IID.~

---I'---'+'-~--...REFERENCE
PLANE rl~~~~PREST,RESSING ~
STEEL

(e)

(d)

(e)

Actual and Idealized Cross Section

A Prestressing Steel Segment in an Element

Elevation of an Element and Displacement Componenents

FIG. 15

ANALYTICAL MODEL FOR COMPUTER PROGRAM PCFRAME

145

and a constant cross-sectional area along its length can be

included in the analytical model. The actual curved prestressing
tendons is idealized (Fig. 15) by a series of segments each of
which is straight, spans an element and is assumed to have a
constant prestress force.
An efficient numerical procedure is used for the material
and geometric nonlinear analysis which includes the time-dependent
effects due to load history, temperature history, creep, shrinkage
and aging of the concrete and friction and relaxation of the prestress. The procedure is capable of predicting the response of
planar structures throughout their service load history as well as
throughout elastic, cracking, inelastic and ultimate load ranges.
A step forward integration is performed by dividing the time
domain into a discrete number of intervals for the quasi-static
time dependent analysis. For each time interval, nonlinear
equilibrium equations which are valid for the current geometry and
material properties are set up and solved by the finite element
method based on the displacement formulation. An incremental load
method combined with unbalanced load iterations for each load
increment is used for the solution of the nonlinear equilibrium
equations.
Time-dependent variation of concrete properties is recognized.
Concrete strain is assumed to consist of the mechanical component
and the non-mechanical component due to creep, shrinkage, aging
and temperature variations. Parabolic-linear, bilinear and multilinear approximations of the stress-strain curves are utilized for
concrete, reinforcing steel and prestressing steel, respectively.
A simple model for any inelastic load reversal is incorporated.
Creep strain is evaluated by an efficient numerical procedure
based on the age and temperature dependent integral formulation.
In the procedure, the history effect is incorporated by updating
the values of only two variables instead of the storage of all the
previous stress or strain histories.
Nonlinear creep effect at
high stress levels is also considered.
output from the computer program PCFRAME for each load or
time step includes the following:
(1) joint displacements and
rotations; (2) support reactions; (3) element axial forces and
moments; (4) strains, stresses and forces for each prestressing
steel segment; and (5) strains and stresses for each concrete and
reinforcing steel layer.
9.2

Curved, Segmentally Erected, Single Cell, Prestressed Concrete

Box Girder Bridges
Deflection and stress prediction are important in the design

146

of segmentally erected bridges. A method is presented for the

analysis of curved segmentally erected prestressed concrete box
girder bridges including time-dependent effects which can be
performed using the computer program SEGAN (1978) [A21, B19]. This
program was developed as part of a Ph.D. dissertation completed by
s. F. Van Zyl in 1978.
The bridge cross section must consist of a single box section
with vertical or inclined webs and can have cantilever flanges.
width and depth dimensions can vary along the bridge span (Fig. 16a).
The segments can be either precast or cast-in-place and can be
erected at various specified times.
The analytical model consists of a series of joints interconnected by a series of special straight one dimensional elements.
These are skew-ended finite elements (Fig. l6b) with eight degrees
of freedom at each of the two end nodes.
In addition to the usual
three translational and three rotational degrees of freedom, a
transverse distortional and a longitudinal warping degree of
freedom are used at each end. Because the element end cross section
need not be normal to the longitudinal axis, it can be made to
coincide with the end of the adjacent element meeting it on angle.
Diaphragm and support spring stiffnesses are added to the structure
stiffness matrix before solving for the displacements by the direct
stiffness method.
Time-dependent material changes such as creep and shrinkage of
concrete as well as environmental effects such as humidity and
temperature changes are accounted for.
However, other concrete
material nonlinearities such as cracking or inelastic behavior are
not included in the analysis. A super-position method where the
previous strain history is contained in a limited number of
variables is used in the creep analysis which is similar to that
used in PCFRAME.
Prestressing in the longitudinal direction can be included
in the analysis. Tendons can be located in either the webs or
flanges and can have a linear or parabolic profile. Stressing
operations encountered in segmental construction such as stress
and release operations, slip-in of tendons, prestressing and
removal of tendons are included. Prestress losses due to creep,
shrinkage and elastic deformations of the concrete as well as
friction and relaxation of prestressing steel are computed and the
effect included in the analysis.
A procedure is presented for specifying the structural
configuration and the erection sequence. Operations used in
segmental construction such as addition of segments, prestressing,
changing of support boundary conditions, can be analyzed. At each
construction stage the current structure is analyzed by the direct

(I)

-cr

1.--' ]

(I)

LLJ

FIG. 16a

('1)

('1)

\T

'--IT-'

(4)

.-T"T""I---' .-r---.-, ,-

(~)

LT

CURVED SEGMENTAL BRIDGE

Cross-S~ctlons

~)

-,;;::::r-

ElevatIons

~)

z-.----zz

G)

('5)

'J""

,~--.--.

i..-~=--l"7

Plan

.j>.

.....

148

-----';1:

a
Box Cross Sectlon

rC.

!&

\;

-x"

Plan

FIG. l(,b

Lc

SKEW ENDED FINITE ELEMENT FOR COMPUTER PROGRAN SEGAN

149

stiffness method and the incremental displacements, stresses and

strains are obtained. By superposition the total displacements,
stresses and strains are obtained. The complete structure also
can be analyzed for a variety of live loads and time steps.
9.3

Cast-in-place, Multicell, Reinforced Concrete Box Girder

Bridges of Arbitrary Plan Geometry

A numerical method of analysis has been developed by F. Seible,

in a Ph.D. dissertation [A28] completed in 1982, to trace the
nonlinear response of multicell reinforced concrete box girder
bridges under stepwise increasing static loads. This analysis
forms the basis for the computer program NOBOX (1982) [A28, B27,
B28]. Nonlinearities considered are material nonlinearities
inherent in reinforced concrete structural members under short
term loading, such as cracking of the concrete, yielding of the
reinforcement and formation of plastic hinges due to shear and
flexure.
The analytical model is developed for multi-cellular
structures of arbitrary plan geometry and constant height. The
analytical method can trace the complete nonlinear behavior of the
bridge beyond the working stress range all the way up to the
ultimate failure and collapse of the structure. Both collapse
mechanisms and failure loads can be determined.
A three-dimensional grillage model (Fig. 17a) comprised of
special elements, has been chosen in order to minimize the computational effort.
All the elements used in this displacement
model satisfy external and internal equilibrium at all times.
Individual primary actions in a multi-cellular structure such as
longitudinal bending and shear, transverse bending and torsion are,
with some engineering judgment, directly assigned to the individual
elements comprising the model which minimizes the number of DOF's
necessary to capture the complex cellular behavior. The structure
is modeled by longitudinal beam elements, transverse bending frames
and special torsional shear panels. The DOF's necessary for this
three-dimensional grillage model are shown in Fig. 17a. No axial
deformations are considered, i.e., all translational DOF's (Fig.
17a) are assumed to act uniformly over the height of the webs or
the width of the top and bottom slabs of the structure.
The longitudinal beam elements along the girder lines (Fig.
17b) represent the longitudinal bending and shear resistance in
the bridge structure. Transverse bending frames (Fig. l7c)
comprised of individual beam elements are arranged transversely
at discrete intervals (Fig. l7a) in order to model the transverse
bending in top and bottom slabs, and in the vertical webs, as well
as the transverse distortion in the box section. Geometric

150

.t
I

---

",~

, 2 '

"
~]
TRANSFORMATION TO GLOBAL COORDINATES
DIRlCTIOllS OF

1'IWISVERSE SECTIONS

--- LOCAL

-GLOBAL

(al

-?- t:t]
r t13i': ~ t 3
B

ELElIENT DOF' S

(bl

Displacement Degrees of Freedom

2"L 4
\J

(c)

FIG. 17

d'-----

'2

.~
I

---- LOCAL
_
GLOBAL

3 2

Beam Element

SHEAR
PANEL

TOP

C;

Transverse Bending Frame

(d)

Torsional Shear Panels

3-D GRILLAGE KlDEL AND STRUCTlJRAL ELEMENTS FOR COMPUTER PROGRAM NOBOX

151

compatibility between adjacent longitudinal beam elements or

adjacent transverse bending frames is established through shear
panels, which can only represent a uniform state of shear. These
are introduced between girder lines and bending frames at the top
and bottom slab levels (Fig. 17d). These shear panels are also
used to model the torsional properties of the box section by simply
requiring an equal and opposite shear flow in top and bottom slabs
at a certain section. Optional transverse diaphragms in the form of
transverse beam elements (Fig. 17a) can also be introduced in the
mid-plane of the structural model.
Material nonlinearities in the form of member force-deformation
relationships for reinforced concrete members can be characterized
from numerous experimental investigations by three major events:
(1) cracking of the concrete; (2) yielding of the steel reinforcement;
and (3) crushing of the concrete. These are approximated in each of
the three types of elements used by simple trilinear force-deformation
relationships.
Based on a mixed model formulation, an analytical scheme is
developed which can treat the perfectly plastic branch of the force
deformation characteristics by physically introducing mechanisms
into the analytical model. A simple separation of the force and
displacement unknowns in the mixed model leads to an analytical
scheme in which the nonlinear element states are considered by a
modification of the structural system, while the nonlinear
element history is accounted for in a corrected load vector. Plastic
deformations in the introduced mechanisms are readily available at
the element level from the mixed model formulation.
The discretization of the analytical model as a grillage,
consisting of structural elements which satisfy internal equilibrium and capture the dominant structural behavior, allows an
economical solution due to the minimization of the number of degrees
of freedom.
The computer program NOBOX can be used to give directly as
output the displacements, the longitudinal girder moments and
shears, transverse slab moments and torsional shear forces in the
top and bottom slabs at any stage of loading.

10.0

NUMERICAL EXAMPLES

In order to demonstrate, test and verify the analytical

solutions and computer programs developed for the linear and nonlinear analysis of concrete box girder bridges, a large number of
numerical studies have been performed. Complete details on these
studies for each of the computer programs developed may be found
in the pertinent research reports given as references in Appendix

152
A.2. In general, the numerical results from the computer programs
have been verified by one or more of the following ways:
(1) comparison with experimental results from physical model studies;
(2) comparison with numerical results from other analytical models;
and (3) performing statical checks.
Five representative numerical examples will be presented here
to illustrate some of the solutions which can be obtained using the
Berkeley computer programs. Additional numerical examples may be
found in the references given in Appendix A2.
10.1

Example 1 -- Small Scale Aluminum Model Studies

Extensive experimental studies have been conducted on a series

of small scale elastic aluminum models (Fig. 18) [A14, A15, A16,
BIO, B14]. All models had the same cross section (Fig. 18e) which
represented a 1:29 scale model or a typical two lane bridge in
California.
Instrumentation of the models made it possible to determine
experimental values of reactions, deflections, longitudinal
membrane forces and transverse plate bending moments (Fig. 2), which
could then be compared to theoretical values found using the computer
programs CELL for the straight and skew bridges and CURSTR for the
curved bridges. Each model was subjected to a series of vertical
point loads at various locations.
For the straight and skew bridges, ten models were tested. The
model reference numbers and plan geometries including spans and
angle of skew are shown in Fig. '18a. Starting with the largest
span rectangular bridge (Model OB), the model was cut back in five
stages to study two spans at 30 0 skew (Models 2 and 4), two spans
at 45 0 skew (Models 1 and 3), and a final short span rectangular
model (Model 5). Model references A and B were, respectively,
models tested without and with a midspan right transverse diaphragm.
All models had end diaphragms and had only four reaction supports
Rl ~o R4 at the corners (Fig. 18b).
For the curved bridges, six models were tested (Figs. 18c, d)
using a similar cutting back procedure. The radius of curvature
selected represented the sharpest curvature normally used in
California bridges.
Comparisons of theoretical and experimental results for all
bridges tested under a variety of load positions permit the
following conclusions:
(1) the computer programs CELL and CURSTR
accurately predict the magnitude and distribution of longitudinal
membrane ~orces Nx ' transverse plate bending moments My, and
reactions Rl to R4; (2) the transverse distribution of Nx at midspan,
for a given load position, has essentially a constant form of

0 Jill
~ L.-

"

-:"'-";===,R4

!;==;..;=..;::WEB::NO~--2::~:!.={WSV::ERSE::~IA;HRAGM------ ~~::- '

WEB NO. 1 :,

-,;-..;::c::::-:..-.-.-:.:.-.-.-.'ii/P-::::o;::::::::;::.-:.__-.-_-.::.:.:::_-.:..;

b)

Rl

r-~ I - - - - - . - - - - - - - - -

RJ

FIG. 18

d)

L _ _ _ _ _ _-1

0 AT 3" , '2"

0 49

I/B

--

-:.-i

/,

No-f--~-"- ' 1. /

STRAIN GAGE AXIS

F"'

Typical Curved Box Girder Bridge Model

110 DIAPHRAGM

I ~,'

Cross-sectional Dimensions of All Models

'" I-

S~AN

EXAMPLE 1 - DIMENSIONS OF SMALL SCALE ALUMINUM BOX GIRDER BRIDGE MODELS

e)

10"

--------l

Dimensions of Curved Models

60"

j~Fi~~~~:~:i~~tl~;~:;l.

c)

lLJCJ~B6pr~

Typical Skew Box Girder Bridge Model

STRAIN GAGE AXIS ---..OM, MIDSPAN

---.---------.------WEB NO.

Ci:~~~-:~~~~~~~~:~::;~~~~H[~::~~~~~~~::~:~;::?-END DIAPHRA~

-'

_R2

SPAN L - - - - - - 1

Dimensions of Straight and Skew Models

3"SKEW ANGLE

1"
J _

a)

I,

VI
W

154

0.60

a)

Skew Bridge Model 3A (Skew Angle

= 45,

-0.90

_ _ EXPERIMENT

-o.eo.

--<>_

0-.0..

in)

I>

.'

"0

..0"

= 35.5

J~"

THEo.RY

All
~

-0.70.

Span

~-O-

0.-0-

.0'.-

-0..40

-0.30
-0.20.
~

-0..10.
0.

w
&

w
&

Curved Model lA (Radius

= 116

lOOOOr lOOOOr lOOOOr

0..10.
0..20
0..30
o.Ao.
O.~O

0.60.
0..70
0.80

0..90

FIG. 19

b)

in, Span

= 60

in)

EXAMPLE 1 - TRANSVERSE DISTRIBUTION OF LONGITUDINAL MEMBRANE

FORCES N /W AT MIDSPAN
x

155
distribution that' is a function of span, skew angle or curvature;
and (3) the transverse distribution of My at midspan, for a given
load position, is essentially independent of span, skew angle or
curvature.
Typical theoretical and experimental results for the transverse
distribution of Nx at midspan for three midspan load positions are
presented in Fig. 19 for skew Model 3A having a skew angle = 45
and a span = 35.5 in. (902 mm) Fig. IBa) and for curved Model lA
having a radius of lIB in. and a span = 60 in. (1524 mm) (Fig. IBc).
Note the good agreement between theoretical and experimental values.
It can also be seen that the distribution curves for the two bridges
are eseentially parallel, as indicated in conclusion (2) above,
indicating that the departure from a uniform transverse distribution
is primarily a function of the local plate bending moments in the
bridge.
10.2

Example 2 -- Large Scale Reinforced Concrete Model Studies

Extensive studies of the structural behavior of three reinforced

concrete 1:2.B2 scale models of a typical two-lane bridge in
California having two 101.5 ft (31 m) spans with a single column
bent at the center have been performed at the University of
California. The elevations, plan and section dimensions of the
three models are shown in Fig. 20. All three had the same crosssectional and longitudinal span dimensions, but in plan one was
straight, one was curved and one was skew. Amounts of steel reinforcement used in the straight and curved models were somewhat
similar, but in the skew model the amount was considerably less.
The 72 ft. (21.1 m) long, 12 ft. (3.7 m) wide, and 1 ft.,
8-9/16 in. (0.52 m) deep models shown in Fig. 20 were built and
tested in the Structural Engineering Laboratory of the University
of California, Berkeley, using standard high-strength [60 ksi
(410 MN/m2) yield] deformed bars as reinforcement and concrete with
a 3/8 in. (10 mm) aggregate instead of a mortar mix as the model
material.
A detailed presentation of the investigation on each of these
models including their description, instrumentations, construction,
automatic data recording and reduction, test programs and comparisons of theoretical and experimental results may be found in a
series of publications listed in Appendix A2. The experimental
program was divided into two parts to study the response of the
bridges under: (1) dead load and working loads; and (2) overloads
including loading to failure.
The straight bridge was loaded to failure in 1970 [A9, AlO,
All, B7, Bll, B12, B13], the curved bridge in 1973 [A17, A18, A19,
B17, B18, B20] and the skew bridge in 1979 [A22, A23, A24, A25,

156

A X

II

WESTI

Y 0

lEAST

~======~F=======~========~=F======~
SPAN I

SPAN

n
3600

GI~ER

A X

-r
~
I
~) r, f-t

1442

~DIAPHRAGM+358
\300 30e.
1500

1
1

1500

T
I

358'

