V 110 KM/H 146 M

v = 110 km/h
146 m
335 m
146 m
- 25
15
5
-5
-15
-25
-35
with tuned mass damper (TMD)
-45
without tuned mass damper (TMD)
the truck leaves the bridge
Mid-point vertical displacement (mm) -55
0
10
20
30
40
Time (s)
Response of Cable-Stayed and

Suspension Bridges to Moving Vehicles
Analysis methods and practical modeling techniques
Raid Karoumi
Royal Institute of Technology
Department of Structural Engineering
TRITA-BKN. Bulletin 44, 1998
ISSN 1103-4270
ISRN KTH/BKN/B--44--SE
Doctoral Thesis
Response of Cable-Stayed and Suspension

Bridges to Moving Vehicles
Raid Karoumi
S-100 44 Stockholm, Sweden
Akademisk avhandling
Som med tillstnd av Kungl Tekniska Hgskolan i Stockholm framlgges till offentlig
granskning fr avlggande av teknologie doktorsexamen fredagen den 12 februari
1999 kl 10.00 i Kollegiesalen, Valhallavgen 79, Stockholm. Avhandlingen frsvaras p
svenska.
Fakultetsopponent:
Docent Sven Ohlsson

Huvudhandledare: Professor Hkan Sundquist
ISSN 1103-4270
Stockholm 1999
Response of Cable-Stayed and Suspension

Bridges to Moving Vehicles
Raid Karoumi
S-100 44 Stockholm, Sweden
_____________________________________________________________________
ISSN 1103-4270
Doctoral Thesis
To my wife, Lena,
to my daughter and son, Maria and Marcus,
and to my parents, Faiza and Sabah.
Akademisk avhandling som med tillstnd av Kungliga Tekniska Hgskolan i

Stockholm framlgges till offentlig granskning fr avlggande av teknologie
doktorsexamen fredagen den 12 februari 1999.
Raid Karoumi 1999
KTH, TS- Tryck & Kopiering, Stockholm 1999
______________________________________________________________________
Abstract
______________________________________________________________________
This thesis presents a state-of-the-art-review and two different approaches for solving the
moving load problem of cable-stayed and suspension bridges.
The first approach uses a simplified analysis method to study the dynamic response of
simple cable-stayed bridge models. The bridge is idealized as a Bernoulli-Euler beam on elastic
supports with varying support stiffness. To solve the equation of motion of the bridge, the finite
difference method and the mode superposition technique are used.
The second approach is based on the nonlinear finite element method and is used to study
the response of more realistic cable-stayed and suspension bridge models
considering exact cable behavior and nonlinear geometric effects. The cables are
modeled using a two-node catenary cable element derived using exact analytical
expressions for the elastic catenary. Two methods for evaluating the dynamic response
are presented. The first for evaluating the linear traffic load response using the mode
superposition technique and the deformed dead load tangent stiffness matrix, and the second for
the nonlinear traffic load response using the Newton-Newmark algorithm.
The implemented programs have been verified by comparing analysis results with
those found in the literature and with results obtained using a commercial finite
element code. Several numerical examples are presented including one for the Great Belt
suspension bridge in Denmark. Parametric studies have been conducted to
investigate the effect of, among others, bridge damping, bridge-vehicle interaction, cables
vibration, road surface roughness, vehicle speed, and tuned mass dampers.
From the numerical study, it was concluded that road surface roughness has great
influence on the dynamic response and should always be considered. It was also found
that utilizing the dead load tangent stiffness matrix, linear dynamic traffic load analysis give
sufficiently accurate results from the engineering point of view.
Key words: cable-stayed bridge, suspension bridge, Great Belt suspension bridge, bridge,
moving loads, traffic-induced vibrations, bridge-vehicle interaction, dynamic analysis, cable
element, finite element analysis, finite difference method, tuned mass damper.
i
ii
______________________________________________________________________
Preface
__________________________________________________________________
____
The research presented in this thesis was carried out at the Department of Structural
Engineering, Structural Design and Bridges group, at the Royal Institute of
Technology (KTH) in Stockholm. The project has been financed by KTH and the Axel
and Margaret Ax:son Johnson Foundation. The work was conducted under the
supervision of Professor Hkan Sundquist to whom I want to express my sincere
appreciation and gratitude for his encouragement, valuable advice and for always
having time for discussions. I also wish to thank Dr. Costin Pacoste for reviewing the
manuscript of this report and providing valuable comments for improvement.
Finally, I would like to thank my wife Lena Karoumi, my daughter and son, and my
parents for their love, understanding, support and encouragement.
Stockholm, January 1999

Raid Karoumi
iii
iv
______________________________________________________________________
Contents
______________________________________________________________________
Abstract i
Preface iii
General Introduction and Summary
1
Part A State-of-the-art Review and a Simplified Analysis Method for Cable7
Stayed Bridges
1 Introduction
9
1.1 General. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2 Review of previous research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.2.1
Research on cable-stayed bridges . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.2.2
Research on other bridge types. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.3 General aims of the present study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2 Vehicle and Structure Modeling
29
2.1 Vehicle
models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.2 Bridge
structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.2.1 Major
assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2.2
Differential equation of motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2.3 Spring
stiffness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.3 Bridge deck surface roughness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3 Response
Analysis
43
3.1 Dynamic
analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.1.1 Eigenmode
extraction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
v
3.1.2
Response of the bridge. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.2 Static
analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4 Numerical Examples and Model Verifications
51
4.1 General. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.2 Simply supported bridge, moving force model . . . . . . . . . . . . . . . . . . . . . . . . 52
4.3 Multi-span continuous bridge with rough road surface . . . . . . . . . . . . . . . . . 57
4.4 Simple cable-stayed bridge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.5 Three-span cable-stayed bridge. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.6 Discussion of the numerical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5 Conclusions and Suggestions for Further Research
83
5.1 Conclusions of Part A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.2 Suggestions for further research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Bibliography of Part A
87
Part B Refined Analysis Utilizing the Nonlinear Finite Element Method
97
6 Introduction
99
6.1 General ......................................................................................................... 99
6.2 Cable structures and cable modeling techniques ....................................... 101
6.3 General aims of the present study .............................................................. 103
7 Nonlinear
Finite
Elements
105
7.1 General ....................................................................................................... 105
7.2 Modeling
of
cables ..................................................................................... 106
7.2.1 Cable element formulation............................................................ 107
7.2.2 Analytical
verification................................................................... 111
7.3 Modeling of bridge deck and pylons.......................................................... 113
vi
8 Vehicle and Structure Modeling
117
8.1 Vehicle
models........................................................................................... 117
8.2 Vehicle load modeling and the moving load algorithm............................. 121
8.3 Bridge
structure .......................................................................................... 123
8.3.1 Modeling of damping in cable supported bridges......................... 123
8.3.2 Bridge deck surface roughness...................................................... 126
8.4 Tuned vibration absorbers.......................................................................... 127
9 Response
Analysis
133
9.1 Dynamic
Analysis ...................................................................................... 133
9.1.1 Linear dynamic analysis................................................................ 134
9.1.1.1 Eigenmode extraction and normalization of eigenvectors..... 135
9.1.1.2 Mode
superposition technique ............................................... 136
9.1.2 Nonlinear
dynamic
analysis .......................................................... 138
9.2 Static
analysis............................................................................................. 141
10 Numerical
Examples
143
10.1 Simply supported bridge ............................................................................ 144
10.2 The Great Belt suspension bridge .............................................................. 149
10.2.1 Static response during erection and natural frequency analysis ... 151
10.2.2 Dynamic response due to moving vehicles................................... 154
10.3 Medium span cable-stayed bridge.............................................................. 158
10.3.1 Static response and natural frequency analysis............................. 159
10.3.2 Dynamic response due to moving vehicles parametric study.... 162
10.3.2.1 Response due to a single moving vehicle .............................. 163
10.3.2.2 Response due to a train of moving vehicles, effect of bridge-
vehicle interaction and cable modeling.................................. 165

10.3.2.3 Speed and bridge damping effect........................................... 166
10.3.2.4 Effect of surface irregularities at the bridge entrance ............ 167
10.3.2.5 Effect of tuned vibration absorbers ........................................ 168
vii
11 Conclusions and Suggestions for Further Research
181
11.1 Conclusions of Part B................................................................................. 181
11.1.1 Nonlinear finite element modeling technique............................... 181
11.1.2 Response due to moving vehicles ................................................. 182
11.2 Suggestions for further research................................................................. 184
A Maple
Procedures
187
A.1 Cable
element............................................................................................. 187
A.2 Beam
element............................................................................................. 188
Bibliography of Part B
189
viii
______________________________________________________________________
General Introduction and Summary
______________________________________________________________________
Due to their aesthetic appearance, efficient utilization of structural materials and other
notable advantages, cable supported bridges, i.e. cable-stayed and suspension bridges, have
gained much popularity in recent decades. Among bridge engineers the
popularity of cable-stayed bridges has increased tremendously. Bridges of this type are
now entering a new era with main span lengths reaching 1000 m. This fact is due, on one hand to
the relatively small size of the substructures required and on the other hand to the development
of efficient construction techniques and to the rapid progress in the analysis and design of this
type of bridges.
Ever since the dramatic collapse of the first Tacoma Narrows Bridge in 1940, much
attention has been given to the dynamic behavior of cable supported bridges. During the
last fifty-eight years, great deal of theoretical and experimental research was conducted in order
to gain more knowledge about the different aspects that affect the behavior of this type of
structures to wind and earthquake loading. The recent
developments in design technology, material qualities, and efficient construction
techniques in bridge engineering enable the construction of lighter, longer, and more
slender bridges. Thus nowadays, very long span cable supported bridges are being
built, and the ambition is to further increase the span length and use shallower and more
slender girders for future bridges. To achieve this, accurate procedures need to be developed that
can lead to a thorough understanding and a realistic prediction of the structural response due to
not only wind and earthquake loading but also traffic

loading. It is well known that large deflections and vibrations caused by dynamic tire
forces of heavy vehicles can lead to bridge deterioration and eventually increasing maintenance
costs and decreasing service life of the bridge structure.
The recent developments in bridge engineering have also affected damping capacity of
bridge structures. Major sources of damping in conventional bridgework have been
largely eliminated in modern bridge designs reducing the damping to undesirably low
levels. As an example, welded joints are extensively used nowadays in modern bridge designs.
This has greatly reduced the hysteresis that was provided in riveted or bolted
1
joints in earlier bridges. For cable supported bridges and in particular long span cablestayed bridges, energy dissipation is very low and is often not enough on its own to suppress
vibrations. To increase the overall damping capacity of the bridge structure, one possible option
is to incorporate external dampers (discrete damping devices such as viscous dampers and tuned
mass dampers) into the system. Such devices are
frequently used today for cable supported bridges. However, it is not believed that this is
always the most effective and the most economic solution. Therefore, a great deal of research is
needed to investigate the damping capacity of modern cable supported
bridges and to find new alternatives to increase the overall damping of the bridge
structure.
To consider dynamic effects due to moving traffic on bridges, structural engineers
worldwide rely on dynamic amplification factors specified in bridge design codes.
These factors are usually a function of the bridge fundamental natural frequency or span
length and states how many times the static effects must be magnified in order to cover the
additional dynamic loads. This is the traditional method used today for
design purpose and can yield a conservative and expensive design for some bridges
but might underestimate the dynamic effects for others. In addition, design codes
disagree on how this factor should be evaluated and today, when comparing different
national codes, a wide range of variation is found for the dynamic amplification factor.
Thus, improved analytical techniques that consider all the important parameters that
influence the dynamic response, such as bridge-vehicle interaction and road surface roughness,
are required in order to check the true capacity of existing bridges to heavier traffic and for
proper design of new bridges.
Various studies, of the dynamic response due to moving vehicles, have been conducted
on ordinary bridges. However, they cannot be directly applied to cable supported
bridges, as cable supported bridges are more complex structures consisting of various
structural components with different properties. Consequently, more research is
required on cable supported bridges to take account of the complex structural response
and to realistically predict their response due to moving vehicles. Not only the
dynamic behavior of new bridges need to be studied and understood but also the
response of existing bridges, as governments and the industry are seeking
improvements in transport efficiency and our aging and deteriorating bridge
infrastructure is being asked to carry ever increasing loads.
2
The aim of this work is to study the moving load problem of cable supported bridges
using different analysis methods and modeling techniques. The applicability of the implemented
solution procedures is examined and guidelines for future analysis are proposed. Moreover, the
influence of different parameters on the response of cable supported bridges is investigated.
However, it should be noted that the aim is not to completely solve the moving load problem and
develop new formulas for the dynamic
amplification factors. It is to the authors opinion that one must conduct more
comprehensive parametric studies than what is done here and perform extensive
testing on existing bridges before introducing new formulas for design.
This thesis contains two separate parts, Part A (Chapter 1-5) and Part B (Chapter 6-11),
where each has its own introduction, conclusions, and reference list. These two parts present two
different approaches for solving the moving load problem of
ordinary and cable supported bridges.
Part A, which is a slightly modified version of the licentiate thesis presented by the author
in November 96, presents a state-of-the-art review and proposes a simplified analysis method for
evaluating the dynamic response of cable-stayed bridges. The
bridge is idealized as a Bernoulli-Euler beam on elastic supports with varying support
stiffness. To solve the equation of motion of the bridge, the finite difference method and the
mode superposition technique are used. The utilization of the beam on elastic bed analogy makes
the presented approach also suitable for analysis of the dynamic response of railway tracks
subjected to moving trains.
In Part B, a more general approach, based on the nonlinear finite element method, is
adopted to study more realistic cable-stayed and suspension bridge models
considering, e.g., exact cable behavior and nonlinear geometric effects. A beam
element is used for modeling the girder and the pylons, and a catenary cable element,
derived using exact analytical expressions for the elastic catenary, is used for modeling the
cables. This cable element has the distinct advantage over the
traditionally used elements in being able to approximate the curved catenary of the real
cable with high accuracy using only one element. Two methods for evaluating the
dynamic response are presented. The first for evaluating the linear traffic load
response using the mode superposition technique and the deformed dead load tangent
stiffness matrix, and the second for the nonlinear traffic load response using the NewtonNewmark algorithm. Damping characteristics and damping ratios of cable
supported bridges are discussed and a practical technique for deriving the damping
3
matrix from modal damping ratios, is presented. Among other things, the effectiveness of
using a tuned mass damper to suppress traffic-induced vibrations and the effect of including
cables motion and modes of vibration on the dynamic response are
investigated.
To study the dynamic response of the bridge-vehicle system in Part A and B, two sets of
equations of motion are written one for the vehicle and one for the bridge. The two sets of
equations are coupled through the interaction forces existing at the contact points of the two
subsystems. To solve these two sets of equations, an iterative
procedure is adopted. The implemented codes fully consider the bridge-vehicle
dynamic interaction and have been verified by comparing analysis results with those
found in the literature and with results obtained using a commercial finite element code.
The following basic assumptions and restrictions are made:
elastic structural material
two-dimensional bridge models. Consequently, the torsional behavior caused by
eccentric loading of the bridge deck is disregarded
as the damage to bridges is done mostly by heavy moving trucks rather than
passenger cars, only vehicle models of heavy trucks are used
simple one dimensional vehicle models are used consisting of masses, springs, and
viscous dampers. Consequently, only vertical modes of vibration of the vehicles
are considered
it is assumed that the vehicles never loses contact with the bridge, the springs and the
viscous dampers of the vehicles have linear characteristics, the bridge-vehicle interaction forces
act in the vertical direction, and the contact between the bridge and each moving vehicle is
assumed to be a point contact. Moreover, longitudinal
forces generated by the moving vehicles are neglected.
Based on the study conducted in Part A and B, the following guidelines for future
analysis and practical recommendations can be made:
for preliminary studies using very simple cable-stayed bridge models to determine the
feasibility of different design alternatives, the approach presented in Part A can
4
be adopted as it is found to be simple and accurate enough for the analysis of the dynamic
response. However, for analysis of more realistic bridge models where
e.g. exact cable behavior, nonlinear geometric effects, or non-uniform crosssections are to be considered, this approach becomes difficult and cumbersome.
For such problems, the finite element approach presented in Part B is found to be
more suitable as it can easily handle such analysis difficulties
for cable supported bridges, nonlinear static analysis is essential to determine the dead
load deformed condition. However, starting from this position and utilizing
the dead load tangent stiffness matrix, linear static and linear dynamic traffic load
analysis give sufficiently accurate results from the engineering point of view
it is recommended to use the mode superposition technique for such analysis
especially if large bridge models with many degrees of freedom are to be analyzed.
For most cases, sufficiently accurate results are obtained including only the first 25
to 30 modes of vibration
correct and accurate representation of the true dynamic response is obtained only if road
surface roughness, bridge-vehicle interaction, bridge damping, and cables
vibration are considered. For the analysis, realistic bridge damping values, e.g.
based on results from tests on similar bridges, must be used
care should be taken when the dynamic amplification factors given in the different
design codes and specifications are used for cable supported bridges, as it is not believed that
these can be used for such bridges. For some cases it is found that
design codes underestimate the additional dynamic loads due to moving vehicles.
Consequently, each bridge of this type, particularly those with long spans, should be
analyzed as made in Part B of this thesis. For the final design, such analysis
should be performed more accurately using a 3D bridge and vehicle models and
with more realistic traffic conditions
to reduce damage to bridges not only maintenance of the bridge deck surface is
important but also the elimination of irregularities (unevenness) in the approach
pavements and over bearings. It is also suggested that the formulas for dynamic
amplification factors specified in bridge design codes should not only be a function of the
fundamental natural frequency or span length (as in many present design
codes) but also should consider the road surface condition.
5
It is believed that Part A presents the first study of the moving load problem of cablestayed bridges where this simple modeling and analysis technique is utilized. For Part B of this
thesis, it is believed that this is the first study of the moving load problem of cable-stayed and
suspension bridges where results from linear and nonlinear dynamic traffic load analysis are
compared. In addition, such analyses have not been performed earlier taking into account exact
cable behavior and fully considering the bridge-vehicle dynamic interaction.
Most certainly this study has not provided a complete answer to the moving load
problem of cable supported bridges. However, the author hopes that the results of this
study will be a help to bridge designers and researchers, and provide a basis for future work.
6
Part A
State-of-the-art Review and a
Simplified Analysis Method
for Cable-Stayed Bridges
7
8
Chapter
______________________________________________________________________
Introduction
______________________________________________________________________
1.1 General
Studies of the dynamic effects on bridges subjected to moving loads have been carried
out ever since the first railway bridges were built in the early 19th century. Since that time
vehicle speed and vehicle mass to the bridge mass ratio have been increased,
resulting in much greater dynamic effects. In recent years, the interest in traffic induced
vibrations has been increasing due to the introduction of high-speed vehicles, like the TGV train
in France and the Shinkansen train in Japan with speeds exceeding 300 km/h. The increasing
dynamic effects are not only imposing severe conditions
upon bridge design but also upon vehicle design, in order to give an acceptable level of
comfort for the passengers.
Modern cable-stayed bridges with their long spans are relatively new and have been
introduced widely only since the 1950, see Table 1.1 and Figure 1.2. The first modern cablestayed bridge was the Strmsund Bridge in Sweden opened to traffic in 1956.
For the study of the concept, design and construction of cable-stayed bridges, see the
excellent book by Gimsing [27] and also [28, 68, 75, 76, 79]. Cable supported bridges are special
because they are of the geometric-hardening type, as shown in Figure 1.3
on page 16, which means that the overall stiffness of the bridge increases with the
increase in the displacements as well as the forces. This is mainly due to the decrease of the cable
sag and increase of the cable stiffness as the cable tension increases.
Compared to other types of bridges, the dynamic response of cable-stayed bridges
subjected to moving loads is given less attention in theoretical studies. Static analysis and
dynamic response analysis of cable-stayed bridges due to earthquake and wind
loading, received, and have been receiving most of the attention, while only few
9
studies, see section 1.2.1, have been carried out to investigate the dynamic effects of
moving loads on cable-stayed bridges. However, with increasing span length and
increasing slenderness of the stiffening girder, great attention must be paid not only to the
behavior of such bridges under earthquake and wind loading but also under
dynamic traffic loading as well.
The dynamic response of bridges subjected to moving vehicles is complicated. This is
because the dynamic effects induced by moving vehicles on the bridge are greatly
influenced by the interaction between vehicles and the bridge structure. The important
parameters that influence the dynamic response are (according to previous research conducted in
this field, see section 1.2):
vehicle speed
road (or rail) surface roughness
characteristics of the vehicle, such as the number of axles, axle spacing, axle load,
natural frequencies, and damping and stiffness of the vehicle suspension system
the number of vehicles and their travel paths
characteristics of the bridge structure, such as the bridge geometry, support
conditions, bridge mass and stiffness, and natural frequencies.

For design purpose, structural engineers worldwide rely on dynamic amplification
factors (DAF), which are usually related to the first vibration frequency of the bridge or
to its span length. The DAF states how many times the static effects must be
magnified in order to cover additional dynamic loads resulting from the moving traffic
(DAF is usually defined as the ratio of the absolute maximum dynamic response to the absolute
maximum static response). Because of the simplicity of the DAF expressions specified in current
bridge design codes, these expressions cannot characterize the effect of all the above listed
parameters. Moreover, as these expressions are originally developed for ordinary bridges, it is
believed that for long span bridges like cable-stayed bridges the additional dynamic loads must
be determined in more accurate way in order to guarantee the planned lifetime and economical
dimensioning.
Figure 1.1 shows the variation of the DAF with respect to the fundamental frequency of
the bridge, recommended by different standards [66]. For cases where the DAF was related to the
span length, the fundamental frequency was approximated from the span length. It is apparent
from Figure 1.1 that the national design codes disagree on the
10
evaluation of the dynamic amplification factors, and although the specified traffic loads
vary in these codes, this does not explain such a wide range of variation for the DAF. In the
Swedish design code for new bridges, the Swedish National Road
Administration (Vgverket) includes the additional dynamic loads, due to moving
vehicles, in the traffic loads specified for the different types of vehicles. This gives a
constant DAF that is totally independent on the characteristics of the bridge. For bridges like
cable-stayed bridges that are more complex and behave differently
compared to ordinary bridges, this approach can lead to incorrect traffic loads to be used
for designing the bridge.
This part of the thesis presents a state-of-the-art review and a simplified analysis method
for evaluating the dynamic response of cable-stayed bridges. The bridge is
idealized as a Bernoulli-Euler beam on elastic supports with varying support stiffness.
To solve the equation of motion of the bridge, the finite difference method and the mode
superposition technique are used. The utilization of the beam on elastic bed
analogy makes the presented approach also suitable for analysis of the dynamic
response of railway tracks subjected to moving trains.
2.0
Canada CSA-S6-88m OHBDC
Swiss SIA-88, single vehicle
)
Swiss SIA-88, lane load
1.8
AASHTO-1989
(DAF
India, IRC
or
Germany, DIN1075
act
U.K. - BS5400 (1978)
1.6
France LCPC D/L=0.5
ion f
France LCPC D/L=5
plificat
D/L = Dead load / Live load
1.4
c am
ami
Dyn 1.2
1.0
0 1 2 3 4 5 6 7 8 9 10
Bridge fundamental frequency (Hz)
Figure 1.1
Dynamic amplification factors used in different national codes [66]
11
Bridge name
Country
Center span
Year of
Girder
(m)
completion
material
Tatara Japan
890
1999
Steel
Pont de Normandie
France
856
1995
Steel
Qingzhou Minjiang
China (Fuzhou)
605
1996
Composite
Yangpu China
(Shanghai)
602
1993
Composite
Xupu China
(Shanghai)
590
1996
Composite
Meiko-Chuo Japan
590
1997
Steel
Skarnsund Norway
530
1991
Concrete
Tsurumi Tsubasa
Japan
510
1994
Steel
resund Sweden/Denmark
490
2000
Steel
Ikuchi Japan
490
1991
Steel
Higashi-Kobe Japan
485
1994
Steel
Ting Kau
Hong Kong
475
1997
Steel
Seohae South
Korea
470
1998
unknown
Annacis Island
Canada
465
1986
Composite
Yokohama Bay
Japan
460
1989
Steel
Second Hooghly
India (Calcutta)
457
1992
Composite
Second Severn
England
456
1996
Composite
Queen Elizabeth II
England
450
1991
Composite
Rama IX
Thailand (Bangk.)
450
1987
Steel
Chongqing Second
China (Sichuan)
444
1996
Concrete
Barrios de Luna
Spain
440
1983
Concrete
Tongling China
(Anhui)
432
1995
Concrete
Kap Shui Mun
Hong Kong
430
1997
Composite
Helgeland Norway
425
1991
Concrete
Nanpu China
(Shanghai)
423
1991
Composite
Vasco da Gama
Portugal
420
1998
unknown
Hitsushijima Japan
420
1988
Steel
Iwagurujima Japan
420
1988
Steel
Yuanyang Hanjiang
China (Hubei)
414
1993
Concrete
Uddevalla Sweden
414
2000
Composite
Meiko-Nishi Ohashi
Japan
405
1986
Steel
S:t Nazarine
France
404
1975
Steel
Elorn France
400
1994
Concrete
Vigo-Rande Spain
400
1978
Steel
Table 1.1
Major cable-stayed bridges in the world
12
Dame Point
USA (Florida)
396
1989
Concrete
Houston Ship Channel
USA (Texas)
381
1995
Composite
Luling, Mississippi
USA
372
1982
Steel
Duesseldorf-Flehe Germany
368
1979
Steel
Tjrn (new)
Sweden
366
1981
Steel
Sunshine Skyway
USA (Florida)
366
1987
Concrete
Yamatogawa Japan
355
1982
Steel
Neuenkamp Germany
350
1970
Steel
Ajigawa (Tempozan)
Japan
350
1990
Steel
Glebe Island
Australia
345
1990
Concrete
ALRT Fraser
Canada
340
1985
Concrete
West Gate
Australia
336
1974
Steel
Talmadge Memorial
USA (Georgia)
335
1990
Concrete
Rio Parana (2 bridges)
Argentina
330
1978
Steel
Karnali Nepal
325
1993
Composite
Khlbrand Germany
325
1974
Steel
Guadiana Portugal/Spain
324
1991
Concrete
Kniebruecke Germany
320
1969
Steel
Brotonne France
320
1977
Concrete
Mezcala Mexico
311
1993
Composite
Erskine Scotland
305
1971
Steel
Bratislava Slovakia
305
1972
Steel
Severin Germany
302
1959
Steel
Moscovsky Ukraine
(Kiev)
300
1976
Steel
Faro Denmark
290
1985
Steel
Dongying China
(Shandong)
288
1987
Steel
Mannheim Germany
287
1971
Steel
Wadi Kuf
Libya
282
1972
Concrete
Leverkusen Germany
280
1965
Steel
Bonn Nord
Germany
280
1967
Steel
Speyer Germany
275
1974
Steel
East Huntington
USA
274
1985
Concrete
Bayview USA
274
1990
Composite
River Waal
Holland
267
1974
Concrete
Theodor Heuss
Germany
260
1958
Steel
Yonghe China
(Tianjin)
260
1987
Concrete
Table 1.1
(continued)
13
Oberkassel Germany
258
1975
Steel
Rees-Kalkar Germany
255
1967
Steel
Weirton-Steubenville USA 250
1986
Steel
Chaco/Corrientes Argentina
245
1973
Concrete
Papineau-Leblanc Canada
241
1971
Steel
Krkistensalmi Finland
240
1996
Composite
Maracaibo Venezuela
235
1962
Concrete
Pasco Kennewick
USA
229
1978
Concrete
Jinan Yellow River
China (Shandong)
220
1983
Concrete
Toyosato-Ohashi Japan
216
1970
Steel
Onomichi-Ohashi Japan
215
1968
Steel
Strmsund Sweden
183
1956
Steel
Table 1.1
(continued)
1000
900
Steel girder
800
Composite girder
Concrete girder
700
600
500
400
Length of center span (m)
300
200
100
1950
1960
1970
1980
1990
2000
Year of completion
Figure 1.2
Span length increase of cable-stayed bridges in the last fifty years
14
1.2
Review of previous research
1.2.1 Research on cable-stayed bridges
In recent years the dynamic behavior of cable-stayed bridges has been a source of
interesting research. This includes free vibration and forced vibration due to wind and
earthquakes, see for example [2, 9, 47]. However, literature dealing with the dynamics of these
bridges due to moving vehicles is relatively scarce.
For a cable-stayed footbridge, theoretical and experimental study on the effectiveness of
tuned mass dampers, TMDs, was carried out in [6]. In this study, tests with one and two persons
jumping or running were performed, and acceleration responses with the TMD locked and
unlocked were compared. In [59, 60], modal testing of the Tjrn
bridge, a cable-stayed bridge in Sweden with a 366 m main span, is described. And in
[11], dynamic load testing on the Riddes-Leytron bridge, a cable-stayed bridge in
Switzerland with a 60 m main span, is presented.
Previous investigations on the dynamic response of cable-stayed bridges subjected to
moving loads are summarised in the following:
Fleming and Egeseli (1980) [21, 22] compared linear and nonlinear dynamic analysis
results for a cable-stayed bridge subjected to seismic and wind loads. The nonlinear dynamic
response due to a single moving constant force was also studied. A two-dimensional (2-D) harp
system cable-stayed bridge model with a main span of 260 m
was adopted, and the bridge was discretized using the finite element method. The
nonlinear behavior of the cables due to sag effect and the nonlinear behavior of the
bending members due to the interaction of axial and bending deformations, were
considered. Fleming et al. showed that although there is significant nonlinear behavior
during the static application of the dead load, the structure can be assumed to behave as a linear
system starting from the dead load deformed state for both static and
dynamic loads, as illustrated in Figure 1.3. This means that influence lines and
superposition technique can be used in the design process.
Considering only seismic loading a similar comparison was conducted in [2] and the
same conclusion was made.
15
Generalized force
ic
dynamic
namic
load
eigenvalue
problem
ear dy
lin
linear dynamic
nonlinear dynam
cable structures
dead
load
non-cable structures
linear static
nonlinear static
Generalized displacement
Figure 1.3
Schematic diagram showing the difference between the behavior of cable
structures and non-cable structures and also the accuracy in the results
from different analysis procedures
Wilson and Barbas (1980) [89] performed theoretical and experimental works on cablestayed bridge models to determine the dynamic effects due to a moving vehicle.
For the theoretical bridge model, a 2-D undamped continuous Bernoulli-Euler beam
resting on discrete evenly spaced elastic supports, was adopted. The vehicle was
modeled using one or two constant forces travelling at constant speeds. For the
solution of the problem, mode superposition technique was used. All bridge cables
were approximated by linear springs with equal stiffness, and solutions with two to five
cables in the main span were presented. Only the main span was considered in this study, and the
road surface roughness was neglected. The experimental models
consisted of straight steel beams (cross section = 0.0492 cm 1.97 cm and length =
2.36 m) spliced end to end at the supports (springs) so that continuous spans of up to 23.6
m could be tested. By prestressing the bridge model, an initial flatness to within
0.2 cm under self-weight was achieved. For the interior span supports, coil springs
were used. One or two linear induction motors running in a separate track above the bridge
model were used to move the point load vehicle at constant speeds in the range of 1.22 m/s to
8.85 m/s. The total vehicle model weight was about 1.2 kg. Wilson et al.
presented diagrams showing, for both the theoretical and the experimental models, the
influence of the speed parameter on the DAF values for displacements and bending
16
moments. To show the influence of cable stiffness, diagrams with different values for the
spring stiffness were also presented. The results showed good agreement between the theoretical
and the experimental work. According to Wilson et al., the main
reasons for the differences in the results were due to the inability of the experimental
system to maintain constant speed, and the neglection of the inertia effects of the experimental
transit load in the theoretical model. Wilson et al. concluded also that increasing the spring
stiffness at the supports will for most cases lead to an increase in the bridge dynamic response.
Rasoul (1981) [69] used the structural impedance method1 and studied the dynamic
response of bridges due to moving vehicles. The bridge flexibility functions were
evaluated by using a static analysis of the bridge subjected to unit loads. A simply
supported beam, a continuous beam, and very simple cable-stayed bridges were
studied. For the cable-stayed bridges, two different analysis methods were used,
namely an approximate method using the concept of continuous beam with
intermediate elastic supports, fixed pylon heads and with the cables approximated by
springs, and a more exact method taken into account the effect of the axial force in the girder and
the transverse displacement of the pylons by using the reduction method.
Solutions with different girder damping ratios for a simple 2-D cable-stayed bridge with
only two cables were presented. The traffic load was modeled as a series of
vehicles traversing along the bridge. Each vehicle was modeled with a sprung mass
and an unsprung mass giving a vehicle model with two degrees of freedom (2 DOF).
Different traffic conditions were studied, and the effect of vehicle speed and bridge
damping on DAF was presented. Rasoul concluded that bridge damping was one of
the important parameters affecting the DAF, and that the DAF was considerably
higher for the cables than for other elements of the bridge. Rasoul found also that for a
single vehicle travelling at constant speed, the moving force solutions are good
approximations of the exact solutions. The road surface roughness was totally
neglected in this study.
Alessandrini, Brancaleoni and Petrangeli (1984) [3] studied the dynamic response of
railway cable-stayed bridges subjected to a moving train. The bridge was
discretized using the finite element method, and geometric nonlinearities for the cables
were considered by using an equivalent modulus of elasticity. The solution was carried 1 In this
study, the equation of motion of the bridge was formulated in an integral form using the
flexibility function (Greens function) for the bridge.
17
out using a direct time integration procedure (explicit algorithm). 2-D fan type cablestayed bridges with steel deck and center spans of about 160, 260, and 412 m were
adopted. Five different train lengths of 12-260 m and three different values for the mass
per unit length of the train to the mass per unit length of the bridge were
considered. The train was simulated using moving masses at three different speeds of 60,
120, and 200 km/h. DAF values for mid-span vertical displacement, axial force in the longest
center span cable, and axial force in the anchor cables, were presented and compared with those
obtained by the Italian Railways Steel Bridge Code. Alessandrini et al. concluded that, for most
cases, the standard expression for DAF given in the Italian Railway Code were not admissible
for cable-stayed bridges. It was also found that for speeds of up to about 120 km/h, the dynamic
effects were small if not
negligible. For speeds higher than 120 km/h the DAF values increase rapidly and for
speeds of about 200 km/h, DAF values greater than those prescribed by the Italian
Railway Code were observed. The rail surface roughness was neglected in this study.
Brancaleoni, Petrangeli and Villatico (1987) [8] presented solutions for the dynamic
response of a railway cable-stayed bridge subjected to a single moving high-speed
locomotive. The bridge was discretized using the finite element method and geometric
nonlinearities were considered in the analysis. The analysis was carried out using a direct time
integration procedure (explicit algorithm). A 2-D modified fan type cable-stayed bridge with
concrete deck and a main span of 150 m, was adopted. The bridge deck and the pylons were
modeled using beam elements, while nonlinear cable
elements with parabolic shape functions were adopted for the cables. For the bridge, a
Rayleigh type damping producing 2 % of the critical on the first mode has been used.
Solutions for a total train weight of about 95 tons, treated as a set of moving forces, a set
of moving masses, and a four axles 6 DOF sprung mass model, were presented.
Three different train speeds were considered, 60, 120, and 200 km/h. Diagrams
showing the variation of DAF with speed for the three different vehicle models, and time
histories for the mid-span vertical displacements, were presented. The rail surface roughness was
neglected in this study. Brancaleoni et al. concluded that treating the train as a set of moving
forces or moving masses results in lower DAF values for the girder bending moments and the
cable axial forces, and higher DAF values for the
center span vertical displacements. Brancaleoni et al. showed also that bending
moment amplification factors were greater than those for cable axial forces and center
span vertical displacements. The rail surface roughness was neglected in this study.
18
Walther (1988) [80] performed experimental study on a cable-stayed bridge model with
slender deck to determine the dynamic displacements produced by the passage of a 250 kN
vehicle at different speeds. The bridge model, which was equipped with rails and a launching
ramp, represented a 3 span modified fan type cable-stayed bridge with a 200 m main span and
about 100 m side spans. The deck and the two A-shaped
pylons were made of reinforced microconcrete, while piano cord wires with a diameter of
2 to 3 mm were used for the cables. The scale adopted was 1/20 giving a total
length of about 20 m for the bridge model and a model vehicle weight of 62.5 kg.
Different model vehicle speeds from 0.6 to 3.8 m/s (corresponds to real vehicle speeds of
about 10 to 61 km/h) were used, and tests with and without a plank in the main span were
undertaken to simulate different road surface conditions. Time histories for mid-span vertical
displacements were presented, for centric and eccentric vehicle
movements, with or without a plank, and for fixed joint and free joint at mid-span.
Based on measured data, vertical accelerations were calculated and a study of
physiological effects (human sensitivity to vibrations) was undertaken. Walther
concluded that from the physiological effects point of view, the structure could be
considered acceptable to tolerable depending on the road surface condition. The
maximum DAF value for mid-span vertical displacement was found to be 1.3. Walther
found also that placing a joint at the center of the bridge deck only give very local effects
and have little influence on the global dynamic behavior of the model.
Indrawan (1989) [45] studied the dynamic behavior of Rama IX cable-stayed bridge in
Bangkok due to an idealized single axle vehicle travelling over the bridge at
constant speeds. The 450 m main span, modified fan type, single plane, cable-stayed
bridge, was modeled in 2-D. The dynamic response was analyzed using the finite
element method and mode superposition technique, including only the first 10 modes of
vibration. All analyses were carried out in the frequency domain and time domain responses
were calculated using the fast Fourier transform (FFT) technique. The
bridge deck and pylons were modeled using beam elements while truss elements were
used for the cables. When evaluating the stiffness of each cable, the cable sag was
considered by using an equivalent tangent modulus of elasticity. Time histories
showing cable forces, mid-span vertical displacements, and pylon tops horizontal
displacements, were presented for different types of vehicle models moving over a
smooth surface, a rough surface, and a bumpy surface, at speeds of 36 to 540 km/h.
The single axle vehicle was modeled as a constant force, an unsprung mass, and a
sprung mass (1 DOF system). For the sprung mass vehicle model the assumed natural
frequency and damping ratio were 1.39 Hz and 3.5 % respectively. The inertial effect
19
in the vehicle due to bridge vibrations was totally neglected by the author. The road
surface roughness was generated from a power spectral density function (PSD) (the
same as the one used here in sec. 2.3). Since Rama IX bridge is equipped with tuned mass
dampers (TMD) to suppress wind induced oscillations, a comparison was made
between the dynamic response with and without the presence of a TMD. The TMD
was assumed to be installed at mid-span and tuned to the first flexural mode of
vibration. Indrawan found that the TMD was very effective in reducing the vibration
level of cables anchored in the vicinity of the mid-span. But he suggested that, instead of using
TMDs, viscous dampers should be installed in all cables to more effectively increase the fatigue
life of the cables. The analysis results showed also that the DAF
increases with increasing vehicle speed and can for bumpy surface reach very high
values.
Khalifa (1991) [49] carried out an analytical study on two cable-stayed bridges with
main spans of 335 m and 670 m. The 3 spans cable-stayed bridges were of the double plane
modified fan type, and were modeled in 3-D and discretized using the finite
element method. The dynamic response was evaluated using the mode superposition
technique, where each equation was solved adopting the Wilson- numerical
integration scheme. The linear dynamic analysis, based on geometrically nonlinear
static analysis (see Figure 1.3), was conducted using the deformed dead load tangent
stiffness matrix. The effect of including cable modes on the overall bridge dynamics was
investigated by discretizing each cable of the longer bridge as one element and as eight equal
elements. The dynamic response was evaluated for a single moving vehicle and a train of
vehicles moving in one direction or in both directions. The vehicles, travelling with constant
speeds of about 43 to 130 km/h over a smooth and a rough
surface, were approximated using a constant moving force model and a sprung mass
model. For the sprung mass vehicle model the assumed natural frequency and damping
ratio were 1 or 3 Hz and 3 %, respectively. The road surface roughness was generated from a
power spectral density function (PSD) (the same as the one used here in sec.
2.3). Diagrams showing the influence of bridge damping ratio, cable vibrations,
vehicle model type, vehicle speed, number of vehicles, traffic direction, and deck
condition, on the bridge dynamic response, were presented. A stress-life fatigue
analysis was also conducted to estimate the virtual cable life under continuous moving
traffic loads. Khalifa found that the fatigue life of stays cables were relatively very short if they
were subjected to extreme vibrational stresses resulting from a continuous fluctuating heavy
traffic. The results also showed that the magnitude of the dynamic response was influenced by
the bridge damping ratio, the type of vehicle model, and
20
the roughness of the bridge deck. The author recommended discretizing each cable
into small elements when calculating the dynamic response due to environmental and
service dynamic loads.
Wang and Huang (1992) [84] studied the dynamic response of a cable-stayed bridge due
to a vehicle moving across rough bridge decks. The vehicle was simulated by a
nonlinear vehicle model with 3-axles and seven degrees of freedom. A 2-D modified
fan type cable-stayed bridge with concrete deck and a main span of 128 m, was
adopted. The bridge deck roughness was generated using PSD functions. The dynamic
response was analyzed using the finite element method and the geometric nonlinear
behavior of the bridge due to dead load was considered. The equation of motion for the
vehicle was solved using the fourth-order Runge-Kutta integration scheme, and an iterative
procedure with mode superposition technique was used for solving the
equation of motion for the bridge. Wang et al. concluded that the mode superposition
procedure used was effective and involved much less computation, because accurate
results of the bridge dynamic response could be obtained based on solving only 8 to 12
equations of motion of the bridge. Wang et al. noted that the DAF of all components of
the bridge were generally less than 1.2 for very good road surface, but increased tremendously
with increasing road surface roughness. High values of DAF were noted at the girder near the
pylons and at the lower ends of the pylons and piers, but
comparatively small DAF values were noted at the girder adjacent to the mid-span of the
bridge.
Miyazaki et al. (1993) [55] carried out an analytical study on the dynamic response and
train running quality of a prestressed concrete multicable-stayed railway bridge planned for
future use on the high-speed Shinkansen line. For the analysis, the
simulation program DIASTARS, developed at the Japanese Railway Technical
Research Institute, was used. The railway track and the bridge structure were modeled
using the finite element method. In this study, a 2-D and a 3-D bridge models of a two span
cable-stayed bridge, were used. The 2-D bridge model together with a simple 12
cars train model consisting of only constant forces were used to evaluate the dynamic
response of the bridge, while the 3-D bridge and the 3-D train model were used to
evaluate the train running quality. The 3-D Shinkansen train model consisted of 12
cars where each car consisted of a body, two bogies, and four wheelsets giving 23
DOF. The track was assumed to be directly placed on the bridge deck surface, and the rail
surface roughness was neglected. The 3-D bridge deck was modeled by 3-D beam
elements connected to the cables through transversely extended rigid beams. In the
21
study, a comparison was also made with the design value of DAF specified in the
Japanese Design Standards for Railway Concrete Structures. Miyazaki et al. presented
diagrams showing the speed, 0-400 km/h, influence on the DAF for the deck and
pylons bending moments, deck and pylons shear forces, deck and pylons axial forces, and
axial forces in cables. For the vehicle, diagrams were presented showing wheel load variations
and vertical car body accelerations. Miyazaki et al. concluded that the examined PC cable-stayed
bridge had a satisfactory train running quality (acceptable riding comfort). For the different
bridge members, the authors recommended different values for the coefficient included in the
DAF expression in the Japanese design
standard.
Chatterjee, Datta and Surana (1994) [14] presented a continuum approach for
analyzing the dynamic response of cable-stayed bridges. The effects of the pylons
flexibility, coupling of the vertical and torsional motion of the bridge deck due to
eccentric vehicle movement, and the roughness of the bridge surface, were considered.
The vehicle was simulated using a vehicle model with 3 DOF and 3-axles. A PSD
function was used to generate the road surface roughness and mode superposition
technique was adopted for solving the equation of motion of the bridge. Chatterjee et al.
investigated the influence of vehicle speed, eccentrically placed vehicle, spacing between first
and second vehicle axles, and bridge damping ratios on the dynamic
behavior of a double-plane harp type cable-stayed bridge with roller type cable-pylon
connections and a main span of 335 m. Chatterjee et al. concluded that pylon rigidity and the
nature of cable-pylon connection have significant effect on the natural
frequencies of vertical vibration, but no effect on those of torsional vibration.
Chatterjee et al. noted that idealizing the vehicle as a constant force leads to
overestimation of the DAF compared to the sprung mass model. The same conclusion
was found when assuming that there is no eccentricity in the vehicle path. And finely, it
was noted that increasing the axle spacing of the vehicle, or not including the roughness of the
bridge surface, decreases the DAF values.
1.2.2 Research on other bridge types
The dynamic effects of moving vehicles on bridges have been investigated by various
researchers, using bridge and vehicle models of varying degrees of sophistication.
22
A review of the early work on the dynamic response of structures under moving loads
was presented in the paper by Filho [20]. For a thorough treatment of the analytical methods
used for problems of moving loads with and without mass in both structures and solids, see the
excellent book by Frba [23]. In this book, analysis of sprung and unsprung mass systems
moving along a beam covered with elastic layer of variable
stiffness and surface irregularities, were presented. The dynamics of railway bridges and
railway vehicle modeling are described in the book by Frba [24] and the book by Garg and
Dukkipati [25]. Interesting research was also presented by Olsson, see Table 1.2, where he
derived a structure-vehicle finite element by eliminating the contact degrees of freedom of the
vehicle. The stiffness and damping matrices thus became
time-variant and non-symmetric.
Previous investigations on the dynamic response of other bridge types subjected to
moving loads are summarized in Table 1.2 below.
Author(s) Bridge
type Vehicle
model Surface Other remarks like
roughness
analysis methods used etc.
function
Hillerborg (1951) SSB
SMS-1-1-2
not considered
theoretical & experimental
[34]
study
Hirai et al. (1967) suspension
MF, moving pulsating
not considered
[36]
bridge
force
study
Veletsos et al.
3-SB cantilever SMS-3-3-2
not considered
lumped mass method
(1970) [77]
, SSB
Yoshida et al.
SSB, SS slab
MF, MM
not considered
FEM
(1971) [93]
Nagaraju et al.
3-SB
MF, SMS-1-1-2
not considered
continuum approach, mode
(1973) [57]
cantilever
superposition
Ting et al. (1974)
SSB
MM
not considered
structural impedance
[72]
method
Table 1.2
Previous investigations on the dynamic response of other bridge types
subjected to moving loads. SMS-x-y-z=sprung mass system with x-axles, y
degrees of freedom, and in z dimensions, MF=moving force, MM=moving
mass, SSB=simply supported beam, x-SB=x span beam, SS xx=simply
supported xx, FEM=finite element method
23
Genin et al.
SSB,
MF, SMS-1-1-2,
harmonic
(1975) [26]
2-SB
air cushion system
sinusoidal
method
Ginsberg (1976)
SSB multiple
not considered
[29]
SMS-1-1-2
method
Filho (1978) [20]
SSB
SMS-1-2-2
not considered
FEM
Blejwas et al.
SSB MM,
SMS-1-2-2
harmonic
Lagranges eqn. with
(1979) [7]
sinusoidal
multipliers
Chu et al. (1979)
SS girder &
SMS-4-3-3 for
not considered
lumped mass method
[16]
truss railway
each railcar
Gupta et al.
SS orthotr.
SMS-2-3-2 not
considered
vehicle braking, eccentric
(1980) [31]
plate, SSB
loading
Ting et al.
SSB
MF, MM,
not considered
review, different analysis
(1980,1983)
SMS-1-2-2
procedures and vehicle
[73, 74]
models
Hayashikawa et
SSB, 2-SB,
MF
not considered
eigen stiffness matrix
al. (1981) [32]
3-SB
method
Hayashikawa et
suspension
MF
not considered
al. (1982) [33]

