You are on page 1of 254

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
Analysis methods and practical modeling techniques

Raid Karoumi
Department of Structural Engineering
Royal Institute of Technology
S-100 44 Stockholm, Sweden

Akademisk avhandling
Som med tillstånd av Kungl Tekniska Högskolan i Stockholm framlägges till offentlig
granskning för avläggande av teknologie doktorsexamen fredagen den 12 februari
1999 kl 10.00 i Kollegiesalen, Valhallavägen 79, Stockholm. Avhandlingen försvaras på
svenska.
Fakultetsopponent:

Docent Sven Ohlsson
Huvudhandledare: Professor Håkan Sundquist
TRITA-BKN. Bulletin 44, 1998
ISSN 1103-4270
ISRN KTH/BKN/B--44--SE
Stockholm 1999

Response of Cable-Stayed and Suspension
Bridges to Moving Vehicles
Analysis methods and practical modeling techniques

Raid Karoumi
Department of Structural Engineering
Royal Institute of Technology
S-100 44 Stockholm, Sweden

_____________________________________________________________________
TRITA-BKN. Bulletin 44, 1998
ISSN 1103-4270
ISRN KTH/BKN/B--44--SE
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 tillstånd av Kungliga Tekniska Högskolan i
Stockholm framlägges till offentlig granskning för avläggande 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.

Parametric studies have been conducted to investigate the effect of. dynamic analysis. 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. and the second for the nonlinear traffic load response using the Newton-Newmark algorithm. The bridge is idealized as a Bernoulli-Euler beam on elastic supports with varying support stiffness. the finite difference method and the mode superposition technique are used. moving loads. cable element. Two methods for evaluating the dynamic response are presented. The work was conducted under the supervision of Professor Håkan Sundquist to whom I want to express my sincere appreciation and gratitude for his encouragement. vehicle speed. The first for evaluating the linear traffic load response using the mode superposition technique and the deformed dead load tangent stiffness matrix. bridge-vehicle interaction. Several numerical examples are presented including one for the Great Belt suspension bridge in Denmark. 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. valuable advice and for always . finite difference method. finite element analysis. To solve the equation of motion of the bridge. It was also found that utilizing the dead load tangent stiffness matrix. road surface roughness. From the numerical study. and tuned mass dampers. suspension bridge. traffic-induced vibrations. bridge-vehicle interaction. Great Belt suspension bridge. –i– – ii – ______________________________________________________________________ Preface __________________________________________________________________ ____ The research presented in this thesis was carried out at the Department of Structural Engineering. The cables are modeled using a two-node catenary cable element derived using “exact” analytical expressions for the elastic catenary. at the Royal Institute of Technology (KTH) in Stockholm. bridge damping. cables vibration. tuned mass damper. linear dynamic traffic load analysis give sufficiently accurate results from the engineering point of view. The project has been financed by KTH and the Axel and Margaret Ax:son Johnson Foundation.The first approach uses a simplified analysis method to study the dynamic response of simple cable-stayed bridge models. among others. Key words: cable-stayed bridge. it was concluded that road surface roughness has great influence on the dynamic response and should always be considered. bridge. Structural Design and Bridges group.

. . . . . . . . . . . . . . . . . . . .having time for discussions. 15 1.1 Vehicle models. . . . . . . . . . . . . . .1 Major . . 27 2 Vehicle and Structure Modeling 29 2. . I would like to thank my wife Lena Karoumi. . . . . . . and my parents for their love. . . . . . . . . . .2 Research on other bridge types. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stockholm. . . . . . . . . . . . . . . .2. . . . . . . . . . . . . . . my daughter and son. . . 22 1. . .2 Review of previous research . . . . . . . . . . . I also wish to thank Dr. Finally. . . . . . . . . . . . . . . . . . . . . . . . . . . . understanding. . . . . . . . . . . . . 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. . 29 2. . . . . . . . . . . 31 2. . . . . . . . . . Costin Pacoste for reviewing the manuscript of this report and providing valuable comments for improvement. . . . .1 Research on cable-stayed bridges . . . . . . . . support and encouragement. . . . . . . . . . . . . . . . . . . . 15 1. . . . 9 1. . .2. . . . . . . . . . . . . . . . . . . .2. . . . . . . . . . . . . . . . . . . . . . . . . .3 General aims of the present study. . . . . . .2 Bridge structure . . . . . . . . .

. .2 Static analysis .. . . . . .. . . .. .. . . .. ... . . .. . . ... .. . . . .. . . . . . .. . .. .. . . ... . .. . . . . . . .. . . . . .. .. . .. . . . . . . . ... . . . . .1 Dynamic analysis. . . . .. .. . ... . . .. 101 6. . .. . . .. .. .3 Spring stiffness . . . . . .... .. . . .... .. . . 83 5.. . ... ..3 Bridge deck surface roughness . . ... . . . .. .. .... . . . . . . . . 72 4. . . . . .. . . ..... . . . .... . . . .. .6 Discussion of the numerical results . .. ... .. . . . . .. . .. .. . . . ... . . .. 38 3 Response Analysis 43 3.. . . .. .... .. . ... . . . . ... . . . .. .. ... . 49 4 Numerical Examples and Model Verifications 51 4. ..1. . . .. . . .. . . . . .. . . . . . . . .. ..... 43 3. .. . . 34 2... . . .. . . .. . ... . . . . . . ... . .... ..... . . . . . ... . . . . . . . . .. .. .. ... . . .. . . ... .. . . . . ... .... . .. . . . .. . .. .. .. .1 Eigenmode extraction. . ... .... .. . .. .. . ... moving force model . ... .. . . . . . . . . . . . . . .. ... .1 General .. . . . .. . .. . . . 57 4. . . .. .. .2. .5 Three-span cable-stayed bridge.2 Suggestions for further research . . . .. . . . .. . .1 General.... 80 5 Conclusions and Suggestions for Further Research 83 5.... . ... . ... . . .3 General aims of the present study . 103 7 Nonlinear Finite Elements 105 7. .. . . ... . . . . 52 4.... . . . .. . . . .. .. . .. .. ... .... . . ... . . 85 Bibliography of Part A 87 Part B Refined Analysis Utilizing the Nonlinear Finite Element Method 97 6 Introduction 99 6... . ... .. .1. ....1 General . . .. . . .3 Multi-span continuous bridge with rough road surface . .. . .. . .. .. .. .. . . . . . .. .. . . ..2. ... . . . . . .. . .. . . . . .. . . . .. . . . . . . .. . .... . . . . . . .2 Response of the bridge.. . . . . .. . . . . . . .. . . . .. . . . . . . .. .. . . . .. . . . . . 43 –v– 3. . . . . . .. . . .. ... .. . . . .2 Differential equation of motion . . 99 6.. . . . . .. .. . . .. . .... .. 45 3. ...1 Conclusions of Part A .. . .. . . . .. . .. . . . 63 4. . . .. .. . . . . . ... . . . . . 32 2. . .. . ... . . .. .. .... . . 105 .. . ... . . . . . . .... . . ... .. . . . ..4 Simple cable-stayed bridge .. . .. . . . .. ... .. . . .. ... . . .. ... . .. . . . . .. 51 4. . .. . . . .... ... ... . . . . . . .. . .. .. . . .. .. . . . . .. .2 Simply supported bridge. .. . . . ... . . . .. .. . .assumptions .. .. . .. . . .. .. 33 2. . .2 Cable structures and cable modeling techniques .. . . . . . . .. ... . . . . . ... . . . .. . . .. .. . ..

.....................................3 Medium span cable-stayed bridge......2 Bridge deck surface roughness.......1 Response due to a single moving vehicle .......... 126 8..................... 151 10............................... 113 – vi – 8 Vehicle and Structure Modeling 117 8......................2..............................2 Static analysis.......2.....3....1 Static response during erection and natural frequency analysis ........................................2 Nonlinear dynamic analysis .........1 Eigenmode extraction and normalization of eigenvectors... 162 10.....................1 Static response and natural frequency analysis..........2.......... 123 8......... effect of bridge- ..2 Mode superposition technique .............................................1............... 149 10.2...2 Dynamic response due to moving vehicles – parametric study........................................3................................................................................................................................... 163 10..........2............................. 111 7..... 154 10...................1.........................................................................................3.............................................. 121 8..3 Bridge structure .........1 Modeling of damping in cable supported bridges.......... 117 8...............1................1 Simply supported bridge ................................. 107 7... 159 10.. 133 9...............3..........................2 Response due to a train of moving vehicles................3 Modeling of bridge deck and pylons.....3......3..................... 134 9........... 123 8....1 Dynamic Analysis .....................2 Vehicle load modeling and the moving load algorithm...................................2.... 136 9................................... 144 10.................... 135 9...............4 Tuned vibration absorbers................... 158 10..............................................1 Linear dynamic analysis....... 106 7................. 141 10 Numerical Examples 143 10.................................. 127 9 Response Analysis 133 9......1 Vehicle models..1........... 138 9....................................................2 Dynamic response due to moving vehicles......7....1....2 Modeling of cables ..2 The Great Belt suspension bridge ..........................1 Cable element formulation...............................1.......2 Analytical verification.................

.........e.3 Speed and bridge damping effect..................................... 165 10..........2...... and efficient construction techniques in bridge engineering enable the construction of lighter.............2. 166 10. Bridges of this type are now entering a new era with main span lengths reaching 1000 m........ and more slender bridges.3............1 Nonlinear finite element modeling technique...2 Suggestions for further research........... 188 Bibliography of Part B 189 – viii – ______________________________________________________________________ General Introduction and Summary ______________________________________________________________________ Due to their aesthetic appearance......... 181 11...vehicle interaction and cable modeling.. 181 11.... This fact is due.3......................................... To achieve this......1.............................. cable-stayed and suspension bridges. 167 10...........2............... 168 – vii – 11 Conclusions and Suggestions for Further Research 181 11..2 Response due to moving vehicles ..........................................3....... 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..1 Cable element............... The recent developments in design technology.................................................. and the ambition is to further increase the span length and use shallower and more slender girders for future bridges.........5 Effect of tuned vibration absorbers ... have gained much popularity in recent decades... 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.............. Thus nowadays.............. efficient utilization of structural materials and other notable advantages..2 Beam element.. cable supported bridges............................. very long span cable supported bridges are being built.. 187 A.. Among bridge engineers the popularity of cable-stayed bridges has increased tremendously.. 184 A Maple Procedures 187 A............ accurate procedures need to be developed that can lead to a thorough understanding and a realistic prediction of the structural response due to . During the last fifty-eight years.....4 Effect of surface irregularities at the bridge entrance .1 Conclusions of Part B..... i................................. much attention has been given to the dynamic behavior of cable supported bridges... 182 11......... material qualities....................................1......................... longer....