-r
I

1442

}
~~~==~~~~~~~~~~
_~~~~__~~~__~~~~~~__~~__~~~__~~___ ~

--~-----+~~-----4~----~-~!~
04MENSIONS

FIG. 20a

EXAMPLE 2 -

IN FT (1 FT .0305 m )

ELEVATION AND PLAN DIMENSIONS OF STI<AI('HT, CURVED AND SKEW

BRIDGE MODELS

35/8"

~.I

FIG. 20b

5'-0"

II

i
L.

"

2'-6114"

FT.

IN.

=25.4 mm
=0.305 m

4' - 6" SQ. FOOTING

EXAMPLE 2 - TYPICAL SECTION OF STRAIGHT, CURVED AND SKEW BRIDGE MODELS

L,

l' -8"

l' -6"
ROUND COLUMN

iOiOI

12' - O
4(@2'-6 7/8"=10-3/1/2"

v.

-"

158
B21, B24, B25]. Extensive correlative studies of experimental
vAlues for reactions, deflections, internal forces and moments with
theoretical values predicted by various computer programs are
presented in the above references.
Linear elastic analyses were carried out for the straight
bridge using FINPLA2 and for the curved and skew bridges using
CELL. Only a typical comparison of theoretical results for the
three bridge models [A23] is given in Table 1 for midspan deflections
and in Table 2 for longitudinal membrane forces Nx at Section A
near midspan, for midspan point loads of 100 kips (445 kN)
The transverse distribution of deflections at midspan sections
X and Y given in Table 1 indicates that for point loads on the
longitudinal centerline at 3X + 3Y the skew bridge deflections are
generally smaller than the corresponding deflections for the
straight and the curved bridge. For a point load on the acute
side of the span the skew bridge deflections at the loaded point
are higher than for the straight or curved bridge, while for
loads o~ the obtuse side of the bridge the opposite is the case.
From the transverse distribution of longitudinal membrane
forces Nx at Section A given in Table 2, it can be observed that
for loads along center girder 3, the membrane forces in the
longitudinal direction in the skew bridge are of the same order
of magnitude as for the straight and the curved bridges. It is
interesting to note that higher membrane forces exist at Section
A on the acute side of the span than on the obtuse side for the
skew bridge. For loads along girder 1 the membrane forces Nx at
Section A are considerably higher for the skew bridge than the
straight or the curved bridge models, while for loads along girder
5 the membrane forces at Section A of the skew bridge are much
smaller than in the corresponding straight and curved bridge cases.
This emphasizes the sensitivity of the skew bridge to transverse
load position as compared to the straight or curved bridge.
Nonlinear analyses were also carried out for the straight and
curved bridge models using NOBOX [A28, B27, B28). For both box
girder bridge models a longitudinal collapse load analysis was
performed first in order to obtain an upper bound on the ultimate
collapse load and an indication of the ultimate collapse mechanism.
A complete nonlinear analysis was then done for both examples and
results were compared with the experimental findings. Finally
eccentric midspan point loads were applied in both cases as
opposed to the more uniform midspan loadings used for the
experimental ultimate load test, and again longitudinal collapse
load analyses and complete nonlinear analyses were performed.
Only typical results for load vs. midspan deflections are
presented in Fig. 21 and 22 for the curved bridge model. Fig. 21

BRIDGE
MODEL

2
3

GIRDERS

SECTION X
4

2
3

GIRDERS

SECTION Y
4

STRAIGHT 5.86 5.96 6.10 5.96 5.86 -1.98 -1.97 -1.97 -1. 97 -1.98
CURVED 5.25 5.61 6.01 6.13 6.31 -1.84 -1.91 -1.97 -2.06 -2.16
6.81 6.27 5.77 4.66 3.58 -.99 -1.48 -1.99 -2.51 -3.03
SKEW

STRAIGHT 4.17 4.94 5.86 6.97 8.30 -2.55 -2.28 -1.99 -1.71 -1.44
CURVED 4.06 5.11 6.31 7.73 9.36 -2.55 -2.37 -2.18 -2.02 -1.87
2.59 3.01 3.58 4.30 5.90 -1.24 -1.12 -1.02 -.91 -.78
SKEW

3X

5X

IIFT~O.305ml

STRAIGHT 8.29 6.97 5.85 4.93 4.17 -1.43 -1.71 -1.99 -2.28 -2.55
CURVED 7.14 6.11 5.25 4.59 4.06 -1.16 -1.49 -1.81 -2.15 -2.50
SKEW 12.46 9.18 6.80 4.60 2.56 -.76 -1.88 -3.02 -4.1B -5.37

IX

STRAIGHT 1.61 2.68 3.88 5.28 6.86 1.36 2.25 3.50 5.30 8.00
5X + 5Y CURVED 1.56 2.78 4.15 5.72 7.49 1.27 2.30 3.69 5.63 8.46
SKEW -2.82 -1.18 .55 2.42 5.12 1.10 ~.16 5.55 8.50 12.33

STRAIGHT 3.86 3.98 4.13 3.98 3.86 3.50 3.89 4.90 3.89 3.50
3X + 3V CURVED 3.44 3.72 4.04 4.07 4.12 3.15 3.67 4.81 3.95 3.71
SKEW
3.78 3.77 3.78 3.16 2.56 2.36 3.06 4.36 3.65 3.55

AX

YD

YD

,<

II I II I;

AX

/1 II I II

~I

COMPARISON OF TRANSVERSE DISTRIBUTION OF DEFLECTIONS (FT x 10- 2 ) AT tnoSPAN

SECTIONS X AND V FOR STRAIGHT. CURVED AND SKEW BRIDGE MODELS

STRAIGHT 6.86 5.27 3.88 3.59 1.61 8.00 5.30 3.50 2.25 1.36
1X + lY CURVED 5.79 4.61 3.41 2.40 1.50 4.77 4.77 3.18 2.01 1.33
SKEW 11.69 8.30 5.81 3.99 1.32 5.75 2.63 0.33 -1.48 -3.04

LOAD
CASE

TABLE 1

'-0

UI

5X+ 5Y

3X+ 3Y

1X+ 1Y

GIRDER

12
3

5
7
9

11

23.2 26.1 32.1 39.9 47.5

26.8 27.6 32.0 38.0 43.3
-6.0 7.9 8.1 21.9 42.9

WEB

-20.9 -20.9 -22.5 -24.5 -25.6

-19.6 -21.2 -24.6 -28.2 -29.3
-5.8 -7.4 -6.7 -16.6 -29.0

STRAIGHT -23.2 -21.6 -20.7 -22.1 -24.1 -26.2 -28.1 15.6 17.6 21. 6 27.0 32.3
CURVED -24.3'-19.7 -18.0 -20.7 -23.8 -25.9 -37.3 18.1 18.6 21. 5 25.6 29.3
SKEW
-6.5 -5.2 -5.6 -5.3 -12.7 -26.0 -34.2 4.0 5.5 5.9 15.1 28.6

PLATE

STRAIGHT
CURVED
SKEW

STRAIGHT
CURVED
SKEW

31.8 32.2 33.5 33.2 31.8

33.4 34.0 34.2 33.6 31.1
39.4 37.9 30.7 29.2 25.7

47.5 39.9 32.1 22.8 21.7

48.0 39.5 32.2 27.0 25.4
52.3 57.8 52.9 41.8 40.8

WEB

-21.4 -22.9 -23.6 -22.9 -21.4

-23.9 -25.6 -26.4 -25.3 -22.9
-28.5 -31.4 -26.1 -24.1 -20.7

-25.6 -24.5 -22.5 -20.9 -20.9

-32.0 -29.1 -24.6 -21.2 -19.3
-41. 7 -46.2 -44.1 -34.7 -31.5

STRAIGHT -21.6 -22.1 -23.0 -23.8 -23.0 -22.1 -21.6 21.7 22.8 23.1 22.8 21.7
CURVED -27.2 -24.2 -20.1 -23.0 -21.8 -23.1 -24.7 22.8 23.3' 23.5 23.0 21.2
SKEW -30.1 -25.4 -24.0 -20.4 -18.7 -18.6 -19.2 26.4 25.7 21. 2 20.1 17.4

10

BOTTOM NODES

1FT :O.305m

lIKIP: 4.448 kNJ

PLATE

II

STRAIGHT
CURVED
SKEW

(!)

WEB

TOP NOOES

G>

STRAIGHT -2B.1 -26.2 -24.1 -22.1 -20.7 -21.6 -23.2 32.3 27.0 21.6 17.6 15.6
CURVED 41.6 -32.3 -24.7 -20.8 -18.0 -19.4 -22.5 32.5 26.7 21.8 18.6 17.2
SKEW -42.2 -37.2 -34.2 -32.5 -25.7 -28.0 -33.4 34.8 38.1 34.6 27.6 26.1

<D

81012

PLATE

24

-1:1:1:1:1-

NODE I

COMPARISON OF TRANSVERSE DISTRIBUTION OF LONGITUDINAL MEMBRANE FORCES

Nx (KIPS/FT) AT SECTION A FOR STRAIGHT CURVED AND SKEW BRIDGE MODELS

LOAD
BRIDGE
CASE LOCATION MODEL

TABLE 2

161

compares theoretical values from NOBOX with experimental values

obtained in the test of the bridge to failure during which three
increasingly equal point loads on girders 2, 3 and 4 were applied
at both midspans X and Y in addition to the dead load. Very close
agreement can be observed between theoretical and experimental
results. Also note the smaller deflections at the inner girder
IX as compared to the outer girder 5X, under this loading.
'Fig. 22 compares only the theoretical values from NOBOX of
midspan deflections for three different point loadings, symmetrically
placed longitudinally at the two midspan sections X and Y:
(1) the
actual failure loading used in the test program of point loads on
girders 2, 3 and 4 (note that a different horizontal scale has been
used for the deflection in Fig. 22 compared to that in Fig. 21);
(2) an eccentric point load on inner girder 1; and (3) an eccentric
point load on outer girder 5. A comparison of the midspan deflection under the exterior girders for all three load cases for which
a complete nonlinear analysis was performed (Fig. 22) shows the
enormous influence of the transverse stiffness deterioration in
the bridge model. Not only is the ultimate load level and the
ultimate deflection for the eccentric loadings drastically
different from the more uniform experimental failure loading in
the curved bridge model, but the eccentric loading cases feature
also a torsional rotation of the bridge model at midspan, i.e.,
a negative deflection or uplift of the unloaded side of the span.
This uplift would not occur without the transverse stiffness
deterioration, since the downward loading would be distributed
transversely all the way to the other side of the span. This can
be seen for the lower load levels in Fig. 22 and also from the
results of the linear analyses given in Tables 1 and 2.
10.3

Example 3 -- Two Cell Aribtrary Quadrilateral Box Girder

A linear analysis of a single span, two cell box girder of

arbitrary plan geometry with diaphragms at the ends only (Fig. 23a)
was performed using the grillage model of NOBOX (termed BOX for a
linear analysis) and the finite element model of CELL. The same
discretization (Fig. 23a) was used in both cases. The analysis
was performed for an eccentric point load of I kip (4.448 kN) on
girder 3 at Section 6 rather than at midspan Section 7 in order to
avoid possible disturbances at the point of load application in
the finite element model.
The vertical deflections ~f the three girder lines obtained
by NOBOX and CELL (Fig. 23b) are in close agreement. A similar
conclusion can be reached for the transverse bending moments in
the top slab at midspan Section 7 (Fig. 24a) and the percentage
of the total midspan longitudinal moment distributed to each of
the three girders (Fig. 24b).

162

(KIP)

180

.... ,.;=g

160
:z

""

CVI

m
~

140

C-

120

c>:

:z
CVI

'"
w 100
C-

80

""-'
0

EXPERIMENT

'"w

NOBOX

( KIP)
176 K
173 K

160
140
120

C-

c>:
0

60

D
W

180

-'

EXPERIMENT
NOBOX

Cl

u.J

-'

-'
C-

CC-

C-

c>:

c(

__~__~--~--~
0.5 1.0 1.5 2.0 2.5
DEFLECTION AT IX (IN.)

O~--~--

FIG. 21

0.5 1.0 1.5 2.0 2.5

DEFLECTION AT 5X (IN.)

EXAMPLE 2 - MIDSPAN DEFLECTIONS AT EXTERIOR GIRDERS OF CURVED BRIDGE

DURING FAILURE LOADING

(KJ P)
mDSPAN
LO.~D 1BO

mDSPM
( KIP)
LO.~D 180

160

.~
"
w

140

'-

l!,:..

120
100

LOAD AT

2X,3X,4X

160
140
120
100

LOAD AT
5X

LOAD AT
5X

LOAD AT
lX

20
2

MIDSPAN DEFLECTION AT IX
FIG. 22

00

(IN.)

"1IDSPAN DEFLECTION AT 5X

(I N. )

EXAMPLE 2 - THEORETICAL MIDSPAN DEFLECTIONS AT EXTERIOR GIRDERS OF

CURVED BRIDGE UNDER DIFFERENT LOADINGS

163

10

0.667'

-1t

20

30

40

: O. 542'

II

50

60

1K1P

0.548',

~}.O'

'>

70 (FT.)

432000 KSF
r,

188000 KSF

9.0,L
t--9.0'
(!)
GIRDER <D
/
'-- (!)
FOR SECTION 7
(a)

Dimensions and Discretization

SECTION 1

GIRDER

il'

CELL

BOX

(b)
FIG. 23

Girder Deflections

EXAMPLE 3 - DIMENSIONS, DISCRETIZATION AND DEFLECTIQNS OF ARBITRARY

QUADRILATERlIL niO-CELL BOX GIRDER BRIDGE

164

-4
-2

CELL
~--- ...

BOX

2
4

My (FT x KIP/FT x lO-2)

(a)

Transverse Bendlng MOments in Top Slab

O'I~
C'.

>
M

0 CELL
SECTION
GIRDEll ,
(b)
FIG. 24

BOX

Percentage Distribution of Girder MOments

EXAMPLE 3 - TRANSVERSE SLAB MOMENTS AND LONGITUDINAL GIRDER MOMENTS

AT MIDSPAN SECTION 7

165
A large difference exists, however, in the computer time
necessary for the solution by the grillage model of NOBOX as compared
to the finite element model of CELL. CELL requires about six times
the central processor time required by NOBOX.
In addition, it
should be noted that NOBOX yields directly design quantities such
as girder moments and shears, which have to be calculated in an
additional step from the internal nodal forces when CELL is used.
10.4

Example 4
Single Cell, Curved, Two Span, Prestressed Box
Girder Model

Linear analyses due to external loading; and prestress,

including friction, of a curved bridge model (Fig. 25, 26), have
been performed using the skewed finite segment model of SEGAN
[A21, B19] and the finite strip model of CURDIP [A6]. The model
was constructed by Hood of New Zealand [A2l], who also conducted
an extensive experimental and analytical investigation of its
behavior.
It was made of an epoxy resin. The cross section of the
actual experimental model as well as those assumed in the analyses
by CURDIP and SEGAN are shown in Fig. 26. The two span bridge had
vertical supports at the two ends and at its center (Fig. 25a).
It
had single prestressing tendons in each web with the profile shown
in Fig. 25b. The tendon geometry formed part of the input into the
computer analyses.
Only typical analytical and experimental results for the case
of prestress alone are presented in Fig. 27. Friction along the
tendon was accounted for in the analyses. Fig. 27a depicts the
longitudinal (tangential) stress distribution at a midspan section
between the end and center support. The stresses shown are membrane
stresses, therefore the average through the thickness of each plate.
The agreement of the analytical results by CURDIP and SEGAN with
those found experimentally are quite good. A check of the statical
moment at the cross-section reveals that the internal moment is
within 2 to 3 percent of the external moment, which is quite
satisfactory.
In Fig. 27b, the transverse (radial) stresses at the outer
and inner faces of the top slab at the midspan section are plotted.
Note that because of the analytical models used (Fig. 26) only two
plotted points are available from SEGAN, while five are available
from CURDIP. The agreement between analytical and experimental
results is satisfactory.
10.5

Example 5 -- Single Cell, Straight, Three Span, Prestressed

Box Girder Bridge

A nonlinear material and time dependent analysis has been

applied to a three span, continuous, post-tensioned, single cell,

166

FIG. 25"

EXAMPLE 4 - PLAN OF CURVED, TWO SPAN, PRESTRESSED SINGLE CELL

BOX GIRDER MODEL TESTED BY HOOD

~1 f1+U"11111111
~tmml
.

CI>

8 0 $rI
~

; ..

:n-P"

CD

c i

FIG. 25b

~I~
~ ~

!:l P:\ ~
N
~ r:

rII

S~ ~ ~
.'S r,; i h =,
jII

18~

~ !:!

!!!

l!

~ ~
!!? ~

..

lIS
N

~, ~, ~ , ~, ~ ~ ~
~ r;, ,..

\!l !!! ~ ~ ~

rJI :!' P (iii

~ ~ ~ ~ ii>

...r-ill

"S 'S,

r;;

~ ~
y.;, .'
(VI

Po

EXAMPLE 4 - PRESTRESSING LAYOUT FOR SINGLE TENDON IN EACH WEB

167

r._'7~
(a)

Actual Experimental Model

",

'P

(b)

F1nite Strip Analysis by CURDIP

lOS

(e)
FIG. 26

Finite Element Analysis by SEGAN

EXAMPLE 4 - CROSS-SECTIONAL DIMENSIONS OF ACTUAL MODEL AND

THOSE USED IN ANALYSES (mm)

168

- - - EXPERII1ENT

-1.0

Go --oi) CURDIP

A,MEGAN

_1.5 ..._ ....... - . - _.....

cr (MPa)
(a)

Longitud1nal (Tangential) Membrane Stresses (MPa) at Middle Plane of Plates

ELEMENT

33

y! / p,,-'p
I

0.4 cr (MPa)
0.3

pI, ""-

0.2

0.1

2224 N

-0.1
-0.2
-0.3
-0.4
(b)

Transverse (Radial) Stresses (MPa) at OUter and Inner Surface of Top Slab

FIG. 27

EXAMPLE 4 - COMPARISON OF' ANALYTICAL (CURDIP AND SEGAN) AND EXPERIMENTAL

RESULTS AT CENTER OF ELEMENT 33 DUE TO PRESTRESSING ALONE
(P = 2224N IN EACH WEB)

169

concrete box girder bridge (Fig. 28) using the computer program
PCFRAME [A29, B23] in order to determine its response to time
dependent effects and to increasing truck overloads up to failure.
The cross section (Fig. 28), with a two lane roadway width, is
assumed constant over the entire length of the bridge.
It is
assumed that simple support conditions exist at the ends of the
structure and vertical bearing supports exist at the two interior
supports.
The cast-in-place, post-tensioned bridge is designed using
California highway design criteria. One prestressing (PS) posttensioned tendon in each vertical web is assumed to have the
profile shown in Fig. 28. Two bridges are analyzed for comparison,
one without and one with additional longitudinal mild reinforcing
steel (RS), uniformly distributed over the cross-section. This
steel, not required for strength, is provided for construction
purposes.
For PCFRAME analysis the cross section is discretized into
ten concrete layers, two for each of the top and bottom slabs and
six for the webs. Where included the mild RS is modeled by three
steel layers, one in each of the top and bottom slabs and one at
the midheight of the web.
The time dependent analysis traces the response of the bridge
under dead load only from its initial prestressing at 28 days
through a period of 10,000 days (27 years). Time dependent
variations of deflections, curvatures, moments, and stresses as
well as prestressing forces are traced. Only typical results for
the vertical deflection (Fig. 29a) and the prestressing tendon
stress (Fig. 29b) versus time are presented here.
Notably evident are the smaller deflections for the bridge
with, as compared to without, mild RS (Fig. 29a). The ratios of
the dead load deflection after 10,000 days to the initial deflection
at 28 days are about 3.0 and 4.2 for the two cases. As expected
the prestressing tendon stress decreases with time (Fig. 29b) due
to creep and shrinkage of the concrete and relaxation of the stress
in the prestressing steel. Most of the prestress loss occurs within
the first 100 days. The analysis by PCFRAME predicts considerably
smaller losses than code values, probably because the latter are a
conservative estimate for design.
In order to study the response of the bridge to increasinq
load levels up to ultimate failure, a nonlinear material and time
dependent analysis was performed with PCFRAME in which the bridge
was loaded to failure using increasing levels of truck live loads.
The truck live loads were applied at two different ages after the
initial application of dead load plus prestress, first at an early
age of 30 days and second at a much later age of 10,000 days (27

48.8
I!I

AI_

II

6
1 m

-A.71'T"

48 8

1.1..

J..

10.4
I!I

..

_I

FIG. 28

.ld:.

EXAMPLE 5 - DIMENSIONS OF THREE SPAN PRESTRESSED CONCRETE BOX GIRDER BRIDGE

5.5m

ro'"Pf:~:": )0"
~
I

25 em

l~oor~

..

f'

::;

171

14
E
u

...>3
5o

--'

UJ

I-

c::

...

l-

e<

.L

UJ

:>

::0....

:a:

2
O~----__----~~----~

o~----~----~----~~
10
100
1000
10000

10

TIME IN DAYS

(al

Vertical Displacement

~v

(bl

VS. T~e

'"

I-

u.

10000

IN !lAYS

a vs. Time

Tendon Stress

6
a

0-

""-

1000

Tl~'E

EXAMPLE 5 - PCFRAME RESULTS FOR TIME DEPENDENT ANALYSIS

FIG. 29

0-

100

0-

;:;:-5

e<

~ 4

~
o 3

c::

C3
--'

--'

.JL

:x

!lL!

6V

O~~--~--~~--~~
o 0.2 0.4 0.6 0.8 1.0 1.2
DISPLACEt~ENT

(al

BV, em

Vertical Displacement
vs. OVerload P/PO
FIG. 30

6v

12

14

16

TENDON STRESS

(bl

18

If, MPa

Tendon Stress fI
VS. OVerload PlPo

EXAMPLE 5 - PCFRAME RESULTS FOR OVERLOAD ANALYSIS TO FAILURE

20

172
years) .
The truck live load consisted of an overload vehicle requiring
a special permit to be used on California highways. Total length
of the vehicle is 108 ft (33m), total weight Po= 107 tons (952 kN),
which is carried on 13 axles. In the overload analysis one truck
only is positioned in the middle of the center span of the bridge,
and the structural load vector representing Po is incremented until
ultimate failure occurs. Only the bridge without mild RS was
analyzed since its contribution to the ultimate capacity is small.
In general, overload response of the bridge at the two loading
ages proved to be quite similar. Typical results are presented for
the vertical deflection (Fig. 30a) for both loading ages and for the
prestressing tendon stress (Fig. 30b) loaded only at an age of 30
days, since loading at 10,000 days results in similar curves, but
shifted slightly.
The load-deflection curves (Fig. 30a) show that the structure
is slightly stiffer when loaded at 30 days than when loaded at
10,000 days. This is a result of the reduction in prestressing
force, due to time dependent effects, and the decrease in the
cracking load of the bridge. The large overload capacity of the
bridge is evident with ultimate failure occuring under dead load
plus 7.2 times the live load due to one vehicle.
It can also be
seen in Fig. 30a that the ultimate load does not vary significantly
with age at loading.
Fig. 30b plots tendon stresses at two critical locations
along the girder under increasing vehicle overload. Response is
shown only for loading at the age of 30 days.
The initiation of
cracking at the critical locations is readily identified in this
figure. After dead load plus three times the load of one vehicle,
cracking is initiated at the center of the main span, resulting in
a large increase in tendon stress at that location. After dead load
plus five times the load of one vehicle, cracking is initiated over
the interior support, resulting in a similarly large increase in
tendon stress at that location, and a further increase in tendon
stress at center span due to a redistribution of moment. Tendon
stress continues to increase under increasing overload until ultimate
failure of ,the bridge occurs due to rupture of the tendons at the
center of the main span.

11.0

SUMMARY AND CONCLUSIONS

A review of the analytical methods and computer programs which

have been developed at th;University of California at Berkeley,
for the linear and nonlinear analysis of concrete box girder
bridges has been presented. Short descriptions of 15 different

173

computer programs and pertinent references, where details on their

theoretical basis, input-output capabilities and numerous numerical
examples can be found, are given in Appendix 1 and 2.
In the present paper, five numerical examples were presented
to illustrate the wide range of problems, which can be solved and
the type of output results obtained.
It can be concluded that the available cOMputer programs
described are valuable computational aids for the analysis of single
or multi-cell, reinforced or prestressed, concrete box girder
bridges which are straight, curved, skew or of arbitrary geometry
in plan.

12.0

ACKNOWLEDGEMENTS

Many graduate students, listed as authors or co-authors in the

publications referenced in Appendix 2, participated extensively in
the development of the computer programs listed in Appendix 1.
Their excellent contributions to this research effort are gratefully
acknowledged.
Most of the computer programs for the linear analysis of box
girder bridges were developed as part of a continuing research
program sponsored by the Division of Highways, Department of
Transportation, State of California and the Federal Highway
Administration, united States Department of Transportation.
Much of the work on the computer programs for the nonlinear
analysis of box girder bridges was done as part of a continuing
research program sponsored by the National Science Foundation.
The support of these sponsoring agencies is acknowledged
with many thanks. However, the content of this paper reflects
the views of the author and not necessarily the sponsors.

174

APPENDIX 1.

COMPUTER PROGRAMS FOR BOX GIRDER BRIDGES

This Appendix 1 contains a list of the computer programs

developed at the University of California since 1965 for the
analysis of single or multi-cell box girder bridges. All of the
programs, except the last three (SEGAN, PCFRAME, NOBOX), are based
on linear elastic analyses of uncracked homogeneous structures.
The last three are nonlinear analysis programs, which trace the
response of bridge systems under various nonlinear effects.
The programs are grouped in the list under the following
methods of analysis:
A.
B.
C.
D.
E.

Folded Plate Elasticity Method

Finite Strip Method
Finite Element Method
Finite Segment Method
Nonlinear Analysis Method

For each computer program a short computer program name of

four to seven letters is given followed by a brief description of
the program. Figures in parentheses indicate the year the program
was first used and figures in brackets indicate references in
Appendix 2 where detailed descriptions of the programs and their
theoretical development may be found.
Anyone interested in any of the programs listed should first
read the detailed references indicated and then if they wish to
obtain a copy of the program itself, they should write for further
information to:
Professor A. C. Scordelis
Department of Civil Engineering
729 Davis Hall
UNIVERSITY OF CALIFORNIA
Berkeley, California 94720

175

COMPUTER PROGRAMS FOR THE ANALYSIS OF BOX GIRDER BRIDGES DEVELOPED

AT THE UNIVERSITY OF CALIFORNIA, BERKELEY, UNDER THE SUPERVISION
OF PROFESSOR A. C. SCORDELIS
Note:

A.

All of the computer programs listed below are written in

FORTRAN IV language.

FOLDED PLATE ELASTICITY METHOD

MULTPL Program (1965)

[AI, Bl, B2, B6]

This program provides a rapid solution for open or cellular

folded plate structures having single simple spans. A direct
stiffness harmonic analysis is performed in the solution. The
elasticity theory is used for the isotropic plate elements.
Uniform or partial surface loads as well as line loads and
concentrated loads may be applied anywhere on the structure. Up
to 100 non-zero terms of the appropriate Fourier series may be
used to approximate the loading. Internal forces, moments and
displacements in each plate element at any desired point may be
found.
MUPDI Program (1966)

[AI, B3, B6]

This program extends the MULTPL program to continuous span

bridges and can be utilized to analyze open or cellular folded
plate structures simply supported at the two ends and having up to
four interior rigid diaphragms or supports between the two ends. A
direct stiffness harmonic analysis based on the elasticity theory
is used for the folded plate system. Compatibility at the interior
rigid diaphragms or supports is accomplished by a force method of
analysis.
Loads and redundant forces may be approximated by up
to 100 non-zero terms of the appropriate Fourier series.
MUPDI3 Program (1971)

[A13, B6, B16]

This program extends the original MUPDI program such that up

to twelve interior diaphragms or supports may be used and they
need no longer be rigid. Diaphragms may be defined by flexible
beams and supports may be defined by two dimensional planar rigid
frame bents. Options permit evaluation of internal forces and
moments in the diaphragms and bents as well as in the plate
elements of the bridge.
B.

FINITE STRIP METHOD

MULSTR Program (1970)

[A4, B5, B6, B16]

This program is capable of analyzing straight prismatic folded

plates made up of orthotropic plate elements with eccentric stiffen-

176
ers. The structures must be simply supported at its two ends.
Each plate element is idealized by a number of longitudinal finite
strips in which the properties of the longitudinal and transverse
stiffeners are distributed uniformly over the area of each strip
and are accounted for in the analysis. The finite strip method
is used to determine the strip stiffness. The displacement
patterns are assumed to vary as harmonics longitudinally.
In the
transverse direction, a linear variation of the in-plane displacements and a cubic variation of the normal displacements are chosen.
A direct stiffness harmonic analysis is used to analyze the
assembled structure.
CURSTR Program (1970)

[A6, B6, B8, B14, B16]

This program is capable of analyzing prismatic folded plate

structures which are circular in plan and made up of orthotropic
plate elements. The structure must be simply supported by radial
diaphragms at its two ends. Each plate element, which may be a
segment from a general cone, a cylindrical shell, or a circular
ring plate, is idealized by a number of circumferential finite
strips. The finite strip method is used to determine the strip
stiffness. The displacement patterns are assumed to vary as
harmonics in a circumferential direction. In the transverse
direction, a linear variation of the in-plane displacements and
a cubic variation of the normal displacements are chosen. A direct
stiffness harmonic analysis is used to analyze the assembled
structure.
CURDI Program (1974)

[A20]

This program extends the CURSTR program to continuous span

bridges and can be utilized to analyze circularly curved structures
with up to twelve interior radial diaphragms or supports. Diaphragms
may be defined by flexible beams and supports may be defined by two
dimensional planar rigid frame bents. A direct stiffness harmonic
analysis is used for the folded plate systems. Compatibility at the
interior diaphragms or supporting bents is accomplished by a force
method of analysis.
Loads and redundant forces may be approximated by up to 100
non-zero circular harmonics of the appropriate Fourier series.
Internal forces, moments and displacements in the folded plate
elements as well as the one dimensional rigid frame elements may be
found at any desired point.
CURDIP Program (1980)

[A26]

This program extends the CURDI program to automatically

include prestressing as well as external loads. Up to ten posttensioned longitudinal prestressing tendons of general three

177

dimensional profiles may be specified with a minimum of input data.

Longitudinal tendons may exist in the top or bottom slabs and the
webs of the bridge. Friction between the prestressing tendons and
the ducts is accounted for and jacking may be specified from one or
both ends of the tendon. Equivalent loads due to prestressing are
calculated automatically in the program using vector algebra.
COSLAB Program (1980)

[A27]

This program provides a rapid solution for circularly

curved orthotropic slabs s~mply supported at the two ends and
having up to twelve interior column supports. Uniform or partial
surface loads, line loads and concentrated loads may be applied
anywhere on the structure. Up to ten longitudinally post-tensioned
prestressing tendons with profiles over the depth of the slab made
up of parabolic and straight portions may be defined with a
minimum of input data.
Resulting joint displacements and the
internal forces, moments and displacements in the finite strip
elements are calculated.
A direct stiffness harmonic analysis is used for the slab
system. Compatibility at the column supports is accomplished by a
force method of analysis. The finite strip method is used to
evaluate plate edge forces, stiffnesses and final internal forces,
moments and displacements. Fourier series are used for the loads.
Geometry and tendon forces are given as input for the prestressing
tendons, and equivalent loads are computed. Friction between
tendon and duct is neglected. Column supports are approximated as
vertical one-dimensional springs, thus neglecting the influence of
the bending properties of the columns.
C.

FINITE ELEMENT METHOD

CELL Program (1970)

[A7, B6, B9, BIO, B16]

This program analyses cellular structures of constant depth

with arbitrary plan geometry. The structure must be made up of
top and bottom decks and vertical webs. Two different finite
element types are used to capture the main behavior of the deck
and web components. The top and bottom deck slabs are idealized
by quadrilateral elements having a total of five degrees of
freedom (DOF) at each corner node. These elements possess both
membrane and plate bending stiffnesses. The vertical webs are
idealized by special rectangular spar elements having a total of
five DOF at each corner mode. A single element over the entire
depth of the bridge can be used to capture the essential in-plane
behavior and transverse bending of the web.
Orthotropic plate properties and arbitrary loadings and
boundary conditions can be treated. Automatic element and

178
coordinate generation options minimize the required input data.
FINPLA Program (1967)

[A2, AS, B6]

This program utilizes the finite element method to analyze

straight prismatic folded plate structures made up of rectangular
plate finite elements reinforced by one dimensional transverse
or longitudinal beam elements and supported by single or multiple
column rigid frame bent supports.
Internal diaphragms can also
be included. Arbitrary loadings and boundary conditions can be
treated.
The rectangular plate finite elements possess both membrane
and plate bending stiffnesses. They have six degree~ of freedom
(DOF) at each corner mode. The beam and frame one dimensional
elements possess the standard six DOF at each end and may be
eccentric with respect to the plate system. The program has been
written primarily for box girder bridge structures and requires
a minimum of input data to define the structure and the loading.
Displacements and internal forces and moments are output for all
elements specified.
FINPLA 2 Program (1971)

[A12,B6, B16]

This program utilizes the finite element mothod to analyze

general nonprismatic box girder bridge structures of varying width
and depth and may have an integrated three-dimensional frame.
The
structure is discretized by dividing it longitudinally into a
certain number of structure segments by vertical sections, and
by subdividing each such segment into finite elements. The
structure alignment is described by a longitudinal reference line
which may be a straight line, a circular curve or an arbitrary
planar string polygon and cross sections are defined with respect
to this line.
A structure of this type is made up of quadrilateral two
dimensional elements and one dimensional frame elements, each of
which is assumed to have six DOF at each node, three translations
and three rotations.
The quadrilateral elements for both the decks and the webs
possess both membrane and plate bending stiffnesses and they may
have an arbitrary orientation in space. They are taken as flat
plate elements which give the best least squares fit through the
actual location of the corner nodes. For box girder bridges of
general plan and elevation geometry this assumption should be quite
good.
Orthogropic plate properties and arbitrary loadings and
boundary conditions can be treated. Automatic element and coordinate

179
generations options minimize the required input data.
D.

FINITE SEGMENT METHOD

SIMPLA Program (1967)

[A2, B4]

The program provides a complete analysis of straight prismatic

cellular or open folded plate structures. The structure may be
composed of one or more (up to 15) types of plates. The folded
plate structure is defined transversely by the cross section in
terms of the dimensions of its plate elements and their joint
interconnections, and longitudinally by the number of segments.
Arbitrary support conditions can be accounted for.
The input data
is so arranged that only the properties of a typical cross section
need be specified, and any repeating segment can be indicated by
a simple input format.
The solution is based on the finite segment concept in which
each plate element is divided into a finite number of segments
longitudinally. Compatibility and equilibrium conditions are then
satisfied along the four edges of each segment. Each segment of
the plate is assumed to obey the ordinary theory of folded plate
analysis. This means, slab action is defined by the behavior of a
transverse one-way slab spanning between longitudinal joints, and
membrane stresses produced in each plate by longitudinal plate
action are calculated by the elementary beam theory. A transfer
matrix method is used to solve the resulting set of linear equations
for the unknown displacements and then the internal plate forces and
moments can be found.
E.

NONLINEAR ANALYSIS METHOD

PCFRAME Program (1977)

[A29, B23, B26]

This program provides an efficient numerical procedure for

the material and geometric nonlinear analysis of planar reinforced
and prestressed concrete frames including the time dependent effects
due to load history, temperature history, creep, shrinkaqe and
aging of concrete and relaxation of prestressing tendons.
The
analysis is capable of predicting the response of these structures
throughout their service load history as well as throughout the
elastic cracking, inelastic and ultimate load ranges.
The analytical model consists of a series of interconnected
beam elements with three degrees of freedom at each end. Varied
material properties within a beam element are accounted for by a
composite concrete and reinforcing steel layer system. Pretensioned or post-tensioned bonded or unbonded tendons of general
profile can be specified and results for stresses in the concrete,
reinforcing and prestressing steel and displacements can be obtained

180
before, at and anytime after the transfer of prestress.
SEGAN Program (1978)

[A21, B19]

This program permits the analysis of curved segmentally

erected prestressed concrete box girder bridges, including time
dependent effects, so that deflections and stresses can be
computed at any stage during the construction or service life of
the bridge.
The segments can be either precast or cast in place. The
bridge cross section must consist of a single box section with
vertical or inclined webs and have cantilever flanges. Width and
depth dimensions can vary along the bridge span. Time-dependent
material changes, such as creep and shrinkage of concrete, as
well as environmental effects such as humidity and temperature
changes, are accounted for.
However, other concrete material nonlinearities such as cracking or inelastic behavior and geometric
nonlinearities are not included in the analysis.
Prestressing in the longitudinal direction can be included in
the analysis. Tendons can be located in either the webs or
flanges and can have a linear or parabolic profile. Stressing
operations encountered in segmental construction, such as stress
and release operations, slip in of tendons, and restressing and
removal of tendons are included. A procedure is included for
specifying the structural configuration and the erection sequence.
Operations used in segmental construction, such as addition of
segments, prestressing, changing of support boundary conditions,
application or removal of construction loads, and prescribed
displacements can be analyzed. At each construction stage, the
current structure is analyzed and the incremental and total
displacements, stresses, and strains are obtained. The complete
structure can also be analyzed for a variety of live loads and time
steps.
NOBOX Program (1982)

[A28, B27, B28]

This program can be used to trace the nonlinear response of

multi-cell reinforced concrete box girder bridges under stepwise
increasing static loads. Nonlinearities considered are material
nonlinearities inherent in reinforced concrete structural members
under short term loading such as cracking of the concrete, yielding
of the reinforcement and formation of plastic hinges due to shear
and flexure.
The analytical model is developed for multi-cellular structures
of arbitrary plan geometry and constant height. The analytical
method can trace the complete nonlinear behavior of the bridge
beyond the working stress range all the way up to the ultimate

181
failure and collapse of the structure.
and failure loads can be determined.

Both collapse

mech~nisms

A three-dimensional grillage, consisting of longitudinal beam

elements, transverse diaphragm elements, transverse bending frames
and specially developed torsional shear panels, forms the basis
for the analytical model. Material nonlinearities in the form of
member end force-deformation relationships for reinforced concrete
members subjected to flexure, shear or torsion, as obtained from
numerous experimental investigations, are approximated by simple
trilinear elastic-perfectly plastic force-deformation characteristics.

182

APPENDIX 2.

REFERENCES ON BOX GIRDER BRIDGE RESEARCH PERFORMED

AT THE UNIVERSITY OF CALIFORNIA, BERKELEY

This Appendix 2 contains a list of research reports (part A)

and technical papers (part B) on analytical and experimental
research investigations performed in the Division of Structural
Engineering and Structural Mechanics of the University of
California, since 1965.
The technical papers (part B) are publications in technical
society journals and proceedings, which can be found in most
engineering libraries.
The research reports (part A) are published as part of a
Structural Engineering and structural Mechanics (UC-SESM) report
series of the University of California. Information on their
availability from the University of California may be obtained by
writing to:
Professor A. C. Scordelis
Department of Civil Engineering
729 Davis Hall
University of California
Berkeley, California 94720
Unfortunately, the original supply of some of the research
reports has been exhausted. However, most of them have been placed
on file with the U.S. Department of Commerce and may be obtained on
request for cost of reproduction by writing to:
National Technical Information Service
Operations Division
Springfield, Virginia 22161
The accession number (shown in parentheses on reference list
A) should be specified when ordering a particular report.

183

REFERENCES ON BOX GIRDER BRIDGE RESEARCH PERFORMED AT THE

UNIVERSITY OF CALIFORNIA, BERKELEY
A.

Research Reports
1.

Scordelis, A.C., "Analysis of Simply Supported Box Girder

Bridges," Structural Engineering and Structural Mechanics
Report No. SESM 66-17, University of California, Berkeley,
October 1966 (PB 175 646) .

2.

Scordelis, A. C., "Analysis of Continuous Box Girder

Bridges," Structural Engineering and Structural Mechanics
Report No. SESM 67-25, University of California, Berkeley,
November 1967 (PB 178 355).

3.

Scordelis, A.C., and Meyer, C., "Wheel Load Distribution

in Concrete Box Girder Bridges," Structural Engineering
and Structural Mechanics Report No. SESM 69-1, University
of California, Berkeley, January 1969 (PB 183 923).

4.

Willam, K. J., and Scordelis, A.C., "Analysis of Orthotropic Folded Plates with Eccentric Stiffeners,"
Structural Engineering and Structural Mechanics Report
No. SESM 70-2, University of California, Berkeley,
February 1970 (PB 191 051).

5.

Meyer, C., and Scordelis, A.C., "Computer Program for

Prismatic Folded Plates with Plate and Beam Elements,"
structural Engineering and Structural Mechanics Report
No. SESM 70-3, University of California, Berkeley,
February 1970 (PB 191 050).

6.

Meyer, C., and Scordelis, A.C., "Analysis of Curved Folded

Plate Structures," Structural Engineering and Structural
Mechanics Report No. UC SESM 70-8, University of
California, Berkeley, June 1970 (PB 193 535).

7.

Willam, K. J., and Scordelis, A.C., "Computer Program for

Cellular structures of Arbitrary Plan Geometry," Structural
Engineering and Structural Mechanics Report No. UC SESM
70-10, University of California, Berkeley, September 1970
(PB 196 143).

8.

Meyer, C., "Analysis and Design of Curved" Box Girder

Bridges," Structural Engineering and Structural Mechanics
Report No. UC SESM 70-22, University of California,
Berkeley, December 1970 (PB 197 289).

9.

"Bouwkamp, J.G., Scordelis, A.C., and Wasti, S.T.,

"Structural Behavior of a Two Span Reinforced Concrete Box

184
Girder Bridge Model, Volumn I, "Structural Engineering and
Structural Mechanics Report No. UC SESM 71-5, University
of California, Berkeley, April 1971 (PB 199 187).
10.

Scordelis, A.C., Bouwkamp, J.G., and wasti, S.T.,

"Structural Behavior of a Two Span Reinforced Concrete Box
Girder Bridge Model, Volume II," Structural Engineering and
Structural Mechanics Report No. UC SESM 71-16, University
of California, Berkeley, October 1971 (PB 210 431).

11.

Scordelis, A. C., Bouwkamp, J. G., and Wasti, S.T.,

"Structural Behavior of a Two Span Reinforced Concrete
Box Girder Bridge Model, Volume III," Structural
Engineering and Structural Mechanics Report, No. UC SESM
71-17, University of California, Berkeley, October 1971.

12.

Meyer, C., and Scordelis, A.C., "Computer Program for

Non-Prismatic Folded Plates with Plate and Beam Elements,"
Structural Engineering and Structural Mechanics Report
No. UC SESM 71-23, University of California, Berkeley,
December 1971 (PB 220 197).

13.

Lin, C.S., and Scordelis, A.C., "Computer Program for

Bridges on Flexible Bents," Structural Engineering and
Structural Mechanics Report No. UC SESM 71-24, University
of California, Berkeley, December 1971 (PB 210 171).

14.

Godden, W. G., and Aslam, M., "Model Studies of Skew

Box Girder Bridges," Structural Engineering and Structural
Mechanics Report No. UC SESM 71-26, University of
California, Berkeley, December 1971 (PB 223 120).

15.

Comartin, C.D., and Scordelis, A.C., "Analysis and Design

of Skew Box Girder Bridges," Structural Engineering and
Structural Mechanics Report No. UC SESM 72-14, university
of California, Berkeley, December 1972. (FB 226 793).

16.

Godden, W. G., and Aslam, M., "Model Studies of Curved

Box Girder Bridges," Structural Engineering and Structural
Mechanics Report No. UC SESM 73-5, University of
California, Berkeley, March 1973 (PB 226 842/AS) .

17.

Scordelis, A.C., Bouwkamp, J.G., and Larsen, P.K.,

"Structural Behavior of a Curved Two Span Reinforced
Concrete Box girder Bridge Model, Volume I," Structural
Engineering and Structural Mechanics Report No. UC SESM
74-5, University of California, Berkeley, September 1974
(PB 242 523/AS).

18.

Scordelis, A.C., Bouwkamp, J.G., and Larsen, P.K.,

185
"Structural Behavior of a Curved Two Span Reinforced
Concrete Box Girder Bridge Model, Volume II," structural
Engineering and Structural Mechanics Report No. UC SESM
74-6, University of California, Berkeley, September 1974
(PB 242 524/AS).
19.

Scordelis, A.C., Bouwkamp, J.G., and Larsen, P. K.,

"Structural Behavior of a Curved Two Span Reinforced
concrete Box Girder Bridge Model, Volume III," Structural
Engineering and Structural Mechanics Report No. UC SESM
74-7, university of California, Berkeley, September 1974
(PB 242 525/AS).

20.

Kabir, A.F., and Scordelis, A.C., "Computer Programs for

Curved Bridges on Flexible Bents," Structural Engineering
and Structural Mechanics Report No. UC SESM 74-10,
University of California, Berkeley, September 1974
(PB 242 470/AS).

21.

Van Zyl, S., "Analysis of Curved Segmentally Erected

Prestressed Concrete Box Girder Bridges," Structural
Engineering and Structural Mechanics, Report No. UC/SESM
78-2, University of California, Berkeley, January 1978
(PB 284 939/AS).

22.

scordelis, A.C., Bouwkamp, J.G., Wasti, S.T., and Anicic,

D. "Structural Behavior of A Skew Two Span Reinforced
Concrete Box Girder Bridge Model," Vol. I, Structural
Engineering and Structural Mechanics Report No. UC SESM
80-1, University of California, Berkeley, June 1980.

23.

Scordelis, A.C., Bouwkamp, J.G., Wasti, S.T., and Anicic,

D., "Structural Behavior of a Skew Two Span Reinforced
Concrete Box Girder Bridge Model," Vol. II, Structural
Engineering and Structural Mechanics Report No. UC SESM
80-2, University of California, Berkeley, June 1980.

24.

Scordelis, A.C., Bouwkamp, J. G. wasti, S.T., and Anicic,

D., "Structural Behavior of a Skew Two Span Reinforced
Concrete Box Girder Bridge Model," Vol. III, Structural
Engineering and Structural Mechanics Report No. UC SESM
80-3, University of California, Berkeley, June 1980.

25.

Scordelis, A.C., Bouwkamp, J.G., Wasti, S.T., and Anicic,

D., "Structural Behavior of a Skew Two Span Reinforced
Concrete Box Girder Bridge Model," Vol. IV,
Structural
Engineering and Structural Mechanics Report No. UC SESM
80-4, University of California, Berkeley, June 1980.

26.

Sorensen, S.I., and Scordelis, A.C., "Computer Program

186
for Curved Prestressed Box Girder Bridges," Structures
and Materials Research Report No. UC SESM 80-10, Department
of Civil Engineering, University of California, Berkeley,
December 1980.

B.

27.

Sorensen, S.I., and Scordelis, A.C., "Computer Program for

Curved Prestressed Slabs," Structures and Materials
Research Report No. UC SESM 80-11, Department of Civil
Engineering, University of California, Berkeley, December
1980.

28.

Seible, F., "Nonlinear Analysis and Ultimate Strength of

Multi-Cell Reinforced Concrete Box Girder Bridges,"
Structural Engineering and Structural Mechanics Report
No. UC SESM 82-2, University of California, Berkeley,
February 1982.

29.

Kang, Y.J., "Nonlinear Geometric, Material and Time

Dependent Analysis of Reinforced and Prestressed Concrete
Frames," Structural Engineering and Structural Mechanics
Report No. UC SESM 77-1, university of California,
Berkeley, January 1977 (PB 272 930jAS)

Technical Papers
1.

DeFries-Skene, A., and Scordelis, A.C., "Direct Stiffness

Solution for Folded Plates," Journal of the Structural
Division, Proceedings of American Society of Civil
Engineers, Volume 90, No. ST-3, June 1964.

2.

Scordelis, A.C., Davis, R.E., and Lo, K.S., "Load

Distribution in Concrete Box Girder Bridges," ACI
Proceedings of First International Symposium on Concrete
Bridge Design, Toronto, Canada, April 1967, ACI
Publication SP-23, 1969.

3.

Scordelis, A.C., and Davis, E.R., "Stresses in Continuous

Concrete Box Girder Bridges," ACI Proceedings of Second
International Symposium on Concrete Bridge Design,
Chicago, April, 1969, ACI Publication SP-26, 1971.

4.

Lo, K. S., and Scordelis, A.C., "Finite Segment Analysis

of Folded Plates," Journal of the Structural Division,
Proceedings of American Society of Civil Engineers, Volume
95, No. ST5, May 1969.

5.

Willam, K. J., and scordelis, A.C., "Analysis of

Eccentrically Stiffened Folded Plates," Proceedings of
IASS Symposium on Folded Plates and Prismatic Structures,

187
Vienna, September 1970.
6.

Scordelis, A.C., "Analytical Solutions for Box Girder

Bridges," Proceedings, Conference on Modern Developments
in Bridge Design and Construction, Cardiff, Great Britain,
April I 1971.

7.

Bouwkamp, J.G., Scordelis, A.C., and Wasti, S.T.,

"Structural Behavior of a Reinforced Concrete Box Girder
Bridge," Proceedings, Conference on Modern Developments
in Bridge Design and Construction, Cardiff, Great
Britain, April 1971.

8.

Meyer, C., and Scordelis, A.C., "Analysis of Curved Folded

Plate Structures," Journal of the Structural Division,
Proceedings of American Society of Civil Engineers,
Volume 98, No. STl, January 1972.

9.

Willam, K. J., and Scordelis, A.C., "Cellular Structures

of Arbitrary Plan Geometry," Journal of the Structural
Division, proceedings of American Society of civil
Engineers, Volume 98, No. ST7, July 1972.

10.

Godden, W.G., and Aslam, M., "Model Studies of Skew

Multicell Girder Bridges," Journal of the Engineering
Mechanics Division, Proceedings of the American Society
of civil Engineering, Volume 99, No. EM-I, February 1973.

11.

Scordelis, A.C., Bouwkamp, J. G., and Wasti, S.T.,

"Study of AASHO Loadings on a Concrete Box Girder Bridge,"
Highway Research Record No. 428, Highway Research Board,
Washington, D.C. 1973.

12.

Scordelis, A.C., Bouwkamp, J.G., and Wasti, S.T.,

"Structural Response of a Concrete Box Girder Bridge,"
Journal of the Structural Division, Proceedings of the
American Society of civil Engineers, Vol. 99, No. STIO,
October 1973.

13.

Scordelis, A.C., Bouwkamp, J.G., and Wasti, S.T.,

"Ultimate Strength of a Concrete Box Girder Bridge,"
Journal of the Structural Division, Proceedings of the
American Society of Civil Engineers, Vol. 100, STl,
January 1974.

14.

Aslam, M., Godden, W.G., "Model Studies of Multicell

Curved Box Girder Bridges," Journal of the Engineering
Mechanics Division, Proceedings of the American Society
of Civil Engineering, Vol. 101, No. EM3, June 1975.

188
15.

Scordelis, A.C., "Folded Plates for Bridges," Bulletin of

the International Association for Shell and Spatial
Structures, Madrid, No. 57, April 1975.

16.

Scordelis, A.C., "Analytical and Experimental Studies of

Multicell Concrete Box Girder Bridges," Bulletin of the
International Association for Shell and Spatial Structures,
Madrid, No. 58, August 1975.

17.

Scordelis, A.C. and Larsen, P.K., "Structural Response of

Curved RC Box Girder Bridges," Journal of the Structural
Division, Proceedings of the American Society of Civil
Engineers, Vol. 103, No. ST8, August 1977.

18.

Scordelis, A.C., Larsen, P.K. and Elfgren, L.G.,

"Ultimate strength of Curved RC Box Girder Bridge," Journal
of the Structural Division, Proceedings of the American
Society of Civil Engineers, Vol. 103, No. ST8, August
1977 .

19.

Van Zyl, S.F., and Scordelis, A.C., "Analysis of Curved

Prestressed Segmental Bridges," Journal of the Structural
Division, Proceedings of the American Society of Civil
Engineers, Vol. 105, No. ST 11, November 1979.

20.

Scordelis, A.C., Elfgren, L.G., and Larsen, P.K.,

"Time Dependent Behavior of Concrete Box Girder Bridges,"
Journal of the American Concrete Institute, Title No.
76-9, January 1979.

21.

Bouwkamp, J.G., Scordelis, A.C., and Wasti, S.T., "Failure

Study of a Skew Box Girder Bridge Model," Final Report,
11th Congress, Vienna, International Association for
Bridge and Structural Engineering, August 31 - September
5, 1980.

22.

Hellesland, J., and Scordelis, A.C., "Analysis of RC

Bridge Columns Under Imposed Deformations," Final Report,
IABSE Colloquium on Advanced Mechanics of Reinforced
Concrete, Delft, June 1981.

23.

Ketchum, M.A., and Scordelis, A.C., "Nonlinear Analysis

of a Prestressed Concrete Bridge," Final Report, IABSE
Colloquium on Advanced Mechanics of Reinforced Concrete,
Delft, June 1981.

24.

Scordelis, A.C., Wasti, S.T., and Seible, F., "Structural

Response of Skew RC Box Girder Bridges," Journal of the
Structural Division, Proceedings ASCE, Vol. 108, No. STl,
January 1982.

189

25.

Scordelis, A.C., Bouwkamp, J.G., Wasti, S.T., and Seible,

F., "Ultimate Strength of Skew RC Box Girder Bridge,"
Journal of the Structural Division, Proceedings ASCE,
Vol. 108, No. ST 1, January 1982.

26.

Kang, Y.J., and Scordelis, A.C., "Nonlinear Analysis of

Prestressed concrete Frames," Journal of the Structural
Division, Proceedings of the American Society of Civil
Engineers, Vol. 106, No. ST 2, February 1980.

27.

Seible, F., and Scordelis, A.C., "Nonlinear Behavior and

Failure Analysis of Multicell RC Box Girder Bridges,"
proceedings, CSCE International Conference on Short and
Medium Span Bridges, Toronto, Canada, August 1982.

28.

Seible, F., and Scordelis, A. C., "Nonlinear Analysis

of Multicell Reinforced Concrete Box Girder Bridges,"
Journal of Engineering Structures, IPC Science and
Technology Press Limited, January 1983.

191

COMPARATIVE STRUCTURAL BEHAVIOR OF STRAIGHT, CURVED AND SKEW

REINFORCED CONCRETE BOX GIRDER BRIDGE MODELS

S. Tanvir Wasti

Alex C. Scordelis

Professor of Civil Engineering

Middle East Technical University
Ankara, Turkey

Professor of Civil Engineering

University of California
Berkeley, California USA

SUMMARY
Comparisons of theoretical and experimental values of selected
structural quantities, such as reactions, deflections and moments
are presented for three large scale (1:2.8), two span, four cell,
reinforced concrete box girder bridge models tested under dead load,
working loads, and overloads, including loading to failure. All
three models had the same cross-sectional and longitudinal span
dimensions, but in plan one was straight, one was curved and one
was skew. The models were 72 ft (21 m) long along the longitudinal
centerline, 12 ft (3.7 m) wide and 1.71 ft (0.52 m) deep.
Amounts of steel reinforcement used in the straight and curved
model were similar, but in the skew model the amount was considerably less.
General agreement was found between theoretical and experimental results. The structural response of the straight and curved
models to point loads at midspan sections was similar, however that
of the skew bridge was different and highly dependent on the transverse position of the applied point load. All three bridge models
exhibited excellent behavior with high overload capacities before
ultimate failure.

192

1.

INTRODUCTION

Box girder bridges are extensively used in highway construction

because of their structural efficiency and because they offer an
economical alternative in many cases. Although plan geometries may
be straight, curved or skew because of lack of space or the growing
number of complex intersections, design procedures, e.g., the AASHTO
and State of California wheel load formulas, do not incorporate any
parameters to account for curvature or skew. The purpose of this
paper is to present comparisons of theoretical and experimental
values for selected structural quantities such as reactions,
deflections and moments to assess the effects of curvature or skew
on the behavior of otherwise similar large scale reinforced concrete
box girder bridge models.

2.

DESCRIPTION OF BOX GIRDER BRIDGE MODELS

An in-depth program of analytical and experimental research on

box girder bridges has been conducted at the University of California, Berkeley, since 1965. Especially relevant to the present
paper are a series of technical reports and papers [1-9] that
describe detailed experimental and analytical studies of 1:2.8
scale continuous two span, four cell straight, curved and skew
reinforced concrete box girder bridge models.
In each case the
scale was chosen so as to allow the representation of a typical No.
11 reinforcement bar of area 1.56 sq. in. (1006 rnrn 2 ) in the prototype by a No.4 reinforcement bar of area 0.196 sq. in. (126 rnrn 2 )
in the model. Construction of the model was a replica of prototype
construction in the field, with the bottom slab and webs being cast
first and the top slab being cast subsequently.
In order to obtain proper simulation of prototype behavior and
ensure that strains and stresses measured in the model would
represent those in the prototype, it was necessary in each case to
add extra dead weight which, along with the self-weight of the model,
would result in 2.8 times the weight of the model itself. various
schemes for implementing the placing of this extra weight (equivalent in each of the three models to about 95 kips or 423 kN) were
examined. For the straight and curved bridge models, it was decided to place steel billets inside the cells and to distribute
their weight by means of wooden supports and layers of sand. For
the skew bridge model it was decided to use 80 concrete blocks on
the top deck of the bridge as the required extra weight.
Main longitudinal reinforcement for the straight and curved
models comprised 55 No. 4 rebars of area 10.8 sq. in. (6968 rnrn 2 )
in the midspan regions of the bottom slabs and 82 No.4 rebars of
area 16.1 sq. in. :10387 rnrn 2 ) in the top slab over the center bent.
Comparative figures for the skew model were 50 No. 4 rebars of area

193

9.8 sq. in. (6323 mm 2 ) in the bottom slabs and 68 No.4 rebars of
area 13.4 sq. in. (8645 mm 2 ) in the top slab over the center bent.
The reductions represent smaller design moments for skew bridges
because of the observed decrease in dead load moments in skew
bridges [7]. It should, however, be pointed out that the main
longitudinal reinforcement quantities comprise only a part of the
total steel in box girder bridges because of the additional requirements for transverse steel and web reinforcement.
The elevation and plan views of the three bridge models are
given in Fig. 1 with basic dimensions and designations of significant longitudinal and transverse sections. The common transverse
section is shown in Fig. 2. In each model A, B, C and D represent
the instrumented sections, Wand E the West and East abutment
sections and X and Y the midspan sections. Z is the central right
section showing the orientation of the center bent diaphragm for
the straight and curved models, whereas the skew center bent section
along the diaphragm for the skew model is designated as T. Span I
for each of the models possesses a midspan diaphragm which, for the
skew model, is staggered as shown in Fig. 1. In the straight and
curved models it may be observed that the distances between sections
A and X and sections D and Yare 3.58 ft (1.09 m) and that sections
Band C are in each case 3 ft (0.91 m) on either side of section Z.
For the skew model it was decided to divide the sections in mUltiples
of 2.57 ft (0.78 m). In each case it was considered preferable to
locate the instrumented sections A and D close to but not at the
midspan loaded sections.
For all models, locations were designated by the intersection
of the girder and the associated transverse (right or skew) section.
Thus IX refers to the location where girder 1 and the midspan section
X cut each other. For the skew bridge model, the acute side and
obtuse side refer to the bridge portions on either side of girder 3,
these portions are skew-symmetrically placed in the two spans of
the bridge.

3.

EXPERIMENTAL PROGRAM

The model had a loading frame at each of the midspan sections

X and Y, enabling point loads to be applied at each of the girders
1-5 either singly or in combination. Instrumentation and data
acquisition were planned to allow the measurement of girder and
central column reactions, deflections at the sections of interest
and internal tensile and compressive strains.
Each phase of the live load program for each bridge model
comprised firstly the application of equal loads called "conditioning loads" on all girders at both midspans to produce the samp
nominal total steel stresses (24, 30, 40, 50 and 60 ksi = 166, 207,

194

A X

II

WESTI

Y 0

~~~~~~==~==~
SPAN I

SPAN

36.00

A X

~I

..- DIAPHRAGM

1442

358

Y 0

r-

t---o300 30P

1500

358

1500

III
N

1442

r-6

oil

GIRDER

J'

~l
~1543 ~

G>;

.!

2'/\45.

X \" DIAPHRAGM

2S

=---- ~

VY/

ltj
1800

1//

// /

SCHEMATIC
SKEW

BOX

REPRESENTATION
GIRDER

Y D
I

1543

III

//

!@
oil

=0.305 m )

OF ELEVATION

BRIDGE MODELS

~ '"
r-

1800

///

DIMENSIONS IN FT (1 FT

FIG. 1

) }}

AND PLANS OF STRAIGHT

CURVED

AND

=,0' - 311ft

'2' - O
M

35/8M

FIG. 2

5' -cf

TYPICAL

SECTION

L, "

" _8 H

ROUND COLUMN

" -6"

OF BOX

... i

GIRDER

2' - 6 114

BRIDGE

MODELS

= 25.4mm

=0.305 m

IN.

FT.

4' - 6" SQ. FOOTING

~~i--~--------~--------~----~~~--~--~~~

iOiOI

4 (@ 2'-6 7/8

\0

v.

196
276, 345 and 414 MPa) at sections of maximum positive and negative
moment. After the removal of the conditioning loads, point loads
were applied in each phase in several combinations, to produce
total stresses in the tensile reinforcement of each girder (considered separately) of the order of the working stresses, i.e.,
24 and 30 ksi (166 and 207 MPa). The conditioning loads represented
the successive deterioration of the bridge models under the effects
of overloading.
Details of the very extensive loading program for each model,
incorporating different support restraints, truck and construction
vehicle loads and also loading to failure shall not be dealt with
herein, but may be studied in the above mentioned references.

4.

THEORETICAL ANALYSES

Theoretical results for the straight bridge model were obtained

using a finite element computer program entitled FINPLA2 [10] whereas both the curved and skew bridge models were analyzed using the
finite element program CELL [11]. Both FINPLA2 and CELL assume that
the plates forming the box girder are elastic, isotropic and homogeneous and that a linear relationship exists between forces and
deformations.
For each model the results obtained from the chosen finite
element programs were compared in detail with those from several
other analytical models. Extensive comparisons indicated the
superiority of the specially developed FINPLA2 and CELL programs in
predicting reactions, deflections and moments for box girders and also
the distribution of the section moments along the different girders.
The FINPLA2 model for the straight bridge consisted of a
relatively coarse mesh that resulted in a total of 180 top deck,
120 bottom deck and 150 web elements. For the curved bridge, the
analytical model comprised a total of 534 elements. The CELL
finite element model for the skew bridge consisted of 576 deck
elements each for the top and bottom slabs and 312 vertical web
elements.

5.

5.1

COMPARISONS OF THEORETICAL AND WORKING STRESS EXPERIMENTAL

RESULTS
Preliminary Results

Pertinent comparisons of theoretical and experimental values

for end abutment and center footing reactions and for moments
(total section moment and distribution to girders) will be presented

197
for the straight, curved and skew bridge models. The load cases
treated here will be restricted for lack of space mainly to the
following:
IX, 3X, 5X, IX + lY, 3X + 3Y and 5X + 5Y. The experimental values will comprise results recorded for these point loadings after the 30 ksi (207 MPa) conditioning load level. Because
the analytical models do not allow for cracking, comparisons of
deflections will be limited to experimental results only.
In order to enable an assessment of the response of the three
bridge models to overload stresses, some results of reactions,
deflections and moments at the 60 ksi (414 MPa) stress level will
be included in Section" 6.
5.2

Reactions

The theoretical (T) and experimental (E) values of the vertical

reactions RW for the west end, RF for the center footing and RE for
the east end with the associated bending moments MW, Mp and ME and
torques Tw, Tp and TE are listed in Table 1 for the six load cases
IX, 3X, 5X, Ix + lY, 3X + 3Y and 5X + 5Y.
Por all three models the boundary conditions were in general
similar with the exception that flexible vertical end supports were
used for the skew bridge instead of the rigid vertical end supports
used for the straight and curved bridges and a horizontal transverse
restraint at the end supports was used for the straight bridge.
These differences, however, are not likely to influence the results
considerably.
An observation of the results in Table I indicates that good
agreement exists between the theoretical and experimental values
in virtually all cases.
The total vertical reactions RW' Rp and RE change only very
insignificantly for point loads moving transversely across the
straight or curved bridge.
The skew bridge, however, shows significant changes in the vertical reactions depending on the transverse
positioning of the load. While for load cases where the acute side
of a span is loaded, the adjacent end reaction is smaller than for
the corresponding straight and curved bridge cases, this reaction
increases to a value much larger than for the straight and curved
bridge models as the loads shift from the acute to the obtuse side
of the skew span. Por the load cases where only the center girder
3 is loaded, the skew bridge features slightly higher end reactions
and slightly lower vertical center bent reactions than the curved
or straight bridge.
In addition to this, however, a significant
negative end moment can also be noted for the skew bridge which
decreases the total positive midspan moment. This negative end
moment increases in magnitude as the loads move towards the obtuse
side of the span, but decreases and even changes sign for loads