bridge
superposition
Mulcahy (1983)
SS orthotr.
SMS-2-4-3,
10 mm bump
finite strip method, vehicle
[56]
plate
SMS-3-7-3
braking
Olsson (1983,
SSB
MF, MM,
harmonic
FEM, special bridge1985) [63, 62]
SMS-1-2-2
cosine
vehicle element
Schneider et al.
SSB
MF, MM
not considered
used the FEM package
(1983) [71]
ADINA
Arpe (1984)
SSB
SMS-2-4-2
not considered
[4, 5]
study
Hino et al. (1984) 1-SB cantilever SMS-1-1-2
not considered
FEM, direct time
[35]
integration
Palamas et al.
SSB, 2-SB
SMS-1-1-2
sinusoidal,
Rayleigh-Ritz method
(1985) [65]
pothole
Chu et al. (1986)
SS PC railway
SMS-4-23-3
PSD
lumped mass method
[17]
Honda et al.
2-SB, 3-SB, 4SMS-1-2-2
PSD, bump at
1 vehicle & multiple
(1986) [37]
SB, 5-SB, SSB
entrance
groups of vehicles
Table 1.2
(continued)
Olsson (1986)
SSB, 2-SB,
MF, MM, SMS-1-2-2,
not considered
FEM, special bridge 24
[64] 6-SB
SMS-2-4-2,
vehicle element, vehicle
SMS-2-6-2, SMS-2-7-2
braking
Inbanathan et al.
SSB
MF, MM
considered
FEM, PSD for interaction
(1987) [44]
force
Bryja et al.
suspension
multiple MF
not considered
random highway traffic
(1988) [10]
bridge
Diana et al.
suspension
SMS-4-23-3 for each
not considered
FEM, different traffic
(1988) [19]
bridge
railcar
conditions
Coussy et al.
SSB
SMS-2-2-2
PSD
(1989) [18]
superposition
Wang (1990) [81] SS PC railway
SMS-4-23-3 for each
PSD
influence of ramp/ bridge
railcar
track stiffness
Hwang et al.
SSB SMS-2-4-2, PSD
traffic simulations, one and
(1991) [43]
SMS-3-7-2
two trucks
Olsson (1991)
SSB
MF
not considered
compared analytical
[61]
solution with FEM
Wang et al.
SS truss
SMS-4-23-3 for each
PSD
lumped mass method
(1991) [82]
railway
railcar
Huang et al.
continuous
SMS-3-12-3
PSD
FEM, one and two trucks

(1992) [39]
multigirder
Wang et al.
SS multigirder SMS-2-7-3,
PSD
(1992) [85]
SMS-3-12-3
Wang et al.
SSB SMS-2-7-3, bump, PSD
FEM, validation of vehicle
(1992) [83]
SMS-3-12-3
models
Knothe et al.
review of dynamic modeling of railway track and of vehicle-track interaction
(1993) [50]
Nielsen (1993)
beam on elastic MM, SMS-1-3-2,
harmonic sinusrailway structures,
[58]
foundation,3-D SMS-2-4-2, SMS-2-6-2 oidal for railcompared theoretical and
track model
head, wheelflat
experimental results
Saadeghvaziri
SSB,
MF
not considered
used the FEM package
(1993) [70]
3-SB
ADINA
Wang et al.
no bridge
SMS-2-7-3,
bump, PSD
only validation of the
(1993) [86]
SMS-3-12-3
vehicle models
Table 1.2
(continued)
Wang (1993) [87] SS truss

SMS-4-23-3 for each
PSD
lumped mass method
railway
railcar
25
Cai et al. (1994)
SSB, 2-SB
moving pulsating force, not considered
[12]
SMS-1-2-2
superposition
Chatterjee et al.
suspension
SMS-1-1-2,
PSD
(1994) [15]
bridge
SMS-3-3-2, SMS-3-6-3
superposition
Wakui et al.
describes a computer program developed using FEM and mode superposition to solve the
(1994) [78]
dynamic interaction problem between high speed railway vehicles, each of SMS-4-31-3,
and railway structures
Yener et al.
slab on SSBs
MF, SMS-1-3-2,
not considered
(1994) [92]
SMS-2-6-2
conditions
Chatterjee et al.
arch bridge
MF
not considered
mixed and lumped mass
(1995) [13]
method
Green et al.
3-SB, 4-SB
SMS-4-11-2
PSD, 20 mm
compared leaf sprung with
(1995) [30]
bump
air sprung vehicles
Huang et al.
thin walled
SMS-3-12-3 PSD FEM
(1995) [40]
box-girder
Huang et al.
hor. curved
SMS-3-12-3
PSD
(1995) [41]
I-girder
Humar et al.
SS orthotr.
SMS-1-2-2
not considered
(1995) [42]
plate
conditions
Lee (1995) [51]
2-SB, 3-SB,
MF
not considered
beams on one-sided point
4-SB
constraints
Lee (1995) [52]
SSB
rigid wheel
not considered
unknown wheel nominal
motion, FEM
Paultre et al.
arch, box
ambient & controlled
dynamic bridge testing
(1995) [67]
girder
traffic
Yang et al. (1995) SSB, 3-SB,
MF, MM, SMS-1-2-2,
PSD
FEM, special bridge[90, 91]
5-SB
SMS-3-6-2
vehicle element
Table 1.2
(continued)
1.3
General aims of the present study
In all the aforementioned studies on the dynamic behavior of cable-stayed bridges,
authors either used very simple vehicle models, or very complicated and time 26
consuming vehicle and bridge models. In [21, 22, 89], the vehicle was modeled as a
constant moving force, neglecting the vehicle inertial effects, and in [69, 3, 8, 55], the road (or
rail) surface roughness was neglected and only the elastic displacements of the bridge, caused by
the varying position of the vehicle, were considered. The
opposite assumption was made in [45], where the bridge elastic displacements were
neglected and only the excitation caused by the road surface roughness was
considered. Of course, the assumptions made by those authors are acceptable, if for
example the vehicle is travelling at low speed, the road surface is smooth, and the vehicle mass
to the bridge mass ratio is low.
The vehicle inertial effects, the road surface roughness, and the bridge displacements
were considered in [49]. However, the formulations for the coupling equations
(equations (2.4a-c) in section 2.1) are, according to the authors opinion, incorrect.
Only the models developed in [84, 14] are believed to be general and handle the
bridge-vehicle contact problem correctly. On the other hand, the vehicle models used are
very complicated and, as Frba [24] pointed out, very detailed and complicated vehicle models
are unnecessary, if the main purpose is to study the bridge dynamic response. In the work
presented here, the most detailed vehicle model used consists of two degrees of freedom, as this
is adequate for large span bridges, according to Frba.
The main aims of this study are as follows:
to develop a general but simple analysis tool which fully consider the bridgevehicle interaction, including all inertial terms, in evaluating the dynamic response of
bridges subjected to moving vehicles

to investigate on the applicability of the beam on elastic bed analogy and the finite
difference method for dynamic analysis of cable-stayed bridges. Moreover, to
show that the proposed simplified analysis method, which uses the finite difference
method and the mode superposition technique for dynamic response evaluation, is
very efficient and is easy to implement and understand
to analyze the dynamic response of simple cable-stayed bridge models and to study the
influence of different vehicle models and the influence of different parameters, such as vehicle
speed and bridge deck surface roughness, on the dynamic response.
27
For this purpose a computer code has been developed using the MATLAB language
[53], where the dynamic interaction between the bridge and the vehicle is included by
utilizing an iterative scheme. Time histories and dynamic amplification factors are presented as
functions of a limited set of parameters for quite simple but representative bridge and vehicle
models. The implemented code has been verified by comparing
analysis results with those obtained using the commercial finite element code
ABAQUS [1]. Special emphasis is put on verification of the proposed model and on
investigating the effects of local and global irregularities on the dynamic response.
Part of this work was presented earlier at the 15th Congress of IABSE, Copenhagen,
1996 [48].
28
Chapter
______________________________________________________________________
Vehicle and Structure Modeling
_____________________________________________________________________
_
2.1 Vehicle
models
Heavy vehicles consist of several major components, such as tractors, trailers and
suspension systems, and can be modeled by a set of lumped masses, springs and
dampers. As illustrated in Figure 2.1, the vehicle models used in this study include a
moving force model, a moving mass model, and a sprung mass model with two
degrees of freedom. The moving force model (constant force magnitude) is sufficient if
the inertia forces of the vehicle are much smaller than the dead weight of the
vehicle. For a vehicle moving along a straight path at a constant speed, these inertia
effects are mainly caused by bridge deformations (bridge-vehicle interaction) and
bridge surface irregularities. Hence factors that are believed to contribute in creating
vehicle inertia effects include: high vehicle speed, flexible bridge structure, large vehicle mass,
small bridge mass, stiff vehicle suspension system and large surface irregularities. In the present
study, the adopted sprung mass model is a one-axle
vehicle model of a real multi-axle vehicle. This model is acceptable, when the bridge
span is considerably larger than the vehicle axle base [24], as the case is for cable supported
bridges. The author believes that the use of simplified models may be more effective in
identifying correlation between the governing bridge-vehicle interaction parameters and the
bridge response. Very detailed vehicle models are unnecessary and will not bring any great
advantage, when the main purpose is to study the dynamic
response of bridges.
Heavy roadway vehicles generate most of their dynamic wheel loads in two distinct
frequency ranges [30]: body-bounce and pitch motions at 1.5-4 Hz and wheel-hop
29
motion at 8-15 Hz. This explains the increase of some of the specified DAF in Figure 1.1,
for bridges with a fundamental frequency in the range of 1 to 5 Hz.
v( t)
w 2( t)
m2
v( t)
v( t)
w 1( t)
w 1( t)
(m
k
c
S
S
1+ m 2) g
m1
m 1+ m 2
Moving force model Moving mass model Sprung mass model Figure 2.1
Vehicle modeling
Considering the sprung mass model, shown in Figure 2.1, and denoting the contact
force between the bridge and the vehicle by F( t) , defined positive when it acts
downward on the bridge, the following equations of motion can be established
[23, 63]:
2
(
dw
dw
dw
m+m
1
2
1
1
2)gm
+k
1
(ww)+
c
F( t
S
S
) 0 (2.1)
dt 2
2
1
d t
d t +
=
d2 2
w
w
w
m
kwwc
S
S
=
2
( 2 1)
d
d
2
1
0 (2.2)
d2
t
dt
dt
Equations (2.1) and (2.2) are the dynamic equilibrium equations for the unsprung mass
and the sprung mass, respectively. Referring to Figure 2.1, w ( t
1 ) and w ( t
2
) are the
displacements of the vehicle unsprung mass m 1 and the vehicle sprung mass m 2,
respectively, kS the stiffness of the linear spring connecting the two masses, cS the damping
coefficient of the viscous damper, and g the acceleration of gravity. It should be noted that w ( t
2
) is measured from the equilibrium position under the dead weight
m 2 g. The contact force may be expressed by use of equations (2.1) and (2.2) giving:
30
2
dw
2
dw
F( t) = ( m + m
1
2
1
2)g+m
+m
1
(2.3)
dt 2
2dt2
where the first term on the right-hand side is the dead weight (static part) of the contact
force and the other terms represent the inertia effects.
d2 w
The contact force for the moving mass model will be F( t) = ( m + m ) g +
1
1
2
,
dt 2
and for the moving force model F = ( m + m ) g
1
2
.
Assuming that the vehicle never loses contact with the bridge (that is F( t) > 0), and that
the deformation between the unsprung mass center and the bridge deck center line may be
neglected, the following coupling equations for the point of contact,
x ( t) = x ( t
v
) (see Figure 2.2), must be fulfilled [20, 63]:
w t = y( x t ,t ) + r( x t
1( )
()
( )) (2.4a)
y
y
r
& w ( t) =
v+
+
v
1
(2.4b)
x
2
y
2
2
2
2
y
y
yr 2r
& w ( t) =
v+
2v+
a+
+
v+
a
1
(2.4c)
x2
x t
x
t2
x2
x
where & w ( t)
1
and & w ( t) denote the unsprung mass vertical velocity and acceleration, 1
respectively, v and a the vehicle velocity and acceleration in the longitudinal direction,
respectively, y ( x , t ) the bridge vertical displacement, and r( x) the surface irregularity function.
The first term on the right-hand side of equation (2.4c) represent the influence of the bridge deck
curvature (centripetal acceleration), the second term the influence of Coriolis acceleration, and
the fourth term the influence of the
acceleration of the point of contact in the vertical direction.

2.2 Bridge
structure
For the present study, the fan-shaped self or earth anchored cable-stayed bridge
scheme shown in Figure 2.2 is adopted. To make the presentation of the model more
31
clear, the derivation of the equations in this section will be presented including only the
main span of the bridge as shown in Figure 2.2, and assuming that the stiffening girder, having a
uniform mass and flexural rigidity, is simply supported at the pylons.
Of course the developed computer code is very general and capable of handling the
more realistic case including side spans, suspended or not suspended, and as many
supports as needed.
Figure 2.2
Idealized vehicle in contact with a cable-stayed bridge
2.2.1 Major
assumptions
The following assumptions are made:
multicable system with small stays spacing compared to the bridge length
negligible cable mass
the cables are idealized as vertical springs continuously distributed along the length of
the stiffening girder
according to the usual erection procedures, the bridge in its initial configuration under
dead load is free from bending moments, while only axial forces are present
32
cable forces under dead load are so adjusted that all displacements remain zero
axial girder forces have negligible effect on the frequencies and mode shapes and are
therefore neglected
only in-plane flexural behavior of the bridge is considered. The torsional behavior
caused by eccentric loading of the bridge deck is disregarded in this study
bridge damping is small and therefore neglected
when the vehicle enters the bridge, the vertical deflection and the vertical velocity of the
moving vehicle are assumed to be zero.
2.2.2 Differential equation of motion
The governing equation of motion for vertical vibration of the bridge at any section of the
stiffening girder (idealized as a Bernoulli-Euler beam on elastic supports) is given by [23]:
4 y( x,t)
2 y x,t
EI
+ k( x) y( x,t)
()
+m
=
g
( x xv) F( t
gg
) (2.5)
x
4
t 2
where is the Dirac delta function, Eg the modulus of elasticity, I g the moment of
inertia, mg the mass per unit length, and k( x) the spring stiffness (to study ordinary beam type
bridges k( x) is set to zero). The effects of rotatory inertia and shear deformation are neglected as
the cross-sectional dimensions of the stiffening girder are small in comparison with its length and
the higher vibration modes are not
significantly excited.
The boundary conditions are:
2
2
(
0,
,
0, )
y( t)
yLt
= 0,
= 0, y( L, t)
()
y
t
= 0,
= 0 (2.6a-d)
x2
x2
and the initial conditions are:
y(
yx
x,0)
( ,0)
=
,
0
= 0 (2.7a,b)
t
33
2.2.3 Spring
stiffness
Using the notations of Figure 2.2, the stiffness of the spring idealizing cable i is given by
[75]:
c
EA
k
ii
i
=
sin2
i
(2.8)
Li
Denoting the allowable cable stress by a , the dead load and the live load per unit length
by q and q
g
q , the cross-sectional area of cable i is given by [9]:
(q+q
g
q)s
A=
i
(2.9)
sin
a
i
Due to its own dead weight, a stay cable actually takes the shape of a curved line, rather
than a straight one, between the two anchorage points. When the cable tension increases, the sag
decreases, and the apparent axial stiffness of the inclined cable increases. In the present study,
the cable geometric nonlinearity, due to the change of the sag and shape under varying stresses
(forces), is approximately taken into account by introducing the following equivalent tangent
modulus of elasticity [27, 9, 84]: E
L
E( x) = E
c
=
(2.10)
22
0x
i
x
2
c
+
1
E
3
c
12o
where Ec is the modulus of elasticity for the straight cable, c the specific weight of the
cable material, and o the initial tensile stress in the cable. As cable forces caused by the vehicle
load are small when compared to those created by dead load, the
starting equlibrium configuration under dead load is used [9, 84, 14] and o is here set
equal to g , which is the tension stress due to dead load qg and is given by [9]: q
=
g
g
a
(2.11)
(q+
g
qq )
34
After substituting (2.9) into (2.8), the following equation can be established for the spring
stiffness per unit length due to the elongation of the cables in the main span: E( x)
+
c
(qq
g
q)
1
L
k ( x) =
(2.12)
2
0x

H
a
x
2
1 +
H
The horizontal force on the pylon top due to the tensile force Fi in the main span cable i
is:
(q+q
+
g
q)s
(qq
g
q)s
T = F cos =
=
x
i
i
i
(2.13)
tan
H
i
Neglecting the stiffness of the pylon, the area of the anchor cable per unit length of the
main span and the total area of each anchor cable can be expressed as:
q+q
A
=
o( x)
g
q
x (2.14)
H cos
a
o
L/2
q+q
2
tot
g
q
L
A
=
A
d=
o
o( x) x
(2.15)
H
8
0
a
cos o
In equation (2.15), the dead weight of the side spans are not included because the side
spans are not considered in this derivation.
The elongation of the anchor cable, the horizontal displacement of the pylon top, and the
vertical displacement in the main span at joint i, due the force Fi in cable i are:
FL
L
i o
i
=
cos
o
(2.16)
E A cos
oo
o
L
FL
cos
b
o
i o
i
=
=
(2.17)
2
coso
EA
o o cos
o
35
x
=
i
b
(2.18)
H
The internal force in cable i due to the vertical displacement =
i
1, and the spring
stiffness are:
2
( = )
1
E A H cos 1
Fi
=oo
o
i
(2.19)
L cos
x
o
i
( i = )
1
2
2
a
F
sin
E A H cos
1
k
i
i
=
=oo
o
i
(2.20)
1
L
2
o
x
After substituting (2.14) into (2.20), the following equation can be established for the
spring stiffness per unit length due to the elongation of the anchor cables:
ELH
+
a
os
(qq
g
q)1
L
k ( x) =
0 x (2.21)
L2
x
2
a o
Where E o is evaluated according to equation (2.10).
Referring now to the pylons, a horizontal force at the top gives the horizontal
displacement:
TH3
b
i
p
=
(2.22)
3EI
pp
Thinking of the pylon as a fictitious anchor cable having the fictitious area Ap , the force
Ti gives the horizontal displacement:
L
TL
b
o
i
=
=
o
(2.23)
cos
2
o
EA

o
p cos
o
Equating both equations (2.22) and (2.23) for b, we find that the fictitious cable area is:
36
3EIL
A
pp
=
o
p
(2.24)
EH3
2
o
p cos o
To approximately include the effect of the pylons, equation (2.21) is modified to give: E
LH
+
tot
o
+
a
s
(qq
g
q)A
A
o
p 1
L
k ( x) =
0x
(2.25)
L2
A tot
x
2
a o
o
And finally, the resulting spring stiffness that includes the effect of the elongation of the
cables in the main span, the elongation of the anchor cables, and the stiffness of the pylons is:
1