In addition. structural engineers worldwide rely on dynamic amplification factors specified in bridge design codes. Thus. have been conducted on ordinary bridges. they cannot be directly applied to cable supported bridges. Such devices are frequently used today for cable supported bridges. energy dissipation is very low and is often not enough on its own to suppress vibrations. of the dynamic response due to moving vehicles. Various studies. This has greatly reduced the hysteresis that was provided in riveted or bolted –1– joints in earlier bridges. To increase the overall damping capacity of the bridge structure. Therefore. when comparing different national codes. Consequently.not only wind and earthquake loading but also traffic loading. 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. 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. As an example. one possible option is to incorporate external dampers (discrete damping devices such as viscous dampers and tuned mass dampers) into the system. The recent developments in bridge engineering have also affected damping capacity of bridge structures. To consider dynamic effects due to moving traffic on 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. are required in order to check the true capacity of existing bridges to heavier traffic and for proper design of new bridges. However. it is not believed that this is always the most effective and the most economic solution. 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. such as bridge-vehicle interaction and road surface roughness. However. 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. a wide range of variation is found for the dynamic amplification factor. For cable supported bridges and in particular long span cablestayed bridges. . improved analytical techniques that consider all the important parameters that influence the dynamic response. design codes disagree on how this factor should be evaluated and today. as cable supported bridges are more complex structures consisting of various structural components with different properties. welded joints are extensively used nowadays in modern bridge designs. 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. Not only the dynamic behavior of new bridges need to be studied and understood but also the response of existing bridges. Major sources of damping in conventional bridgework have been largely eliminated in modern bridge designs reducing the damping to undesirably low levels.

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. These two parts present two different approaches for solving the moving load problem of ordinary and cable supported bridges. and a catenary cable element. e. Moreover.g. and the second for the nonlinear traffic load response using the NewtonNewmark algorithm. based on the nonlinear finite element method. 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. The first for evaluating the linear traffic load response using the mode superposition technique and the deformed dead load tangent stiffness matrix. exact cable behavior and nonlinear geometric effects. and reference list. Two methods for evaluating the dynamic response are presented.. a more general approach. conclusions. where each has its own introduction. This thesis contains two separate parts. 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. 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. the influence of different parameters on the response of cable supported bridges is investigated. is adopted to study more realistic cable-stayed and suspension bridge models considering. The applicability of the implemented solution procedures is examined and guidelines for future analysis are proposed. 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. However. two sets of . Part A. is used for modeling the cables. To solve the equation of motion of the bridge. Among other things. To study the dynamic response of the bridge-vehicle system in Part A and B. In Part B. the finite difference method and the mode superposition technique are used. is presented. It is to the author’s 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. Part A (Chapter 1-5) and Part B (Chapter 6-11).–2– The aim of this work is to study the moving load problem of cable supported bridges using different analysis methods and modeling techniques. derived using “exact” analytical expressions for the elastic catenary. A beam element is used for modeling the girder and the pylons. which is a slightly modified version of the licentiate thesis presented by the author in November 96. 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.

For most cases. and the contact between the bridge and each moving vehicle is assumed to be a point contact. The two sets of equations are coupled through the interaction forces existing at the contact points of the two subsystems. based on results from tests on similar bridges. longitudinal forces generated by the moving vehicles are neglected. To solve these two sets of equations. However. only vertical modes of vibration of the vehicles are considered • it is assumed that the vehicles never loses contact with the bridge. Based on the study conducted in Part A and B. the springs and the viscous dampers of the vehicles have linear characteristics. only vehicle models of heavy trucks are used • simple one dimensional vehicle models are used consisting of masses.equations of motion are written one for the vehicle and one for the bridge. 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. bridge-vehicle interaction.g. exact cable behavior. However. 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. For such problems. 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. and viscous dampers.g. nonlinear static analysis is essential to determine the dead load deformed condition. 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. Consequently. bridge damping. this approach becomes difficult and cumbersome. The following basic assumptions and restrictions are made: • elastic structural material • two-dimensional bridge models. nonlinear geometric effects. realistic bridge damping values. an iterative procedure is adopted. 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. Consequently. the bridge-vehicle interaction forces act in the vertical direction. For the analysis. e. must be used • care should be taken when the dynamic amplification factors given in the different . Moreover. springs. starting from this position and utilizing the dead load tangent stiffness matrix. or non-uniform crosssections are to be considered. and cables vibration are considered. 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. for analysis of more realistic bridge models where e. 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.

particularly those with long spans. Most certainly this study has not provided a complete answer to the moving load problem of cable supported bridges. 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. 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 . For Part B of this thesis. For some cases it is found that design codes underestimate the additional dynamic loads due to moving vehicles. as it is not believed that these can be used for such bridges. 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. However. In addition. the author hopes that the results of this study will be a help to bridge designers and researchers. 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. For the final design. each bridge of this type. –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. such analyses have not been performed earlier taking into account exact cable behavior and fully considering the bridge-vehicle dynamic interaction.design codes and specifications are used for cable supported bridges. Consequently. should be analyzed as made in Part B of this thesis.

However. while only few –9– studies.2. with increasing span length and increasing slenderness of the stiffening girder. 76.3 on page 16.______________________________________________________________________ 1. like the TGV train in France and the Shinkansen train in Japan with speeds exceeding 300 km/h. the interest in traffic induced vibrations has been increasing due to the introduction of high-speed vehicles. 75. The first modern cablestayed bridge was the Strömsund Bridge in Sweden opened to traffic in 1956. In recent years. Cable supported bridges are special because they are of the geometric-hardening type. For the study of the concept. 79]. 68. The important parameters that influence the dynamic response are (according to previous research conducted in this field. Compared to other types of bridges.1 and Figure 1. the dynamic response of cable-stayed bridges subjected to moving loads is given less attention in theoretical studies. Since that time vehicle speed and vehicle mass to the bridge mass ratio have been increased. which means that the overall stiffness of the bridge increases with the increase in the displacements as well as the forces. The dynamic response of bridges subjected to moving vehicles is complicated. and have been receiving most of the attention. see section 1. see the excellent book by Gimsing [27] and also [28. design and construction of cable-stayed bridges. such as the number of axles. as shown in Figure 1.2. support . axle load. 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. see Table 1. received.1. see section 1.2): • vehicle speed • road (or rail) surface roughness • characteristics of the vehicle. have been carried out to investigate the dynamic effects of moving loads on cable-stayed bridges. resulting in much greater dynamic effects. Static analysis and dynamic response analysis of cable-stayed bridges due to earthquake and wind loading. The increasing dynamic effects are not only imposing severe conditions upon bridge design but also upon vehicle design. Modern cable-stayed bridges with their long spans are relatively new and have been introduced widely only since the 1950. and damping and stiffness of the vehicle suspension system • the number of vehicles and their travel paths • characteristics of the bridge structure. axle spacing. natural frequencies. in order to give an acceptable level of comfort for the passengers. 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. such as the bridge geometry. This is mainly due to the decrease of the cable sag and increase of the cable stiffness as the cable tension increases.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.

This gives a constant DAF that is totally independent on the characteristics of the bridge. the Swedish National Road Administration (Vägverket) includes the additional dynamic loads. the finite difference method and the mode superposition technique are used. The bridge is idealized as a Bernoulli-Euler beam on elastic supports with varying support stiffness.conditions. due to moving vehicles.1 that the national design codes disagree on the – 10 – evaluation of the dynamic amplification factors. Moreover. For bridges like cable-stayed bridges that are more complex and behave differently compared to ordinary bridges. lane load 1. Figure 1. structural engineers worldwide rely on dynamic amplification factors (DAF). 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. 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. these expressions cannot characterize the effect of all the above listed parameters.0 Canada CSA-S6-88m OHBDC Swiss SIA-88. which are usually related to the first vibration frequency of the bridge or to its span length. and natural frequencies. 2. this approach can lead to incorrect traffic loads to be used for designing the bridge. 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).1 shows the variation of the DAF with respect to the fundamental frequency of the bridge. For design purpose. this does not explain such a wide range of variation for the DAF. as these expressions are originally developed for ordinary bridges. 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. It is apparent from Figure 1. bridge mass and stiffness. To solve the equation of motion of the bridge. single vehicle ) Swiss SIA-88. in the traffic loads specified for the different types of vehicles. recommended by different standards [66]. Because of the simplicity of the DAF expressions specified in current bridge design codes. the fundamental frequency was approximated from the span length.8 AASHTO-1989 (DAF India. IRC or . For cases where the DAF was related to the span length. In the Swedish design code for new bridges. and although the specified traffic loads vary in these codes.

Germany. .6 France LCPC D/L=0.K.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 .5 ion f France LCPC D/L=5 plificat D/L = Dead load / Live load 1.BS5400 (1978) 1. DIN1075 act U.4 c am ami Dyn 1.

Composite Xupu China (Shanghai) 590 1996 Composite Meiko-Chuo Japan 590 1997 Steel Skarnsund Norway 530 1991 Concrete Tsurumi Tsubasa Japan 510 1994 Steel Öresund 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 .

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.1978 Steel Table 1. Mississippi USA 372 1982 Steel Duesseldorf-Flehe Germany 368 1979 Steel Tjörn (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 Köhlbrand 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 .

1 (continued) 1000 900 Steel girder 800 Composite girder Concrete girder .1973 Concrete Papineau-Leblanc Canada 241 1971 Steel Kärkistensalmi 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 Strömsund Sweden 183 1956 Steel Table 1.

showed that although there is significant nonlinear behavior . TMD’s. theoretical and experimental study on the effectiveness of tuned mass dampers. And in [11]. see for example [2. and acceleration responses with the TMD locked and unlocked were compared. a cable-stayed bridge in Switzerland with a 60 m main span. is described.1 Research on cable-stayed bridges In recent years the dynamic behavior of cable-stayed bridges has been a source of interesting research.2 Span length increase of cable-stayed bridges in the last fifty years – 14 – 1. literature dealing with the dynamics of these bridges due to moving vehicles is relatively scarce. 47].2. For a cable-stayed footbridge. In this study. dynamic load testing on the Riddes-Leytron bridge. was carried out in [6]. modal testing of the Tjörn bridge. Previous investigations on the dynamic response of cable-stayed bridges subjected to moving loads are summarised in the following: Fleming and Egeseli (1980) [21. is presented. However. 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. In [59. Fleming et al. tests with one and two persons jumping or running were performed. a cable-stayed bridge in Sweden with a 366 m main span. and the bridge was discretized using the finite element method.700 600 500 400 Length of center span (m) 300 200 100 1950 1960 1970 1980 1990 2000 Year of completion Figure 1. 60].2 Review of previous research 1. 9. This includes free vibration and forced vibration due to wind and earthquakes. A two-dimensional (2-D) harp system cable-stayed bridge model with a main span of 260 m was adopted. 22] compared linear and nonlinear dynamic analysis results for a cable-stayed bridge subjected to seismic and wind loads. were considered. The nonlinear dynamic response due to a single moving constant force was also studied.

The experimental models consisted of straight steel beams (cross section = 0. This means that influence lines and superposition technique can be used in the design process. All bridge cables were approximated by linear springs with equal stiffness. and solutions with two to five cables in the main span were presented.2 cm under self-weight was achieved. a 2-D undamped continuous Bernoulli-Euler beam resting on discrete evenly spaced elastic supports. as illustrated in Figure 1.3. the structure can be assumed to behave as a linear system starting from the dead load deformed state for both static and dynamic loads.during the static application of the dead load.22 m/s to .0492 cm × 1. – 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. The vehicle was modeled using one or two constant forces travelling at constant speeds. By prestressing the bridge model. an initial flatness to within ± 0. For the interior span supports. 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. and the road surface roughness was neglected. Considering only seismic loading a similar comparison was conducted in [2] and the same conclusion was made. For the solution of the problem. mode superposition technique was used.6 m could be tested. Only the main span was considered in this study. coil springs were used.36 m) spliced end to end at the supports (springs) so that continuous spans of up to 23. was adopted.97 cm and length = 2. For the theoretical bridge model.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.