198

TABLE'
COMPARISON OF REACTIONS FOR STRAIGHT. CURVED AHO SKEW BRIDGE MODELS

~~~TE

\Mw

'

ME

\:_4iY=fIt2~r'
t

REACTIONS (KIPS AND FT. KIPS)

0
D

LOAD

WEST END

E
L

CASE

Mw

TW

39.2

38.0

MF

TF

RE

-355

69.0

96

- 42

-8.2

- :35

-300

69.0

135

- 66

-8.3

-1 <'2

41.2

-283

64.5

110

- 83

-5.7

- 99

41.2

-280

68.1

24

- 78

-7.1

- 96

32.5

193

-193

72.9

117

-196

-5.4

125

-12S

33.6

181

-181

7ol.n

133

-216

-6.0

126

-L.. o

39.3

68.8

86

-S.l

38.1

70.3

135

-8.3

-2

38.8

57

69.1

122

-7.9

- 17

40.0

67

6~.7

103

-7.2

- 20

42.S

-93

93

64.4

91

- 69

-6.9

35

- 35

42.2

-78

78

69.1

107

- 65

-8.1

10

- 10

39.2

335

6~.O

96

42

-8.2

95

37.9

306

69.5

128

72

-8.3

116

36.2

395

74.1

133

82

-10.2

63

36.4

400

76.0

145

74

-9.7

.)5

52.6

-390

390

55.9

65

51

-8.5

-73

73

57.0 -373

373

62.3

91

68

-9.1

-83

83

CU

ST

CU

SK

ST

5X

CU

SK

1 KIP

EAST END

RF

SK

3X

CENTER FOOTING

RW

ST

lX

= 4.448 kN

, FT .0.305 m

ME

1FT-KIP =1.356kN-m

TE

199

TABLE 1 (CONT)
COMPARISON OF REACTIONS

FOR STRAIGHT. CURVED AND SKEW

REACTIONS

0
0

LOAD

T
ST

1X
+
1Y

3X
+
3Y

CU

SK

SX
+
SY

1 KIP

END

RW

MW

TW

RF

MF

TF

31. J

-432

133.0

- 84

3' . J

-_ 31

30.1

-374

1 3~. 2

14

-140

30 <1

-<l5 ~~

Re

ME

Te

-332

129.3

-168

35.3

35.5

-381

132.5

-54

-149

33.2

- 346

24.1

264

-264

12H.7

54

-247

47 3

517

-317

24.5

244

-244

132.3

58

-277

47 2

5~S

-5)5

?'.~2

.-I...

31. 2

137.7

31.2

29.7

30.5

30.9

"

138.4

39

138.3

-3

30.8

39

31. 3

37

141.4

72

30.1

35

35.6

-118

118

128.9

-2

35.5

117

-117

36.0

- 41

99

137.6

35.9

85

- 85

31. 0

432

138.0

84

31.0

431

30.3

389

140.2

-2

144

31. 5

468

26.0

458

147.9

166

26.1

4,8

28.7

461

151.4

-31

162

27 .1

<117

47.1

-515

515

128.9

-52

249

24.0

-265

-265

47 .8

-475

475

140.1

-49

300

24.0

-277

!. 77

CU

SK

EAST

T
ST

KIPS)

35.3

T
ST

FT

MODelS

CU

SK

AND

CENTER FOOTING

WEST END

CASE

( KIPS

BRIDGE

= 4448 kN

1 FT

=0.305 m

1 FT- KIP = 1.356 kN - m

200
moving towards the acute side of the span. Note that there is no
end moment for the simply supported straight or curved bridges and
that the end moment for the skew bridge is only introduced by the
skewness of the simply supported end abutment.
The torque at the end abutments of the skew bridge has the
same magnitude but opposite sign as the end moment and is smaller
for loads on the acute side but larger for loads on the obtuse side
of the skew span than in corresponding straight or curved bridge
cases.
5.3

Deflections

In Figs. 3, 4 and 5 the experimental vertical deflections along

girders 1, 3 and 5 have been plotted for the straight, curved and
skew bridge models for a total of nine different load cases. The
load values have been normalized to 100 kips (445 kN).
The number
of experimental points for the skew bridge model is much greater
than for the other two cases because several more linear potentiometers were used for deflection measurements. No modifications
have been made in the plotted experimental results to reflect the
fact that the modulus of elasticity of the concrete was slightly
different for the straight, curved and skew bridge models.
Considering the deflection under the application point of the
load, Figs. 3, 4 and 5 show that the skew bridge features the
highest deflection at the acute side of a span and the lowest
deflection of all three bridge types at the obtuse side of the
span.
The influence of the midspan diaphragm in Span I can also be
seen by comparing the deflections at section X and section Y. Thus
for the load cases IX + lY and 5X + 5Y it is observed that the
maximum deflection in the diaphragmed span (between 0 and 36 ft) is
less for both the straight and curved bridges than the maximum
deflection in the undiaphragmed span.
For the load at IX it is normal to expect the curved bridge to
have the smallest deflection under the load because of the smaller
length of girder 1. Conversely, for the load at 5X, the curved
bridge should feature the largest deflection. This is borne out by
Figs. 3, 4 and 5. Finally, the measured deflections for the load
cases 3X, 3Y and 3X + 3Y which comprise loads on the center girder
only are observed not to be significantly different for the three
types of bridge models.
While no theoretical values for deflections have been plotted
in the above figures because the analytical model assumes a homogeneous material with completely linear elastic behavior, the
observed deflections lie in all cases between the limits of 1.0 to

201

: I

-I

I-~-S
,-3

I~

,-

0.0

90
I

180
I

210
I

45.0

36.0
I

54.0
I

63.0
I

-0.5

1.0
STRAIGHT
CURVED

o SKEW

-05
LOAD AT IV

1.0

2.0
LOAD AT IX .. IV

10

2.0
1 IN ... 25.4 mm

FIG 3

DISTANCE IN FT

GIRDER

W'

R3

3 I"

VERTICAL

1 FT.

= o 3048 m

DEFLECTIONS (INCHES) ALONG GIRDER 1 FOR 100 KIP

LOADS AT 1X.W. 1X .. 1Y

72.0
IE

202

II-~-II
3-3
,-

y-'

,,
II-

GIRDER 3
aD

11.0

I
X

/\ 3

DISTANCE IN FT.
18.0

27.0

36.0

45.D

54.0

63.0

72.0

IE

W~I----~I------~I------+------+------+------+------+-----~

-0.5
ao~

__~lO~A~D~A~T~3~X~__________~~____________________~~_

1.0

STRAIGHT
CURVED
o SKEW

2.0

-0.5
ao~~----~~--~~----~~~--------------------~

1.0

2.0

2.0

1 IN . 25.4 mm
FIG. 4

VERTICAl

1 FT. 0.3048 m

DEFLECTIONS (INCHES> ALONG GIRDER 3 FOR 100 KIP

LOADS AT 3X. 3V 3X +3V

203

Ia-I
,-

-I
IY I-,a

I-~'I
a'3
,

I-

3 <<

~, 3

GIRDER S

OD
Wi

DISTANCE IN FT

11.0
I

11.0
I

'n.0

36.0
I

45.0
I

54D
I

63.0
I

720
IE

-S.O

STRAIGHT
CURVED

o SKEW

2.0
-O.S

0.ok~~~2:;=::~.,...---------,
LOAD AT SY

1.0

2.0

1 IN . 2S.4

AG. 5

VERTICAL

mm

1 FT 0.3048 m

DEFLECTIONS (INCHES) ALONG GIRDER 5 FOR 100 KIP

LOADS AT 5X .5V. 5X.5V

204
1.5 times the analytically predicted values.
5.4

Internal Moments

A comparison of theoretical and experimental internal moments

at Sections A and D for the six load cases IX, 3X, 5X, IX + lY,
3X + 3Y is presented in Table 2 for the straight, curved and skew
bridge models, respectively. The agreement between theory and
experiment, while not as close as in the case of reactions, is
adequate.
While the total moments for the straight and curved bridge
change only a little for point loads moving from girder 1 to girder
5, the skew bridge total midspan moments vary significantly. For
loads on the acute side of the skew span, the total moments are much
higher, and for loads on the obtuse side of the span much lower than
for the straight or curved bridges. This is especially important
for the positioning of high concentrated live loads such as heavy
construction vehicles. As might be expected, the differences in
moment values for the three different bridge types are least when
the loads are placed over the central girder 3.
Two further points indicating the difference in the behavior
of the skew bridge model compared to the straight and curved models
are not featured in Table 2 but will be mentioned here. One is
that for the important load cases of dead load and conditioning
loads the skew bridge possesses substantially lower internal moments
than the straight or curved bridges. Secondly, the negative moment
over the center support is found to be smaller for the skew bridge
than for the straight or curved bridges, independently of the load
position on the bridge.
The transverse distribution of the maximum total positive
moments for the straight, curved and skew bridge models to the
individual girders provides an assessment of the load distribution
properties of the box girders. While restrictions of space prevent
a detailed presentation here of the theoretical and experimental
percentages of the total moment taken by the individual girders for
different load cases, it can be stated that reasonable agreement
was found between theoretical and experimental values.
Analytical and experimental results indicate that the distribution of maximum total moments to individual girders for the
straight vs. the curved bridge are almost identical at Section A
for the diaphragmed Span I and only slightly different at Section D
for the undiaphragmed Span II. However, they differ from those of
the skew bridge which shows larger percentages especially of the
experimental internal moments being taken by the exterior girders.
This is especially true in the vicinity of location lA near the
staggered midspan diaphragm in the skew bridge.

205

TABLE
THEORETICAL
(LOAOS

ANO

EXPERIMENTAL

NORMALIZEO

TOTAL

MOMENTS

AT

1X

3X

5X

1X+1Y

3X+3Y

5X+5Y

INTERNAL
LOAD CASE

A ANO 0

SECTIONS

TO 100 KIPS)

BRIDGE

MODEL

MOMENTS (FT - KIPS)

SECTION A

SECTION

THEORY

EXPER.

THEORY

STRAIGHT

565

524

-118

-125

CURVED

557

515

- 97

- 75

SKEW

693

657

-20~

-183

STRAIGHT

561

535

-116

-123

CURVED

569

538

-116

-124

SKEW

561

556

-131

-123

STRAIGHT

565

542

-118

-127

CURVED

582

586

-139

-143

SKEW

420

407

- 58

- 62

STRAIGHT

448

404

443

461

CURVED

458

427

457

431

SKEW

632

597

212

268

STRAIGHT

446

406

448

443

CURVED

453

467

452

399

SKEW

430

433

430

437

STRAIGHT

448

429

443

475

CURVED

444

440

445

467

SKEW

211

273

634

576

1 KIP. 4.448 kN

1 FT -KIP.1.356 kN-m

EXPER

206

6.

6.1

COMPARISONS OF THEORETICAL AND OVERLOAD STRESS EXPERIMENTAL

RESULTS
Preliminary Remarks

In order to provide an indication of the modification of the

structural response of the straight, curved and skew box girder
bridge models with overload, selected comparisons of reactions,
deflections and moments are presented in this section. Details are
available in References [1-9]. The chosen results are from the undiaphragmed Span II with loads at Section Y so as to eliminate the
effects of the midspan diaphragms in Span I.
6.2

Reactions and Deflections

In Fig. 6 the behavior of the total vertical east end reaction

and deflection at the location 5Y for a point load applied at 5Y
after the removal of the 30 ksi (207 MPa) and 60 ksi (414 MPa)
conditioning load is depicted. As mentioned in section 3, this
point load was actually applied at several different locations to
produce nominal stresses in the tensile reinforcement of the girder
where it acted of 30 ksi (207 MPa). The value of this point load
was about 20 kips (89 kN), and it has not been normalized so that it
is retained at its actual value.
The reaction graphs show only a minor change in the values of
the reactions after overload stresses of up to 60 ksi (414 MPa) .
The deflections at 5Y, however, increase when the same point load
is reapplied after increased cracking and deterioration of the
bridge models. As expected, the deflection at 5Y of the skew model
is greater even than that of the curved model where girder 5 is the
longest (outermost) girder and should show large deflection.
6.3

Moments

The point load of about 20 kips (89 kN) at location 5Y has been
taken as a basis of comparison for the assessment of changes in the
transverse distribution of the moments at sections A and D after the
successive application of conditioning loads resulting in nominal
maximum tensile stresses in the steel reinforcement at Sections X
and Y of 24 ksi to 60 ksi (166 MPa to 414 MPa).
In Fig. 7a the
theoretical distribution in percentage of the total section moment
to each girder for a point load at 5Y is given for all three bridge
models. This distribution does not change with an increase in
stress level because of the assumption of linear elastic behavior
in the analytical models. The experimentally measured percentages
in Fig. 7b indicate fluctuati~ns of up to 4% in the girder distributions, but comparisons with theoretical percentage predictions
continue to be reasonable.

....Z

-'

'"

g
....

z
~

Ie

....

FIG.6

14

16

18

1 KIP. 4.44BkN
1 IN .254mm

THEORY

0.3
DEFLECTION

0.2

AT 5Y

0.4

LOAD AT 5Y AFTER 30 AND 60 KSI

CONDITIONING LOADS

0.5

0.6

FOR A POINT

STRAIGHT
CURVED
SKEW

COMPARISON OF TOTAL VERTICAL EAST REACTION (KIPS) AND DEFLECTION (INCHES) AT 5Y

EAST EN!) REACTION (KIPS)

10D

10

30 KSI

~~

20

10

STRAIGHT
CURVED
SKEW
I
8.0

l f..v/

12

6D

60 KSI ~

12

14

16

18

~O

2~

-.]

208

I-

3-1

1-

1-5

1~~-53

I I II 1~

AX

3~

,~

YD

AX

YD

~s::~

SECT.
A

SECT.
D

(0) THEORY

EI STRAIGHT

oCURVED

fjSKEW

~
SECT
A

...

~'"

~",,,,

1 1

"5!~
1 1 1

00 . . . .

;~~

SECT

(b) EXPERIMENT
FIG. 7

00 ... '"
~~'"

MAX AND"" ""LUES FOIl 24.30.40.50

AND 50 KSI LEVEL

COMPARISON OF PERCENTAGES OF TOTAL MOMENT AT SECTIONS A AND D

EACH GIRDER FOR A POINT LOAD AT 5 Y AFTER 24 TO 60 KSI
CONDITIONING

LOADS

W9

7.

CONCLUSIONS

The finite element computer programs based on linear elastic

analysis of an uncracked homogeneous concrete structure are observed
to predict the structural response of all three types of box girder
bridges satisfactorily and accurately, especially with regard to
reactions and moments at sections of interest. Experimental values
of deflections at working stress levels lie between 1.0 and 1.S
times the theoretical values but the experimental trend is suitably
predicted by the analytical model.
The straight, curved and skew box girder bridge models do
indicate broad similarities of overall structural behavior. However,
the structural behavior of the curved box bridge closely resembles
that of the straight bridge, while the skew bridge features marked
differences. All loadings produce end moments and torques only in
the case of the skew bridge. Furthermore, the skew bridge is
sensitive to the transverse location of the load on the top deck,
showing higher deflections and moments for loads on the acute side
of the span and lower moments for loads on the obtuse side of the
span. Similarly the outermost girder registers the highest
deflections in the case of the curved bridge model.
The general response of the skew bridge is favorable under
conditioning loads (and also under dead load) in that smaller
total section moments are produced compared to the straight and
curved bridges. The skew center bent diaphragm also offers the
advantage of participating in the longitudinal bending and thus
attenuating the internal negative moments in the center bent
region.
In
towards
For the
greater

all cases the midspan diaphragms in Span I contribute

reducing the maximum deflections under eccentric loads.
skew bridge the diaphragm has the effect of transferring a
portion of the section moment to the exterior girders.

The level of deterioration in the bridge models as measured

herein does not produce any major deviation in structural behavior
over the range 30 ksi - 60 ksi (207 MPa - 414 MPa).
The reactions
can be adequately predicted and the transverse distribution of
moments to each girder is substantially unaffected. The deflections given by theory need to be augmented by a suitable magnification factor.
Finally, it should be emphasized that all three bridge models
exhibited excellent load carrying abilities and structural
behavior even after high overloads. All three models had ultimate
capacities well above DL + 4LL even though the skew bridge had
substantially less longitudinal reinforcement than the straight or
curved bridges.

210

S.

REFERENCES

1.

Scordelis, A. C., Bouwkamp, J.G. and Wasti, S.T., "Structural

Behavior of a Two Span Reinforced Concrete Box Girder Bridge
Model," Vols. I, II, and III, SESM Reports Nos. 71-5, 71-6
and 71-7, university of California, Berkeley, Calif., 1971.

2.

Scordelis, A.C., Bouwkamp, J.G., and Wasti, S.T., "Structural

Response of Concrete Box Girder Bridge," Journal of the
Structural Division, Proceedings of the American Society of
civil Engineers, Volume 99, No. STIO, October 1973.

3.

Bouwkamp, J.G., Scordelis, A.C., and Wasti, S.T., "Ultimate

Strength of Concrete Box Girder Bridge," Journal of the
Structural Division, Proceedings of the American Society of
Civil Engineers, Vol. 100, No. ST1, January 1973.

4.

Scordelis, A.C., Bouwkamp, J.G., and Larsen, P.K., "Structural

Behavior of a Curved Two Span Reinforced Concrete Box Girder
Bridge Model," Vols. I, II and III, SESM Reports Nos. 74-5,
74-6 and 74-7, University of California, Berkeley, Calif., 1974.

5.

Scordelis, A.C., and Larsen, P.K., "Structural Response of

Curved RC Box Girder Bridge," Journal of the Structural
Division, Proceedings of the American Society of Civil Engineers,
Vol. 103, No. ST8, August 1977.

6.

Scordelis, A.C., Larsen, P.K., and Elfgren, L.G., "Ultimate

Strength of Curved RC Box Girder Bridge," Journal of the
Structural Division, Proceedings of the American Society of
civil Engineers, Vol. 103, No. STS, August 1977.

7.

Scordelis, A.C., Bouwkamp, J.G., Wasti, S.T., and Anicic, D.,

"Structural Behavior of a Skew Two Span Reinforced Concrete
Box Girder Bridge Model," Vols. I, II, III and IV, SESM
Reports Nos. SO-l, SO-2, 80-3, and SO-4, University of
California, Berkeley, Calif., 19S0.

S.

Scordelis, A.C., Wasti, S.T., and Seible, F., "Structural

Response of Skew RC Box Girder Bridge," Journal of the
Structural Division, Proceedings of the American Society of
Civil Engineers, Vol. lOS, No. ST1, January 19S2.

9.

Scordelis, A.C., Bouwkamp, J.G., Wasti, S.T., and Seible, F.,

"Ultimate strength of Skew RC Box Girder Bridge," Journal of
the Structural Division, Proceedings of the American Society
of Civil Engineers, Vol. 108, No. ST1, January 1982.

211

10.

Meyer, C., and Scordelis, A.C., "Computer Program for Prismatic

Folded Plates with Plate and Beam Elements," Structural
Engineering and Structural Mechanics Report No. SESM 70-3,
University of California, Berkeley, February 1970.

11.

Willam, K.J., and Scordelis, A.C., "Computer Program for

Cellular Structures of Arbitrary Plan Geometry," Structural
Engineering and Structural Mechanics Report No. UC SESM 70-10,
University of California, Berkeley, Calif., September 1970.

213

LATERAL LIVE LOAD DISTRIBUTION IN PRESTRESSED CONCRETE HIGHWAY

BRIDGES

Celal N. Kostem, Professor of Civil Engineering

Fritz Engineering Laboratory, 13, Lehigh University
Bethlehem, Pennsylvania, 18015 USA

INTRODUCTION

The design of prestressed concrete highway bridges in the USA

is governed by the provisions of the "Standard Specifications for
Highway Bridges" of The American Association of State Highway and
Transportation Officials (AASHTO) (7). These specifications, or
slightly modified versions, have been extensively used in the design
of bridges throughout the world (4). The AASHTO Specifications are
known to be rather conservative, and in some areas ambiguous. Some,
if not most, bridge engineers tend to implement the provisions faithfully, without due consideration to the first paragraph of the
Specifications, which reads "In any case where the specifications
provide an empirical formula as a design convenience, a rational
analysis based on a theory accepted by the Committee on Bridge
Superstructures of the American Association of State Highway and
Transportation Officials, with stresses in accordance with the
specifications, will be considered as compliance with the Specifications." The flexibility provided by these statements should be
considered by bridge engineers in the review of generally accepted
research findings and their implementation in the design of bridges.
This paper summarizes the results of extensive analytical
research programs which were aimed at the identification of the
design parameters which effect the lateral live load distribution in
simple span highway bridges, with or without skew, and with prestressed concrete I-beams and spread box-beams.
1.1

Design Loading
The design loading employed in major bridges follow the HS

214

20-44 design vehicle (7). For simplified design purposes this loading vehicle is replaced by a lane loading and a concentrated loading. The research reported herein have investigated the practicality of this design vehicle. This vehicle is a pragmatic load.
However, it is known that this loading should be subject to careful
scrutiny in view of the changes in the frequently encountered
vehicular axle combinations and weights (4). The research employed
the design vehicle, rather than the simplified equivalent lane
loadings. Thus, the results reflect the intent of the AASHTO
Specifications more realistically (7).
1.2

Distribution Factor

In the design of bridges beams are individually designed through

the use of the distribution factor, such that the fraction of the
vehicular loading is applied to the single isolated simple beam.
This fraction is called the lateral live load distribution factor,
or simply the distribution factor, and has been defined as "s/k,"
(7), where "s" is center to center spacing of the beams and k is
a constant. If the beam spacing is defined in feet, then for prestressed concrete I-beam bridges the value of "k" is 5.S (s/1.676,
if "s" is in meters). The implication of this formula is that
regardless of the geometry of the bridge, that is, regardless of the
span length, width, skew, etc., the load distribution does not vary
so long as the beam spacing is constant. It has been found by
previous research on the prestressed concrete spread box-beam
bridges without skew that such a simplification can not be warranted
for many bridge geometries (4). More complex, but highly realistic
formulae, were developed and were adopted by the AASHTO Specifications (7). A similar work was conducted for I-beam bridges and the
formulae were developed, pending the official adoption into the
AASHTO Specifications (5,6).
Since the detailed formulae for the new distribution factor
provisions are to appear in the future AASHTO Specifications, this
paper will focus on the pertinent aspects of the distribution fac-
tors but not necessarily the formulae.
2

EFFECTS OF BRIDGE GEOMETRY AND DIAPHRAGMS

The research have indicated that for simple span right prestressed concrete I-beam bridges the design parameters that govern
the load distribution are the beam spacing, as it is already
recognized by the current Specifications, span length of the bridge,
and width of the bridge. (Spacing/span length) ratio is critical
in the definition of the effective participatory area of a given
beam. This ratio also facilitates the consideration of the extent
of the single or bi-directional bending of the deck slab. Furthermore, the width of the bridge also needs to be considered. The

215

width can be taken into account either as is, or through the specification of the number of beams and the beam spacing.
It is also
noted that depending upon whether the load distribution formulae
is to be used for old bridges as a rating tool, or for the design
of new bridges, the number of loaded traffic lanes also needs to
be considered as a secondary factor.
In view of the many independent design parameters that need
to be investigated in assessing their full contribution to the
load distribution, the conduct of a test program using full scale
bridges became a highly improbable proposition. However, an analytical research program that can include the contribution of all
design parameters was expeditiously executed.
In the reported
research it has been found that an analytical experiment can be
designed and executed using a finite element software package.
In
excess of 300 right bridges were analyzed having span lengths from
about 40 ft. (12.20 m) to 140 ft. (42.70 m), and with the number
of beams varying from 3 to 15.
Since the computation of the distribution factors require the definition of the mid-span transverse
influence line for the definition of loaded lanes, each bridge had
to be analyzed for ten different loading conditions.
Through the
full computerization of the input and output options of the finite
element analysis, and automatic file transfers from program to
program, it was possible to conduct the extensive parametric
investigation with great ease, after the development of the required
computer programs (1).
All the theoretical analysis results were benchmarked against
available test results of full scale bridges and the tests conducted in the laboratories under controlled conditions for scaled
down bridges (4). An acceptable agreement, even with the relatively
coarse "production tool" version of the finite element mesh, was
observed, the maximum discrepancy being less than 5% (1).
2.1

Curb and Parapet

The research have indicated that even though in the design of

the superstructure the curb and parapet are not considered as the
primary load carrying members, these structural components still
participate in the load distribution process (1,2).
If the curb
and parapet follow the so-called "New Jersey Barrier" geometry,
about one-half of the moment of inertia of this structural component contributes to the structural stiffness of the bridge.
Thus, maintenance of the structural integrity and provision for
full composite interaction between the curb and parapet, and the
deck slab, will have positive effects on the load distribution.
Figure 1 shows that regardless of how small the contribution might
be, the curb and parapet still provide some stiffness.

216

64'-0

.span
Beam Spacing
Design Lanes

FJ-o"
2-12'-0 Wide

a -Without Curbs, Parapets, Or Diaphragms

b-With Curbs And Parapets Only
e-With Diaphragms Only
d-With Curbs, Parapets, And Diaphragms

1.60-

4-Beam Bridae

5-Beam Brid(le
S/!5.5

cz::

1.20 I-

oz

1= 0.80 I-

be d

abed

abed

abed

abed

;:)
II)

a:

I-

~ 0.40-

Key

Fig. 1

2.2

A
B
B
A
24 fI. Wid. BridQ.

I AI

48ft. Wid. BridQ.

Distribution Factors for the 4-Beam and 5-Beam Bridges

Diaphragms

The structural contribution of the diaphragms is still a

misunderstood phenomenon.
It has been found that if the diaphragm's
stiffness is comparable to that of the beams, and also if the beams
and the diaphragms are monolithically constructed (or post tensioned
in the transverse direction after the construction) the beams and
diaphragms will work as a grillage. However, if the diaphragms
are cast in-situ with the bridge deck, and also if they have limited
depth, the structural contribution of the diaphragms is not as
critical as assumed, but never quantified by the bridge engineers
(1,2). The role of the diaphragms can be noted in Figure 1.

217
Certain bridge specifications require that if the span length
is less than some set length then mid-span diaphragms are to be
used. However, if the span length is larger than this set value,
instead of midspan diaphragm, the use of diaphragms at the third
points is specified. The research have showed that for a real
bridge, with an adverse loading condition as would be commonly
encountered in the traffic stream, the diaphragms at midspan lessen
the portion of the load carried by the beam immediately under the
vehicle (1,6). The effects of the diaphragms are imperceptible
for the beams sufficiently away from the vehicle. Furthermore, at
the maximum moment section of the bridge, i.e. mid-span, more
uniform load distribution can be achieved through the use of midspan diaphragms. Third-point diaphragms do not have any justifiable contribution. Therefore, it is recommended that if any
diaphragm is to be used it should be at the mid-span. Also, if
more diaphragms are to be used due to any reason that may be put
forth, then the diaphragms should be located at mid-span and at
quarter points. These conclusions can easily be observed in
Figure 2.
2.3

Skew

Most bridge specifications tend to leave the inclusion of

the skew in the design computations to the designer. For the skew
angle of 90-75 the bridges can safely be designed as right bridges
(noting that the skew angle of 90 0 corresponds to the right bridge).
In the definition of the skew another parameter that might be
looked into is the distance between the slab and the centroid of
the beam, i.e. "e." Figure 3 shows the effect of the skew and
the eccentricity of the beams on the distribution of the moments
between various beams.
In the figure Beams-A, -B, and -C correspond
to exterior beams, interior beam, and the center beam in a five
beam bridge. The span length was about 70 ft. (21.35 m) and the
bridge carried four traffic lanes. In the figure e/d=O.O corresponds to fully non-composite construction and e/d=0.5 corresponds
to fully composite interaction between the girders and the deck.
It can be noted that depending upon the extent of the composite
interaction there is a substantial change in the torsional properties of the beam, thus the bending moment carried by each beam
changes. From this figure additional observations can be made
(13) :
The beam directly under the load carries a major
portion of the total load as a composite structure.
Increase in moment coefficients in exterior and
center beams is balanced by the decrease in the
moment coefficient of the exterior beam. The
remaining differences are carried by the slab.
The reduction and the rate of reduction in moment
coefficlents for the interior beams seems to be

218

50.0.-----------------.

40.0

~- 30.0
f!?
z

lIJ
U

Li:

Ib

20.0

o Na Diaphragms

A Diaphragms At L/3

IZ

+ Diaphragms At

~ 10.0
o

L/4

X Diaphragms At L/!5

:E

o Diaphragms At Y2

.0

-5.0 L.....!,..------:!:---~:_--::_-___;!:_-~

Roadway Width

Ec
36'-0"

Beam Size Penndot .24/45

Fig. 2

J01

J01

Span
Beem Spacing

g
71'-0"
7'-2"

Influence Lines for Moment, Beam C With and Without Diaphragms

almost the same for both composite and noncomposite

construction.
The effect of the skew in bridges with skew angle
greater than 75 can be neglected. The skew for
bridges with skew angle 75_60 may be neglected,
with the understanding that this will lead to a more
conservative design. At the same time, this results
in adverse plate bending moments at the corners of
the deck slab. For bridges with skew angle less than
60, and especially less than 45, the effect of the
skew can not be neglected.
The correction terms to be applied to the distribution
factors for bridges with non-negligible skew are

219

presented in Reference 1, and a more refined version

can be found in Reference 3.
For skewed prestressed concrete I-beam bridges the distribution factor for interior and exterior beams has the following
general form (2):

where C1 and C2 are constants, sand L are beam spacing and the
span length respectively, and 0 is the skew angle.
The formulae
for the load distribution for prestressed concrete spread boxbeam bridges have a slightly more complex form and the formulae
used for interior and the exterior beams are different (2).

------0-_

OJ2

-a...

0.11

..... .......

0.10

...

Beam C

en 0.09

ffi

0.08

0.07
...80.0&

It

Beam B

Beam A

0.01

45

30

SKEW ANGLE. degrees

f=o.1

Fig. 3

f-o.

~aO
EI
.

t=0.5

-W- '0.035

Moment Coefficients in a 5-Beam Bridge Including

the Beam Eccentricity and Torsional Rigidity

220

DISTRIBUTION FACTORS FOR RIGHT BRIDGES

The realistic distribution factor formulae for simple span

prestressed concrete box-beam bridges without skew has already
been developed by the researchers at Lehigh University and adopted
by AASHTO (7).
In order to provide uniformity between the formulae for spread
box-beam and the I-beam bridges, the following general format of
formula has been employed:

where NL , NB , s, and L correspond to the number of traffic lanes,

number of beams, beam spacing and the span length respectively.
kl and k2 are the algebraic expressions in terms of bridge width,
number or beams and number of traffic lanes; and n is a constant
(5,6). It has been found that even though these formulae are
slightly more complicated then the current s/5.5 expression given
in the AASHTO Specifications (7), they are far more realistic.
The best illustration of these developed formulae, and the
proposed specification, can be seen through inspection of Figs.

2.2
20
a::

1.8

u
~

1.6

14

0
I-

t=

::l

In

ii:

IUI

...

1.2
1.0

Nt.-4
Na=7

NL- 3

0.8

06

1/10

1/8 1/1

1f6

1/4

S/L

Fig. 4

Distribution Factors for Interior Beam

42 Ft. Wide, 7 Beam Bridges

221

4 and 5. These figures contain the finite element study results,

as shown by discrete symbols.
For example, using Fig. 4 for a bridge with span length of 49
ft. (14.95 m), and beam spacing of 7 ft. (2.14 m), it can be found
that the interior beam distribution factors for 3 and 4 traffic
lanes will be 1.24 and 1.42, respectively.
If AASHTO prov~s~ons
are applied, then the distribution factor would be 7/5.5 = 1.27.
The difference at first glance may not be large, however, due to
the large number of bridges that are built of this type, even small
differences will have a contributory effect on the overall cost
and safety of the bridges in question.
A study of Fig. 5 shows that the distribution factors decrease
more rapidly then the decrease in the beam spacing, i.e. to go from
"n+1" to "n" beam bridge.

CONCLUSIONS

Simple span prestressed concrete bridges with in-situ reinforced concrete deck and with prestressed concrete I-beams or
spread box-beams are one of the most commonly used types of bridge

2.2

NB=5

2.0
II:

I-

1.8

1.6

1.4

LL

i=
::l

'!!

1.2

I~

1.0

NB= 6

--;;

NB=8

i"O'

NB=9
NB= 10

.../. )U

0.8
0.6

NB=7

II:
0

-I

1/3

1/30 II.! 0 1/15

S/L

Fig. 5

Distribution Factors for Interior Beam

42 Ft. Wide Bridges (N L = 4)

222

superstructures throughout the world. In the past the design of

the beams has followed the basic distribution factor approach,
using s/k as the distribution factor. The field studies and
analytical research have indicated that a distribution factor of
this simplicity may be good for the preliminary design, but is
too crude for the final design of the bridges. Refined distribution factors should be used. Formulae for both right and skewed
bridges have already been developed and verified, and are available in the literature (1,3,5,6).
The curb and parapet, depending upon the construction practices,
do help in the distribution of the load. Their contribution should,
preferably, not be taken into account in the design process, but
for the maintenance inspection of the bridges attention should be
paid to the structural integrity of these components.
The shallow diaphragms that are not built as a part of the
total bridge structure contribute marginally in the lateral distribution of the live load. If diaphragms of any type are to be
used, there should always be a midspan diaphragm.
To verify the completed design of a bridge "reverse design"
process, i.e. the use of s/k, should not be employed. Instead,
consideration should be given either to the new formulae or to
the finite element analysis of the total superstructure, if the
structure is critical enough and/or if a number of similar bridges
are to be built.

1.

2.

3.

4.

DeCastro, E. S. and C. N. Kostem. Load Distribution in Skewed

Beam-Slab Highway Bridges, Fritz Engineering Laboratory Report
No. 378A.7 (Lehigh University, Bethlehem, Pennsylvania, USA,
1975) .
DeCastro, E. S., C. N. Kostem, and D. A. VanHorn. A Pilot
Study of Factors Affecting Live-Load Distribution in PreStressed Concrete Beam-Slab Bridges, Fritz Engineering Laboratory Report No. 387.4 (Lehigh University, Bethlehem,
Pennsylvania, USA, 1979).
DeCastro, E. S., C. N. Kostem, D. R. Mertz, and D. A. VanHorn.
Live Load Distribution in Skewed Prestressed Concrete I-Beam
and Spread Box-Beam Bridges, Fritz Engineering Laboratory
Report No. 387.3 (Lehigh University, Bethlehem, Pennsylvania,
USA, 1979).
Zellin, M. A., C. N. Kostem and D. A. VanHorn. Structural
Behavior of Beam-Slab Highway Bridges-A Summary of Completed
Research and Bibliography, Fritz Engineering Laboratory Report
No. 387.1 (Lehigh University, Bethlehem, Pennsylvania, USA,
1973).

223

5.

6.

7.

Zellin, M. A., C. N. Kostem, D. A. VanHorn, and J. M. Kulicki.

Lateral Distribution of Live Load in Prestressed Concrete 1Beam Bridges, Fritz Engineering Laboratory Report No. 387.2A
(Lehigh University, Bethlehem, Pennsylvania, USA, 1975).
Zellin, M. A., C. N. Kostem, D. A. VanHorn, and J. M. Kulicki.
Live Load Distribution Factors for Prestressed Concrete I-Beam
Bridges, Fritz Engineering Laboratory Report No. 387.2B (Lehigh
University, Bethlehem, Pennsylvania, USA, 1976).
Standard Specifications for Highway Bridges, The American
Association of State Highway and Transportation Officials
(Washington, D.C., USA, 1977).

225

SKEW SLAB BRIDGES

Christian Menn
Professor fur Baustatik und Konstruktion
Eidgenossische Technische Hochschule
Zurich

SUMMARY
The applications, typical dimensions, structural behavior,
analysis methods, and ultimate strength considerations of skew slab
bridges are presented.
The effects of cracking and inelastic behavior on the
deflections and servicabili ty of skew slab bridges are reviewed.
The need for prestressing in order to control behavior and provide
servicability is discussed.
Theories of elasticity and plasticity are discussed with
respect to their applicability to the design of skew slab bridges.
It is shown that an inherent stress condition can be superimposed on
the computed ultimate section forces with no effect on ultimate
strength.
Finally, the relationships necessary for
slab bridge for ultimate strength are presented.

designing a

skew

226
1.

INTRODUCTION

Skew slab bridges are especially suitable for overpasses with

limited construction
depth. The material
outlay
for
these
structures is relatively high, but their construction is extremely
simple. Slab bridges with a span of 20 m or less for simply
supported systems, or 25 m for continuous systems are economical.
For larger spans, the available construction depth generally
determines whether the cross section is designed as a plate or a
hollow box girder.
1.1

Dimensions and Cross Sections

The plan dimensions of a skew plate (Fig. 1) are labelled as

follows: Span 1, width b, angle of skew ~.
Various types of cross section are possible (Fig. 2). Both
solid plates and voided plates are used, with and without lateral
overhangs. Solid plates of not more than 80 cm thickness are
suitable. Correspondingly the span over which a solid skew plate is
appropriate is one of 25 to 40 m for simply supported bridges,
depending on the relationship between lib and ~ (Fig. 3).
1.2

Load Carrying Behavior

The load carrying behavior of a skew plate is characterized by

the fact that as the width b increases, the direction of the larger
principal moment in the middle of the plate increasingly deviates
from the bridge axis and approaches the perpendicular to the skew
support lines (Fig. 4).
Stress concentrations in the obtuse corners result from the
fixity of the free edges in the rigid support lines. This does not
significantly influence the behavior of the plate. Through
admissible cracking or limited plastic deformation in the obtuse
corners, the stress condition in the plate is only slightly altered.
In addition to economic considerations and crack behavior, the
anticipated deflection of the plate determines whether or not
prestressing should be employed.
The cracking moment MR = (bh2 /6) f ct for slab bridges is
approximately five times smaller than the dead load moment in the
middle of the plate. Therefore, normally reinforced slab bridges
should be considered as thoroughly cracked, and they behave in
principle like a normally reinforced girder.
The deflection of a girder is computed as follows:
or

227

Fig. 1

FIG. 1:

DIMENSIONS OF A SKEW PLATE

~.:::::::::::::::::;::::::::::::::::::::::::::

6:::::::::::::::::-::::::::::::::::::::::::::::::::::::1

It)o:c.to::f~AJ

1.0.0.:0.0:.0]

Fig. 2

FIG. 2:

TYPES OF CROSS SECTIONS

[m]
40

30

20}:::===========

<p = 30'
<p = 60'
<p = 90'

10

Fig. 3

FIG. 3:

RECOMMENDED SPAN FOR SIMPLY SUPPORTED BRIDGES

228

m2
-------

- - - - me
y

6~O~

lo"b = 0.63

J0.3;;J

t/m2

~x
mt

a) narrow plate

lo/b= 1.67
.,

b) wide plate

narrow
plate

m1

ml

0.42l

()(

wide

plote

m1

ml

()(

1.000

0.164

290

1.060

0.016

22.0

1048

-0.104

18.3

-0.096

16.8

0.448

-0480

51.0

1.088
0.272

-0.304

475

ql~18

ql:/8

[0]

ql~/8

ql~/8

[0]

c) points of design

Fig 4
FIG. 4:

LOAD CARRYING BEHAVIOR UNDER UNIFORMLY

DISTRIBUTED LOAD

229
Consequently, l4/h3 is a deflection index which gi ves an
indication of the extent of the deflection. An 80 cm plate with a
span of 20 m exhibits a deflection value of

which is a very high value. A girder with a slenderness ratio l/h =

20 acquires such a high value only when the spans reach 40 m or more.
Of course no one would design such a girder in normally reinforced
concrete. Thus slab bridges of substantial span should always be
prestressed. Normally reinforced plates frequently suffer inadmissibly large deflections.

2.

CALCULATION OF SECTION FORCES IN SKEW PLATES

The controlling position of live load on a skew plate depends

on the pattern of reinforcement and is different for each
reinforcement direction. Simplifications are therefore indispensable. Generally the section forces at a few characteristic points in
the plate (Fig. 5) are calculated for several specific load
positions. These results are then used to obtain a simplified
reinforcement pattern for the slab.
A plate element is subjected to the following section forces:

The shear stresses due to the shear forces v x and v

are
usually small and generally require no special reinforcement? The
determination of the necessary reinforcement is primarily dependent
on the moments mx ' ~, ffix y '
2.1

Theory of Plasticity

Any reliable analysis of the ultimate limit state must be based

on the theory of plasticity, since an accurate non-linear theory,
taking into account the real behavior of the structure past the
quasi-elastic range, is too complicated for practice.
A check
follows:
where

for

ultimate

strength

is

normally

formulated

as

<
associated safety factors:

dead load
live load
characteristic concrete strength
characteristic steel strength

'14

F IG . 5 :

114

AND
OF DESIGN
TIC POINTS
IS
R
TE
C
A
R
A
CH
ENT ZONES
REINFORCEM

LI2

--------

_ _ -1. ~

b /6

<::)

IN

tv

231

The ultimate strength check may be expressed as

design value of forces

design value of capacity.

For the calculation of the design internal forces, the static

method of the theory of plasticity must be used, since it yields a
lower bound on the value of the ultimate load capacity of the system.
The kinematic (yield line) method of the theory of plasticity
for slabs is unsuitable and should be avoided. Particularly in the
case of skew slabs it is very difficult to prove that the chosen
mechanism does not overestimate the ultimate load, in other words
that the chosen mechanism fulfills the conditions of plasticity.
This method is both obsolete and dangerous since it yields an upper
bound on the ultimate load capacity of the system.
According to the preferred static method, the determination of
the reinforcement may be based on any distribution of internal forces
which satisfies the conditions of equilibrium.
Today the internal forces of a relatively difficult system such
as a skew slab are almost always calculated with a finite element
computer program. The programs are based on the theory of elasticity
in that they assume linear elastic material behavior. Under high
stresses this assumption is no longer valid. As a result of
cracking, the rigidity of the structure is altered, resulting in a
redistribution of internal forces. An inherent stress condition is
therefore superimposed on the elastic stress conditions. This
reduces high local stresses and exploits the structure's available
load carrying reserves.
2.2

Inherent Stress Conditions

The internal forces under an inherent stress condition are not

a consequence of a load. The number of independent inherent stress
conditions of a system corresponds to the degree of statical
indeterminacy. A plate, as an infinitely statically indeterminate
system, exhibits an infinite number of independent inherent stress
conditions.
Inherent
follows:
for beams:

iN

dx

stress

conditions
or

avy
avx
for plates: - - +
dx
ay
Such an inherent

or

are

d2 M
dx2

32 mx

defined

in

principle

as

a2m

+ 2.~ +
ax dy
ax2

a2 m

stress condition can result

from

support

232

a)

support displacement

I I

I I I

J1

.Jf

b)

I I qo

qo' - qo acting on different static systems

Fig 6

FIG 6:

INHERENT STRESS CONDITIONS

233

displacements (Fig. 6a) or curvature of elements, due for example to

variations in temperature. An inherent stress condition also
results when two different statical models of the same system are
loaded with ~ and - ~ respectively (Fig. 6b), and the results are
superimposed.
According to the theory of plasticity, an inherent stress
condition does not change the load carrying capacity of a system.
Thus, in the calculation of the ultimate capacity of a system, any
state of inherent stresses may be superimposed on the stresses
resulting from an elastic analysis of the structure under the applied
loads (Fig. 7).
Therefore the design forces may be formulated as follows:

where F d,E

represents the design force according to the

theory of elasticity.
represents an inherent stress condition.

or:

where F d,i

2.3

represents the design force according to statical

model number 'i'.

Prestressing and Servicability

In principle, the design forces must be calculated in such a

way that the corresponding reinforcement provides optimum structural
behavior with respect to servicability.
For normally reinforced structures this is customarily the
case when the design forces are computed on the basis of the theory
of elasticity and the stiffness of the homogeneous concrete cross
section. Nonetheless it is possible to superimpose an inherent
stress condition which takes into consideration the arrangement of
the reinforcement and approximates the stiffness contribution of the
cracked concrete section.
In prestressed structures the redundant moments due to
prestress generally create a favorable inherent stress condition
which should be taken into account in the design of the
reinforcement.
The

redundant

moments

represent

part

of

the

entire

234

i(Z----------#~B~7
R
~
L_._B
A

h4=f-'-'-'-'-'-'-f
'I

l = 2b

.............

_----

my

m x y .
Section

A- A

Section

Fig 7

_e_e_e_e_.

FIG 7:

Fd

DESIGN INTERNAL FORCE DISTRIBUTIONS

B-B

235

prestressing moment

where IDap
mzp

~:

- p'e
represents the redundant moment due to prestress.

For plates the total prestressing moments m., are generally

computed with the same computer program as used lor the applied
loads, with the prestressing represented by anchorage forces and the
deviation forces of the tendons. This yields the redundant moment

The value chosen for the width over which a prestressing tendon
is considered effective is not of great importance. More important
is the necessity for using the same width (for example b p '" 2' h)
throughout the design calculation.
In the formula for the design ultimate forces, the redundant
moments due to prestressing (an inherent stress condition) are
generally included with a factor of 1.0 (Fig. 8). Taking into
account a possible moment redistribution, however, this factor may
actually vary from about 0.8 to 1.2.

This inherent stress need not necessarily cOLrespond exactly

to the redundant moments from the effective prestress. The inherent
stress can also be determined from a fictitious prestress. This
fictitious prestress must not, however, differ substantially from
the effective prestress, particularly with respect to the reinforced
zone and the section of the tendons.
A proposed prestressing arrangement
supported skew slab is shown in Fig. g.

3.

for

narrow

simply

DESIGN CHECKING FOR ULTIMATE STRENGTH

The reinforcement in a skew slab bridge normally consists of

prestressing tendons and a grid of normal mild steel reinforcement.
proportioning of the reinforcement for strength generally
requires several trials in order to satisfy the design requirements.
Each trial proceeds as follows.
First, the mechanical reinforcement content in the x and y
directions is estimated based on the computed internal forces and the

236

+=~l==~~~';;;;;";';;;;;========l+

L = 2b

-----

............ _.-

-.- -

Section

8- 8

~
-'-'

Section

A- A

Fig 8

----1.8 m

g+p

_._-_._. 1.8 m

g+p

FIG. B:

+ 1.0 m
zp

DESIGN INTERNAL FORCE DISTRIBUTIONS CONSIDERING mzp

237

A-

...A/~=============/47

k ' - / - - - - - - - - - - - - - - - -...// /

/
/

-c

_I

A-A

B-B

F'-'- _._._._._._.-. .,

Hypothetical arrangement
of prestressing for the
calculation of mzp '

h
AlBIC

Fig 9

FIG. 9:

PRESTRESSING ARRANGEMENT FOR NARROW SIMPLY SUPPORTED SLABS

238
chosen reinforcement directions.
W

x'

y'

xy

A fictitious reinforcement in the principal directions is then

calculated (Fig. 10).
Wi;;

Wx oCOS 2 C01 + WyOsin2 C01 + Wxy osin2 C01

W
11

Wx oSin2col + Wy ocos2COl - Wxy osin2 <P1

From this fictitious reinforcement,

resisting moments may be calculatedo
Wi;;
~i;; d2 .f cu .W( 1~)
w
~ll
d2 ofcu OWll O( 1 y )

the

ultimate

internal

The design ultimate applied moments, expressed as

m*

1.a Omg+q + (1.0 7 1.3)Omzp

must now be transformed to the principal directions for comparison

with the provided ultimate internal resisting moments.

m~

m:ocos2COl + m;osin2COl + m:yOSin2C01

m*
11

*
2 COl + nyocos
*
2 ~1
mxosin

m* = 21 (1Il_* i;;ll
:t
0

~)

~y sin

2 oC0 1

osin2<Ol +

According to the theory of plasticity, safety against collapse

is guaranteed by satisfying the following inequalities:
positive moments:
(~i;;

*
* - mi;;)O(m
ull - mil)

~~
> 0
>0

Negative moments:

> 0

239

FIG 10:

MOHR'S CIRCLE FOR THE TRANSFORMATION OF THE

MECHANICAL REINFORCEMENT

IV. DYNAMICS AND VIBRATION

243

FLUID-STRUCTURE INTERACTION IN ANALYSIS OF BRIDGE PIERS. PART I

THEORETICAL PRELIMINARIES

Nuri Akka and

~etin Y~lmaz

Department of Civil Engineering

Middle East Technical University, Ankara, TurKey

ABSTRACT. The theoretical preliminaries needed for the analysis of

fluid-structure interaction are presented. After the derivation of
tne wave equation which governs the response of an acoustic fluid,
the effect of the surrounding fluid on the response of a spring-mass
system is studied. A submerged rigid cylinder, which may be considered
as the model of a bridge pier, is subjected to a sinusoidal excitation. The hydrodynamic force acting on the cylinder is calculated
analytically.
1.

INTRODUCTION

The dynamic analysis of structural systems immersed in fluid

media is of significant interest in many fields of engineering.
Fluid-structure interaction occurs in the earthquake response of
bridge piers. The dynamic response of such immersed structural
systems is coupled to the fluid response through dependence of the
effective mass, damping, and stiffness of these systems on the
acceleration, velocity, and displacement of the fluid field,
respectively. In the range of frequencies characteristic of the
free vibration of these systems, the primary effect of the fluid ~s
to increase the effective mass of the structural system. In the
analysis of the forced or transient response of these systems, the
effects of the fluid on the effective damping and stiffness of the
system may be of considerable importance also.
In the present work we are concerned with the analysis of
fluid-strpcture systems in which the physical behavior of the fluid,
to a first approximation, is characteristic of an acoustic medium.
Tne constraining assumptions are that the fluid is inviscid,

244

irrotational, compressible and subject only to small displacements

from or oscillations about some equilibrium position. The behavior
of the structural systems is assumed to be linearly elastic. In the
presence of fluid free surfaces, it is generally considered consistent
with the above assumptions to use the simplified, high frequency
approximation to the free surface condition which corresponds to
the absence of gravity waves. Within these constraints, the analytical
methods described in the present work may be routinely applied to
approximate the vibratory response of submerged bridge piers.
As piers respond to earthquake ground motions, they cause
corresponding motions in the water in contact with the pier surfaces.
This results in changes of water pressure acting on the piers which,
in turn, affect the dynamic response of the piers. This coupled
hydroelastic phenomenon is termed fluid-structure interaction.
In the following discussion, the equations of motion governing
the response of an acoustic fluid medium are derived from the
principles of conservation of mass and momentum. Similar derivations
can be found in many textbooks (Lamb, 1945). Following the derivation of the governing equations, the analytical solutions for two
simple problems are presented. The applicability of analytical
teChniques to realistic fluid-structure interaction problems is
limited to simple geometries. The solution of more realistic problems
is feasible generally through numerical teChniques which will be
the topic of Part II of this presentation. Analytical solutions,
on the other hand, are not completely useless. They provide one
with an insight to the interaction phenomenon. The experience and
insight gained from analytical solutions prove to be extremely
useful in the interpretation of the results obtained from numerical
techniques.
2.

EQUATIONS OF FLUID MOTION

The equations governing the motion of a homogeneous, isentropic,

clewtonian fluid medium are given by (Temkin, 1981)
( 1)

(2)

where
p
p(t) = density of the
u = u(X, Y, Z) = (ul, u2'
velocity of the fluid