L
k( x) =
0x
(2.26)
1+1
2
k c( x) ka( x)
The pylons shortening effect is very small and is therefore not considered in equation
(2.26).
Figure 2.3 shows the spring stiffness, k( x), for a fan-shaped cable-stayed bridge with
L=150
m, and Figure 2.4 shows the fixed pylon top curve in Figure 2.3 but with the suspended
side spans included.
2.0
)2
fixed pylon top
N/m
free pylon top
6
1.5
free pylon top, I p =0
) (10( x k 1.0
0.5
Spring stiffness,
0.0
0
25
50
75
100
125
150
x (m)
37
Figure 2.3
Spring stiffness for a fan-shaped cable-stayed bridge with L=150 m, H=30
m, Hp=60 m, Ec =2.11011 N/m2, Ep=0.351011 N/m2, Ip=25 m4, a =720106
N/m2, qg=12104 N/m, qq=5104 N/m, and c =9104 N/m3
2.0
)2
N/m6 1.5
) (10( xk 1.0
0.5
Spring stiffness,
0.0
0
50
100
150
200
250
x (m)
Figure 2.4
Spring stiffness for a fan-shaped cable-stayed bridge (fixed pylon top) with
the same input data as in Figure 2.3 but with the side spans, Ls=50 m,
included
Equations (2.1-2.5) and (2.26), together with the boundary and initial conditions, define
the analytical model of the problem.
2.3
Bridge deck surface roughness
The characteristics of road surface roughness is expressed generally by the power
spectral density (PSD) function which is assumed from a stationary normal probability
process with a zero mean value [38]. The PSD, S() , of road surface roughness is approximated
generally by an exponential function [38, 46] as:
S( ) = a n
(2.27)
in which is the wave number, a is the spectral roughness coefficient, and n is the
spectral roughness exponent.
38
In the present study, the bridge road surface roughness is simulated using the onesided
PSD function:
35.8
10 4 0 < < 0.05
S() =
(2.28)
00
. 107
10 4 1 9.4 0.05 <
<1
presented in [38], and shown in log-log by the bold line in Figure 2.5. The exponent n
and the coefficient a were calculated by the least square method using measurements on 56
bridges.
The random surface roughness profile is assumed to be the sum of series of sinusoidal
waves and is generated by the following formula [49]:
M
r( x) = si (
nx+
i
i
i ) (2.29)
i=1
where i is the amplitude of the sinusoidal wave, i is a random phase angle with
uniform distribution in the interval {0, 2 }
, and i is the i th wave number within the
specified PSD interval and is given by:
=
+
i
min
( i )1 (2.30)
in which the wave number increment is defined as:
= max
min (2.31)
M
where M is the total number of wave number increments in the interval between the
minimum wave number min and the maximum wave number max in the defined
spectra. The unit of the wave number is (m-1) and the unit of the spectrum is (m3). The
amplitude
i is related to the PSD function S(

i ) by:
2
S(
i
=
i)
(2.32)
2
39
10
)3
Analytical
m-4
Simulated
1
) (10
0.1
0.01
PSD of surface roughness, S(
0.001
0.01
0.1
1
Wave number, (1/m)
Figure 2.5
Power spectral density of road surface roughness
0.02
0.01
0
-0.01
-0.02
Surface roughness (m)
0
50
100
150
Distance along the bridge (m)
Figure 2.6
Simulated road surface roughness profile for a bridge with L=150 m
40
The simulated road surface roughness profiles are different depending on the random
phase angles i , used in equation (2.29). Figure 2.6 shows a sample of the simulated random
road surface roughness profile used in this study for bridges with L=150 m.
To compare the analytical PSD, equation (2.28), with the PSD of the sample shown in
Figure 2.6, a power spectrum analysis of the sample is performed using the following
relationship [83]:
1
S
() =
F 2
sim
(2.33)
N
where F is the length N fast Fourier transform (FFT) of the sample, and S
sim (
) is the
resulting PSD estimate for the sample. Using the symmetry property of real FFTs, the
spectral estimates corresponding to negative wave numbers are removed and a
compensation is made for them in the positive side. The resulting one-sided PSD
estimate for the sample in Figure 2.6 is indicated by the fine solid line in Figure 2.5.
The bridge-vehicle interaction forces can now be calculated including the effect of road
surface roughness by using the simulated profile of the bridge road.
41
42
Chapter
______________________________________________________________________
Response Analysis
_____________________________________________________________________
_
3.1 Dynamic
analysis
3.1.1 Eigenmode
extraction
Before analyzing the response of the bridge to moving vehicles it is important to study
the free vibration of the bridge model. The free vibration analysis is here an essential first step in
obtaining the forced vibration response of the bridge and the parameters that affect the free
vibration also affect the response to moving vehicles.
The differential equation governing the free vibration of the bridge model is obtained by
setting the right-hand side in equation (2.5) equal to zero, giving:
4 y( x,t)
2 y x,t
+k
+
=
gg
( x) y( x,t)
()
EI
m
0 (3.1)
x4
g
t
2
The displacement function y( x, t) can be expressed as a product of two functions, one
involving only the space coordinate x, called the mode shape function or eigenfunction z( x) , and
the other one involving the variable time and called the generalized coordinates (
t) . Then when the bridge vibrates in its i th natural bending mode, the displacement at
any location varies harmonically with time and can be expressed as
[88]:
y ( x, t) = z ( x) ( t) = z ( x)( a cos t + b sin t i
i
i
i
i
i
i
i ) (3.2)
43
where i is the i th circular frequency of the free motion of the bridge. By expressing
as in equation (3.2) and substituting into equation (3.1) gives:
i(t)
d 4 zi( x)
EI
+ k( x) z
2
= 0 (3.3)
4
i ( x)
m
z
g
ii(x
gg
)
dx
Dividing the bridge model into n equal segments each of length h, giving n-1
unknowns, and replacing the derivative in equation (3.3) with its finite difference
approximation, the following eigenvalue problem is obtained2:
A Z = Z (3.4)
where
c
4 1
0
1
4 c
2
4
1
1 4 c 4 1
A=
OOOOO
,=
O
(3.5a,b)
1
4 c
3
4
1
n1( n1 x n )
4 c
4
0
1
4 c 1 ( n1 x n )1
z,
L
z
11
1 ,n1
Z = [z , z , L, z
1
2
n1 ] = M
M
(3.5c)
zn11 , L zn1 ,n1( n1 x n )1
4
g
mh
k
2
( x 1)
k
4
( x 2 , 3 , K)
=
,
=5+
and
=6+
4
i
i
c
h
c
1
2,
h (3.5d,f)
E
3,K
g Ig
EgIg
EgIg
where i is the mode number and column i in Z gives the i th mode shape. The first term
in equation (3.5e) for c 1 must be changed from 5 to 7, if fixed end supports are assumed. If
rotational springs with stiffness Km are introduced at end supports, the first term mentioned
above is evaluated as: 6+(-1)/(+1), where = K h/(2 E I
m
g g).
2 If nothing else is mentioned, bold upper case letters are used for matrices and bold
lower case letters for vectors. The transpose of a matrix or a vector is denoted with the
superscript T.
44
Mode shapes and corresponding circular frequencies are now obtained numerically by
solving equation (3.4).
3.1.2 Response of the bridge
Using mode superposition technique, the solution of equation (2.5) can be written as:
y( x, t) = z ( x) i ( t
i
) (3.6)
i=1
Substituting equation (3.6) into equation (2.5) gives:
( E I z IV + k( x) z + m z & ) = ( x xv) F
ggi
i
i
i
gi
i
(3.7)
i=1
Further, both sides of the above equation are multiplied with eigenmode z j and
integrated3 over the length of the bridge. Using the orthogonality properties of the
eigenfunctions and setting the normalization constant4 equal to unity, the following equation is
obtained (see ref. [88] for more details on derivation):
2
F
& + =
i
i
i
zi ( xv ) (3.8)
mg
where the contact force, F, is time depended and therefore evaluated at each time step
using the expressions given in section 2.1.
When dividing the bridge model into n equal segments each of length h and setting the
time step to t
= h / v , which corresponds to having the vehicle load applied only at the segment
joints, the term z ( x
i
v ) in equation (3.8) will be the value in the i th
eigenvector z i and the row number corresponding to the point of contact (i.e. where the
contact force is applied).
L
3 The properties of the Dirac function gives
( x x ) z ( x) d x= zi( x
v
i
v)
0
L
4
2
Normalization constant is defined as z
( x) d x
T
i
or in vector form z
z
i
i
0
45
The solution of the simple second order differential equation in (3.8) is:
Fz
+ 2
o
&
i ( xv )
( Fzi( xv)
m
i
i
g)
=
+
cos t
oi
+
sin t
i
(3.9)
2 m
2 m
i
i
i
g
i
g
i
where and &
o i o i are the initial values for the step and can, when using the orthogonality and
normalization relationships, be evaluated as [88]:
T
=
oi
i
z yo (3.10a)
&
T
=
oi
i
z & yo (3.10b)
where y and y
o
& o are the bridge vertical displacement and velocity vectors at the
previous time step.
Substituting into equation (3.6), the bridge displacement vector at each time step is
obtained from the sum:

s
Fz x
Fzx
2m
&
i(v)
( i( v)+ i
o
i
g)
y=
io
z
+
cos t +
sin
i
i
i
t (3.11)
2
2
i 1
m
i
g
i
g
i
where s is the number of included modes, s n 1. The expressions for the vertical
velocity and acceleration vectors for the bridge are obtained, by differentiating
equation (3.11), as:
s
( Fz
2
i ( xv ) +
oi
i mg )
&y=
z
&
sin t + cos
i
i
oi
i
t (3.12)
i 1
i mg
s
( Fz
2
i ( xv ) +
i
o
i mg )
&y=
z
&
cos t sin
i
i
i
o
i
i
t (3.13)
m
i=1
46
Using the MATLAB language [53], a computer code has been developed for analysis of
the dynamic response. The dynamic interaction between the bridge and the vehicle is included by
utilizing the iterative scheme shown in Figure 3.1.
The recalculation of bridge displacements is repeated in each step until convergence is
obtained. In this study, the following convergence criterion is considered:
max( y y
j
j1 ) Tolerance (3.14)
max( y j )
where y j is the j th estimate for the bridge displacement vector at the current time step.
For the results presented in this study, the tolerance was set to 5 10 6
.
The mode superposition technique is also used to obtain the dynamic bending moment
vector at each time step, giving:
s
m=E
y=
z
gIg
Eg Ig
i i (3.15)
i=1
where
z i is evaluated using the finite difference approximation.
The obtained displacement, moment, velocity, and acceleration vectors are now
collected in matrices for processing and plotting.
47
Simulation
start
Evaluate the natural frequencies and the
corresponding vibration modes of the bridge
Determine the longitudinal position of the
vehicle on the bridge, x = x + t
v
vv
Based on previously determined quantities, set
initial values for the vehicle sprung mass and
for the unsprung mass predict
1
w , &1
w , and 1
w&
Solve the equations of motion of the vehicle
and calculate the interaction force F( t)
Move vehicle to
next position
Solve equ. (3.11-3.13) and (3.15) to determine
Calculate new
the bridge response using mode superposition
1
w , &1
w , and 1
w&
technique
Check
No
convergence
Yes
Vehicle leaves the right
No
end of the bridge
Yes
Calculate static displacements and bending
moments using equ. (3.16) and (3.18)
End
Figure 3.1
Overall computational scheme
48
3.2 Static
analysis
Using the same discretization technique as in section 3.1, the static displacement vector is
obtained from the expression:
+
3
st
(mm
1
2)gh
y=
A1 p
j
j (3.16)
Eg Ig
where A1is the inverse of the matrix given in equation (3.5a) and:
M
0
j th row
p = 1
j
(3.17)
0
M
0
where j is the joint number where the vehicle load is applied. The obtained yst j is a
column vector containing the bridge static displacements corresponding to a loading case where
the vehicle static load is applied at joint number j.
The static bending moment is computed using the expression given in equation (3.15).
Approximating the second derivative of the displacement by its finite difference, gives
the i th element in the static moment vector st
m j to be equal to:
EI
st
gg
=
i
m,j
( st
st
st
yi+
+
,
1
2
j
yi, j yi ,1 j ) (3.18)
2
h
49
50
Chapter
______________________________________________________________________
Numerical Examples and Model Verifications
______________________________________________________________________
4.1 General
In this chapter, two beam examples and two cable-stayed bridge examples are studied and
results obtained using the present model, described in chapter 2 and 3, are
presented. For verification, the present model results are compared with those obtained
using a commercial finite element code. The finite element code ABAQUS/Standard 5.4 [1] was
used for this comparison and was run on a SUN SPARCstation 20. This code is considered as
one of the most powerful finite element codes available on the market today . In ABAQUS, 2-D
geometrically nonlinear analyses were performed starting from the dead load configuration and
direct time integration of the dynamic response was selected with a fixed time increment. The
time increment was set equal to the total time needed for the vehicle to cross the bridge divided
by the number of increments. The number of increments was chosen for each numerical example
so
convergent results were obtained.
From ABAQUS element library, a T2D2 (2-node 2-D linear) truss element was selected to
model the cables and a B23 (2-node 2-D cubic Euler-Bernoulli) beam
element for the rest. The number of elements, in each numerical example, was
increased until a converged solution was obtained. For each cable in the cable-stayed
bridge examples, the equivalent tangent modulus of elasticity, calculated according to equation
(2.10), and the cable density =
c
9000kg / m3 , were given in the ABAQUS
input files. The vehicle was modeled using SPRING2, DASHPOT2 and MASS
elements.
For the interaction problem the SLIDE LINE option was used to define the surface on
which the vehicle and the bridge may interact, and the NO SEPARATION parameter
51
was included in the SURFACE BEHAVIOR option. At the node where the vehicle
and the bridge interact, a special purpose contact element ITT21 was used. As the
position of this node is time dependent, the AMPLITUDE and the BOUNDARY
options were used to modify this position for each increment.
The modeling approach used in ABAQUS for the bridge-vehicle contact problem,
described above, is the simplest one, but still very complicated and requires a lot of time and
experience in ABAQUS and in finite element modeling. Other modeling approaches, e.g.,
developing user subroutines in FORTRAN and including them in the input file for ABAQUS or
using another type of contact definition, were tested for simple problems. The results obtained
from these approaches agreed well with those from the above described modeling procedure but
the implementation became much
more complex. Therefore, results using these more complex approaches are not
presented here, and all the ABAQUS curves presented in this study are those obtained
using the SLIDE LINE and the ITT21 contact element approach described above.
The program developed for the present model, using the MATLAB language [53], was
also run on the same SUN SPARCstation 20 machine. This made it possible to not only compare
the obtained solutions but also the time consumed by the computer. In
MATLAB the CPUTIME command was used to measure the CPU time needed to run
the program. This value was then compared with the USER TIME value given at the
end of the ABAQUS output file. For the present model solutions , the number of modes
that participate in the analysis and need to be considered and the number of segments, n, which
also determines the time step used for the analysis, were found using a trial and error procedure.
In the following examples, dynamic amplification factors5 for displacements (DAF d)
and for bending moments (DAF m), are presented. The definition for DAF d and DAF m adopted
in this study is the ratio of the absolute maximum live load dynamic response to the absolute
maximum live load static response.
4.2
Simply supported bridge, moving force model
The simply supported bridge subjected to a constant force, F, moving at constant speed,
v, is the easiest problem to tackle and belongs to the very few moving load problems that can be
solved analytically.
5 In the literature, the impact factor defined as I=100*(DAF-1) (%) is found sometimes
instead.
52
Adopting the moving force model means that the influence of the inertia of the vehicle
mass and the bridge-vehicle interaction are neglected. The dynamic effects are then caused only
by the varying position of the force. Still this is a good approximation, see
[64], for low values of the speed parameter (defined later in equation (4.4)) and low
values of the vehicle mass to bridge mass ratio.
v = 68.1 m/s ( =0.25)
y( x, t)
F =347000 N
x
E I =9.92 10
. 10 Nm2
gg
x ( t)
m =11400 kg/m
v
g
L =34 m
Figure 4.1 Simply supported bridge subjected to a constant moving force Using the
notations of Figure 4.1, the analytical solutions (here referred to as the exact solutions) for the
displacement and bending moment are given as [23]:
FL 3
96
1
i x
y( x, t) =
(4.1)
4
sin i
t
sin
t sin
48 E I
i
4
2
2
=1 i (1 / i )
()
L i
L
gg
i
FL
8
1
i x
M(
mx
( x,t)
,t)=
(4.2)
2
sin i
t
sin
t
i
sin
4
2
2
2
=1 i (1 / i )
()
L i
L
i
where i is the mode number, i the circular frequency for the i th mode of vibration, and
the non-dimensional speed parameter. i and are defined as:
2 E I
i
gg
=
i
(4.3)
L
g
m
=
v (4.4)
L
1
53
The problem defined by Figure 4.1 was solved using the exact analytical model,
equations (4.1-4.4), the present model, and the ABAQUS model. The first 20 natural
frequencies, the normalized vertical mid-span displacement, and the normalized mid-span
moment are presented in Table 4.1, Figure 4.2, and Figure 4.3, respectively. One can notice from
Table 4.1 that the ABAQUS model, which uses the finite element method, always gives higher
natural frequency values (stiffer solution) compared to the exact ones, whereas the present
method, which uses the finite difference
approximation, always gives lower values than the exact ones.
Mode
Exact
Present
ABAQUS
number
70
150
200
300
34
70
(bending)
segments segments segments segments elements elements
1
4.01
4.01
4.01
4.01
4.01
4.01
4.01
2
16.03
16.02
16.03
16.03
16.03
16.03
16.03
3
36.08
36.02
36.06
36.07
36.07
36.08
36.08
4
64.13
63.96
64.10
64.11
64.12
64.13
64.13
5
100.21
99.79
100.12
100.16
100.19
100.21
100.21
6
144.30
143.43
144.11
144.19
144.25
144.31
144.30
7
196.41
194.80
196.06
196.21
196.32
196.43
196.41
8
256.53
253.79
255.93
256.20
256.38
256.59
256.54
9
324.68
320.29
323.72
324.14
324.44
324.78
324.68
10
400.83
394.15
399.37
400.01
400.47
401.03
400.85
11
485.01
475.24
482.87
483.80
484.47
485.36
485.03
12
577.20
563.39
574.17
575.50
576.44
577.79
577.24
13
677.41
658.41
673.24
675.06
676.37
678.36
677.47
14
785.64
760.13
780.02
782.48
784.23
787.11
785.72
15
901.88
868.33
894.49
897.71
900.03
904.10
902.01
16
1026.14
982.80
1016.57 1020.75 1023.74
1029.40
1026.30
17
1158.41 1103.30 1146.23 1151.55 1155.36
1163.10
1158.70
18
1298.70 1229.60 1283.40 1290.08 1294.86
1305.20
1299.10
19
1447.01 1361.43 1428.02 1436.30 1442.25
1456.00
1447.50
20
1603.34 1498.54 1580.03 1590.20 1597.49
1615.40
1604.00
Table 4.1 Comparison of the first 20 natural frequencies (Hz) for the problem defined by
Figure 4.1
54
A convergence study was conducted to determine the time step for the exact model,
and the number of segments and elements needed for the present and the ABAQUS
models. For the following numerical investigation, the 150 segments solution was
chosen for the present model and the 70 elements solution with 70 increments for the
ABAQUS model. For the exact model, a time step of t = L/(100 v) = 0.005 seconds was chosen
and the mid-span vertical displacement and the mid-span moment were
obtained from equations (4.1) and (4.2) by introducing x = L/2. For the exact model and
the present model, the first 20 modes of vibration were considered.
To make the results easily understood, all moment and displacement curves presented in
Figure 4.2 and Figure 4.3 are normalized by dividing with the maximum static
values obtained using the three different methods. For the static response, only the results
from the present model are presented, in Figure 4.2 and Figure 4.3, as all static curves from the
two other solution methods coincide.
-0.2
0
Exact
Present
0.2
Static (present)
ABAQUS
0.4
0.6
0.8
1
1.2
Normalized vertical mid-span displacement
1.4
0
0.2
0.4
0.6
0.8
1
Vehicle position, x v / L
Figure 4.2 Vertical displacement at mid-span (normalized with respect to maximum
static displacement) versus vehicle position for the
problem defined by Figure 4.1
55
-0.2
Exact
0
Present
Static (present)
0.2
ABAQUS
0.4
0.6
0.8
Normalized mid-span moment
1
1.2
0
0.2
0.4
0.6
0.8
1
Figure 4.3 Moment at mid-span (normalized with respect to maximum static moment)
versus vehicle position for the problem defined by Figure
4.1
It can be seen from the figures above that a negative dynamic mid-span moment is
obtained when the vehicle enters the bridge, and that the dynamic moment curves
looks similar to the dynamic displacement curves but are more irregular.
The maximum normalized displacement (corresponds to DAF d) in Figure 4.2 are 1.258,
1.258, and 1.259 for the exact model, the present model, and the ABAQUS
model, respectively. The corresponding data for the bending moment (corresponds to
DAF m) are 1.093, 1.093, and 1.083, respectively. The time consumed by the computer to solve
the problem using the exact model, the present model, and the ABAQUS
model was about 2, 14, and 70 CPU seconds, respectively.
One should remember that in this example only the mid-span ( x = L/2) vertical
displacements and mid-span moments are presented, for different vehicle positions, in Figure 4.2
and Figure 4.3. However, the x-value where the maximum response occurs
56
may differ from L/2. From the present model solution, values of 1.260 for x/ L = 0.48
and 1.179 for x/ L= 0.41 were obtained for the maximum normalized displacement and
the maximum normalized moment.
For the present model, the solution of equation (3.11) converges very fast and may
therefore be truncated after few modes (5 modes for the current example). However, as well
known, the convergence of equation (3.15) for the bending moment is much
slower. In reference [57], a better method for calculating the dynamic bending moment is
described.
From Figure 4.2, Figure 4.3, and Table 4.1 one can conclude that the results from the
present model are in very good agreement with those obtained using ABAQUS and the exact
expressions. In addition, the results from the present model for the maximum normalized midspan displacement can also be compared with the results 1.259, 1.258, and 1.251 given in
reference [64], [20], and [93], respectively.
When comparing with the exact solution, the results also indicate that the moving load
problem in ABAQUS was correctly modeled.
4.3
Multi-span continuous bridge with rough road surface
A four-span continuous bridge model with one fixed end, as shown in Figure 4.4, was
adopted for this investigation. This problem was chosen to illustrate the flexibility of the
proposed analysis method.
v = 30 m/s
y( x, t)
m =31700 kg c =8.6 . 104 Ns/m
2
s
m =3000 kg k =9.12 10
. 6 N/m
1
s
x
EI
.
=9.92 1010 Nm2
gg
xv( t)
m =11400 kg/m
g
20 m 25 m
60 m
45 m
Figure 4.4 Idealized four-span continuous bridge with one fixed end and rough road
surface subjected to an idealized 2 DOF vehicle. See chapter 2
for the description of the symbols
57
In this example, the road surface roughness was considered by using the profile shown in
Figure 2.6. This means that the excitation of the dynamic system was not only
caused by the elastic displacements of the bridge but also by the roughness of the road
surface. The 2 DOF vehicle model, consisting of one sprung mass and one unsprung
mass, was adopted. The vehicle characteristics are similar to those used in reference
[65] and correspond to a fundamental frequency of 2.7 Hz and a damping ratio of 8 %.
As in the simply supported bridge example, section 4.2, comparison was also made
here between the present model solution and the one obtained using ABAQUS. After
conducting some convergence study, the solution with 300 segments and 25 included
modes was chosen for the present model and a 150 elements solution with 300
increments for the ABAQUS model. For the ABAQUS model, the computer needed
approximately 570 CPU seconds to solve the problem. The corresponding time for the present
model was about 240 CPU seconds which is only about 42 % of the time
consumed by ABAQUS.
Considering the entire bridge, the dynamic amplification factors for displacement and
bending moment, calculated using the present model, were found to be DAF d =1.396
and DAF m =1.296, respectively. These values can be compared with the DAF values for
the center point6 of the bridge DAF d =1.387 and DAF m =1.126. For the fixed end, the DAF m
was as high as 1.705, which means that the dynamic moment at the fixed end is about 70 %
larger than the static one. Therefore it is very important for the designer not to neglect these
additional dynamic loads in the design process. Studying Figure 1.1, it is interesting to note that
only the Swiss SIA-88, single vehicle curve considers such high DAF values.
In the following pages, mode shapes, natural frequencies, some time histories of
vertical displacements, bending moments, and vertical acceleration, are presented. The
presented results indicate very good agreement between the two different solutions and much
less computer time consumption for the present model. One can observe that the agreement for
the dynamic curves in Figures 4.6 and 4.7 is not as good as for the first numerical example, and it
is not possible to infer which solution is closer to the correct one. The differences in the two
solution methods are believed to cause the differences in the results, see section 4.6 for the
discussion of the numerical results.
6 For a certain point P, the DAF is defined as the ratio of the absolute maximum live load
dynamic response at P to the absolute maximum live load static response at the same point.
58
0
7
14
21
28
35
42
49
56
63
70
77
84
91
98
105
112
119
126
133
140
147
Mode 1 (2.04 Hz)
Mode 2 (4.27 Hz)
Mode 3 (6.48 Hz)
Mode 4 (10.00 Hz)
Mode 5 (12.26 Hz)
Figure 4.5 The first 5 mode shapes calculated using the present model
Mode
Present ABAQUS
Mode
Present ABAQUS
number
300
150
number
300
150
(bending) segments elements (bending) segments elements
1
2.04
2.04
14
53.39
53.56
2
4.27
4.27
15
63.79
64.00
3
6.48
6.48
16
68.77
69.03
4
10.00
10.00
17
77.87
78.20
5
12.26
12.27
18
87.07
87.46
6
14.41
14.42
19
93.79
94.28
7
16.58
16.60
20
107.87
108.42
8
22.64
22.67
21
112.06
112.72
9
26.56
26.60
22
123.04
123.82
10
33.57
33.63
23
131.21
132.10
11
38.12
38.20
24
139.45
140.53
12
43.60
43.71
25
156.52
157.73
13
50.25
50.38
Figure 4.4
59
4.0
(a)
2.0
0.0
-2.0
-4.0
Present
-6.0
ABAQUS
Vertical displacement (mm)
Static (present &
-8.0
ABAQUS )
-10.0
0
0.2
0.4
0.6
0.8
1
4.0
(b)
2.0
0.0
Present
ABAQUS
-2.0
Static (present &
ABAQUS )
-4.0
0
0.2
0.4
0.6
0.8
1
Figure 4.6 Vertical displacement histories for the problem defined by Figure 4.4: (a) at
the center of the bridge (x = 75 m); (b) at the center of
the last span (x = 127.5 m)
60
1000
(a)
0
-1000
-2000
Present
Bending moment (kNm)
ABAQUS
-3000
Static (present &
ABAQUS )
-4000
0
0.2
0.4
0.6
0.8
1
1500
(b)
500
-500
-1500
Present
ABAQUS
-2500
Static (present &
ABAQUS )
-3500
0
0.2
0.4
0.6
0.8
1
Figure 4.7 Bending moment histories for the problem defined by Figure 4.4: (a) at the
center of the bridge (x = 75 m); (b) at the center of the last
span (x = 127.5 m)
61
15.0
(a)
Present
5.0
ABAQUS
-5.0
-15.0
-25.0
0
0.2
0.4
0.6
0.8
1
4.0
(b)
)2 2.0
0.0
-2.0
Vertical acceleration (m/s
Present
ABAQUS
-4.0
0
0.2
0.4
0.6
0.8
1
Figure 4.8 Displacement and acceleration histories for the sprung mass of the vehicle:
(a) sprung mass vertical displacement; (b) sprung mass
vertical acceleration
62
4.4
Simple cable-stayed bridge
A simple fan-shaped cable-stayed bridge with rough road surface and with only the
main span suspended, was chosen for this numerical investigation. The geometrical
configuration and the mechanical properties are shown in Figure 4.9. Optimal values, see
[9], were chosen for the ratios of side span to the main span length and pylon height to main span
length, corresponding to 1/3 and 1/5, respectively. Rotational springs with stiffness Km were
introduced at end supports of the main span girder to simulate the influence of the approach
spans stiffness. The girder is also supported vertically at the pylons but is totally independent of
the pylons. The chosen dead weight, qg, (including surfacing, cables, anchorages, railings, etc.)
is equivalent, for example, to 0.50 m of concrete for a bridge width of 14 m. The chosen live
load, qq, corresponds to a uniformly distributed load of 4.3 kN/m2, which also allows for point
loads. This gives a load ratio of qq /q
g = 0.35, close to that of practice. The allowable
cable stress, a , corresponds to 45 % of the chosen ultimate tensile strength for the
cables which is 1670 MPa. The cross-sectional area of the cables and the
corresponding cable spring stiffness were evaluated using the expressions given in
section 2.2.3.
The characteristics of the 2 DOF vehicle model and the road surface roughness profile are
similar to those used in the previous example. If nothing else is stated in
connection with the presented figures in the following pages, this 2 DOF vehicle
model and the speed of 30 m/s (108 km/h) were used when calculating the response.
y( x, t)
v = 30 m/s
K
K
m
m
40 m 30 m
x
10 m
xv( t)
50 m
150 m
50 m
Figure 4.9 Idealized fan-shaped cable-stayed bridge with rough road surface subjected
to an idealized 2 DOF vehicle. Ec =2.01011 N/m2,
Eg=0.301011 N/m2, Ig=4 m4, Ep=0.301011 N/m2, Ip=10 m4,
a =750106 N/m2, qg=17104 N/m, qq=6104 N/m, and c=9104

N/m3. See chapter 2 for the description of the symbols
63
Satisfactory results were obtained by discretizing the bridge girder into 150 elements in
ABAQUS and using 300 increments, while the present model required 300 segments with 25
included modes. The computer time consumed for the two models was about
625 and 325 CPU seconds, respectively.
Mode shapes, natural frequencies, some time histories of vertical displacements and
bending moments, and the influence of the rotational spring stiffness, vehicle speed, different
vehicle models and road surface roughness on the dynamic response, are
presented. The results are shown for two alternatives, namely, fixed and free pylon top.
Studying Table 4.3, Figures 4.11 and 4.12, one can observe good agreement between
the solutions of the two different methods, specially for the fixed pylon top alternative.
For the free pylon top curves, the simple expressions derived in section 2.2.3, which
consider the anchor cable stiffness and the pylon stiffness, gave larger static and dynamic
displacements for the present model, compared with the ABAQUS model.
The bending moment responses shown in Figure 4.12 are underestimated, by the
present model, at xv / L=0.5. However, by increasing the number of modes considered
to 50, a better agreement will be obtained at this vehicle position, and of course more
CPU seconds will be consumed. See section 4.6 for discussion of the results.
0
7
14
21
28
35
42
49
56
63
70
77
84
91
98
105
112
119
126
133
140
147
Mode 1 (0.94 Hz)
Mode 2 (1.56 Hz)
Mode 3 (2.53 Hz)
Mode 4 (3.93 Hz)
Mode 5 (5.73 Hz)
Figure 4.10 The first 5 mode shapes for the girder calculated using the present model
for fixed pylon top and with Km=
64
Fixed pylon top

Free pylon top
Mode
Present ABAQUS Present ABAQUS
number
300
150
300
150
(bending) segments elements segments elements
1
0.94
0.94
0.76
0.76
2
1.56
1.55
1.38
1.42
3
2.53
2.53
2.41
2.50
4
3.93
3.92
3.85
3.92
5
5.73
5.71
5.68
5.71
6
7.92
7.90
7.88
7.89
7
10.49
10.46
10.47
10.46
8
13.44
13.40
13.42
13.40
9
16.77
16.72
16.75
16.72
10
20.46
20.41
20.45
20.41
11
24.52
24.47
24.51
24.47
12
28.96
28.90
28.95
28.90
13
33.76
33.72
33.75
33.72
14
38.93
38.78
38.92
38.78
15
44.47
44.52
44.46
44.52
16
50.37
50.30
50.37
50.30
17
56.64
56.62
56.64
56.62
18
63.27
63.26
63.27
63.26
19
70.27
70.29
70.27
70.29
20
77.63
77.68
77.63
77.68
21
85.35
85.44
85.35
85.44
22
93.44
93.57
93.44
93.57
23
101.88
102.07
101.88
102.07
24
110.68
110.94
110.68
110.94
25
119.85
120.18
119.85
120.18
Figure 4.9 with Km= . The number of ABAQUS elements
includes only elements for the girder
65
2.0
(a)
0.0
Present
-2.0
ABAQUS
Static (present &
-4.0
ABAQUS )
-6.0
-8.0
-10.0
-12.0
-14.0
0
0.2
0.4
0.6
0.8
1
4.0
(b)
Present
0.0
ABAQUS
-4.0
Static (present)
Static ( ABAQUS )
-8.0
-12.0
-16.0
-20.0
0
0.2
0.4
0.6
0.8
1
Figure 4.11 Vertical displacement histories at the center of the bridge (x = 75 m)
calculated for rough road surface and with Km= : (a) fixed pylon
top; (b) free pylon top
66
1000
(a)
0
-1000
-2000
Present
ABAQUS
-3000
Static (present &
ABAQUS )
-4000
0
0.2
0.4
0.6
0.8
1
1500
(b)
500
-500
-1500
-2500
Present
ABAQUS
-3500
Static (present)
Static ( ABAQUS )
-4500
0
0.2
0.4
0.6
0.8
1
Figure 4.12 Bending moment histories at the center of the bridge (x = 75 m) calculated
for rough road surface and with Km= : (a) fixed pylon
top; (b) free pylon top
67
1.125
(a)
For the entire bridge
At point x=75 m
1.115
d 1.105
DAF
1.095
1.085
1.075
0
0.2
0.4
0.6
0.8
1
1.2
Spring stiffness, K m (1012 Nm)
2.0
(b)
1.8
1.6
m
DAF 1.4
At point x=75 m
At point x=150 m
1.2
1.0
0
0.2
0.4
0.6
0.8
1
1.2
Spring stiffness, K m (1012 Nm)

Figure 4.13 The influence of the rotational spring stiffness, Km, on the dynamic
amplification factors for displacement and bending moment (rough
road surface and fixed pylon top): (a) DAF d; (b) DAF m
68
1.40
(a)
Fixed pylon top, rough road
1.35
Fixed pylon top, smooth road
=30 m/s
Free pylon top, rough road
1.30
v
1.25
d 1.20
DAF
1.15
1.10
1.05
1.00
35
50
65
80
95
110
125
140
155
170
185
200
215
Vehicle speed (km/h)
2.0
(b)
Fixed pylon top, rough road
1.8
=30 m/s
Fixed pylon top, smooth road
v
Free pylon top, rough road
1.6
m
DAF 1.4
1.2
1.0
35
50
65
80
95
110
125
140
155
170
185
200
215
Figure 4.14 The vehicle speed influence on the dynamic amplification factors for
displacement and bending moment calculated for Km= : (a) DAF d for the entire bridge; (b)
DAF m at point x =150 m
69
2.0
(a)
0.0
Sprung mass model
Moving mass model
-2.0
Moving force model
Static
-4.0
-6.0
-8.0
-10.0
-12.0
-14.0
0
0.2
0.4
0.6
0.8
1
1.0
-1.0
Sprung mass model
(b)
Moving mass model
Moving force model
-3.0
Static
-5.0
-7.0
-9.0
-11.0
-13.0
0
0.2
0.4
0.6
0.8
1
Figure 4.15 Vertical displacement histories at the center of the bridge (x = 75 m)
calculated using different vehicle models (fixed pylon top and
Km= ): (a) rough road surface; (b) road surface with no roughness
70
1.25
)g
Sprung mass, rough road
2 m+
Sprung mass, smooth road
1 1.15
m
Moving force
( t) / ( F 1.05
0.95
0.85
Normalized contact force,
0.75
0
0.2
0.4
0.6
0.8
1
Figure 4.16 Normalized contact force calculated for fixed pylon top and with Km=
The presented curves, in Figure 4.13, for the influence of the rotational springs on the
DAF values, indicate fast convergence with increasing Km value. The same DAF
values, as if totally fixed end supports were assumed, are obtained already for a Km value
of 1.21012 Nm. This Km value corresponds to a value of 6.43 for the first term for c 1 in
equation (3.5e). As mentioned earlier in section 3.1.1, a value of 5
corresponds to having a free end support and 7 to a fixed end support.
From Figure 4.14, it can be seen that very high DAF values can be obtained already at
normal vehicle speeds. For example, for a speed of 108 km/h, fixed pylon top, rough road
surface and Km
=
, the obtained DAF values for displacement and bending
moment, considering the entire bridge, are DAF d =1.093 and DAF m =1.531,
respectively. For the fixed end, the DAF m obtained was as high as 1.745. Figure 4.15 and 4.16
indicate that for normal vehicle speeds, most of the excitation of the dynamic system is caused
by the roughness of the road surface and very little is caused by the elastic displacement of the
bridge itself. It can also be concluded that, when assuming a road surface with no roughness and
a normal vehicle speed, the moving force model
71
is fully adequate for this example and there is no need for using complicated vehicle
models.
4.5
Three-span cable-stayed bridge
For this fourth and last example, a symmetric three-span fan-shaped cable-stayed
bridge was adopted with the geometrical configuration and the mechanical properties
shown in Figure 4.17. The horizontal displacements of the pylon tops were neglected in this
study as it was assumed that these displacements are small due to having the side spans vertically
fixed at four points. The cross-sectional area of the cables and the corresponding cable spring
stiffness were evaluated using the expressions given in section 2.2.3. The 2 DOF vehicle model
was used with the same characteristics as for the last two examples.
The effect of local irregularities was investigated by simulating a 3 cm pot-hole
located at about x=150 m. This pot-hole represents a local defect in the road surface or
an initial joint defect.
v = 30 m/s
y( x, t)
80
x
3 cm pot-hole
xv( t)
128 m
400 m
128 m
(8 x 16 m)
(25 x 16 m)
(8 x 16 m)
Figure 4.17 Idealized fan-shaped cable-stayed bridge subjected to an idealized 2
DOF vehicle. Ec =2.01011 N/m2, Eg=0.301011 N/m2, Ig=10 m4,
a =750106 N/m2, qg=17104 N/m, qq=6104 N/m, and c=9104
N/m3. See chapter 2 for the description of the symbols
If nothing else is stated in connection with the presented figures in the following pages,
the bridge characteristics given in Figure 4.17 and the vehicle characteristics given in Figure 4.4
were used when calculating the response.
Satisfactory results were obtained by discretizing the bridge girder into 369 elements in
ABAQUS and using 738 increments, while the present model required 533 segments with 25
included modes. The computer time consumed for the two models was about
72
4000 CPU seconds for the ABAQUS model and 1000 CPU seconds for the present model.
This indicates that the present method involves much less computation, and
therefore is much more computationally efficient, compared to the finite element
method with direct time integration.
Mode shapes, natural frequencies, some time histories of vertical displacements and
bending moments, and the influence of some parameters (vehicle speed, pylon height and vehicle
spring stiffness) on the dynamic amplification factors, are shown in the following pages.
It can be seen from Table 4.4 that the natural frequencies of this cable-stayed bridge are
very close to one another, and that excellent agreement is obtained especially when comparing
with the ABAQUS values calculated neglecting cable mass. The ABAQUS
frequencies calculated neglecting cable mass ( c =0) are only presented in Table 4.4
for comparison and have not been used for dynamic response calculations.
The obtained DAF values, for the problem defined by Figure 4.17, are:
for the entire bridge DAF d =1.056 and DAF m =1.118; for the center point DAF
d =1.046
and DAF m =1.074; for the point at x =128 m DAF =1.179; and for the point at
m
x =528 m DAF =1.264.
m
Mode 1 (0.50 Hz)
Mode 2 (0.66 Hz)
Mode 3 (0.84 Hz)
Mode 4 (1.07 Hz)
Mode 5 (1.39 Hz)
0
32 64 96 128 160 192 224 256 288 320 352 384 416 448 480 512 544 576 608 640
Figure 4.18 The first 5 mode shapes for the girder calculated using the present model
73
Mode
Present
ABAQUS
ABAQUS
number
533 segments
369 elements
369 elements
(bending)
c = 9104 N/m3
c=0
1
0.50
0.49
0.50
2
0.66
0.65
0.66
3
0.84
0.83
0.84
4
1.07
1.06
1.07
5
1.39
1.37
1.39
6
1.80
1.78
1.80
7
2.31
2.29
2.31
8
2.91
2.88
2.91
9
3.59
3.56
3.60
10
4.36
4.32
4.36
11
5.20
5.15
5.20
12
6.08
6.04
6.09
13
6.65
6.61
6.66
14
6.73
6.68
6.74
15
7.13
7.07
7.14
16
7.90
7.85
7.91
17
8.29
8.25
8.31
18
8.69
8.64
8.71
19
9.52
9.45
9.54
20
10.34
10.29
10.37
21
10.79
10.75
10.83
22
11.34
11.29
11.38
23
12.33
12.25
12.36
24
13.10
13.06
13.16
25
13.41
13.38
13.47
Figure 4.17. The number of ABAQUS elements includes only
elements for the girder
74
Studying Figure 4.19 and Figure 4.20, it is noted that the present solution is somewhat
stiffer than the ABAQUS solution. This is probably due to the differences in the two models. In
the ABAQUS model the cables are placed 16 meters apart, whereas in the present model there is
always a spring7 supporting the girder beneath each vehicle position. Still, from an engineering
point of view, one can say that the ABAQUS and the present model results exhibit good matches,
because the important values in Figure 4.19 and Figure 4.20 are those giving the maximum
response.
Figure 4.21 shows the variation of the DAF with the speed of the vehicle for three
different girder moment of inertia. It can be seen that high DAF values are obtained for low
vehicle speeds. Thus for this case, limitation of the speed of vehicles will not necessarily avoid
damage effects on bridges.
The effect of pylon height on the DAF is shown in Figure 4.22. It is obvious that
increasing pylon height results in lower DAF values.
0
-5
Present
-10
ABAQUS
-15
Static (present &
ABAQUS )
Vertical displacement (mm) -20
-25
0
0.2
0.4
0.6
0.8
1
Figure 4.19 Vertical displacement histories at the center of the bridge (x=328
m)
7 Used to idealize the cables.
75
2000
(a)
1000
0
-1000
-2000
Present
-3000
ABAQUS
Bending moment (kNm) -4000
Static (present &
ABAQUS )
-5000
-6000
0
0.2
0.4
0.6
0.8
1
4000
(b)
Present
3000
ABAQUS
2000
Static (present &
ABAQUS )
1000
0
-1000
-2000
0
0.2
0.4
0.6
0.8
1
Figure 4.20 Bending moment histories: (a) at the center of the bridge (x=328 m);
(b) at the first pylon (x=128 m)
76
1.12
(a)
Ig
I=
g 5m
=5 4
m4
1.11
/s m
Ig
I =10 m4
g=10 m4
1.10
30=
Ig
I =30 m4
v
g=30 m4
1.09
d 1.08
DAF 1.07
1.06
1.05
1.04
1.03
50
70
90
110
130
150
170
1.80
(b)
1.70
Ig
I =5 m4
g=5 m4
Ig
I=
g 10 m4
=10 m4
1.60
=30 m/s
Ig
I=
g 30 m4
v
=30 m4
1.50
m 1.40
DAF
1.30
1.20
1.10
1.00
50
70
90
110
130
150
170