– 17 – out using a direct time integration procedure (explicit algorithm). the moving force solutions are good approximations of the exact solutions. Rasoul concluded that bridge damping was one of the important parameters affecting the DAF.. The total vehicle model weight was about 1. Each vehicle was modeled with a sprung mass and an unsprung mass giving a vehicle model with two degrees of freedom (2 DOF). and the effect of vehicle speed and bridge damping on DAF was presented. for both the theoretical and the experimental models. The traffic load was modeled as a series of vehicles traversing along the bridge. The road surface roughness was totally neglected in this study. The results showed good agreement between the theoretical and the experimental work. Alessandrini. For the cable-stayed bridges. axial force in the longest . A simply supported beam. Different traffic conditions were studied. presented diagrams showing. 120.85 m/s. and that the DAF was considerably higher for the cables than for other elements of the bridge. Rasoul (1981) [69] used the structural impedance method1 and studied the dynamic response of bridges due to moving vehicles. and geometric nonlinearities for the cables were considered by using an equivalent modulus of elasticity. The bridge was discretized using the finite element method. the main reasons for the differences in the results were due to the inability of the experimental system to maintain constant speed. 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. fixed pylon heads and with the cables approximated by springs. The train was simulated using moving masses at three different speeds of 60. and 200 km/h. The bridge flexibility functions were evaluated by using a static analysis of the bridge subjected to unit loads. Wilson et al. 2-D fan type cablestayed bridges with steel deck and center spans of about 160. and very simple cable-stayed bridges were studied. the equation of motion of the bridge was formulated in an integral form using the flexibility function (Green’s function) for the bridge. Solutions with different girder damping ratios for a simple 2-D cable-stayed bridge with only two cables were presented.8. a continuous beam. diagrams with different values for the spring stiffness were also presented. The solution was carried 1 In this study. According to Wilson et al. the influence of the speed parameter on the DAF values for displacements and bending – 16 – moments. Wilson et al. To show the influence of cable stiffness. two different analysis methods were used. Brancaleoni and Petrangeli (1984) [3] studied the dynamic response of railway cable-stayed bridges subjected to a moving train. DAF values for mid-span vertical displacement.2 kg. concluded also that increasing the spring stiffness at the supports will for most cases lead to an increase in the bridge dynamic response. 260. namely an approximate method using the concept of continuous beam with intermediate elastic supports. 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. Rasoul found also that for a single vehicle travelling at constant speed. and 412 m were adopted. and the neglection of the inertia effects of the experimental transit load in the theoretical model.

a set of moving masses. the dynamic effects were small if not negligible. The bridge model. Solutions for a total train weight of about 95 tons. For speeds higher than 120 km/h the DAF values increase rapidly and for speeds of about 200 km/h. It was also found that for speeds of up to about 120 km/h. while piano cord wires with a diameter of 2 to 3 mm were used for the cables. for centric and eccentric vehicle movements. were presented. and tests with and without a plank in the main span were undertaken to simulate different road surface conditions.center span cable. A 2-D modified fan type cable-stayed bridge with concrete deck and a main span of 150 m. showed also that bending moment amplification factors were greater than those for cable axial forces and center span vertical displacements. and 200 km/h. represented a 3 span modified fan type cable-stayed bridge with a 200 m main span and about 100 m side spans. 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. Alessandrini et al. were presented. The rail surface roughness was neglected in this study. Based on measured data. 120. concluded that. Three different train speeds were considered. The rail surface roughness was neglected in this study.8 m/s (corresponds to real vehicle speeds of about 10 to 61 km/h) were used. The analysis was carried out using a direct time integration procedure (explicit algorithm). 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. and a four axles 6 DOF sprung mass model. The deck and the two A-shaped pylons were made of reinforced microconcrete. – 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 was discretized using the finite element method and geometric nonlinearities were considered in the analysis. and higher DAF values for the center span vertical displacements. was adopted. Diagrams showing the variation of DAF with speed for the three different vehicle models. Brancaleoni. with or without a plank. which was equipped with rails and a launching ramp. for most cases. Different model vehicle speeds from 0. Brancaleoni et al. For the bridge.5 kg. and for fixed joint and free joint at mid-span. while nonlinear cable elements with parabolic shape functions were adopted for the cables. Walther . treated as a set of moving forces. a Rayleigh type damping producing 2 % of the critical on the first mode has been used. and time histories for the mid-span vertical displacements. and axial force in the anchor cables. Time histories for mid-span vertical displacements were presented. Brancaleoni et al. The bridge deck and the pylons were modeled using beam elements. The rail surface roughness was neglected in this study. were presented and compared with those obtained by the Italian Railways Steel Bridge Code. 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. 60.6 to 3. vertical accelerations were calculated and a study of physiological effects (human sensitivity to vibrations) was undertaken. DAF values greater than those prescribed by the Italian Railway Code were observed. the standard expression for DAF given in the Italian Railway Code were not admissible for cable-stayed bridges.

But he suggested that. a rough surface. based on geometrically nonlinear static analysis (see Figure 1. were presented for different types of vehicle models moving over a smooth surface. Time histories showing cable forces. The maximum DAF value for mid-span vertical displacement was found to be 1. For the sprung mass vehicle model the assumed natural frequency and damping ratio were 1.3). When evaluating the stiffness of each cable. cable-stayed bridge. 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. 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. including only the first 10 modes of vibration.concluded that from the physiological effects point of view. and a sprung mass (1 DOF system). 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. instead of using TMD’s. and pylon tops horizontal displacements. The dynamic response was evaluated using the mode superposition technique. and a bumpy surface. was conducted using the deformed dead load tangent stiffness matrix. was modeled in 2-D.5 % respectively.3.39 Hz and 3. a comparison was made between the dynamic response with and without the presence of a TMD. the structure could be considered acceptable to tolerable depending on the road surface condition. The analysis results showed also that the DAF increases with increasing vehicle speed and can for bumpy surface reach very high values. 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. Khalifa (1991) [49] carried out an analytical study on two cable-stayed bridges with main spans of 335 m and 670 m. Indrawan found that the TMD was very effective in reducing the vibration level of cables anchored in the vicinity of the mid-span. The single axle vehicle was modeled as a constant force. Since Rama IX bridge is equipped with tuned mass dampers (TMD) to suppress wind induced oscillations. and were modeled in 3-D and discretized using the finite element method.3). The dynamic response was evaluated for a single moving vehicle and a train of . viscous dampers should be installed in all cables to more effectively increase the fatigue life of the cables. All analyses were carried out in the frequency domain and time domain responses were calculated using the fast Fourier transform (FFT) technique. The 3 spans cable-stayed bridges were of the double plane modified fan type. The 450 m main span. mid-span vertical displacements. The bridge deck and pylons were modeled using beam elements while truss elements were used for the cables. an unsprung mass. The dynamic response was analyzed using the finite element method and mode superposition technique. the cable sag was considered by using an equivalent tangent modulus of elasticity. where each equation was solved adopting the Wilson-Θ numerical integration scheme. single plane. 2. The linear dynamic analysis. modified fan type. at speeds of 36 to 540 km/h. The TMD was assumed to be installed at mid-span and tuned to the first flexural mode of vibration.

was used. (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. because accurate results of the bridge dynamic response could be obtained based on solving only 8 to 12 equations of motion of the bridge. the type of vehicle model. and deck condition. The results also showed that the magnitude of the dynamic response was influenced by the bridge damping ratio. on the bridge dynamic response. the simulation program DIASTARS. In this study. but comparatively small DAF values were noted at the girder adjacent to the mid-span of the bridge. but increased tremendously with increasing road surface roughness. and an iterative procedure with mode superposition technique was used for solving the equation of motion for the bridge. and – 20 – the roughness of the bridge deck.3). was adopted. The bridge deck roughness was generated using PSD functions. A stress-life fatigue analysis was also conducted to estimate the virtual cable life under continuous moving traffic loads. developed at the Japanese Railway Technical Research Institute. respectively. vehicle speed. 2. vehicle model type. The road surface roughness was generated from a power spectral density function (PSD) (the same as the one used here in sec. The author recommended discretizing each cable into small elements when calculating the dynamic response due to environmental and service dynamic loads. traffic direction. noted that the DAF of all components of the bridge were generally less than 1. travelling with constant speeds of about 43 to 130 km/h over a smooth and a rough surface. Wang and Huang (1992) [84] studied the dynamic response of a cable-stayed bridge due to a vehicle moving across rough bridge decks. A 2-D modified fan type cable-stayed bridge with concrete deck and a main span of 128 m. High values of DAF were noted at the girder near the pylons and at the lower ends of the pylons and piers.vehicles moving in one direction or in both directions. The railway track and the bridge structure were modeled using the finite element method. For the analysis. The dynamic response was analyzed using the finite element method and the geometric nonlinear behavior of the bridge due to dead load was considered. For the sprung mass vehicle model the assumed natural frequency and damping ratio were 1 or 3 Hz and 3 %. Miyazaki et al.2 for very good road surface. The equation of motion for the vehicle was solved using the fourth-order Runge-Kutta integration scheme. The 2-D bridge model together with a simple 12 cars train model consisting of only constant forces were used to evaluate the dynamic . Wang et al. were presented. a 2-D and a 3-D bridge models of a two span cable-stayed bridge. The vehicle was simulated by a nonlinear vehicle model with 3-axles and seven degrees of freedom. 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. cable vibrations. Diagrams showing the influence of bridge damping ratio. The vehicles. were used. Wang et al. concluded that the mode superposition procedure used was effective and involved much less computation. number of vehicles. were approximated using a constant moving force model and a sprung mass model.

while the 3-D bridge and the 3-D train model were used to evaluate the train running quality. decreases the DAF values. 0-400 km/h. see the . noted that idealizing the vehicle as a constant force leads to overestimation of the DAF compared to the sprung mass model. The effects of the pylons flexibility. The 3-D bridge deck was modeled by 3-D beam elements connected to the cables through transversely extended rigid beams. or not including the roughness of the bridge surface. Miyazaki et al. Datta and Surana (1994) [14] presented a continuum approach for analyzing the dynamic response of cable-stayed bridges. the authors recommended different values for the coefficient included in the DAF expression in the Japanese design standard. investigated the influence of vehicle speed. The 3-D Shinkansen train model consisted of 12 cars where each car consisted of a body. 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. For the different bridge members. and the roughness of the bridge surface. and four wheelsets giving 23 DOF. concluded that pylon rigidity and the nature of cable-pylon connection have significant effect on the natural frequencies of vertical vibration. and axial forces in cables. Chatterjee et al. 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. 1. two bogies. using bridge and vehicle models of varying degrees of sophistication. concluded that the examined PC cable-stayed bridge had a satisfactory train running quality (acceptable riding comfort). presented diagrams showing the speed.response of the bridge. The same conclusion was found when assuming that there is no eccentricity in the vehicle path. – 22 – A review of the early work on the dynamic response of structures under moving loads was presented in the paper by Filho [20]. Chatterjee et al. The track was assumed to be directly placed on the bridge deck surface. were considered. it was noted that increasing the axle spacing of the vehicle. deck and pylons shear forces. spacing between first and second vehicle axles. For a thorough treatment of the analytical methods used for problems of moving loads with and without mass in both structures and solids. deck and pylons axial forces.2. coupling of the vertical and torsional motion of the bridge deck due to eccentric vehicle movement. but no effect on those of torsional vibration. In the – 21 – study. The vehicle was simulated using a vehicle model with 3 DOF and 3-axles. Miyazaki et al. eccentrically placed vehicle. And finely.2 Research on other bridge types The dynamic effects of moving vehicles on bridges have been investigated by various researchers. diagrams were presented showing wheel load variations and vertical car body accelerations. For the vehicle. influence on the DAF for the deck and pylons bending moments. Chatterjee et al. and the rail surface roughness was neglected. a comparison was also made with the design value of DAF specified in the Japanese Design Standards for Railway Concrete Structures.