Fi= body force at a point
0ij= dynamic stress in the

fluid
u 3)
at a point (X, Y, Z)
(X, Y, Z) of the fluid
fluid at a point (X, Y, Z)

245

Finally,
D( ) _ ~+

at

1ft -

a()
ax.-

uj

(3)

is the material derivative. As usual, repeated indices imply summation.

In characterizing the behavior of an acoustic fluid, we assume
tnat the fluid is initially at rest with uniform density, p. The
fluid is then disturbed by some or all material elements being
displaced small amounts with corresponding changes in density. The
fluid is subsequently allowed to return to some equilibrium position
and oscillate about it. The gradients of both the fluid velocity and
density are assumed to be negligibly small. Thus a consistent
approximation to equations (2) is seen to be

au.

ClO

~J

( 4)

Pat=~+Fi
J

The dynamic stress, Oij' in a Newtonian fluid can be related

to rate of strain in the fluid by the relationship
0 .. =
~J

-pO ..
~J

21l(e ..
~J

- 3" ekko ij )

(5)

where Oij is the Kroenecker delta function, 11 is the viscous shear

coefficient, and eij are rates of fluid strain.The pressure is p.
For a homogeneous, isentropic, elastic medium the stress, Tij
is related to the material strain by
T ..
~J

KEkko ..
q

2G(E .. - -3 Ekko .. )
~J

~J

(6)

where K is the bulk modulus and G is the shear modulus of the

material. Following the assumptions which characterize the acoustic
fluid medium, we may further approximate the behavior of the fluid
as being analogous to that of an elastic medium with zero shear
modulus. We write
ij

KE:kkO ij

(7)

Stress is now defined as positive in compression and E:kk is the

divergence of a material fluid volume given by
(8)
-+

where h

-+

h(X,Y,Z,t)

246

from equilibrium at a point (X,Y,Z).

In order to make use of the above relations we must return to
equation (4) and ex~ress the fluid velocity in terms of the
displacement field h. We write
(9)
+

for which we have assumed the body forces, F, to be uniform

tnroughout the fluid domain and have set them equal to zero.
On combining equations (7), (8), and (9) to eliminate either
displacement or stress we obtain, respectively,
(10)
and
2 "2+

cf(V h

curlcurl h)

(11)

where
(12)
is the speed of sound in the fluid.
Equations (10) and (11) are both forms of the wave equation
for which various methods of solutipn are well established. Equation
(10) is a scalar equation in one unknown, p, whose solution yields
the dynamic pressure field in the fluid domain. Equation (11) is a
vector equation in one, two, or three unknowns governing the timevarying displacements of the fluid medium.
Recall that equations (10) and (11) have been derived under
the assumption that the acoustic fluid is compressible. This implies
that the speed of sound in the fluid, cf, is finite. If the fluid
is assumed to be incompressible, which is a reasonable assumption
for water, the speed of sound becomes infinitely large. In this
case, the response of the fluid is governed by either one of the
following equations

o
V2 h +

curlcurl

(13)

h=

( 14)

In the following sections, the pressure dependent formulation

of tne wave equation, equation (10), will be used to show the effect

247

of the interacting fluid on the response of a structure.

3.

SINGLE-DOF MASS-SPRING SYSTEM

I'"

r--.

Fluid

~IC

Fig. 1

A subrN>l'g~ singl. dogrH - of - fr dom

mass - spring syst.m

Consider a single-degree-of-freedom mass-spring system interacting with an acoustic fluid medium as shown in Fig. 1. The hydrodynamic pressure acting on the face of the mass is denoted by p.
Assuming that this face has a unit area p is also equal to the
hydrodynamic force acting on the mass. The displacement of the mass
m from its position at rest is denoted by x. The spring constant is
k. The mass is assumed to be at a distance L from the origin of the
y axis fixed in space.
The equation of motion governing the displacement x(t) of this
idealized structure is

mX

kx

(15)

= -p

in which dots denote differentiation with respect to time. The

response of the acoustic fluid, under the assumptions stated in
the previous section, is governed by the classical one-dimensional
wave equation

a2 q,
ay

1"

(16)

;r=Z<P
cf

in which q,(y,t) is the velocity potential. The fluid is assumed to

be compressible. The hydrodynamic pressure on the mass is related
to the velocity potential through the following relation :
p

-p~

at y = L,

(17)

in which p is the mass density of the fluid. The kinematic (or

interface) boundary condition between the mass and the fluid states

248

that the velocities of the mass and the fluid particles at the
interface must be equal so that no gap can occur between the two.
This condition is stated mathematically as follows :
at y = L

(18)

The assumption that displacements are small has been used in equation
(18). From equation (16) it is seen that we need two boundary
conditions on in y. Equation (18) provides one of them. The second
one is obtained from the fact that, for propagating waves, the
radiation condition must be satisfied at y = 00. This implies that
as y + 00, must go to zero. The definition of the problem will be
completed by assigning the initial conditions which are taken as
x

0,

= V,

0,

at

0,

(19)

in which V is a constant.
The problem is now well-defined and the solution can be obtained
by various means. The Laplace Transform method used in the present
work yields the following
x=

~ e -a.t SinQt

in which
Q = w

i _a~

w=1f
m'

(20)

(21a)
(2lb)
(2lc)

Note that w is the natural circular frequency of vibration

(rads/sec) of the single-degree-of-freedom mass-spring system in
vacuo. The frequency of the same system interacting with the fluid
medium shall be called the modified frequency and denoted by Q.
It is obvious from a study of equation (20) that the solution
is similar to that for a single-degree-of-freedom mass-spring-damper
system in vacuo. In other words, the fluid acts as a damper as
expected. The term a/w corresponds to the well-known damping ratio
~ = c/2mW of structural dynamics (Chopra, 1981). Here c is the
damping coefficient for the structure. The presence of the fluid
has the effect of decreasing the natural circular frequency of
vibration and, thuS, increasing the natural period of vibration.
One significant difference between the response characteristics
of a submerged mass-spring system and a mass-spring-damper system

249
in vacuo is the following: According to Chopra (1981), damping
ratios less than or equal to 0.2 constitute a range which includes
most structures in vacuo. Thus, damping effects on the frequency and
period of such structures are negligible. On the other hand, the
same can not be said for a mass-spring system in a fluid medium.
The a./w for tne interacting system is rIOt necessarily less than
0.2. It is commonly larger than 0.2 and it may even be larger than
1.0. Recall that, for p=lOOO kg/m J , c f =1500 m/sec and w=200 rad/sec,
a.
pc f
3750
m

ThU1l, for a mass m > 3750 kg the a./w ratio will be less than 1.0.
For smaller masses it will be even larger than 1.0. The gE'neralization stated by Chopra (1981) is, therefore, not applicable for
damping caused by the presence of the fluid. The difference between
structural damping and fluid damping can be explained as follows :
Recall that

t,;

and

= 2rrw

a.
w

(22)

which are the damping ratios for a damped structure in vacuo and a
structurally undamped structure in fluid. t,; depends on-cne-ratio
between c and m wnich are both related to the properties of~
same structure. On tne other hand, a./w depends on the ratio between
the quantities which are related to the structure and the surrounding
fluid. Thus, it is, in principle, possible to Come up with a./w
ratios which are much different from corresponding ~ ratios. In case
a./w turns out to be greater than 1.0, tne solution becomes
x

witn

V e -at Sinh rlt

IT

/a.~
L
W

(23)

_ 1

rhis solution is not even sinusoidal.

As a second example, let us now study the response of a single
degree-of-freedom mass-spring system interacting with a fluid
subjected to base motion. The governing equation is

mx +

kx

(24)

in which xB is the displacement of the support shown in Fig .1. The

fluid response is governed by the wave equation (16) as before but,
now, it is more advantageous to use it in its plane wave form
Cl<!>

iF;

1
cf

=0.

(25)

250
I

Tne kinematic boundary condition becomes

at

(26)

y = L,

berause the fluid particle velocity is equal to the absolute velociLY

of tne mass. Using equation (17) we obtain the hydrodynamic force as
at

y = L.

(27)

Tnus, tne final form of the governing equation

rnX + pCf~ +

kx = -rnx B - pcfx B

LS

(21:3)

Let us assume that the base displacement is

xB=Be

Wt

(29)

Tne solution will be of the form

x = x e

irlt

(30)

substituting equations (29) and (30) into (28) we obtain the real
part of x and it is
2 2
pC f 2
2
P cf
rl) --2] Cos~lt + liJf W Sinrlt}
m
(31)
Real x
2
t2
[ (w

Here,

W =

Nm as

~
m

before.

Nmv cO',ldder tne r('3por.se of a single-degree-of-freedom massspring-damper system in vacuo subjected to a base motion. The
governing equation is (Chopra, 1981)

mx

cx

kx

-mxB

(32)

A comparison of equations (28) and (32) shows that, in fluidstructure interaction problems with base motion, base velocity x
enters the equation in addition to the base acceleration xB. Thi~
is another important difference between the responses of a structure
in vacuo and of that in a fluid medium. The damping force proportional
to tne base velocity is added to the damping force proportional to
tne relative velocity of the mass. Having Ui.sc1lssed seme significant
aspects of fluid-structure interaction using a single-degree-offreedom system, we can now consider a slightly more complicated and
more realistic model which is studied below.

251

4.

SUBMERGED LONG RIGID CYLINDER

A long rigid cylinder, surrounded by an infinite fluid medium,

subjected to a rigid body motion in the rz-plane as shown in
Fig. 2.The motion of the cylinder is described by the velocity Vg .
The fluid is compressible. Small vibrations of the fluid are
governed by tne wave equation in cylindrical coordinates :
LS

d2 <Jl

1 Cl<Jl

Cl 2 <Jl

r or

~8-L

-+--,,-+-,,-~

or 2

0 -

1 Cl <Jl
------.,.~ L
c 2 at

(33)

- - t -__ Vg

Fig. 2

subm~

long rigid

cylinder

in which 1> is t~1.: vel.ocic} pctential and cf is the speed of sound

in the fluid. Fluid particle velocity in radial direction is Cl1>/br
and the hydrodynamic excess pressure is p = -p3/Clt where p is the
fluid mass density.
Boundary conditions are
V

at

r = a,

(34a)

252

3 = 0
3r

as

r -+

(34b)

00.

Here, a is the radius of the cylinder. Note that Vg corresponds to

tne rigid body motion and for a steady state analysis we have :

-iwt

(35)

From (34a) :
3
ar
=

v g CosO e

-iwt

at

(36)

Accordingly,
(r,O,t) = l(r) CosO e

-iwt

(37)

Substituting (37) into (33) :

<P"
1

+ .!:.

+~

CP' _ 1 <p
r12l
r

21

=- 0

(38)

cf

in which the derivatives are with respect to r. The solution of

(3d) for outgoing waves is (Reuwood, 1960).

(39)

in wr.ich .\ is a constant to be detE'cn:ined fro1!\ ("I.e boundary condition at r = a and Hfl) is the Hankel function of the first kind
and fi~5t order. The general solution of (33) is, therefore,
(1)

<P(r,8,t)=A Hi

Wr
(-) CosO

-iwt

(40)

cf
The boundary condition (36) yields

A = v / {~
g dr

H(l) (wr)}
1

cf

(41)

r = a

The hydrodynamic pressure at r

r=a

- ipw{}

a is

r=a

Substituting (40) and (41) into (42), one obtains

{Hil) (t;;)

}t;;~-wa/c

f
- - - - - - - - - " - - l.:os8 ,
d
(1)

{d~ Hi
in which

S=

wr/c f

(t;;)}'=wa/c

Recall that (Redwood, 1960)

'"

( 43)

253
= H(l) (s) -.!:.. H(l) (0
l o s 1

~ H(l) (0
ds

( 44)

Substituting (44) into (43), one obtains

H(l)(wa/c)
1
f
p = ipc f Vg
(1)

cf

{Ho

(1)

(wa/c f ) - wa HI

------ Case

( 45)

(wa/c f )}

Call the terms on the right hand side of (45) S, except for the
Case term. Thus,
( 46)

Pa = S Cos8

Tne total hydrodynamic force, PT , acting on a cylinder of height

H is

P
T

4aHSf

11/Z
a

= TIaHS

cosZe de

( 47)

Finally,
iTIaHpc

{H(l) (wa/c f
a

~
Hl(l) (wa/c f )}
wa

( 48)

The following two approximations bring the above equation into a

simpler form to handle.
a) High frequency approximation
It is assumed that wa/c f 1. In this case, the second term
in the denominator in (48) can be neglected. Thus,
H(!) (wa)
1

cf

( 49)

H (1) (wa)
a
c
f

For

1, from Abramowitz and Stegun (1965),

exp{i(~ - ~) }
4

cf

-iTT/Z

-1

(50)

254

substituting (50) into (49) one obtains

PT = naHpc V
f

(51)

Assuming that the highest frequency that is encountered in earthquakes is about 10 cps; thus, for (19) to be valid, one must have
2nfa/c f
Tnis yields,
cf
a 21Tf

or

24 m.

This is very unrealistic. Therefore, high frequency approximation

~s not a reasonable one.
b) Low frequency approximation
It is assumed that walcf
to the following form

1. In this case, (40) can be brought

(52)

This yields the added mass results because (na 2Hp) is the mass of
the fluid displaced by the cylinder. For this assumption,

and this snJ'I~ that, for cEasonahle frequencies, realistic dimensions

will always satisfy the above equation. Therefore, the low frequency
approximation,AIhich is equivalent to the added mass approach, is
satisfactory for many practical cases.
After noting the limitations of the high and low frequency
approximation as given above, now let us go back to the exact selut ion (48). Le t
PT = inaHPc f Vg /1.

(53)

l.n 'AIhich

H( 1 ) (wa/c )
1

(54)

Using the followiTlg identities (Abramowitz and Stegun, 1965)

H~l) (z)

= J)z)

i Y)z)

(55)

255

we obtain
J 1 -I- i Yl
/\ =

------,-

(J

(56)

iY ) - ~(J
o
z 1

l.n which z = wa/c f and the argument of t,le Bessel functions has
been d:opped. Equation (56) can be brought to the following form
Jl

A=

i Y1

(57)

-----,-----------,,-

(J

~
z

J \

l'

iCY

~
z

Y )

By multiplying and dividing (57) \"ith complex conjugate of the

denominator one obtains
1\

+
y

is

(5tl)

in which
(59a)

(59b)
(59c)

Accordingly, (53) becomes

PT

iTIaHPC f Vg

(a

+
y

is)

(60)

Using equation (35) one obtains

PT = TIaHpc f vg(Sinwt
Let

~ =

TIaHpc
=

iCoswt)

(ex

+y

is)

(61)

v. Thus

~{(~ Sinwt - ~ Coswt) + i(~ Coswt + ~ Sinwt)

y
Y
y
y

(62)

Finally,
(63)

in which

.
= -~( O,Sl.nwt
y

-(O,Coswt
y

- SCoswt) ,

(64a)

(64b)

SSinwt) .

l'londimensionalize PT as follows :

256

na Hpg

(65)

'

in which g is the gravitational acceleration. Let X = na 2Hpg which

~s the weight of the fluid displaced by the cylinder of height H.
Then,
(66)
~n

which

.
-to ( aS~nwt

(67 a)

SCoswt),

XY
to
-- (aCoswt + SSinwt) .
XY

(67b)

The amplitude o! Re(PT) is obtained by setting wt = Arctan(-a/S)

and that of Im(PT) by setting wt = Arctan(B/u). The results are

i Re(PT) i

X~{aSin(Arctan

iIm(PT)i =

~{aCos(Arctan(~ + SSin(Arctan(~}

(- F

- SCos(Arctan(-

B-} ,

(68)
(69)

A computer program has been coded for the calculation of the above
results. The numerical results obtained from the approximate solution (52) and the exact solution (68) are compared in Fig. 3. It is
seen that, for the range of frequencies realistic for earthquakes,
the low frequency (or the added mass) approximation yields results
very close to the exact solution.
5.

SUMMARY AND CONCLUSIONS

The effect of the surrounding fluid on the response of a singledegree-of-freedom spring-mass system is studied analytically. An
analytical method of solution is used also for the steady state
response analysis of a submerged rigid cylinder. The flexibility
of the cylinder has not been considered to keep the analytical
solution feasible. The solution of more realistic problems, in
wnich, for instance, flexibility of the pier is considered, generally
requires numerical techniques which will be the topic of Part II of
the present paper.
Acknowledgement: This work has been supported by Yliksel Proje
architect-engineers for the Karakaya Dam Bridge project. The
authors are indebted to Mr. Erdal Erkakan, project manager.

A.~.,

257

~
~

15

"'.>'"

.~

.3

lit

cb

'"

tl
0

)(

""

;I

"

~
~

...J

'"

""'
~
.....

""

'!I.!I.

"0
0

c:

.~

Ill."""

258

REFERENCES
1. Abramowitz, M., and Stegun, I.A., (editors), Handbook of
Mathematical Functions, Dover Publ., New York, 1965.
2. Chopra, A., Dynamics of Structures, A Primer, EERI, California,
1981.
3. Lamb, H., Hydrodynamics, Dover publ., 6th edn., New York, 1945.

4. Redwood, M., Mechanical Waveguides, Pergamon Press, Oxford, 1960.

5. Temkin, S., Elements of Acoustics, John Wiley and Sons, New York,
1981.

259

FLUID-STRUCTURE INTERACTIW IN ANALYSIS OF BRIDGE PIERS. PART II:

APPROACH

l~UMERICAL

<;:etin Ydmaz and

i~uri

Akka

Department of Civil Engineering

Middle East Technical University, Ankara, Turkey

ABSTRACT. The development of fluid finite element is presented.

Eartnquake response analysis of a submerged thick cylindrical shell
is studied. The response of the same shell in vacuo is also
investigated. Tne effect of the fluid-snell interaction on the
displacements, stresses and the base moment of the snell is
discussed. It was noted that the presence of tne fluid may increase
tne base moment by quite a significant percentage.

1.

INTRODUCTION

In Part I of tile present work the theoretical preliminaries

needed for the analysis of fluid-structure interaction have been
presented. In Part I it was stated that the solution of realistic
problems is feasible generally through numerical techniques. The
finite element formulat ion for the numerical analys is of the fluidstructure interaction problems will be presented in this part of
tne worK.
l~umerical solution met,lOds which have been applied to interaction problems were initially dependent on approximate decoupling
of the structural response from the fluid motion. Solutions for the
fluid response were based on potential theory (Geers, 1971 ; Clark,
1973) or, in the case of snock response, on plane and spherical
wave approximations (Bedrossian et al., 1972, 1972). Tnese
approximations nave the advantage of not requiring expl icit or
discrete modeling of the fluid volume. They are, however, limited
to applications in wilich the structure is deeply submerged and the
effects of tue free surface are negligible. Zienkiewicz and Newton
(1909) developed acoustic-fluid finite elements based on a Galerkin

260
approximation to the wave equation using pressure as the dependent
field variable. This meti10d was used to compute the natural
frequencies of submerged structures in finite fluid regions and in
unbounded fluid regions in which the fluid-structure interaction
effects are negligible beyond finite distances from tne structure.
This method was extended by Hunt (1974) and Zarda (1976) to applications in wi1ich the far field effects are important by a procedure
which matches finite element solutions at finite external fluid
boundaries with known analytical solutions for infinite fluid Cases,
A similar formulation is presented in (Nastran, 1976) and implemented
in the finite element, structural analysis computer program, NASTRAN.
In this Case tne finite element equations were derived from variational metnods using a form of the wave equation in which the fluid
pressure is the dependent field variable. Pressure analog methods
nave also been developed which use either "lumped" (Everstine et
al., 197) or consistent (Schroeder and Marcus, 1974,1976) formulations equivalent to the previous finite element pressure methods.
The pressure analog methods are derived from an analogy between
fluid pressure and one of the Cartesian displacement components in
finite element formulation of classic linear elasticity. These
methods have been implemented using readily available finite element,
structural analysis programs such as iiastran.
Further investigations have produced structurally compatible,
acoustic fluid finite elements for linear two-dimensional analysis
based on pressure as the dependent variable (Nahavandi et al. 1975;
Marcus, 1973). Bathe and Hahn (1979) have implemented two-and
three-dimensional, acoustic fluid elements in the structural
analysis program ADIHA which is capable of nonlinear and large
deformation analyses. These methods require the addition of separate
acoustic fluid finite elements to the element libraries of existing
finite element, structural analysis programs.
In all of tne preceding finite element methods, the boundary
conditions at the fluid-structure interfaces are satisfied such that
the acoustic-fluid finite elements are compatible with structural
elements. The coupling of the fluid-structure interaction is
achieved directly throug;1 the assembly of the system mass, damping,
and stiffness matrices. The generality and versatility of the
finite element method in purely structural applications is thus
extended to the acoustic fluid-structure application with relative
ease.
This present work represents an extension of techniques
developed in (Akay et al., 1977; Akka et al., 1979) for the
implementation of two-dimensional and axisy~tric, acoustic fluid
finite elements in the structural analysis, finite element program
SAPIV. The finite element formulation is derived from the application of the method of weighted residuals using Galerkins criterion
to a form of the wave equation based on fluid displacement as the
independent field variable. Implementation of acoustic fluid finite

261

elements is achieved with a m~n~mum of computer programming and

without addition of separate elements to the existing SAPIV library
by recognition of the equivalence between the above displacement
formulation and the classical linear elasticity formulation used
by SAPIV.
In tt1e derivation of acoustic fluid fm~te element equations,
several crucial assumptions are made which approximate the physical
properties and behavior of the fluid medium. The most important of
these assumptions in regard to the structural compatibility of
acoustic fluid elements, is that fluid displacements are small and
oscillatory about some equilibrium position. This assumption allows
the fluid stress tensor to be derived based on an analogy between
the stress state in a linearly elastic medium and that of an
acoustic fluid medium. In order to further simplify the governing
equations of fluid motion, the acoustic fluid medium is assumed to
be inviscid, irrotational, and isentropic. Changes in fluid density
due to compression of the fluid are assumed to be negligibly small.
In the following discussion, three different finite element
formulations are derived using either variational or weighted
residual methods. Two of these formulations use pressure as the
dependent field variable and the third uses displacement. The
displacement formulation is shown to be equivalent to the classic
linear elasticity formulation commonly used in finite element,
structural analysis programs such as SAPIV. The present method for
implementation of acoustic fluid, finite elements in SAPIV is then
discussed.
2.

PRESSURE - DEPENDENT FINITE ELEMENT FORMULATION

Finite element formulations for acoustiC-fluid elements may be

readily obtained from equation (10) of Part I by either variational
methods or the method of weighted residuals (Huebner, 1975). In the
classical variational formulation, the problem is to find the unknown
function or functions which extremize (maximize, minimize) or make
stationary a functional or system of functionals subject to given
boundary conditions. In order to apply the method we must obtain a
variational statement for the continuum problem described by
equation (10) of Part I. We begin by rewriting this equation in the
form

pp -

->-

KIJ

->-

IJp =

(1)

We define the sum of the terms on the left hand side of equation
(1) to be the gene ral ized force,
J

->-

->-

= pp - KV Vp

(2)

262

For tne case of snock or pressure wave propagation in an inviscid

fluid initially at rest, the flow field may be described as
nomentropic. Hence the volume integral of the product of the
variation of pressure, op, times the generalized force will produce
a variation of tile generalized energy:
(3)

We may substitute tne identity

-+

-+

-+

-+--+

-+

'l . {op(V'p)} "' op'l . 'lp + V'( op) . 'lp

( 4)

into equation (3) to obtain

-+

'lp dxdydz

-+

-+

(5)

- f~KV' .{op(V'p)} dxdydz

Applying the divergence theorem to the ti1ird integral in equation

(5) produces an integral over the surface s,
-+

-+

f~K'l. {op('lp)} dxdydz "'

-+

-+

Ids KopV'p . n ds

where tne vector ri is tne outward unit normal to tile surface s. Tne
first and second integrals in equation (5) may be rewritten using
tne ident it ies

(3t)

dP 2

..

dp

-+

-+

"' 2 dt (2f)op "' Zpop,o(V'p.V'p)

-+-+

= 2'lp

.V'(op)

(7)

Applying equations (6) and (7) to equation (5) results in

oH

1
2
o{/~(2 P (p)

1-+

+ 2 KV'p
-

-+

V'p) dxdydz}

-+

-+

fdS Kop(V'p).n ds

(d)

Tue generalized energy internal to tile fluid is the value of

tne first integral of equation (5) in which we can identify a term
representing kinetic energy and a term representing potential energy.
Tne energy flow tilrough tile boundary is given by the second integral.
In the finite element method of solut ion, tile solution domain
is now divided into m elements of r nodes. We then express the
benavior of tlle unknown function p within each element as

(9)

263

where p. is tne nodal value of p at node i. A stiffness matrix

Ikij I i~ formed from the internal potential energy
V =

-;.

-;.

f rI 2" Kl7p. I7p dxdyd z

( 10)

by tne equation
2

d V _
dP dp. 1- J

K .

1-J

( 11)

A ma5S matrix 1m. 1 1-S formed from tile internal kinetic energy

1-J

T = f rI

2
p(p) dxdydz

(12 )

by the equation
2

m .

1-J

_d_T_ = f p i~ .l~. dxdydz

dp . Clp.
rI
1- J
1- J

(13)

Tile set of simultaneous equations describing the fluid pressure at

discrete points may then be written in matrix form as
Im.1 Hi.} + Ik .. 1{p.} = {f.}
1-J
J
1-J
J
1-

( 14)

wnere {p.} is the vector of fluid pressures at the nodal points.

The matrix {fi} is the vector of generalized forces transmitted to
tne fluid and is given by the equation

acl.

f.
1-

acl.

1-,,, + __
1-N +
f "sKN(
a
i ax I X
Cly y

aN.

az1-N)d
Z
S

(15)

An alternative to the preceeding formulation is given by the

method of weighted residuals. The resulting equations governing
tne fluid pressure field are identical to those previously derived.
Tne mathematics of the weighted residual formulation are, however,
a bit more direct and we do not have to rely on the abstract
identification of energy terms as was previously necessary.

Applying the method of weighted residuals involves basically

two steps. The first step is to assume the general functional
behavior of t11e dependent field variable so as to satisfy the
boundary conditions and to approximately satisfy the governing
differential equation. Substitution of this approximation into the
governing differential equation and boundary conditions then results
in some error called a residual. This residual is required to
vanish in some average sense over the entire solution domain.

264

The second step is to solve the equation or equations resulting

from the first step and thereby specialize the general functional
form to a particular function. which then becomes the solution to
tne field problem.
Thus we want to find an approximate functional representation
for the field variable p governed by the differential equation
2

K'V P -

~
dt

;0

L(p)

;0

(16)

in the domain ~ bounded by the surface s and subject to the

boundary conditions
(17)

on fluid-structure interfaces. and

..?E.;o 0
dn

->-

on tne fluid free surfaces. In equation (17). n is the outward unit

normal from the fluid-structure interface and ~n is the normal
acceleration of the structure surface.
The unknown exact solution p is approximated by

Ii '"

m
l: NiPi
i;ol

(18)

where the m functions iL are chosen to satisfy the boundary conditions

(17) and the m parametefs Pi are unknown. Substitution of Ii into
equation (16) produces
(19)

where R is the error or residual that results from approximating p

by P. Minimization of R over the domain ~ is accomplished by forming
a weighted average of the error which is made to vanish over the
solution domain. The error distribution principle most often used
to derive finite elements is Galerkin's method which requires that
i;ol.2 ... m

(20)

Since the equation (16) is valid at any point in the solution

domain. it also holds for any collection of points defining an
arbitrary element of the whole domain. Thus the functions Ni may be

265

recognized as interpolation functions Ni(e) defined over the domain

of the element and the Pi are approximate nodal pressures.
Thus from Galerkin's method we can write the equations
governing the behavior of an acoustic-fluid finite element as

51

(e)

N. (e)
1

a2 (e)

a2 (e)

ax

ay

'i

a2 (e)

(e)

IK(---?-z + ---?-z + ---?-z)-p ~ I

az

at

dxdydz Ce ) = 0,

i = 1,2, .. ,r

(21)

wnere r is the number of nodes for the given element. We now write
(22)
Upon substitution of equation (22) into equation (21) and integrating
the Laplacian terms by parts we obtain (dropping the (e) superscripts)
t

I "s K11~1 I7p.~ds - I

o

ClN.
K(_1

aN.

"N

I~I +_1
ax ax
ay

eM.

,,-.T

I~I

ay

+ az1 I ~z I) {p} dxdydz

- In pNilNI {ii} dxdydz

(23)

or, in matrix form,

(24)
In writing equation (24) we have implied the following definitions.
m. .
1J

k .. =
1J

I"

f. =
1

51

"

p N .i~. dxdydz

(25)

1 J

a,L

ai~.

dX

ax

1 _J
KC_

al~.

aN. aN.
_1

ay

aN.

_J
dy

aN. aN.
_1

dZ

_J) d d

az

x ydz

(26)

aN.

I "s Kl'l. (-,,1 n + -,,1 n + -,,1

n Z) ds
0
1
oX
X
oy
y
oZ

(27)

We observe that this is the same result as obtained previously in

equations (11), (13), and (15).

266
3.

DISPLACEMENT-DEPENDENT FINITE ELEMENT

FORMULATIOi~

We now proceed to derive the finite element equations of fluid

motion for acoustic fluid elements based on the displacement
formulation, equations (11) of Part I. The method of weighted
residuals using Galerkin's criterion is applied to the boundary
value problem defined by these equations, the kinematic boundary
condition

-+

-+

h = s

on s

(28)

and the dynamic free surface boundary condition

+-+

V.h

on fluid free surfaces

(29)

-+

where sn is the normal velocity of the structure surface. We begin

by writing the field displacements as
r
(e)
{h}(e) =
l:i~o(x,y,z),h oCt)
hx(x,y,z ,t)
1 l
Xl
r

h Cx,y,z,t)
y

IN (x,y ,z)h

h/x,y,z,t)

ILL (x ,y ,z) h

1 l
0

yl
0

(t)

Zl

t)

r)
J

(30)

~quations

(11) of Part I may be rewritten as

Cl 2h

,?h

Cl 2 h

Cl 2 h

Clx

ClxCly

ClxClz'

Clt

a2 h

Cl 2 h

Cly

CiyCiz

j{C--;' + __
x + __x)_p--;, =
Cl 2 h
K(--y

ClxCly

+ ~+ --y)

'1

Cl 2 n

Cl 2
+ __
z +

dXClz

ClyClz

K( __z_

,lh

(31)

-p----f = 0

(32)

Cl 2 h
Cit

Ci 2h

---T-)
-P---T=0
Ciz
Cit

C33)

267
Applying the method of weighted residuals using Galerkin IS criterion
to equation (31) yields (dropping the (e) superscripts)
,?h

3Zh

32h

:lh

3x3y

dXdZ

IS-Pi IK(~ + __
x + __
X)_ p~1
3x

dxdydz

(34)

dt

Consider, for a moment, the integral of spacial derivatives

in (34),

dZh
__
x) dxdyd z

dXdZ
d
dh x
= I nl~ ~ (-" ~ dX
Clx

ah

ah

+ ---..:. + ~)
ely

(35)

dxdydz

dZ

We now subst i tute for ilx from equat ions (30) and inte grate
(35) by parts to obtain

IN. ( I ClN I +
dS

dX

I dN I
dy

_ I

I dN I){h }n ds
dZ
x x

~.

r;:

_...2( I 31~
dX

dX

"

+ I dN

dy

+ I dl~
3z

I ) {h } dxdydz
x

( 36)

In a similar manner, after applying the metilOd of weighted

residuals with Galerkin's criterion to equations (3Z) and (33), we
obtain the results

IN. (I eM
dS

dX

+ I Cli~ I + I ,h~
dy

- I

3H.

r;:

I ) {h }n

3z

"'"

ds

'"

'"

~
I~I
+ I~I + I~I
dy
dX
dy
dZ

) {h }dxdydz
y

(37)

and

I",

oS

d. ( I ~N I
~

oX

+ I oy
~N I + I aZ
~i~ I
- I

3N.

r;:

'"

) {h}n
Z

-~
(I~I
+
dZ
dX

ds

IClNI +

dy

I~Nzl
a

){h }dxdydz
Z

(38)

Substituting the results (36), (37), and (38) into equation

(34) and the corresponding expressions derived from equations (32)
and (33), respectively, yields the following result.

268

f as
~ K.t~.L

3N
I"I
aX

3N I + I,,)
3N I {h}.n
+ I"oy
oZ

ds

3N;

3N;

dN;

I ~N I

~N

~N

dX

3y

dZ

dX

3y

dZ

- f K(-~ + -~ + -~)( -"'--- + 1-"'---1 + 1-"'---1

II

){h}

(39)

The first integral in equation (39) may be identified as the

vector of generalized forces, {f.}, upon substitution of the
relation
L
( 40)
We write
{f. } =
L

( 41)

~ N. {p }ds
as
L
n

wnere {p } are the generalized normal pressures on the surface of

tile elerrRnt.
Furthermore, it can be snown that the second integral Ln
equation (3~) may be written as

IK LJ.. 1

fll

IB.IL t lei IB.I

J

wuere

IB L1

3N.
L

3X

0
31~

.
L

dY

3i~

.
L

3N.
L

dxdydz

( 42)

o1

( 43)

0
3N.
L

8Z

3Y

8X

az

3H.
L

3N.
L

3Y

ax

3N.
L

8Z

3N.
L

269

and
IC I

( 44)

Defining
1m.l.J1

= f~p

( 45)

N.N. dxdydz
l. J

allows us, finally, to write the simultaneous equations (31) to

(33) in the form
Im.1 {h} + Ik .. 1{h}
l.J

l.J

{f.}

(46)

l.

Cast in this form we can see that if we replace the constitutive

matrix

Icl

(l+\!) (l-2v)
I-\!

\!

I-v

\!

I-v

(1-2v) /2

(1-2v)/2

Sym.

( 47)

(1-2v) /2

of classic linear elasticity with that of equation (44), the

classic linear elastic finite element formulation beco~s equivalent
to the displace~nt dependent, acoustic fluid finite ele~nt
formulation.

270
4.

k~

APPLICATION

k~D

CONCLUSIONS

The numerical method described in the previous sections is

now applied for the analysis of a bridge pier. With the impoundment
of water in the Karakaya Dam, currently under construction in Turkey,
a part of railway linking the cities of Malatya and ElazLg will be
inundated. A new route with an approximately 2.1 km long crossing
over the reservoir has been planned. The earthquake analysis of
tnis mUltispan railway bridge on the Euphrates River has already
been reported (Glilkan et al., 1982). Determination of the design
earthquake ground motion from seismic hazard and geotechnical
studies and derivation of the equivalent static force spectrum for
design purposes nave been presented in the mentioned reference .
The emphasis in this work will be on the dynamic response analysis
of the bridge piers including hydrodynamic effects. In the numerical
analysis of the problem a computer program coded for earthquake
response analysis of axisymmetric tower structures surrounded by
water is used (Liaw and Chopra, 1973). The program is based on a
pressure-dependent finite element formulation and the fluid is
assumed to be incompressible.
A typical tall pier elevation and the pier cross section are
illustrated in Fig.l. It is obvious that under operational conditions
water will seep into the center of the box section, so part holes
will be left in the walls to facilitate this action. Thus, it becomes
necessary to consider the water in the box section in the finite
element modelling.
As seen in Fig. 1 the pier cross section is of a box type. On
the other hand, the computer program utilized is for axisymmetric
structures only. It is necessary, therefore, to transform the actual
cross section to an equivalent circular section (Fig.2). This can be
achieved by requiring that the equivalent circular s8ction have
cross sectional area and moment of inertia equal to those of the
actual section. It should be noted that the actual section has two
different moments of inertia about its principal axes. We will have,
indeed, two different equivalent circular sections. However, in the
present work, the results for one of the equivalent sections only
will be given. The actual pier, its equivalent model and the
corresponding finite element model are shown in Fig. 3.
Figure 4 illustrates the first three
with and without water. The corresponding
in the same figure. It is noted that each
decreases when the effect of the water is

mode shapes for the pier

frequencies are also given
one of the frequencies
taken into account.

A probability study of past seismLcLty, the known fault

structures in the vicinity of the site and the attenuation
characteristics of earthquakes in Turkey led to the establishment
of O.16g as the effective peak ground acceleration on firm ground

271

fOr the present problem. It was also decided tnat the Taft record
of 1952 and the EI-Centro record of 1940 modified with respect to
tne appropriate peak acceleration should be employed in the studies
of the superstructure and its interaction with the fluid medium.
Tne use of two different earthquake records allows one to study the
effect of the frequency content of an earthquake on the response of
the system. In Fig. 5 the north-south component of the EI-Centro
record of 1940 which is modified for 0.16g is given. The Taft record
~s not snown but it is also modified for 0.16g.
Tne maximum responses of the pier model subjected to the
modified Taft earthquake and to the modified EI-Centro earthquake
are illustrated in Figs. 6a and 6b, respectively. In these figures,
maximum displacement, maximum bending stress and maximum shear
stress are plotted along the length of the pier for both the in
vacuo and submerged cases. The effect of the fluid interaction-is
clearly seen in these figures. For all the cases considered, the
presence of tne fluid amplifies the response of the structure. The
maximum responses resulting from the modified EI-Centro record are
generally higher than those resulting from the modified Taft record.
The displacement at the top and the mOment at the base of the
pier subjected to the modified EI-Centro record are given in Figs.
7 and 0, respectively, as functions of time. The corresponding
responses for the modified Taft record are given in Fig. 9 and 10.
The amplifying effect of the presence of water is clearly noted in
these figures. Respective comparisons of Figs. 7 and 9 and Figs. 8
and 10 reveal the effect of the frequency content of the earthquake
record on the structural response.
The numerical results given in this section are for a specific
pier model only. The results indi~ate that the presence of water
interacting with the structure affects the response of the latter
significantly. The maximum responses are, in sOme cases, almost
doubled. However, it should be emphasized that these conclusions
are valid for the specific pier considered only. For another pier
with different cross sectional dimensions and/or height the effect
of water may turn out to be not so much significant. Thus it is
suggested that each case be studied on its own and the conclusions
tnereof should not be generalized.
ACKNOWLEDGMENTS
The work described in this paper was performed in conjunction
witn consultancy services provided to Yliksel Proje A.$., architectengineers for the project. The encouragement of Mr. Erdal Erktakan,
Project Manager, is appreciated. The sections on the finite element
formulation are essentially based on a report by Mr. J.G. Bennett of
tne University of Michigan submitted to N .Akka as part of the
requirements for a graduate course.

272

6i7.719

~
.:1.619

Sl.....-

IL

..........

698.084

~
~.284

~!I.OOO

693.000

.~

~.

A-A

~.OOO

I. ' I f

'f/lfff

Ifill

C - C
FIG. 1 ELEVATION AND CROSS SECTION OF THE PIER

273

t-.
ACTUAL

EQUIVALENT

FIG. 2 ACTUAL AND EQUIVALENT CROSS SECTIONS

:sz

..

WATER LEVER

I:

I:
I
II
II
III
II
II

II
II
II

II
11
II

-1

I
I
I
I
I

--I
44(

EQUIVALENT PIER

I
I
I
I
I
I
I
I
I
I
I
I
I

m.

z
40

37

34

31

28

25

22

19

16

13

10

13

12

11

10

41

38

35

32

29

26

23

20

17

14

11

Is

:;

42

39

36

33

30

gr

24

21

18

B.

26

25

24

23

22

21

20

19

18

17

16

15

14

FINITE ELEMENT MODEL

5. 3 m.

~
L

ACTUAL AND EOUIVALENT PIERS AND THE FINITE ELEMENT MODEL

51 m.(16732 3 It)

..'\7

'"

FIG. 3

ACTUAL PIER

II
II 1"II

II

I"

II

II

II
II
II
" I"II
II
II j II
II
II
II
II
II
II
II III
II
II
II
II
II
II
II
II
II
II
II

II
II
II
II
II
II

11

8 It )

+>

-.)

40

37

34

28

..!1!

..!.

..!Q

...1

--!

13

12

11

10

11

21

24

33

30

.!!

J1

....

41

26

!I

r
r
I

,~

MOOE 1

r
,,

MODE 2

0>-- -

IN VACUO

M00E3

-0

0.10 0.20 0.30

FIG. 4

1
2

CASE

0.10 0.20 0.30

FREQUENCY (CPS)
MODE 1 MOllE 2
MODE 3
10.03 T :51.401 118.40
7.98
40.28
91.:54

-0.30 -0.20 -0.10

0.10 020

MODE SHAPES AND FREQUENCIES

WATER LEVEL
MIlE
OUTSIDE
ZERO
I ZERO
_MAL
NORMAL

-0.30 -020 -O.ll

42~__~4=~========~==~~~~F=~~========~==~==~~F=~-=====__

3825 ~

3$

23

129

22

26

23

32

19

20

20

17

18

17

, 14
15
16

14

....,

IV

vo

276

....

o
o
a:
o
~
a:

40l

37

~l

31

28!

2S!

22~

19!

16;

~!

10~

7~

4!

13

12

11

10

19

21

./

42~,

FIG. 60

3e

J11

41

1\'

"

21 2 42 4 848 1060
BENDING STRESS (k~/e",)

"I

.........

IN VACUO

WITH WATER

1.40 :1 80
g.60
SHEAR STRESS (k~/em')

~-----o

0----00

MAX4MUM RESPONSES OF THE PIER TO THE MODIFIED TAFT RECORD

260 S20 t80 10.40

DISPLACEMENT ( em)

II,

II

24

II

3O

27

24

/
/

.
.
:

.I

~
I

-I!

21JI

18 II

IS

12

..j

..!

38 25

3:1

32

23

29

22

26

23

18'

17

20

20

17

14

11

16

15

141

1 :I

tv
...,
...,

4' ~

3[

3~

3!

2!

2~

16!

.!Q!

13

12

11

10

26

25

24

23

22

21

20

19

18

17

16

gil

---I

42

!!

24

3!

18

~.IO

: :
7.6~ 10.2

DISPLACEMENT ( cm )

2.~~

iI

~JI

I
I

~J r

225
675
1I.2~
8ENOING STRESS ( kg/cm~ )

0---0
0 - _ _ -0

IN VACUO

WITH WATER

3.77
11.31
SHEAR STRESS (kg/c",)

.,

II

'I

FIG.6b MAXIMUM RESPONSES OF THE PIER TO THE MODIFIED EL-CENTRO RECORD

41

38

32

29

26

23

20

17

14

11

1 ~ 141
15
2

4~

00

--.J

- 9

- 6

- 3

31

i:1

'

FIG. 7

t .,
,

:~

:Ii'1':i1'l

II

"II'

" I',I

"

,III
(, ,t

em.

[,

WITH WATER

(%

5 DAMPING)

IN VACUO

hrne(sec)

DISPLACEMENT AT TOP OF THE PIER SUBJECTED TO THE MODIFIED

EL - CENTRO RECORD

DISPLACEMENT

tv
.....

\0

30.000\

- 5Q<XXj

_40000\

_ 20.000 \

_ 10.,

100001

20.000\

300001

40 000\

50=\

FIG. 8

"

I.

,,

{%

5 DAMPING}

IN VACUO

WITH WA1'ER

MOMENT AT BASE OF THE

PIER SUBJECTED TO THE
MODIFIED

EI.-- CENTRO RECORD

"\':I ':1:\:' 1\1'

\ ~ ~1!\ \'~ U\I r
11111 '\1\llr
1111
',III,.'"
,\'1"\'I 1'1'1:

MOMENT Tm.

_ 6

- 3

'3

f'I

I
I

I.

"

;~

I
I'

em,

FIG, 9

:,

',I

~~

"

I',

\I

TAFT

",'

,>;11'

i'

, ,

RECORD

l >il\
'

.... '~:c' r',,! ' ," ':

'.v''V' ' ' 'V' ;1: ::,'
'1':, I I', ' ,

::

! \ :: \

:\

DISP LACE MEN T

'.

Ii

'

'

!.: \:if

' ,I,

' I
"
,

hi

::\1

'

"
,!'
.I,

'I'

(0/0 5 DAMPINGl

'/11TH WATER

IN VACUO

' '

i~
'

'

>,'
I 11~I\i, ' ,

",

',j'

;1,

, ,

'llme(s

: V' I

'~\

"

"

' \ , ~ ,{
I: , :

____ -

"!

\.

,! ,: I

,:".', i
>"
"
',21
I

!,

__

~'

\~"

"

:' \

I'

! 'I

(i.

I\.~ 0.i' !:'N ,:'

U'JJ:
, I,:;~'\;I~'." ''"":\~I",\i \

I:

eel

00

'

- :i0.000

- 40.000

- 30.000

-20.000

-Io,ooof

10.000

20000

\!

'

\11

'II

~
I

II

I I

rill)'"
I

if'J

It
il

I'

,I

I,"
I

I,

\1

V:: :1~ \Y

,i "

I' I: ::

'I

1t-.;lt{III::'~

:11:
. I' , If. ,
I
I I I I i Iill, I"'
I
I

II I 'i~'i\

II!

, " L
I,,,

'\'
I
I'PII
,! ' I

1 "

I \

IiI

II

'!

(0/0

~ OAMPING I

IN VACUO

WliH WATER

I'

If

rV ~ I

,Ill I
II' " I I
II
J,
I
"
, ' I II I
I '/
~
I

\~\"

II I, ~
11111",
1:1 " II II JI
, '
t
I II II /1" I
,
~
II /1 II
I
I
!
f I 11
I I ' \ /1
I
J

D
MOMENT AT BASE OF TO PIER SUBJECTED TO THE MODIFIE
RECORD
TAFT

:
I

II
I I

I , II II

I ,\ ,
' Ii I
I,

:'/1
jl"11
'I'
,

'i'!

1\

::I!I~:

\,

Tm.

'II!~I"

'II

FIG. 10

"V ''\ l'

';

: 1\

MOMENT

l
oo,oooL,j,~~~'\1

30,000

40,000

50000

I
\

\}

'~i

,I

l\7Ttime (SeCI

'.

t..>
co
t..>

283

REFERENCES
1.

Akay, H.U., Akka, i~., and Y~lmaz, <;., "Analysis of Solid-Fluid

Interaction Problems with SAPIV," Proc. Second Sap Users Conf.,
Paper No. 13, University of Southern California, 1977.

2.

Akka, N., Akay, H.U., and Y~lmaz, C;., "Applicability of

General-Purpose Finite Element Programs in Solid-Fluid Interaction Problems," Computers and Structures, Vol. 10, 1979,
pp. 773-7133.

3.

Bathe, K.J., and Hahn, W.F., "On Transient Analysis of FluidStructure Systems ," Computers and Structures, Vol. 10, 1979,
pp. 303-391.

4.

Bedrossian, Bedros and Di Maggio, Frank L., "Transient Response

of Submerged Spheroidal Shells ," Int. Journal Solids Structures,
Vol. d, 1':172, pp. 111-129.

5.

Bedross ian, Bedros and Di Maggio, Frank L., "Acoust ic Approximations in Fluid-Shell Interactions ," Journal Engr. Mech. Div.,
A.S.C.E., Vol. 93, No. EM3, June, 1972, pp. 731-743.

o.

Clark, A. V., Jr., "A Study of Fluid-Structure Interaction and

Decoupling Approximations," Naval Research Laboratory Report
7590, 1973.

7.

Everstine, G.C., Schroeder, E .A., and Marcus, M.S., "The Dynamic

Analysis of Submerged Structures," Nastran : Users' Experiences,
NASA TM X-3273, National Space and Aeronautics Administration,
1975, pp. 419-429.

d.

Geers, T.L., "Residual Potential and Approximate Methods for

Three-Dimensional Fluid-Structure Interaction Problems ," Journal
of the Acoustical Society of America, Vol. 49,1971, pp. 15051510.

9.

Glilkan, P., Erdik, M., Akka, N., Y~lmaz, <;;., and Qner, M.,
"Seismic Analysis of a Multispan Railway Bridge: A Case Study,"
Proc. 7th European Conference on Earthquake Engineering,
September 20-25,1982, Athens, Greece. Vol. 6, pp. 163-170.

10.

Huebner, K.H., "The Finite Element Method for Engineers," John

Wiley and Sons, Inc., 1975.

284
11.

Hunt, J.T., Knittel, M.R., and Barach, D., "Finite Elemont

Approach to Acoustic Radiation from Elastic Structures,"
Journal of tne Acoustical Society of Amorica, Vol. 55, 1974,
pp. 269-280.

12.

Liaw, C.Y., and Chopra, A.K., "Earthquake Response of

Axisymmotric Tower Structures Surrounded by Water," EERC 73-25,
Un ive rsity of Cal iforn ia, Be rkeley, 1973.

13.

Marcus, M. S., "A Finite Elemont Method Appl ied to ti1e Vibration
of Subm2rged Plates," Journal of Ship Research, Vol. 22, No.2,
June 1978, pp. 94-99.

14.

Nanavandi, A.N., Bomm, G.J., and Pedrido, R.R., "Structurally

Compatible Fluid Finite Element for Solid-Fluid Interaction
Studies,"Nuc1. Engng. Design, Vol. 35, 1975, pp. 335-347.

15.

"The NASTRAl.'l Theoretical Manual, II NASA SP-222(03), ,-lational

and Space Administration, Washington, D.C., 1976.

Aeronaut~cs

10.

Schroeder, E .A., and Marcus, M.S., "Natural Frequencies of a

Submorged Ring," CMD Report 27-74, David W. Taylor "aval Ship
Research and Developmont Center, 1974.

17.

SChroeder, E .A., and Marcus, M.S., "Finite Elem2nt Solution of

Fluid-Structure Interaction Problems," David W. Taylor Naval
Ship Research and Developmont Center Report 76-0145, 1976.

lS.

Zarda, P.R., "A Finite Element-Analytical Method for Modeling

a Structure in an Infinite Fluid," Nastran : Users"
Experiences, NASA TM X-327S, National Aeronautics and Space
Adminis trat ion, 1976, pp. 251-272.

19.

Zienkiewicz, O.C., and l~ewton, R.E., "Coupled Vibrations of a

Structure Submorged in a Compressible Fluid," Proceedings,
International Symposium on Finite Elemont Techniques, Stuttgart,
1969, pp. 359-379.

285

VIBRATIONS AND DYNAMIC BEHAVIOUR OF ACTUAL BRIDGES

E. ZELLER
Associate Professor
Department of Civil Engineering - Institut Universitaire
de Technologie - Universite STRASBOURG-III (FRANCE)
INTRODUCTION
Bridges have been built since ancient times. As
crossing and linking needs increased and developed,bridges have
multiplied and improved.
In the beginning, the know-how was closely related to
previous experience. From the point-of-view of materials used, as
well as the knowledge of structural behaviour and an understanding
of building techniques, bridges sum up the science and know-how
of an era.
If we observe the evolution of the most important
structures in history, we will be able to follow up the series of
progressive innovations and improvements. Moreover, it can clearly
be seen that those bridges nowadays "old", were modern and even
ahead of their time since the most advanced scientific and
technical elements had to be worked out and then assembled.
Nowadays, industry regularly provides high specification
materials, and present technology supplies powerful means for a
quick setting up and construction. But undoubtedly the most
outstanding progress in the last twenty years is due to the
elaboration of mathematical modelling of bridges, thus allowing
be,..tter analysis.
If the hand solution procedures have permitted predicting fairly accurately the deflection and the stresses of
structures understatic or quite static loads, when it comes to
deal with complex structures or carry out a dynamic analysis, the

286

long numerical calculations required are only feasible by

computer.
While the Finite Element Method and the processing power
of computers were being developed, giving way to a real structural
language, new measuring techniques have appeared which, together, by
means of data acquisition and processing, enable us to closely
determine the real structural behaviour of bridges and other
structures.
If a bridge is subjected to a dynamic excitation load,
the response varies with the time; the practical vibration
analysis must take into account all mechanical properties of the
system.
Generally the designer is most interested in obtaining
the maxima of the response data: stress and deflection, in order
to chose the appropriate section and rigidity. Nevertheless it is
also interestjng to note that the dynamic problem does not have
a single solution which can be approximated in the unique term of
the theoretical maxima but all the elements of the response history
are of interest.
Even if the objectives of a designer differ from those
of a mechanical engineer, many aspects of the dynamic mechanism
of a bridge are of direct interest to him. In fact the assessment
of vibrational behavior and durability of bridges is the result of
2 types of investigation:
- observation and measuring of dynamic characteristics
of bridges in service or subjected to dynamic tests.
- in laboratory simulated tests on bridge elements or
material.
Due to the extent of the subject of investigation large variety of structures, duration of a measuring period, cost
of material ... - the present results go beyond the possibilities
of an engineer of the Civil Engineering Department of the
University of Strasbourg. Thus the author could count on the
experience and cooperation of some researchers, responsible for
tne advanced investigation carried out in the different Laboratories and Institutes.
~M.De

Back J. Kolstein, M.H. Van Leeuwen J.L.

University of Technologie, DELFT (Netherlands),
Lewin Institut.

MM.Caracilli J., Theuillout

Laboratoire Central des Ponts et
PARIS (FRANCE)

Chauss~es,

LCPC,

287

Mr Cochet D.and Guillard Y.

Laboratoire Regional des Ponts et Chaussees,
LCPC, STRASBOURG (FRANCE)
Mr Gregoire R.
Centre d'Etudes Techniques de l'Industrie Mecanique,
CETIM, SENLIS (FRANCE)
Mr Lieurade H.P.
Institut de Recherche sur la Siderurgie,
IRSID, St-Germain/Paris (FRANCE)
Mr Matsumoto Y.
University of Tokyo, Japan
Mr Pfeiffer M.R.
Frauennofer Institut fUr Betriebsfestigkeit
LBF, DARMSTADT (GERMANY)
Mr Van Koten H.
Lab. for dynamics of structures
TNO, DELFT (Netherland).
CONTENTS
Introduction
I. Method of measurement and analysis process
1. Dynamic test on actual bridges
2. Analysis process of measuring data
3. Dynamic response components
II. Dynamic behaviour of bridges subjected to traffic
1. Simplified approach
2. Current research
a) road traffic loads on a steel bridge
b) railway bridges
III. Dynamic test on actual bridges
1. Dynamic test with vertical impulse. Auscultation of fullconcrete slab bridge
2. Sinusoidal excitation test
IV. Dynamic test in laboratory
1. Accumulative damage
2. Damping
3. Mechanical interaction railway-longspan suspension bridge

288

v.

Simulation and modelling

1. Choice of mechanical properties of bearings in order to
improve the safety of a boxgirder bridge subjected to
seismic vibrations
a) mechanical characteristics of bearings
b) choice of elastomer bearings in order to improve the
mechanical behaviour
2. Effect of profile imperfection of the bridge subjected to
moving loads

CONCLUSIONS
I. METHOD OF MEASUREMENT AND ANALYSIS

PROCES~

1. Dynamic test on actual bridges

A vertical force sufficient to produce meaningful
amplitudes in bending can be applied in the central parts and in
the symmetrical plan. According to the needs, the excitation force
must be perfectly well-known, so the measurement of the response
characteristics produces the transfer function of the mechanical
system.
Otherwise, it will be sufficient to generate a movement
by any means available, and then analyse the natural evolution
(amplitudes, modes, frequencies, damping). In case of a random
excitation due to road or railway traffic or wind, the analysis
can only be carried out in statistical terms. The range of dynamic
tests for the acquisition of the mechanical characteristics is
shown in figure 1.
2. Analysis process of measuring data
For all measurements on actual structures, the apparatus
and techniques for the acquisition and processing of data have to
be made as compatible as Dossible. The complete equipment must be
able to operate precisely all along the acquisition and treatment
p~o~ess. Gauges.and transducers, data recording (including analog
dlgltal converslon) data analyses (including editing or automatic
plotting) equipment provided with numerous options are now available.

289

Exciting Force Impulse

F (t)

Excitation Method

J F (t)

A---

A [tr-

A>

rp

LL__

:QAtLet go Load

-.

~
Vehicle

+ Obstacle

\:0
Shaki ng machi ne

~QQQ---