Figure 4.21 The influence of vehicle speed on the dynamic amplification factors for
displacement and bending moment calculated for different Ig
values: (a) DAF d for the entire bridge; (b) DAF m at point x =128 m
77
1.060
(a)
1.055
At point x=328 m
1.050
d
1.045
DAF
1.040
m 80= H
1.035
1.030
50
60
70
80
90
100
110
120
130
140
150
Pylon height, H (m)
1.30
At point x=128 m
(
At point x=328 m
1.25
At point x=528 m
(b)
=80 m
1.20
H
m 1.15
DAF
1.10
1.05
1.00
50
60
70
80
90
100
110
120
130
140
150
Pylon height, H (m)
Figure 4.22 The influence of pylon height on the dynamic amplification factors for
displacement and bending moment: (a) DAF d; (b) DAF m
78
1.07
(a)
1.06
1.05
d
DAF 1.04
At point x=328 m
N/m610.
1.03
=9.12
kS
1.02
2.28
4.56
6.84
9.12
11.4
13.68
15.96
18.24
Vehicle spring stiffness, k S (106 N/m)
(1.35 Hz) (1.91 Hz) (2.34 Hz) (2.70 Hz) (3.02 Hz) (3.31 Hz) (3.57 Hz) (3.82 Hz) 1.50
(b)
N/m6
1.40
10.
1.30
=9.12
kS
m 1.20
DAF
1.10
At point x=128 m
At point x=328 m
1.00
At point x=528 m
0.90
2.28
4.56
6.84
9.12
11.4
13.68
15.96
18.24
Vehicle spring stiffness, k S (106 N/m)
Figure 4.23 The influence of vehicle spring stiffness on the dynamic
amplification factors for displacement and bending moment:
(a) DAF d; (b) DAF m
79
Figure 4.23 shows the variation of the DAF with the vehicle spring stiffness. It is clear
that increasing vehicle spring stiffness results in increasing contact forces and
increasing DAF values. The reason for this is believed to be that the vehicle sprung mass
will more and more become unsprung as the spring stiffness increases.
From Figure 4.22 and Figure 4.23 we can infer that the DAF can reach high values for
the girder near the pylons, and that the DAF at the center of the bridge are
comparatively small. This situation should be considered in the design practice of the
bridge. These high DAF values, for the girder near the pylons, are believed to be
caused by the higher modes (mode number 8 and 7), which have natural frequencies
close to that of the vehicle (2.7 Hz). Mode number 8 has a zero point at the center of the
bridge ( x = 328 m) and maximum points close to the pylons.
4.6
Discussion of the numerical results
In this chapter, four problems have been studied in order to demonstrate the efficiency
and the validity of the proposed model. The results obtained using the present model are
generally in good agreement with those obtained using the ABAQUS model.
For the first numerical example, extremely good agreement is observed between the
three different solutions, and it was not easy to produce Figures 4.2 and 4.3 so all curves
can be distinguished. The reason for this extremely good agreement is the
simplicity of the problem, as no surface roughness was included and the moving force
model was adopted.
As the problems studied got more complicated due to having rough road surface and
more complex bridge and vehicle models, the differences between the ABAQUS
solutions and the present model solutions increased. As a general remark, one should bear
in mind that both methods are approximate, and it is not always possible to infer which solution
is closer to the correct one.
The main parameters that are believed to cause the differences in the dynamic
response are:
cable spacing. For large cable spacing compared to bridge length, the present model
will give stiffer solutions due to using continuously distributed vertical
spring to idealize the cables
80
location of the point studied. Differences can also be the result of studying the response
of a point situated between two cables instead of a point adjacent to a
cable. For the same reason as above, the present model will give stiffer solutions for
points between cables
cable mass. The cable mass is neglected in the present model but not in the ABAQUS
model. Using the present model, this will also lead to stiffer solutions
the surface roughness. For the present model, the surface roughness profile, r( x), and
its derivatives were given at each segment joint, while for the ABAQUS model the roughness
profile was considered when giving the vertical location of each
node. If more increments are needed than the number of nodes, the vehicle locations will
not always coincide with one of those nodes, and it is not clear how ABAQUS
evaluates the vertical locations of these inter nodal points situated on the SLIDE
LINE
numerical damping. The influence of the automatically introduced numerical
(artificial) damping on the ABAQUS solution. For problems tested with no
numerical damping, bad numerical stability was observed.
81
82
Chapter
______________________________________________________________________
Conclusions and Suggestions for Further

Research
_____________________________________________________________________
_
5.1
Conclusions of Part A
The present work has reviewed previous research conducted in the field of dynamic
response of bridges subjected to moving vehicles, and presented a simplified analysis
method for evaluating the dynamic response of cable-stayed bridges.
For the evaluation of the dynamic response due to moving vehicles, a computer code was
developed which fully consider the dynamic interaction between the bridge and
the vehicle. Time histories and the influence of some parameters, e.g. vehicle speed, on
the dynamic amplification factors were presented. The results were compared with results
obtained using the commercial finite element code ABAQUS.
Good agreement was obtained between the ABAQUS results and the results from the
present model. Moreover, the presented model proves to be simple to use,
computationally economical, and the finite difference method serves as an efficient tool
for analysis of such problems. However, the major drawback for the present
model is that it is not capable of considering the interaction between the side spans and
the main span through the cables. As a conclusion, referring to the discussion in
section 4.6 of the numerical results, it has been shown that the theoretical formulations
used are valid and that the present model correctly handles the moving load problem.
There are several conclusions that can be drawn from the numerical results presented in
the previous chapter:
83
the surface roughness has great effect on the dynamic response. Thus, when
calculating the DAF, the roughness of the road surface should be considered. To
reduce damage to bridges, attempts should be made to eliminate irregularities in the deck,
in the approach pavements and over bearings
it was found that limitation of the speed of vehicles will not necessarily avoid damage
effects on bridges
in some cases greater dynamic amplification factors can be obtained than those
given by some of the current bridge design codes
as expected, if the vehicle travels at a normal speed and smooth bridge deck surface is
assumed, and if the ratio of the vehicle mass to bridge mass is very small, the effect of bridgevehicle interaction is small compared to the bridge inertia.
Therefore, for this case, the moving force solution represents an acceptable
approximation for the sprung mass solution.
The simplified analysis method presented in this part of the thesis can be used for
dynamic analysis of conventional type of bridges, and for comparing different design alternatives
of cable-stayed bridges in the feasibility design stage. For the final design stage, and specially for
long span cable-stayed bridges, the author recommends

carrying out a three-dimensional finite element analysis, as the nonlinearities, cable
masses, axial forces, and the three-dimensional motion of the structure may no longer be
ignored.
For the present model, it is clear that the modeling assumptions ignore several factors that
may significantly affect the response, particularly those related to cable mass, bridge damping,
and vehicle configuration. Moreover, only fan-shaped cable-stayed
bridges were considered in this study. However, the developed computer code is
general and flexible and can be improved to handle cable-stayed bridges with other cable
configurations (harp-shaped or modified fan-shaped), more realistic traffic loads such as a
sequence of moving vehicles, or more realistic bridge structures including damping and with
variable mass and flexural rigidity of the stiffening girder.
The results obtained in the previous chapter indicate that the dynamic behavior of bridges
due to moving vehicles is too complicated to be approximated using a simple formula for
dynamic amplification factors, as adopted in many of the current bridge design codes. This study
and previous studies have shown that parameters such as
84
span length (or fundamental frequency of the bridge), road surface condition, bridge
damping, and design vehicle speed should be considered in these DAF formulas.
Conducting this study, it was also found that very few commercial finite element
codes are capable of handling the bridge-vehicle interaction problem and the moving load
problem. The commercial finite element codes capable of handling such problems are very
expensive and require powerful computers. The proposed model on the other hand has shown to
be computationally efficient and can be simply implemented, e.g.
using FORTRAN, MATLAB [53], or even Microsoft Excel, on an ordinary PC, giving a
much more economical solution.
5.2
Suggestions for further research
Based on the present study, the following suggestions for further research can be
given:
referring to the discussion in section 4.6 of the numerical results, a sensitivity study
need to be carried out to investigate the effects of cable spacing, cable mass
neglection, variation in natural frequencies, etc., on the present model results. This is
necessary to find the limits of the proposed simplified analysis method
further work is needed to develop the expressions derived in chapter 2 and the
implemented code, in order to study the response of cable-stayed bridges with other cable
configurations and to include the effect of bridge damping and axial forces
investigate the effect of span length, bridge damping, bridge-vehicle mass ratio,
multiple vehicles, and torsional loading on the dynamic response of more realistic cable-stayed
bridge models
study the effect of vehicle braking on the dynamic response of cable-stayed bridges
study the effectiveness of tuned mass dampers (TMD) on suppressing vibrations

induced by moving vehicles
perform extensive instrumentation and testing on an existing cable-stayed bridge, or on
a laboratory model, to further enhance the understanding of the behavior of this type of bridge
structures, and to verify theoretical models.
To study the above listed theoretical topics, the author recommends using the finite
element method. This is necessary if more realistic three-dimensional bridge and
85
vehicle models are to be used for analysis. Axial forces, geometric nonlinearities,
torsional modes of vibration, cable modes of vibration, and bridge damping can then easily be
considered in the analysis.
86
______________________________________________________________________
Bibliography of Part A
______________________________________________________________________
[1]
ABAQUS Users Manual, Hibbitt, Karlsson & Sorensen, Inc., Providence, Rhode Island,
1994.
[2]
Abdel-Ghaffar A.M., Nazmy A.S., 3-D Nonlinear Seismic Behaviour of CableStayed
Bridges, J. Struct. Eng., ASCE, Vol. 117, No. 11, Nov. 1991, pp.
3456-3476.
[3]
Alessandrini L., Brancaleoni F., Petrangeli M.P., On the Dynamic Response of
Cable- Stayed Bridges under moving loads, Costruzioni Metalliche, No. 2,
Oct., 1984, pp. 67-82.
[4] Arpe
R. , Dynamic Response of Bridges due to Traffic Loading - Part 1
Theoretical
Study, Serie R, No. 194, Dept. of Struct. Eng., Techn. Univ. of
Denmark, Lyngby, 1984. (in Danish)
[5] Arpe
R. , Dynamic Response of Bridges due to Traffic Loading - Part 2
Experimental
Study, Serie R, No. 195, Dept. of Struct. Eng., Techn. Univ. of
Denmark, Lyngby, 1984. (in Danish)
[6] Bachmann H., Weber B. , Tuned Vibration Absorbers for Lively Structures,
Struct. Eng. Int., Vol. 5, No. 1, 1995, pp. 31-36.
[7]
Blejwas T.E., Feng C.C., Ayre R.S. , Dynamic Interaction of Moving Vehicles
and
Structures, J. of Sound and Vibration, 67(4), 1979, pp. 513-521.
[8]
Brancaleoni F., Petrangeli M.P., Villatico C., Train Structure Interaction in
Railway Cable Stayed Bridge Dynamics, Proc. Int. Conference on CableStayed Bridges, Bangkok, Nov. 1987, pp. 452-463.
[9] Bruno D., Colotti V. , Vibration Analysis of Cable-Stayed Bridges, Struct.
Eng. Int., Vol. 4, No. 1, 1994, pp. 23-28.
87
87
[10] Bryja D., Sniady P. , Random Vibration of a Suspension Bridge due to
Highway
Traffic, J. of Sound and Vibration, 125(2), 1988, pp. 379-387.
[11] Burdet O.L., Corthay S. , Dynamic Load Testing of Swiss Bridges, IABSE
Symposium, San Francisco, 1995, pp. 1123-1128.
[12] Cai Y., Chen S.S., Rote D.M., Coffey H.T. ,Vehicle/Guideway Interaction for
High Speed Vehicles on a Flexible Guideway, J. of Sound and Vibration,
175(5), 1994, pp. 625-646.
[13] Chatterjee P.K., Datta T.K. , Dynamic Analysis of Arch Bridges Under
Travelling
Loads, Int. J. Solids Struct., Vol. 32, No. 11, 1995, pp.1585-1594.
[14] Chatterjee P.K., Datta T.K., Surana C.S ., Vibration of Cable-Stayed Bridges
under Moving Vehicles, Struct. Eng. Int., No. 2, 1994, pp. 116-121.
[15] Chatterjee P.K., Datta T.K., Surana C.S ., Vibration of Suspension Bridges
under Vehicular Movement, J. Struct. Eng., ASCE, Vol. 120, No. 3, March
1994,
pp.681-703.
[16] Chu K.H., Garg V.K., Dhar C.L. , Railway-Bridge Impact: Simplified Train and
Bridge Model, J. Struct. Division, ASCE, Vol. 105, No. ST9, Sept. 1979, pp.1823-1844.
[17] Chu K.H., Garg V.K., Wang T.L. , Impact in Railway Prestressed Concrete
Bridges, J. Struct. Eng., ASCE, Vol. 112, No. 5, May 1986, pp.1036-1051.
[18] Coussy O., Said M., Van Hoove J.P. , The Influence of Random Surface
Irregularities on the Dynamic Response of Bridges Under Suspended Moving
Loads, J. of Sound and Vibration, 130(2), 1989, pp. 313-320.
[19] Diana G., Cheli F. , A Numerical Method to Define the Dynamic Behaviour of a
Train Running on a Deformable Structure, Meccanica, Vol. 23, 1988, pp. 27-42.
[20] Filho F.V. , Finite Element Analysis of Structures Under Moving Loads, Shock and
Vibration Digest, Vol. 10, No. 8, 1978, pp. 27-35.

[21] Fleming J.F., Egeseli E.A. , Dynamic Behaviour of a Cable-Stayed Bridge,
Earthquake Eng. and Struct. Dynamics, Vol. 8, 1980, pp. 1-16.
88
88
[22] Fleming J.F., Egeseli E.A. , Dynamic Response of Cable-Stayed Bridge
Structures, Cable-Stayed Bridges, a compilation of papers presented at Eng.
Symposium Cable-Stayed Bridges, Pasco, Washington, Dec. 1977, pp. 146168.
[23] Frba
L. , Vibration of Solids and Structures under Moving Loads, Noordhoff
International
Publishing, Groningen, 1972.
[24] Frba
L. , Dynamics of Railway Bridges, Thomas Telford, London, 1996.
[25] Garg V.K., Dukkipati R.V. , Dynamics of Railway Vehicle Systems, Academic Press,
Toronto, 1984.
[26] Genin J., Ginsberg J.H., Ting E.C. , A Complete Formulation of Inertial Effects
in the Guideway-Vehicle Interaction Problem, J. of Sound and Vibration,
38(1), 1975, pp. 15-26.
[27] Gimsing
N.J. , Cable Supported Bridges, second edition, Wiley, Chichester,
1997.
[28] Gimsing N.J., Lockwood J.D., Song J. , Analysis of Erection Procedures for CableStayed
Bridges, Serie R, No. 247, Dept. of Struct. Eng., Techn. Univ. of
Denmark, Lyngby, 1989.
[29] Ginsberg J.H., Ting E.C., Genin J., Parametric Study of the Interaction of
Bridges and Moving Vehicles, J. Appl. Sci. Res., 32, 1976, pp. 355-370.
[30] Green M.F., Cebon D., Cole D.J., Effects of Vehicle Suspension Design on
Dynamics of Highway Bridges, J. Struct. Eng., ASCE, Vol. 121, No. 2, Feb.
1995, pp. 272-282.
[31] Gupta R.K., Traill-Nash R.W., Vehicle Braking on Highway Bridges, J. Eng.
Mechanics Division, ASCE, Vol. 106, No. EM4, Aug. 1980, pp. 641-658.
[32] Hayashikawa T., Watanabe N., Dynamic Behavior of Continuous Beams with
Moving
Loads, J. Eng. Mechanics Division, ASCE, Vol. 107, No. EM1, Feb.
1981, pp. 229-246.

89
89
[33] Hayashikawa T., Watanabe N., Suspension Bridge Response to Moving
Loads,
J. Eng. Mechanics Division, ASCE, Vol. 108, No. EM6, Dec. 1982, pp.
1051- 1066.
[34] Hillerborg A., Dynamic Influences of Smoothly Running Loads on Simply
Supported
Girders, Doctoral Thesis, Dept. of Struct. Eng., Royal Institute of
Technology, Stockholm, 1951.
[35] Hino J., Yoshimura T., Konishi K., Ananthanarayana N. , A Finite Element
Method Prediction of the Vibration of a Bridge Subjected to a Moving Vehicle
Load, J. of Sound and Vibration, 96(1), 1984, pp. 45-53.
[36] Hirai A., Ito M., Response of Suspension Bridges to Moving Vehicles, J. of the
Faculty of Eng., University of Tokyo (B), Vol. XXIX, No. 1, 1967.
[37] Honda H., Kobori T., Yamada Y., Dynamic Factor of Highway Steel Girder
Bridges, IABSE Proceedings P-98/86, Periodica 2/1986, pp. 57-75.
[38] Honda H., Kajikawa Y., Kobori T., Spectra of Road Surface Roughness on
Bridges, J. Struct. Division, ASCE, Vol. 108, No. ST9, Sep. 1982, pp. 1956-1966.
[39] Huang D., Wang T.L., Shahawy M., Impact Analysis of Continuous
Multigirder Bridges due to Moving Vehicles, J. Struct. Eng., ASCE, Vol. 118, No. 12,
Dec. 1992, pp. 3427-3443.
[40] Huang D., Wang T.L., Shahawy M., Vibration of Thin-Walled Box-Girder
Bridges Excited by Vehicles, J. Struct. Eng., ASCE, Vol. 121, No. 9, Sep.
1995, pp. 1330-1337.
[41] Huang D., Wang T.L., Shahawy M., Dynamic Behavior of Horizontally
Curved
I-Girder Bridges, Computer & Structures, Vol. 57, No. 4, 1995, pp.
703-714.
[42] Humar J.L., Kashif A.H., Dynamic Response Analysis of Slab-Type Bridges, J.
Struct. Eng., ASCE, Vol. 121, No. 1, Jan. 1995, pp. 48-62.
[43] Hwang E.S., Nowak A.S., Simulation of Dynamic Load for Bridges, J.
Struct. Eng., ASCE, Vol. 117, No. 5, May 1991, pp. 1413-1434.
90
90
[44] Inbanathan M.J., Wieland M., Bridge Vibrations due to Vehicle Moving Over
Rough
Surface, J. Struct. Eng., ASCE, Vol. 113, No. 9, Sep. 1987, pp. 19942008.
[45] Indrawan B., Vibrations of a Cable-Stayed Bridge due to Vehicle Moving over
Rough
Surface, Master Thesis, Asian Institute of Technology, Bangkok,
March
1989.
[46] ISO
8608:1995(E),
Mechanical vibration- Road surface profiles- Reporting of
measured
data.
[47] Karoumi
R. , Aerodynamic Stability, Wind Forces on Suspension and CableStayed
Bridges, TRITA-BKN Report No. 11, Dept. of Struct. Eng., Royal
Institute of Technology, Stockholm, 1994. (in Swedish)
[48] Karoumi
R. , Dynamic Response of Cable-Stayed Bridges Subjected to Moving
Vehicles, IABSE 15th. Congress, Denmark, June 1996, pp. 87-92.
[49] Khalifa M.A.H., Analysis of Traffic Induced Vibration and Fatigue of CableStayed
Bridges, Doctoral Thesis, Faculty of the Graduate School, University of
Southern California, Aug. 1991.
[50] Knothe KL., Grassie S.L. , Modelling of Railway Track and Vehicle/Track
Interaction at High Frequencies, Vehicle System Dynamics, Vol. 22, No. 3-4,
1993, pp. 209-262.
[51] Lee
H.P.,
Dynamic Response of a Multi-Span Beam on One-Sided Point
Constraints Subjected to a Moving Load, Computer & Structures, Vol. 55, No.
4, 1995, pp. 615-623.
[52] Lee K., A Numerical Solution for Dynamic Interaction Between Rigid Wheel
and Flexible Beam, Communications in Numerical Methods in Eng., Vol. 11,
1995, pp. 267-279.
[53] Matlab Reference Guide, version 4.2, The MathWorks, Inc., Natick, Mass., 1994.
[54] Menn
C.,
Prestressed Concrete Bridges, Birkhuser Verlag, Boston, c1990.
91
91
[55] Miyazaki S., Kanamori M., Wakui H., Matsumoto N., Sogabe M. , Analytical
Study on Dynamic Response and Riding Comfort of PC Multicable-Stayed
Railway
Bridge, Int. Conference on Speedup Technology for Railway and
Maglev Vehicles, Yokohama, Nov. 1993, pp. 424-429.
[56] Mulcahy N.L. , Bridge Response with Tractor-Trailer Vehicle Loading, Earthquake
Eng. and Struct. Dynamics, Vol. 11, 1983, pp. 649-665.
[57] Nagaraju N., Jagadish K.S., Sundara K.T., Dynamic Behaviour of Cantilever
Bridges Under Moving Loads, IABSE Publications 33-II, 1973, pp. 149-171.
[58] Nielsen J. C. O., Train/Track Interaction, Coupling of Moving and Stationary
Dynamic Systems - Theoretical and Experimental Analysis of Railway
Structures Considering Wheel and Track Imperfections, Doctoral Thesis, Div.
of Solid Mechanics, Chalmers Univ. of Technology, Gteborg, Sweden, 1993.
[59] Ohlsson S., Dynamic Characteristics of Cable-Stayed Bridges - Nonlinearities
and Weakly Coupled Modes of Vibration, Proc. Int. Conference on CableStayed Bridges, Bangkok, Nov. 1987, pp. 421-430.
[60] Ohlsson
S. , Modal Testing of the Tjrn Bridge, 4th Int. Modal Analysis
Conference, Los Angeles, Feb. 1986, pp. 599-605.
[61] Olsson
M. , On the Fundamental Moving Load Problem, J. of Sound and
Vibration, 145(2), 1991, pp. 299-307.
[62] Olsson M. , Finite Element, Modal Co-ordinate Analysis of Structures
Subjected to Moving Loads, J. of Sound and Vibration, 99(1), 1985, pp. 1-12.
[63] Olsson
M. , Finite Element Analysis of Structures Subjected to Moving Loads,
Report TVSM-3004, Division of Struct. Mechanics, Lund Institute of
Technology, Lund, Sweden, 1983.
[64] Olsson M., Analysis of Structures Subjected to Moving Loads, Doctoral Thesis,
Report TVSM-1003, Division of Struct. Mechanics, Lund Institute of
Technology, Lund, Sweden, 1986.
[65] Palamas J., Coussy O., Bamberger Y. , Effects of Surface Irregularities Upon
the Dynamic Response of Bridges Under Suspended Moving Loads, J. of
Sound
and Vibration, 99(2), 1985, pp. 235-245.
92
92
[66] Paultre P., Chaallal O., Proulx J. , Bridge Dynamics and Dynamic
Amplification Factors - a Review of Analytical and Experimental Findings, Canadian J.
of Civil Eng., 19 (2), 1992, pp. 260-278.
[67] Paultre P., Proulx J., Talbot M. , Dynamic Testing Procedures for Highway
Bridges Using Traffic Loads, J. Struct. Eng., ASCE, Vol. 121, No. 2, Feb.
1995, pp. 362-376.
[68] Podolny W. Jr., Scalzi J.B. , Construction and Design of Cable-Stayed Bridges, 2nd
ed., Wiley & Sons, New York, NY, 1986.
[69] Rasoul M.,
Dynamic Analysis of Cable-Stayed Bridges, Doctoral Thesis,
Purdue University, Dec. 1981.
[70] Saadeghvaziri M.A., Finite Element Analysis of Highway Bridges Subjected to
Moving
Loads, Computer & Structures, Vol. 49, No. 5, 1993, pp. 837-842.
[71] Schneider H.J., Elf H.P., Klle P., Modeling of Travelling-Loads and TimeDependent Masses with Adina, Computer & Structures, Vol. 17, No. 5-6, 1983, pp. 749755.
[72] Ting E.C., Genin J., Ginsberg J.H. , A General Algorithm for Moving Mass
Problems, J. of Sound and Vibration, 33(1), 1974, pp. 49-58.
[73] Ting E.C., Genin J., Dynamics of Bridge Structures, Structural Mechanics
Archives, Vol. 5, Issue 3, Aug. 1980, pp. 217-252.
[74] Ting E.C., Yener M., Vehicle-Structure Interaction in Bridge Dynamics, Shock and
Vibration Digest, Vol. 15, No. 2, 1983, pp. 3-9.
[75] Troitsky M.S. , Cable-Stayed Bridges, 2nd ed., BSP Professional Books, London,
1988.
[76] Ulstrup C.C. et al. , Guidelines for the Design of Cable-Stayed Bridges, Reported by
the ASCE Committee on Cable-Stayed Bridges, ASCE, New
York, NY, 1992.
[77] Veletsos A.S., Huang T., Analysis of Dynamic Response of Highway Bridges, J.
Eng. Mechanics Division, ASCE, Vol. 96, No. EM5, Oct. 1970, pp. 593-620.
93
93
[78] Wakui H., Matsumoto N., Tanabe M., A Study on Dynamic Interaction
Analysis for Railway Vehicles and Structures - Mechanical Model and

Practical
Analysis Method, QR of Railway Tech. Research Inst., Tokyo, Vol. 35,
No. 2, May. 1994, pp. 96-104.
[79] Walther R., Houriet B., Isler W., Moa P. , Cable Stayed Bridges, Thomas Telford,
London, 1988.
[80] Walther R., Cable Stayed Bridges with Slender Deck, Swiss Federal Inst. of
Technology - IBAP, Lausanne, Test Report No. 81.11.03, Sept. 1988.
[81] Wang T.L., Ramp/Bridge Interface of Railway Prestressed Concrete Bridge, J.
Struct. Eng., ASCE, Vol. 116, No. 6, June 1990, pp. 1648-1659.
[82] Wang T.L., Garg V.K., Chu K.H., Railway Bridge/Vehicle Interaction Studies
with New Vehicle Model, J. Struct. Eng., ASCE, Vol. 117, No. 7, July 1991, pp. 20992116.
[83] Wang T.L., Huang D., Computer Modeling Analysis in Bridge Evaluation, Report
No. FL/DOT/RMC/0542-3394, Florida Dept. of Transp. Tallahasee,
Florida,
1992.
[84] Wang T.L., Huang D., Cable-Stayed Bridge Vibration due to Road Surface
Roughness, J. Struct. Eng., ASCE, Vol. 118, No. 5, May 1992, pp. 13541374.
[85] Wang T.L., Huang D., Shahawy M., Dynamic Response of Multigirder
Bridges, J. Struct. Eng., ASCE, Vol. 118, No. 8, Aug. 1992, pp. 2222-2238.
[86] Wang T.L., Shahawy M., Huang D.Z., Dynamic Response of Highway Trucks
due to Road Surface Roughness, Computer & Structures, Vol. 49, No. 6, 1993, pp.
1055-1067.
[87] Wang T.L., Impact in a Railway Truss Bridge, Computer & Structures, Vol.
49, No. 6, 1993, pp. 1045-1054.
[88] Weaver W., Timoshenko S.P., Young D.H. , Vibration Problems in
Engineering, 5th edition, Wiley & Sons, New York, 1990.
94
94
[89] Wilson J.F., Barbas S.T. , Responses of Continuous Elastically Supported Beam
Guideways to Transit Loads, J. Dynamic System, Measurement, and Control, Vol. 102,
Dec.1980, pp. 247-254.
[90] Yang Y.B., Liao S.S., Lin B.H., Impact Formulas for Vehicles Moving Over
Simple and Continuous Beams, J. Struct. Eng., ASCE, Vol. 121, No. 11, Nov.
1995, pp. 1644-1650.
[91] Yang Y.B., Lin B.H., Vehicle-Bridge Interaction Analysis by Dynamic
Condensation
Method, J. Struct. Eng., ASCE, Vol. 121, No. 11, Nov. 1995,
pp. 1636-1643.
[92] Yener M., Chompooming K., Numerical Method of Lines for Analysis of
Vehicle-Bridge
Dynamic
Interaction, Computer & Structures, Vol. 53, No. 3,
1994, pp. 709-726.
[93] Yoshida D.M., Weaver W., Finite-Element Analysis of Beams and Plates with
Moving
Loads, Publ. Int. Assoc. Bridge Struct. Eng., Vol. 31, No. 1, 1971, pp.
179-195.
95
95
96
96
Part B
Refined Analysis Utilizing
the Nonlinear Finite
Element Method
97
98
Chapter
______________________________________________________________________
Introduction
______________________________________________________________________
6.1
General
Although several long span cable-stayed and suspension bridges are being build or
proposed for future bridges, little is known about their dynamic behavior under the action
of moving traffic loads. Cable supported bridges are complex structures
consisting of various structural components with different properties. Thus, to take
account of the complex structural response and to more realistically predict their response due to
traffic loading, a detailed structural analysis is required. Various studies of the dynamic response
due to moving vehicles have been conducted on
ordinary bridges, see the state-of-the-art review in Part A of this thesis. However, due to
reasons stated above, they cannot be directly applied to modern cable supported bridges.
Due to the low damping, lightweight and high flexibility of modern long span cablestayed and suspension bridges, vibrations induced by traffic can be a serious problem.
Vibration effects should play a much more dominating role for the design and not be
underestimated, as the ratio between traffic load and dead load can be larger for these modern
bridges. It is well known that long-term vibration of the bridge deck and
cables (in particular the cables in cable-stayed bridges) might enhance and accelerate the
fatigue damage on the bridge. Costly repairs and modifications have been
undertaken on relatively new suspension and cable-stayed bridges, because the
possibility of fatigue caused by traffic-induced vibrations had not been sufficiently
investigated at the design stage. In addition, dynamic forces of heavy vehicles can lead to bridge
deterioration and eventually increasing maintenance costs and decreasing service life of the
bridge structure. Therefore it ought to be evident that it is important for engineers to not only use
design code formulas but also be able to accurately
99
investigate and understand the vibration effects at the design stage. At this stage the
bridge structure can readily be modified, rather than having to make costly
modifications later on.
For cable supported bridges and in particular long span cable-stayed bridges, energy
dissipation is very low and is often not enough on its own to suppress vibrations. To increase the
overall damping capacity of the bridge structure, one possible option is to incorporate external
dampers1 into the system. Such dampers can be found today on
many existing cable supported bridges [33, 58]. However, it is not always easy to find a
location with significant relative movements and enough space to accommodate
these devices. Moreover, this is not always the most effective and economic solution.
The application of such devices for bridge structures is discussed later on in Chapter 8.
The damping characteristics and damping ratios of cable supported bridges are also
discussed and a practical technique for deriving the damping matrix from modal
damping ratios, is presented.
Since the cable supported bridge structures to be analyzed in this part of the thesis are
flexible and can undergo large displacements, the nonlinear finite element method is utilized
considering all sources of geometric nonlinearity. A beam element, which
includes geometrically nonlinear effects and is derived using a consistent mass
formulation, is adopted for modeling the girder and the pylons. Whereas, a two-node
cable element derived using exact analytical expressions for the elastic catenary, is adopted for
modeling the cables.
Two approaches for evaluating the dynamic response are adopted. In the first
approach, the response is evaluated using the mode superposition technique utilizing the
deformed dead load tangent stiffness matrix. This is a linear dynamic procedure based on results
from a nonlinear static analysis. This approach has won considerable popularity in spite of its
limitations, as one usually only need to consider the first dominant modes of vibration to obtain
sufficiently accurate results. The second
approach evaluates the nonlinear dynamic response using a direct time integration
method combined with a nonlinear solution procedure. Here, the so-called NewtonNewmark algorithm is adopted. This is a much more CPU time-consuming approach.
Still, nonlinear dynamic analysis is essential if it is believed that the bridge will not 1
Discrete damping devices such as viscous dampers and tuned mass dampers. A tuned mass
damper (TMD) is a vibration absorber tuned to a particular mode of the bridge and consists of a
mass, a viscous damper and a linear spring.
100
behave linearly during the application of traffic loads. If this is the case, the natural
frequencies will vary with the amplitude of response and linear dynamic analysis will
consequently be inadequate.
6.2
Cable structures and cable modeling techniques
The increasing attention on cable structures is not only due to their inherent beauty but
also to their stubborn nature in not easily revealing the secret of their nonlinear behavior. Cable
structures exhibit geometrically nonlinear behavior, they are very flexible and undergo large
displacements before attaining their equilibrium
configuration. As an example, due to this inherently nonlinear behavior, conventional
linear dead load analysis which assumes small displacements is often not applicable
[49, 55], except in special cases in the feasibility design stage. In the final design stages
however, a refined nonlinear dead load analysis procedure should be adopted.
A brief early history of the research on the behavior of cables has been published in
[34], as well as a more recent history in [39]. Methods of static and dynamic analysis and
the behavior of cable structures are thoroughly presented in [20, 27, 34, 49]. In
[31, 32, 39], trial-and-error search procedures have been proposed for the nonlinear
computer analysis of simple cable problems. For cable roof structures, analyses
methods and several very illustrative design details, e.g. details of connections, are
presented in [14].
For cable supported bridges, i.e. suspension and cable-stayed bridges, the trend today is
to use more shallow and slender stiffening girders combined with increasing span lengths. For
that reason, it is highly desirable in bridge engineering to develop accurate procedures that can
lead to a thorough understanding and a realistic prediction of the structural response. Although
several investigators [2, 6, 7, 12, 13, 22, 40, 41, 55]
studied the nonlinear behavior of cable supported bridges, very few [2, 7] tackled the
problem of using cable elements for modeling the cables.
Commercial finite element codes used in civil engineering today cannot be readily
used for modeling and analysis of modern cable structures as they lack suitable cable
elements that can accurately model the actual cable curvature. As the cable represents a flexible
member with virtually no resistance to applied moments, the idea of
replacing each cable by a bar element with equivalent cable stiffness or by several
101
beam elements with negligible moment of inertia has found wide acceptance and has
been adopted by many investigators and designers using commercial codes.
As the popularity of cable structures has increased, the search for more efficient methods
has intensified and today various other cable modeling techniques, than the crude modeling with
a bar element mentioned above, can be found in the literature. In
[25], a 2-node curved finite element was developed, using cubic polynomial
interpolation functions, and used for the static and dynamic analysis of 3-D prestressed
cable nets. In [60], another 2-node curved finite element was developed using
Lagrangian functions for the interpolation of element geometry. In [11, 49],
derivations of isoparametric cable elements which includes the element curvature are
presented, and in [7] a four-node isoparametric cable element is presented and used for modeling
cables in cable-stayed bridges.
An iterative analysis procedure for cables, based on using exact analytical
expressions for the elastic catenary, was suggested in [56, 57]. This approach was later
adopted by other investigators, developed and used for the analysis of very simple cable
structures [35, 63] and of power transmission lines [64]. In [34], the same
approach was also suggested for the analysis of cable structures with appreciable sag and
the applicability of this method was later demonstrated in [2] on numerical
examples of cable supported bridges.
The cable element used in this thesis is derived using the exact analytical
expressions for the elastic catenary given in [35]. The procedure presented later in section
7.2.1, to derive the element matrices, is similar to that described in [2, 34].
However, the analytical expressions for the elastic catenary adopted here are
somewhat simpler and therefore easier to handle. The presented element can be used for
modeling large sag cables such as suspension bridge main cables, cables in long span cablestayed bridges, cables in large cable roofs, etc., where straight elements are not readily
applicable. Despite the fact that the cable modeling technique based on the expressions given in
[35] has been available for many years it has, at least to the authors knowledge, very seldom
been used for analysis of cable supported bridges.
The expressions of the internal force vectors and tangent stiffness matrices for the
elements used were derived using the Maple software package for symbolic
computations [52]. Samples of these Maple procedures are given in Appendix A.
102
6.3
General aims of the present study
The main objective of the work presented in this part is to study the response of more
realistic, than those studied in Part A, two-dimensional bridge models considering bridge
damping, exact cable behavior, nonlinear geometric effects, non-uniform cross-sections, and
variable material properties. For this reason a more straightforward and general approach, based
on the nonlinear finite element method, is developed to handle such analysis difficulties and
allow a thorough study of the moving load problem of cable supported bridges.
The main aims of this study are as follows:

to implement two approaches (linear dynamic and nonlinear dynamic) for
evaluating the response and to find whether linear dynamic traffic load analysis is
adequate when investigating the behavior of cable supported bridges under the
action of moving traffic
to better understand and outline the influence of different parameters on the
behavior of cable-stayed bridge. The parameters that are believed to significantly
influence the dynamic response and therefore considered in this study are: bridge-vehicle
interaction, road surface roughness, vehicle speed, number of vehicles on
the bridge, and bridge damping
to investigate the influence of the bridge girder supporting condition on the
response of the bridge. In addition, to study the effect of multi-element cable
discretization2 on the dynamic response
to study the efficiency of a so-called tuned mass damper (TMD) on suppressing
traffic-induced vibrations and increasing the overall damping of the bridge.
Since, in this study, the main concern is to investigate the dynamic response of bridges
and not the dynamics of the vehicle itself and since the spans of cable supported
bridges are considerably larger than the vehicle axle base, a very simple vehicle model is
adopted. This model is an improved version of the one used in Part A as it here includes both
primary and secondary vehicle suspension systems. Since the main aim is not to develop design
formulas for calculating the dynamic amplification factors, 2 Each cable is modeled with several
catenary cable elements to include cable modes of vibration and the dynamic interaction between
the vibrating cables and the bridge.
103
only hypothetical trains of moving traffic are simulated and used for analysis. The effect
of using more complex vehicle models, than the one mentioned above, or finer bridge models
(more elements for discretizing the bridge girder and pylons) have not been investigated in this
study.
For the purpose of this study, computer codes, fully capable of handling the above stated
important factors and parameters, are implemented using the MATLAB language
[53]. The developed codes have been tested by comparing results against those
obtained using the commercial finite element code ABAQUS [1].
Parts of the work presented in the following chapters, concerning nonlinear finite element
modeling, static dead load analysis, and frequency analysis of cable supported bridges, have been
included in the paper Some Modeling Aspects in the Nonlinear
Finite Element Analysis of Cable Supported Bridges. This paper has been accepted
for publication in the journal Computer & Structures.
104
Chapter
______________________________________________________________________
Nonlinear Finite Elements
______________________________________________________________________
7.1
General
The cable supported structures considered in this study are cable-stayed and
suspension bridges. Such bridges consist of cables, pylons and girders (bridge decks) and
are usually modeled using beam and bar elements for the analysis of the global structural
response [28, 41]. In the following, an alternative approach is presented where accurate and
efficient cable and beam elements are used for the modeling. All sources of geometric
nonlinearity, i.e. change of cable geometry under different
tension load levels (cable sag effect), change of the bridge geometry due to large
displacements, and axial force-bending moment interaction in the bridge deck and
pylons (P- effect), are considered in the present study.
For simplicity the present study focuses on two-dimensional problems. Consequently,
torsional effects and torsional modes of vibration are disregarded. As modern cable supported
structures are flexible three-dimensional structures, two-dimensional models are of course not
adequate when studying the response of such structures under the action of environmental loads
like wind, traffic, and earthquakes [4, 28]. However, simplified two-dimensional bridge models
are still very useful for bridge designers in the preliminary design stage, e.g. for investigating the
feasibility of alternative structural solutions. For the interested reader, accurate three-dimensional
cable and beam elements can be found in [35, 61].
The formulation of cable and beam element matrices will be described in the following
sections. These are included for the sake of clarity and also for the purpose of having a thesis as
self-contained as possible. The matrices will be given in the element local coordinate system. For
each individual element in the model, the evaluated element
105
matrices in the local coordinate system are transformed to global coordinate system by
the usual coordinate transformation technique [11]. The structure matrices (i.e. the global tangent
stiffness matrix Kt , global mass matrix M, and global internal force vector p) are constructed
from the transformed matrices of the individual elements of the structure by the general assembly
procedure [11]. The superscript e, used later to denote e.g. the element length Le or the element
nodal displacement vector q e, is omitted in this chapter for notational convenience.
7.2
Modeling of cables
The problem of analyzing cables under different configurations and loading conditions is
very complex. This is because the behavior of cables is inherently nonlinear and also because
large displacements introduce nonlinearities in the geometric sense.
For cable supported bridges, there are mainly three approaches used today to consider the
nonlinear behavior of the cables. In the first approach each cable is replaced by one bar element
with equivalent cable stiffness. This approach, often adopted when
modeling cables in cable-stayed bridges, is referred to as the equivalent modulus

approach and has been used by several investigators [6, 12, 13, 22, 42, 55, 68] and also in
Part A of this thesis. Derivation of the equivalent tangent modulus of elasticity for the bar
element, equation (2.10), is presented in [27]. It has been shown in [7] that the equivalent
modulus approach results in softer cable response as it accounts for the sag effect but does not
account for the stiffening effect due to large displacements. Still, for some cases, e.g. for short
span cable-stayed bridges, linear analysis utilizing the equivalent modulus approach is often
sufficient [27, 41], especially in the feasibility design stage. Whereas, long span cable-stayed
bridges built today or proposed for
future bridges are very flexible, they undergo large displacements, and should
therefore be analyzed taking into account all sources of geometric nonlinearity.
The second approach is to divide each cable into several straight elements, as in [3], in
order to adequately model the curved geometry of the cable. This would introduce
many added degrees of freedom with a consequent increase in computer storage
requirements and computational cost. In addition, numerical problems can occur and
spurious results can be obtained if equilibrium conditions, at those nodes, are not fulfilled.
106
The third approach to model cables is to use isoparametric elements. In [7, 11, 49],
derivations of isoparametric cable elements, which include the element curvature, are presented.
Using such elements one can model the curved geometry of a cable with
fewer elements compared to using straight elements and obtain a better convergence
[49]. However, those elements are stiffer and require numerical integration to
formulate the element stiffness matrix [49].
The alternative approach presented in this thesis is based on exact analytical
expressions for the elastic catenary. In contrast to other modeling techniques
mentioned above, each cable may be represented by a single 2-node finite element,
which accurately consider the curved geometry of the cable, making this method very
attractive for static response calculations. Even if each cable must be divided into several
catenary cable elements, to include cable modes of vibration in the dynamic analysis or external
loads acting between cable ends, the author still believes that this approach is more efficient to
adopt. This is mainly because fewer internal nodes need to be defined for each cable in the
model. The main advantages of the proposed cable element are the reduction of degrees of
freedom, the simplicity of finding the dead load geometry of the cable system, the exact
treatment of cable sag, the exact
treatment of cable weight as it is included in the equations used for element
formulation, and the simplicity of including the effect of pretension of the cable by
simply giving the unstressed cable length.
7.2.1 Cable element formulation
The procedure presented in this section determines the complete geometry of the
cable, the cable element internal force vector, and its tangent stiffness matrix from a
given unstressed cable length and given positions of the ends of the cable.
Consider an elastic cable element, stretched in the vertical plane as shown in Figure 7.1,
with an unstressed length u
L , modulus of elasticity E, cross section area A, and
weight per unit length w (uniformly distributed along the unstressed length). For the
elastic catenary, the exact relations between the element projections and cable force
components at the ends of the element are [35]:
107
u4,P4
u3,P3
node j
y
Ly
Lu,E,A,w
u2,P2
u
x
1,P1
node i
Lx
Figure 7.1
Catenary cable element
L
1
P+T
u
4
j
L=
x
1
P
+
ln
(7.1)
EA w
T
i
2
P
1
L=
y
(T2T2
j
i)
T
T
j
i
+
(7.2)
2 EAw
w
where i
T and T j are the cable tension forces at the two nodes of the element. For the above
expressions it is assumed that the cable is perfectly flexible and Hookes law is applicable to the
cable material. The expressions for x
L and y
L in equations (7.1) and
(7.2) may be written, in terms of the end forces 1
P and 2
P only, as:
x
L=x
L(1
P,2
P ); y
L=y
L(1
P,2
P ) (7.3)
because P , P , P , P ,
1
2
3
4
i
T
T
and
j are related by the following equations:
4
P=
w u
L2
P ;3
P=1
P (7.4)
108
2
2
T
2
i=
1
P+2
P;
2
Tj=
3
P+4
P (7.5)
Differentiating equation (7.3) and rewriting the results using matrix notation gives: L
x
x
d
y
y
x
L=
d1
P+
d2
P;dL=
dP+
d P (7.6)
y
1
2
1
P
2
P
1
P
2
P
L
x
x
L
dL
x
P
P d P
dP
1
2
1
1
=
F
(7.7)
d L
L
L
y
y d 2
P
d 2
P
P
P
1
2
where F is the flexibility matrix. The stiffness matrix K is given by the inverse of F as: 1
1
k
k 2
K=
F=
(7.8)
3
k
k 4
The tangent stiffness matrix Kt and the corresponding internal force vector p for the
cable element can now be obtained in terms of the four nodal degrees of freedom as (noting that
k 2 = k 3 ):
k
1
k2
1
k
k2
1
P
k
k
k
2
P
K=
4
2
4
t
;p=
(7.9)
k
k
P
1
2
3
sym.
k
P
4
4
The element tangent stiffness matrix Kt relates the incremental element nodal force
vector { 1
P,2
P,3
P,P
}T
4
to the incremental element nodal displacement vector
{ 1 u, u
2, u 3, u
}T
4
. The Maple software package for symbolic computations [52]
was used to perform the above mentioned operations and produce the necessary
Fortran code. This Maple procedure is listed in Appendix A. However, if this package is
not available the following expressions, obtained by derivation of equation (7.3), may be used to
evaluate the matrices K, Kt , and p:
1
L
1P
P
k=
u + 4 + 2
1
(7.10a)
det F
EA
wT
T
j
i
109
1
1
1
k=k=
P 1
2
3
(7.10b)
det F w T
T
j
i
1
L
1P
P
k=
x + 4 + 2
4
(7.10c)
det F
P
wT
T
1
j
i
2
L
P
P L
P
P P
u
1
4
2
F=
+
x
1
4
2
+
+
1 1
1
det
(7.10d)
EA
wT
T
P
wT
T
wT
T
j
i
j
i
j
i
To evaluate the tangent stiffness matrix Kt , the end forces 1
P and 2
P must be
determined first. Those forces are adopted as the redundant forces and are determined,
from given positions of cable end nodes, using an iterative stiffness procedure. This procedure
requires starting values for the redundant forces. Based on the catenary relationships the
following expressions will be used for the starting values [35]: w
x
L
w
cosh
1
P=
;P=
2
L
+
y
u
L (7.11)
2
2
sinh
where
2
L2
L
u
y
=3
(7.12)
2
x
L
In cases where equation (7.12) cannot be used because the unstressed cable length is less
than the chord length, a conservative value of 0.2 for is assumed [35]. Another difficulty arises
in equation (7.12) for vertical cables. In that case an arbitrary large value of 106 for is used.
Using equations (7.4) and (7.5), new cable projections corresponding to the assumed end forces
1
P and 2
P are now determined directly from
equations (7.1) and (7.2) and the misclosure vector { L x , L y}T is evaluated as the
positions of the end nodes are given. Corrections to the assumed end forces can now be made
using the computed misclosure vector as:
i+1
i
1
P
x
L
1
P
1
P
1
P
=K
;
=+
(7.13)
2
P
y
L
2
P
2
P 2
P
110
where the stiffness matrix K is given in equation (7.8) and i is the iteration number.
For the present study, this iteration process continued until L x and L y are less than 5
1 10
. As will be demonstrated later, this iterative procedure converges very rapidly.

To determine the unstressed cable length, u
L , for cases where the initial cable tension
is known instead, a similar iteration procedure can be adopted. A starting value for the
unstressed cable length is assumed, e.g. equal to the cable chord length, and cable end forces 1
P and 2
P are computed using the iterative procedure described above. Using
equation (7.5), cable tension can now be computed. This is then compared with the
given initial tension to obtain a better approximation for u
L for the next iteration step.
If the complete geometry of the cable is to be determined, coordinates for a number of
points along the cable must be computed. This is very simple because 1
P and 2
P are
now known, so equations (7.1) and (7.2) can be used to compute the coordinates of
any new point along the cable by simply replacing u
L by any fraction of u
L.
For the dynamic analysis, mass discretization is simply done by static lumping of the
element mass at both ends giving the following lumped mass matrix ( is the mass
density of the cable element):
1 0 0
AL
0100
M=
u
(7.14)
2
0 0 1
0 0 0 1
7.2.2 Analytical verification
A cable hanging under its own weight and subjected to a tensile force at both ends along
its chord, as shown in Figure 7.2, was studied to verify the cable element and the analysis
procedure described in section 7.2.1. This problem was earlier studied in [7]
using isoparametric cable elements and the published results can now be compared to
results obtained here.
A cable with an unstressed length u
L = 312.7 m, modulus of elasticity E = 1.311011
N/m2, cross section area A = 5.4810-4 m2, and weight per unit length w = 46.11 N/m,
111
was studied using two different models. For the two models: the cable was replaced by
one catenary cable element, and by twenty beam elements with negligible moment of
inertia. The beam element used is described in the next section.
To
To
304.8 m
Figure 7.2
Cable under its own weight subjected to tensile force at both ends
To span the distance of 304.8 m,
using a cable with the above given
1.5
(a)
properties, a horizontal force of
To = 1.7794104 N was needed at
both ends. This force gave a mid1.0
span cable sag of 30.48 m and was
1 catenary cable element
adopted as the initial force when
20 beam elements, I =10-5
0.5
20 beam elements, I
calculating the curves in Figure 7.3.
=10-4
Using the two models, the sag and
0.0
the longitudinal displacement along
0
5
10
15
20
(Displacement / Horizontal length) x100
Tension / Initial tension
the chord of the cable were deter10.0
mined for different values of the
(b)
tensile force T and the results are
8.0
1 catenary cable element
plotted in Figure 7.3. Good agree20 beam elements, I =10-5
ment is observed when comparing
20 beam elements,
6.0
I =10-4
the curves for the adopted catenary
cable element with those for the iso4.0
parametric cable element presented in
2.0
[7]. Figure 7.3 shows also that the
replacement of the curved cable by
(Sag / Horizontal length) x100 0.0
several beam elements with negligi0
5
10
15
20
ble flexural stiffness can give
Tension / Initial tension
acceptable results. As the number of
Figure 7.3 Response of the cable defined
in Figure 7.2
112
beam elements increase and the moment of inertia decreases, the results should
converge to those of the catenary cable element model. One should only bear in mind that
replacing the cables by several straight elements will give a stiffer structural model and
consequently an underestimation of the displacements.
The results from this simple numerical experiment provide confidence in the
application of the catenary cable element for modeling cables in cable supported
bridges.
7.3
Modeling of bridge deck and pylons
The pylons and the bridge deck girder or stiffening girder as it is also called are
modeled using beam elements able to resist bending, shear, and axial forces. For the present
study, the simplest 2D beam element introduced in [61] is adopted and the
treatment given there will be followed below when deriving the element matrices. This
finite element is developed following the total Lagrangian approach and using a linear
interpolation scheme for the displacement components. Previous studies, reported in
[61], have shown that the element is efficient and accurate. This element is chosen also
because it can handle large displacements and shear deformations and because it is simple to
formulate the element matrices.
Referring to Figure 7.4, the current deformed configuration of the beam axis is
described by a regular curve defined by the position vector:
s ( x)
o
= [ x + u( x)]i + (
w x )j (7.15)
where the abscissa x[0, L] is measured on the straight reference configuration of the
beam, u( x), w( x) represent the axial and transversal displacement components and i and j are
unit axis vectors. By introducing the angle ( x) as the rotation of the cross sections (S') in the
deformed configuration, the unit vectors orthogonal and parallel to the cross sections for each
point on the deformed beam are obtained as:
a( x) = cos i + sin j; b( x) = sin i + cos j (7.16) Further, defining the deformation
measures , , according to:
113
u5
y, w
u6
u4
b( x)
s
a( x)
o, x
deformed beam
u2
( x)
u3
u1
S'
s
s( x)
w( x)
o( x)
u( x)
j
S
i
undeformed beam
x, u
x
L
Figure 7.4
Deformed and undeformed configuration of the beam element
ds
d
s
o
o, =
=
x
(1+ )a + b ; =
= (7.17)
dx
x
x
,
d
and using equations (7.15) and (7.16), the following expressions are obtained:
= (1+ , ux)cos
+ w sin 1
,x
(7.18)
= w cos
,x
(1+ , ux)sin
(7.19)
= , x (7.20)
If the constitutive relations are assumed as linear, the strain energy can be written as: 1 L
(u) =
i
(
2
2
2
EA + GA + EI )d
x (7.21)
20
114
where EA, GA and EI represent the axial, shear and flexural rigidities. For u, w and a
linear interpolation scheme is used according to:
u = Ni ( x q
)i+Nj(xq
) j (7.22)
where T
u={u ,w
, }, T
q={
T
1
u , u 2 , u 3} and q = { u 4 , u 5 , u 6} contain the corresponding i
j
values of the displacements at the two nodes of the element, and N ( x) = 1 x /
L
i
;
N ( x) = x L
j
/ are the interpolation functions.
Finally, the expressions for the internal force vector p and the element tangent stiffness
matrix Kt are obtained through successive differentiation of the expression for the strain energy
according to:
2
i
p
i
p=
;
Kt=
=
(7.23)
2
q
where q is the nodal displacement vector { 1

u ,u2,u3,u4,5
u , u }T
6
. The Maple procedure
which performs the above mentioned operations and produces the necessary Fortran
code for p and Kt , is listed in Appendix A.
The kinetic energy is expressed as the integral over the volume V :
=1
k
s&( x)Ts&( x) V
d (7.24)
V
2
where is the mass density and s&( x) represents the velocity in a general point of the
beam. The position of this point is defined by the vector s( x) as shown in Figure 7.4.
For this element, the final expression for the kinetic energy becomes [23]:
L
2
L
2
L
=
Au&( x) dx +
A w&( x) dx +
I &( x 2
) dx
k
(7.25)
2
2
2
0
0
0
Using the interpolation functions, as in equation (7.22), the kinetic energy is written as a
function of the velocity components in the nodal degrees of freedom of the element.
From the resulting expression for the kinetic energy, the consistent element mass
matrix is evaluated as:
115
2 A
0
0
A
0
0
0
2A
0
0
A
0
2
L 0
0
2I
0
0
I
M=
k=
(7.26)
2
q
6
&
A
0
0
2A
0
0
0
A
0
0
2A
0
0
0
I
0
0
2 I
For more details concerning the formulation of the kinetic energy and mass matrix and
the performance of this element in dynamic problems, the reader is referred to [23].
116
Chapter
______________________________________________________________________
Vehicle and Structure Modeling
______________________________________________________________________
8.1
Vehicle models
Vehicles have at least two suspension systems. Thus, in order to improve the earlier
model adopted in Part A, an additional suspension system is introduced. Consequently, the
improved model includes not only the body-bounce motion, as the one in Part A, but also the
wheel-hop motion. The two suspension systems, each consisting of a
spring and a damper element, can be seen as filters. The first suspension system
(primary suspension) reduces the road input into the vehicle structure and creates
isolation for frequencies higher than about 15 Hz [9]. The second suspension system (secondary
or chassis suspension) produces isolation for frequencies higher than about 2 Hz for air
suspended and 3 Hz for leaf spring suspended chassis. This system should also give sufficient
damping to the axle and tire system to prevent the tire leaving the ground on very rough road
surfaces. Other suspension systems, e.g. cab suspension

which is used to reduce the acceleration levels in and the forces on the cab structure, are
also present in modern heavy vehicles. However, the simple vehicle model adopted does not
contain all suspension systems, detailed suspension nonlinearities and
complexities of vehicle body motion that are typical of heavy vehicles. Despite that, it is
believed that this model is sufficiently realistic, as was discussed in Part A, for the purpose of
this study.
For the current study, it is assumed that the vehicle never loses contact with the bridge,
the spring and the viscous damper have linear characteristics, and the contact between the bridge
and the moving vehicle is assumed to be a point contact. Referring to Figure 8.1a, and denoting
the contact force between the bridge and the vehicle by F( t), defined positive when it acts
downward on the bridge, the following dynamic
equilibrium equations for the three masses can be established:
117
w ( t)
3
v( t)
m3
(a)
(b)
k
c
ss
w ( t)
2
m2
k
c
pp
w ( t)
1
m1
node i
xc
node j
mode 1
mode 2
Le
1.5-4 Hz
8-15 Hz
Figure 8.1
(a) vehicle model on a bridge element; (b) typical modes of vibration
3
m 3
w& + ks( 3
w2
w ) + cs( &3
w w& ) = 0
2
(8.1a)
2
m 2
w& + ks( 2
w3
w ) + cs( &2
w &3
w ) + kp( 2
w1
w ) + cp( &2
w w& ) = 0
1
(8.1b)
1
m 1
w& k p ( 1
w2
w ) c p ( &1
w w&2 ) ( 1
m + m 2 + m ) g + F( t) = 0
3
(8.1c)
Where 1
w,2
w and 3
w are the vertical displacements of the masses measured from the
static equilibrium position, kp and ks the stiffness of the linear springs connecting the
masses, c and c the damping coefficients of the viscous dampers, and g the p
s
acceleration of gravity. A dot superscript denotes differentiation with respect to time.
Using the equations above, the contact force may be expressed as:
F( t) = ( 1
m+2
m+3
m)g+1
m 1
w& + m 2 2
w& + 3
m 3
w& (8.2)
where the first term on the right-hand side is the dead weight (static part) of the contact
force while the other terms represent the inertia effects. The moving force and the moving mass
vehicle models can now be obtained, as shown in Part A, by modifying
the contact force expression, equation (8.2). When modeling suspended roadway
vehicles, the unsprung mass, m 1, is always set to zero in this study. This mass is still
included in the model to be able to model a vehicle as an unsprung moving mass (so-called
moving mass model) or for modeling railway vehicles having unsprung wheels.
118
With the assumption that the road profile cannot be rough enough to make the vehicle
jump or leave the road surface, the displacement w 1( t) is not an independent variable but can be
coupled at each time step to the displacement of the contact point on the bridge deck. In this way
this degree of freedom is eliminated and, as in Part A, the displacement w 1( t) and its derivatives
are expressed in terms of the nodal degrees of freedom of the bridge as:
e
w=N
1
cq+c
r (8.3a)
w
e
&=Nq v
e
+N q& + r v
1
c, x
c
c, x
(8.3b)
w
e
&=N
q v2
e
+ N q& v
e
2+N
q a
e
+N q& + r
v2+ra
1
c, xx
c, x
c, x
c
c, xx
c, x
(8.3c)
where N c is a row vector containing linear interpolation functions for the vertical
displacement of the beam element evaluated at the contact point xc. As shown in Figure 8.1a, xc
is the distance from left node of the element to the contact point. q e is the nodal displacement
column vector for the element on which the vehicle is
positioned, v and a the vehicle velocity and acceleration in the longitudinal direction, and
rc the surface irregularity evaluated at the contact point. The subscript x denotes derivation with
respect to x.
Equations (8.1a) and (8.1b) may be written using matrix notation as:
3
m
0 &3
wc
c
s
s & 3
wk
k
s
s
3
w
0
+
+
=
(8.4)
0
m 2 &2
w c c + c
s
p
s & 2
w k
k+k
s
p
s 2
w c w& +
p 1
kp 1
w
The vehicle model has consequently two degrees of freedom and its equation of
motion can be written as:
M vw& + C vw& + K vw = f v (8.5) In order to solve this second order equation of
motion, it is transformed into a system of first order equations by complementing it with the
equality Iw& Iw& = 0 , where I is a 2x2 identity matrix. This will give the extended dynamic
equilibrium equation:
119
I
0 w& 0
Iw 0
+
= (8.6)
0 M v w& K v C v w& f v
which can be rewritten as:
w&
0
I
w 0
=
or & =
+ (8.7)
1
1
+ 1
u
Au
B
v
v
w& M K
v
v
M v C v w& M v f v
For the present study, this first order differential equation is solved using the Matlab
algorithm ode45 [53] which uses automatic step size 4th and 5th order pair Runge-KuttaFehlberg integration method.
The undamped natural frequencies for this simple sprung mass vehicle model is
determined from the eigenvalue equation:
det(
2
K
)=0
v
Mv
(8.8)
giving the eigenvalues:
2
1 kp
k
m
k
s
p
k
m
4kk
3
s
3
ps
=
1,2
(8.9)
2
+
1+
+
1+

m
m
m
m
m
m
mm
2
3
2
2
3
2
2 3
For typical vehicle mass and stiffness values, the first vibration mode (body-bounce)
excites mainly the vehicle body mass, m 3, while the second mode (wheel-hop) almost only
excites the wheel mass, m 2, se Figure 8.1b.
Heavy roadway vehicles generate most of their dynamic wheel loads in two distinct
frequency ranges, as mentioned earlier in Part A of this thesis: body-bounce and pitch
motions at 1.5-4 Hz and wheel-hop motion at 8-15 Hz, se Figure 8.1b. The body-bounce mode of
a vehicle is excited by relatively long and the wheel-hop mode by
relatively short wavelengths of the road surface irregularities. As an example, the bodybounce mode for a vehicle traveling at 70 km/h is excited by irregularities with wavelengths of
4.9-13 m and the wheel-hop mode by irregularity wavelengths of 1.3-2.4 m. Moreover,
depending on the vehicle speed, a surface irregularity of a certain wavelength may be effective in
both frequency ranges.
120
8.2
Vehicle load modeling and the moving load algorithm
The moving load problem is more complicated than other problems in structural
dynamics, as the external force vector, containing the interaction forces existing at the
contact points between the vehicles and the bridge, is time-dependent. This vector is totally
unknown beforehand (except for the moving force problem which is
uncoupled), as interaction forces are dependent on the motion of both the bridge
structure and the vehicles.
The external force vectors for the elements where vehicles are positioned are obtained by
adding the dead load and the moving traffic (live) nodal load vectors, i.e.
e
e
e
f = f dead + f live . Using, as before (see Part A), the Dirac function ( x ci x ) to
characterize the action of a unit force concentrated in point x = ci
x , the traffic part can
be expressed as:
a
Le
e
f live = i
F ( t)
a
T
N ( x x

ci ) d x = i
F ( t) T
N ci (8.10)
i 1
=
0
i 1
=
where a is the total number of vehicles on the element, Le the element length (see Figure
8.1a), Fi( t) the interaction force between the bridge and the i th vehicle wheel, x is the distance
measured from left node of the element, and N ci the row vector:
x
x
N=,
0 1
,
,
0
,
0
, 0 (8.11)
e
e
L
L
containing the element interpolation functions evaluated at the contact point of vehicle i,
i.e. for x = ci
x . For consistency reasons, the same linear interpolation functions are
adopted here as those used in section 7.3 for deriving the beam element matrices.
The moving load algorithm, illustrated in Figure 8.2, calculates the bridge-vehicle contact
force, for each vehicle on the bridge, and prepares the global external force vector caused by the
moving traffic. The code developed is capable of handling
unlimited number of vehicles, all having the same velocity (independent on the
deflected shape of the bridge deck), and fully considers the dynamic interaction
between the vehicles and the bridge.
121
START
Call algorithm for

each iteration.
Input: initial vehicle
vectors w
Initialize global external
0 and w
&0,
traffic load vector, f
=0
new vehicle positions,
live
and surface roughness
parameters rc, rc,x, rc,xx
Loop through all
vehicles on the bridge
For vehicle i, initialize
w0 i, w&0 i , M vi, C vi and K vi Identify the loaded element. Evaluate
xci, the vectors N ci, N ci,x, N ci,xx and determine the loaded elements dof
No
Interaction
problem
Yes
Determine e e e
q , q& , q& and calculate
w i, w
1
& i, w
1
& i 1 form equations 8.3
Solve differential equation 8.7 to
Moving force model
F
calculate w i, w& i, w& i . Evaluate contact i = ( m + m + m
1
2
3)
g
force, Fi( t), from equation 8.2
Accumulate nodal forces
f live( dof 1 dof 6) = f live( dof 1 dof ) i F ( t) T
6
N ci
Next vehicle
RETURN f live
Figure 8.2
The moving load algorithm
122
8.3
Bridge structure
The bridge structures are discretized, for the nonlinear finite element analysis, using a
catenary cable element and a simple beam element able to resist bending, shear and axial forces.
As discussed earlier, this cable element is adopted in order to simplify cable modeling and to
more accurately predict the response of cable supported bridge structures. The modeling
technique and derivation of the elements are described in detail in Chapter 7. Only twodimensional bridge models are considered in this study, consequently, the torsional behavior
caused by eccentric loading of the bridge deck is disregarded. Furthermore, the nonlinearity
considered is of the geometric type, as linear material behavior is assumed.
8.3.1 Modeling of damping in cable supported bridges
The assumption that bridges have no damping (adopted in Part A) is rather restrictive
since vibration energy is dissipated in all structures even in very low damped cable-stayed
bridges. There are various factors causing energy dissipation in bridge
structures. Energy dissipation is generally developed by material nonlinearity, opening
and closing of hair cracks (in reinforced concrete structures), structural damping such as friction
at movable bearings and in the joints of the structure, loss of energy from foundation to ground,
and aerodynamic damping by friction with air. The different
forces that contribute to the damping of a structure may vary with vibration amplitude,
velocity, acceleration, and stress intensity. Thus, theoretical evaluation of damping is extremely
difficult and hence at present we have to rely upon the empirical approach.
Attempts on theoretical approach to damping evaluation of cable-stayed bridges have
been made in [43, 44] where a method is proposed to determine the damping ratio for the desired
mode shapes by evaluating the overall energy dissipation and strain energy in the whole bridge.
It was shown that the proposed method predicted the dependency of the damping ratio on
oscillation amplitude and cable type with reasonable accuracy.
For practical use however, damping models that represent more or less satisfactory
approximations have been introduced. The most commonly employed damping model
in the field of structural engineering is the viscous one, in which the damping force is
proportional to the velocity. This model, also adopted for the present study, is very popular as it
leads to the simplest mathematical treatment and generally gives the most satisfactory results
[15].
123
In the case of cable-stayed bridges, the dissipation of energy in cables is potentially a
major contribution to the overall damping of the bridge. As pointed out in [65], the damping ratio
depends on the type of cables and on the cable configuration used.
Using spiral or locked coil strands instead of parallel wire strands (PWS), increases the
total damping capacity of the bridge as such strand exhibits significant hysteresis in its
longitudinal load-displacement behavior. Despite this, parallel wire strands have been favored in
resent years due to economic and durability reasons.
A limited amount of information is available on damping of cable supported bridges.
However, as more and more forced-excitation and ambient-vibration tests are
conducted, fundamental data on the damping ratio is increasing drastically.
Observations made from various tests suggest that the damping ratio for a suspension
bridge decreases with increasing frequency (mode number) but for cable-stayed
bridges it is found to be invariant with frequency [19, 37, 43]. In [19], this trend is
discussed by interpreting many observed data and it was concluded that friction
damping in the main cables of suspension bridges could be a possible explanation that
distinguishes cable-stayed form suspension bridges.
To derive the damping matrix, one frequently used technique is to assume the damping
matrix proportional to mass and stiffness matrices giving the so-called Rayleigh
damping [8, 15], C = K + M . For practical problems the parameters and are often
chosen based on the knowledge of the damping property of a similar structure.
One disadvantage of the Rayleigh damping is the fact that damping behavior of the
complete bridge structure is described only by the two parameters and , which are
taken as constant values over the entire bridge model. On the other hand, this method gives
damping matrices that have the same orthogonal properties with respect to the eigenvectors of
the undamped system, as the mass and stiffness matrices. Thus, by
premultiplying and postmultiplying with the mode shape matrix, the damping matrix
can be diagonalized giving the matrix of generalized (or modal) damping values3 (see
also section 9.1.1.2):
3 In the present study the eigenvectors are normalized such that the generalized (or
modal) mass is set to unity.
124
2 1
1
0
0
2]
0
222
0
=
= ZT CZ
(8.12)
0
0
2
n n
Due to the earlier mentioned limitation of the Rayleigh damping, the damping matrix
derived gives a poor simulation of the real damping characteristics of the bridge
structure. An improvement may be achieved by ignoring the proportionality condition
and instead establish directly the diagonalized matrix, [
2
], of equation (8.12). To
derive a damping matrix with orthogonal properties, this method is much more flexible
than the Rayleigh damping and is preferred if modal damping ratios are available.
Another advantage, when compared to Rayleigh damping, is that only the damping
ratios of the required modes can be accounted for. The actual expression for the
damping matrix may be deduced from the assumed matrix [
2
], as:
T
C=Z
[2] 1
Z (8.13)
The damping matrix obtained is full regardless of whether the stiffness and mass
matrices are banded or not, but this is of little significance. In practice, the inversion of
the mode shape matrix requires a large computational effort. This can be avoided by taking
advantage of the orthogonality properties of the mode shapes relative to the mass matrix. Thus,
by using the expression3 ZTM Z = I to express T
1
Z
and
Z in
equation (8.13), one can obtain:
C= M Z[
] ZT
2
M (8.14)
For the present study, the i th mode damping ratio in equation (8.12) is assumed as:
0 84
.
i
= 0.0042
i
(for suspension bridges [19])