Author(s) Bridge type Vehicle model Surface Other remarks like roughness analysis methods used etc. SSB. SS slab MF. Previous investigations on the dynamic response of other bridge types subjected to moving loads are summarized in Table 1.2 below. Interesting research was also presented by Olsson. analysis of sprung and unsprung mass systems moving along a beam covered with elastic layer of variable stiffness and surface irregularities.excellent book by Frýba [23]. 3-SB cantilever SMS-3-3-2 not considered lumped mass method (1970) [77] . SSB Yoshida et al. moving pulsating not considered theoretical & experimental [36] bridge force study Veletsos et al. MM not considered FEM (1971) [93] Nagaraju et al. (1967) suspension MF. In this book. . were presented.2. The dynamics of railway bridges and railway vehicle modeling are described in the book by Frýba [24] and the book by Garg and Dukkipati [25]. where he derived a structure-vehicle finite element by eliminating the contact degrees of freedom of the vehicle. function Hillerborg (1951) SSB SMS-1-1-2 not considered theoretical & experimental [34] study Hirai et al. see Table 1. The stiffness and damping matrices thus became time-variant and non-symmetric.

SMS-1-1-2 not considered continuum approach. harmonic structural impedance (1975) [26] 2-SB air cushion system sinusoidal method Ginsberg (1976) SSB multiple not considered structural impedance [29] SMS-1-1-2 method Filho (1978) [20] SSB SMS-1-2-2 not considered FEM Blejwas et al. y degrees of freedom. (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. FEM=finite element method – 23 – Genin et al.3-SB MF. SMS-1-1-2. SSB. x-SB=x span beam. . MF=moving force. mode (1973) [57] cantilever superposition Ting et al. MM=moving mass. SSB=simply supported beam. SMS-x-y-z=sprung mass system with x-axles. SS xx=simply supported xx. and in z dimensions. MF.

(1981) [32] 3-SB method Hayashikawa et suspension MF not considered continuum approach. MM. (1979) SS girder & SMS-4-3-3 for not considered lumped mass method [16] truss railway each railcar Gupta et al. SMS-1-2-2 harmonic Lagrange’s eqn. SMS-2-3-2 not considered vehicle braking. 74] models Hayashikawa et SSB.1983) SMS-1-2-2 procedures and vehicle [73. eccentric (1980) [31] plate. SS orthotr. SSB MF. not considered review.SSB MM. 2-SB. MF not considered eigen stiffness matrix al. SSB loading Ting et al. different analysis (1980. with (1979) [7] sinusoidal multipliers Chu et al. mode .

MM. 2-SB SMS-1-1-2 sinusoidal. harmonic FEM. SSB. SSB MF. 10 mm bump finite strip method.al. Rayleigh-Ritz method (1985) [65] . SSB MF. 62] SMS-1-2-2 cosine vehicle element Schneider et al. vehicle [56] plate SMS-3-7-3 braking Olsson (1983. direct time [35] integration Palamas et al. SMS-2-4-3. (1982) [33] bridge superposition Mulcahy (1983) SS orthotr. special bridge1985) [63. 5] study Hino et al. MM not considered used the FEM package (1983) [71] ADINA Arpe (1984) SSB SMS-2-4-2 not considered theoretical & experimental [4. (1984) 1-SB cantilever SMS-1-1-2 not considered FEM.

. 5-SB. 2-SB.2 (continued) Olsson (1986) SSB. (1986) SS PC railway SMS-4-23-3 PSD lumped mass method [17] Honda et al. bump at 1 vehicle & multiple (1986) [37] SB. special bridge– 24 – [64] 6-SB SMS-2-4-2. MF.pothole Chu et al. vehicle element. MM. SSB entrance groups of vehicles Table 1. vehicle SMS-2-6-2. PSD for interaction (1987) [44] force Bryja et al. SMS-2-7-2 braking Inbanathan et al. SSB MF. 3-SB. SMS-1-2-2. 2-SB. not considered FEM. MM considered FEM. 4SMS-1-2-2 PSD. suspension multiple MF not considered random highway traffic (1988) [10] bridge Diana et al.

mode (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. SS truss SMS-4-23-3 for each PSD lumped mass method (1991) [82] railway railcar Huang et al. 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.suspension SMS-4-23-3 for each not considered FEM. SSB SMS-2-2-2 PSD continuum approach. continuous SMS-3-12-3 PSD . different traffic (1988) [19] bridge railcar conditions Coussy et al. PSD traffic simulations. SSB SMS-2-4-2.

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

describes a computer program developed using FEM and mode superposition to solve the (1994) [78] dynamic interaction problem between high speed railway vehicles. different traffic (1994) [92] SMS-2-6-2 conditions Chatterjee et al. SMS-1-3-2. not considered FEM. PSD continuum approach. mode [12] SMS-1-2-2 superposition Chatterjee et al. mode (1994) [15] bridge SMS-3-3-2. slab on SSB’s MF. each of SMS-4-31-3. 4-SB . SMS-3-6-3 superposition Wakui et al. (1994) SSB. arch bridge MF not considered mixed and lumped mass (1995) [13] method Green et al. 2-SB moving pulsating force. suspension SMS-1-1-2.Wang (1993) [87] SS truss SMS-4-23-3 for each PSD lumped mass method railway railcar – 25 – Cai et al. not considered continuum approach. and railway structures Yener et al. 3-SB.

different traffic (1995) [42] plate conditions Lee (1995) [51] 2-SB. SS orthotr.SMS-4-11-2 PSD. 20 mm compared leaf sprung with (1995) [30] bump air sprung vehicles Huang et al. arch. MF not considered beams on one-sided point 4-SB constraints Lee (1995) [52] SSB rigid wheel not considered unknown wheel nominal motion. box ambient & controlled dynamic bridge testing (1995) [67] girder . one and two trucks (1995) [41] I-girder Humar et al. hor. SMS-1-2-2 not considered FEM. curved SMS-3-12-3 PSD FEM. thin walled SMS-3-12-3 PSD FEM (1995) [40] box-girder Huang et al. 3-SB. FEM Paultre et al.

as this is adequate for large span bridges. the assumptions made by those authors are acceptable. On the other hand. 91] 5-SB SMS-3-6-2 vehicle element Table 1. where the bridge elastic displacements were neglected and only the excitation caused by the road surface roughness was considered. The opposite assumption was made in [45]. and the vehicle mass to the bridge mass ratio is low. 55]. and in [69.2 (continued) 1. neglecting the vehicle inertial effects. according to Frýba. in evaluating the dynamic response of . 3-SB. the formulations for the coupling equations (equations (2. (1995) SSB. Of course. as Frýba [24] pointed out. the road (or rail) surface roughness was neglected and only the elastic displacements of the bridge. The vehicle inertial effects. incorrect. However. 22. SMS-1-2-2. including all inertial terms.4a-c) in section 2. the road surface is smooth. 3. Only the models developed in [84. the most detailed vehicle model used consists of two degrees of freedom. were considered.3 General aims of the present study In all the aforementioned studies on the dynamic behavior of cable-stayed bridges. and the bridge displacements were considered in [49]. 89]. In the work presented here. 8. the road surface roughness. the vehicle models used are very complicated and. PSD FEM. MF. special bridge[90. MM. The main aims of this study are as follows: • to develop a general but simple analysis tool which fully consider the bridgevehicle interaction. caused by the varying position of the vehicle. In [21. if the main purpose is to study the bridge dynamic response. according to the author’s opinion. 14] are believed to be general and handle the bridge-vehicle contact problem correctly. if for example the vehicle is travelling at low speed. the vehicle was modeled as a constant moving force. authors either used very simple vehicle models.1) are. very detailed and complicated vehicle models are unnecessary. or very complicated and time– 26 – consuming vehicle and bridge models.traffic Yang et al.

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. where the dynamic interaction between the bridge and the vehicle is included by utilizing an iterative scheme. 1996 [48]. a moving mass model. as the case is for cable supported . Moreover. – 28 – Chapter ______________________________________________________________________ Vehicle and Structure Modeling _____________________________________________________________________ _ 2. springs and dampers. on the dynamic response. these inertia effects are mainly caused by bridge deformations (bridge-vehicle interaction) and bridge surface irregularities. when the bridge span is considerably larger than the vehicle axle base [24]. large vehicle mass. In the present study. such as vehicle speed and bridge deck surface roughness. which uses the finite difference method and the mode superposition technique for dynamic response evaluation. – 27 – For this purpose a computer code has been developed using the MATLAB language [53]. For a vehicle moving along a straight path at a constant speed. and a sprung mass model with two degrees of freedom. The implemented code has been verified by comparing analysis results with those obtained using the commercial finite element code ABAQUS [1]. the vehicle models used in this study include a moving force model. flexible bridge structure. and can be modeled by a set of lumped masses. 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. to show that the proposed simplified analysis method. small bridge mass. This model is acceptable. Part of this work was presented earlier at the 15th Congress of IABSE. Hence factors that are believed to contribute in creating vehicle inertia effects include: high vehicle speed. trailers and suspension systems.1 Vehicle models Heavy vehicles consist of several major components.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. stiff vehicle suspension system and large surface irregularities. As illustrated in Figure 2. 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.1. Copenhagen. Special emphasis is put on verification of the proposed model and on investigating the effects of local and global irregularities on the dynamic response. the adopted sprung mass model is a one-axle vehicle model of a real multi-axle vehicle. such as tractors.

and denoting the contact force between the bridge and the vehicle by F( t) .bridges. the following equations of motion can be established [23.1.1 Vehicle modeling Considering the sprung mass model. 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. Heavy roadway vehicles generate most of their dynamic wheel loads in two distinct frequency ranges [30]: body-bounce and pitch motions at 1. Very detailed vehicle models are unnecessary and will not bring any great advantage. 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. defined positive when it acts downward on the bridge.5-4 Hz and wheel-hop – 29 – motion at 8-15 Hz. 63]: 2 −( dw dw dw m+m 1 2 1 1 2)g−m +k 1 . This explains the increase of some of the specified DAF in Figure 1.1. when the main purpose is to study the dynamic response of bridges. shown in Figure 2. for bridges with a fundamental frequency in the range of 1 to 5 Hz.

respectively.2) are the dynamic equilibrium equations for the unsprung mass and the sprung mass.2) giving: – 30 – . and g the acceleration of gravity. w ( t 1 ) and w ( t 2 ) are the displacements of the vehicle unsprung mass m 1 and the vehicle sprung mass m 2.1.1) dt 2 2 1  d t d t  + = d2 2 w w w −m −kw−w−c S S − = 2 ( 2 1) d d 2 1 0 (2.(w−w)+ c −  F( t S S ) 0 (2.1) and (2. cS the damping coefficient of the viscous damper.2) d2 t dt dt Equations (2. Referring to Figure 2. kS the stiffness of the linear spring connecting the two masses. 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. respectively.1) and (2.

x ( t) = x ( t v ) (see Figure 2.2).  dt 2 and for the moving force model F = ( m + m ) g 1 2 . 63]: w t = y( x t . the following coupling equations for the point of contact.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. Assuming that the vehicle never loses contact with the bridge (that is F( t) > 0). must be fulfilled [20.t ) + r( x t 1( ) () ( )) (2.4a) y ∂ ∂y r ∂ .  d2 w  The contact force for the moving mass model will be F( t) = ( m + m ) g + 1 1 2 .2 dw 2 dw F( t) = ( m + m 1 2 1 2)g+m +m 1 (2. and that the deformation between the unsprung mass center and the bridge deck center line may be neglected.