~~~

Train Road Traffic

-,

-----

tt

-- ----

.t

'~

Vertica I Load Impact

IIYdraul . electric Vibrator

Output Signal A (t)

Oi sp I. Acce I. Stress Fre .

~
h .

~~:
,

14fV -

4A--- ~

hl\ ~~'~ !I~l~l i

V

a.
rI.

t)l{V+v-

t~.

Wind Excitation

Fi g. 1

t~
I

IAt~.(\,~
TV 'PV"

290

Position of the point of application of the excitation force

The position of the exciting force is a priori
independent of the obtained characteristics, as long as it
remains in the verticaJ symmetrical plan. Nevertheless, the
further the spacing from the vibration knots of the corresponding
mode, the bigger the amplitude response. Moreover, if we are
dea~ing with a fixed frequency sinusoidal force, the initial
excited mode generally degenerates quite fast to give way to all
the other modes. However, when the system is excited in its
central parts with a large energy the fundamental mode appears
firstly with minute precision.
Position of the measuring gauges
In order to detect with preclslon the vibrating
characteristics, one has to keep away from the vibration knot of
the modes to be analysed. A systematic precaution consists in
moving the qau~es during the successive measurements.
~1ain

types of gauges

Different types of measuring gauges, which provide

enough precision, are presently available and i t is easy to
express the signal in millivolts with transducers. Numerical
transcription of all data guarantees a perfect storage of the
recorded data. All analytical studies can be carried out later.
Torsional behaviour
Though generally less significant, torsional behaviour
combines in fact with behaviour in longitudinal vibration. In
order to reveal the torsional characteristics of a bridge, the
excitation force must be applied off the axis of the bridge;
otherwise, the simple presence of a relatively important load
as an integral part of the bridge during the excitation process
is sufficient to degenerate the pure vertical movement and will
reveal its torsional components.
An interesting technique for torsional analysis consists
in placing 2 movement gauges in A and B and then assembling them
in oppos iti on.

291

Cut pu< 5ig-~"lal.

lfethod of' measurements t-1easuring appara
tus
mC'nsuY"cd d"\tns
TlOtli. tiN'
QccE"leration

~"

accelerometer

n $&i .mometar
~-----

==:c--

.~..

magnetic

Cl

II
:/~~~':

:frequencies

moment
di.splaceL'1ent

~
Ji

stress str..-.in
stra:l.n gagee

/'

:frequencies
di.!lplac6mt'nt

#.

e1ectronic theo
dolite

"trackercoil or
capnc:l.ty induc-

tion

"

~F

displacemont.Creq~eDcies

s tntic.d)"1l.:u,1iC

opticnl measure

displacell1\)n.t

laser
hologram

de:f'ormation

.~-~liGht
.,x-~

stat1c + dynamic
!'requencies

f'requencies

weighbridge

axle load

s'trnin sage

tr,,:fNc ..,tn'tisticlf

quartz

Nw:tber ot: ve!uc(e.s

IIUltpletic 0011s

speod. intern\l:J

Fig. 2

~ j

~~
Fig. 3

I
I

IF

292

Frequency analysis
For the treatment of vibrational characteristics
different types of spectrum analysis (Narrow Band) are now
popular. The common mathematical basis is the Fourier Transform.
The discrete equivalent of Fourier's integral, called Discrete
Fourier Transform D.F.T. can be written in two equations:
n-l
_ 1
f(n) exp (_j 2~nk)
forward transform
I:
F(k) -"N
n=O
N-1
inverse transform
f(x)
I: F(k) exp (j 2~nk)
k=O
The Fast Fourier Transform F.F.T. is more efficient in
digital computing ; th~s algorithm reduces the number of complex
multiplication from N to N lo92N.
Analysis bandwidth T
Spectrum or

S~gnal.

period~oally

ra pea ted signal

~t ~i

....J.-7'

The analysis bandwidth

must be greater than
the Fourier 1ine
spacing B > ___1_
Trep
(T rep = repetition time) .

Fig. 4
A typical bandwidth choice would be approximately 1/5 of the
bandwidth BAff = iE ; TE is the length of the pulse.
Nevertheless it is possible to increase Trep to obtain a better
likelihood of encountering problems with dynamic range and crest
factor. The value of bandwidth Beff can in most cases only be
fixed by analysis of the data (If we gradually vary the bandwidth,
a stage is reached where the results remain unchanged. Only if
we want to express the results as an "energy spectral density"
the bandwidth must be less than the value of the transient itself
B :< Beff .
Window length and incremental step
It is always desirable to make the step somewhat shorter
than T so that there is a certain amount of redundant information in the succession of analysis

293

Fi g. 5

Window type
The choice of the window type is also important.
The Hanning function as well as the rectangular function are
easily generated and give good results.
o t)'"'-~~_f~_":f.I.-1X :1.
-to ~:i'." I ! ! ' , ~~ . l.,
20 :

i:::: I I i II

:;0

i Li.

bo

~H(:

10 ;:;;:':.

III

! III

;~ ~l ii~i; i: !

dB

Frequency

.:

Att.nuat~on

j.'-

20 dB

Decade

60 dB / Decado

~
1-1--\

-{I\i----Corrtperi.on

or

F1nt woiehtin~
(rectansular)

~ctlon

Hanning
(cooine)

tho ,.per.tra of' two cOmn'lon 'Window :functions

(rectangu1ar and Hanning)

Fig. 6

Advanced analysis ana "Zoom FFT"

It has been a long time since the F F T became well
known. The normal analysis is made from 0 to the Nyquist frequency
F and the frequency lines up to F give an acceptable resolution.
Howewer in some special cases of analysis it is necessary to have
a finer resolution over a limited range of the spectrum. This fine
resolution of the "Zoom F F T" is particulary useful for sharp
resonances which are too closely spaced to be identified by normal

294

analysis: Multiple assembled spans showing similar mechanical

properties, deck elements presenting numerous harmonics, superposition of vibration modes at very close frequencies, flexion,
torsion ...
Signal and noise
FIt)

At

.r

HEASUllli

Fig. 7

A vertical impulse applied at a point A

on a bridge structure gives at B : the
dynamic characteristics of the structure
(periodic) and the "noisiness" (random).
Measurement procedures used to render
the analysis clearer, and which guarantee that a mode, though not widely
represented, will not be forgotten,
consist in summing up the n sample data
successively recorded.

If for 1 measure Signal

Noise -- a , by addition of n
response data (acceleration, stress ... )
1 the periodical expected signal becomes more apparent
2 the average effect of the noise produces an

attenuation
Consequently the above mentioned report becomes much
more favorable.

Si~nal = aln and the accuracy of the

NOlse
parameter of interest is greatly improved.
'I

,
s

.AMPLITUDE
Accelerltion spectnn CIIftCnte bridg. :
30 II $'ngl. ,pin SUbJKted to rood

tl'lff!e

'" ",sponse .ftel' pass.ge 0' 1 Yehle Ie

- . response .ftel' pus.g. 0' 100 YeMe les
'n t.tII eoSll 1...,1 ng T. 0,011

Fig. 8
'1

II

FREQUENCIES

295

3. Dynamic Response Components

Design specialists as well as mechanical engineers tend
to interpret as clearly as possible the compon~nts of dynamic
bridge behaviour. Different pictorial interpretations of Fourier
Analysis greatly facilitate the comprehension of the relationship
dynamic excitation response.
Spectral Density
The Forward Fourier Transform of a periodical signal
becomes a continuous function of the frequency and in this case
the spectral components have the dimension of spectral density.
If we square the amplitudes we then represent the unit of energy
per frequency unit; the finite energy in a frequency range is
obtained by integration of the corresponding power spectral surface.
The power spectral density is specially adapted to stationary random
function with a continuous spectrum and which has a statically
constant and finite power. This is the case of the traffic loads.
Various spectrum representations can be made.
Amplituda

hmpl1.tude

~\,~
-

A'
A.

<111

~.'"j ~ k;-"~'<O
III

A.

t:

Fig. 9
Autocorrelation function, Convolution, Transfer function
This function is obtained by multiplying a function
fx(t) by the function fx(t + T) as follows:
Rxx (T)

1 JT/2

lim f

T-+-oo

fx(t) fx(t+T)dt

- T/2

So Rxx (T) gives a precise measure of the extent to which a

signal fx (t) and a displaced version of itself fx (t + T)
correlate. If T = 0 ; Rxx (0) gives us the mean square value of
the function fx (t). The power spectrum is the Forward Fourier
Transform of the autocorrelation function.
- Let us consider 2 time-functions
convolution is defined as:

f(t) and

h(t). The

296

+OO

9 (t)

-00

written in a symbolic form ( '" - - - convolved with)

g (t)

f(t) '" h (t)

f (t)

1-1----.-..-

~_--,g (t)

input

~I

h(t)

out put

h (t) impulse response of the system f (t) can be

considered as a succession of impulses (delta function) and each
of them generates an impulse response with two caracteristics :
- the scaling is proportional to the level of f (t)
- the delay is constant, the time interval excitationresponse
The complete response at time (t) is the sum of these impulse
responses (integration over a dummy time variable T).
- There is a parallelism between convolution in time and
convolution in frequency domain.
If we call G (f)
respectively
F (f)

the Forward Fourier Transform of g (t)

of f (t)

and H (f)

(f)

= F (f) .

of h (t)
H

and

then

(f)

The frequency spectrum of the output is obtained by multiplying

the input spectrum by the transfer function.
Now let us consider :
Fx (f) Fourier spectrum of the entry signal fx (t)
applied to point x
Fy (f) Fourier spectrum of the response signal fy (t)
obtained at the point y.
The transform function from x to y is the complex ratio
Hxy (f) = ~
Fx~rf)
by multiplying by FX'" (f), the complex conjugate of Fx (t),
numerator and denominator of the second member
Fxy ill
Hxy (f) = ~~ f~~
x
~~
Fxx

fB

297

Hxy (f) - Cross spectrum between x and y

- Power spectrum at point x
Therefore the sample records fx (t) and fy (t).give us Fx (f)
and Fy (f).
From these last two functions we can easily obtain the cross
and power spectrum and also the transfer function (or frequency
response).
In the case of traffic loads, wind cha~ges, earthquake,
vibrations, excitations are random; the transfer function will be
determining and this means that between input and output for a
given frequency the phase angle is always the same. It should also
be noticed that the transfer function applied to a given exciting
random force can only provide the envelope of maximums and not all
the intermediate steps so that the transfer function cannot be
used directly for fatigue or prediction of the life-time of a
structure.
II. DYNAMIC BEHAVIOUR OF BRIDGES SUBJECTED TO TRAFFIC
1. Simplified approach:
Until recently, bridges were heavier and had more
limited spans, and at the same time the dynamic requirements to
be taken into consideration were relatively less important.
The design methods were meant to deal with the
structural problem in statistical terms, by using the so-called
dynamic increase coefficient 0 > 1 or dynamic factor.
Atcording to different design codes, simplified
approaches were suggested, based on actual experience.These, at
any rate, give bridges sufficient safety.
Considering 20 countries whose experience in bridge
design is recognized, the dynamic coefficient diagrams according
to the bridge span are located within a wide band shown in
figure 10.
For example, for a road bridge (class B), in France,
we have the following coefficient of dynamic action
0=1

298

=a

factor of elastic
sensibility which
depends on the length
L of the element
0,4
a

= 'l-+"--;0",""2-;-L

1.5

----SO

100
BRIDGE SPAN m

Fi g. 10

factor similar to a
mechanical admittance ;

dynamic load
(vehicles) ;

dead load of the

bridge
_
0,6
S- 1+ 4 G

2. Current research
Traffic conditions as well as higher stresses in
material as lighter bridge decks lead to greater stress changes.
Therefore the vibrational behavior and the risk of fatigue
damage must be taken into account.
In order to study the relation between traffic load
and stress in the different bridges components, the European Coal
and Steel Community sponsored a large investigation program.
Some interesting research on existing bridges has already been
undertaken by various laboratories.
a) Road-traffic loads on a steel bridge
A single carriage way bridge with only two lanes, built
with a simple structural system has been chosen and studied by
Mr. J. DE BACK and MR. M.H. KOLSTEIN, Delft University. The
Netherlands.
The bridge
A movable bascule bridge crossing a river has been
chosen : "the Haagsche Schouw Bri dge" on Hi ghway 44 Den Haag -

299

Amsterdam.
A
B
C
D
E

The bridge structure is as follows

two main steel girders
cross girders
longitudinal stringers
deck made of wooden boards
asphalt top layer

k- _1._ _ _ _ _

7.40 m _ _ _ _

9.50 ..

Fi g. 11
Measuring experiment
Strain gages were applied to a cross girder (bottom
flange) and to the stringer located under the traffic lanes.
----,"l)OO~

DRI VE DIRECTION

FAST LANE

"

"

Fi g. 12
Weightbridges and magnetic coils were installed a distance from
the bridge in order to measure axle load, vehicle speed spacing,
and to give all statistical traffic data.
All required data for the traffic load as well as the
resulting loading effects are recorded and processed with a
minicomputer. All the electronic equipment is situated in a
measuring lorry in order to produce immediately 6 histograms of

300

time function: Frequency distribution of amplitudes, level

crossing counts.
Principal results

we shall first see what happens when there is a single

vehicle on the bridge, then in normal traffic conditions .
. A single vehicle:
- Determination of influence plans from static measurements
The influence surface for each measuring point can be
determined experimentally

A lorry with calibrated wheel loads was placed in different positions

and the stress at the measuring point was recorded. The calculated
stresses are about 30% higher than the measured values.
- Dynamic response of the bridge to moving vehicles
- Moving vehicles transmit to the bridge-deck a local dynamic
action which fluctuates around the static load. The direct
impact action can be defined by an impact factor: maximum
force between wheel and road divided by the corresponding
static load; the factor for trucks and trailers can vary from
1. 2 to 2.1
- Moving vehicles generate vlbrations in the bridge structure,
and the resulting stress fluctuations at a measured point also
depend on the general oscillation movement of the bridge even
if the vehicle is far away.
This can be seen in the signal obtained by the passage of a truck
and trailer over the bridge. This analogous recording enables us
to clearly see the 4 maxima in spite of the superposition of

301

higher frequency vibrations.

Afterwards a more detailed
analysis shows the influence
of speed or lateral positioning of the vehicle. The
ratio between the dynamic
influence factor and the
static influence factor
reached i. 44.

Fi g. 14
Bridges subjected to all road traffic vehicles
The real problem in dynamic requirements of
structures is the passage of vehicles at different speeds, lateral
position, load, spacing.
Even though the axle-load of the traffic and stresses
are measured simultaneously at any given moment, because of the
random superposition principle, the mechanism of loading actionloading effects must be statistically studied. The vehicles
composing the traffic sample have been grouped according to
position and number of axles in 21 groups.
The main factor in the traffic statistics are vehicle
group load, spacing, velocity and lateral position.
In regard to stress, the number of level crossings
is of major importance. These measurements, carried out on the
cross girder, clearly show that the slow lane bears the
strongest stress.

Fig. 15

10 01

"'ll/Nl:l

The number of levelcrosning of a given stress is rneasured at

2 pOints on the cross girder.

The longitudinal stringers show very different stress values

according to their lateral position.

302

Counting methods
For a given composition of traffic, the experimental
stresses in the structure are analysed for a chosen time period.
Two counting methods can be used in order to work out the stresshistogram. The classical levelcrossing method and the rainflowcounting method.
In the rainflow counting method, the collection of the
stress variation with time (vertically represented) is similar to
the different roof surfaces on which the rain flows. Each half
fluctuation of the signal corresponds to a roof surface which
leads the rainflow from a maximum to a minimum and inversely.

o, - -_ _ _ _ _ STRESS
Fig. 16
o

1'>

min and max of half cycle

whol fluctuation

l'

TI C

In the rainflow counts, graphs are plotted in the

range(s) in N/mm2 against the number of oversteps of a range
divided by the total sum of axle-loads greater than 10 KN.
Results and conclusions
Stress in material
Every passage of an axle-load produces a pulse with a
maximum at a given measure point. The amplitude of this maximum
is mainly proportional to the static axle-load. Different
factors modify this equivalence.
Axle-load and stress: dynamic effect of the bridge
suspension of the axle and profile imperfections of the road
surface, lateral position of the vehicle, simultaneous action of
other vehicles.
Computed level crossing and rainflow counts
For the design of steel bridges, in terms of fatigue,
loading and durability, it is necessary to calculate the
theoretical number of axle loads and from them to compute the

303

stress frequency curve for a given per,;od.

The influence lines associated with a measuring point
and the axle-load count make it possible to establish mathematic~lly model level-crossing and rainflow counts.
Comparison with the measured data
In order to test the above theoretical approach during
a period of several hours, measurements of axle-loads and stresses
have been made. The results of computed and measured level
crossing and rain flow are in good agreement especially in the
high stress cycles.

.,

"

Lorry position and date (measuring

points on stringers).

Measured and computed curve ......

Fig. 17
Each vehicle type has a statistical frequency for
crossing the bridge in the slow or fast lane in a given period of
time, and it contributes to an increase in fatigue. In order to
estimate with precision the damage potential, a new concept of
prime importance has been used.
Characteristic axle-load
based on the equivalence of the fatigue damage potential
between
[characteristic axle-load]N times = N axles of the axle load
spectrum

304

For each vehicle, this value can be calculated. For a bridge having a defined structure and a given distribution of traffic the
potential damage by each vehicle type can be estimated for a
given duration.
characteristic axle-load
Obviously the sum of the potential damage induced by the
different vehicle types increases with time. The time needed for
this cumulative damage to reach the value 1 is nothing more than
the theoretical durability of the bridge under the considered
traffic data.
The application of the notion of characteristic axleload is extremely viable. It shows that for this bridge, 91% of
the accumulated damage in the slow lane is caused by only 5% of
all traffic.
In particular, it points out the most damaging types of
vehicles. Several other bridges have been studied by the same
laboratory. Even though, the structure of the bridge and the
traffic data are very different, the results are similar. The
above power relationship made it possible to calculate the damage
potential of a load range class or a stress range class with the
Palgren Miner damage rule.
For a given vehicle type the characteristic axle-load
can considerably differ for two different bridges, but also-for
each bridge type, the fatigue life expectancy can be obtained.
This datum is fundamental for future bridge codes.
b) Railway bridges
Some interesting research on bridges required for traincrossing has been realised by M. M.R. Pfeiffer at the Frauenhofer
Institute of Darmstadt (W. Germany).
Axle-loads are measured on the rail itself due to an
ingenious strain gage system which only delivers an electric
tension when the wheel is situated between points 1 and 2. At the
same time this system gives the count and 2 of these systems give
the speed.

Fi g. 18
Rail + wheel

strain gages

305

The axle-loads are much more homogeneous and the axleload spectrum, obtained by the passage of a passenger train is
shown by Fig. 19.

Fig. 19
log
dB

20

Hz

10
10 1
10 2
Main girder

longitudinal stringer

Proportional to the inverse of the axle-distances,

a,b,c, frequencies can be recognized on the curves, as the third
superior harmonic.
The main girder chiefly provides a response for the
grouped charges, while the longitudinal beam gives a punctual
response for each passage of an axle-load.
As the dispersion of the axle-load is not as wide as for
road-traffic, the distribution of the stress by the level crossing
method is situated in a narrower range. For example, the levelcrossing count of stress for one day, made on the 3 principal
elements of a steel bridge (freight and passenger trains) is
represented in fig. 20.

70

STRESS
N/ ...2
~.

60

LONGITUDINAL BEAll

- - - - TRANSVERSAL GIRDER
_ _ MAIN GIRDER

_.

-- .--._.

~J -~~))

-10

Fi g. 20

~-----=r~'::' --,-~

10- 1

10

10 1

10 2

10 3

Log

~f

104

Level crossing countage for 1 day train traffic

306

Conclusion
For railway as for road traffic subjected bridges,
the dynamic response can be written in terms of stress cycles.
The bridge design method, in order to take into account the
potential fatigue damage of structural :.elements is inseparable
from the stress spectrum which must be converted into a design
spectrum. Finally for a given traffic, it is the design spectrum
which improves the fatigue life expectancy calculations.
This type of research on bridges subjected to traffic
clearly shows the insufficiency of the classical notion of
dynamic factor or homogeneous stress for ensuring saf~ty
This research introduces a more efficient approach: it
means attempting to level the cumulated damage for all points of
the structure. With this concept all elements of a bridge
structure have the same life time expectancy.
III. DYNAMIC TEST ON ACTUAL BRIDGES
1. Dynamic test with vertical impulse: Auscultation of full
concrete slab bridge
It concerns a two-lane road-bridge, with short span.
This full slab bridge (cast-in-place) lies on 4 elastomer
bearings with a dead load of 120 metrical tons. The principal
geometrical characteristics are given in fig. 21 and 22.
The dynamic test was made with a mechanical system.
It consists in dropping a 50 Kg-steel plate from a hei9ht of
0.7 m on springs.
The dynamic characteristics have been obtained
through the classical method: accelerometer, recorder, analysis.

Cross section
Fig. 21

9.00

307

/,

-;
/:

?
/

,..

/,

"

'/

/
A:

2 x 3 x 2 12 pilln

, 10 <II
L. 570 ~

: .. terttght 3 ell
ISptNil t top lIyer 9 CII

C :

x Z " .l.stener bearings

300 " 500 . . x 3 (10 + 3)

Fi g. 22
The two main frequencies very clearly appeared, and
whatever the conditions of measures might be, they hQve the
following value.
2,95 Hz (calculated 3)
Fundamental mode frequency
"
10,3)
11,0 Hz (
Second
"
Third
c
23
Hz (
23)
The damping obtained by logarithmic decrement has also
given a constant value, similar for the 2 modes 0 = 0,01.
As in most concrete slab bridges, geometrical imperfections, errors of concrete quantity (mass), errors of steelbar
positioning (inertia) ... give experimental frequencies slightly
different to the theoretical ones. The amplitude of the vibrations
depends on the position of the measuring position M. Fig. 23
shows the response for 2 characteristic positions of M : in the
middle and at 1/3 of the bridge span.
It would be extremely interesting to redo in identical
conditions this test in order to follow the mechanical behavior
through the years.

308

Acceleration
M
,

tt,

i},-t
'"'--:la

20

Fi g. 23
10

FREQ~ ESC 1ES

Spectral response for 2 positiooSof measuring point M

The frequence spectrum evolution in the few seconds

immediatdy after the impulse application is given in Fig. 24.
This vibrational "survey picture" by itself sums up the main
mechanical characteristics of the bridge. A further control
carried out after n years can be compared with this initial survey
picture taken as reference so that any modification in the response data can be detected immediately.
A

Aropl itude

A three dimensional figure showing the evolution of the frequency

spectrum with the time

Fi g. 24

309

2. Sinusoidal excitation test

Comparison between normal and light concrete deck on
mechanical properties of 2 identical bridge-structures.
In this test, dynamic excitation is obtained by a
hydraulic vibrator which produces a sinusoidal force applied at
mid span. Two types of methods can then be used.
logarithmic decremental method:
When the vi bra tor works with a constant frequency close
to the fundamental frequency of the bridge (forcing vibration),
the amplitude is much greater than in the impulse method.
The excitation is stopped and the natural response is
obtained. If A.1 and A.1 +1 are two successive maxima then
.
Ai
8 . Log A.+l

* The
* the

structural damping is written

viscous damping

. resonance method
A constant sinusoidal force with a progressive
variable frequency is applied. For each frequency N. the
1
corresponding amplitude Al is measured. For the resonance,
frequency N2 ' the maximum of amplitude reaches Ar
If 2 different frequencies NI and N2 (so that Nl < N2 < N2 )
produce the same amplitude A = A = Ar
1
2
-= the damping can be
/2
obtained:
Nr
N2-N 1

Q surtension factor
1

the structural damping

'V

=Q

the viscous damping

= 2Q

With these two methods, a very interesting experiment

allowed the study of the influence of the concrete-slab nature on
the dynamical behaviour of a 77 m span steel bridge structure.

310

AMPLITUDE RATIO A
AD

AMPLITUDE
1

...

.,...:

1[2
0.5

...
...

"
".

FREQUENCY

...

~REQUENCY

N1Nr N2

No

resonance

Impulse spectrum

. the main structure of each bridge is constituted of

2 boxgirders with variable inertia and 2310 to 2330 mm in height .
. The bridge deck is constructed in a concrete slab
9.360 m wide and 12 cm thick in 2 versions (normal and light
concrete) .

Cross section:

I~
Fig. 25

311

The bridge was sinusoidally excited. Measures have been

made for the main girder without deck and with the 2 versions of
the bri dge deck.
FUNDAMENTAL
FREQUENCY (rio)

DAMPING

resonance

steel girder

1.19

1.23

109.
decrement
2.35

.. wi th nonnal

1.19

1.15

2.60

2.84

with lei Qht

1.18

1.15

2.50

2.50

mea::.ured

concrete deck

concrete deck

computed

2.24

Fig. 26

The bridge was excited by a hydraulic vibrator and the

natural frequencies have been obtained with an accelerometer. The
damping was measured with the logar~thmic decrement method and
also with the forming resonance.
The 2 methods have approximately given the same results
and if 0 decrement log differs, it means that the decreasing
amplitudes are not striccly linear and in fact, depend on the
considered number of oscillations.
The damping value obtained with 20 oscillations is
given in Fig. 27.
It is interesting to note that the calculated frequencies,
with the method of lumped masses and with the equivalent beam method,
give values near to those measured.
The bridge has in fact a slenderness of ~:75 = 7.5
which is chiefly supported by the steel girder, the participation
at the deck is sufficient to keep frequency at the same value in
spite of its mass. In fact in the case of light concrete, the
structure is overdimensioned.
The resonance method needs more implements than the
impulse method, also much more time and the performances are not
better in precision; but the injected energy can be higher with
a changeable moving load of the vibrator from 0.365 to 3.4 tons.
With an increase of the vibrational energy, the amplitude grows so
that the observed damping mechanism of the bridge corresponds
effectively more to that which works in traffic conditions. In
spite of the relatively high vibrational energy level, the
evolution of the damping value with the decreasing amplitudes of
the bridge shows dispersion.

312

;I

Damping

8 %

'. '.c .....

stee 1 qi rder
'-- . ....c

with normal concrete

deck
3

..mm

I
6

Amplitude

..

............. p

with 1 i ght concrete

deck

.... ....

3
2

.... - ....

2.5 mm

. . . ...
'

10

1'3

mm

Fig. 27
IV. DYNAMIC TEST IN LABORATORY
The dynamic behaviour of a large number of bridge
elements can be usefully studied inthe laboratory with dynamic tests
to assess durability.
1. Accumulative damage
If n1 n2 .. , are the number of stress cycles applied
to the material at amplitudes of stress 01 02 .. The o-N Miner
relationship give us the linear accumulative damage criterion D
in the discrete form

N. is the number of cycles (with a stress level 0.) to failure.

Often D can be written in a continuouS form(harrOwband system)

313

'-:I"-::!---:---;----,;+
"1 nj
number of cycles"
stress spectrum

stress histogram

c~mulative

Fi g. 28

Fig. 29

Fig. 30

damage

In some cases the modified Miner law givesa better approximation

of the tlamage Di obtained by the application of ni cycles
with a stress o. : the Miner element Di of the Miner sum can
1
n. x .
be wri tten Di
(-2)
Wl th
x > 1
Ni
Fatigue testing methods for traffic simulation.
For a structural element of a bridge, the engineer can
only make the design with a given security and life expectance if
he can solve the problem of damage under vibrational random forces.
For this,different experimental tests allow the determination of
the fatigue line curves.
The statistical estimation of the traffic data can be
translated in a load spectrum applied by the vehicles running the
future bridge. The service inputs are then applied by dynamical
tests using di fferent types of cycles in order to obtain the actual
service behaviour of the bridge element and especially the accumulation damage.
The single and constant amplitude cycles method,the
most popular, is nowadays replaced by a better approach allowed
with the new test machines: constant amplitude cycle blockS
which can be programmed in advance fixed sequences or in a random
process.

Fig. 31

constant amplitude
cycling

in adviln,=e' fixed
bloc. eyel'!!..

l'"3ndcm proces$

c./cl1n']

314

Markov matrix method

The forecasting of durability and damage provlslon on
the basis of the loadhistogram can be more explicit if the process
respectsthe distribution of the different stress levels and also
the mean stress random. For this the total stress range is
divided in n levels. The number of passage from a given minimum
stress level 0. for example to a given maximum level 0. is
1
J
called the transition frequency Tij from the initial to the
final considered levels.
For each transition possibility from a low criterion to a
higher criterion, the frequence is written in the corresponding case
in the transition matrix (with n rows and n columns) on the right of
the diagonal. The reverse transition possibilities figure under the
diagonal. The use of this Markov matrix allows the simulation for
whole statistical data of any dynamical loading process.
In figure 32 for an initial stress level O'i ' with
0i < On in the row i , all corresponding transition
frequencies possibly one to the higher stress level (from OJ to
0' ) figure in the
rows i in the right part of the diagonal.
Anstatistical process allows us to explore the totality of the
matrix elements as shown in the figure. The stress levels in
regard to the corresponding transition frequencies can then be
applied in a structural element with a hydraulic load system.
A more schematic and concise representation of the
transition frequencies can be made and for this 12 stress level
numerical matrix in Fig. 33 a hachure map can be obtained. For
this vie can choose for example six classes of transition frequencies
such as for the class lOa < Tas < lOa+l the corresponding zone
can be drawn with a unique hachure surface. With a little experience,
this map gives immediately some indication of the severity of a
cyclic loading process; and for different tested elements, a classification of cumulative damage in fatigue can be made. With the transition frequency method, the whole characteristic of a dynamic loading
process can be simulated in a laboratory in order to provide qualitative observation of actual behaviour and quantitative data to the
design engineer.
0'1 <

315

Markov matrix with frequencies of

transition bet""en stress levels

Fi g. 33

Fig. 32

The numerous research studied on damage of concrete

under cyclic or random loading made with acoustic emission
control, give similar results in structural behaviour.
Phase

I : microcracks appear immediately corresponding to

structural modification

Phase

II

a linear but moderated

variation of the material
damage with the number of
cycles can be observed.
The corresponding acoustic
emission showed a low level
of acoustic emission

Phase

III

after the limited damage

evolution of the extended
phase II the damage in
phase III is growing
quickly to the failure.
The corresponding high
emission show a instability of the crack evolution which shorten the
life time.

Cumulative damage
ratio 0i

i l FAILURf

O.S

O.S

Fig. 34

1 N
cycle ratl0

2. Damping
Themodal analyses of a bridge furnish the frequencies
and mode from the mass and stiffness matrixes without necessitating other hypotheses. On the contrary the damping value has to
be introduced arbitrarily. This value is based on an experimental
damping behaviour in similar bridges, generally small. The errors

316

of the damping estimation have little effect on the value of

the different modal frequencies, the pseudo-period remaining
always close to undamped period.
During vibrations the alternative transfer from kinetic
to potential energy is carried out with internal loss. This
damping effect can be obtained in two different ways:
In the case of forced vibration the energy dissipation
per cycle is delimited by the hysteresis curve. This latter
shows the deviation from Hooke'slaw according to the type of
damping.

-f ~
+ -fFORCE

FORCING S Y S T E M .

AMPLITUDE

Fi g. 35

The system of forces can be written

mx+ cx + kx = Fo sin wt
and the corresponding vectorial representation directly gives the
phase angle.

Fig. 36

317

When the exciting forces cease, the amplitude evolution

with the time can be followed by the logarithmic decrement 0 .
For a complex structure, 0 results in general from the superposing
of the different damping types: Coulomb friction viscous and
structural damping. Considering the envelope of the successive
maximum of amplitude according to the order n, different curves
are obtained. Fig. 37.
Moreover,6 is also a function of the emrlitude, Fia. 3d.
Ampl,tude

n=3

n 2
n I
n 0

'--_ _ _ _ _ _ _ _ tillle

Amp itu

Fig. 37
Fi g. 38
Considering now the influence of the prestressing stress
on the value of 0, tests on the prestressed beams show clearly
that from a certain stress level 0p stabilisation of 0 is
obtained,Fig. 39.
6 S