(8.15)
2
.
0 645
1
= 2 =
L
237
.
0
=
L
n
(for cable-stayed bridges [43])
(8.16)
where L is the main span length in meter. The above two approximations are based on
field forced-excitation tests.
125
8.3.2 Bridge deck surface roughness
Two types of roughness profiles are considered in this study. The first type is a random
profile described using a power spectral density function. The generation of such a profile is
described in detail in Part A, Chapter 2.
For the second roughness profile, a bump, assumed to vary harmonically as shown in
Figure 8.3, is generated to simulate roughness of the local type. The profile is
evaluated, for x
x(x
+L
) , as:
bump
bump
bump
1
2(xx
)
r( x) =
h
1
bump
cos
(8.17)
bump
bump
L
where xbump is the position where the bump starts, hbump the maximum height, and
Lbump the length of the bump. It should be noted that here x and xbump are measured starting
from the left end support of the bridge.
h
r( x)
bump
x
xbump
Lbump
Figure 8.3
Bump dimensions
To obtain a realistic simulation of vehicles moving over a bump, the developed
algorithm temporary reduce the time step as a vehicle reaches the bump. This is done
automatically, if the time step is too large, to ensure that at least a specified number of contact
points (10 for the results reported here) are obtained along the bump.
126
8.4
Tuned vibration absorbers
A vibration absorber is a device that reduces the vibration level of a protected
structure. Usually such a device consists of an additional mass connected by means of an
elastic and a damping element to the structure needed protection. The invention of the vibration
absorber is usually associated with the name of Frahm, who in 1909 first patented a vibration
absorber design [21].
There are innumerable examples of vibration absorbers being applied to control
vibrations in various engineering structures. For civil engineering structures like
chimneys, TV towers, and high-rise buildings, vibration absorbers are nowadays
widely used mainly to reduce vibrations caused by wind or earthquakes, see for
example [50, 54, 66]. For bridges, such devices are still not that common. However, as
modern bridges become longer and more flexible and as energy dissipation in these
bridges is very low (see section 8.3.1) and often not enough on its own to suppress
vibrations, it is realized that additional vibration control measures are needed. The primary
intention, in the few examples found where vibration absorbers are practically applied on long
span bridges, was to control wind-induced vibrations. As such bridges not only are vulnerable to
vibrations caused by wind and earthquakes but also to those caused by traffic, some
investigations have been carried out to find if such devices are also efficient for controlling
traffic-induced vibrations [10, 18, 33, 38, 45]. In [45] the efficiency of a tuned mass damper
(TMD4) on suppressing vibrations induced by high-speed trains moving over a simple three-span
bridge was studied and in [10] the
efficiency of TMDs for footbridges was investigated theoretically and experimentally
using walking, jumping and running tests. In [33] the effect of a TMD on the dynamic response
of the Rama IX cable-stayed bridge5 in Bangkok, subjected to a moving
vehicle, was studied. In the present study, the efficiency of a TMD is also investigated but
this is done more correctly here as, among other things, the bridge-vehicle
interaction was not fully considered in [33].
Other types of absorbers and devices used to control vibrations can also be found in 4 A
TMD is a vibration absorber tuned to a particular mode of the bridge and usually consists of a
mass, a viscous damper and a linear spring as illustrated in Figure 8.6. For convenience, the
abbreviation TMD is used hereafter for such vibration absorbers.
5 The Rama IX cable-stayed bridge is equipped with TMDs for suppressing windinduced flexural and torsional vibrations in the bridge deck as well as flexural vibrations in the
pylons [33].
127
bridge structures. As an example Figure

8.4 illustrate how cable vibrations in
bridges can be controlled. Such design
(a)
measures for suppression of cable
vibrations can be found on existing
bridges. For example, auxiliary ropes
are found on the Far cable-stayed
bridge in Denmark and the Normandie
(b)
bridge in France, dashpot dampers
(shock absorbers) are installed between
bridge deck and stay cables on the
Brotonne bridge in France and the
Sunshine Skyway bridge in Florida, and
(c)
stockbridge dampers are mounted on
the long hangers of the Humber
suspension bridge in England [58].
Figure 8.4 Design measures for suppression of cable vibrations [58]
TMDs are also used temporarily during
erection of long span bridges, as free-standing pylons and cantilevered spans are very
sensitive to wind and often need vibration control. For example, two TMDs each
weighting about 40 tonnes, see Figure 8.5a, were employed during the erection of the 856
m main span (currently word record) Normandie cable-stayed bridge to stabilize the bridge
structure during erection and limit the moment in the pylons. Wind tunnel tests and calculations
showed that, by using TMDs, the dynamic response due to wind could be reduced by
approximately 35 % [16]. As construction progressed, these
TMDs were relocated and tuned to maximum efficiency. The TMDs were designed so
their eigenfrequency could be adjusted from 0.1 Hz to 0.2 Hz and the relative internal
damping from 3 % to 18 %. More information can be found in [51]. Most recently, in the two
approach bridges for the Great Belt suspension bridge in Denmark, 32 TMDs
have been installed, see Figure 8.5b, each with a mass of 8 tonnes (which
approximately equals 0.5 % of the modal mass) [59].
There are two important roles of a TMD: firstly, it reduces the resonance response of the
main structure, and secondly, the attached dashpot increases the overall damping of the structure
by providing an additional source of energy dissipation. The efficiency of the TMD depends on
the correct tuning of its parameters (eigenfrequency) relative to
128
(a)
(b)
Figure 8.5
TMD of the (a) Normandie Bridge during erection [16]; (b) Great Belt
Bridge [59]
the eigenfrequancy of the mode of the structure it is designed to suppress. Thus, one
disadvantage with such a passive device is that its performance significantly
deteriorates when the dynamic characteristics of the structure are different from the
original values assumed during the optimal design of the TMD. On the other hand,
TMDs can still be advantageous as they work without requiring any connection to the
ground or external power supply. In the literature, a great deal of attention is given to the
optimization of parameters and evaluation of the efficiency of the TMD. For the present study,
the following most often used optimum tuning parameters, derived in
[21] for a structure with no damping, are adopted6:
=
i
tmd
(8.18a)
1+
6 These equations are used in the present study to determine the TMD parameters without
considering available standard components and dimensions. However, in practice dimensions of
available standard springs etc. often dictate the final choice of TMD parameters.
129
3
=
tmd
(8.18b)
8(+)3
1
where tmd
and i are the circular frequencies of the TMD and the dominant bridge
mode to be tuned to, tmd the damping ratio of the TMD, and is the mass ratio which
relates the TMD mass to the modal mass of the dominant bridge mode to be
tuned to, = tmd
m
/i
m . These equations are derived for undamped structures affected
by a stationary harmonic load. Consequently, they may not be valid for a damped
bridge structure subjected to moving traffic loads and one maybe should, using
equations (8.18a-b) as starting values, perform a trail and error computer study to find the
optimal TMD parameters. However, in [38] it was found that the difference is
small and for practical design purpose these optimization formulas are suitable, as far as
the bridge is lightly damped. But as expected, the response reduction was not as great as for the
case with a stationary harmonic load.
The exact tuning of the TMD frequency is, as mentioned earlier, very important,
whereas, the damping needs only be assumed approximately as its efficiency is
relatively less sensitive for variations in tmd [10]. The mass ratio , which is usually
small and lies between 0.01-0.04 for bridges [10, 45], is another important design parameter as it
directly influences the response of the bridge and the relative
movement of the TMD mass (Lower values of lead to large relative movement of
the TMD mass). Therefore, when designing a TMD it is important to carefully select a
suitable mass ratio, as it is not always easy to find a location with significant relative movements
and enough space to accommodate the TMD. Another engineering
problem is to provide a very low friction bearing surface for the TMD mass, so that the
damper mass can also respond to the bridge movement at low levels of excitation.
ktmd
ctmd
wtmd
mtmd
Figure 8.6
Cross section of bridge girder with a tuned mass damper, TMD
130
Now referring to Figure 8.6, the following equilibrium equation can be established for the
TMD mass including only the inertia effect:
m
w&
+
tmd tmd
tmd
c
( w&
e
tmd
N tmdq& )+ ktmd ( w

e
tmd
N tmdq )= 0 (8.19)
where mtmd is the mass of the TMD, ctmd the damping coefficient of the viscous
dampers, ktmd the spring stiffness, wtmd the vertical displacement of the TMD mass measured
from the static equilibrium position, q e the nodal displacement vector for the element on which
the TMD is attached, and N tmd the interpolation vector given in equation (8.11). This vector is
evaluated for x = tmd
x
where xtmd is the distance from
the left node of the TMD loaded element to the point where the TMD is attached. The
spring stiffness and damping coefficient are evaluated using equations (8.18a-b). To include the
force applied to this beam element by the TMD, an addition to the external force vector given in
equation (8.10) is made as:
e
e
T
f live = f live tmd
m
tmd
w&
N tmd (8.20)
It is interesting to note that when a vehicle is modeled as a sprung mass system instead of
a constant moving force (i.e. including bridge-vehicle interaction) one can for some cases
observe a reduction in the bridge dynamic response as the vehicle acts as a
vibration absorber [29]. Thus, for such cases the dynamic response, calculated by
ignoring interaction, is conservative.
131
132
Chapter
______________________________________________________________________
Response Analysis
______________________________________________________________________
9.1
Dynamic Analysis
In the previous chapter, the vehicle equation of motion, equation (8.5), and an
algorithm for assembling the global external force vector are presented. In the
following, the general equation of motion for a bridge under the action of moving
traffic loads is formulated.
Based on the element property matrices and the external nodal force vector derived in the
foregoing chapters, one may obtain the following general equation of motion for the entire
bridge by means of conventional FEM assemblage:

M q& + Cq& + p(q) = f (q, q&, q,& t) (9.1) q , q&, q
&
bridge node displacement, velocity, and acceleration vectors
M
bridge mass matrix
C
bridge damping matrix
p(q)
vector of internal elastic forces
f( q, q&, q& , t) external force vector resulting from the dead load, the moving traffic,
and the tuned mass dampers
As indicated, the external force vector is not only time dependent but is also dependent
on the bridge displacements, velocities and accelerations. This vector contains the interaction
forces existing at the contact points between the vehicles and the bridge and thereby couples the
bridge equation of motion with those of the vehicles.
Consequently, the external force vector is dependent on the motion of both the bridge
structure and the vehicles. To consider the bridge-vehicle interaction, an iterative
133
procedure is adopted. First the displacements etc. of the contact points are assumed and
the vehicle equations are solved to obtain the interaction forces and the external force vector.
Then the bridge equation of motion, equation (9.1), is solved to obtain improved values of
displacements etc. for the contact points. If the convergence
criteria are not fulfilled, a new iteration is preformed recalculating the external force
vector and resolving the bridge equation. The type of convergence criterion used in Part A is also
adopted here to check the errors in not only the displacement vector, as in Part A equation (3.14),
but also the velocity and acceleration vectors. The time step and the number of considered
eigenmodes (for the mode superposition procedure) are problem-dependent. Accordingly, for
each numerical example studied here, a
convergence study was carried out with a view to getting reasonably converged
reliable solutions with an optimum number of increments and optimum number of
considered eigenmodes.
In the present study, two approaches are adopted for solving equation (9.1) in the time
domain: one for evaluating the linear dynamic response and one for the nonlinear
dynamic response. These are described in detail in the following sections.
9.1.1 Linear dynamic analysis
If the structure is assumed to respond linearly during the application of traffic loads, it is
possible to evaluate the dynamic response using the mode superposition technique starting from
the deformed dead load state. Thus, this linear dynamic procedure
utilizes the dead load tangent stiffness matrix, Kt . The internal force vector in equation
(9.1) is therefore evaluated as:
pq
()=Kq
t
(9.2)
where Kt is obtained from a nonlinear static dead load analysis. It is believed that this
approach is adequate for short and medium span cable supported bridges as far as
traffic load to dead load ratios are small. Moreover, it is well known that this approach
give sufficiently accurate results with minimum consumption of CPU time, as usually one only
need to consider the first dominant modes of vibration. On the other hand, this approach requires
frequency analysis and eigenmode extraction to start with,
which can be expensive and time consuming for large systems. Furthermore, the
134
effectiveness of mode superposition technique is reduced, as the required number of
eigenmodes for satisfactory convergence is difficult to estimate beforehand.
The eigenmode extraction and the mode superposition technique are well described in
many textbooks on structural dynamics [11]. Nevertheless, in order for this thesis to be fully
contained and to facilitate understanding, the above mathematical concepts will be described in
some depth in the following subsections.
9.1.1.1 Eigenmode extraction and normalization of eigenvectors
The first step of this approach is to determine the natural frequencies and mode shapes of
the bridge structure. For this, the undamped free vibration is considered, as the effect of damping
on the natural frequencies of most real structures is small. Thus equation (9.1) is reduced to:
M q& + K q 0
t
= (9.3)
in which 0 is a zero vector. Assuming harmonic motion, which may be expressed as
[11]:
q = z sin( t ) (9.4)
where z is a vector with components independent of time, is a circular frequency, and
is a phase angle, and substituting into equation (9.3), we obtain the following eigenvalue
problem:
(K2
M) z 0
t
= (9.5)
In the present study, this eigenvalue problem is solved using the Matlab algorithm eig
[53] giving n (= number of d.o.f.) eigensolutions ( 2
, z ),( 2
, z ),L,( 2
,)
1
1
2
2
n z n , where
z i is called the i th mode shape vector (eigenvector) and i is the corresponding
circular frequency of vibration ( 2
i is the eigenvalue). Storing the obtained eigensolutions in two matrices Z and 2
, we can now write the n solutions to equation (9.5)
as:
K Z M Z 2
0
t
= (9.6)
135
where the matrix Z contains the eigenvectors z i in columns and the matrix 2
is a
diagonal matrix, which stores the eigenvalues on its diagonal, i.e.:
2
Z=[1
z , z2,L, z n]
2
and =
2
(9.7)
2
n
The obtained eigenvectors are then normalized such that the modal mass is set to
unity, i.e.
T
=zMz=1
i
m
i
i
. From this M-normalization and the orthogonality properties
of eigenvectors [11] it follows that:
T
T
2
ZMZ=I
(9.6)
equation
from
and
ZKZ
t
=
(9.8)
where I is identity or unit matrix. As noted, one of the advantages of this normalization
method is that the modal stiffnesses will be equal to the eigenvalues.
9.1.1.2 Mode superposition technique
When the natural frequencies and mode shapes are determined, the equation of motion
can be solved. The vector of nodal displacement, q, can be approximated by a linear combination
of s eigenvectors, with s much less than n, giving: s
q z i i ( t) = Z s s (9.9) i=1
where i in vector
s are the modal amplitudes and state the proportion of each
eigenvector in the transformation. Substitution of the above expression into equation
(9.1) and postmultiplication of each term by T
Zs yields:
ZT
T
T
T
&
sMZ s
&
s+Z sCZ ss+Z sKZ
t
s s = Z s f live (9.10)
Note that the external force vector in the above equation only contains the external forces
resulting from the moving traffic and the tuned mass dampers. From the
136
properties of M-orthonormalized eigenvectors, equation (9.8), the above equation is

reduced to:
I
T
2
T
&
&
s + Z s C Z s s + s = Z s f live (9.11) The damping matrix C is introduced to
approximate the overall energy dissipation of the bridge structure during vibration and cannot
generally be constructed from element damping matrices, such as the mass and stiffness
matrices. To obtain a system of s uncoupled differential equations the matrix C must have the
same orthogonal properties, with respect to the eigenvectors of the undamped structure, as the
mass and stiffness matrices. When this is the case, the damping matrix will be diagonalized by
premultiplying and postmultiplying with the mode shape matrix giving:
T
Z s CZ s =
{
diag c , c ,L, c
1
2
s}=
{
diag
2,
2 ,L,
2
11
22
s s } (9.12)
where c , c ,L, c
1
2
s are the modal damping values and are here expressed using the
damping ratios7 , ,L,
1
2
s . The obtained matrix is referred to as the matrix of
generalized (or modal) damping values. One of the advantages of this method when
compared to other methods like Rayleigh damping is that there is no need to assemble a
damping matrix with the same dimension as the stiffness and mass matrix. With the introduction
of this matrix we can obtain a system of s independent equations which can be written as:
2
& + 2 & + = zT f
1
111
1
1
1 live
2
& + 2 & + = zT f
2
222
2
2
2 live
......
..........
..........
..........
..........
(9.13)
......
..........
..........
..........
..........
2
T
&
&
s + 2 s s s + s s = z s f live Using the Maple software package for symbolic
computations [52], a program code for the exact solution of the above equations was generated.
This code is incorporated in the developed computer program for linear dynamic analysis. Initial
values for
equations (9.13) are obtained, from known initial values of q and q& , by premultiplying
both sides of equation (9.9) with zT
i
and taking note of equation (9.8) giving:
7 is the fraction of critical damping in mode i.
i
137
T
T
=
&
o
zMq
i
i
o
and
o = z Mq
i
i
& o (9.14)
Finally, having determined the elements in vector s by solving the above s equations,
the response due to live loads (i.e. traffic and TMD) is found from equation (9.9). By adding the
dead load response, determined using nonlinear static analysis, the total response of the bridge is
obtained.
9.1.2 Nonlinear dynamic analysis
A procedure for evaluating the nonlinear dynamic response using a direct time
integration method combined with a nonlinear solution procedure, is described in the
following. This method is significantly more expensive (i.e. CPU time-consuming)
than the mode superposition method which is usually chosen for linear studies. Still,
nonlinear dynamic analysis is essential for the present study if its is believed that the bridge will
not behave linearly during the application of traffic loads. If this is the case, the natural
frequencies and mode shapes will vary with the amplitude of response and the earlier described
mode superposition technique will consequently be
inadequate.
For this procedure, it is necessary to assemble not only the stiffness and mass matrices,
but also the damping matrix. A method for modeling the damping in matrix form in
terms of modal damping ratios is presented in Chapter 8. As the stiffness is a function of
the response amplitude, frequencies and damping ratios will also change as the
structure deforms, thus theoretically such damping matrix ought to be updated for each
time step. In practice, however, this is usually not necessary, because the damping ratios used
will in most cases be only approximate values evaluated from field tests, taken from codes of
practice or the literature.
For this study, an implicit procedure based on Newmarks modified average
acceleration method combined with a full Newton-Raphson solution procedure is
adopted for solving the bridge equation of motion. Among engineers and researchers in
the field of structural engineering, the Newmark method presented 1959 is the most popular
algorithm for numerical solution of the equation of motion and is generally the most suitable for
nonlinear analysis [7, 23]. In the following, the adopted procedure will be briefly described. For
the interested reader, more details concerning the
derivation etc. can be found in [26].
138
Denoting the time step by t, for the Newmark method the following expressions hold
for the displacements and for the velocities:
21
2
t+ t
=qt+t
q& t + t
q& t + t
q& t+ t (9.15a)
2
q& t+ t
= q& t + t
(1 )q& t + t
q& t+ t (9.15b)
where
and
are the Newmarks integration parameters and are evaluated for the
modified average acceleration method as = 5
.
0 + and
2
= 0.25 1
( + ) , with > 0 .
The parameter introduces numerical damping to the solution but reduces the
accuracy to the first order. The second order accurate average acceleration method is
obtained by setting = 0 (no numerical energy loss).
In analogy to the static nonlinear analysis, the residual (error in nodal forces or dynamical
out-of-balance forces) can be specified by rewriting equation (9.1) to give (note that f also
includes the dead load):
err
r
(q) = Mq& + Cq& + p(q) f(q, q&, q,& t) (9.16) These should vanish at the end of each
incremental time step, i.e. r err (q t+ t ) = 0 , by means of an iterative solution technique. For
the present study, a procedure using full8
Newton-Raphson iterations is adopted for solving the residual equation. Denoting
k
q t+ t
as the approximate value of q t+ t resulting from iteration k, the residual after k+1
iterations is approximated by linearization around the previous iterated solution as: err
r
( k+1
+
q
1
t+ t
)
err
=r(k
q t+ t
)+ S( k
q t+ t
)( k
k
q t+ t q t+ t
) (9.17)
The matrix S in the above equation is the iteration (jacobian) matrix and is deduced from:
8 For the full Newton-Raphson iteration procedure, the tangent stiffness and iteration
matrices are updated in each iteration.
139
&
&
S(
r
q
q
f
q t+ )
err
kt=
= M
+C
+K
t
(9.18)
q
q
q
q k
k
q
q t+ t
t+ t

The last term f
/q
reproduces the effect of displacement-dependent external loads
and is, for simplicity, omitted in the present study, see [26]. Making now use of
equation (9.15), the iteration matrix can be specified in detail as:
1
S(q) =
M+
C + K t (9.19)
t 2
t
At each time step, the iteration has to be started from a prediction of the
displacements, velocities and accelerations. This is done here by using equation (9.15)
and setting q& t+ t = q
& t . In each iteration, the displacement corrections are computed

using equation (9.17) and setting the residual at iteration k+1 equals zero, giving: k
1 err
q = S r ( k
q t+ t
) (9.20)
The corrected displacements, velocities, and accelerations are then computed as:
k+1
k
k
q t+ t = q t+ t + q
(9.21a)
k+
k
q
&
k
k
k
q& 1
t+ t = q
& t+ t +
q
=q
& t+ t +
(9.21b)
q
k+1
k
q
&
k
k
1
k
q& t+ t = q& t+ t +
q
=q
& t+ t +
q
(9.21c)
q
t
2
By iteratively updating the solution using the corrections
k
q
, the error in nodal
forces, i.e. the residual vector err
r
, is reduced when proceeding form iteration k to k+1.
At each time step, this iteration process continues until the ratio of the Euclidian norm of
the residual vector to that of the internal force vector has dropped below a certain chosen
tolerance value. For the described procedure, choosing an adequate time step is a critical issue, as
this not only drives the accuracy of the integration but also governs the stability of the iteration
process. As mentioned earlier, since the time step is
140
problem-dependent, a convergence study was carried out for each numerical example
to find the optimum number of increment.
9.2
Static analysis
In this study, the dead load response is always evaluated using a nonlinear procedure,
while the live load static response is evaluated either linearly based on the dead load tangent
stiffness matrix or nonlinearly, i.e.:
nonlin
q static, tot = q dead+ live (9.22)
and/or
nonlin
lin
nonlin
q
1
static, tot = q dead + q live = q dead + K

f
t
live (9.23)
To evaluate the nonlinear static response, the stiffness matrix is reformulated as the
bridge deforms. There are several procedures for evaluating the nonlinear static
response and the most frequently used are the Newton-Raphson iteration schemes
[11, 17, 55]. In the present study, an incremental-iterative procedure using full
Newton-Raphson iterations is adopted. This procedure is generally expected to give
quadratic convergence [17].
141
142
Chapter
______________________________________________________________________
Numerical Examples
______________________________________________________________________
In this chapter, static and dynamic responses of a simply supported bridge, a long span
suspension bridge (the Great Belt suspension bridge), and a medium span cable-stayed bridge,
under the action of moving vehicles, are presented. Frequency analysis was also conducted for
all three examples. In the last two examples, the applicability of the catenary cable element, for
modeling cables in cable supported bridges, is
demonstrated. The nonlinear behavior of cable supported bridges during erection and
application of the dead load is also presented. To verify and illustrate the efficiency of the
developed finite element code, some of the obtained results are compared to those obtained using
the commercial finite element code ABAQUS [1] and to results reported by other researchers.
Details concerning the modeling technique used in ABAQUS to model the moving loads and the
bridge structures are given in Chapter 4. The

calculations were preformed on a Pentium Pro 200 MHz computer with 128 MB RAM
and for each numerical example the CPU time used by the MATLAB process is given.
Throughout the dynamic investigation, the bridge initial velocity and acceleration vectors
at time t = 0 were assumed to be zero. The dead load displacements obtained from the nonlinear
static analysis were used as the bridge initial displacements.
However, if nothing else is mentioned, the bridge deck initial dead load vertical
displacements were assumed to be zero, when evaluating the bridge-vehicle contact
forces. Zero initial values were also assumed for the moving vehicles vertical
displacements and vertical velocities. Further, for convenience, the shear modulus of all
beam elements in the models was evaluated as G = E / 2.6 .
A convergence study was carried out for each numerical example with a view to
getting reasonably converged reliable solutions with optimum number of increments
143
and optimum number of considered eigenmodes (for the mode superposition
procedure). For the convergence study, the analysis was repeated using a smaller time
step (more increments) until good agreement is shown between the last two analysis results. To
find the optimum number of eigenmodes, to be considered in the mode
superposition procedure, the results were also checked after repeating the analysis
including more eigenmodes.
For the linear and nonlinear dynamic analysis results presented in the following study, the
tolerances for the three criteria, used to control the convergence of the bridge-vehicle dynamic
interaction (similar to equation (3.14) in Part A, see also section 9.1), were set to 0.01. For the
nonlinear dynamic procedure, the tolerance for the residual based convergence criterion (see
section 9.1.2) was set to 10-5. These tolerance values were chosen from the practical engineering
point of view.
As the main aim of this study is not to develop formulas to be used in design
specifications for calculating the dynamic amplification factors, not much emphasis was
put on simulating realistic trains of moving traffic or finding the most critical points on the
bridge with the highest dynamic amplification factors.
10.1 Simply supported bridge
The simply supported bridge studied earlier in Part A of this thesis was also adopted for
this investigation. The response of this simple bridge model was mainly studied in order to verify
the developed algorithms. This was done by comparing the results with those obtained using the
commercial finite element code ABAQUS. The bridge was discretized, for both the ABAQUS and
the present model, using 70 beam elements. For the present bridge model, the first four bending
natural frequencies obtained utilizing the dead load tangent stiffness matrix are: 3.97, 15.41,
33.19, and 55.91 Hz. As
expected, due to considering shear effects and the utilization of the deformed dead load
tangent stiffness matrix, these are lower than the frequencies given in Table 4.1.
The properties for the bridge and the vehicle models are given in Figure 10.1a. The
modeled vehicle, permitted in Sweden but not within the European community, is a 24
m long truck and trailer with a total weight of 60 ton. As seen, this vehicle was
modeled as two separate sprung mass systems one for the truck and one for the trailer.
Thus, as a result of this simplification, the interaction between the truck and trailer was
144
disregarded and only vertical modes of vibration of the vehicle were considered. The
body-bounce and wheel-hop frequencies for the truck and the trailer models were
chosen as 1.89 Hz and 11.35 Hz, with the corresponding mode shapes shown in Figure
8.1b. The vehicle model was assumed to move, from the left to the right, at the
constant speed of 25 m/s (90 km/h) on a rough road surface having the profile shown in
Figure 10.2a.
In Figure 10.1b-d and 10.2b-d, the ABAQUS and the present mode superposition
solutions, due to only traffic load, are plotted on the same diagrams and the results are found to
be in very good agreement with each other. About 30 % increase in the
vehicle weight is observed in Figure 10.2b as a result of having about 9 mm higher
surface at the bridge entrance than the approach pavement. Only 4 modes were
considered when producing the present solution curves, except for the bending
moment, which required 10 modes to converge. No bridge damping was included,
= 0 , except the automatically introduced numerical (artificial) damping in the
ABAQUS solution. This numerical damping is believed to cause the difference in Figure
10.1d for the mid-point vertical acceleration. The present linear and nonlinear static traffic load
responses were found to be identical. The problem was also solved using the present direct
integration method but the results are not shown as these were found to be nearly identical to the
mode superposition results, except for the mid-point vertical acceleration. For the present direct
integration solution, to obtain a stable mid-point vertical acceleration curve with good agreement
to the curves in Figure 10.1d, it was essential to introduce numerical damping by setting the
parameter to be equal to 0.05. For the three solutions, 300 increments were chosen with a time
step of about 6 ms. In Table 10.1, dynamic amplification factors (DAF) for mid-point9 vertical
displacement and bending moment are given together with the corresponding CPU
time and maximal iterations per increment required for solving the problem.
For the results in Figures 10.1 and 10.2 it was assumed that, as this is usually the case, no
initial dead load vertical displacements in the bridge deck were present. It is normally attempted
to compensate by precambering the bridge deck during
construction so the dead load displacements are cancelled by the specified camber. If no
compensation is made during construction, the initial displaced geometry of the 9 For a certain
point P, the DAF is defined as the ratio of the absolute maximum live load dynamic response at P
to the absolute maximum live load static response at the same point.
145
(a)
v = 25 m/s
Truck and Trailer
0.0
(b)
m = 0 kg
ABAQUS
1
m = 4400 kg
Present
2
m = 35600 kg
-1.0
Static (present)
3
24 m
c = 1.6 10
. 4 Ns/m
p
c = 12.0 10
. 4 Ns/m
s
-2.0
.
k = 14.0 106 N/m
p
m
m /2
3
3
k = 8.0 10
. 6 N/m
s
k
c
k /2
c /2
-3.0
ss
s
s
Bridge
m
m
2
2 /2
E = 3.0 10
. 10 N/m2
k
c /2
k
c
pp
p /2 p
I = 3.307 m4
-4.0
m
m
1
1 /2
12 m
A = 4.75 m2
-5.0
m = 11400 kg/m
34 m
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
Time (s)
0.0
0.5
(d)
ABAQUS
)2
-1.0
Present (10 modes)
ABAQUS
(c)
Static (present)
0.3
Present
-2.0
0.0
-3.0
-0.3
-4.0
Bending moment (MNm)
-5.0
-0.5
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
Time (s)
Time (s)
Figure 10.1 Mid-point vertical displacement (b), bending moment (c), and acceleration
histories (d) for the problem defined in (a). The road surface roughness profile is shown in
Figure 10.2a. The dashed vertical lines indicate when the trailer enters and the truck leaves the
bridge
15.0
1.4
(a)
(b)
ABAQUS-truck
1.3
ABAQUS-trailer
10.0
Present-trailer
1.2
Present-truck
5.0
1.1
1.0
0.0
Normalized contact force 0.9
Surface roughness (mm)
-5.0
0.8
0
2
4
6
8 10 12 14 16 18 20 22 24 26 28 30 32 34
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
Distance along the bridge (m)
Time (s)
25.0
ABAQUS-truck
ABAQUS-truck
(c)
2.0
(d)
ABAQUS-trailer
)
ABAQUS-trailer
2
Present-trailer
15.0
Present-trailer
Present-truck
Present-truck
1.0
5.0
0.0
-5.0
-1.0
-15.0
-2.0
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
Time (s)
Time (s)
Figure 10.2 Normalized bridge-vehicle contact force (b), truck and trailer body vertical
displacement (c), and acceleration histories (d), when travelling over a bridge having the road
surface profile shown in (a).
The dashed vertical lines indicate when the trailer enters and the truck leaves the bridge
0.0
(a)
No initial displacement (linear)
20.0
(b)
No initial displacement (linear)
With initial displacement (linear)
With initial displacement (linear)
With initial displacement (nonlinear)
-1.0
With initial displacement (nonlinear)
10.0
-2.0
0.0
-3.0
-10.0
-4.0
-20.0
-5.0
-30.0
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
0.00
0.25
0.50
0.75
1.00
1.25
Time (s)
Time (s)
Figure 10.3 Effect of initial dead load displacements on mid-point vertical bridge
displacement (a) and on vertical displacement of the truck body (b)
0.0
(a)
= 0 (linear)
-0.5
=
0 (linear)
= 0.05 (linear)
=
0.05 (linear)
(b)
-1.0
= 0.05 (nonlinear)
= 0.05 (nonlinear)
(MNm) -1.5
-2.0
-2.5
-3.0
ng moment
-3.5
-4.0
Bendi
Vertical displacement (mm) -5.0

-4.5
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
Time (s)
Time (s)
Figure 10.4 Effect of damping on mid-point vertical bridge displacement (a) and
bending moment (b) bridge deck must be considered, as this will influence the bridge-vehicle
contact force and thereby the bridge and vehicle responses. This is shown, using the two present
analysis methods, in Figure 10.3 where mid-point vertical bridge displacements, due to traffic
load only, and truck body displacements, are plotted. The effect of damping is shown in Figure
10.4, where the undamped solution and the solution with 5 %
damping ratio, assumed constant for all modes, are plotted.
Mode Superposition (present) Direct Integration

4 modes
10 modes
(present)
DAF d 1.142
1.145 1.143
DAF m 1.095
1.151 1.154
CPU time (s)
60
85
830
Max. iterations/inc.
3
3
2
Table 10.1
Dynamic amplification factors for mid-point vertical displacement and
bending moment with the corresponding required CPU time and
maximal iterations per increment
The comparisons, made using this simple numerical example, provide some
confidence in the application of the implemented direct integration and mode
superposition algorithms. Therefore, in the following sections, only analysis results
obtained using these two algorithms will be presented and no comparison will be made to results
from commercial codes like ABAQUS.
10.2 The Great Belt suspension bridge
In this numerical example, the behavior of the Great Belt suspension bridge during girder
erection and under the action of moving vehicles is presented. A frequency
analysis for the completed bridge was also conducted.
The Great Belt (Storeblt) suspension bridge in Denmark has, with its 1624 m main
span, the second largest span in the world10. The bridge deck the girder or stiffening 10
Akashi Kaikyo suspension bridge in Japan holds currently the worlds record with a main span
length of 1991 m. Both the Great Belt bridge and the Akashi Kaikyo bridge were opened for
traffic in 1998.
149
girder as it is called is a 31 m wide and 4 m deep steel box. This girder is continuous
over the full cable supported length of 2694 m, i.e. without expansion joints at the pylons. Like
many other modern suspension bridges, the girder section is streamlined (almost wing shaped) to
resist strong wind action. The two main cables, each include a total of 18648 high strength
galvanized steel wires 5.38 mm in diameter and have a final diameter of 0.85 m, were erected in
the autumn of 1996 by the air-spinning
method11 [27]. To minimize deflections under asymmetric traffic load, the main cables
are fixed to the stiffening girder through rigid clamps at mid-span. The ratio of cable sag to main
span length was chosen to be 1/9 as this was found economically
favorable. This gives higher pylons than usual and therefore a more flexible structure.
The total height of the concrete pylons, including cable and cable saddle, is
approximately 258 m.
180
77.6
535 m
1624 m
535 m
Figure 10.5 Geometry of the Great Belt suspension bridge