& w ( t) =
v+
+
v
1
(2.4b)
x

t

x

2
∂y
2
2
2
2
∂y
∂y
∂y∂r 2∂r
& 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
=−δ

g
( x xv) F( t
gg
) (2.5)
x
∂4
∂t 2
where δ 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 α
i
(2.8)
Li
Denoting the allowable cable stress by σ 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)
σ sinα
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
0≤x≤
i

γx
2
c
+


1
E
3
c
12σo
where Ec is the modulus of elasticity for the straight cable, γ c the specific weight of the
cable material, and σo 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 σo is here set
equal to σ g , which is the tension stress due to dead load qg and is given by [9]: q
σ=σ
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
0≤x≤

13) tanα 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. σH 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 cosα = = x i i i (2.14) σ H cosα a o L/2 q+q 2 tot g q L .

A = A d= o ∫ o( x) x (2. The elongation of the anchor cable. the dead weight of the side spans are not included because the side spans are not considered in this derivation.15).15) σH α8 0 a cos o In equation (2. and the vertical displacement in the main span at joint i.17) 2 cosαo EA α o o cos o . the horizontal displacement of the pylon top.16) E A cosα oo o L ∆ FL cosα b o i o i = = (2. due the force Fi in cable i are: α ∆ FL L i o i = cos o (2.

20) 1 L 2 o .19) L cosα x o i (δ i = ) 1 2 2 a F sinα E A H cos α 1 k i i = =oo o i (2. and the spring stiffness are: 2 (δ = ) 1 E A H cos α 1 Fi =oo o i (2.18) H The internal force in cable i due to the vertical displacement δ = i 1.– 35 – δ x = i b (2.

x After substituting (2. 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 ≤  (2.14) into (2. the force Ti gives the horizontal displacement: L ∆ TL b o i = = o (2.22) 3EI pp Thinking of the pylon as a fictitious anchor cable having the fictitious area Ap .20). a horizontal force at the top gives the horizontal displacement: TH3 b i p = (2. Referring now to the pylons.21) σL2 x  2 a o Where E o is evaluated according to equation (2.10).23) cosα 2 o EA .

the resulting spring stiffness that includes the effect of the elongation of the cables in the main span.23) for b. and the stiffness of the pylons is: 1 . equation (2.25) σL2 A tot x  2 a o o And finally. the elongation of the anchor cables.22) and (2.21) is modified to give: E LH + tot o + a s (qq g q)A A o p 1  L k ( x) = 0≤x≤  (2. we find that the fictitious cable area is: – 36 – 3EIL A pp = o p (2.24) EH3 2 o p cos αo To approximately include the effect of the pylons.α o p cos o Equating both equations (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. qg=12·104 N/m. H=30 m. qq=5·104 N/m.26). L k( x) = 0≤x≤  (2. Ec =2.5 free pylon top. Hp=60 m.0 0.3 Spring stiffness for a fan-shaped cable-stayed bridge with L=150 m. I p =0 ) (10( x k 1. k( x).4 shows the fixed pylon top curve in Figure 2.3 shows the spring stiffness. Ep=0. and Figure 2. 0. Figure 2. for a fan-shaped cable-stayed bridge with L=150 m.0 0 25 50 75 100 125 150 x (m) – 37 – Figure 2.0 )2 fixed pylon top N/m free pylon top 6 1. and γ c =9·104 N/m3 2. σ a =720·106 N/m2.5 Spring stiffness.0 )2 .1·1011 N/m2.3 but with the suspended side spans included. 2. Ip=25 m4.35·1011 N/m2.

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]. Ls=50 m. S(Ω) . 0.8⋅ − 10 4 0 < Ω < 0. and n is the spectral roughness exponent. the bridge road surface roughness is simulated using the onesided PSD function:  35.1-2.4 0. included Equations (2. define the analytical model of the problem.5) and (2.05 < Ω  .3 but with the side spans. 2.4 Spring stiffness for a fan-shaped cable-stayed bridge (fixed pylon top) with the same input data as in Figure 2.0 0 50 100 150 200 250 x (m) Figure 2. – 38 – In the present study. of road surface roughness is approximated generally by an exponential function [38.27) in which Ω is the wave number. The PSD.28) 00 .5 Spring stiffness. 107 ⋅ − 10 4 Ω−1 9.05 S(Ω) =  (2.26). a is the spectral roughness coefficient. together with the boundary and initial conditions. 46] as: S( ) = a − n Ω Ω (2.N/m6 1.5 ) (10( xk 1.0 0.

31) M where M is the total number of wave number increments in the interval between the minimum wave number Ωmin and the maximum wave number Ωmax in the defined spectra. and Ω i is the i th wave number within the specified PSD interval and is given by: Ω=Ω + ∆Ω − i min ( i )1 (2.5.30) in which the wave number increment ∆Ω is defined as: Ω −Ω ∆Ω = max min (2. The amplitude α Ω i is related to the PSD function S( i ) by: α2 S( i Ω ∆Ω = i) (2. and shown in log-log by the bold line in Figure 2.  i is a random phase angle with uniform distribution in the interval {0.29) i=1 where α i is the amplitude of the sinusoidal wave. The exponent n and the coefficient a were calculated by the least square method using measurements on 56 bridges.32) 2 – 39 – 10 )3 .<1 presented in [38]. 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) = ∑α si ( nΩx+ i i i ) (2. 2 } π . The unit of the wave number is (m-1) and the unit of the spectrum is (m3).

6.02 Surface roughness (m) 0 50 100 150 Distance along the bridge (m) Figure 2. To compare the analytical PSD.02 0. S( 0.6 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 .01 0. equation (2.1 1 Wave number.6 shows a sample of the simulated random road surface roughness profile used in this study for bridges with L=150 m.01 PSD of surface roughness.001 0. Figure 2.Analytical m-4 Simulated 1 ) (10Ω 0. used in equation (2.01 0 -0.28).29). a power spectrum analysis of the sample is performed using the following relationship [83]: 1 S .01 -0. with the PSD of the sample shown in Figure 2.1 0. Ω (1/m) Figure 2.5 Power spectral density of road surface roughness 0.

the spectral estimates corresponding to negative wave numbers are removed and a compensation is made for them in the positive side. The differential equation governing the free vibration of the bridge model is obtained by setting the right-hand side in equation (2.t) . and S Ω sim ( ) is the resulting PSD estimate for the sample.t +k + = gg ( x) y( x. The resulting one-sided PSD estimate for the sample in Figure 2.5) equal to zero.33) N  where F is the length N fast Fourier transform (FFT) of the sample.1 Dynamic analysis 3. 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.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.6 is indicated by the fine solid line in Figure 2.(Ω) =  F 2 sim  (2. giving: 4 y( x.t) ∂ ∂ 2 y x.5.1. 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. Using the symmetry property of real FFTs.

1) gives: i(t) d 4 zi( x) EI + k( x) z − ω2 = 0 (3. giving n-1 unknowns.3) 4 i ( x) m z g ii(x gg ) dx Dividing the bridge model into n equal segments each of length h.4) . t) = z ( x) φ ( t) = z ( x)( a cosω t + b sinω t i i i i i i i i ) (3. t) can be expressed as a product of two functions.3) with its finite difference approximation.() EI m 0 (3. By expressing φ as in equation (3. the following eigenvalue problem is obtained2: A Z = λ Z (3.1) x4 g ∂ t ∂2 The displacement function y( x. called the mode shape function or eigenfunction z( x) . the displacement at any location varies harmonically with time and can be expressed as [88]: y ( x.2) and substituting into equation (3. one involving only the space coordinate x. and replacing the derivative in equation (3.2) – 43 – where ω i is the i th circular frequency of the free motion of the bridge. Then when the bridge vibrates in its i th natural bending mode. and the other one involving the variable time and called the generalized coordinates ( φ t) .

5a.b)  1 −4 c −  0 λ 3 4 1  n−1( n−1 x n− )   1  1 .where c −4 1 0 1 −4 c −   2 4 1   1 −4 c −4 1  λ 0  3  1  A=  OOOOO  .λ= O  (3.

L zn−1 . L.−4 c − 2  4 0 1 −   4 c 1 ( n−1 x n− )1 z. K) λ= ω. 3 .n−1( n−1 x n− )1 4 g mh k 2 ( x 1) k 4 ( x 2 .5c)  zn−11 . z 1 2 n−1 ] =  M M  (3. =5+ and =6+ 4 i i c h c 1 . z . L z   11 1 .n−1  Z = [z .

Using the orthogonality properties of the eigenfunctions and setting the normalization constant4 equal to unity.6) i=1 Substituting equation (3. h (3.4). [88] for more details on derivation): 2 . t) = ∑ z ( x) φ i ( t i ) (3. The first term in equation (3.2.7) i=1 Further. 2 If nothing else is mentioned.5d. – 44 – Mode shapes and corresponding circular frequencies are now obtained numerically by solving equation (3. the first term mentioned above is evaluated as: 6+(η-1)/(η+1). the following equation is obtained (see ref.6) into equation (2.5) gives: ∞ ∑ ( E I z IVφ + k( x) z φ + m z φ& ) = −δ ( x − xv) F ggi i i i gi i (3. 3. both sides of the above equation are multiplied with eigenmode z j and integrated3 over the length of the bridge.2 Response of the bridge Using mode superposition technique.f) E 3.5) can be written as: ∞ y( x.K g Ig EgIg EgIg where i is the mode number and column i in Z gives the i th mode shape.5e) for c 1 must be changed from 5 to 7. bold upper case letters are used for matrices and bold lower case letters for vectors.1. the solution of equation (2. The transpose of a matrix or a vector is denoted with the superscript T. where η= K h/(2 E I m g g). if fixed end supports are assumed. If rotational springs with stiffness Km are introduced at end supports.

8) will be the value in the i th eigenvector z i and the row number corresponding to the point of contact (i.1. which corresponds to having the vehicle load applied only at the segment joints.F φ& + ω φ = − i i i zi ( xv ) (3. When dividing the bridge model into n equal segments each of length h and setting the time step to t ∆ = h / v . the term z ( x i v ) in equation (3.8) is: Fz + φ ω2 o φ& i ( xv ) . L 3 The properties of the Dirac function gives δ ∫ ( 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) mg where the contact force.e. where the contact force is applied). is time depended and therefore evaluated at each time step using the expressions given in section 2. F.

when using the orthogonality and normalization relationships. Substituting into equation (3.6).( Fzi( xv) m i i g) φ=− + cos ω t oi ∆+ sin ω t ∆ i (3. be evaluated as [88]: φ 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.10a) &φ T = oi i z & yo (3.9) ω2 m ω2 m i i ω i g i g i where φ and φ& o i o i are the initial values for the step and can. the bridge displacement vector at each time step is .

by differentiating equation (3.obtained from the sum: s  Fz x Fzx 2m & i(v) ( i( v)+φ ω i o i g)  y=∑ φio  z− + cos ω ∆ t + sin ω ∆ i i i t (3. s ≤ n − 1. as: s  ( Fz 2 .11).11) 2 2 i 1 ωm ω  =  m ω i g i g i  where s is the number of included modes. The expressions for the vertical velocity and acceleration vectors for the bridge are obtained.