-1.5

Fi g. 39

3.5

2.5

~~~-~~~~-----------+----~--~
100

Variation of

..

with the prestressing

0pdaN/cm 2

O"p

o is also dependent on this amplitude but less, as long

as the elastic domain is explored.
Concerning the damping, the prestressed and poststressed
beams have identical behaviour.R~ decR~ases when the stress
increases (pre or post). When -- or -- are exceeded 0 increases
suddenly; this phenomenon is felate8 to the microcrack exten~
sion. especially in the areas under tension.

318

The correspondence between a and the state of the

microcracks in the material can be presented schematically in
Fig. 40.

FAILURE

"-

n 0

I!

Fig. 40

II!

n I

AMPLITUDE MAX.

In the domain I, the level of stress is low, the

damping results from Coulomb friction. Progressively with the
increase of the amplitude, the viscous damping, as the structural
damping, appear in the domain II of visco-elasticity the
deformation law is quasi-elastic. With higher stress application
the internal m~crocrack propagation grows as the damping which
reaches the order n = 2 (hydrodynamic). With only a small number
of higher stress cycles the damping grows to order 3. Simultaneonsly the microcrack development generates visible cracks and
the cumulative damage reaches the value 1 (failure).
For concrete elements tested in laboratories, the
damping evolution is a useful
indicator of the internal microcracking mechanism. Two complimentary procedures allow currently
to follow directly the mechanical and structural state of the
material : In the acoustic emission, damage is assumed, if a
certain level of emitted acoustic pulses is exceeded. In the
volume alteration measuring system, a constant amplitude repeated
load first produces a decrease in volume domain I in the increase
in damping. Then, the volume decrease dimin1shes rapidily, as soon
as a low level of cumulative damage is attained. This corresponds
to the passably large domain II. After the transition from the
domain IT to III the crack propagation is so intense that it
produces an increase in volume and the damage progress is then
rapid.
Acoustic emission, volume alteration and damping variation
are indicators of the microstructure concerning cracks, that means
of the cumulative damage. In the laboratory microelastic ~s w~ll as
thermodynamic approaches tend to define the energy dissipation in
the material.

319

From tne laboratory to the bridge

The damping is the macro indicator for tne eval uation of
tne level of internal microcracking state of the concrete or its
cumulated damage. For a bridge, in so far as domain II is
explored, small alterations of 8 risk being hidden by measurement
imprecis ion, affecting tne damping. However, if in a gi ven area,
a diminution of the constraining force or an increasing of an
external load occurs; so that, at minimum, one of the stress value
in tne extreme fiber leaves tne elastic domain, the measure of the
corresponding damping shows an important increase.
Additionally the position of this corresponding area can
be ootained : the damping evolution affects firstly modes for
wnich the altered area is located at tne antimode. A high
prestress value delays the appearence of cracks. It is important
to note tne similarity of curve deformation-prestressing load with
tne Curve prestress val ue-crack width.
The damping survey equally for a bridge span as for a
local sections allows the estimation of the structural damage
consequence of nigh stress cycles, or a loss of cable tension.
Current metnods of control of the vibrational properties
snould be completed by a damping diagnostic in order to estimate
tne damage before external cracks can be visually observed.
Thus a prestressed box girder bridge ldO m long witn 3
spans nas been tested. The vertical cracks (1 to 3 mm) in the
bottom part between tne adjacent elements and the horizontal cracks
between bridge deck and girder are the result of the combined
action of tne daily temperature gradient with strong stress cycles.
6

10%.

The obtained damping value of the damaged bridge was

3. Mecnanical interaction railway-longspan suspension bridge

Mucn researcn has been accomplished on the problem of
bridges required by railway traffic. However, for suspension
bridges, tne low rigidity causes great deformations at the time
of tne passage of trains. Otherwise the mechanical interaction bridge - carriage can cause vertical and horizontal vibrations :
Tracks supported by suspension bridges are so flexible,
that they may be influenced by the running vehicle itself (as by
wind or seismic actions).

320

Some very important research has been conducted in this

direction by the Professors Yasoshima Y. and Matsumoto Y. of the
Civil Engineering Department-University of Tokyo. The question
was (for Japan Railway Construction Corporation) to improve
carriage suspension, and at the same time, to study the mechanical
behaviour of a long span suspension bridge, exposed to the passage
of an express train.
Vibrational behavior of trains
Railway vehicles, running on tracks, are always
affected by displacements of the positions of the wheel axles.
Track irregularities, especially of a constant wave length,
affect all axles with a different phase. However, track alignment
which has a longer wavelength, affects a great portion or the
total number of axles simultaneously.
Railway vehicles, running at a constant velocity on a
track with irregularities of a constant wavelength produce a
periodical force on the track and the bridge deck.
Lateral instabilities also have to be taken into
consideration, especially wheel sets and flange action behaviour
which can induce transversal periodic acceleration.

40
)J",i/e

~H. ..I it're9U la,"~

..\

Fig. 41
Characteristics such as rail-wheel creep coefficient, stiffness
of the vertical and horizontal wagon suspension, hysteresis
damping must be taken into account.

321

~
._./

,.

"

Lat.. ,.al

.s

Fig. 42

J"ne ..,."

Stiffness and hysteresis of the bogie suspension

The wh0le dynamic behaviour of the wagon on rails has been

tested on a two dimensional, 36 m long shaking-table. The model
of test vehicle could be accelerated laterally and vertically in
a frequency range up to 10 Hz.
The tests have made it possible to determine all the
mechanical characteristics of the movement (amplitude and frequency
of stationary lateral snake motion, of rolling motion, response of
wheelset with velocity increase ... ).
The experimental system Wei:> l.0" 'cu VUI. uy ~dKlng tne
vibration modes of the suspension bridge (1st to the 6th mode).
Feedback to the shaking-table, made it possibleto take
into account the calculated response of the suspension bridge.
The vertical acceleration of the bridge does not have a so great
importance, as the lateral one. The lateral acceleration possibilities have been simulated, in orderto produce vehicle instability
for a better knowledge of the derailment conditions.
- lateral action: for a 1000 m long span bridge, the
lateral vibration is small because of the small mass ratio and
the great difference in frequencies. A small amplitude produced
by wheel set appeared only for the 6th vibrational mode of the
bridge.
- vertical action: the running vehicles produce
impulsive vibration due to the irregularities of
tracks, railjoints, movements of vehicles (rolling,
snake)
dynamic motion of the trusses which affects the
complete bridge system
The bridge
Several types of suspension bridges have been studied.
The 1700 m suspension bridge project, joining Honshu to Shikoku,
has been investigated in order to improve the serviceability of
running trains and roadcars . With a main span of 1100 m and two

322

side-spans of 300 m, the stiffening truss has a constant depth

of 10~7 m ; a width of 33 m with a sectional area of 0.68 m2 and
an inertia of 2.687 m4.
When a train Crosses this suspension bridqe, the
stiffening trusses undergo significant deformation~
- vertical deflection
- elongation and contraction of truss
extremities due to deflection and temperature change
(117 cm)
- vertical curvature at main towers with deflection
change.
The deflection of stiffening trusses depends greatly on the type
of the bridge suspension. Thus, besides the classical cable suspension bridges (type 1). two other solutions have been proposed
auxiliary cable. continuous suspension bridge having hangers at
branching points (type 2) or having no hangers at branching points
(type 3).
In the auxiliary cable continuous suspension bridge,
the stiffenening trusses are not supported on the main tower
bearings but only on the auxiliary cables and on the main cables.
Vertical deflection of a concentrated moving load
A theoretical analysis shows that the velocity effect
on dynamic deflection under distributed loads (road traffic) is
far less than under a concentrated load.
By considering the local deformation of the bridge,
there are two different gradients to consider : the gradient in
the direction of the vehicle movement, the gradient of the
track under the vehicle. For a concentrated load of 100 t, the
running and track gradient for a bridge type 1 with similar
length are given in figure 43.

Fig. 43

running gradient
track grad i ent
for V: 100 km/hr

323

So it appears that, if the track gradient is an

angle a , the trajectory of the moving loaa is in the direction
a . For a simple supported beam, the running gradient is
double
the track gradient a = 2a . In addition these gradients
are greatly affected by the velocity of the moving load: Fig. 44.
The dynamic amplification factor can reach 2. The vertical
acceleration of the stiffening truss caused by the moving load
with only smooth running is given in Fig. 45.
2

Speed

.........

\""".j.":":

....;'
...........
__ 4

65 kmthr

AmI,
'Ih,. ,.

0 5 0 '00 15021>0250
Dynamic amplification
- running gradient
.... track gradient

Max vert. ace.

0.022 9

no

0.030 9

250

0.162 9

Fig. 44

Fig. 45

Deflection change due to train passage

In the case of a train with a length of 400 m and a
linear load of 1.2t/m a numerical application was done to
investigate the essential behaviour of auxiliary cable suspension
bridge. The experimental results are in good agreement with the
theoretical values for all bridges types and all load conditions.
The deflection change due to train passage is given in Fig. 46.
Vibration tests
The vibration modes and natural frequencies have been
studied experimentaly (bridge model of 48 m) and analytically
(multimass system). The general vibration properties of an
auxiliary cable suspended bridge are similar to those of a
classical type 1 except the damping effect which is better for
the type 2 and 3.
Vertical resonance : with increasing the train velocity in the
experiment it was found that the first symmetric mode is
obtained at 328 km/hr,far from real traffic conditions.
The very difficult problem of mechanical interactions
bridge-traffic has been followed up in this research without
needing to refer to hypotheses. All basic effects have been
simulated including earthquake shaking and wind excitation (wind
tunnel). An optimum auxiliary cable geometry improves greatly the
serviceability of train and highway traffic:

324

400 m train

6 m
5

--r---.,~---

4
3.l---~'"

,,

o
\
\

lBridge Type 2

~------~----- -...
,- - - -/

./

.'

Bridge Type 3

"

300m

1l00m

Fi g. 46

""

""' "-

",

""

300m

'.

""

325

~/.~

,
~ia

'8r ;J!FZ IfYOt>J<v.>

ordtZr
ty~

4oo~

-f~vtZ.ncitt!)
('t>""we..'.! ,"~'>IJrc.! Damp/~

frtz-9 v rt.n cie<:J

(''''''D.fd hl4116vreJ
0. 651.1 0100

0.64'''

~683

.t

0.827

0.4'(.1

0.11"

0.1'7

1.3/0

/.361

1./U4

1.3.B

0133

0.100

o.4'J?

ad'6;t

o.(.r! 0120
0.199 0.8'-10

1.3'3

1.34'3

U9J

1.300

a.f.].!

a100

0161

O.sr-o

0.of1e:.

ao~!>

.z

"3

o.,ro

o.no

aJ{'1

1./5'

(J,

~-

1.21/

'----'----

a.ol"

---

Dalnp;"..!!
c.ot,s"

0.0.288

aal4~J

I. aU 1.233
- - ._------

Fig. 47

a milder change in deflection of stiffening trusses and

compared with a classical suspension bridge only a third of
the maximum negative bending m~ment.
In the central part of the bridge, the mechanical behaviour is
similar and the damping value is higher.

V. SIMULATION AND MODELING

1. Choice of mechanical properties of bearings in order to
improve the safety
of a boxgirder bridge subjected to seismic
vibrations.

*"F1,.

a) Mechanical characteristics of bearings

For the vertical transmission of bridge-deck load,
restrained elastomer bearings are used more and more. They-are
composed of a succession of elastomer layers vulcanised with

,t"l plate, .(rec;,",ar or

road

Ci;~:ar

railway - bridge bCilrings

type)

*Te-tr()n-

circular type

48

326

In France, these bearings have been sized accordin0 to a

static design calculation.
The experimentation and the
normalisation of the different types of bearings have been led
by ORE* and SETRA * .
The mechanical characteristics of the elastomer material
vary with the type of chemical formula. For the bearings, the
recommended shore hardness reaches 60 and the advised modulus is
G = 0.8 MPa for road bridges and G = 1.1 MPa for railwaybridges.
Vertical compression :
A vertical load N on a bearing produces an average
stress for a rectangular plate (a x b)
N and a
oM = a.x-t)
shearing 'N at the interface steelplate-elastomer. The
maximum value is obtained in the middle of the length a.
_ 1,5 om
--S-

S is a shapefactor obtained experimentaUy. The shearing stress

has increasing values from the middle of the bearing plate to
the sides where it reaches the maximum values. If there are local
damages for a rectangular bearing, they always appear towards the
middle of the length. ('N = max)

~ ~'
-- !r

~
/

I a..

1.".-

---

l"

subjected to a vertlcal compression

* ORE
* SETRA

Fig. 49

[JJ1D" ;...,.,..-1
I~

OfstrifJution of shearing stress

7:

'l~

A_

of a bearing

Office de Recherches et d'Essais de l'Union Internationale

des Chemins de Fer
Service d'Etudes Techniques des Routes et des
Autoroutes

327

Horizontal forces
All the dimensional variation of the bridge (temperature, creep
... ) with norizontal forces applied produce distortion in the oearing and a shearing stress a H is developed.

~
.

II

Ii

"___ 2.~

f~

Fig. 50
Rotation
In addition an elastomer plate can be a subjected to a
rotation a and the corresponding stress is T . When the
linear or angular displacements are too large, a the bearing has
to slide so that overly larger efforts can be avoided on the
bridge deck as on the foundation ; the friction coefficient on
a PTFE plate is experimentally given (approximately a few percent).
Dimensioning
Whatever the types of actions applied to the elastomer
bearings the shear-stress is limited T = TN + TH + Ta < 5G
as well as normal ~)stress aM < 15 MPa
For a railway bridge the values T = 1,5 TN + 1,5 Tare
a
recommended because of the fatigue phenomenon resulting from the
great variation of the vertical ,loads.
By their elasto-plastic adaptation, the bearing allows
a good distribution of loads even with small differential
movements of the foundation; after an initial static deformation,
a supplementary adaptation must be taken into consideration.
Dynamic Behaviour
The visco elastic properties interfere at the time of
the transmission of dynamic charges. However,in case of road
traffic the dynamic components remain limited, even if there is
a small vertical oscillation around the average position, the
amplitudes remain low compared to the dynamic bridge deflection.
The influence on proper frequencies of the bridge consequently
remains insignificant. In any case the damping effect on the
bridge structure will be limited and can not easily be separated

328

from the foundation damping.

The elastomer bearing~ ar~ designed for long service in
normal conditions. However, in case of seismic vibrations, the
stress and deformation might be much stronger and can reach the
point of damage.
In the limit conditions, the choice of the mechanical
characteristics of the bearings has an influence on its own
serviceability,
as we shall show in a real example on mechanical coupling between foundations abutment and bridge-deck
mOllements.
b} Choice of elastomer bearing of a boxgirder bridge (in
prestressed concrete) in order to improve the mechanical behaviour
under seismic action.
The bridge
It is a big highway bridge with several spans, achieved
by prefabricated concrete bo~-elements. The elements situated in
the central zone of the spans are made of light concrete (fig. 52).
Seismic risk, Shake-program
the bridge is situated in a low risk seismic area. The
spectrum has been obtained from surface recorded seismic movement
in a nearby area with similar seismicity. The characteristics of
the seismic wave have been computed for a height. midway up the
foundation piles (fig. 51).
The ground has the following characteristics :
- powdery so i 1
- C Menard ~ = 3.5
- Pressiometric modulus 285 kg/cm 2
- G = 357 kg/cm 2 horizontal
" = 0,4
- E = 1000 kg/cm 2

Fig. 51

PERIOIl(

Ir--.--.----,.--__
on-"

0.24

:c:

AA'

~~----.----

.,

II

lr ~-~.

-.

I ~

O:'~UJ ___,j---

o.7'l

..

"'OJC

- ,.--,

1~QQ.~

172 m

320 bars
21 bars
Density = 1.B5
Mod. E = 21.0 x 106 KN/m2
(+)

1 110 tons

"28 (-) =

"28

4J 37.5

350 bars
28 bars
Mod. E = 39.3 x 106 KN/m2
"28 (-)

"28 (+) =

_6~~---1

144 m

TOTAL WEIGHT BRIDGE DECK

prefab. elem

U gh:t concrtte

Traditional concrete
prefab. elemt

Fig. 52

__ ~ ___ _

-L.

00

-L

n0

330

. For horizontal action:

It appears that the acceleration Yo = 0,4 g is not
easily transmitted to topsoil Layer. A more realistic acceleration
has been applied y = 0,15 g It corresponds to a 6.5 seismic
o
magnitude with its epicenter 40 Kilometers away .
For vertical action:
It is supposed that the vertical charges are diffused
from the bottom of the abutment (6 x 10 m) into a pyramidal slope
model of 45.
The Shake-program yielded the following G moduli
from the top to the bottom of the pi er : G va ryi ng from 550 to
770.

Dynami c behavi our of the bridge

All characteristics of the real bridge are well-known
(dimensions, masses ... ). A dynamic modulus for the concrete of the
pier E = 30x 10 6 KN/m2 = 300000 kg/cm 2 has been admitted.
The ANSYS program was used with the specific purpose of
determining the influence of the mechanical characteristics of
bearings on the stresses in the foundations as well as on the
bridge deck.
The first oscillatory modes are represented in Fig. 53.
Comparative test of two solutions of bearings
The ANSYS program has made it possible to determine all
the actions on the different elements of a bridge with the results
of the ground analysis (Shake). The problem was to determine
which of the 2 solutions I and II for the bearing makes the
foundations and bridge a safer unit.
The 2 analyses differ only in the stiffness matrix of
the bearings. The "tetron" sliding bearing types (bi or multidirectional) are supposed to have a grip coefficient of 2 % (constant with temperature and vertical load).

331

I
~!L
i~ ~,,=o--I--::=*I:=x
:::::~-=~-=t-A---:====---A-::::I"
.42

-----,I

'

.60

.67

Fi g. 53
4

.85

1.03

Z 1.26

I--~~~---"""-<-

1.38

- -- -

- - ----

--~~=---

1.60

The types of bearings used are shown in Fig. 54.

Results
The forces and moments produced by the precedent
seismic action are plotted for solutions I and II for the
diff~rent bearings (from a to 5) and given in Fig. 55.
Analysis and comment
Solution I :
The bearings of pile 2 and 3 are over-loaded compared to
the maximum tolerated. As the neoprenebearingsare not strong,
the maximum seismic acceleration values acceptable are:
Yo = 0.08 g.

332

PERHAN t
LOAD

TYPE OF BEARINGS
~

c:

tons

SOLUTION II

SOLUTION I

::.:
0

2 TETRONS o2E 400

1 TETR~N 02E 2500

1 TETRON D2E 2500

+
+----

Neopren block5BOO x BOO

4 [12 + 3]

ident,cal

1 TETRON 02E 2000

1 TETRO~ O:E 2000

Z TETRONS 02E lZS

;,. .s-,
sS-'
0

1
2
3
4
5

identical I

Neop. blocks
00 x 900

9[20+3]

--L,",I;>

--

Ix

4090

Fi g. 54
identical I

2450

identical I

2500

4 Neop

block~

500 x 500

520

-t--+-

'2910

300

identical I

r~

11,

l1y

151,:'11

WOq

II

177B

bJ ':$3

Ii 116

.7-:5

2305

6472

40

II

7324

21,7

189?

175

Si34

2b67

4~O4

2B~

181
l6,

_I

IS"

11
I

~b91

l411

4419

190

:bl

II

6458

2276

4810

337

356

IS GIl

4432

:39

JS"

I} ,"(,1

27

ZO 173

J~:';

11

MuD

~1~!J

3701

7609

II

3825

2475

1691

2~6

110

.....

I
II

2430

19 79;

2418

1 338

Dynamic action on bearings

Fx : kN

Fig. 55

"x

kll.m

I":.

333

The foundations at 1 and 4 are strongly crosswise loaded

because of the great s ti ffness of the "tetron" beari ngs, wh i ch do not
allow transversal sliding.
Lengthwise, by contrast, bearings 2 and 3 transmit the
quasi-totality of the charge. The bridge-deck will be submitted
to relatively weak tractions in a zone normally under compression.
The fissured zone will remain very limited without being able to
provoke rupture.
Solution II
A sensible modification of the dynamic behaviour is afforded by
elastomer block-bearings. The system having a more homogenous
transversal flexibility, the dynamical behaviour is outrightly
improved transversally(G dyn = 160 t/m2). Lengthwise, supports
1 and 4 must underpin the dimensional variations of the bridgedeck. To avoid a too large distorsion in elastomer a larger
hei ght of. the beari ng is necessary.
A weaker longitudinal stiffness results from this and
it is again piles 2 and 3 which take the majority of the
longitudinal seismic force.
Conclusion
The most important benefit of solution II is obtained
on the foundations. The forces on the different abutment piers
of the bridge are considerably reduced compared to solution I. We
can outline the most damaging vertical and hori~ontal actions due
to dead-load and seismic-load for each solution, in the table
of Fig 56
BEARING
SOLUTION

IA,-,",

~fi/61"11C.
ACTION/F"I!~

1
t N-

N+ ton

T-

~
II

N.

1807

1433

1471

16~

203

409

453

:;85

163

142

H':

191

1632

1376

114

liZ

t NT-

Fi g. 56

Solution II shows a better distribution of the

horizontal charges, which are brought down to an acceptable
level.
In both solutions, the bearings situated in 0 and 5 are
subjected to the same actions (pendular counterweight) and the
vertical loads applied on the different abutments have similar

334

values.
The dynamic behaviour of the girder-foundations unit is
inseparable from the mechanical characteristics of the bearings,
especially in case of seismic action.
The dynamic analysis of the mechanical system with
solution I and II makes an advised choice of the bearings possible.
Safety is increased by limitation of the forces in the whole
structure.
2. Effect of profile imperfection of the bridge subjected to
moving loads
The dynamic effects due to traffic are generally
introduced in calculations with a dynamic amplification coefficient.
This coefficient is not always the consequence of a
dynamic approach of mechanical response or of experimental
measures.
These coefficients depend on different parameters, the
different national code values show significant differences
because they are not based on the same parameters.
Most of the codes give a global dynamic coefficient
which allowsthe obtention from a static response of the so-called
dynami c moment and defl ection of the bridge.
A more subtle approach has been conducted by Mf.Palamas
(L.C.P.C. Paris). Mr. Palamas shows in fact, that the
coefficient has to be defined for each section of the bridge. The
dynamic majoration coefficient of the vertical deflection (Cd)
and of the moments (Cm) for a given section then depends not only
on the mechanical characteristics of a bridge and of the moving
load, but is also greatly influenced by the profile imperfections
and the vehicle speed.
Calculation of the bridge-moving load system was
realised by a computer program "Palmares", the numerical application was done for a classical concrete bridge and one vehicle.
The essential characteristics are given in Fig. 57.
When the profile is without irregularities and when the vehicle
arrives at the entrance of the bridge without vertical oscillation,
the force exerced is constant and Cd at midspan, scarcely
exceeds the unit value (Fig. 58).

335
llechanical charaterist1cs :
Bndge ..pan 34. or 11

2 ..

fondalllental Freq. 1f34 .. 4H_Z

Dead Load 11.4

Yehtcule mass

+---~dL---~--~Q-----r

----~~z=-==-s=~~~==-====~~

"17" 16.7 HZ

1. 9.92 x 10 10 "/ra2

RlgtcJUy

t/~

31.7 t

suspension stiffness

k 9.123 ZOO N/_

" 85 600 N sec/II

suspension dal!ping

vertical period
veloclt,yV

To 0,37 sec

o< V<

100 KlD/hr

Fig. 57
1t,1).

S'"

JIIc"UC~.1II'

"tn/,.,

11.55
L. ]q",

Fig. 58

If before arrlvlng on the bridge-deck the vehicle runs over an

obstacle which induces oscillations, the bridge is also subjected
to this effect and Cd reaches about 2 for 100 km/hr (Fig. 59).
If the bridge-deck presents a global imperfection,
the bridge-motion is attenuated compared to the precedent case
(Fig. 60).
If a two-span bridge is affected by global defects,
the influence of these two imperfections on the deflection of
the span are given in Fig. 61.

Fig. 59

Fig. 60

,.1

336

~~
rUD
I

Fig. 61

$~",

tit,."""...

L"

-I

A local imperfection at mid-span

The vehicle, this time, gives a great impulse and the
bridge shows an important couplage. The graph-representation of
the dynamic coefficient variations in the different transversal
sections of the bridge shows clearly how a unique value
(generally accepted in codes) is of poor signification.
For a local obstacle, situated at mid-span, the
velocity at the vehicle has a great influence on the dynamic
moment CM (Fig. 62 and 63).
As it concerns an energy transfer from the vehicle to
the bridge, the vehicle damping as well as the bridge's has great
effect on the value of CM . The velocity of the vehicle is also
important as shownin Fig. 64 and Fig. 65.
d

f-.V= 1001,,,,/),,._

Fig. 62

Fig. 64

./

Fig. 63

Fig. 65

337

This study shows how the profile imperfections can

affect the dynamic bridge behaviour. A local imperfection can
increase deflection and moments much more than the dynamic code
coefficients let us suppose. The immediate mechanical consequence
of profile imperfection consists in a more severe stress cycle
(level and number of cycles) so that the damage process can
affect considerably the life time of the bridge.
During the final stage of a bridge construction heavy
vehicles crossing the uneven deck can develop higher stress than
in the future life time.
Moreover, in certain conditions road surface profiles
can be modified, for example when differential melting snow
occurs. This is why in different circumstances, in certain
countries, in addition to the classical dynamic coefficient curve
a second curve with higher values has been introduced in the codes.

338

CON C L U S ION
The problem of the dynamic behaviour of a bridge is
complex. Different methods of measurements and analysis process
have been presented. The control of the bridge response under
traffic demands important infrastructures. Whereas the dynamic
tests, such as vertical impulse or sinusoidal excitation which
can be repeated in the same conditions, are the indicators for
survey and maintenance.
This paper shows that on the one hand different bridge
types must be tested in situ in order to establish the diagnostic,
and on the other hand, the dynamic behaviour of new elements and
material must be evaluated in advance by a more scientific
approach through laboratory tests and mathematical simulation.
These 2 aspects are complimentary. In the stage of bridge design
the life time prediction must be obtained with an acceptable level
of threshold. That means that the whole structure must resist
simultaneously two types of phenomena :
- mechanical loading, which introduces stress cycles with the consequence of fatigue and cumulative damage
- physical and chemical environmental effects and their interaction with the first one.
The mechanical characteristics can be considered as
indicators of the vibrational behaviour so that the durability of
a bridge or bridge element is nothing more than the maintenance
of performance in time. The value of frequencies and damping of
the different modes are with the stress the main indicator. Of
course their evolution depends on the environmental characteristics such as temperature effect, chemical action, corrosion,
time.
In laboratories,these characteristics can be more or less
simulated but the correlation with the in situ behaviour is not
easy when the time factor is of great importance.
In fact, concerning the environmental agents which
can attain the mechanical performance, the life time cannot be
correctly estimated. Exterior conditions, bridge type, load
cycles ... vary to such a great degree that present experience
is almost always founded on numerous spot observations made in
situ.

339

More precise methods of damping control must be

established to facilitate the appreciation of the state of
cracks and the cumulative damage. These methods work well under
laboratory conditions at the present time as in situ for damaged
bridges. A methodology situated between the 2 extremes has yet
to be elaborated for a better survey of a given section or of
the whole bridge.
So the measure of the evolution of all indicators amplitude, frequency, damping - must be accomplished to improve
the in situ diagnosti~and permit recognition of the cause in
order to choose the appropriate maintenance intervention : girder
bearing foundation ... This also contributes to improvement of
codes and standards concerning safety and life-time.

REFERENCES
CHAPTER I
- Bendat J.S., Piersol A.G. - Random data: analysis and measurements procedures. Wiley Interscience, New York, 1971.
- Guillard Y. - Vibration dues au trafic routier, degradation de
l'environnement. These 3e cycle. Fac. Science, Universite Le
Mans, 1978.
- Zeller E. - Comportement vibratoire d'une structure metallique
test impulsion et excitation sinusoldale. Etude Juin 1982.
CHAPTER II
- De Back J., Ko I stei n M. H., Van Leeuwen J. L. - r1easurements and
interpretation of traffic loading in bridges. Lewin Inst.
Univ. Technology Civil Eng., DELFT, 1981.
- R. Pfeiffer, J.M. Zaschel - Ermittlung der Verkehrslasten und
der Verkersbeanspruchungen von Stahlbrucken als Grundlage fur
kunftige Bemessungsregeln, LBF Fraunhofer Institut, Darmstadt
FhG, Berichte n 4, 1979.
CHAPTER III
- J. Caracilli - Constatation sur ouvrages metalliques en
construction et en service. Etude Paris 1980.
- L. Ferron, E. Zeller - Comportement vibratoire d'un plancher
industriel, du calcul aux tests dynamiques. Construction
Metallique nO 4 1982.

340

CHAPTER IV
- H.P. Lieurade - Essais de fatigue sous sollicitation d'amplitude variable, RFP 255, IRSID 1978.
- C.P. Heins - Bridge response and damage, J. Washington Ac.
Science 1979, 69 n 1.
- M. Gregoire - Interpretation des resultats d'analyse statistique de sollicitations relevees en service, CETIM. 1961
- B.T. Yen - Minimizing fatigue and fracture in steel bridges,
AS ME 1979 Conference, Washington.
- J. Caracilli, A. Aloustaz - Influence de la precontrainte
sur 1 'amortissement des vibrations transversales des poutres
en Beton Precontraint, Bulletin Liaison Labo P.C., Mai-Juin
1978.
- J.L. Van Leeuwen, AJM Siemens - Miner's rule with respect to
plain concrete, Heron vol. 24, 1979.
- H. Van Koten - Structural damping, TNO CUR report n 75, 1977.
- Y. Yasoshima, Y. Matsumoto, T. Nishioka - Studies on the
running stability of railway vehicles on suspension bridges.
Journal of Fac. of Eng. UnivcrsiLY of Tokyo, n 1, 1981.
CHAPTER V
- Appareils d'appui en elastomere frette, Bulletin technique
n 4, 1974.
- Palamas J. - Imperfections du profil d'un pont et vibrations
sous trafic, recherche effectuee au LCPC Paris, These 1982.
- Y. Guillard, E. Zeller - Initiation Dynamique des Structures,
constructions parasismiques; mesures sur ossatures industrielles
et ponts sous trafic routier. Cours Ecole d'Ingenieurs
Geophysiciens - Institut de Physique du Globe, Strasbourg.

v.

EARTHQUAKE AND WIND EFFECTS

343

SEISMIC DESIGN OF HIGHWAY BRIDGES

C. P. Heins, Professor
Institute for Physical Science and Technology and
Civil Engineering
University of Maryland, College Park, MD. 20742
INTRODUCTION
With the

~dvent

of the 1964 Alaskan Earthquake, the 1971 San

Fernando Earthquake and more recently the 1978 Santa Barbara Earthquake
(1), bridge structures in the USA have undergone considerable destructive
forces.
The earthquakes have caused the bridge professionals to reassess the
design techniques that have been applied, up until that time, for seismic
design.
The present 1977 AASHTO bridge code (2), as related to seismic
design, suggests an equivalent static force method for simple structures
and when the structure is complex, for example curved bridges, a
computer based response spectrum or dynamic analysis should be considered.
In this present 1977 AASHTO code, most engineers would utilize the
seismic coefficient method (SCM), however, the utilization of the SCM,
may give erroneous results when designing under seismic conditions (3).

3M

Because of such errors and from experience gained from recent earthquake
the 1975 AASHTO code was reassessed in 1977 by the ATC (4) research group.
The work of this council, developed a new specification (5), which will be
given in part herein, relative to straight girder bridges.
Although the ATC-6 design guide (5) has advanced the state of seismic
design of highway bridges, it was recognized (6) that additional research
was necessary to determine the seismic response of bridges which are
influenced by geometry and curvature.

Such a study has been conducted (7),

resulting in a proposed design guide for curved steel box girder bridges.
These results will be given in detail in the following section.
STRAIGHT HIGHWAY BRIDGES
The analysis techniques, as recommended in the proposed design guide (6),
are:
1.

Single-Mode Spectral Method (SMSM)

2.

Multi-Mode Spectral Method (MMSM)

The (SMSM) procedure can be applied by considering;

i)
ii)

Distributed mass and equivalent static loading (5)

Equivalent structural stiffness (k) and equivalent static
loading (8, 9)

in which both methods can employ simple analytical techniques not requiring
dynamic computer programs (10).
The (MMSM) requires modeling of the structure as a lumped mass or
distributed mass continuum and application of a given response spectrum
or time history.

The difficulty involved in evaluating the various node

345
shapes and eigen vectors and actions requires use of a computer program (1).
Such application is generally warrented when the structure has an unusual
configuration.
The details to be presented herein will only be relative to the (SMSM)
technique and direct design criteria.
General - The development of the (SMSM) technique requires utilization
of the generalized displacement function concept.

In order to explain this

idea, consider the static displacement of a general girder given as;

(1)

Eq.

(1) can also be written as;

(2)

where;

Amplitude terms

~n

displacement functions

In the case of dynamic response the displacement function can be

described as;

l\(x, t)

where:

~(x)

Z(t)

Z(t)

Amplitude of motion

~(x)

Shape function

(3)

The generalized displacement function as given by Eq. (3), will now be used
to develope design oriented equations.
Formulation of SDOF Equations - Consider a SDOF system, as shown in Fig. 1,
subjected to a ground motion.

Summation of the forces gives;

346

mli

where: {f1

(4)

= cl:J.

fD

(5)

fS = kI:J.
The virtual work method will now be used to obtain the general equations
of motion; and is given by;
oW =
where: JrF

ol:J.

=0

(6)

Applied Forces

lLol:J.

IF.

= Virtual

Displacement

The generalized displacement functions. including ground displacement. can

be written as;
1jJ(X)

Z(t)

+ /).

(t)

(7)

Substituting Eq. (7) into Eq. (5) gives;

f1 = m[1jJ(x)Z(t) + li g (t)]

fD

c[1jJ(x)Z(t)]

fS

k[1jJ(x)Z(t)]

1jJ(x)oZ(t)

(8)

347

Substituting Eq. (8) into Eq. (6) gives;

6w

= f:m(x)[~(X)Z(t) +

JL

~g(t)]~(x)dx6Z(t)

+ oc(x)[w(x) Z(t)]W(x)dx6Z(t)

Since

6~

f:k(X)[W(X)Z(t)]~(X)dX6Z(t)

~(x)

6z(t)

is arbitrary, then set equal to any value

and also assume;

y = f:(m(x

s=

[w(x)2]dx

f:m(X) [W(x)]dx
(9)

J:C(X)

c*

k*

= J:k(X)

[~(x)

+ c*

Z(t)

] dx

therefore;
y Z(t)

+ k*

Z(t)

= -!J. g (t)

or
Z(t) + (c*)
y

Z(t) + (k*)
y

Z(t) =

..

-~

(t)-

(10)

Let:
~

c*
2m (x)w

=-;:---::;~

Then the general equation of motion is;

=

~ (t)~
g
Y

(11)

348

The solution of Eq. (11) is obtained using Duhamel integral;

Z(t) = ~ t;""(t)

l1(t) = t:"(t)

where:
and

III

III

= JtX (T)exp[-~(t-T)sinlll(t-T)dx
og

t:"(t)

which is the earthquake response integral.

The local displacements are therefore;
l1(xt) =

l/J(x)

Z(t)

l/J(x)~

!let)
III

(12)

Local Forces:
Neglecting damping:

= m(x)

fl(x,t)

i;(x,t)

for harmonic motion;

11 = A sinw t
{

b.

III

A cos

-Ii

Substituting
..

III

A sin lIJt

b.

= A sinwt

gives

11=-1Il11
2

therefore

fl(x,t) = -m(x)1Il b.(x,t)

or

fS(x,t)

fS(x.t)

= m(x)

= -fS(x,t)

[m(x)
(13)

349

Maximum effects:
Displacements
Using the response spectrum, the spectral velocity, which by definition
is a maximum of the response integral or

also from the response spectrum

~(x,t)max

1
s
w v
n:uS

sd
mS

therefore:
~

max

and
S

cs

.. t::max

c s g/w
(14)

Forces

a-

fs(x,t)
fs(x,t)

m(x)w(x)yw~(t)
=

m(x)

w(x)~

m(x)

. w(x)~
y

Sa

wOJ

However

Sa = g
{

m(x) = w(x)

and

350

=~
y s w(x) ~(x)

fS(x)

(15)

where:
Cs

= Elastic

Seismic Response Coefficient

Natural Frequency:
The natural frequency can be determined by equating external and
internal work.

Consider a simply supported beam subjected to a load

as shown in Fig. 2. the external work is;

f Fll

WE

where:

= Po
2

JL t.(x)dx
0

po. a

2"

f:t.(X)dX

the kinetic energy is given by;

T

2
00
~ --

max:

max:

m(x)v 200 2dx

~ 2dx
w(x)(t.)
s

max:

= !!LL

fo
L

w(x)(t. ) dx

2g

Equating WE = Tmax: gives;

U

max:

P
o

2"

2
2g

_~

a -

max:
and

00

21T

Therefore the natural frequency is

T

(16)

Po

351

Elastic Seismic Response Coefficient Cs

Computation of the SRC is utilized in conjunction with the displacements
given by Eq. (14) and the local forces given by Eq. (15).

The coefficient

Cs ' given as follows, is a function of the acceleration coefficients, site

conditions and natural frequency (given by Eq. (16, and is;

where:
A

= Acceleration

Coefficient (Fig. 3)

Site Coefficient Table

Natural Frequency (Eq. 16)

General Coefficients A, S:
The coefficient A and S given in Eq. (17) are prescribed as follows:
Acceleration Coefficient A. - The acceleration coefficient A, is given
in Fig. 3, as applied to the United States.
Site Coefficient S. - The site coefficient approximates the effects of the
site conditions on the elastic response coefficient and is;
(S) Coefficient
Soil
I

1.0

Profile

Type

II

III

1.2

1.5

The soil profile type identification I, II, III is prescribed as

foll~'s:

SOIL PROFILE TYPE I is a profile with either

1.

Rock of any characteristic, either shale-like or crystaline

in nature (such material may be characterized by a shear
wave velocity greater then 2,500 ft/sec (762 m/sec), or by

352

other appropriate means of classification); or

2.

Stiff soil conditions where the soil depth is less than 200 ft
(61 m) and the soil types overlying rock are stable deposits of
sands, gravels, or stiff clays.

SOIL PROFILE TYPE II is a profile with stiff clay or deep cohesionless

conditions where the soil septh exceeds 200 ft (61 m) and the soil types
overlying rock are stable deposits of sands, gravels, or stiff clays.
SOIL PROFILE TYPE III is a profile with soft to mediUm-stiff clays and
sands, characterized by 30 ft (9 m) or more of soft to medium-stiff clays
with or without intervening layers of sand or other cohesionless soils.
Design Procedure - The design technique utilizing the previously given
equations, is performed as follows (5):
STEP 1
-----Apply unit uniform loading

Po

and evaluate displacement functions

in transverse direction Fig. 4(a)

W(x) , similarly apply unit

longitudinal loading, Fig. 4(b), and evaluate

~(x)

STEP 2
Calculate parameters
a

y
where

J ~(x)dx
J m(x) w(x)dx
J m(x)~(x) 2dx

a,

or
or

m(x) is the mass and

structure.

8, and y

f w(x)~(x)dx
J w(x)~(x) 2dx
w(x)

weight of dead load of bridge super

353

STEP 3
Calculate natural period;
unit load

STEP 4
Calculate equivalent static earthquake loading:
f s (x).

=~
w(x)$(x)
y s

(force/unit length)

STEP 5
Apply loading fs(x) to structure and determine resulting member
faces and displacements.

354

CURVED BOX GIRDER HIGHWAY BRIDGES

General - As indicated previously, the proposed highway design code (5)
does not include curved girder bridges.

Therefore a study (7) has been

undertaken to develope such a design technique, which will now be described.

THEORY
The general static response of curved bridge structures requires
incorporation of the interaction between the bending and torsional forces.
Such interaction can be considered by solving Vlasov equations (11), or by
development of ths stiffness matrix (10) and appropriate restraint conditions.
The matrix oriented technique, however, is more versatile, in that a
three dimensional model (space frame) can be considered.

This then permits

modeling of the structure such that the support conditions can represent the
physical restraints.

Also. the induced actions can be applied in three

directions. and thus simulate the various earthquake induced actions.

Therefore. the study of the induced actions on a structure, when
subjected to earthquakes. was confined to utilization of a space frame
matrii simulation.

The basic modeling consists of series of typical

elements attached rigidly together to form a continuous curved box

girder bridge.
The basic properties of each beam element will consist of I I

~.

Although warping and distortional properties

and

(Iw,Wn,Wa) can be

computed. they were not considered in this study. as it has been shown
(12). that with proper bracing of the box girder, warping and distortional
effects are negligible.

355

Therefore, by proper evaluation of the stiffness of each beam element,

and identifying each joint load, the static response of the continuous
curved girder can be determined.

The static response can then be utilized

to determine the effective earthquake effects, using the response spectrum curves.
The general procedure using this method will now be described.
The support restraints, to be imposed on the bridge model, can be identified
as releases in the computer model.

Because a space frame model is being utilized,

six releases or rrstraints were considered.

Equivalent Dynamic Analysis.The natural frequency response of a single degree of freedom system can be
predicted by the following;
w

=1 ~

(18)

where
k

spring constant

mass (wIg) or mass moment of inertia

If the natural frequency

w of the system, as computed from Eq. (18)

is reliable, the the corresponding induced acceleration of the

the mass

created by an earthquake can be predicted by using the response

spectrum curve given in Figures 5, 6, and 7.

The resultant accelerations ob-

tained therein are then used to determine the induced dynamic force

M . ~.S.

356

where
~.S.

linear acceleration obtained from response spectrum

curves.
If the system is subjected to angular accelerations

e,

then the induced

dynamic torque is
M
where:

(20)

laRS

pr 2 dA

mass moment of interia

eRS

rotational acceleration obtained from the response curve

mass per unit area.

1J

steel

J<x2+y2)dA +p

concrete

J(X 2+/)dA

Natural Frequency.As indicated by Equation (18 ), the natural frequency of a SDOF system is
given as a function of the spring stiffness (k) and the spring mass (m).

If the

system is a bridge structure the spring constant can be represented by:

k

(21 )

where:
~

= induced

maximum displacement caused by a unit load.

Therefore by determining the response of a given box girder bridge, when subjected
to a unit load, an equivalent spring constant can be obtained.

This constant,

k , and the total mass of the bridge, will then permit evaluation of the natural
frequency as given by Eq. (18).
In the instance of curved structures, the dynamic action can occur in three

principal directions, and one primary rotation as shown in Figure 8.

The

357

resulting maximum displacement. induced by these unit loads. will then give the
correspond1ng equivalent stiffnesses

kx k_". k
y
z

The corresponding natural frequencies can then be determined by applying Eq. (18).

BRIDGE STUDIES
Typical Sections.In order to develop a simplified design technique. the response of
various curved box girder bridges must be examined.

Such box girders.

which have been used in previous studies (13). were utilized in this
parametric study.

In this study only the three-lane, three girder

system were considered because this is most typical of the curved box
girder structures (12).
The basic span length configurations that were examined, are shown
in Fig. 9, where L=50' to 150' and N=l,2.

The radius for these various

structures that were used, varied from 200 feet to infinity.

General.Utilizing the basic box geometry and the support spring constants of

k = 0 ~ 2 x 10 3 k/in; K
x
y

co

0.5 x 10 3 k/in.

The equivalent seismic resonse of the single, two, three, and four span
structures were examined.

The resulting natural frequencies (w , w ,

x
y

for all bridge spans and their corresponding induced accelerations were
then obtained (7).
given by

and

For the continuous spans, the pier flexibilities, as

k

were also included as a variable.

wt )

358

The induced accelerations, as determined from the response spectra,

L

have been plotted as a function of (i) and the span length (7), for
the single span structure.

The relationship between the corresponding

accelerations for the continuous spans and the Single span however have been
plotted as a function
F

F, where

Ax (continuous span value)

A (single span value)

(22)

versus the number of spans, radius, stiffness and span lengths.

For all the span lengths of 50',100', and 150', similar relationships
between F , F , and F t have been plotted (7).
y

These data were then used to

develop appropriate design criteria.

DESIGN CRITERIA
Trends.The seismic design of continuous curved box girders were related to
the response of single span curved girders.

Therefore, the single span

accelerations (A , A ,A and A ) were determined with respect to tbe basic

z
y
z
t
bridge geometry, giving the following:
A

2.2 (~)2+0.0llL+0.45

(23)

-0.016(L)+4.7

(24)

3.8

(25)

for 100' < L < 150'

Ax

for L > 100'

for L > 100'

....L
Y

A
z

359

(26)

at = 1. 5

for L > 100'

The continuity factors F have similarly been determined in analytical

form, resulting in the following four continuity factors:
F

(longitudinal)
F

= -0.02

where:

F
J...

(L)

+ 3.75 +

= 0.00125

(27)

for R < 600'

for R > 600'

1.0

(vertical)

(28)

F = -0.125 (NS) -0.002L + 1.35

Y
where: NS = number of spans (2, 3, or 4)
F

(transverse)
F

(29)

-0.005L + 1.5

F t (torsion)
Ft

(30)

= -0.075

where:

NS

(NS) + 1.15

= number

of spans (2, 3, or 4)

Design Approach.The equivalent seismic design of curved box girder bridges will
incorporate the primarily developed equations and the effective peak
acceleration map (Kp )' given in AASHTO.

The general design equation

is of the form:
EQ n

= Fn

EQ n

= Fn

A
n

Translation

(31)

K
p

Rotation

(32)

360

where:
Fn
An

continuity factor in x, y, z or t directions

= single

span acceleration

Kp = effective peak acceleration modifying factor

= wI = total

weight of structure
gravity

rotational mass moment of section

= mass per unit area

EQn

total applied seismic force in x, y, z, or t directions

for the specific direction n, the continuity factor F and single span
acceleration A are given by:
n

Longitudinal Direction (x)

Fx

-0.02 (L) + 3.75 + K

K = 0.00125 R
K = 1.0

for R

<

600'

>

600'

A = 2.2 (1.)2 + 0.011 L + 0.45

x
R
Vertical Direction (y)

-0.125 (NS) - 0.002L + 1.35

-0.016 (L) + 4.7

Transverse Direction. (z)

-0.005 (L) + 1.5

Fz
Az

3.8

Torsional Direction (6)

Ft

At =

-0.075 (NS) + 1.15

1.5

361

EXAMPLE
In order to illustrate the application of these proposed equations,
consider a two span curved box structure of spans 100'-100' and
radius R = 600'

the single span accelerations are:

A

x
Ax= 2.2

(*

)2 + 0.011 L

+ 0.45

100
2.2 ( 600
)2 + 0.011 (100) + 0.45
1.61
A

J....
A

-0.016 (L) + 4.7

-0.016 (100) + 4.7
3.1

z
Az = 3.8

At = 1.5
the continuity factors can now be computed as:
F

x
F

-0.02 (L) + 3.75 + K

(K

0.00125 R,

-0.02 (100) + 3.75 + 0.00125 (600)

2.5

600')

362
F

J...

Fy

-0.125 (N.S.) - 0.002 L + 1.35

-0.125(2) - 0.002 (100) + 1.35
0.9

Fz =

-0.005 (L) + 1.5

-0.005 (100) + 1.5
1.0

Ft

-0.075 (N.S.) + 1.15

-0.075 (2) + 1.15
1.0

the weight of the bridge is computed as

w

= 442.15

Kips

and the mass moment of inertia ( j

2273.63 Kip-in - sec 2

therefore, the induced seismic forces are computed as:

EQ x

= Fx

W
A

(g)

(2.5) (1.61 g)

4.025 W
EQ

=F.A
Y
Y
=

(3.1) (0.9 g)

2.79 W

w
g

= (1.0) (3.8

g)

w
g

363

3.8 W
EQt

Ft. At

(1.0) (1.5) (i)

1.51

the equivalent induced uniform seismic forces qx' qy' qz' qt' are therefore:
E~
4.025 x 442.15
0.7415 kips/in
qx
-TX'"""L
2 x 100 x 12
EQ
2.79 x 442.15
Y
q
0.5140 kips/in
2 x 100 x 12
y
2 x L
EQz
3.8 x 442.15
q
0.7 kips/in
2 x 100 x 12
z
lXL
EQt
1.5 x 2273.63
1.4210 kip-in/in
qt
2 x 100 x 12
2 x L
Evaluation of the reactions of curved box girder bridge, using a
straight girder gives:
i)

x direction
I' E E E E (

qx

(( (C (E C E ( ' (( l ( ' I (E C E' ((

Il

.
2

Rx 1

qx

Rx2
Rx 3

0
0

(2L)

EQx

1779.65 kips

364

ii)

Y direction

rf--~L----ft--~L~----f

= 0.375q y L
Ryl

= 0.375qy L

1. 25qyL

0.375q y L
0.375 (0.5140) (100 x 12)

Ry

231.3 kips

1.25qy L

1.25 (0.5140) (100 x 12)

771 kips

Ry3
iii)

Ry1

231.3 kips

z direction

)~+~+~~~~+~~+~~+~~~~~~+~~~++++++++++

~l

~2

' _ _L=--_--+'

Rzl

Rz2

0.3759, L
z
Rz1

~3

=
=

Rz3

= 1.25q'L
,Z

.075q zL
0.375 (0.7) (100 x 12)
315 kips

Rz2

1. 25q z L
1.25 (0.7) (100 x 12)

= 0.375q L
z

365

1050 kips
Rz3

Rz1

= 315

kips

iv) Torsion

0.375QtL

Rt 1

= 1. 25Qt L

0.37SQtL
0.375 (1.4210) (100 x 12)
639.45 kip - in

Rt 2

1. 25QtL

1.25 (1.4210) (100 x 12)

2131. 5 kip - in
Rt 3

Rt 1

= 639.45

kip

in

In the interior support. we use Ry x d = T.

Ry

=d

2131.5
48

d = 48".

44.41 kips

comparing these results with the results obtained from the (SFS) computer
program gives:

366

~
Direction

No.

Eqs

2
program

E.qs

program

Eqs

program

x direction

Rx

1779.65

1298

223.9

y direction

Ry

231.3

250

771-

775.3

231.3

250

z direction

Rz

315

417.5

967.7

315

417.5

Torsion

639.45

869.8

639.45

869.8

* Ry x

d= T

which shows reasonable agreement.

1050
44.41

55,5*

367

P(t)

or

grd motion

m.6 g

Fig. 1

o(x)

J~~----,~ ~

rt t f t

f f f t f t r~

Po
Fig. 2

368

F'GURE 3.

ACCELERATION COEFFICIENT - CONTINENTAL UNITED STATES

IAn enl.il(ged verSIOn 01 thIS map. Including Countaes,
IS Olven In the back cOlier of the repon)

,/(x)

~I I f I ! 1 t
-- - -I"'(X)

~x

(a)

Po

r r
H"'(X)

Po _ _ x

(b)

Fig.

roo

...,

lJ>

'1

"..

()Q

>

...>-

!:

0.25

'" ,"

3.6

\: ,"

FREQUENCY( CPS )

< ""","

33

,,',<'"

10

i 1000

1000

" " '" "" '" ,,<L

VERTICAL RESPONSE SPECTRUM FOR tOg MAXIMUM GROUND ACCELERATION

01

\OX, ""X"

...,

.....

0'

(1)

"...

()Q

''', """,,,"

'"

,,','"

"'"

< < "

<

'"
< ,.lOOCI

>!:
u

~
5

025

2.:>

9
FREQUENCY (CPS)

l3

1000

HORIZOTAL RESPONSE SPECTRUM FOR tOg MAXIMUM GROUND ACCELERATION