For the results presented here, a simple bridge model consisting of 126 elements, 89
nodes and a total of 215 active degrees of freedom, was used. The bridge geometry
shown in Figure 10.5 and the properties given in Table 10.2 were taken from [28, 47].
The section properties for the pylons were assumed by the author to give the first and
second pylon frequencies 0.147 Hz and 0.803 Hz, which are acceptable when
compared to the measured pylon frequencies reported in [46]. In the bridge model, the
girder was assumed to be simply supported at the ends, and every third hanger from the original
bridge was included and modeled using catenary cable elements. The main cables were assumed
to be fixed at the pylon tops and the pylons were rigidly fixed to the piers. To simplify the data
input process, all internal cable nodes were vertically 11 Much the same method was used for
suspension bridges more than 100 years ago, e.g. for the Brooklyn bridge in New York dating
from 1883.
150
positioned at the same level as the girder, and the configuration of the main cable under
dead load was determined accurately, during analysis, after few iterations.
E (N/m2)
A (m2)
I (m4)
w (t/m)
Girder
2.11011 1.00
3.32
14.78
(31 m 4 m)
Pylons
0.41011
2 37.5
2 750
2 90
0 75.5 m
Pylons
0.41011 2 (32.5, 30, 25) 2 (275, 200, 150)
2 (78, 72, 60)
75.5 257.6 m
Cable
2.11011
2 0.41
2 3.45
side spans
Cable
2.11011
2 0.40
2 3.36
main span
Hangers
2.11011
2 0.025
Mass of hangers and clamps are considered distributed uniformly along the main cable
and included in the cable mass.
Table 10.2
Parameters for the model of the Great Belt suspension bridge
10.2.1 Static response during erection and natural frequency analysis
Erection of a suspension bridge involves many challenging problems, especially
aerodynamic stability problems, which relate to the fact that the bridge structure is
incomplete, thus various structural components do not receive or render the kind of support
intended in the complete structure. The erection of the girder may proceed in a number of
different ways [36, 47, 48]. For the analysis presented in this thesis it was assumed that the
erection proceeded simultaneously from mid-span and anchor blocks towards the pylons. To
study the erection procedure of the Great Belt suspension
bridge and calculate the initial profile for the free hanging cables (i.e. the initial cable
lengths, the cable sag at mid-span, and the pylon tops horizontal displacement), the simple 2D
model described earlier is used.
To start the analysis, the initial side span and main span cable lengths were determined
using a trial and error procedure. Thus the initial cable lengths were estimated and the
151
final dead load profile, i.e. cable sag at mid-span and pylon top displacements, were
determined and compared with the desired dead load profile shown in Figure 10.5. The estimated
values were then improved and the calculation was repeated until the
required dead load profile was obtained. It was found that if the initial lengths of the side
span and main span cables were chosen as 564.8 m and 1672.7 m, the calculated mid-span cable
sag and pylon tops horizontal displacements would be 180.09 m and
0.04 m. Those initial cable lengths were therefore considered to be good enough and were
used for all the following results presented for this example.
When the initial cable lengths were
known, different erection stages
185
were analyzed and the results are
(a)
plotted in Figure 10.6. Studying this
figure one can notice that, to arrive
180
specified final cable sag under
at the desired dead load profile, the
dead load =180 m
pylon tops must be displaced about
0.85 m outwards and the cable sag
175
should be about 173 m, for 0 %
erected girder. This pylon top
Cable sag at main span (m)
displacement is needed to counteract
170
0
20
40
60
80
100
the displacement caused by the
% of bridge deck erected
elongation of the side span main
1.5
cable during the subsequent erection
(b)
of the girder. Consequently, to
compensate for the later erection of
1.0
the girder the saddles at the pylon
tops must therefore, prior to cable
erection, be horizontally displaced in
0.5
relation to the vertical pylon axis.
This is done either by displacing the
saddles in relation to the pylon tops
0.0
or by pulling back the pylons with
0
20
40
60
80
100
Pylon tops horizontal displacement (m)

so called tie-back cables or a
% of bridge deck erected
combination of the two methods Figure 10.6 Cable sag variation (a) and
[36]. It is therefore necessary to horizontal displacement of pylon specify this
displacement of the tops (b) during girder erection saddles to arrive at vertical pylons
152
with zero bending in the final dead load condition. During the constructing of the Great
Belt suspension bridge, tie-back cables running from each pylon top to the
nearest anchor block were used pulling back each pylon 1.24 m [28] prior to main
cable erection. When the main cables were erected and the tie-back cables were
dismantled, the pylon tops moved back about 0.20 m [28]. Thus, the remaining
displacement before erecting the girder was about 1 m. Acceptable agreement is found,
according to the authors opinion, when comparing this value with the one obtained from the
present analysis, i.e. 0.85 m. It is worth noting that the aims and the essential goals of this
investigation were to study the bridge response using a simple model and to check the efficiency
and applicability of the presented finite elements for modeling cable supported bridges. For this
reason some simplifications and assumptions were made when modeling the bridge structure and
not much emphasis was put on using the exact properties etc. for each bridge member. This is
believed to be the major
explanation for the differences in the results.
Frequency analysis was also conducted for the completed bridge and the first three
natural frequencies and mode shapes are given in Figure 10.7.
mode 1: 0.099 Hz (0.100 Hz)

mode 2: 0.112 Hz (0.115 Hz)
mode 3: 0.130 Hz (0.135 Hz)
Figure 10.7 Natural frequencies and mode shapes for the lowest three vertical bending
modes of vibration. Values inside brackets are reported in
[48]
153
In [47, 48], analytical frequency results from a 3D finite element model and results from
a 1:200 scale aeroelastic bridge model, made for wind tunnel testing, are
presented. These results are also given in Figure 10.7 within brackets and as can be seen
the agreement is very good when comparing with the result obtained from the
present analysis.
The CPU time used by the MATLAB process, to find the tangent stiffness matrix at the
dead load deformed state and to solve the system eigenvalue problem determining all 215 modes
of vibration, was about 200 seconds.
10.2.2 Dynamic response due to moving vehicles

The simple suspension bridge model described earlier was also adopted to study the
response of the Great Belt Suspension bridge under the action of two moving trucks.
Each truck, moving on a smooth road surface from the left to the right at the constant
speed of 25 m/s (90 km/h), was assumed to have a total weight of 44 ton and a length of 18.75 m.
The free distance between the two trucks was assumed to be 40 m. The
body-bounce and wheel-hop frequencies, for each truck model, were chosen as 1.89
and 11.35 Hz. The corresponding mode shapes are shown in Figure 8.1b and the
model properties are given in Figure 10.8.
The problem was solved using both the
v = 25 m/s
implemented linear dynamic (mode
m
m
m = 0 kg
3
3
1
k
c
k
c
m = 4840 kg
superposition) and nonlinear dynamic
ss
2
ss
m = 39160 kg
m
m
3
2
2
(direct integration) procedures. For the
c = 1.76 104 Ns/m
k
c
p
k
c
pp
pp
c = 13.2 104 Ns/m
s
m
m
two procedures, reasonably converged
1
1
k = 15.4 106 N/m
p
k = 8.8 106 N/m
reliable solutions were obtained using
58.75 m
s
600 increments corresponding to a time Figure 10.8 Model of the two trucks step of
about 0.18 s. For the mode superposition procedure, the first 25 modes were found to be
sufficient for calculating the dynamic response of the bridge and the trucks, as 25- and 30-mode
solutions did not differ significantly. Bridge damping ratios were evaluated according to equation
(8.15) giving: =
;
029
.
0
=
;
026
.
0
1
2
3 =
L
023
.
0
etc. No numerical damping was introduced
in the direct integration procedure.
154
Some interesting diagrams are shown in Figure 10.9 and 10.10 where the linear and
nonlinear solutions are plotted. These figures do not show very significant difference
between the two solutions. Although some differences can be observed between the
linear and nonlinear dynamic moment, acceleration, and contact force curves in
Figures 10.9 and 10.10, these are insignificant from the engineering point of view.
In Table 10.3, dynamic amplification factors for the horizontal displacement of the left
pylon top and the bending moment at the fixed end of the left pylon are given together with the
corresponding CPU time and maximal iterations per increment required for
solving the problem. Even though the excitation of the dynamic system was only
caused by the elastic displacement of the bridge itself, as road surface with no
roughness was assumed, the maximum dynamic moment for the fixed end of the left
pylon was found to be nearly 30 % larger than the static one, see Figure 10.10a.
Linear
Dynamic
Nonlinear Dynamic
(mode superposition)
(direct integration)
DAF d (pylon top) 1.127 1.123
DAF m (pylon fixed end) 1.250
1.274
CPU time (s)
845
5930
Max. iterations/inc.
4
3
Table 10.3
Dynamic amplification factors for the horizontal displacement of the
left pylon top and the bending moment at the fixed end of the left pylon
with the corresponding required CPU time and maximal iterations per
increment
It can be concluded from the numerical results obtained in this study that, utilizing the
dead load tangent stiffness matrix, linear static and linear dynamic traffic load
response analysis of long span suspension bridges is adequate.
The response was also evaluated neglecting bridge damping and it was found that
correct estimation of bridge damping is very important, as this will greatly affect the
dynamic response of the bridge-vehicle system. As an example, the dynamic
amplification factor for the deck vertical displacement at the center of the bridge will
increase from about 1.10 (Figure 10.9b) to about 1.22, if bridge damping is not
considered.
155
30
20
(a)
10
0
-10
-20
Dynamic (nonlinear)
Dynamic (linear)
-30
Static (linear & nonlinear)
Horizontal displacement (mm) -40
0
10
20
30
40
50
60
70
80
90
100
110
Time (s)
100
50
0
-50
(b)
-100
-150
Dynamic (nonlinear)
-200
Dynamic (linear)
-250
0
10
20
30
40
50
60
70
80
90
100
110
Time (s)
0.06
)
(c)
2
0.04
0.02
0.00
-0.02
-0.04
Nonlinear
Linear
-0.06
0
10
20
30
40
50
60
70
80
90
100
110
Time (s)
Figure 10.9 Horizontal displacement of the left pylon top (a); bridge deck vertical
displacement (b) and vertical acceleration (c) at the center of the bridge.
The curves are not easy to distinguish, as the responses are almost identical
156
40
Dynamic (nonlinear)
(a)
20
Dynamic (linear)
0
-20
-40
-60
0
10
20
30
40
50
60
70
80
90
100
110
Time (s)
15
Dynamic (nonlinear)
(b)
10
Dynamic (linear)
5
0
-5
-10
-15
0
10
20
30
40
50
60
70
80
90
100
110
Time (s)
1.08
(c)
Nonlinear
1.06
Linear
1.04
1.02
1.00
0.98
0.96
0.94
Normalized contact force
0.92
0
10
20
30
40
50
60
70
80
90
100
110
Time (s)
Figure 10.10 Bending moment at the fixed end of the left pylon (a) and for the bridge
deck at the left pylon (b); normalized bridge-vehicle contact force for the
first truck (c)
157
10.3 Medium span cable-stayed bridge
A 2D model of the cable-stayed bridge described in [55] was adopted for this
investigation. The bridge geometry is shown in Figure 10.11 and the properties are given
in Table 10.4. This bridge is similar in configuration to an existing bridge in Japan (The MeikoNishi Bridge in Nagoya) with some modifications in dimension.
The static and dynamic behavior of this bridge has been studied earlier by several other
investigators [4, 40, 55].
E (N/m2) A (m2) I (m4) w (t/m)

Girder
2.01011 0.93 0.26 19.64
Girder
2.01011 1.11 1.29 19.64
central part
Pylons
2.81010 13.01 34.52 30.65
above deck level
Pylons
2.81010 18.58 86.31 43.78
below deck level
Links
2.01011 0.56 0.10 4.38
deck to pylons
Including weight of cross beams.

Cable no.
E (N/m2) A (m2)
u
L (m) w (t/m)
1, 24
2.01011 0.0362 158.13 0.398
2, 11, 14, 23
2.01011 0.0232 134.66 0.255
3,10, 15, 22
2.01011 0.0204 111.64 0.225
4, 9, 16, 21
2.01011 0.0176 89.43 0.194
5, 8, 17, 20
2.01011 0.0139 68.80 0.153
6, 7, 18, 19
2.01011 0.0113 51.69 0.125
12, 13
2.01011 0.0372 158.12 0.409
Table 10.4
Parameters for the cable-stayed bridge model defined in Figure 10.11
158
node 43 .
5x3
element 61
cable16
45.7
cable12 cable13
.
..
.
30.5
cable 6
node 7 elem. 36
node 14
element 52
node 36
node 13
node 18
.
element 30
146.3 m
335.3 m
146.3 m
Figure 10.11 Geometry of the cable-stayed bridge. The cables are numbered from the
left to the right starting with cable 1
For the model, it was assumed that the girder was pinned at the ends, i.e. only rotations
were allowed, and elastically connected to the pylons by vertical links. The pylons were assumed
to be rigidly fixed to the piers, and all cables were assumed fixed to the pylons and to the girder
at their joints of attachment. The simplest model analyzed in this example, i.e. the model with
one element per cable, was composed of 66 elements and 43 nodal points.
10.3.1 Static response and natural frequency analysis
Figure 10.12 shows the nonlinear behavior of the model under static dead load,
described in terms of the vertical displacement of the girder at the center of the bridge
and the tension in cable 12 and 13. Examining this figure, a hardening characteristic with respect
to the applied load is apparent. It is also evident that at the start there is a significant nonlinear
behavior during the static application of the dead load. Thus, a nonlinear static analysis under
dead load is essential to arrive at the deformed dead load tangent stiffness matrix. For this cablestayed bridge with modest main span
length, as the nonlinearity is not so strong above this dead load equilibrium point, it is
believed that this bridge will behave as a linear system, when affected by live static and dynamic
loads, starting from this dead load deformed state. This means that
influence lines and superposition technique can be used in the design process.
However, as the span length increases this nonlinearity will get more pronounced [55]
and linear live load analysis might no longer be adequate.
For the frequency analysis, to include cable motions, natural frequencies were also
determined replacing each stay cable by 3, 5, and 7 catenary cable elements. This was
159
5
5
(a)
(b)
4
4
3
3
tangent
tangent
2
eigenvalue
2
problem
Dead load multiplier 1
Dead load multiplier 1
0
0
-3
-2
0
2
3
5
6
0
10
20
30
40
50
60
70
Vertical displacement of the girder at the
Tension in cable 12 and 13 (MN)
center of the bridge (m)
Figure 10.12 Nonlinear behavior of the cable-stayed bridge defined in Figure 10.11: (a)
vertical displacement; (b) cable tension
mode 1: 0.334 Hz (0.311 Hz)

mode 2: 0.437 Hz (0.411 Hz)
mode 3: 0.703 Hz (0.650 Hz)
Figure 10.13 Natural frequencies and mode shapes for the first three vertical bending
modes of vibration. Values inside brackets are reported in
[4]
160
easily done as the preprocessor code developed can automatically refine the model if the
user requests it. For the simplest model, i.e. the model with one element per cable, the resulting
modes of vibration only include the vibrating girder and pylons. Thus, cable modes and the
dynamic interaction between the vibrating cables and the bridge were disregarded. For the finer
models, pure cable modes, i.e. additional new mode shapes characterized only by vibrating
cables, were obtained between those basic
bending modes. Moreover, it is evident that cable motions are associated with every
mode of vibration, as can be noticed in Figure 10.13. For the four alternative models, Table 10.5
presents a comparative frequency study of the first ten vertical bending modes of vibration. The
order in which these modes appear is given inside brackets.
Vertical
Natural frequencies (Hz) and mode order
bending
1
3
5
7
mode no. element/cable elements/cable elements/cable elements/cable
1
0.332 (1)
0.334 (1)
0.334 (1)
0.334 (1)
2
0.436 (2)
0.437 (2)
0.437 (2)
0.437 (2)
3
0.692 (3)
0.700 (7)
0.703 (7)
0.703 (7)
4
0.734 (4)
0.739 (8)
0.741 (8)
0.742 (8)
5
0.868 (5)
0.874 (13)
0.875 (13)
0.875 (13)
6
1.044 (6)
1.051 (18)
1.053 (18)
1.053 (18)
7
1.212 (7)
1.218 (27)
1.220 (23)
1.220 (23)
8
1.214 (8)
1.220 (28)
1.222 (24)
1.222 (24)
9
1.379 (9)
1.388 (33)
1.387 (29)
1.388 (29)
10
1.671 (10)
1.690 (42)
1.680 (38)
1.687 (38)
Table 10.5
Comparison of the first ten natural frequencies for vertical bending
modes of vibration. Each cable was modeled using 1, 3, 5 and 7
catenary cable elements. Mode order is given inside brackets
Satisfactory agreement is found when comparing the results from the static and
frequency analysis presented here with those reported in [4, 40, 55]. The disagreement in
the frequency results shown in Figure 10.13 is believed to be due to the fact that the
161
catenary cable element used in the present study is stiffer12 than the one bar element with
an equivalent modulus used in [4, 55]. Moreover, the girder and pylons were
modeled in [4, 55] using conventional beam elements modified by the stability
functions and a diagonal lumped mass matrix was adopted for all elements.
For the simplest model with a total of 119 active degrees of freedom, the CPU time used
by the MATLAB process, to find the tangent stiffness matrix at the dead load deformed state and
to solve the system eigenvalue problem determining all 119 modes of vibration, was about 15
seconds. This indicates high efficiency of the presented elements.
10.3.2 Dynamic response due to moving vehicles parametric study
For the cable-stayed bridge shown in Figure 10.11, a study was conducted to assess the
importance of different factors that influence the dynamic response due to moving
vehicles. The investigation was focused on the following factors that are believed to have
some kind of influence on the dynamic amplification factor: the effect of bridge-vehicle
interaction, number of vehicles on the bridge, surface irregularities at the bridge entrance, vehicle
speed, bridge damping, cable modeling, tuned vibration
absorbers, and girder supporting conditions.
Some interesting figures are presented at the end of this numerical example, where
element and node numbers are referred to. The locations of these nodes and elements are shown
in Figure 10.11. The corresponding dynamic amplification factors (DAF)
for the different studied cases are given in Table 10.6 at the end of this numerical
example. Also, the absolute maximum bridge deck vertical acceleration at the center of the
bridge and the maximum normalized bridge-vehicle contact force are given for
each case in this table. The plotted cable axial forces are the average of the axial forces at
the two ends of the cable. Also, one should observe that the dead load response (always
evaluated using a nonlinear procedure) was subtracted from the solutions, thus only the
responses due to traffic loads are plotted. The linear and nonlinear static traffic load responses
were found to be almost identical and are not plotted separately for the sake of making the
figures more clear. However, these static solutions
corresponding to the linear and nonlinear dynamic solutions were used when
12 This can be concluded when comparing Figure 7.3a with the one published in [7]
showing the response for the one bar element with an equivalent modulus of elasticity.
162
calculating the DAF. The following was adopted for the parametric study unless
otherwise specified:
The simplest bridge model, i.e. the model with one element per cable.
Vehicles moving from the left to the right at the constant speed of 25 m/s (90 km/h) on a
smooth road surface.
The same 44 ton vehicle model as for the Great Belt suspension bridge example.
The body-bounce and wheel-hop frequencies for the truck model were chosen as
1.89 and 11.35 Hz. See Figure 10.8 for the model properties.
1000 increments corresponding to a time step of 0.025 s.
The first 30 modes were considered for the mode superposition solution.
No numerical damping was introduced in the direct integration procedure.
Bridge damping ratios were evaluated according to equation (8.16) giving:
= =
L
= .0056
0
1
2
30
.
10.3.2.1 Response due to a single moving vehicle
In this study, linear and nonlinear dynamic responses due to a 44 ton moving truck are
compared. In Table 10.6 (original configuration), some results for different time steps
corresponding to 500, 1000, and 1500 increments are given. A linear mode
superposition analysis, including all 119 modes of vibration and using 1000
increments, was also performed for this case and the results are also given in Table 10.6.
During the convergence study, for the mode superposition procedure, it was
found that the 30- and 35-mode solutions did not differ significantly. However, as seen
from the results in Table 10.6, there are higher modes than 35 that are affecting the response. For
example, for the bending moment at the right pylon fixed end, the DAF
decreases from 1.180 to 1.159 when considering all 119 modes of vibration.
Comparing the 500, 1000, and 1500 increment solutions it is clear that the 500solution is not good enough and the 1000-solution does not give perfectly converged
results. However, from the practical engineering point of view, analyses with 1000
increments and 30 modes give reasonably converged reliable solutions. Consequently,
unless otherwise specified, 1000 increments and 30 modes were adopted for this
investigation.
163
Some linear and nonlinear response curves, for this single moving vehicle case, are
presented in Figures 10.14 10.16. By examining those figures, it is evident that the two
dynamic solutions demonstrate similar behavior throughout the time of loading
and the agreement between them is so good that it is even difficult to see the two curves
in some of the figures. This is an important observation since the linear
dynamic analysis is far less complicated and requires far less computer time than the
nonlinear dynamic analysis. Although there are some minor differences at the peak
regions in, for example, Figures 10.15b and 10.16a, which also lead to differences in the
DAF (see Table 10.6), these are insignificant from the engineering point of view as they are
comparatively small when compared to those resulting from other
simplifications and assumptions made when designing a bridge. For the linear and
nonlinear dynamic solutions, a maximum of 3 and 2 iterations per increment were
required and the corresponding CPU time was 470 s and 1890 s, respectively.
Analysis results indicated that cable 6 has, for the case studied here, the highest DAF
of all cables (1.275). The response curves for the tension in this cable, Figure 10.16a,
show a fluctuation with a frequency of about 1.19 Hz. This fluctuation in cable
tension, also the dominant frequency for the vertical acceleration of the corresponding
cable-girder attachment point, was caused by vibration modes 7 and 8 having a
frequency of about 1.21 Hz.
As mention earlier, for this bridge model it was assumed that the girder was pinned at the
ends. If a simply supported girder at the ends was assumed instead, which is the traditional way
of supporting such a medium long bridge girder, the natural
frequencies for the first three vertical bending modes of vibration will decrease to 0.314,
0.413, and 0.676 Hz. The DAF for this simply supported girder case are also included in Table
10.6. As can be seen, the DAF for the tension in cable 6 and for the vertical displacement at node
18 are lower for this case. However, a significant
increase in the DAF was obtained for the tension in cable 13, the horizontal
displacement of the right pylon top, and for the fixed end moment of the same pylon.
Based upon the presented results, three general conclusions can be made. The first is that
linear dynamic analysis give sufficiently accurate results as there is no apparent nonlinear
behavior of the bridge during application of the moving traffic load. The second conclusion is
that high DAF values can be reached (1.39 for the axial force in the deck at the left pylon) even
though a road surface with no roughness was assumed.
164
The third conclusion is that the supporting condition of the bridge girder has a
significant influence on the dynamic response.
10.3.2.2 Response due to a train of moving vehicles, effect of bridge-vehicle
interaction and cable modeling
To investigate the effect of having a number of vehicles simultaneously on the bridge, a
hypothetical train of four moving 44 ton trucks was adopted. The distance between truck models
and model properties are shown in Figure 10.8. According to the study conducted in Part A, one
of the important factors affecting the bridge response is the dynamic interaction between the
vehicles and the bridge deck. Therefore, an
investigation of the bridge-vehicle interaction effect was also conducted for this case,
where the moving vehicles were modeled either as constant moving forces (i.e.
ignoring interaction) or as sprung masses moving on a smooth road surface. In
addition, an investigation of the effect of including cable motions and modes of
vibrations on the dynamic response was conducted where each cable was either
modeled using 1 or 5 catenary cable elements. 1500 increments corresponding to a
time step of 0.021 s was required for this analysis.
For the study of bridge-vehicle interaction effect, Figure 10.17 shows some response
curves obtained from a nonlinear dynamic and nonlinear static traffic load analysis.
The DAF values are given in Table 10.6. Examining this figure and also Table 10.6, it
was found that the moving force model, which usually gives negligible differences in results
compared to the sprung mass results when the road surface roughness is
ignored, gives for this flexible bridge significant differences especially for the tension in
cable 6. As also noticed the DAF decreases for the moving force model except for the tension in
cable 6 where the moving force model increases the DAF tremendously.
It is believed that this difference in response occurs, as mentioned earlier in section 8.4,
as the result of having the vehicles acting as vibration absorbers when modeled as sprung mass
systems.
Comparing the nonlinear dynamic results of the one truck case (original configuration)
with the case of four trucks modeled as sprung masses, it was found that the maximal dynamic
part of the contact force increases by 87 %, and the absolute maximal vertical acceleration of the
girder at the bridge center by 125 %. On the other hand, no
particular trend was found when comparing the DAF of the two cases, as the DAF
165
increases significantly for some of the studied elements and nodes but decreases for
others.
The case of four moving trucks modeled as sprung masses was also solved for the
bridge model with 5 catenary cable elements per cable. Consequently, cable motions and
cable vibration modes were considered in this analysis. In Figure 10.18, some
response curves obtained from a nonlinear dynamic and nonlinear static traffic load
analysis are presented. Mixed results were obtained when the new solution was
compared with the 1 element per cable solution, see Table 10.6. As an example, the 1
element per cable case underestimated the dynamic response and therefore the DAF
for the tension in cable 6 but overestimated it for the tension in cable 13. Furthermore, no
particular trend could be found for the cables, as the tension in all cables were effected regardless
of their length or position (side or main span). Evaluation of the DAF for the entire bridge model
revealed that the largest difference in response
occurred for the axial force in deck element 36 (see Figure 10.18c), where the DAF
increased from 1.388 to 1.658, when the 5 element per cable model was adopted. For the
cable tension, the largest difference was found for cable 16 where the DAF
increased from 1.333 to 1.422. The fluctuation of tension in this cable is shown in Figure
10.18d. A significant increase in DAF (from 1.400 to 1.596), see Figure 10.18b, was also found
for the bending moment in pylon element 61.
As a conclusion, the dynamic interaction between the vehicles and the bridge deck
should always be taken into account even if a road surface with no roughness is
assumed. Moreover, to avoid an underestimation of the dynamic response, the cables
should always be modeled so their motion and modes of vibration are considered in
the analysis.
10.3.2.3 Speed and bridge damping effect
To investigate the effect of vehicle speed and bridge damping on the dynamic
response, a single 44 ton moving truck was adopted. This is the same case as the socalled original configuration but is solved here for the new speeds of 50, 70, 110, and 130 km/h.
1500 increments were used for the solution of the 50 and the 70 km/h cases, and 1000 increments
for the rest. The original configuration ( v = 90 km/h) was also
analyzed here for new bridge damping ratios, namely: 0, 0.01, 0.015, 0.02, 0.03. Only
166
nonlinear dynamic analysis was performed and the results are presented in Figure
10.19 and in Table 10.6.
The vertical displacement of the girder at the center of the bridge is shown in Figures
10.19a and 10.19b for different speeds and damping ratios. The variation of the DAF, for some
selected elements and nodes, with respect to different speeds and damping ratios are shown in
Figures 10.19c and 10.19d. As seen, even though the DAF tends to reduce at certain speeds, it is
evident that the response generally increases with the increase in vehicle speed.
As expected, damping reduces the DAF. The effect of damping on the reduction of the
bridge dynamic response was found to be considerable. Among the selected elements
and nodes, the amount of reduction is largest for the axial force in pylon element 52
and the vertical displacement at node 18, while the influence of damping on the axial
force in deck element 30 and the tension in cable 6 are comparatively small. In
addition, it is noted in Figure 10.19d that the relationship between bridge damping ratio
and the DAF is not always linear.
Form this study one can conclude that the response generally increases with the
increase in vehicle speed and that bridge damping have a significant effect upon the
response and should be considered if accurate representation of the true dynamic
response is required.
10.3.2.4 Effect of surface irregularities at the bridge entrance
To roughly study the effect of irregularities in the deck and over the bearings at the
entrance to the bridge, a 3 cm high and 2.5 m long bump located at the left end of the bridge, was
generated as described in section 8.3.2. Results from the nonlinear
dynamic analysis are presented in Figure 10.20 and in Table 10.6. Figures 10.20a-c show
response curves for a vehicle speed of 90 km/h, and in Figure 10.20d some
curves are plotted to show the variation of the DAF with the vehicle speed.
From those results we can infer that the effect of this type of irregularity, on the bridgevehicle response, is considerable. As noted from Table 10.6, a 45 % larger
contact force than the static one was obtained. Also, for as little bump height as 3 cm,
which is not unusual, a significant increase in the axial force in deck element 30 was
167
found, increasing the DAF from 1.39 to 1.60. Of course, the response of nodes and
elements that are located close to the bridge entrance were affected the most by this
bump, whereas, e.g., the bridge deck vertical displacement and acceleration at the center of the
bridge were not affected at all. Again the highest DAF for all cables was found for cable 6.
Moreover this bump affected the response of this cable more than the response of other cables.
From this study one can conclude that not only maintenance of the bridge road surface is
important to reduce damage to bridges but also the elimination of irregularities (unevenness) in
the approach pavements and over bearings is important as these
strongly influence the response.
10.3.2.5 Effect of tuned vibration absorbers
The effectiveness of a tuned mass damper (TMD) in suppressing vibrations due to a
single 44 ton moving truck is investigated in this study. The truck was assumed to move
on a smooth road surface at the constant speed of 110 km/h. The TMD was
positioned at the center of the bridge (node 14) and tuned to the first bending mode of
vibration (0.33 Hz). The optimum tuning parameters given in equation (8.18) were
adopted, where the mass ratio was set to 0.005 giving a TMD mass of approximately
15.6 ton. Some results from the nonlinear dynamic analysis are presented in Table 10.6
and in Figures 10.21-10.23. The results for this case presented in Table 10.6 should be
compared with those corresponding to v =110 km/h.
Examining the results, it was found that the two anchorage cables 1 and 24 had the largest
reduction in the DAF value. The DAF for cable 1 (highest DAF among all
cables) was reduced by the TMD from 1.348 to 1.306. For cable 24, the tension due to
traffic load is shown in Figure 10.22c, and the DAF was reduced from 1.130 to 1.063.
A considerable reduction was also found for the horizontal displacement of node 43
and the moment in pylon element 52, see Table 10.6. However, The results also
revealed that the TMD not always is very effective in reducing the maximum dynamic
response during the forced vibration period (i.e. when the vehicle is on the bridge). In fact, due to
the interaction between the bridge-vehicle-TMD systems, the response and the DAF for certain
elements and nodes can even increase due to the TMD. As an
example, the DAF for cable 14 increased from 1.049 to 1.079, when the TMD was
considered. However, it is evident that the TMD is very effective in reducing the
168
vibration level in the free vibration period, even for the case of cable 14. This is due to
the increase of the overall damping of the bridge by the TMD.
Power spectral densities (PSD) of the vertical acceleration response of the girder at the
center of the bridge (node 14) and the horizontal acceleration response of the right pylon top
(node 43), with and without TMD, are presented in Figures 10.23b and
10.23c. The corresponding bending modes of vibration are also indicated in those
figures. It is clear from those figures that several modes were excited, but the first mode
has the largest contribution to the response of the two chosen nodes. To further reduce the
vibration level of the bridge one might install extra TMDs, e.g. tuned to the third mode and
installed at the top of the two pylons. As expected and observed in those figures, a TMD tuned to
the first mode is only effective in reducing the
contribution of this mode to the dynamic response. This is one of the disadvantages of a
TMD, see the discussion in section 8.4.
In addition, it is noted that the PSD for the vertical response at the center of the bridge
does not have peaks corresponding to the antisymmetric modes as these mode shapes
have zero displacements at this node, e.g. see mode 2 in Figure 10.13. On the other hand,
as seen in Figure 10.23c and also Figure 10.13, these antisymmetric modes do contribute for the
pylon tops horizontal response. This should be taken into
consideration when conducting field measurements to estimate the vibration
frequencies of a bridge of this type, as an accelerometer installed at the pylon top will
identify not only the symmetric modes but also the antisymmetric ones.
From this investigation it is concluded that a TMD, despite the fact that it is not always
effective in reducing the maximum dynamic response during the forced vibration
period, increases the overall damping of the bridge by working as an additional energy
dissipater.
169
0.14
20
(a)
Dynamic (nonlinear)
)
Nonlinear
(b)
Dynamic (linear)
2
0.10
Linear
10
0.06
0
0.02
-10
-20
-0.02
-30
-0.06
-40
-0.10
-50
-0.14
0
5
10
15
20
25
0
5
10
15
20
25
Time (s)
Time (s)
20
0.50
Dynamic (nonlinear)
(c)
(d)
Dynamic (linear)
0.25
10
0.00
0
-0.25
-10
-0.50
Dynamic (nonlinear)
Axial force (MN)
Dynamic (linear)
-20
-0.75
-30
-1.00
0
5
10
15
20
25
0
5
10
15
20
25
Time (s)
Time (s)
Figure 10.14 Vertical displacement (a) and acceleration histories (b) at node 14;
vertical displacement at node 18 (c); axial force in deck element 30 at node 7 (d). The dashed
vertical lines indicate when the vehicle is at the left and right pylon
15
10
(a)
Dynamic (nonlinear)
(b)
10
Dynamic (linear)
5
5
(MNm)
0
0
-5
ng moment -5
Dynamic (nonlinear)
-10
Dynamic (linear)
Bendi
-10
0
5
10
15
20
25
0
5
10
15
20
25
Time (s)
Time (s)
0.80
0.80
(c)
(d)
Dynamic (nonlinear)
0.60
Dynamic (linear)
0.40
0.40
0.20
0.00
0.00
Axial force (MN)
-0.40
Dynamic (nonlinear)
Shear force (MN)
-0.20
Dynamic (linear)

-0.40
-0.80
0
5
10
15
20
25
0
5
10
15
20
25
Time (s)
Time (s)
Figure 10.15 Horizontal displacement of node 43 (a); bending moment (b), axial force
(c), and shear force (d) in element 52 at node 36. The dashed vertical lines indicate when the
vehicle is at the left and right pylon 0.45
0.20
Dynamic (nonlinear)
Dynamic (nonlinear)
(a)
(b)
Dynamic (linear)
Dynamic (linear)
0.35
0.16
0.12
0.25
0.08
0.15
0.04
0.05
0.00
Axial force (MN)
Axial force (MN) -0.05
-0.04
-0.08
-0.15
0
5
10
15
20
25
0
5
10
15
20
25
Time (s)
Time (s)
0.50
0
(c)
Nonlinear
)
(d)
Nonlinear
2
Linear
Linear
-10
0.25
-20
0.00
-30
-0.25
-40
-50
-0.50
0
5
10
15
20
25
0
5
10
15
20
25
Time (s)
Time (s)
Figure 10.16 Axial force in cable 6 (a); axial force in cable 13 (b); truck body vertical
displacement (c), and acceleration histories (d). The dashed vertical lines indicate when the
vehicle is at the left and right pylon
30
25
Dynamic (sprung mass model)
20
(a)
(b)
Dynamic (moving force model)
20
10
Static
15
0
(MNm) 10
-10
5
-20
0
-30
ng moment
-40
-5
-50
Bendi -10
Static
-60
-15
0
5
10
15
20
25
30
0
5
10
15
20
25
30
Time (s)
Time (s)
2.5
0.30
(c)
0.25
(d)
2.0
0.20
Static
1.5
0.15
1.0
0.10
0.05
0.5
Axial force (MN)
0.00
Axial force (MN)
0.0
-0.05
Static
-0.5
-0.10
0
5
10
15
20
25
30
0
5
10
15
20
25
30
Time (s)
Time (s)
Figure 10.17 Vertical displacement at node 18 (a); bending moment (b), and axial force
(c) in element 52 at node 36; axial force in cable 6 (d)
40
3.5
Dynamic (1 element/cable)
(a)
2.5
(b)
20
Dynamic (5 elements/cable)
Static
1.5
0
0.5
(MNm)
-20
-0.5
-1.5
-40
-2.5
-60
ng moment -3.5
-4.5
-80
Bendi -5.5
Static
-100
-6.5
0
5
10
15
20
25
30
0
5
10
15
20
25
30
Time (s)
Time (s)
0.6
0.5
(c)
0.5
0.4
Static
Static
0.4
0.3
(d)
0.3
0.2
0.1
0.2
0.0
0.1
0.0
Axial force (MN)
-0.2
-0.1
-0.3
-0.2
0
5
10
15
20
25
30
0
5
10
15
20
25
30
Time (s)
Time (s)
Figure 10.18 Vertical displacement at node 14 (a); bending moment in pylon element 61
at the attachment point of cable 17; (b), axial force in element 36 at node 13 (c); axial force in
cable 16 (d) 25
25
(a)
(b)
15
15
5
5
-5
-5
-15
-15
-25
-25
= 50
v
km/h
=
0
-35
= 90
v
km/h
-35
=
0.0056
= 110
v
km/h
-45