12) i 1 ω  =  i mg  s  ( Fz 2 i ( xv ) + φ ω i o i mg )  &y=∑ z &  − cos ω ∆ t − φ ω sin ω ∆ i i i o i i t (3.13) m  .i ( xv ) + φ ω oi i mg )  &y=∑ z &  − sin ω ∆ t + φ cos ω ∆ i i oi i t (3.

and acceleration vectors are now collected in matrices for processing and plotting. the tolerance was set to 5 10 6 ⋅−.15) i=1 where ′′ z i is evaluated using the finite difference approximation. the following convergence criterion is considered: max( y − y j j−1 ) ≤ Tolerance (3. The recalculation of bridge displacements is repeated in each step until convergence is obtained. giving: s m=−E y′=− ∑ z ′φ gIg Eg Ig i i (3. The mode superposition technique is also used to obtain the dynamic bending moment vector at each time step. velocity. The dynamic interaction between the bridge and the vehicle is included by utilizing the iterative scheme shown in Figure 3. set initial values for the vehicle sprung mass and for the unsprung mass predict 1 . a computer code has been developed for analysis of the dynamic response. For the results presented in this study.14) max( y j ) where y j is the j th estimate for the bridge displacement vector at the current time step. x = x + ∆t v vv Based on previously determined quantities.1.i=1  g  – 46 – Using the MATLAB language [53]. The obtained displacement. In this study. – 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. moment.

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.15) to determine Calculate new the bridge response using mode superposition 1 w .2 Static analysis Using the same discretization technique as in section 3.16) and (3. (3.18) End Figure 3. (3.13) and (3.1. &1 w .5a) and: . the static displacement vector is obtained from the expression: + 3 st (mm 1 2)gh y=− A−1 p j j (3.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.11-3.1 Overall computational scheme – 48 – 3.16) Eg Ig where A−1is the inverse of the matrix given in equation (3. &1 w .

1 j ) (3.18) 2 h – 49 – – 50 – Chapter .j ( st st st yi+ + . gives the i th element in the static moment vector st m j to be equal to: EI st gg =− − i m. Approximating the second derivative of the displacement by its finite difference. The static bending moment is computed using the expression given in equation (3.15).17)  0 M   0 where j is the joint number where the vehicle load is applied.0  M  0  j th row p = 1 ← j  (3. 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. j yi− . 1 2 j yi.

is the simplest one. As the position of this node is time dependent. DASHPOT2 and MASS elements. This code is considered as one of the most powerful finite element codes available on the market today . calculated according to equation (2. For the interaction problem the SLIDE LINE option was used to define the surface on which the vehicle and the bridge may interact.. The number of elements. was increased until a converged solution was obtained. From ABAQUS element library. and the cable density γ = c 9000kg / m3 . The results obtained from these approaches agreed well with those from the above described modeling procedure but the implementation became much more complex.______________________________________________________________________ Numerical Examples and Model Verifications ______________________________________________________________________ 4. the present model results are compared with those obtained using a commercial finite element code.1 General In this chapter. 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. and the NO SEPARATION parameter – 51 – was included in the SURFACE BEHAVIOR option. The finite element code ABAQUS/Standard 5. e. described in chapter 2 and 3.4 [1] was used for this comparison and was run on a SUN SPARCstation 20. described above. Therefore. In ABAQUS. At the node where the vehicle and the bridge interact.10). Other modeling approaches.g. The modeling approach used in ABAQUS for the bridge-vehicle contact problem. The vehicle was modeled using SPRING2. The time increment was set equal to the total time needed for the vehicle to cross the bridge divided by the number of increments. were tested for simple problems. but still very complicated and requires a lot of time and experience in ABAQUS and in finite element modeling. the equivalent tangent modulus of elasticity. a special purpose contact element ITT21 was used. two beam examples and two cable-stayed bridge examples are studied and results obtained using the present model. 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. developing user subroutines in FORTRAN and including them in the input file for ABAQUS or using another type of contact definition. For verification. results using these more complex approaches are not . are presented. The number of increments was chosen for each numerical example so convergent results were obtained. were given in the ABAQUS input files. the AMPLITUDE and the BOUNDARY options were used to modify this position for each increment. For each cable in the cable-stayed bridge examples. in each numerical example.

In MATLAB the CPUTIME command was used to measure the CPU time needed to run the program. the impact factor defined as I=100*(DAF-1) (%) is found sometimes instead.1 m/s (α =0. The dynamic effects are then caused only by the varying position of the force. n. moving at constant speed. is the easiest problem to tackle and belongs to the very few moving load problems that can be solved analytically. v. t) F =347000 N x E I =9. This value was then compared with the USER TIME value given at the end of the ABAQUS output file. The program developed for the present model. v = 68. using the MATLAB language [53]. see [64]. was also run on the same SUN SPARCstation 20 machine. – 52 – Adopting the moving force model means that the influence of the inertia of the vehicle mass and the bridge-vehicle interaction are neglected. moving force model The simply supported bridge subjected to a constant force. for low values of the speed parameter α (defined later in equation (4. In the following examples.1 Simply supported bridge subjected to a constant moving force Using the . This made it possible to not only compare the obtained solutions but also the time consumed by the computer. 4. were found using a trial and error procedure. are presented. dynamic amplification factors5 for displacements (DAF d) and for bending moments (DAF m). 10 Nm2 gg x ( t) m =11400 kg/m v g L =34 m Figure 4. which also determines the time step used for the analysis.4)) and low values of the vehicle mass to bridge mass ratio. the number of modes that participate in the analysis and need to be considered and the number of segments. and all the ABAQUS curves presented in this study are those obtained using the SLIDE LINE and the ITT21 contact element approach described above.2 Simply supported bridge. 5 In the literature. For the present model solutions . F.presented here. Still this is a good approximation.92 10 .25) y( x. 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.

t) = ω (4.2) .t)= ω (4.t) .1) 4∑ sin i t sin t sin 48 E I i 4 2 2 π =1 i (1− α / i ) ()   L  − i   L  gg i FL ∞ 8 1  πvα   iπ x M( mx ( x.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  πvα   iπ x y( x.notations of Figure 4.

2∑
sin i
t
sin
t
i
sin
4
2
2
2
π =1 i (1− α / i )
()

 L  − i

 L 
i
where i is the mode number, ω i the circular frequency for the i th mode of vibration, and
α the non-dimensional speed parameter. ω i and α are defined as:
π 2 E I
ω
i
gg
=
i

(4.3)
L 
g
m
π
α=
v (4.4)
ωL
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

4 0.1 It can be seen from the figures above that a negative dynamic mid-span moment is obtained when the vehicle enters the bridge.2 Normalized vertical mid-span displacement 1.1 – 55 – -0.4 0.2 0.8 1 Vehicle position.3 Moment at mid-span (normalized with respect to maximum static moment) versus vehicle position for the problem defined by Figure 4.0.2 0 0.6 0.2 ABAQUS 0. and that the dynamic moment curves looks similar to the dynamic displacement curves but are more irregular. .2 Vertical displacement at mid-span (normalized with respect to maximum static displacement) versus vehicle position for the problem defined by Figure 4.8 1 1.6 0.4 0 0.8 1 Vehicle position.8 Normalized mid-span moment 1 1. x v / L Figure 4.2 Exact 0 Present Static (present) 0.2 0. x v / L Figure 4.4 0.6 0.6 0.

258. When comparing with the exact solution.48 and 1.41 were obtained for the maximum normalized displacement and the maximum normalized moment. as shown in Figure 4. The time consumed by the computer to solve the problem using the exact model. and Table 4.258.179 for x/ L= 0. respectively. The corresponding data for the bending moment (corresponds to DAF m) are 1. the convergence of equation (3. However.093. [20]. the results also indicate that the moving load problem in ABAQUS was correctly modeled. and 70 CPU seconds.259 for the exact model.3 Multi-span continuous bridge with rough road surface A four-span continuous bridge model with one fixed end. respectively.3.11) converges very fast and may therefore be truncated after few modes (5 modes for the current example).2. the results from the present model for the maximum normalized midspan displacement can also be compared with the results 1. From the present model solution. the present model. and 1. From Figure 4. 4.260 for x/ L = 0. and 1.2 are 1. respectively.258. a better method for calculating the dynamic bending moment is described. Figure 4. for different vehicle positions. in Figure 4. However. values of 1. 1. the solution of equation (3. respectively.2 and Figure 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. This problem was chosen to illustrate the flexibility of the proposed analysis method. the x-value where the maximum response occurs – 56 – may differ from L/2. and the ABAQUS model.6 .3.15) for the bending moment is much slower. and [93]. One should remember that in this example only the mid-span ( x = L/2) vertical displacements and mid-span moments are presented.251 given in reference [64].093.259.The maximum normalized displacement (corresponds to DAF d) in Figure 4. For the present model. 1. the present model. 14.083.12 10 . In addition. was adopted for this investigation. v = 30 m/s y( x. t) m =31700 kg c =8. and 1. 1.4. as well known. and the ABAQUS model was about 2. In reference [57]. 104 Ns/m 2 s m =3000 kg k =9.

92 1010 Nm2 gg xv( t) m =11400 kg/m g 20 m 25 m 60 m 45 m Figure 4. was adopted. 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. The vehicle characteristics are similar to those used in reference [65] and correspond to a fundamental frequency of 2.4 Idealized four-span continuous bridge with one fixed end and rough road surface subjected to an idealized 2 DOF vehicle. The corresponding time for the present model was about 240 CPU seconds which is only about 42 % of the time consumed by ABAQUS. respectively. For the ABAQUS model. After conducting some convergence study.7 Hz and a damping ratio of 8 %. the dynamic amplification factors for displacement and bending moment.2. These values can be compared with the DAF values for the center point6 of the bridge DAF d =1. 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. See chapter 2 for the description of the symbols – 57 – In this example.705. it is interesting to note that only the “Swiss SIA-88. were found to be DAF d =1. 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. =9. For the fixed end. . The 2 DOF vehicle model.. comparison was also made here between the present model solution and the one obtained using ABAQUS.126. 6 N/m 1 s x EI . the DAF m was as high as 1. consisting of one sprung mass and one unsprung mass. section 4.296.1. calculated using the present model.6.387 and DAF m =1. single vehicle” curve considers such high DAF values. the road surface roughness was considered by using the profile shown in Figure 2. Studying Figure 1. Considering the entire bridge. As in the simply supported bridge example. the computer needed approximately 570 CPU seconds to solve the problem.396 and DAF m =1.

and it is not possible to infer which solution is closer to the correct one. – 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. 6 For a certain point P. bending moments. see section 4.6 for the discussion of the numerical results.00 Hz) Mode 5 (12. natural frequencies.27 Hz) Mode 3 (6.26 Hz) Figure 4. and vertical acceleration. The presented results indicate very good agreement between the two different solutions and much less computer time consumption for the present model.04 Hz) Mode 2 (4. 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. are presented.5 The first 5 mode shapes calculated using the present model . some time histories of vertical displacements. The differences in the two solution methods are believed to cause the differences in the results. mode shapes. One can observe that the agreement for the dynamic curves in Figures 4.In the following pages.7 is not as good as for the first numerical example.48 Hz) Mode 4 (10.6 and 4.