01

__ h.'"'"'"~'"~'" '"'"'"'"'"~'"~ "':--..'-'"'"'"'"~ '"'"'"'"'"'"'"~ '"'i K.l

_100~~~~100

tOOO,
-..J

....

371

0.002r---t-~--~~r-~+-~~~~~~~~-4~~

0.001

L....__J.__ _~~~__J._~:.......J~_~_.l_~~~~

0.1

0.2

0.5

1.0

2.0

FREQUENCY (CPS)

MEAN

TORSIONAL SPECTRUM (AIlERAGE SHEAR VELOCITY- 570 FT/S)

Figure 7

372

(b)

(0)
F

Fig. B

Fig. g

373

REFERENCES
1.

Miller, R. K., Felszeghy, S. F., "Engineering Features of the Santa

Barbara Earthquake of Aug. 1978", BBRI, Berkeley, Calif., 1978.

2.

"Eleventh Edition-AASHTO Bridge Design Specification", Washington, D. C.

3.

Imbsen, R. A., Nutt,' R. V., Pensien, J., "Evaluation of Analytical

Procedure
Used in Bridge Seismic Design Practice", Proceedings:
Earthquake Resistance of Highway Bridges", ATC, Jan. 29/31, 1979.

4.

Sharpe, R. L., Mayes, R. L., "Development of the Highway Bridge Seismic

Design Criteria for the U. S.". Proceedings: "Earth<luake Resistance of
Highway Bridges", ATC, Palo Alto, Calif., Jan. 29.31, 1979.

5.

"Seismic Design Guidelines for Highway Bridges:, ATC-6, Palo Alto, Calif.
October, 1981.

6.

Proceedings: "Earthquake Resistance of Highway Bridges", ATC, Palo Alto,

Calif., Jan. 29/31, 1979.

7.

Lin, I. C., "Equivalent Seismic Design of Curved Box Girder Bridges",

M. S. Thesis, Civil Eng. Dept., University of Maryland, College Park,
Md., April 1981.

8.

Robinson, R. R. et aI, "Structural Analysis and Retrofitting of Existing

Highway Bridges Subjected to Strong Motion Seismic Loading", FHWA, May
1975.

9.

Robinson, R. R., Longinow, A., Chu, K. H., "Seismic Retrofit Measures

for Highway Bridges", Vol. 1, FHWA TS-79-2l6, Washington, D. C., April
1979.

10. Bathe, K, Wilson, E, Peterson, F., "SAP IV , A Structural Analysis Program

for Static and Dynamic Response of Linear Systems", C. E. Dept Univ.
of Calif Berkeley, Calif., June 1973.
11. Heins, C. P.. "Bending and Torsional Design in Structural Members",
Lexington Books, D. C. Heath Co., Mass., 1975.
12. Heins, C. P . "Box Girder Design, State of the Art".
Vol. 15, No.4, 1978.

AISC Engr. Journal.

13. Heins, C. P., Sahin, M. A., "Natural Frequency of Curved Box Girder
Bridges", ASCE Structural Journal, Vol. 105, No. ST12, Dec., 1979.

375

ANALYSIS AND DESIGN OF BRIDGES FOR EARTHQUAKE EFFECTS

Polat Giilkan
Faculty of Engineering, Hacettepe University, Ankara

ABSTRACT
The emergence of procedures related to increasingly more
precise earthquake structural analysis has necessitated the
implementation of similar procedures regarding the definition of
the seismic input. It is evident that bridges, which form a
significant groLlp within lifeline structures, should also be given
progressively greater attention during design and construction if
earthquake induced loads are among the primary load conditions.
Following a review of damage to bridges during past earthquakes, ani outline of methods on dynamic structural analysis is
presented. A survey of code requirements in effect in several
countries is presented next to emphasize the manner in which
theoretical considerations are reflected in regulatory documents.
This is followed by a thorough discussion of procedures for
estimating earthquake ground motions. The report is concluded with
the details of the earthquake analysis of a long multispan railway
bridge planned to cross over a large darn reservoir.

376

1.

INTRODUCTION

with the emergence of earthquake engineering as an independent

branch ot science from what was initially regarded as an esoteric
area within structural engineering, the seismic design of bridges
has also progressed from its infancy to a point where considerably
more insight goes into the design of components than the mere
application of an "equivalent" force at some convenient location.
Much of this has been due to the impetus provided by the 1971 San
Fernando event near Los Angeles which caused ~6.5 million damage in
the highway overcrossing near the epicenter. By comparison, the
damage to bridges caused by all previous e~rthquakes in California
had been minimal, and this had inevitably led to the comforting
belief that the exacting requirements of superimposed gravity loads
somehow led to structures with an adequate margin of safety against
the effects produced by strong ground motion.
With particular reference to highway bridges, examination of
damage patterns and reconstruction of the behavior scenarios
suggested that the then standard requirements for earthquake design
of bridges was inadequate and needed to be revised in conformity
with the state of t~e knowledge (1). Specific conclusions derived
from the San Fernando event were that in the epicentral region
earthquake force levels greatly exceeded those specified by the
design criteria, that vertical acceleration played a major role in
some of the collapses, that the customary deficiencies in detailing
connections may have led to some of the spectacular failures.
The establishment of vibratory design criteria for bridges is
not a straightforward exercise, although code requirements frequently
seem to be precisely that when viewed against the backdrop of the
vast array of other standardized provisions. Particularly for
"lifeline" class bridge structures considerably more thinking should
ideally go into the process whereby earthquake effects are
established during the design stage.
Advances in engineering sebIIPlogy have made it possible to
state what the general features of the earthquake ground motion at
a particular site will contain. Establishing the seismic design
parameters for a site requires consideration of a large body of
geologic, geophysical and geotechnical information. Consideration
of the site characteristics is frequently completed by information
that characterizes the geologic area where possible future activity
occurs, the earthquake source and wave propagation path and the
local ground response.
It is not the purpose of this paper to advocate the implementation of detailed and sophisticated investigations for the purpose
of determining the design earthquake criteria for every bridge
structure. The majority of bridges can in fact be designed and

377
detailed according to code requirements, and this process will
result in perfectly acceptable structures. For the more important
cases, however, the procedures described in the following should
be considered, and code provisions should only be used for checking
the results of the more detailed dynamic analyses or for "calibrating"
them.
The rest of this paper is devoted to several considerations.
First, observations based on the past seismic behavior of bridges
and damage patterns will be described. This will be followed by a
discussion of the code requirements in effect in several countries
regarding bridge design for seismic effects. Considerable attention
will be paid to the methodology for developing vibratory ground
motion criteria at a specific site in a format suitable for use in
analysis. In the last section, an example will be provided for a
railway bridge currently under study and sample results will be
presented.
2.

BRIDGE BEHAVIOR DURING STRONG GROUND MOTIONS

By far the most extensive experience with the obberved behavior

of bridges during earthquakes has been accumulated in Japan because
of the frequency of seismic activity in that country and the
relatively longer period of bridge construction. To somewhat lesser
degrees, damage to bridge superstructures observed during the Alaska
earthquake of 1964, the Chilean earthquake of 1971 and the San
Fernando earthquake of 1971 have enhanced the designer's understanding of typical deficiencies and modes of behavior. Data from
European earthquakes, while not nonexistent, has been on a much
smaller scale. A li~erature survey describing seismic effects on
highway bridges has been compiled (2), and the following points
have been cited as general conclusions :
1. Seismic damages, particularly to low bridges, are most
commonly caused by foundation failures resulting from
excessive ground defocmation and loss of stability and
bearing capacity of foundation soils. As a result, substructures often tilt, settle, slide, or even overturn
leading to severe cracking or complete failure. These large
displacements may cause relative shifting and damage the
superstructure, and may include failures within the bearing
supports.
2. Backfills exert large forces on abutments whicn can be in
phase witn the dynamic inertia forces developed in the
superstructures. These forces in combination may cause
severe failures, often of a brittle nature, in the substructures. It is common for wing walls to break loose from
their abutments due to excessive backfill forces. Settlements

378
of backfills resulting from compaction is often observed.
3. Seismic damages due to vibration effects are less common
than damages due to other effects; however they may occur
hi taller bridge structures which lack sufficient strength.
stiffness, and strength. A commonly encountered form of
failure is the insufficient length of bearing supports,
hinges and other devices for preventing spans from falling
off.
4. To minimize damages, bridge structures should be designed
with proper recognition of the stability and bearing
capacities of foundation soils, force-deformation and energy
absorption characteristics of substructure, superstructure
and linking elements, the dynamic nature of structural
response, and the dynamic characteri~tics of all forces
acting on the complete soil-structure system.
3.

DYNAMIC ANALYSIS

In the dynamic response analysis of structural systems we seek

the solution of the following set of equilibrium equations :
IMI . {X}

+ I ci

{~}

IKI

. {x} = {R(t)}

( 1)

Here {R(t)} can be a vec;::or of arbitrary loads with prescribed

variations in time or effective loads which result from ground
motion, and {x} is the vector of relative displacements of the
structure with respect to the ground with primes denoting time
differentiation. IMI, ICI, and IKI are the standard expressions for
mass, damping and stiffness matrices. In the case of ground motion,
the right side of the equilibrium equations may be expreRsed as
{R(t)} = -

I MI. {e} x (t)

g

(2)

Here, {e} is a vector of influence coefficients relating those

degrees of freedom that are directly affected by th~ motion of the
ground; if a unit ground displacement causes a unit displacement in
the n'th degree of freedom of the structure, the corresponding
entry into the n'th row of {e} is a one, otherwise it is zero.
xg(t) is the acceleration vs. time function; in ~he majorit~ of
cases only one COl:1pOnent of the ground accelerat1.on record 1.S
employed, and the negative sign in Eq. (2) is ignored because of
the arbitrary definition of the ground motion direction.
For the solution of Eq. (1) time domain or frequency domain
approaches may be adopted. The time domain ~~alysis can be carried
out using mode superposition or step-by-step direct integration.
In the frequency domain or response spectrum analysis it is first

379
necessary to solve the corresponding eigenproblem.
3.1

Response History Analysis by Mode Superposition

In the mode superposition analysis, it is assumed that the

structural response can be described adequately by a linear superposition of the p lowest modes where p is typically a fraction of
the total number of degrees of freedom. Using the transformation
(3)

where II are the lowest p eigenvectors commonly orthonormalized

with respect to the mass matrix IMI , Eq. (1) may be written as
..

2
{y} + diagI2w.I;:I.{y} + diaglwl{y} = II {R(t)}
(4)
1.

1.

1.

In Eq. (4) it is assumed that the damping matrix lei also satisfies
the modal orthogonality condition
{cp.}T lei {cp.} =0
1.

i '" j
i

(5 )

With this transformation Eq. (4) represents p uncoupled second

order ordinary differential equations which may be solved by using
any convenient step-by-step integration scheme (3). The same integration time step is used in the solution all components of Eq. (4) in
order to simplify the calculation ,of stress components at specified
times which are obtained from the nodal displacements.
3.2

Response History Analysis by Direct Integration

In direct integration the equations in (1) are integrated

using a numerical step-by-step procedure, the term "direct" meaning
that prior to the numerical integration, no transformation of the
equations into a different form is performed (3). In essence, direct
numerical integration is based on two ideas. First, rather than
satisfying Eq. (1) at any arbitrary time t, it is aimed at satisfying
it at discrete time points ~t apart. This implies that pseudo-static
equilibrium, which includes inertia and damping effects, is sought
only at discrete points within the interval of solution. Therefore,
all solution techniques applicable to static analysis find broad
application here. The second idea upon which direct integration is
based is that implicit assumptions are made regarding the variation
of displacements, velocities and accelerations within the time step
~t. It is the form of these assumptions that dictate the accuracy,
stability and cost of the solution procedure. Package computer
programs contain efficient built-in step-by-step direct integration

380

algorithms (4).
3.3

Response Spectrum Analysis

In this type of analysis the ground acceleration vector in

Eq. (2) is written as
x(t)=x (t)+x (t)+x (t)
g
gx
gy
gz

,x

where x_

and

(6)

are the ground accelerations in the three

princip~f di~~ctions~zThe equation for the response in the r'th

normal mode is then

+ 21; r wr Yr +

yr
where y
r

rx

W r yr

= r rx + r ry + r rz

(7)

is the r'th element in {y} and

_{cjl }T

r ry

_{cjl } T

r rz

-{cjl } T

r
r

IMI
IMI
i MI

gx

x
x

( 8)

gy
gz

Using the definition of spectral displacement, the maximum absolute

modal displacements of the system subjected to, say, an acceleration
in the x direction are

h(max)} = {cp}
rx
r

{cp}T
r

IMI

{I}
x

Sd (w ,I; )
x r r

(9)

where Sdx(wr,l;r) is the spectral displacement in the x-dinection

corresponding to the frequency wr and the fraction of damping I;r
and {I~ is a null vector except that those rows corresponding to
the x direction translational degrees of freedom are equal to one.
The concept of the response spectrum and its consideration for a
given seismic environment will be given further attention.
Similarly for the responses due to ground acceleration
y- and z-directions
h(max)}
ry
h(max)}
rz

{cjl }

{cjl } T

{cjl }

{cjl } T

r
r

IMI
IMI

~n

the

{I }
y

Sdy (w r ,I; r )

(9a)

{I }
z

S dz (w r ,I; r )

(9b)

The total maximum response in the r'th mode is then assumed to be

{xmax} = hmax} + {xmax} + {xmax}
rx
ry
rz
r

(Ill)

The total response in terms of displacement and stress resultants

may be calculated as the square root of the sum of the squares of

381

the modal maximum responses expressed by Eq. (10) provided the

modes are well separated in the frequency space so that no strong
coupling exists between them (5).
4.

SURVEY OF CODE REQUIREMENTS

The primary function of code requirements in relation to

earthquake analysis is to express the dynamic effects of the ground
motion as a set of "equivalent" static forces which will produce a
"similar" influence on the structure. Given the almost unlimited
variations of the physical parameters that govern the behavior of
any structure, it is evident that gross approximations are inherent
in such approaches.
As a whole, current codes for the earthquake analysis of bridge
structures depend wholly or mostly on the equivalent lateral force
method. As the name implies, this IT~thod replaces the dynamic
forces exerted by the movement of the ground on a structure by
constant horizontal static forces. The substitute force P can be
obtained from
P

aW

a.gM

(11)

where W or M stand for the weight or mass of a structure or a

structural component and a or a.g represent the seismic coefficient
or substitute acceleration, respectively.
In most cases a can be expressed as the product of a series
subfactors
a = a o . VI

(12)

In this equation,
the basic coefficient
factor related to regional seismicity (i.e. zone factor)
factor related to the importance of the structure, which
ideally contains its vulnerability
factor that considers the vibratory characteristics of
the structure, most typically its fundamental period
factor related to structural type (ductility)
soil factor
For superstructures with uniform distribution of mass over the

382

height of the structure most norms portray the distribution of the

substitute forces with consideration of the fundamental mode of
vibration. For bridges with significant concentration of mass along
the pier height this will not be the case.
In the following, short descriptions will be provided for
codes or specifications in effect in the following countries :
Switzerland (6), United States (7), Japan (8), New Zealand (9),
Attention will also be paid to the California Seismic Design
Criteria (Ill).
Switzerland
The basic seismic design coefficient is taken as
b

Cl.g -

3{'T

with

T)- O.ss

(13)

In this equation b o = g/20 for "higher" seismic hazard areas which

are defined as having an annual probability of experiencing a ground
motion with intensity equal to or larger than I = ~ (MSK) and
b o = g/sO for all other areas. T is the computed fundamental period
of the bridge pier, and a seismic hazard map is given together with
the specifications.
United States
Here the lateral force is specified by
CI.

where
F

F C

(14)

framing coefficient, equal to l.ll for single piers

combined response factor which is graphically specified
as being dependent on fundamental period T, on underlying
soil properties and on the earthquake zone (for Zone
I, A = 0.09g, Zone II, A = 0.22g and Zone III, A = O.sOg).

New Zealand
Similar to the U.S. requirements the basic lateral force is
specified by
CI.

= C F

(15)

where, for the direction being considered,

C
basic seismic coefficient. Value of C for fully ductile
structures shown in Fig.I. Composite values CF for "non-

383

(.)

0.2

Zone

.!!

;;:::

Zone B

(.)

...

0.1

Zone

en
u

"iii
0

0.4

0.8

1.2

1.6

Fundamental Period of Structure T (s)

Figure 1.

Variation of the Basic Seismic Coefficient C in the

New Zealand Code

384

ductile" structures are given in Table 1.

importance factor, as in Table 2.

Table 1. Values of CF for "Non-Ductile" Structures,

NZMWD Design Brief
Category
1

2
3

Zone A

Zone B

Zone C

0.24

0.18
0.15
0.13

0.12
0.10
0.09

0.20

0.17

Note : Categories are defined in Table 2

Table 2. Importance Factor, NZMWD Design Brief
Description

Category

Bridges carrying more than 2500

vehicles per day and all
bridges under or over
motorways or railways

Minimum Value
of F

1.0

Bridges carrying between 250

and 2500 vehicles per day

0.85

Bridges carrying less than 250

vehicles per day

0.7

Choice of category in Table 2 should be based on the

average number of vehicles per day current at the
time of design.

Note

California
These criteria specify the following value for the lateral
force coefficient

ARS

= -Z-

(16)

In Eq. (16)
A

peak rock acceleration, determined from seismological

studies of fault activity and attenuation data gathered
from past events

= acceleration spectra in rock, based on actual earthquake

385

data recorded on rock

S

soil amp 1 ification factor, based

and actual recorded data

ductility/risk reduction factor based on observed damage

and assumed ductility.

01

analytical studies

The product A.R.S gives an elastic response spectrum curve

for the site that would result from a maximum credible event on the
closest fault. Division by th2 factor Z then gives a design force
for the portion of the structure under consideration. The factor Z
is component oriented, thus the design force depends not only on
seismicity and site conditions but on the actual structural
component being designed.
5.

PROCEDURES FOR ESTIMATING EARTHQUAKE GROUND MOTIONS

Determination of the seismic design parameters for a particular

site requires consideration of a large body of geologic, geophysical,
seismological and geotechnical information. Basically, seven steps
are followed in specifying the characteristics of the ground mot~on
required for earthquake resistant design (Fig.2).
Determination of the seismicity of the geographic location
where the application is planned,
Identification of the seismotectonic features,
Estimation of the regional seismic attenuation,
Estimation of the vibratory ground motion parameters (for
example peak acceleration or intensity) at the site,
Definition of the ground motion response spectra for the
si te,
Determination of the local amplification effects and
modification of the design response spectra for the site as
necessary, and
Estimation of the uncertainty in the ground motion design
values.

5.1

Determination of Seismicity

The objective of this step is to use the past seismicity record

to define earthquake recurrence relations applicable to the region
of interest, to provide a basis for correlating earthquake epicenters
and tectonic structure, and to define seismic source zones. From the
seismicity record, one attempts to establish such basic information
as :

386

Steps

ESTIMATE SEISMICITY
ON BASIS OF
EXISTING
REGIONAL DATA

IDENTIFY SEIS .. OTECTONIC FEATURES OF AREA

ESTIMATE SEIS .. IC ATTENUATION FOR AREA

ESTIMATE .. AXIMUM INTENSITY

OF SHAKING EXPECTED FOR SITE

USER DEFINES
ACCEPTABLE
LEVEL OF RISK
USER PROVI DES
DAMAGE OR DESIGN
CRITERIA

No

Yes
ESTI .. ATE LOCAL SOIL AMPLIFICATION EFFECTS

MODIFY RESPONSE SPECTRA FOR SITE

(IF NECESSARY)
ESTI .. ATE UNCERTAINTY IN
SEISMIC DESIGN PARA .. ETERS

No
USER PROVIDES
ANALYSIS
CRITERIA

USER DOCUMENTS ANALYSES.

ALTERNATIVES AND RECO .... ENDATIONS

Figure 2.

Steps in Estimating Ground Motion for Earthquake

Resistant Design

387

Date of occurrence
Epicentral intensity, and where possible, magnitude
Epicenter and hypocenter locations
Epicentral maps showing the position of all reported earthquakes within an 80 km radius of the site, and, ideally,
all earthquakes with intensity larger than V (or magnitude
larger than about 4.2)
Correlation of epicenter location with tectonic features
Frequency of occurrence for earthquakes in terms of
magnitudes or intensities.
This task is often undertaken within the format of an earthquake catalog compilation. Catalogs of felt or recorded earthquakes
do not in general cover a long enough time interval to permit
meaningful extrapolations into the future except on a board regional
scale (13).
Earthquake recurrence law is of the inverse logarithmic form
log N(M)

a - bM

( 17)

where N(M) is the number of earthquakes occurring within a region

during a given time period with ma,~itude equal to or larger than
M, and a and b are regression constants. The seismic activity index
b is commonly in the range 0.8 to 1.0 for most parts of the world.
5.2

Identification of the Seismotoctonic Features

The location and cbaracteristics of faults in the region

surrounding the site are essential information along with historical
seismicity for estimating the upper bound magnitude and locations
of potential earthquakes. A fault is a fracture or fracture zone
along which displacement of the two sides relative to one another
has occurred. A capable fault is defined as one that may undergo
future surface displacement. A fault is classified as capable if
(i) it has had Quaternary or more recent movement, or (ii) macroseismic activity has been associated with it, or (iii) it has a
demonstrated\structural relation to a known capable fault such
that movement of one may trigger a movement of the other during
the lifetime of the project under consideration. An analysis of
faulting and tectonic activity in a region can be complex, requiring
data from: (i) geologic data on faults and the ages of rock
formation they displace, (ii) seismological data on the areal
distribution of seismicity in relation to known faults and tectonic
structures, ,and (iii) historical accounts which provide evidence
that a fault has ruptured at the surface.

388
One rationale for classifying fault activity uses the available
geologic, seismological and historical data to specify the fault
activity descriptions. Faults are considered to be significant if
they satis fy one of the following conditions : (i) faults cross ing
the site vicinity are capable of fracturing during the lifetime of
the proposed structure, (ii) faults near the site are recognized as
being capable of generating large damaging earthquakes. Current
criteria indicate that a fault should be considered capable of
permanent surface displacement if movement occurred at least once
during the last 35 000 years or more than once during the last
500 000 years. Empirical relationships relating fault length and
earthquake magnitude have been derived from observations
5.3

Determination of Regional Attenuation

An important consideration in specifying the nature of the

ground motion at a given site is related to the knOWledge of how
ground motion is attenuated with distance away from causative fault.
As seismic waves propagate though the rock formations in the crust
anisotropy and inhomogeneity of the earth, geometric damping,
absorption, scattering, reflection, refraction and wave interference
play complex roles in modifying and reshaping them. Strong motion
records are the best guides in deriving attenuation relations
because statistical information can be cited together with such
information but an adequate number of strong motion records are not
available for most regions. In most cases isoseismal maps from past
earthquakes are currently the best basis for deriving an intensity
attenuation relation. An isuseismal map contains the maximum ground
motion intensity reported at various locations. These contours may
overestimate the ground motion level at a site of interest because
they are usually tainted with non-quantified local soil amplification
effects. Data from a number )f earthquakes which occurred in Turkey
has been anlyzed to give the following relation for intensity
attenuation
I(R)

= 7.56 + 1.07M - 2.04

~n(R+

30)

(18)

In Eq. (18) R is the epicentral distance (with no azimuthal bias)

and M is the magnitude. The equations has a standard deviation of
0.7 in intensity units (14).
The preferred procedure for defining the attenuation function
for a region is based on processing recorded strong motion data so
that a family of empirical acceleration-distance cures can be
derived. Such a family of mean curves proposed for "rock" sites
with data obtained from Western United States events is shown in
Fig. 3.

389

0.7
0.6
~

.
...'"...
..'"
0

;::

0.5

-'
u
u

u
0

...

II:
Q

II)

0.2
0.1
5

10

50

100

DISTANCE FROM FAULT, IN KILOMETERS

Figure 3.

Mean Values of Peak Acceleration vs. Distance for

Sites on Rock

390

5.4

Estimation of the Vibratory Ground Motion Characteristics

After determining the location and magnitude of the design

earthqudke the characteristics of the ground shaking at the site
can Le established with the aid of a proper attenuation function.
Current procedures for this step are empirical in nature.
The characteristics of ground motion produced by an earthquake
are dependent on the source 8echanism, epicentral distance and the
geometry and physical properties of the geologic structures though
which the waves pass as they propagate from the source to the site.
Seismograms recorded in the "near field" are part:icularly complex
and site properties are overshadowed by source properties in defining
the characteristics of the ground shaking.
Ground lOCltjon can be specified in differenc ways. The commonly
used formats include the time records, spectra or peak ground motion
parameters. These will be described in the following.
5.4.1 Ground Motion Recorns. A collection of ground motion seismogrruns constitutes a key element in seismic design procedures because
they contain significant information related to the following items
Magnitude
Fault dimensions and configuration
Seismic energy released
Average displacement across the fault
Fault orientation, depth and sense of slip
Rupture velocity
Elastic constants associated with the fault area
Stress drop across the fault
Complexity of energy release.
5.4.2 Peak Ground Motion Parameters. In spite of its limitations
current practice in the design of earthquake resistant structures
employs the peak ground acceleration as a measure of the severity
of the ground motion. This may be true for stiff structures but for
intermediate and low frequency systems peak ground velocity and
displacement are lOClre applicable.
It is commonly assumed that the maximum ground accelerations
in the two horizontal directions are equal and that the maximum
vertical acceleration is two-thirds of this value. In statistical
terms, this assumption is true for about 80 percent of accelerograms
in current use. There is a growing body of information which appears
to contradict the data contained in Fig. 3.

391

Ground motion can also be characterized by velocity and

displacement histories and their peak values. The peak values seem
to have more stable upper', bounds than acceleration.
5.4.3 Duration. The duration of strong ground shaking has been
recognized as being of paramount importance regarding damage
potential. There are numerous examples of accelerograms with peak
accelerations equal to O.5g or higher with no associated structural
damage. The 1940 Imperial Valley earthquake has been regarded as a
model of a destructive earthquake primarily because of its long
duration. The dominant parameters in the destructiveness of a given
ground shaking have been shown to be (1) the amplitude, (2) the
duration, and (3) dominant frequencies at which the shaking occurs
(16). General trends regarding duration are that as the overall
characterization of the recording site ranges from "hard" to
"soft" the average duration will increase, that larger magnitude
events are of longer duration and that increasing epicentral
distances are accompanied by increasing durations.
5.4.4 Earthquake Spectra. In engineering seismology Fourier
techniques are the most commonly used form for the analysis of
seismograms. A Fourier spectrum decomposes the ground motion record
into its harmonic components in tne frequency domain, and is
uniquely related to the original seismogram in terms of amplitude
and phase relationships.
For structural applications the response spectrum is preferred
because structural systems respond to ground motion within fairly
well defined frequency limits. The motion of simple viscously
damped elastic systems to ground excitations is governed by Eq.
(1). The response spectrum technique is a method for determining
the maximum amplitudes of response for structural systems from
the maximum responses of an ensemble of simple oscillators (which
are essentially narrow band filters) when they are excited by a
given ground motion. There are various response spectra such as
pseudo absolute acceleration (PSAA) , pseudo relative velocity,
absolute acceleration (AA), relative velocity (RV) and relative
displacement (RD). The following relationships exist between
relative displacement and the pseudo quantities :
PSRV
PSAA

w. (RD)

w (PSRV)

(19)
=

w2 (RD)

(20)

5.4.5 Intensity. It has been observed that intensity is a more

stable quantity to consider in'attenuation relationships because
factors which affect this parameter are smoothed and averaged for
a given region. The current accepted procedure in earthquake
resistant design is to establish the expected ground motion
intensity and then to convert this information to maximum ground

392

acceleration so that the design response spectrum can be scaled.

Western United States earthquake data (17) yield the following
empirical relations between Modified Mercalli intensity IMM and
peak horizontal and vertical ground motion parameters.
log

-0.014 + 0.30 IMM

(ern! s )

(2la)

log Av

-0.18 + 0.30 IMM

(cm/s )

(2lb)

log Vh

-0.63 + 0.25 IMM

(cm/s)

(2lc)

log V
v

-1.10 + 0.28 IMM

(cm/s)

(2ld)

log Dh

-0.53 + 0.19 IMM

( ern)

(2le)

log D
v

-1.13+0.24 IMM

(ern)

(2lf)

In these equations A, V and D re~resent the peak acceleration,

velocity and displacements, respectively while the subscripts h
and v denote the horizontal or vertical directions. These average
trends and the spread in the data expressed as l standard deviation are shown in Fig. 4.
5.4.6 Probabilistic Estimates of Effective Ground Motion Parameters.
With the help of probability theory coupled to information on
regional geology, seismic history and attenuation functions one can
make estimates of ground motion parameters for given return periods
(13). Probabilistic estimates are not without their own particular
pitfalls, however, because our knowledge of the seismic past, which
is essentially extrapolated hundreds or even thousands of years
into the future, is frequently too short and incomplete.
5.5. Definition of Design Response Spectra
After the characteristics of ground shaking at a site have
been established design response spectra can be constructed.
Procedures for developing vibratory ground motion parameters in
terms of the design spectra may be considered within two categories:
site independent and site dependent. Both approaches were developed
primarily as a response to stringent nuclear plant siting
requirernents.
There is considerable controversy regarding the definition
of seismic input and site response. This is caused by the debate
about whether available geologic, geophysical, seismological and
geotechnical data are adequate to specify the seismic input and
site response in any precise manner. Controversy also centers on
whether judgments about conservatism in the seismic designs are

-'

II:

......
...~

>=

...

!E.. ~

UII:

II:

II)

10

100

1000

...
t;;o
...~ .,

I!:'

II:

'"

!d

I
I

I
I
I

"~I

.....

-I

Figure 4.

i
U
...>g

...

>=
z

:II

....ffi...

I!:'
.,

II:

8~

It

MERCALLI

INTENSITY

I II III IV V VI VII VIII IX X XI XII

MODI FlED

I~

10

100

--.--r--T
o Vertical component
Horizontal component

ffi

.,...-'

:II

~
,.:

...z
...
...

...

t;;
:!
I-

.,

10

100

~~

I
I

I II III IV V VI VII VIII IX X XI XII

It

I
I

r~t

Correlation of the Modified Mercalli Intensity Scale

with Peak Ground Motion Parameters

I II III IV V VI VII VIII IX X XI XII

Il I

\C

....
....

394

justifiable in view of uncertainties in the data. Information for

evaluating the ground response in the site vicinity is obtained
from a combination of borings, trenches, laboratory measurements
and geophysical investigations (18). The purpose is to (1) determine
the classification, lateral distribution, layering, geologic
structure and physical properties of materials underlying the site,
(2) obtain samples and cores for testing, (3) establish the elevation of underground water level.
Vibratory response of geologic materials is governed by the
following parameters: shear wave velocity, density, shear modulus,
damping, Poisson's ratio, bulk modulus, and shear strength. These
physical properties are based on data from laboratory and field
tests.
5.5.1 Site-Independent Spectra. Response spectra which do not
explicitly take into account the characteristics of the construction site fall into this group. These spectra have well defined
standard shapes, and because they are based on accelerograms
recorded on a wide range of geologic and seismologic conditions, are
considered representative of general conditions when local site
properties are not expected to playa significant role in shaping
the response.
The earliest attempt at obtaining site independent spectra was
by Housner (19). His spectrum shapes were simply the average of
two components each from four large earthquakes available at that
time. Later, armed with a larger collection of ground motion records
Newmark and Hall (2U) proposed a stamdard procedure for drawing
design response spectra. The technique is based on the observation
that the spectral ordinates are proportional to the peak values of
ground motion over well defined frequency ranges. Construction of
the response spectrum is accomplished by multiplying the maximum
ground acceleration, velocity and displacement values by amplification factors dependent on damping, In the absence of data to the
contrary, the maximum values of ground motion parameters may be
assumed to be proportional to each other, and for a 1 g peak
acceleration, peak velocity and displacement values are to be taken
as 122 cm/s and 91 em, respectively. In statistical terms, the
amplification factors shown in Table 3 are mean plus one standard
deviation numbers in a lognormal distribution, so that they may be
exceeded in only 10 percent of the time.
Independent studies performed by different groups were
consolidated into a single document (21) as the official position
of the U.S. Atomic Energy Commision regarding response spectra for
the design of nuclear plants in the United States. To use the R.G.
1.6U spectrum shapes maximum horizontal ground acceleration and
displacement must first established. The maximum displacement is
assumed to be proportional to the maximum acceleration regardless
of the soil conditions and is set as 91 cm for a 1 g acceleration.

395
Table 3. Aplification Factors in the Newmark-Hall
Procedure (20)
Amp 1 ification Factor

~ercent

lmping

0
1
2
5
10
20

Acceleration
(PSAA)

Velocity
(PSRV)

Displacement
(RD)

6.4
5.2
4.3
2.6
1.5
1.2

4.0
3.2
2.8
1.9
1.3
1.1

2.5
2.0
2.0
1.8
1.2
1.0

The response spectra are established by a series of straight lines

but the control points have definite frequencies for all horizontal
spectra. Amplification factors are a function of the percentage
of critical damping, as listed in Table 4. A horizontal response
spectrum drawn for a 1 g maximum acceleration earthquake is shown
in Fig. 5. For the vertical direction design response spectra are
drawn in much the same way as for the horizontal direction.
Table 4. Amplification Factors in R.C. 1.60 for the Horizontal
Direction
Percentage
of Critical
Damping
0.5
2.0
5.0
7.0
10.0

Amplification Factors for Control Points

Acceleration

Displacement

A(33 Hz)

B(9 Hz)

C(2.5 Hz)

1.0
1.0
1.0
1.0
1.0

4.96
3.54
2.61
2.27
1.90

5.95
4.25
3.13
2.72
2.28

D(u.25 Hz)
3.20
2.50
2.05
1.88
1. 70

Note that the factors given for the construction of the vertical
spectra are to be applied to the horizontal components of
acceleration and displacement.
5.5.2 Site Dependent Spectra. The data base used for determining
the site independent spectrum shapes can be reexamined with regard

396

PERIOD, IN SECONDS

5.0

1.0

0.5

0.2

0.02

0.1

0.01

...
0

&oJ

'"

...'"
&oJ

'"
II::
&oJ

....&oJ
~

....
Z

...
&oJ

;!!;

>-

...!::
0

...J
&oJ

.
>

...J

.........
II::
....

'"

10

0.3L-____L -____

0.1

0.2

~~

0.5

___ J_ _ _ _~_ _ _ _ _ _~~_ _~_ _ _ _~_ _ _ _ _ _~~_ _~

20

FREQUENCY, IN HERTZ

Figure 5.

Horizontal Design Spectrum Scaled to 1 g

50

100

397

to representative samples of a variety of site conditions. In some

cases it is possible to select a family of response spectra that
were derived from seismograms for which soil conditions were similar
to those of the proposed site. The basic idea in the site matched
response spectrum concept is to categorize the geological conditions
into one of the broad classes for which a standard shape definition
is available.
A comprehensive study of site dependent spectra was performed
by Seed et al. (22). Ensemble average and mean plus standard deviation spectra were developed from lu4 records classified as "softmedium", "deep cohensionless", "stiff", and "rock". Comparison of
the ensemble average and mean plus one standard deviation spectra
shown in Fig. 6 indicate trends that can be attributed to local
soil conditions. In the low frequency range, spectra from soft to
medium soil sites and deep cohensionless soil sites have larger
amplitudes than the spectra from rock and stiff soil sites. For
frequencies above about 6 Hz spectra from stiff or rock sites exhibit
amplitudes larger than. the soft to medium or the deep cohensionless
soil sites. Both trends are clearly visible in Fig. 6 in which the
R.G. 1.60 spectra are also presented for comparison purposes.
5.6

Determination of Local Ground Response

From the earliest times of scientific observation in relation

to seismic events it has been noted that otherwise similar structures
situated on different soil formations tend to experience different
degrees of damage. For example, structures founded on unconsolidated
soils are frequently more severely damaged by ground shaking, and
damage distribution has been positively related to site characteristics. Buildings of a certain class or type with a certain natural
period are often damaged from ground shaking when they are located
on geologic formations with similar periods. Mismatch of structure
and soil periods has the beneficial effect of attenuating the effect'
of shaking.
Local ground response is complicated because earth materials
behave nonlinearly, even at low strain levels. In the infinitesimal
strain ranges below the order about 10-5 the deformations exhibited
by most soils are nearly elastic and recoverable. Over the intermediate ranges of strain from 10- 4 to 10-2 , the behavior of soils
is elasto-plastic and produces irrecoverable permanent deformation.
When even larger strains are imposed, failure generally occurs (23).
Site amplification modeling is especially required for strong
ground motions. Comprehensive parametric curves cannot be developed
because of the increasing number of parameters that need to be
considered and gaps of knowledge pertaining to natural materials.
Some general curves that have been published permit estimates to be

398
PERIOD. IN SECONDS

10

100

0.1

0.01

o..p cohesion Ie...011.

Stiff .011 depo,",

....
..........

Rock

0
<.>

II:

a::

..

...
;::
...z
t-

0.5

<.>

'!
>-

100

!:

<.>
0

...>
-'

...-'

II:

.......o
t-

10

EXPLANATION
E._blo mIG. pl

0.1

'lGftdord dl.latla. :
Saft-mldlum .oil.
o..p collllla.ll...on.
Stiff .oil

dip.,,,,

Rock
Revul.tory Guidi I 60

0.5~____~~~~______~~-L

0.01

____~__~~________~~

'0.1

10
FREQUENCY. IN HERTZ

Figure 6.

Site Dependent Mean and Mean Plus One Standard

Deviation Response Spectra

100

399

made.
When generalized parametric curves are judged to be
insufficiently precise, it might be desirable to utilize finite
element or similar procedures (24,25). The degree of accuracy with
which the physical characteristics of the earth materials at a site
are known should be evaluated critically before embarking upon a
costly site response analysis investigation.
6.

THE RAILWAY BRIDGE OVER THE KARAKAYA DAM RESERVOIR

STUDY

A CASE

With the impoundment of water behind the Karakaya dam which

is currently under construction, a part of the railway linking the
cities of Malatya and Elaz1g in Southeastern Turkey will be inundated. Economic studies comparing the various possible alternatives
have shown that moving the bridge to point downstream from the dam
site or beyond the upper reaches of the lake were not feasible so
that a new route with an approximately 2.1 km long crossing over
the reservoir has been planned. In the absence of an accepted seismic
design code for bridges, it was decided to develop project specific
criteria for the bridge. In the following these will be outlined
in a format similar to that described in Section 5, and some sample
results based on a response spectrum approach as described in Section
j will be presented.
6.1

Description of the Structure

Cost constructions dictated the form and type of the bridge.

The guiding principle agreed upon at the outset was that it should
be possible to erect the structure within a minimum time and
complete all the structural work prior to the completion of the dam
proper itself. To minimize foreign currency requirements for
construction it was also considered appropriate to impose the
constraint that it should be possible to produce the necessary parts
and materials domestically. Factors regarding easy installation,
maintainability and previous experience with long span railway
bridges led to the adoption of a 29-span simple truss type bridge
supported on a succession of piers which vary in height from
approximately 3U m to 70 m. Figure 7 illustrates the overall view
of the planned structure and in Fig. 8 a typical reinforced concrete
pier elevation and cross section are shown. Each pier supports one
fixed and one roller support, and each span is 70 m long. It is
obvious that under normal operational conditions water will seep
into the closed center of the box section in Fig. 8, so portholes
will be left in the walls of the piers to facilitate this action.
The selected truss type is conventional, and represents what
the supplier declared could be built with existing technology rather

Figure 7.

The Railway Bridge at the Karakaya Reservoir

Artist's Conception of the Overall View

An

401

EL. 711.75

.--_----,-!L

.:r

11

-,1
II
I'
II

Ir-

I:
II

II

II
II

'I

693.00

II

=11

II
II
II

II

II

II
II

0
0

CO>

0
0

....

320

75 165

800
( b)

Section

Dimensions in cm
I

II

I 642.00

"""~"'~
(a)

Figure 8.

Elevation

Typical pier Elevation and Cross Section

402
than the end product of a broad optLmLzation effort aimed at
reducing the number of foundation excavations in difficult conditions and a more economical use of materials.
6.2

Determination of the Seismic Input

Current thinking in relation to meeting seismic requirements

for the construction of critical facilities is based on the twolevel approach inspired by the approach to nuclear design. The
~ower intensity earthquake motion is defined as corresponding to
the event which may reasonably be expected to occur once during
the lifetime of the facility. The more intense second level
corresponds to the strongest earthquake ground motion which could
possibly be recorded at the same site. While the two-level
philosophy has found widespread acceptance among engineers the
exact quantification of these events has been rife with controversy.
The former event controls the elastic limit of behavior while the
latter governs strength requirements. Studies aimed at establishing
these levels were conducted in several steps.
6.2.1 Regional Seismotectonics and Regionalization. Seismotectonic
features of Anatolia (Asia Minor) may be considered within five
major groups: (1) North Anatolian fault zone, (2) Aegean - MarmaraTnrace region, (3) Central Anatolia, (4) Eastern Anatolia fault
structures, and (5) Western Tauroids.
The left-lateral strike slip Eastern Anatolia fault extends
in the northeast-southwest direction, and can be considered within
two main parts. The first is" approximately l2u km long and follows
the Murat river valley while the second forms the 65 km long GOynlik
Valley. These two main extensions are separated by a 15 km long
discontinuity section, and have been formed as a result of interplate compressive forces causing reverse faulting in southwestern
Turkey. This fault zone extends southward as far as the Dead Sea
rift valley. The major seismotectonic features of the area are shown
in Fig. 9.
The effort aimed at establishing characteristics of the ground
motion at the site is based on an expression of the general seismotectonic features in the form of a regionalization. To this end,
seismic events of the last 15 years within a 30u km radius of the
site were reviewed and their relationship to the faulting was
investigated. This study resulted in the definition of the major
sources indicated in Fig. 9.
6.2.2 Earthquake Hazard Estimation. In the two widely used approaches
towards establishing some index of the ground motion such as peak
acceleration or intensity a deterministic or a probabilistic
methodology may be implemented. Calibration of the final results is

403
36'

37'

39'

38'

40'

4"

4"

40'
~.

"ERlURUM

---..............

Foult
Probable foult

Normal fault

~ Overthrust
_
Strike slip fault

-=:::;-

Figure 9.

Seismotectonic Features of Eastern Turkey and

Source Regionalization

404

done by ensuring that both will yield similar results. In hazard

estimates the most important relationship which is utilized is the
attenuation expressionlwhich indicates the variation in the ground
motion index as a function of distance. Given the paucity of
recorded strong ground motion, and the more stable variation of
intensity over distance, it was ju~ged more appropriate to employ
Eq. (18) and to calibrate the design response spectrum through the
peak ground acceleration derived from Eq. (21).
For the bridge under study a maximum ground acceleration
value of U.2U g was calculated for the upper limit event, and this
value was corroborated by both probabilistic and deterministic
procedures. For the lower limit event v.13 g was computed. Consideration of the general tectonic features led to the decision that
either component of the Taft record of 1952 scaled with respect
to the appropriate acceleration could be employed in amplification
studies for the local geotechnical structure as well as the bridge
piers and their interaction with the fluid and foundation media.
6.2.3 Assessment of Local Amplification Effects. An extensive series
of borings and geophysical investigations were made along the bridge
axis. The results of these were carefully evaluated in defining four
major soil profiles for which amplification studies were performed.
The ultimate aim of soil amplification studies is to determine the
ground motion at some control point (typically the free surface)
given the input motion, typically defined at the bedrock level or
to determine the response spectrum at the free surface given the
response spectrum for the motion at the bedrock level and physical
properties of the overlying layers of soil.
A truly nonlinear model was used in ascertaLnLng the modifications in the input response spectrum to the four major soil profiles.
Compared to equivalent linearization techniques, this type of
modeling results in spectra in which one can have a greater degree
of confidence, especially at higher frequencies. All analyses were
performed in the time domain and the N21E component of the 1952
earthquake recorded at Taft was used as the input motion without any
modification.
Viewed in relation to the customary differences in spectrum
ordinates for a given ground motion which are smoothed for design
purposes, the conclusion was reached that for two of the rock or
stiff soil profiles the input motion reached the surface without
substantial modifications, whereas for the other profiles through
cohensionless layers the spectral acceleration ordinates were
amplified by 5U percent. On this basis two design spectra were
obtained. These are illustrated in Fig. 10 : the smooth curve
denoted as "rock" corresponds to the stiff profiles, and the curve
labeled as "deep alluvium" refers to the latter profiles. For

405

Seed - Ugas - Lysmer

Deep cohesionless soils)

100

Seed - Ugas - Lysmer

~ (Hard

.. .....................

soils)

............

-.
...
......

,,"

.~"'z.:-..:::::...
.-~~ . . .".

"""

,',"

',.~.....

Rock

,.~

10

'".\>
.
',' ...

~\'.

\"

.
u
0

>

1.0

0.1

10

1.0
Frequency in

Figure 10.

Hz

Design Response Spectra

406

comparison purposes two design spectra for "hard soils" and "deep
cohensionless" soils reported in (23) scaled to the appropriate
maximum acceleration are shown in Fig. 10. It is noted that the
general dynamic features of the soil columns are reflected in
these curves.
6.3

Analysis of Dynamic Response

For long lifeline structures such as continuous bridges it may

be desirable to investigate the effect of finite wave propagation