-45
= 0.03
-55
-55
0
5
10
15
20
25
30
35
40
45
0
5
10
15
20
25
Time (s)
Time (s)
1.40
1.50
(c)
(d)
1.40
1.30
1.30
1.20
1.20
1.10
1.10
1.00
Dynamic amplification factor
Dynamic amplification factor 1.00
50
70
90
110
130
0
0.005
0.01
0.015
0.02
0.025
0.03
Bridge damping ratio,
Vertical displacement at node 14
Axial force in element 52 at node 36
Axial force in cable 6
Axial force in deck element 30 at node 7
Bending moment in element 52 at node 36
Figure 10.19 Vertical displacement at node 14 calculated for different vehicle speeds (a)
and damping ratios (b); the influence of vehicle speed (c) and damping ratio (d) on some
dynamic amplification factors 0.20
0.40
(a)
Dynamic (0 mm bump)
(b)
0.16
Dynamic (30 mm bump)
0.15
Static
0.12
-0.10
0.08
0.04
-0.35
0.00
Dynamic (0 mm bump)
-0.60
Axial force (MN)
-0.04
Axial force (MN)
Static
-0.85
-0.08
-0.12
-1.10
0
5
10
15
20
25
0
5
10
15
20
25
Time (s)
Time (s)
20
1.6
(c)
Dynamic (0 mm bump)
(d)
10
1.5
0
1.4
-10
1.3
-20
1.2
-30
1.1
-40
Bending moment in element 52 at node 36
-50
Dynamic amplification factor 1.0
0
5
10
15
20
25
50
70
90
110
130
Time (s)
Figure 10.20 Axial force in cable 6 (a); axial force in deck element 30 at node 7 (b);
truck body vertical displacement (c); the influence of vehicle speed on some dynamic
amplification factors (30 mm bump) (d) Effect of tuned vibration absorbers
25
(a)
15
5
-5
-15
-25
-35
with TMD
-45
without TMD
-55
0
10
20
30
40
50
60
70
Time (s)
15
(b)
9
3
-3
-9
-15
-21
with TMD
-27
without TMD
-33
0
10
20
30
40
50
60
70
Time (s)
10
(c)
6
2
-2
-6
with TMD
-10
without TMD
0
10
20
30
40
50
60
70
Time (s)
Figure 10.21 Vertical displacement at node 14 (a) and node 18 (b); horizontal
displacement at node 43 (c). The dashed vertical line indicate when the
truck leaves the bridge
177
Effect of tuned vibration absorbers
90
(a)
60
30
0
-30
-60
-90
0
10
20
30
40
50
60
70
Time (s)
0.40
(b)
)2
0.20
0.00
-0.20
Vertical acceleration (m/s -0.40
0
10
20
30
40
50
60
70
Time (s)
0.60
(c)
with TMD
0.40
without TMD
0.20
0.00
-0.40
0
10
20
30
40
50
60
70
Time (s)
Figure 10.22 Vertical displacement (a) and acceleration (b) of the TMD mass; axial
force in cable 24 (c). The dashed vertical line indicate when the truck
leaves the bridge
178
Effect of tuned vibration absorbers
12
(a)
with TMD
8
without TMD
4
0
-4
-8
0
10
20
30
40
50
60
70
Time (s)
1.E+0
1st
3rd
(b)
with TMD
without TMD
5th
9th
1.E-2
1.E-4
PSD
1.E-6
1.E-8
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0
Frequency (Hz)
1.E+0
1st
with TMD
(c)
3rd
6th
1.E-2
without TMD
2nd
10th
4th
9th
5th
7th,8th
1.E-4
PSD
1.E-6
1.E-8
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0
Frequency (Hz)
Figure 10.23 Bending moment in element 52 at node 36 (a); power spectral densities
(PSD) of the vertical acceleration response at node 14 (b) and the
horizontal acceleration response at node 43 (c) (y axis in log scale)
179
original configuration - convergence study
simply supported 4 trucks
4 trucks moving
number of increments
500
1000
1500
1000
500
1000
1500 girder
force model
linear dynamic - 30 modes all modes
nonlinear dynamic
linear
nonlin. linear nonlin. linear nonlin.
1.207
1.186
1.179
1.186
1.208
1.187
1.180
1.131
1.130
1.060
1.056
1.048
1.048
1.270
1.243
1.234
1.243
1.240
1.235
1.230
1.119
1.112
1.286
1.265
1.171
1.162
Horizontal displacement of node 43
1.086
1.065
1.057
1.065
1.094
1.060
1.053
1.250
1.247
1.279
1.274
1.202
1.199
1.404
1.393
1.391
1.394
1.374
1.390
1.387
1.360
1.364
1.237
1.246
1.205
1.214
Shear force in element 52 at node 36
1.212
1.174
1.164
1.160
1.205
1.154
1.149
1.222
1.087
1.285
1.271
1.198
1.181
1.284
1.270
1.268
1.268
1.279
1.301
1.275
1.270
1.290
1.251
1.202
1.229
1.173
Factors (DAF) Bending moment in element 52 at node 36
1.222
1.180
1.166
1.159
1.210
1.152
1.150
1.431
1.442
1.276
1.262
1.207
1.188
1.316
1.294
1.284
1.284
1.240
1.275
1.268
1.240
1.238
1.086
1.035
1.358
1.345
Dynamic Amplification
1.157
1.133
1.125
1.133
1.164
1.135
1.126
1.232
1.234
1.239
1.242
1.211
1.213
Absolute maximum vertical accel. at node 14 (m/s2)
0.160
0.125
0.123
0.125
0.158
0.140
0.124
0.121
0.136
0.336
0.316
0.275
0.266
Maximum normalized bridge-vehicle contact force
1.049
1.039
1.036
1.039
1.049
1.039
1.036
1.041
1.041
1.071
1.073
1.000
1.000
4 trucks
v = 50
v = 70 v = 110 v = 130
=
=
=
=
=
bump
TMD
5 elem/cable
km/h
km/h
km/h
km/h
0
0.01
0.015
0.02
0.03
30 mm v = 110
nonlinear
nonlin. nonlin. nonlin. nonlin. nonlin. nonlin. nonlin. nonlin. nonlin. nonlin. nonlin.
1.087
1.041
1.132
1.282
1.204
1.220
1.164
1.143
1.124
1.093
1.188
1.245
1.289
1.057
1.003
1.427
1.060
1.285
1.202
1.170
1.143
1.102
1.233
1.378
Horizontal displacement of node 43
1.272
1.178
1.086
1.131
1.205
1.106
1.032
1.010
0.992
0.984
1.059
1.065
1.237
1.112
1.176
1.414
1.427
1.401
1.381
1.372
1.362
1.329
1.601
1.413
Shear force in element 52 at node 36
1.272
1.195
1.100
1.331
1.230
1.211
1.127
1.102
1.097
1.087
1.172
1.266
1.272
1.189
1.241
1.511
1.461
1.395
1.258
1.222
1.193
1.168
1.329
1.463
Factors (DAF) Bending moment in element 52 at node 36
1.268
1.178
1.104
1.323
1.244
1.239
1.123
1.103
1.097
1.087
1.211
1.260
1.101
1.024
1.101
1.206
1.408
1.300
1.258
1.241
1.226
1.202
1.331
1.205
Dynamic Amplification
1.208
1.082
1.008
1.104
1.205
1.180
1.107
1.082
1.063
1.038
1.138
1.083
Absolute maximum vertical accel. at node 14 (m/s2)
0.316
0.067
0.091
0.145
0.202
0.195
0.118
0.099
0.086
0.075
0.141
0.137
Maximum normalized bridge-vehicle contact force
1.078
1.024
1.042
1.059
1.060
1.039
1.039
1.039
1.039
1.040
1.446
1.059
Table 10.6
Dynamic amplification factors (DAF), absolute maximum vertical acceleration at node
14, and maximum normalized bridge-vehicle contact force. Note that even the linear dynamic
analysis referred to in this table is based on the dead load tangent stiffness matrix obtained from
a nonlinear static analysis Chapter
______________________________________________________________________
Conclusions and Suggestions for Further
Research
______________________________________________________________________
11.1 Conclusions of Part B
The conclusions from the study conducted in Part B of this thesis are presented in the
following two subsections. In the first subsection, conclusions concerning the
nonlinear finite element modeling of cable supported bridges are presented, and in the
second subsection, conclusions are presented concerning the response due to moving vehicles.
11.1.1 Nonlinear finite element modeling technique

The present work has presented a method for modeling cable supported bridges for the
nonlinear finite element analysis. A two-node catenary cable element was adopted for modeling
the cables and a beam element for modeling the girder and the pylons. This study has shown that
the adopted elements are accurate and efficient for nonlinear analysis of cable-stayed and
suspension bridges. It has been confirmed that the main advantages of the cable element are the
simplicity of including the effect of pretension of the cable and the exact treatment of cable sag
and cable weight. Moreover, the
iterative process adopted, to find the internal force vector and tangent stiffness matrix for
the cable element, was found to converge very rapidly.
According to the authors opinion, linear analysis utilizing the traditional equivalent
modulus approach, is not satisfactory for modern cable-stayed bridges. Modern cable-stayed
bridges built today or proposed for future bridges are, as they are highly
flexible, subjected to large displacements. The equivalent modulus approach however
181
accounts only for the sag effect but not for the stiffening effect due to large
displacements [7]. It was found that the catenary cable element is simple to formulate,
accurate, and can correctly model the geometric change of the cable at any tension level. This
makes the element very attractive, especially for static response
calculations, and the author strongly recommends the use of this element. However, one
drawback is when using commercial finite element codes for analysis, as only few commercial
codes, e.g. ABAQUS, enable the users to define their own elements. This disadvantage applies
also to the one bar element equivalent modulus approach.
It has been demonstrated that cable supported bridges have a hardening characteristic
with respect to the applied load. Furthermore, due to the highly nonlinear behavior during the
static application of the dead load, a nonlinear static analysis is required to arrive at the deformed
dead load tangent stiffness matrix.
Replacing each cable by several catenary cable elements has demonstrated that, in
addition to obtaining new pure cable modes of vibration, cable motions are also
associated with every bending mode of vibration. To simplify the data input process when
utilizing the multi-element cable discretization, one can start from a straight cable configuration
and during analysis the cable configuration under its own weight is determined accurately after
few iterations.
Finally, this work has only focused on two-dimensional modeling of cable supported
bridges. However, the catenary cable element used in this study is also applicable for modeling
cables in other types of cable structures [35, 63, 64], such as: suspended roofs, guyed masts,
electric transmission lines, moored floating bridges, etc. Moreover, with some minor
modifications of the cable element matrices this element can also be used for modeling cables for
three-dimensional analysis. For such analysis, three-dimensional catenary cable and beam
elements can be found in [35, 61].
11.1.2 Response due to moving vehicles

An investigation was conducted to analyze the response of realistic two-dimensional
cable-stayed and suspension bridge models under the action of moving vehicles. For the analysis
of the dynamic response, two approaches were implemented: one for
evaluating the linear dynamic response and one for the nonlinear dynamic response.
Further, nonlinear geometric effects, exact cable behavior, and realistically
182
estimated bridge damping, were considered. This investigation has mainly focused on
comparing linear and nonlinear traffic load dynamic responses and also on the effect of bridgevehicle interaction, road surface roughness, vehicle speed, bridge damping, cable modeling, and
tuned vibration absorbers. Based on this investigation of the
traffic load response of cable-stayed and suspension bridges, the following conclusions
can be made:
utilizing the tangent stiffness matrix (obtained from a nonlinear static analysis under
dead load), linear static and linear dynamic traffic load analysis of cable
supported bridges give sufficiently accurate results from the engineering point of view.
Moreover, the mode superposition technique was found to be very efficient
as accurate results could be obtained based on only 25 to 30 modes of vibration.
Thus, this linear dynamic procedure is especially appropriate for analyzing bridge models
with many degrees of freedom
bridge deck surface roughness and irregularities in the approach pavements and
over bearings have a tremendous effect on the dynamic response. To reduce
damage to bridges not only maintenance of the bridge deck surface is important but also
the elimination of irregularities (unevenness) in the approach pavements and
over bearings. It is also suggested that the formulas for dynamic amplification
factors specified in bridge design codes should not only be a function of the
fundamental natural frequency or span length (as in many present design codes) but
should also consider the road surface condition
for more detailed and accurate studies where the most accurate representation of the true
dynamic response is required, it is recommended to consider the cables
motion and modes of vibration in the dynamic analysis by utilizing the multielement cable discretization. This is also necessary to avoid an underestimation of the
bridge dynamic response
bridge damping has a significant effect upon the response and should always be
considered in such analysis. Some dynamic amplification factors are very sensitive to
bridge damping ratio and the relationship is not always linear. Bridge damping
ratios should be carefully estimated to insure more correct and accurate
representation of the true dynamic response. To obtain realistic damping ratios,
such estimation should be based on results from tests on similar bridges.
Unfortunately, results from many studies of the dynamic response of cable-stayed
183
bridges found in the literature are not useful, as they have been conducted using
either unrealistically high damping ratios for such bridges or no damping at all
a tuned mass damper is not very effective in reducing the maximum dynamic
response during the forced vibration period (i.e. when the vehicle is on the bridge).
In fact, such a device can even increase some of the dynamic amplification factors.
However, the reduction of the vibration level in the free vibration period is
significant as the tuned mass damper increases the overall damping of the bridge
by working as an additional energy dissipater
the moving force model (constant force idealization of the vehicle load) can lead to
unnecessary overestimation of the dynamic amplification factors compared to the
sprung mass model. It is believed that the sprung mass vehicle models are causing
this by acting as vibration absorbers
the dynamic amplification factors of cable supported bridges can reach high values,
higher than 1.30, even if maintenance of the road surface is made regularly. This
situation should be considered in the design practice of such bridges. For the
studied cable-stayed bridge, high dynamic amplification factors were obtained for
the axial force in the girder near the pylons and for the tension in the shortest
cables in the side spans. For this bridge, the designer should consider installation of cable
dampers especially for the shortest cables to increase the fatigue life of the cables.
11.2 Suggestions for further research
Based on the performed investigation, the following suggestions for further research can
be given:
The effect of cable modeling and tuned mass dampers should be more thoroughly
investigated using realistic trains of moving vehicles and considering road surface
roughness and different vehicle speeds, as this could not be accomplished in this
study due to time limitation. Moreover, for future research it is suggested to use
simulated trains of moving traffic based on collected statistical traffic data.
The dependency of bridge response and dynamic amplification factors on the way
in which the girder is connected to the pylons and on other modern girder
supporting conditions, should be investigated.
184
Further work is needed to study the effect of using finer models (i.e. more elements for
discretizing the bridge girder and pylons) of the two studied cable supported
bridges and also three-dimensional models to include torsional effects and torsional
modes of vibration in the analysis.
Extensive testing on a cable supported bridge should be performed to assess the validity
of the analysis methods and the theoretical findings.
Research is needed to thoroughly study active structural control of cable supported
bridges and examine the effectiveness of active devices on suppressing bridge
vibrations due to moving vehicles.
As discussed earlier in section 8.4, the performance of a tuned mass damper (TMD,
passive device) significantly deteriorates when the dynamic characteristics of the bridge
changes (i.e. are different from the original characteristics assumed during the optimal design of
the TMD). Thus, a superior solution can be obtained by using a so-called active tuned mass
damper (ATMD). Such a damper comprises computer
controlled servo-hydraulic actuators that can, when needed, modify the TMD
properties to improve its efficiency. The computer continuously monitors the dynamic
characteristics of the bridge using e.g. sensors attached to the bridge deck.
Active control of structures using cables was proposed by Freyssinet as early as 1960
[67]. Today, active control is applied in advanced airplanes for suppression of
aerodynamic instability, in high-speed trains like the Swedish train X2000 to improve
riding comfort, and in modern cars like the Mercedes A-class to improve stability.
Active controls, e.g. active modification of bridge deck edge shape to enhance
resistance to aerodynamic instabilities like flutter, are also considered for new cable
supported bridges with very long spans such as for the Messina crossing and the
Gibraltar crossing. It is believed that, as the cost of such active systems is high, they can
only be economical for long span bridges where they can induce big saving in
construction material. Furthermore, since some people are perhaps not ready to rely on
computers when crossing a bridge, active control should as a first step only be used to improve
serviceability aspects such as riding comfort, whereas e.g. the stability of the bridge have to rely
entirely on the bridge structure itself.
For cable-stayed bridges real time vibration control can be achieved by e.g. computer
controlling the tension in some cables, so-called active cables, in order to counteract
185
traffic loads at any time. Such control system is based on the idea of constantly
monitoring movements of the bridge using attached sensors and via computer
controlled tensioning jacks, the pretensioning force in the cables is changed. Such bridges
can be referred to as smart bridges as they have the capability of modifying their behavior during
the dynamic loading.
The author believes that active vibration control of long span cable-stayed and
suspension bridges will be an area of significant interest in the future. Till now analysis
and application of active vibration control of structures excited by moving loads have attracted
limited research efforts. For the interested reader, excellent literature review and state-of-the-art
review on control and monitoring of civil
engineering structures are found in [30, 67]. Recent studies describing active control of
bridges are presented in [5, 62, 69].
186
Appendix
______________________________________________________________________
Maple Procedures
______________________________________________________________________
The Maple procedures, used to generate the Fortran code for the elements presented in
Chapter 7, are given below. Each comment line starts with the symbol #.
A.1 Cable element
# Tangent stiffness matrix Kt for;
# the catenary cable element;
readlib(fortran);
with(linalg);
Ly:=1/(2*E*A*w)*(Tj^2-Ti^2)+(Tj-Ti)/w;
Lx:=-P1*(Lu/E/A+1/w*ln((P4+Tj)/(Ti-P2)));
P3:=-P1;
P4:=w*Lu-P2;
Ti:=sqrt(P1^2+P2^2);
Tj:=sqrt(P3^2+P4^2);
f11:=diff(Lx,P1):
f12:=diff(Lx,P2):
f21:=diff(Ly,P1):
f22:=diff(Ly,P2):
f:=matrix(2,2,[f11,f12,f21,f22]):
k:=inverse(f):
k1:=k[1,1]:
k2:=k[1,2]:
k4:=k[2,2]:
187
Kt:=matrix(4,4,[-k1,-k2,k1,k2,-k2,-k4,k2,k4,k1,k2,-k1,-k2,k2,k4,-k2,-k4]):
fortran(Kt,optimized):
A.2 Beam element
# Internal force vector p and;
# tangent stiffness matrix Kt for;
# the beam element;
readlib(fortran);
with(linalg);
ux:=(u4-u1)/L; wx:=(u5-u2)/L;
t:=(u3+u6)/2; tx:=(u6-u3)/L;
e:=(1+ux)*cos(t)+wx*sin(t)-1;
g:=wx*cos(t)-(1+ux)*sin(t);
k:=tx;
PIe:=1/2*L*E*A*e^2;
PIg:=1/2*L*G*A*g^2;
PIk:=1/2*L*E*I*k^2;
PI:=PIe+PIg+PIk;
p:=grad(PI,[u1,u2,u3,u4,u5,u6]);
Kt:=hessian(PI,[u1,u2,u3,u4,u5,u6]);
fortran(p,optimized);
fortran(Kt,optimized);
188
______________________________________________________________________
Bibliography of Part B
______________________________________________________________________
[1]
ABAQUS Users Manual, Hibbitt, Karlsson & Sorensen, Inc., Providence, Rhode Island,
1994.
[2]
Abbas S., Scordelis A., Nonlinear Analysis of Cable-Stayed Bridges, Proc.
Int. Conference on Cable-Stayed and Suspension Bridges, Vol. 2, Deauville,
France, Oct. 1994, pp. 195-210.
[3]
Abdel-Ghaffar A.M., Khalifa M.A., Importance of Cable Vibration in
Dynamics of Cable-Stayed Bridges, J. Eng. Mechanics, ASCE, 117, 1991, pp.
2571-2589.
[4]
Abdel-Ghaffar A.M., Nazmy A.S., 3-D Nonlinear Seismic Behavior of CableStayed Bridges, J. Struct. Eng., ASCE, 117, 1991, pp. 3456-3476.
[5]
Adeli H., Saleh A., Optimal Control of Adaptive/Smart Bridge Structures, J.
Struct. Eng., ASCE, 123, 1997, pp. 218-226.
[6]
Adeli H., Zhang J., Fully Nonlinear Analysis of Composite Girder CableStayed Bridges, Computers and Structures , 54, 1995, pp. 267-277.
[7]
Ali H.M., Abdel-Ghaffar A.M., Modeling the Nonlinear Seismic Behavior of
Cable-Stayed Bridges with Passive Control Bearings, Computers and
Structures, 54, 1995, pp. 461-492.
[8]
Argyris J., Mlejnek H.P., Dynamics of Structures, North-Holland, Amsterdam, 1991.
[9] Aurell J., Edlund S., Vehicle Dyamics of Commercial Vehicles, Volvo Technology
Report, 1990, pp. 20-35.
[10] Bachmann H., Weber B. , Tuned Vibration Absorbers for Lively Structures,
Struct. Eng. Int., Vol. 5, No. 1, 1995, pp. 31-36.
189
[11] Bathe
K.J. , Finite Element Procedures, Prentice Hall, New Jersey, 1996.
[12] Boonyapinyo V., Yamada H., Miyata T., Wind-Induced Nonlinear LateralTorsional Buckling of Cable-Stayed Bridges, J. Struct. Eng., ASCE, 120, 1994, pp. 486506.
[13] Bruno D., Grimaldi A., Nonlinear Behaviour of Long-Span Cable-Stayed
Bridges, Meccanica, 20, 1985, pp. 303-313.
[14] Buchholdt H.A., An Introduction to Cable Roof Structures, Cambridge
University Press, Cambridge, 1985.
[15] Buchholdt
H.A., Structural Dynamics for Engineers, Thomas Telford, London,
1997.
[16] Conti E., Grillaud G., Jacob J., Cohen N., Wind Effects on the Normandie CableStayed Bridge: Comparison Between Full Aeroelastic Model Tests and
Quasi-Steady Analytical Approach, Proc. Int. Conference on Cable-Stayed and
Suspension Bridges, Vol. 2, Deauville, France, Oct. 1994, pp. 81-90.
[17] Crisfield
M.A. , Non-linear Finite Element Analysis of Solids and Structures,
Wiley, Chichester, 1991.
[18] Das A.K., Dey S.S., Effect of Tuned Mass Dampers on Random Response of
Bridges, Computer & Structures, Vol. 43, No. 4, 1992, pp. 745-750.
[19] Davenport A., Larose G., The Structural Damping of Long Span Bridges: An
Interpretation of Observations, Canada-Japan Workshop on Aerodynamics,
Ottawa, Sept. 1989.
[20] Dean D.L., Static and Dynamic Analysis of Guy Cables, J. Struct. Division, ASCE,
87, 1961, pp. 1-21.
[21] Den
Hartog
J.P. , Mechanical Vibrations, 4th edition, McGraw-Hill, New York,
1956.
[22] Fleming J.F., Egeseli E.A., Dynamic Behaviour of a Cable-Stayed Bridge,
Earthquake Eng. and Struct. Dynamics, 8, 1980, pp. 1-16.
190
[23] Forsell
K. , Dynamic Analyses of Static Instability Phenomena, Licentiate
Thesis, TRITA-BKN Bulletin 34, Dept. of Struct. Eng., Royal Institute of
Technology, Stockholm, 1997.
[24] Frba
L. , Vibration of Solids and Structures under Moving Loads, Noordhoff
International Publishing, Groningen, 1972.
[25] Gambhir M.L., Batchelor B. , A Finite Element for 3-D Prestressed Cablenets, Int.
J. Numer. Methods in Eng., 11, 1977, pp. 1699-1718.
[26] Gradin M., Rixen D. , Mechanical Vibrations, Wiley, Chichester, 1994.
[27] Gimsing
N.J., Cable Supported Bridges, second edition, Wiley, Chichester,
1997.
[28] Gimsing
N.J.,
Technical University of Denmark, Tanaka H ., Danish Maritime
Institute, and Esdahl S., COWIconsult A/S, Personal communication, 1997.
[29] Green M.F., Cebon D., Dynamic Interaction Between Heavy Vehicles and
Highway Bridges, Computer & Structures, Vol. 62, No. 2, 1997, pp. 253-264.
[30] Housner
G.W.,
Bergman L.A., Gaughey T.K., Chassiakos A.G., Claus R.O.,
Masri S.F., Skelton R.E., Soong T.T., Spencer B.F., Yao J.T.P., Structural Control: Past,
Present, and Future, J. Eng. Mechanics, ASCE , 123, 1997, pp.
897-971.
[31] Huddleston
J.V. , Computer Analysis of Extensible Cables, J. Eng. Mechanics Div., ASCE , 107,
1981, pp. 27-37.
[32] Huddleston J.V., Ham H.J., Poisson Effect in Extensible Cables with Both Ends
Fixed, J. Eng. Mechanics Div., ASCE , 120, 1994, pp. 1590-1595.
[33] Indrawan
B.,
Vibrations of a Cable-Stayed Bridge due to Vehicle Moving over
Rough Surface, Master Thesis, Asian Institute of Technology, Bangkok, March 1989.
[34] Irvine
H.M. , Cable Structures, MIT Press, Cambridge, 1992.
[35] Jayaraman H.B., Knudson W.C., A Curved Element for the Analysis of Cable
Structures, Computers and Structures, 14, 1981, pp. 325-333.
191
[36] Jensen G., Petersen A., Erection of Suspension Bridges, Proc. Int. Conference on
Cable-Stayed and Suspension Bridges, Vol. 2, Deauville, France, Oct. 1994,
pp. 351-362.
[37] Johnson R., Larose G., Field Measurements of the Dynamic Response of the Hga
Kusten Bridge During Construction, TRITA-BKN Report 49, Dept. of
Struct. Eng., Royal Institute of Technology, Stockholm, 1998.
[38] Jones R.T., Pretlove A.J., Vibration Absorbers and Bridges, The Highway
Engineer, 26, No. 1, 1979, pp. 2-9.
[39] Judd B.J., Wheen R.J., Nonlinear Cable Behavior, J. Struct. Division , ASCE,
104, 1978, pp. 567-575.
[40] Kanok-Nukulchai W., Hong G., Nonlinear Modelling of Cable-Stayed
Bridges, J. Construct. Steel Research , 26, 1993, pp. 249-266.
[41] Kanok-Nukulchai W., Yiu P.K.A., Brotton D.M., Mathamatical Modelling of
Cable-Stayed Bridges, Struct. Eng. Int., 2, 1992, pp. 108-113.
[42] Karoumi R., Dynamic Response of Cable-Stayed Bridges Subjected to Moving
Vehicles, IABSE 15th Congress, Denmark, 1996, pp. 87-92.
[43] Kawashima K., Unjoh S., Tsunomoto M., Damping Characteristics of Cable
Stayed Bridges for Seismic Design, J. Res. Public Works Res. Inst., Vol. 27, Dec. 1991.
[44] Kawashima K., Unjoh S., Tsunomoto M., Estimation of Damping Ratio of
Cable-Stayed Bridges for Seismic Design, J. Struct. Eng., ASCE, Vol. 119, No.
4, April 1993, pp. 1015-1031.
[45] Kwon H.C., Kim M.C., Lee I.W., Vibration Control of Bridges Under Moving
Loads, Computer & Structures, Vol. 66, No. 4, 1998, pp. 473-480.
[46] Larose G.L., Zasso A., Melelli S., Casanova D., Field Measurements of the WindInduced Response of a 254 m High Free-Standing Bridge Pylon, Proc. of the 2nd European
African Conference on Wind Engineering, Genova, Italy, June
1997, pp. 1-8.
192
[47] Larsen A., Jacobsen A.S., Aerodynamic Design of the Great Belt East Bridge,
Proc. 1st Int. Symposium on Aerodynamics of Large Bridges, Copenhagen,

Denmark, Feb. 1992, pp. 269-283.
[48] Larsen A., Aerodynamic Aspects of the Final Design of the 1624 m Suspension
Bridge Across the Great Belt, J. Wind Eng. and Industrial Aerodynamics,
Elsevier, 48, 1993, pp. 261-285.
[49] Leonard
J.W. , Tension Structures, McGraw-Hill, New York, 1988.
[50] Luft
R.W.,
Optimal Tuned Mass Dampers for Buildings, J. Struct. Division,
ASCE, Vol. 105, No. ST12, Dec. 1979, pp. 2766-2772.
[51] Madsen
B.S.,
Sorensen L.T., Manufacture and Erection of the Steel Main
Span, Proc. Int. Conference on Cable-Stayed and Suspension Bridges, Vol. 1, Deauville,
France, Oct. 1994, pp. 691-697.
[52] MAPLE V Language Reference Manual, Springer-Verlag, New York, 1991.
[53] MATLAB Using Matlab Version 5, The MathWorks Inc., Natick, 1996.
[54] McNamara R.J., Tuned Mass Dampers for Buildings, J. Struct. Division, ASCE,
Vol. 103, No. ST9, Feb. 1977, pp. 1785-1798.
[55] Nazmy A.S., Abdel-Ghaffar A.M., Three-Dimensional Nonlinear Static
Analysis of Cable-Stayed Bridges, Computers and Structures, 34, 1990, pp.
257-271.
[56] O'Brien T., Francis A.J., Cable Movements Under Two-Dimensional Loads, J.
Struct. Division, ASCE, 90, 1964, pp. 89-123.
[57] O'Brien T., General Solution of Suspended Cable Problems, J. Struct.
Division, ASCE, 93, 1967, pp. 1-26.
[58] Ostenfeld K.H., Larsen A., Bridge Engineering and Aerodynamics, Proc. 1st Int.
Symposium on Aerodynamics of Large Bridges, Copenhagen, Denmark,
Feb. 1992, pp. 3-22.
[59] Ostenfeld K.H., Long Span Bridges: State-of-the-art, Vg- och Vatten-byggaren, J.
Swedish Society of Civil Eng., 697, pp. 26-31.
193
[60] Ozdemir H., A Finite Element Approach for Cable Problems, Int. J. Solids
Structures, 15, 1979, pp. 427-437.
[61] Pacoste C., Eriksson A., Beam Elements in Instability Problems, Comput.
Methods Appl. Mech. Eng., 144, 1997, pp. 163-197.
[62] Patten W.N., Sack R.L., He Q., Controlled Semiactive Hydraulic Vibration
Absorber for Bridges, J. Struct. Eng., ASCE, Vol. 122, No. 2, Feb. 1996, pp.
187-192.
[63] Peyrot A.H., Goulois A.M., Analysis of Cable Structures, Computers and
Structures, 10, 1979, pp. 805-813.
[64] Peyrot A.H., Goulois A.M., Analysis of Flexible Transmission Lines, J. Struct.
Division, ASCE, 104, 1978, pp. 763-779.
[65] Raoof
M.,
Estimation of Damping Ratio of Cable-Stayed Bridges for Seismic
Design, J. Struct. Eng., ASCE, Vol. 120, No. 8, Aug. 1994, pp. 2548-2550.
[66] Setareh M., Hanson R.D., Tuned Mass Dampers to Control Floor Vibration
from Humans, J. Struct. Eng., ASCE, Vol. 118, No. 3, March 1992, pp. 741762.
[67] Soong T.T., State-of-the-art Review: Active Structural Control in Civil Engineering
, Eng. Struct., 10, 1988, pp. 74-84.
[68] Walther R., Houriet B., Isler W., Moa P. , Cable Stayed Bridges, Thomas Telford,
London, 1988.
[69] Wells M., Active Structural Control: an Appraisal for a Practical Application.
The Royal Victoria Footbridge. London Docklands, J. Constructional Steel
Research, 46:1-3, 1998, paper no. 59.
194

V 110 KM/H 146 M

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

V 110 KM/H 146 M

Caricato da

Copyright:

Formati disponibili

v = 110 km/h

Response of Cable-Stayed and

Response of Cable-Stayed and Suspension

Docent Sven Ohlsson

Response of Cable-Stayed and Suspension

Akademisk avhandling som med tillstnd av Kungliga Tekniska Hgskolan i

Stockholm, January 1999

vehicle interaction and cable modeling.................................. 165

not only wind and earthquake loading but also traffic

conditions, bridge mass and stiffness, and natural frequencies.

al. (1982) [33]

FEM, one and two trucks

Wang (1993) [87] SS truss

bridges subjected to moving vehicles

acceleration of the point of contact in the vertical direction.

i is related to the PSD function S(

obtained from the sum:

a =750106 N/m2, qg=17104 N/m, qq=6104 N/m, and c=9104

Fixed pylon top

Spring stiffness, K m (1012 Nm)

Vehicle speed (km/h)

Conclusions and Suggestions for Further

long span cable-stayed bridges, the author recommends

study the effectiveness of tuned mass dampers (TMD) on suppressing vibrations

Vibration Digest, Vol. 10, No. 8, 1978, pp. 27-35.

1981, pp. 229-246.

Analysis for Railway Vehicles and Structures - Mechanical Model and

The main aims of this study are as follows:

modeling cables in cable-stayed bridges, is referred to as the equivalent modulus

. As will be demonstrated later, this iterative procedure converges very rapidly.

where q is the nodal displacement vector { 1

surfaces. Other suspension systems, e.g. cab suspension

Call algorithm for

(for suspension bridges [19])

bridge structures. As an example Figure

bridge by means of conventional FEM assemblage:

properties of M-orthonormalized eigenvectors, equation (9.8), the above equation is

& t . In each iteration, the displacement corrections are computed

static, tot = q dead + q live = q dead + K

bridge structures are given in Chapter 4. The

Vertical displacement (mm) -5.0

Mode Superposition (present) Direct Integration

Figure 10.5 Geometry of the Great Belt suspension bridge

Pylon tops horizontal displacement (m)

mode 1: 0.099 Hz (0.100 Hz)

10.2.2 Dynamic response due to moving vehicles

E (N/m2) A (m2) I (m4) w (t/m)

Including weight of cross beams.

mode 1: 0.334 Hz (0.311 Hz)

Static (linear & nonlinear)

Vertical displacement (mm)

11.1.1 Nonlinear finite element modeling technique

11.1.2 Response due to moving vehicles

Proc. 1st Int. Symposium on Aerodynamics of Large Bridges, Copenhagen,

Potrebbero piacerti anche