56 2 4.04 2.00 3 6.77 69.26 12.07 .27 4.87 78.04 14 53.79 64.48 16 68.48 6.20 5 12.Mode Present ABAQUS Mode Present ABAQUS number 300 150 number 300 150 (bending) segments elements (bending) segments elements 1 2.27 18 87.27 15 63.39 53.00 17 77.03 4 10.00 10.

82 10 33.20 24 139.41 14.46 6 14.56 26.53 12 43.58 16.63 23 131.60 .42 8 22.72 9 26.45 140.87 108.60 22 123.79 94.10 11 38.64 22.87.04 123.60 20 107.21 132.67 21 112.12 38.06 112.28 7 16.42 19 93.57 33.

0 -2.2 0.0 ABAQUS Vertical displacement (mm) Static (present & -8.25 50.2 Comparison of the first 25 natural frequencies (Hz) for the problem defined by Figure 4.43.0 0.0 Vertical displacement (mm) Static (present & .4 – 59 – 4.4 0.52 157.71 25 156. x v / L 4.73 13 50.0 Present -6.0 (b) 2.0 -4.0 0.0 ABAQUS ) -10.0 (a) 2.0 0 0.38 Table 4.8 1 Vehicle position.0 Present ABAQUS -2.6 0.

ABAQUS ) -4.0 0 0.6 0.2 0.8 1 Vehicle position.8 1 Vehicle position.4 0.2 0. x v / L Figure 4.6 Vertical displacement histories for the problem defined by Figure 4.5 m) – 60 – 1000 (a) 0 -1000 -2000 Present Bending moment (kNm) ABAQUS -3000 Static (present & ABAQUS ) -4000 0 0. x v / L 1500 (b) 500 -500 -1500 Present Bending moment (kNm) ABAQUS -2500 Static (present & ABAQUS ) . (b) at the center of the last span (x = 127.6 0.4: (a) at the center of the bridge (x = 75 m).4 0.

2 0.4: (a) at the center of the bridge (x = 75 m). x v / L Figure 4. x v / L 4.0 (b) )2 2.0 Vertical acceleration (m/s Present ABAQUS -4.0 ABAQUS -5.5 m) – 61 – 15.0 0 0.4 0.8 1 .8 1 Vehicle position.0 (a) Present 5.0 -2.0 Vertical displacement (mm) -25.6 0.6 0.7 Bending moment histories for the problem defined by Figure 4.6 0.4 0.0 0. (b) at the center of the last span (x = 127.0 0 0.2 0.0 -15.4 0.8 1 Vehicle position.2 0.-3500 0 0.

σ a .3 kN/m2.3. qq. qg. this 2 DOF vehicle model and the speed of 30 m/s (108 km/h) were used when calculating the response.) is equivalent. were chosen for the ratios of side span to the main span length and pylon height to main span length. (b) sprung mass vertical acceleration – 62 – 4. respectively. The girder is also supported vertically at the pylons but is totally independent of the pylons. for example. The characteristics of the 2 DOF vehicle model and the road surface roughness profile are similar to those used in the previous example. to 0. This gives a load ratio of qq /q g = 0. (including surfacing.8 Displacement and acceleration histories for the sprung mass of the vehicle: (a) sprung mass vertical displacement. If nothing else is stated in connection with the presented figures in the following pages. Rotational springs with stiffness Km were introduced at end supports of the main span girder to simulate the influence of the approach spans stiffness. corresponds to 45 % of the chosen ultimate tensile strength for the cables which is 1670 MPa. x v / L Figure 4. The cross-sectional area of the cables and the corresponding cable spring stiffness were evaluated using the expressions given in section 2. railings.35. cables. The geometrical configuration and the mechanical properties are shown in Figure 4. which also allows for point loads. t) v = 30 m/s K K m m 40 m 30 m x 10 m xv( t) 50 m 150 m 50 m . was chosen for this numerical investigation.2. Optimal values. y( x.9. etc. see [9]. corresponds to a uniformly distributed load of 4.50 m of concrete for a bridge width of 14 m. anchorages. The allowable cable stress. close to that of practice. The chosen live load.Vehicle position. corresponding to 1/3 and 1/5.4 Simple cable-stayed bridge A simple fan-shaped cable-stayed bridge with rough road surface and with only the main span suspended. The chosen dead weight.

12.5. The results are shown for two alternatives. respectively. However. Figures 4.12 are underestimated. compared with the ABAQUS model. vehicle speed. which consider the anchor cable stiffness and the pylon stiffness. are presented.11 and 4. the simple expressions derived in section 2. while the present model required 300 segments with 25 included modes. The bending moment responses shown in Figure 4. 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.9 Idealized fan-shaped cable-stayed bridge with rough road surface subjected to an idealized 2 DOF vehicle. a better agreement will be obtained at this vehicle position. 0 7 14 21 28 35 42 49 56 63 70 77 . some time histories of vertical displacements and bending moments.30·1011 N/m2.0·1011 N/m2. and γ c=9·104 N/m3. by increasing the number of modes considered to 50. namely.Figure 4. specially for the fixed pylon top alternative. gave larger static and dynamic displacements for the present model. See section 4.3. Ec =2. The computer time consumed for the two models was about 625 and 325 CPU seconds. Ip=10 m4. at xv / L=0.3. and the influence of the rotational spring stiffness. by the present model.6 for discussion of the results. and of course more CPU seconds will be consumed. Ig=4 m4. qq=6·104 N/m. one can observe good agreement between the solutions of the two different methods.30·1011 N/m2. Ep=0. Mode shapes. different vehicle models and road surface roughness on the dynamic response. Eg=0. σ a =750·106 N/m2. For the free pylon top curves.2. qg=17·104 N/m. natural frequencies. Studying Table 4. fixed and free pylon top.

84 91 98 105 112 119 126 133 140 147 Mode 1 (0.73 Hz) Figure 4.55 .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 0.56 Hz) Mode 3 (2.94 0.53 Hz) Mode 4 (3.76 2 1.76 0.94 Hz) Mode 2 (1.93 Hz) Mode 5 (5.56 1.94 0.

40 9 16.71 5.92 5 5.1.44 13.92 3.73 5.53 2.72 16.72 10 .38 1.46 8 13.88 7.93 3.71 6 7.53 2.41 2.75 16.42 13.50 4 3.49 10.46 10.68 5.90 7.89 7 10.92 7.85 3.47 10.77 16.40 13.42 3 2.

72 14 38.37 50.41 20.47 12 28.47 44.47 24.45 20.37 50.30 50.62 .78 38.51 24.62 56.90 13 33.93 38.52 24.52 44.78 15 44.76 33.20.75 33.64 56.64 56.96 28.41 11 24.95 28.72 33.52 16 50.90 28.30 17 56.46 44.92 38.46 20.

27 70.44 85.88 102.68 77.57 23 101.18 .94 25 119.68 21 85.85 120.18 63.68 110.29 20 77.94 110.44 93.63 77.29 70.07 101.68 110.07 24 110.27 70.27 63.44 22 93.63 77.26 63.88 102.44 93.26 19 70.57 93.35 85.35 85.27 63.

119.0 -10.0 Vertical displacement (mm) -12.0 (b) Present 0.2 0.0 0 0.4 0.0 Static (present) Static ( ABAQUS ) -8.0 0 0.0 ABAQUS -4.0 ABAQUS ) -6.0 -14.2 0.0 Vertical displacement (mm) -16.0 (a) 0.18 Table 4.4 .85 120.0 Present -2.8 1 Vehicle position.3 Comparison of the first 25 natural frequencies (Hz) for the problem defined by Figure 4.0 -20.9 with Km=∞ .0 -8.0 -12.0 ABAQUS Static (present & -4. x v / L 4.6 0. The number of ABAQUS elements includes only elements for the girder – 65 – 2.

6 0. x v / L 1500 (b) 500 -500 -1500 -2500 Present Bending moment (kNm) ABAQUS -3500 Static (present) Static ( ABAQUS ) -4500 0 0.4 0.0. (b) free pylon top – 66 – 1000 (a) 0 -1000 -2000 Present Bending moment (kNm) ABAQUS -3000 Static (present & ABAQUS ) -4000 0 0.8 1 Vehicle position.8 1 Vehicle position. x v / L Figure 4.2 0.11 Vertical displacement histories at the center of the bridge (x = 75 m) calculated for rough road surface and with Km=∞ : (a) fixed pylon top.4 0.6 .6 0.2 0.

0 0 0.4 0.6 0.8 1 Vehicle position.8 1 1.2 0.2 1.4 0.8 1 1.115 d 1.2 .4 For the entire bridge At point x=75 m At point x=150 m 1.12 Bending moment histories at the center of the bridge (x = 75 m) calculated for rough road surface and with Km=∞ : (a) fixed pylon top.8 1.2 Spring stiffness.0.125 (a) For the entire bridge At point x=75 m 1.075 0 0.0 (b) 1.6 0.085 1.2 0. K m (1012 Nm) 2.095 1. x v / L Figure 4.105 DAF 1. (b) free pylon top – 67 – 1.6 m DAF 1.

8 =30 m/s Fixed pylon top.13 The influence of the rotational spring stiffness.20 DAF 1. (b) DAF m – 68 – 1. rough road 1. Km.05 1. rough road 1. K m (1012 Nm) Figure 4.25 d 1.15 1.Spring stiffness.35 Fixed pylon top. smooth road v Free pylon top.30 v 1.00 35 50 65 80 95 110 125 140 155 170 185 200 215 Vehicle speed (km/h) 2. rough road .10 1. on the dynamic amplification factors for displacement and bending moment (rough road surface and fixed pylon top): (a) DAF d. smooth road =30 m/s Free pylon top.40 (a) Fixed pylon top. rough road 1.0 (b) Fixed pylon top.