velocity on dynamic response. The relative motion between successive
piers may be accentuated when out-of-phase ground displacements are
input from the foundations. The same effect may be observed for two
adjacent piers with greatly dissimilar heights. For tue bridge under
consideration this will not be critical in the longitudinal direction because of the discontinuous manner in WhiCh the trusses are
connected to the piers. In the transverse direction tue vibratory
response will be primarily controlled by horizontally polarized
shear waves and Love waves. However, there is negligible inertial
coupling between the piers in this direction because the weight of
the truss structure is only a fraction of the weight of a given
pier-typically of the order of 10 percent or less. This judgment
made it possible to calculate the dynamic response of each pier
in either direction on an individual basis because it is not
dependent upon the behavior of adjoining piers.
The direct use in design of the spectra shown in Fig. 10 in
conjunction with the response spectrum method would lead to
unacceptably large design forces because no modifications have
been made on them to account for the effects of cracking and other
nonlinear actions. Most design codes based on the equivalent lateral
force concept contain requirements which recognize the reduction in
force levels resulting from cracking and yield. Calculations aimed
at the determination of tne design forces on a succession of piers
with heights varying from 30 m to 70 m were performed for both
principal directions and for two different stages of construction
corresponding to empty and full reservoir.
The principal special problem which arises in the evaluation
of system property matrices for the pier-superstructure system
results from the inertial effects of water in the reservoir
surrounding the piers. As piers respond to earthquake ground motion,
they cause corresponding motions in the water in contact with the
pier surfaces, and the resulting changes of water pressure acting
on the piers in turn affect their dynamic response. Tnis phenomenon
is called fluid-structure interaction.
In the case of gravity dams this hydrodynamic effect is
by assuming that the pressure changes at the face of the

co~sidered

407
darn result from the action of a certain volume of water attached
to the surface (26). For
flexible structures interaction effects
are more complex but they are again approximated by an extension
of the added mass concept.
Earthquake induced vibrations of submerged structures can be
studied satisfactorily by assuming that wave action is negligible
and the velocity of the structure relative to the surrounding water
is low. Under these conditions the analysis can be performed by
adding to the mass of the structure the mass of
a certain volume of water which gives a new virtual mass but otherwise treating the structure as if it stood in vacuo. Classical
normal modes do not exist for the piers when hydrodynamic effects
are included, but their displacements in terms of their modes of
vibration in vacuo may still be used to approximate the actual
displacements.
Analysis of the response of structure interacting with fluids
is best carried out in the frequency domain. The response to unit
harmonic ground acceleration .
x.. ()
t
= Re {e iwt}

(22 )

may be expressed by

Y (t)
n

Re{Y (w)e

iwt

(23)

where Yn(w) represents the complex frequency response of the n'th

generalized coordinate. It can then be shown that the governing
equations can be expressed as
Sew)

yew) = L(w)

(24)

In Eq. (24) the frequency dependent coefficient matrix Sew) relates

the generalized displacement vector yew) to the corresponding
generalized loads L(w) which include the effects of hydrodynamic
pressure. Unlike the classical mode-superposition analysis outlined
in Section j.l, the matrix Sew) is not diagonal because the mode
shapes J./ are not the normal modes of the pier-water system.
~

~n

Under a set of reasonable assumptions the motion of the water

the reservoir is governed by the wave equation
2
2
~+ d p
dX 2

-:;;z

+ -322P
3z

32 P

= 2-2

(25 )

dt

in which p(x, y, z, t) is the hydrodynamic pressure and c is the

velocity of sound in water.

408

Evaluation of Eq. (24) and (25) with the appropriate boundary

conditions and for a range of values of w to permit synthesizing
earthquake motions into Fourier components requires considerable
computational effort. Therefore, the alternative of solving the
equations of motion for a combined pier-reservoir system using
finite elements in which the water is represented by plane strain
quadrilateral elements with appropriately modified elasticity
matrices was adopted. In later stages the results obtained from
this approach were checked against a more refined model in which
the piers were modeled as axisymmetric shells, surrounded by eight
noded bricks of water, In both types of analyses the body of water
inside the piers was taken simply as a non-interacting mass, and
the mode superposition technique was employed. The results obtained
from these studies corroborated one another : in general it was
observed that .the inclusion of hydrodynamic effects tended to
increase the elastic forces by about 40 percent as compared with
the case when these are not considered.
The effect of soil structure interaction was also investigated
for the piers which were reduced to single degree systems vibrating
in their wirst generalized modes, and tne pile cap foundations
were modeled as rigid rectangular masses with one translational
and one rocking degree of freedom. The elastic constants of the
springs connecting the foundation to the surrounding soil were
based on a range of values representative of soil type and geophysical investigation results. The simplified model (27) showed
that soil-structure interaction was insignificant. The slight
reduction in base shear and overturning moment was of the order of
5 percent, and the top displacements increased by a maximum of 15
percent.
6.4

Derivation of Equivalent Lateral Forces

The computation outlined in the previous section were performed

in both principal directions for the range of piers considered.
Also included in the studies were the spectra shown in Fig. 10
and the response spectrum of the original Taft record. The average
of these computations for spectra adjusted to 0.2 g zero period
spectral acceleration is indicated by the jagged lines shown in
Fig. 11 drawn for the two principal directions. The ordinate
denotes the ratio of the maximum shear force at the ground level
to the total weight of the pier in vacuo, In these analyses rather
than explicitly considering the water surrounding the piers, the
mass density of the submerged part of the pier was modified in
order to reduce computation time.
Tall flexible and statically determinate structures such as
the piers considered for the Karakaya Bridge are sufficiently
long period from the dynamic point of view, and the modification

409

0.5

direction

I -

'I - direction

0.4

:!!:::

u
.,
..,
.,

0.3

CIt

0.2

CD

0.1

o~----~------~----~------~----~
0 3 5
Period in I

Figure 11.

Equivalent Static Forces

410

or Lilt:! elastic response spectrum can be done by simply dividing

the elastic ordinates by the postulated ductility factor (28). The
design response spectra shown as smooth curves in Fig. 11 are the
attenuated replicas of the computed elastic forces. It is to be
noted that these curves contain the hydrodynamic effects, therefore they were applied to the piers as if they stood in air. The
lateral force curves corresponding to the case when these effects
are not considered would be obtained by further dividing the
ordinates by a factor 'representative of the hydrodynamic influences.
6.5

Sample Results

The design elastic forces for which the cross sections shown
in Fig. 8 were designed were derived from the spectra shown in
Fig. 11. It is of interest to note that good agreement was obtained
between the "enhanced" design spectrum applied to the structure
assumed to be standing in air and the reduced spectrum excluding
hydrodynamic forces applied to the piers with appropriately modified
mass distribution derived from the added mass concept to account
for the submerged portions. Although eight of the lowest modest were
considered the fundamental mode was by far the most dominant, and
results based upon it would have been quite satisfactory.
A sample of typical results is given in Fig. 12 for a 70 m
tall pier in terms of displacements, snear forces and b,ending moments
in the direction transverse to the bridge axis. Reinforcement
requirements were basically determined on the basis of working
stress procedures although the ultimate strength theory was used
in checking the strength against critical modes of failure. Contrary
to most design codes the computations were performed for the case
when half of the design train load was effective at t,le same time
as the design earthquake. This explains the finite value of ,the
shear force at the top of the pier in Fig. 12; it stems from the
mass of the truss and the train lumped at the top of the pier.
Because of these additively conservative assumptions and because
of the rather strong ground motion considered for design substantial
overturning moments were obtained for which the pile foundations
need to be designed.
7.

CONCLUSIONS

Lifeline structures such as the long multispan railway bridge

described in Section 6 require careful analysis regarding the
seismic design aspects. Whether as a consequence of structural
complexity, unusual ground conditions or high cost or risk potential
similar bridges will merit special studies. It is judged appropriate
that dynamic analyses to obtain maximum earthquake forces and
displacements or ductility demand should be carried out for bridges

411

o
(a) Olsplacement In em

( b) Shear Force in t

30000
(c) Overturning Moment in tom

Figure 12.

Computed Results for a Typical Tall Pier

412

all into one or more of the following categories

Major bridge structures considered to be of particular
importance
Mass of any pier, including any allowance for hydrodynamic
effects greater than 20 percent of the mass of the superstructure estimated to contribute to the inertia on the pier
Bridges with superstructure joints that are designed to
allow significant relative movement between the separated
spans
Structures in which the lateral load resistance is provided
by structural systems other than conventional piers and
abutment structures
Suspension, cable stayed or long span arch bridges.
Establishing the design criteria for a bridge structure and
verification of their adequacy requires the mobilization of a
significant range of activities within earthquake engineering.
8.

REFERENCES

1.

Jennings, P.C., (Editor), "Engineering Features of the San

Fernando Earthquake of February 9, 1971 ," California Institute
of Technology Report No. EERL 7l-G2, June, 1971, Pasadena,
California.

2.

Iwasaki, T., J. Penzien, and R. W. Clough, "Literature SurveySeismic Effects on Highway Bridges," Earthquake Engineering
Research Center Report No. EERC 72-11, University of California,
Be rkeley, Novembe r, 1972.

3.

Bathe, K.J., and E.L. Wilson, Numerical Methods

Element Analysis, Prentice Hall;' 1976.

4.

Bathe, K.J., E.L. Wilson, and F.E. Peterson, "SAP: A

Structural Analysis Profram for Static and Dynamic Response of
Linear Systems ," Earthquake Engineering Research Center Report
No. 73-11, University of California, Berkeley, June, 1973,
revised April, 1974.

5.

Wilson, E.L., A. der Kiureghian, and E.P. Bayo, "A Replacement for
the SRSS Method in Se ismic Analysis ," International Journal on
Earthquake Engineering and Structural Dynamics, Vol.9, pp. 187194, 19tH.

6.

"Norm fur die Belastungsannahmen und die Uberwachung der Buaten ,"
SIA 162, Art. 22, Schweizerischen Ingenieur und Architekten
Verein, Zurich, 1970.

~n

Finite

413

7.

American Association of State Highway and Transportation

Officials, "Standard Specifications for Highway Bridges,"
Twelftn idition, Wasl1ington, 1977.

d.

Bridge and Structural Committee, Japan Society of Civil Engineers,

"Earthquake Resistant Design of Bridges," in Earthquake Resistant
Design for Civil Engineering Structures, Earth Structures and
Foundation in Japan, Tokyo, 1980.

9.

!~ew

Zealand Ministry of Works and Development, "Highway Bridge

Design Brief ," !~ZMWD Publication CDP 70l/D, Wellington, New
Zealand, 1978.

10.

Gates, James H., "Factors Considered in the Development of the

California Seismic Design Criteria for Bridges ," Workshop
Proceedings on the Research Needs of Seismic Problems Related
to Bridges, Applied Technology Council, Palo Alto, California,
1979.

11.

Hays, Wal ter H., "Procedures for Estimating Earthquake Ground

Motions," U.S. Geological Survey Professional Paper No. 1114,
Government Printing Office, Washington, D.C., 1980.

12.

Werner, S.D., "Procedures for Developing Vibratory Ground Motion

Criteria at !'luclear Plant Sites," l~uclear Engineering and Design,
Vol. 36, pp. 4llJ- i f41, 1976.

13.

Glilkan, P., and M.S. Ylicemen, "Seismic Risk Analysis for Nuclear
Power Plants ," Journal of Pure and Applied Sciences, Middle East
Technical University, Ankara, April, 1977.

14.

Glilkan, P., M. oner, M. Erdik, and M.S. Ylicemen, "A Methodology

for the Assessment of Ground Motion Parameters for Nuclear Power
Plants," Proceedings of the Seventh European Conference on
Earthquake Engineering, Athens, September, 1982.

15.

Schnabel, P.B., and H.B. Seed, "Accelerations in Rock for Earthquakes in the Western United States ," Bulletin of the
Seismological Society of America, Vol. 62, pp. 501-516, 1973.

16.

Trifunac, M.D., and A.G. Brady, "A Study on the Duration of

Strong Motion ," Bulletin of the Seismological Society of America,
Vol. 6~, pp. )81-626, 1975.

17.

Trifunac, M.D., and A.G. Brady, "On the Correlation of Seismic

Intensity Scales with the Peaks of Recorded Ground Motion,"
Bulletin of the Seismological Society of America, Vol. 65,
pp. 139-162, 1975.

414

18.

Waldron, H.H., R.P. Miller, and S. D. Werner, "Geotechnical

Investigations at Nuclear Power Plant Sites ," liluclear Engineering
and Design, Vol. 30, pp. 397-406,1976.

19.

Housner, G.W., "Behavior of Structures During Earthquakes,"

Journal of the Engineering Mechanics Division, ASCE, Vol. 85,
pp. 109-129,1959.

20.

Newmark, N.M., and W.J. Hall, "Seismic Design Criteria for

Nuclear Reactor Fac il it ies, 11 Proceedings of the Fourth World
Conference on ~arthquake Engineering, Santiago, Chile, Vol. B,
pp. 37-50, 1969.

21.

U.S. Atomic Energy Commission, "Design Response Spectra for

Seismic Design of Nuclear Power Plants (Revision)," Regulatory
Guide 1.60, 1973.

22.

Seed, R.n., C. Ugas, and J. Lysmer, "Site Dependent Spectra for

Earthquake Resistant Design," Bulletin of the Seismological
Society of America, Vol. 66, pp. 221-244, 1~76.

23.

Ishihara, K., "Strength of Cohesive Soils under Transient and

Cyclic Loading Conditions," in State-of-the-Art in Earthquake
Engineering 1981, edited by O. Erglinay and M. Erdik, Turkish
clational Committee on Earthquake Engineering, AnKara, 1981.

24.

Idriss, I.M., J. Lysmer, R. Hwang, and H.B. Seed, "QUAD-4 :

A Computer Program for Evaluating the Seismic Response of Soil
Structures by Variable Damping Finite Element Procedures,"
Earthquake Engineering Research Center Report No. EERC 73-lb,
University of California, Berkeley, 1973.

25.

Joyner, W.B., and A.T.F. Chen, "Calculation of lilonlinear Ground

Response in Earthquakes ," Bulletin of the Seismological Society
of America, Vol. 65, pp. 1315-1336,1975.

26.

Clough, R.W., and A.t... Chopra, "Earthquake Response Analysis

of Concrete Dams," in Structural and Geotechnical Mechanics,
edited by W.J. Hall, Prentice Hall, 1977.

27.

Veletsos, A.S., "Dynamic of Structure-Foundation Systems," in

Structural and Geotechnical Mechanics, edited by W.J. Hall,
Prentice Hall, 1977.

2d.

Chopra, A.K., and N .M. Newmark, "AnaLysis ," in Design of Earthquake Resistant Structures, edited by E. Rosenblueth, Wiley,
198v.

415

WIND LOADS ON BRIDGE STRUCTURES

E. ZELLER
Associate Professor
Department of Civil Engineering - Institut Universitaire
de Technologie - Universite STRASBOURG-III (FRANCE)

Introduction
In the history of construction. men have always had to
deal with the effects of wind. Any time a structure or construction
was unable to resist it, the wind always made a ruin of it. Each
time, man was given a lesson by the confrontation of his structural
concepts with reality.
Thus for ages, the knowledge of wind effects has
remained inseparable from pratical experience. A structure was
considered stable if it could resist ~ind ; thus, in many languages,
transversal stability was expressed in a similar fashion ...
"wi nd-braci ng", "contreventement", "controvento" ...
Certainly from the end of the last century, in dealing
with certain bold projects (the Eiffel Tower, skyscrapers, the
first long span bridges) research had to be done. However there
remained a world of difference between knowledge of structural
wind effects and the otherwise advanced science of aerodynamics
(aviation).
And so, in 1940, when the Tacoma bridge collapsed. other
similar bridges had already met the same fate. But in the case of
the Tacoma bridge~ this occurred during daylight hours. The
resonance effect was observable for a long time, and was even
recorded on film, as is well known. This served as a starting
point in the scientific investigation of the aeroelastic phenomenon.

416

Nowadays, in constructions or structures whose structural characteristics (dimensions, sections, mass and stiffness
distribution, damping ... ) stay in a certain defined range, national wind codes, although sometimes presenting differences, bring
ample security to classic constructions.
In the setting of the European Convention for Constructional Steelwork, a great amount of work has been achieved by
some twenty experts from ten nations This group has just
brought together the main strong points of various wind codes,
the results of wind-tunnel work and in situ measurement, such as
constructive experiments. Because the author has had the privilege of participating in the Technical Committee on Wind Effects
since 1968, the analytical approach in the beginning of his
report, chapter I, reflects the results of this work-group.
Nowadays, aeroelasticity has made notable strides, but
in direct consequence of the bridge sections being made of
sharp edges, the position and intensity of shedding vortices have
not been revealed by a faithful mathematical model.
That is why in great bridges with low frequencies, wind
tunnel study on complete or partial models remains the tool number
one for guaranteeing the required serviceability and security in
design.
When the construction of a bridge has been decided, the
theoretical analysis in relation to the parameters (length,
structure type, load, material ... ) leads inevitably to a solution
in which the two following aspects are sufficient to determine the
behaviour in the wind :
Point 1 - mechanical characteristics of the bridge:
rigidity, frequency, damping ...
Point 2 - a geometrical section determining the wind
flow conditions
When the corresponding aerolastic model leads us to a
structure whose response remains inferior to allowed limit values
in all cases of possible excitation, behaviour will be acceptable.
In the opposite case, when excitation conditions (different wind directions, speed, wind turbulence) produce flexural or
torsional instability of the bridge, the solution must be
modified by changing the frequencies or the section, until the
modified model yields a satisfying aerodynamic response.
The author will show the influence that the two above
aspects can have on actual bridges in chapter II of this report.

417

During the construction of a cable stayed bridge "Pont des

Meules" Mr. J. WIANECKY completed a wind tunnel study on a partial
aeroelastic model in the Laboratories of CEBTP at St Remy les
Chevreuse near Paris. The tests render all wind conditions and made
it possible, by independent adjustment of vertical and torsional
frequencies and with different damping values to foresee the
bridge response and to avoid critical velocity and flutter.
The systematic exploration of all parameters with an interesting
measuring technique also recorded as a film, represents for the
bridge designer a pedagoglcal tool of the highest order.
On the other hand, it may sometimes act to improve the
dynamical behaviour of an already existing bridge. Mr. R.L.
Wardlaw had to solve this problem on the ong's Creek cable
stayed bridge (chapter III). This time with mechanical characteristics already imposed, the section for the existing bridge
had to be modified to reduce oscillations. The investigation of
several sections in the wind tunnel made possible a spectacular
improvement in bridge's behaviour by changing its profile. The
amplitude of the vibrations could be reduced by a factor of more
than three to one.
The author wishes to record his gratitude to
- D. Sfintesco - Chairman Technical Committee T 12 on Wind
Effects, European Convention for Constructional Steelwork,
E.C.C.S.
- H. Van Koten - Director of Dynamic of Structures and
Aerodynamic Department, Institute TNO for Building Materials
and Building Structure, Delft, The Netherlands
- R.L. Wardlaw - Deputy Head, Low Speed Aerodynamics Laboratory
National Research Council of Canada, Ottawa.

- J. Wianecky - Chief En~ineer, Division of Aerodynamique du

Batiment et des Ouvrages d'art, CEBTP, Paris.
for sharing some of their practical knowledge gained in their
long and distinguished research careers.

418

CONTENTS
Introduction
I.

Calculation of wind effects on bridges

1.

2.
3.
4.
I I.

Wi nd tunne 1 approach
1.

2.
3.
III.

Dynamic Simulation on sectional model

Energy evolution/cycle
Damping effect of amplitude, example

Improvements of dynamic behaviour of existing bridges

1.
2.
3.

IV.

Wind characteristics, spectrum

Design wind speed
Wind effect on bridges
Aerodynamic instability

Aerodynamic shape modification

Additional damping device
Example: bridge section and dynamic response

Security for wind bridge design

1.
2.
3.
4.

Design wind speed and structural safety

Dampi ng, criti ca 1 value
Aerodynamic stability of bridge under erection
Security criteria of bridge design

419

I. Calculation of wind effects on bridges

1. Wind characteristics, spectrum
- nature of the wind : air movement to balance the atmospheric
pressure differences. The wind velocity, as well as the
wind direction depend on great number of factors: latitude,
Coriolis acceleration, relief, roughness of the ground, etc.
The variation in time can only be approximated in statistical
terms. For a given point in space corresponding direction and
velocity of wind variations are snO\m in fig. 1.
- statistical distribution: in most countries, statistical data
have now been gathered -- in some instances, for almost 100
years. Generally, only the maximum of the wind velocity, obtained
each day, week, or month, is recorded. One of the greatest difficulties for an accurate definition of wind velocity for the
design derives from the turbulent character of air flow.
- The fluctuation of the wind velocity can be considered as a
random process. It can be written in a Fourier series.
Wind velocity is a function of the frequency
fluctuations as follows:
V

= l:(n) sin (211 nt

f..[:

where
a(n)

6(n) = Arc tg

co, (2. nt) dt] 2 +

[1/:
Jf
o

cos (211 nt) dt]

n of its

+ 6(n) ) dn

[1;

'in (2. nt) dt ] 2

phase angle

V sin (211 nt) dt

The variance of the wind velocity over a period T long enough

in or~er to have
~ V dt)2 negligible with respect to

(+

+Jo V2 dt can be wri tten :

J. .
JoO
1<>0
a~ = +i v2 dt = a 2(n) dn = Sv(n) dn = nSv(n) d (loge n)
2

Sv(n) is the spectrum or the power spectral density of the wind

velocity at the frequency n

420

~+
$

:z
0

IU
t..I

a:
Q

\I}

....

01

II

~
~

-t.

~~

""d
S

~o

>

...

"

>

s:!

421

The analysis of the spectrum of the wind velocity shows

that the wind energy presents some maximum values :
- the first peak at a period of 4 days corresponds to the
transit time of fully developed weather systems (depressions),
- the second peak at a 1/2 day period corresponds to diurnal
heating and cooling,
- the third peak at a period of 1 minute is due to the atmospheric
turbulence.
The wind spectrum measured by DAVENPORT and VAN DEN HOVEN
is nowadays unanimously accepted.
Mean wind
Fluctuations
4 days
4
n Sv (n) 3

(m/s)2

2
1

l/n
Fig. 2
Wind spectrum measured by van den Hoven
In the aerodynamicsof structures the velocity fluctuation
of turbulent wind need to choose a base of time T of the order
of minutes and then the expression of wind speed becomes :
V = Vm + Vf
where

Vm is the mean wind speed defined over a period of 10 minutes

.!fv

Vm = T

Vf

dt

is the fluctuation of the wind speed about its mean value

Vm due to the turbulence of the atmosphere.

422

2. Design wind speed

at high Z during t years Vm, t , z measured over a
long time (~ars) show that the distribution of the extreme
values of the mean wind speed can be approximated in terms of a
low of "FISCHER-TIPET type I", statistical distribution:
V
m

- the mean speed at the height z occuring on an average once

during t years that is of return period t years:
V

m,t,z

U-

.!a 1og e [- 1og e

(1 -

.!)
t ]

The parameters U and a define the statistical

distribution of the speed V , and depend on the considered
country.
m
By considering the probability P that the mean wind
speed occurs at the height Z during N years, this can also
be written :
1
m,t,z = U - -a log e

rL loge

(1 - P)

liN]

For the calculation of the mean speed Vm, t ,z of return

period t one starts from the mean speed Vm,50,z of return
period t = 50 years and applies the relationship
Vm,t,z

kt Vm,50,z with kt

t
t

200
100

YEARS 50

20
10

5
2

0.7

/
0.8

0.9

Fig. 3

/
,/

1.1

1.2

423

Wind direction:
As a general rule. the design speed Vmt z is
determined taking no account of direction. But in all cases the
wind is quasi-parallel to the ground surface.
Turbulence intensity :
The statistical analysis of the anemometer records shows
that the standard deviation 0v of the speed fluctuation is
f

proportional to the mean velocity Vm.t.z

= Vm, t ,z

f
-V--

m.t,z

is the intensity of the turbulence and is a function

of terrain roughness and of the height above ground. but independent of the long-term variations of wind speed.
The spectrum of the speed fluctuation can also be
defined as a function of the mean speed Vm t 10 ' of the
vari ance o2V and of the frequency n by the fo 11 owi ng empirical formula f
n

Sv

(n)

where

S (n)
vf
x

= power

spectral density of wind at

frequency n

1800 n
m.t.10

Vmt 10 as defined hereafter.

424

Influence with height

The mean wind speed V t z increases with height
m
according to some empirical laws up to the height of gradient zG
where this increase shows a point of inflexion. The wind speed
at this point is called the gradient speed VG ; it is independent
on the terrain roughness.
The gradient height zG defines the thickness of the
boundary layer;it varies from 250 to 600 m according to th~
terrain roughness.
Below gradient height. the mean wind speed can be
defined by following relationship
V

iI,t,z = k log ~
e Zo
m,t,10

where

VII

basic speed. i.e. mean speed of return period t

m.t.10 10
m above the ground in zone II

k, z00 b and

at

parameters depending on the terrain roughness.

values of which are given in the table below.

Bridges are mostly constructed near the ground surface

but there are exceptions in whichthe height over the level 0 can
be important.
Z.

z.o .. 2.5
~=

Fig. 4

425

Remark :
Vm,t,z does not take into account some phenomena such as
tornadoes, local wind, local topographic effect (escarpments,
hills, valleys).
The vertical gradient of wind velocity depends on the
terrain roughness of the five conventional roughness classes,
defined as:
sea or lake shore with more than 5 km of fetch
over the water
II flat terrain with some isolated obstacles (trees,
buildings, ... )
III rural areas with low buildings, trees, hedges
IV urban, industrial or forest areas
V

centres of large cities

Roughness class
logartthmlc

law

II

III

IV

0,005

0,05

0,3

2,5

0,16

0,19

0,23

0,26

0,29

{'o

(m)

The turbulence intensity I is a function of height

of the air above the ground and of roughness. I is given by
the following relation
I

= _-..:..:K_ _
k log ~
e

in which the parameter K

zo

K ;s dependent on the terrain roughness class and has the

fo 11 owi ng v.a 1ues :

II

III

IV

0,16

0,19

0,22

0,23

0,23

426

Spatial correlation of the speed fluctuation

The determination of the wind pressure acting on a
bridge can not only be based on the mean wind speed. In fact the
wind speed Vf at a given moment varies from One point to
another in a vertical as well as in a horizontal direction.

(1)

Vf

r--~
~

;JY'
.l

(2)

Fig. 5
This variation can be described by a correlation
function p expressing the simu_ltaneity of occurrence -of gusts
at two points (1 and 2 ) :

It is given by the following empirical relationships

in the vertical direction

(app 1i ca ti on to bri dge towers)

427

=e

in the horizontal direction

(application to bridge spans)
being the scale parameters of the correlation and
being the distances which separate the two particular
points.

The parameters
relationship

Lz and Ly are defined by the empirical

(z in m)

42 ( z ) 0,25
20

The passage from the wind V speed to the dynamic

pressure q is linked by the relation
1
2
q

"2

with q expressed 3in N/m2 , V in m/s , pair density(at 15C and

1 bar)= 1,225 kg/m .
This permits the equivalence of the statistical wind speed with
the statistical wind pressure w = wm + wf with
mean pressure over 10 minutes
pressure fluctuation about the mean value w

3. Wind effect on bridges

A bridge is a flexible structure and the dynamic action
must be taken into account. The developed pressure on a bridge
shows fluctuations which depend:
- on the incident wind characteristics which affect the free
stream incident flow (turbulence, gust).
- on the air flow around the bridge: turbulent shear layer
between the air flow and the value, vortices, turbulence ...
The resultant pressure fluctuation produces oscillations and vibrations of the bridge - The mechanical response
of the bridge shows vibrational modes in bending and torsion:

428

the resultant movement can modify the action (lift, drag,

torsion) of the air flow and the bridge can so be dynamically
exited.
Lift

WIND

Fig. 6
When a slender structure oscillates in bending perpendicular to the wind, a periodic variation of pressure can be
produced and an amplification of vertical oscillation can lead
to aerodynamic instability if critical wind speed is reached.
-+

V wlnd
~~'d
V relative vorl ge

Fi g. 7
The torsion of the bridge section modifies the angle of
incidence; the resultant pressure variation can produce
instability in torsion.

WIND

Fig. 8

429

In addition, when the natural torsional frequency approaches

natural bending frequency, the coupling of both movements can
produce instability.
The theoretical amplitude of the. flutter is not
easily obtained because different nondimensional aerodynamic
coefficients have to be evaluated experimentally over the
corresponding reduced frequency range, and the damping value
which can only be estimated, has a great importance.
Scruton proposed a well-known formula for obtaining the
bending amplitude at the resonance, based on partial model study
in wind tunnel
2

1T

oK

CL ~ S
Co

is non stationary lift action in with CL is

the experimental lift coefficient
damping and K vertical rigidity.

By contrast, it will be easier to determine the critical

wind speed of the different instability types in order to avoid
them.
4. Aerodynamic instability

Schematically let us consider the interaction wind

flow-bridge response as follows: Fig. 9.

II

stable interaction
Fig.

aeroelastic instability
9

430

The wind flow around the bridge section produces the

aerodynamic action on the bridge deck. The different possibilities Ai (position, rotation ... ) are located on curve I. Curve II
represents the resultant mechanical response B. of the bridge
(vertical or horizontal bending, torsion) cor~sponding to the
above mentioned action. From an initial equilibrium point 0, a small
perturbation can displace the bridge slightly to point A .
The corresponding modified wind flow reciprocally induces a 1
new mechanical response B which leads the system at position A '.
The feedback system invol~ed can follow two extreme possibilitias :
1 the energy input in the bridge deck at each cycle is di~sipated
quickly enough so that the amplitude will be limited or ~iminished:
the equil ibrium is stable ;
2 the energy input can not be dissip4ted by the internal damping
the dynamic amplitude with the corresponding stress in the
material grows. and the aerodynamic instability can lead the
bridge to destruction.
Instabilities can arise with cross-sections that are
aerodynamically unstable. Hovewer the general features of
vibration do not basically differ in turbulent wind or in smooth
wind ; the transition from turbulence generated buffeting to
fluttering can be continuous but in any case the instability
occurs only if a critical wind speed is reached. The determination
of the critical wind velocity is of highest importance for the
bridge designer.

431

Oscillation in pure first mode bending:

This must be taken into consideration for a rectangular or
polygonal structure if

<:

Vd eSlgn

c
for a constant cross section

V = 1,63 0
c

with

n1

(J

n1~

x b

first mode bending frequency

m = mass per unit 1ength (kg/m)

b

width of the structure (m) perpendicular to the

wind direction
coef. of aerodynamic stability

(J

Example
(J

=11-

2,5

_--I

2n ~ ~ 2 n ~ if ~ small
logarithmic decrement = ~
(~

% of critical damping)

the damping
. structural damping

Os

depends on material, type of construction, connection,

effect of the ::>restressed concrete elements, bearings ...
mostly
aerodynamical damping

0.01

<

Os

<

0.06

0a

is generally negligible except for cable or very slender

elements or bridge decks

432

Qw = total force produced by the design wind

m

= oscillating mass (kg)

logarithmic decrement 0d due to possible additional damping

devices.
- oscillation in pure torsion:
the critical wind speed for pure torsion can be
approximated with the same calculation as Vc for coupled
oscillations bending -torsion.
- oscillation in bending torsion
the critical wind speed Vc is given by the formula
V 3.14. "1
C

fl ex

rL 1

d 1 .1:1 tors .
1 fl ex

o.s'~ t.s Vmi]

d3

the value is function of the section and is given

in the figure n 10

alon~

wind dimension of the section

m mass per unit length 1

I

torsional stiffness factor per unit length

n is a empirical coefficient equal to the ratio between

the critical speed for the bridge section under
consideration and the critical speed for a flat
plate
V .....
(n =

v'-)
c

r-__________~~d;~~~;-------------Ir---------433
Bri dge secti on

--cP-

-d

---i1r-----------~d
--~I 0.1-0.2

Lr-f

'I

0.2

0.1-0.2

0.2

_N

'20.5 d

- D-

0.3
0.3
0.3

0.15-0.Sd

I--

- c:::J

0.05
0.1-0.15

0.15 - 0.5d

...

-'

<.

0.13

0.11

.,

t-

-I

"1'e

-0) - {3d

......

:--

Fi g. 10

0.05
0.5

0.1

0.06-0.13
0.7

0.1-0.2
0.12

e:
I

---1

434

Remark 1

aerodynamic instabilities should be avoided. This is

not always possible in the first stage of design.
If Vc~ Vdesign it is necessary to check by a more
detailed calculation of the critical speed; or to
check the behaviour from an aeroelastic model in a
windtunnel.

Remark 2

In case of instability for given bending and torsional

frequencies the damping is the mechanical characteristic
number one which can limit the amplitude.

Remark 3

This formulation of Vc is approximate but for the

designer this formulation -allows him to define simply
structures which are certainly not sensitive to tne
instability phenomenon.

435

II.

Wind tunnel approach:

1. Dynamic simulation on sectional model

Wind tunnel tests can be made on full rigid models which

reproduce entire bridgesin a given geometrical scale. Now,more
and more,sectional models mounted on springs allow the complete
simulation of the dynamic behaviour. The vertical frequency and
the torsi ona 1 frequency can be chosen se pa rate ly and all conditions of wind excitation can be systematically explored in
order to establish that the mechanical response will not exceed
fixed limits of acceptability (amplitude, stress ... ). But the
extrapolation from the wind tunnel response to the actual bridge
is only assured if certain conditions are respected.
V2

- all scali ng parameters must be observed : FroudO number Og

(V = wind speed, d bridge width, g gravity acceleration)
Reynolds number, etc ...
- the characteristics of the actual wind flow which act on the
bridge must be introduced in the wind tunnel : properties of
the wind layer, turbulence intensity, scale of the turbulence
relative to the size of the bridge span, angle of incidence
of the wind (horizontal, vertical)
- the buffeting response must be checked, the energy of
turbulence can be important compared with the mass of the
bridge deck ; the buffeting can affect the stress level and
the number of cycles
- Flutter and vortices: for turbulent as well as for uniform
flow, critical wind speed should be above the expected wind
speed in the future site.
The experimental determination of the aerodynamic forces
and moments must be completed with a full dynamic simulation. The
energy input in the successive cycles increases with the
amplitude of vibration (Fig. 11). As the steady state deflection
is a function of the mechanical damping, a small variation of the
introduced value can modify considerably the stored mechanical
energy and the resultant amplitude ,and, furthermore, the critical
velocity. This is why the wind tunnel studies have to be made
with different values of the damping. The cable stayed bridge
"Pont des Meules" was studied with a partial aeroelastic model
a special electrical device allowed precisecontrol of bending
and torsion damping (Fig. 12).

436

2. Energy evolution/cycle

Fig. Ii

Amplitude of the bridge

:.

- curve I : elastoplastic energy of deformation in the bridge

~: successill'! energy input due :ELwind effect - damping effect)
- curve II : total energy stored in the bridge after a cycle
Al Aj amplitude corresponding to the successive incremental
energy input

Fig. 12
A
B
C
D
E
F

partial aeroelastic model (130 x 65 cm)

end plate of the model
spring for suspension (torsion. translation)
electrical damping device with control
5 tra 11 gage
mechanical device for torsional inertia control

mm fig

@'

I/"

10

30

40

mm

...

AMPLl TU DE

o
10

20

30

nvn

AMPLITUDE

/--_.
Vd

TORSION

= 9,6

= 3,9

-Vd

m:

Al1PLlTUDE

0tors

0nex

--- ----

HER
km/h

FLUTIER

245 kili/h

Fig. 13 - Dampino and Amrl i tude of the aeroelastic rrodel (for steel solution)

20

IJ

Vcr flexi on = 62,4 l"nAh

165 ~m/h

1,8 %

195 kmth

:;..-"'FLUHER

0tars = 3,7 %

'desiqn

10flex

Q(=+~.

--I---~--==~

20J.

30T J

42 ....
40

50

~LJ

Dynamic pressure

= 6,5 %
= 14,1 %

AMPLITUDE

0. ___

0flex

) DESIGN VdSPEED

.j:..

w
.....,

438

3. Damping effect of amplitude, example

Increased values of damping in flexion and torsion reduce for a
given wind speed the response amplitude, and the critical
velocity for flutter increases markedly over the design speed
(Fig. 13).
If the bridge deck behaviour has been tested and if
the dynamic behaviour is acceptable, some elements must also be
checked:
- example: the vibration of the cables or the stays.
Wnen the cable stayed bridge "Pont des Meules" (100 km from
Paris) was under construction, a storm struck in Oct. 1976.
As the wind speed reached 15 m/sec blowing with a incidence of
20-30 with the longitudinal bridge axis, the vibration amplitude of a 100 m cable was between t 30 cm for the first modei
after several hours, the storm abated; with 15 m/sec the
cable reached again the critical resonance for 30 minutes.
A tunnel experiment showed that the 165 mm diameter
cable became a sufficient damping if the surrounding duct was
not injected and for the same critical wind speed the amplitude
remained insignificant.
- the

be~aviour

of the mast must also be checked.

III. Improvements of dynamic behaviour of existing bridges

1. Aerodynamic shape modification
The bridge design needs a good choice of mechanical and
aerodynamic data in order to minimize the risk of dynamic
excitation and instabilities. Howewer in some cases, in spite of
a rational selecting of the technical data some initial conditions
may change; and the aerodynamic stability of a bridge deck can
become unacceptable. Solutions must be found to reduce dynamic
response of the existing bridge.
Aerodynamic shape modification :
A small modification of the section can be the best
solution for improving the aerodynamic stability. For example the
Long's Creek bridge
- characteristics : cable stayed orthotropic girder bridge
main span of 217 m,
structural steel weight 4 270 kg/m

439

- vibration in vertical flexion has been produced after opening

the bridge at a velocity of ~ 40-50 km/hr
frequency in vertical bending Nz = 0.59 HZ
vertical acceleration 0.3 g unacceptable
amplitude 5-10 cm
As the handrail was blocked by snow the amplitude reached an
excessive value : ~ 20 cm.
Mr R.L. Wardlaw had to find a solution to improve the
bridge response without making structural modification. The
dimensionless frequency range was between 0.11 and 0.13
N . h
( z
, h girder depth = 2.5 m), and the expected Strouhal
V
number: 0.13 ; the large amplitude was only produced in the
speed range 40-50 km/hr : excitation by vortex shedding. A
corrective modification of the bridge section had to be found
in wind tunnel study but, for safety reasons"a temporary solution
had to be applied in order to reduce the amplitude.
2. Additional damping device
As
was shown in chapter II (for thE' Pont des t1eul es)
an increased damping value reduced the amplitude. The measured
damping log. decrement by a vertical impulse method reached
0.065, was not sufficient. An interesting and simple additional
damping system has been used. A load consisting of rocks was put
in a box (2 x 2 x 0.5 m) and suspended in the water 30 m below
the bridge. Fig. 14.

0+0
a
s
O2

0a+ 0s+ 0d

a = 0 0 s = 6,5 %

Fi g. 14

440

AMPLITUDE

em

10
1

I
BASIC BRIDGE

2,--1_~

SOFFIT PLATE

Fi g. 15

441

The initial damping 6 1 became increased with the temporary damping device 6d . With 02 the pseudo periodical movement
had now an amplitude ~ 3 cm .
Definitive solution :
The investigation on a model in a wind tunnel gave
clearly the same mechanical behaviour as observed on the bridge.
After exploring different shapes for the bridge as well as the
use of perfored girders (permeability modifies wind flow favorably) the definitive solution consisted in a simple external modification of the bridge deck : adding soffit plating. The effectiveness of different dimensions of fairing is illustrated in
Figure 15.
Importance of small details:

Furthermore, such insignificant details for a bridge designer, such as handrails, can considerably modify the amplitude of
the response when blocked by snow. The sectional model response of
the unmodified Long's Creek bridge allowed to reproduce this clearly.
Fig. 16.
Amplitude
on

18

16

!~
'I'
II:I I,
II

Ilt
12.

, rr--Handrails blocked by snow

I I~

10

J \
:
I
I

Handrail clear

I
I
I

't

f \
I

2-

10

20

30

40

50

Fi g. 16
V

km/hr

I~B2m

19ot m

Fi g. 17

}V

Pasco Kennwick interc ity bridge and differe nt

edge configurations

630 m

,::

CO"f~2

0.52 m

L
.lQ

Edge config uration n 3

Am ~

Edge

Edge config uration n 1

i.:T]

tN

443

3. Example: bridge section and dynamic response

It is useful to note how a small change on the external
bridge section can modify the wind flow and consequently the
vortex shedding and flutter velocities. For example different
edge configurations showed in Fig. 17 have been examined by
Mr R.L. Wardlaw for the Pasco-Kennewick intercity bridge. In
spite of small damping values used in the test 0flex = 1.9 %
and 0tors = 0.8 %. and with the largest incidence angles
(a = 8) and highest wind to be expected in the site. a
satisfactory response could be obtained with configuration nO 3.
Figure 18 gives the critical wind speeds (test without groundboard). with the maximum amplitude in bending and torsion.
and the vortex shedding range. The torsion values in the table
18 are given by the vertical displacement at bridge edge.
I

I CONF.
NUMUR

WIND
ANeLi

VORTEX SKtDDIN'C
VELOCITY, MPH

~IAXIMml

A~IPLITUDE
0

LOIo.'ER

UPPER

LI~lT

LIMIT

TORSION

FLUTTER VELOCITY, MPH

FT
BENDINC

179

171

+4

In

122
70

+6

29.0

43.7

+8

25.2

41.7

0.409
0.101

0.469

100
77

60

157
122

148

+6

26.4

37.3

0.261

+8

24.5

51.1

0.139

-8

26.3

44.3

1. 24
0.438

+4

+6

26.3

45.7

+8

26.3

41.1

106
147
110

+4

-6
1

VELOCITY
DECREASING

+2

-6
-8

VELOCITY
INCREASINC

0.335
0.468

80.2

In

152

III

102

155

153

132

126

99

88

-6

114

125

-8

124

117

Fig. 18

IV. Security for Wind Bridge Design

1. Design wind speed and structural safety

Meteorological wind data : velocity over 10 minutes,

turbulence intensity, return period can be properly used to
determine
the design wind speed. The dynamic response is
distributed around the mean pressure response; the range of the
response data is a function of the turbulence spectrum, the
aerodynamic admittance, the mechanical admittance of the structure
and the joint acceptance. Aeroelastic instabilities of the bridge
section, in vertical bending as in torsional mode, must be checked with all actual possibilities of incidence angle, velocities
and turbulence intensity. The resultant amplitude in all cases
must stay below fixed limits. Fig. 19.
Amplitude

Limit
servic

Wind velocity

V
Fig. 19

11 y t he
For the brldge
des Meules,1/ curve I shows schematlca
solution of flexural and curve II that of torsional movement with
the wind speed. In fact,each response type is a combination of
response I and II. In order to improve structural safety,the
range V flexion V torsion has to be large ; bending and
cr
cr
torsion or coupled flutter have to be situated largely over the
design speed.

445

Without structural modification,the two examples in

chapter III illustrate how with only small changes in external
shape,the dynamical response of a bridge section can be optimized
in order to di mi ni sh the amplitude.
2. Damping, critical value
In the successive dynamic oscillation)the wind energy
input tends to increase the elastical energy, therefore the
amplitude. In bending, the bridge model in wind tunnel (chapter
II) showed that the amplitude decreases with increase of
damping, the intersection with the horizontal axis gives
directly the critical damping value, Fig. 20.
Amplitude
resonance

Bendi~g
I'llIft

incidence
= + 4
vm = 8 m/s

Cl

V = Vm + Vf = 17,3 m/s

2.0

,,

,0

,,

\'\

11

IS

Damping
Fig. 20

The benefit obtained by a high degree of damping is

important for safety : decreasing of the maximum stress level and
reduction of the highstress cycles (fatigue). In mechanical
engineering, dampers are frequently used, but in civil
engineering, only a small number of solutions have been proposed
for bridges.

~6

3. Aerodynamic stability of a

brid~e

under erection

It must also be underlined that wind sensitive bridges

are sometimes more vulnerable during the construction stages,
which must be studied with special attention. The progressive
steps of the bridge under erection must be checked, the
vibration modes, the unfinished shape can contribute to
diminish the critical velocity: The different stages must be
studied with special attention so that method and
technological devices allow a safe erection of the bridge.
4. Security criteria of bridge design:
Wind action on bridges is a stochastic process, changing
with time. Mostly, only the peak values of the response under different wind conditions are considered so that, generally, in the
different wind codes the aerodynamic behaviour of bridges, in spite
of statistical wind data, is, in fact, treated in a deterministic
way.
Each load (dead and live load, wind ... ) which acts on a
bridge can be defined by a density of probability of the load
effect ; the structural resistance can be defined with the
density function so that the probability of failure is obtained
by convolution. At present time, design criteria for bridges,
subjected to different loads avoid the full statistical treatment
simplified design rules, load combination, the so called load
factors and characteristics loads are used.
For important bridges a better approach for security is
now developed and based on a more complex but improved
analytical process :
The probability for a bridge to present a failure under
combined actions of time variable loadings is the probability
for the given period to exceed,at. minimum,one of the different
limit conditions. For a structural system the whole limit conditions together define the safe domain. The design life time in
this involved formulation is nothing more than the duration
which needs the corresponding vector process to make the first
excursion from this domain.

447

REFERENCES
CHAPTER I
E.C.C.S. - Recommendations for the calculation of wind effects
on buildings and structures, 1978.
SCRUTON C., NEWBERRY C. - On the estimation of wind loads for
Buildings and structural design, 1963.
SFINTESCO D., ZELLER E. - Prise en compte des effects du vent
sur les constructions, 1975.
VAN KOTEN - Windbelasting, TNO, 1974.
WYATT T.A. - A note on the significance of Cross-Modal
Correlations on the Evaluation of quasi-static
response to wind, 1974.
CHAPTER II
DAVENPORT A.G.

The Use of Taut-Strip Models in the prediction

of the response of long span bridges to
turbulent Wind, 1972.

KLOPPEL K. - Modellversuche im Windkanal zur Bemessung von

BrUcken gegen die Gefahr Winderregter
Schwingungen, 1967.
WIANECKY J. - Wind excited vibration of cable stayed bridge,
1~80.

CHAPTER II I
BORGES J.F., CASTANHETA M., BORGES A.R.J. - Design criteria for
wind loads on statistical Bases,
Proceedings
Wind effects on Buildings and structures, Tokyo,
1971.
WARDLAW R.L. - Approaches to the suppression of wind induced
vibrations on structures, 1979.
Laboratory report on the aerodynamic stability of
the Pasco-Kennewick intercity bridge, 1974.

449

LIST OF PARTICIPANTS
Mr. K. Ak1ll1lar (Turkey)

Dr. N.

Akk~

(Turkey)

Mr. H. Belkaya (Turkey)

Mr. V. Caby (Belgium)
Dr. J. Church (U.K. )
Mr. C. Debuysscher (Belgium)
Mr. E. Erkiakan (Turkey)
Dr. 1:I. R. Evans (U.K.)
Mr. T. Friedrich (Germany)
Prof.Dr. P. Glilkan (Turkey)
Dr. H. Glinerman (Turkey)
Prof.Dr. C.P. Heins (U.S.A.)
Dr. E. Karaesmen (Turkey)
Prof .Dr. 1. Kaya (Turkey)
Prof.Dr. M. K1c1man (Turkey)
Mr. C. Kozanoglu (Turkey)
Mr. N. Kok (Turkey)
ProLDr. C.N. Kostem (U.S.A.)
Dr. A.R.G. Lamas (Portugal)
Prof.Dr. F. Leonhardt (W. Germany)
Mr. S. Loseth (Norway)
Prof. Dr. Y. Mengi (Turkey)
Prof.Dr. C. Menn (Switzerland)
Dr. A. P. Parke r (U. K. )
:?rof. A.C. Scordelis (U.S.A.)
Dr. 1.1. Sungur (Turkey)
Dr. T. Tankut (Turkey)
Dr. C. Trezos (Greece)
ProLDr. O. Ural (u.S.A.)

450

Dr. P. Waldron (U.K.)

Prof.Dr. S.T. Wasti (Turkey)
Dr. A.

YargL~oglu

(Turkey)

Prof.Dr. B.T. Yen (U.S.A.)

Prof.Dr. V. Yerlici (Turkey)
Dr. Q. YLlmaz (Turkey)
Lieut. R. YLlmaz (Turkey)
Prof.Dr. E. Zeller (France)
Dr. H. Sucuoglu (Turkey)