0 Moving force model Static -4. (b) DAF m at point x =150 m – 69 – 2.6 m DAF 1.6 0.0 (a) 0.14 The vehicle speed influence on the dynamic amplification factors for displacement and bending moment calculated for Km=∞ : (a) DAF d for the entire bridge.0 35 50 65 80 95 110 125 140 155 170 185 200 215 Vehicle speed (km/h) Figure 4.1.0 0 0.2 0.0 Sprung mass model Moving mass model -2.0 -8.8 .0 -6.0 -14.4 1.2 1.0 Vertical displacement (mm) -12.4 0.0 -10.

x v / L Figure 4.0 Sprung mass model (b) Moving mass model Moving force model -3.0 -9.15 m Moving force ( t) / ( F 1.75 0 0.8 1 Vehicle position.0 -7.2 0.0 Static -5.0 Vertical displacement (mm) -11. x v / L 1.05 0.4 0.85 Normalized contact force.1 Vehicle position.0 -13.0 -1.6 .6 0.0 0 0.2 0. smooth road 1 1. 0.95 0. (b) road surface with no roughness – 70 – 1.4 0.15 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. rough road 2 m+ Sprung mass.25 )g Sprung mass.

when assuming a road surface with no roughness and a normal vehicle speed.531. a value of 5 corresponds to having a free end support and 7 to a fixed end support. the DAF m obtained was as high as 1. indicate fast convergence with increasing Km value.5e).14. for a speed of 108 km/h. in Figure 4. t) 80 x . The effect of local irregularities was investigated by simulating a 3 cm pot-hole located at about x=150 m.2.13. it can be seen that very high DAF values can be obtained already at normal vehicle speeds. It can also be concluded that. are obtained already for a Km value of 1. This Km value corresponds to a value of 6. the moving force model – 71 – is fully adequate for this example and there is no need for using complicated vehicle models. For the fixed end. fixed pylon top. The 2 DOF vehicle model was used with the same characteristics as for the last two examples. 4. The same DAF values. From Figure 4.8 1 Vehicle position. As mentioned earlier in section 3. 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. for the influence of the rotational springs on the DAF values.16 indicate that for normal vehicle speeds.43 for the first term for c 1 in equation (3. This pot-hole represents a local defect in the road surface or an initial joint defect.093 and DAF m =1.3.1. the obtained DAF values for displacement and bending moment. considering the entire bridge. rough road surface and Km ∞ = .1. The cross-sectional area of the cables and the corresponding cable spring stiffness were evaluated using the expressions given in section 2.17. v = 30 m/s y( x. For example.15 and 4. respectively.2·1012 Nm.16 Normalized contact force calculated for fixed pylon top and with Km=∞ The presented curves. 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. as if totally fixed end supports were assumed.745. are DAF d =1. Figure 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. x v / L Figure 4.0.

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. for the point at x =128 m DAF =1. Ig=10 m4. Ec =2. for the problem defined by Figure 4.17.17 Idealized fan-shaped cable-stayed bridge subjected to an idealized 2 DOF vehicle. The ABAQUS frequencies calculated neglecting cable mass ( γ c =0) are only presented in Table 4. and the influence of some parameters (vehicle speed. and for the point at .4 that the natural frequencies of this cable-stayed bridge are very close to one another.046 and DAF m =1. This indicates that the present method involves much less computation.056 and DAF m =1. It can be seen from Table 4. Mode shapes.17 and the vehicle characteristics given in Figure 4. qq=6·104 N/m. for the center point DAF d =1. and that excellent agreement is obtained especially when comparing with the ABAQUS values calculated neglecting cable mass. some time histories of vertical displacements and bending moments. See chapter 2 for the description of the symbols If nothing else is stated in connection with the presented figures in the following pages. 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. qg=17·104 N/m.0·1011 N/m2. pylon height and vehicle spring stiffness) on the dynamic amplification factors. the bridge characteristics given in Figure 4.4 for comparison and have not been used for dynamic response calculations. are: for the entire bridge DAF d =1.30·1011 N/m2. Eg=0. compared to the finite element method with direct time integration.074. Satisfactory results were obtained by discretizing the bridge girder into 369 elements in ABAQUS and using 738 increments. natural frequencies. and therefore is much more computationally efficient. and γ c=9·104 N/m3.179. The obtained DAF values. σ a =750·106 N/m2.4 were used when calculating the response. are shown in the following pages.118.

66 3 0.49 0.264. m Mode 1 (0.66 Hz) Mode 3 (0.84 .39 Hz) 0 32 64 96 128 160 192 224 256 288 320 352 384 416 448 480 512 544 576 608 640 Figure 4.65 0.50 0.50 Hz) Mode 2 (0.50 2 0.84 Hz) Mode 4 (1.18 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) γ γ c = 9·104 N/m3 c=0 1 0.m x =528 m DAF =1.07 Hz) Mode 5 (1.66 0.

91 2.36 11 5.08 6.37 1.32 4.20 12 6.15 5.31 2.91 9 3.07 5 1.04 .60 10 4.56 3.31 8 2.06 1.59 3.39 6 1.80 1.78 1.0.29 2.36 4.83 0.80 7 2.20 5.84 4 1.07 1.88 2.39 1.

14 16 7.25 8.75 10.61 6.91 17 8.66 14 6.54 20 10.73 6.31 18 8.71 19 9.45 9.37 21 10.13 7.65 6.29 10.69 8.90 7.09 13 6.6.74 15 7.52 9.68 6.83 .29 8.07 7.64 8.85 7.79 10.34 10.

This is probably due to the differences in the two models.20.19 and Figure 4.10 13. Figure 4.22.33 12. Still. It can be seen that high DAF values are obtained for low vehicle speeds. whereas in the present model there is always a spring7 supporting the girder beneath each vehicle position.16 25 13. from an engineering point of view.38 13. The effect of pylon height on the DAF is shown in Figure 4.4 Comparison of the first 25 natural frequencies (Hz) for the problem defined by Figure 4.19 and Figure 4. limitation of the speed of vehicles will not necessarily avoid damage effects on bridges.41 13. 0 -5 Present -10 . It is obvious that increasing pylon height results in lower DAF values.34 11. one can say that the ABAQUS and the present model results exhibit good matches.25 12.36 24 13. it is noted that the present solution is somewhat stiffer than the ABAQUS solution. In the ABAQUS model the cables are placed 16 meters apart.29 11.47 Table 4.22 11. Thus for this case. The number of ABAQUS elements includes only elements for the girder – 74 – Studying Figure 4.17.20 are those giving the maximum response.21 shows the variation of the DAF with the speed of the vehicle for three different girder moment of inertia.38 23 12.06 13. because the important values in Figure 4.

8 1 Vehicle position.6 0.4 0.6 0. x v / L Figure 4.ABAQUS -15 Static (present & ABAQUS ) Vertical displacement (mm) -20 -25 0 0.4 0. x v / L 4000 (b) Present . – 75 – 2000 (a) 1000 0 -1000 -2000 Present -3000 ABAQUS Bending moment (kNm) -4000 Static (present & ABAQUS ) -5000 -6000 0 0.19 Vertical displacement histories at the center of the bridge (x=328 m) 7 Used to idealize the cables.8 1 Vehicle position.2 0.2 0.

4 0.3000 ABAQUS 2000 Static (present & ABAQUS ) 1000 0 Bending moment (kNm) -1000 -2000 0 0. (b) at the first pylon (x=128 m) – 76 – 1.12 (a) Ig I= g 5m =5 4 m4 1.20 Bending moment histories: (a) at the center of the bridge (x=328 m).2 0.6 0.10 30= Ig I =30 m4 v g=30 m4 1.11 /s m Ig I =10 m4 g=10 m4 1.8 1 Vehicle position. x v / L Figure 4.09 .

05 1.20 1.00 50 70 90 110 130 150 170 .50 m 1.04 1.d 1.30 1.80 (b) 1.07 1.60 =30 m/s Ig I= g 30 m4 v =30 m4 1.70 Ig I =5 m4 g=5 m4 Ig I= g 10 m4 =10 m4 1.10 1.40 DAF 1.03 50 70 90 110 130 150 170 Vehicle speed (km/h) 1.08 DAF 1.06 1.

045 DAF 1.30 At point x=128 m ( At point x=328 m 1.21 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.15 DAF .040 m 80= H 1.035 1.Vehicle speed (km/h) Figure 4. H (m) 1.060 (a) For the entire bridge 1.050 d 1.055 At point x=328 m 1. (b) DAF m at point x =128 m – 77 – 1.20 H m 1.030 50 60 70 80 90 100 110 120 130 140 150 Pylon height.25 At point x=528 m (b) For the entire bridge =80 m 1.

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
For the entire bridge
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
For the entire bridge
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

M. 194.Stayed Bridges under moving loads’.Part 1 Theoretical Study’. cable modes of vibration. 1.Part 2 Experimental Study’. ‘Dynamic Interaction of Moving Vehicles .. Petrangeli M. [7] Blejwas T. 117. . No. Rhode Island. torsional modes of vibration. Vol.. Techn.. – 86 – ______________________________________________________________________ Bibliography of Part A ______________________________________________________________________ [1] ABAQUS User’s Manual. Serie R. 1995. Serie R.. No. 11. Eng.. 67-82. No. and bridge damping can then easily be considered in the analysis. No. ‘Dynamic Response of Bridges due to Traffic Loading . ‘ 3-D Nonlinear Seismic Behaviour of CableStayed Bridges’. Karlsson & Sorensen.. Feng C. Brancaleoni F. to further enhance the understanding of the behavior of this type of bridge structures. geometric nonlinearities.. Inc.C.S. pp. Lyngby. To study the above listed theoretical topics. the author recommends using the finite element method. J. .S. Costruzioni Metalliche.. Providence. .. 1984. Dept. of Struct. 1984. Weber B. 31-36. of Struct. [2] Abdel-Ghaffar A. ASCE. pp. Dept. 2. Eng. 5. 3456-3476. Nov.. Struct.P. Univ. No.... Vol.. ‘Tuned Vibration Absorbers for “Lively” Structures’. Eng. Hibbitt. of Denmark. Nazmy A.E. Axial forces. This is necessary if more realistic three-dimensional bridge and – 85 – vehicle models are to be used for analysis. [4] Arpe R. Eng. (in Danish) [5] Arpe R. (in Danish) [6] Bachmann H. 195. ‘On the Dynamic Response of Cable. [3] Alessandrini L. or on a laboratory model. and to verify theoretical models. Ayre R. Lyngby. Int. pp. of Denmark. 1984. 1991. Oct. 1994.• 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. Univ. Techn. ‘Dynamic Response of Bridges due to Traffic Loading . . Struct.

Surana C. pp.681-703. .. 67(4). ASCE. 23-28. 1987.. [13] Chatterjee P. ‘Dynamic Analysis of Arch Bridges Under Travelling Loads’. . 1995. [12] Cai Y. .K. Datta T. 4.L. 1994. pp. ‘A Numerical Method to Define the Dynamic Behaviour of a Train Running on a Deformable Structure’. Sept.. 1995. Struct. J. J. of Sound and Vibration. Vol. Solids Struct.. Struct. 1979. Eng.L. 112. 1988. ‘Vibration Analysis of Cable-Stayed Bridges’. 1979. pp. J. pp. pp. No. IABSE Symposium.P. 23.H. 1.. 32. 1988. 313-320. Conference on CableStayed Bridges. 625-646. J. . Nov. [19] Diana G. Colotti V. . 452-463. J. – 87 – 87 [10] Bryja D. ‘Impact in Railway Prestressed Concrete Bridges’. pp.. Said M. Int. 1989. Dhar C. No..1585-1594. ‘Vibration of Cable-Stayed Bridges under Moving Vehicles’. Chen S.H...V. May 1986. [14] Chatterjee P. 3. Shock and . . Datta T. . Division.‘Vehicle/Guideway Interaction for High Speed Vehicles on a Flexible Guideway’. of Sound and Vibration. pp. [9] Bruno D. 1994. ‘Random Vibration of a Suspension Bridge due to Highway Traffic’. Vol. ASCE.. Int. Struct.L.. 5. Garg V. Int. Int. 11. pp. Bangkok.. Surana C. Struct. 513-521. Garg V. Meccanica. Vol. Villatico C.K. Rote D. Wang T. 130(2). pp. 1994. Eng.T..S. ‘Dynamic Load Testing of Swiss Bridges’. ‘Vibration of Suspension Bridges under Vehicular Movement’. 379-387.. 2. J. J... ‘The Influence of Random Surface Irregularities on the Dynamic Response of Bridges Under Suspended Moving Loads’.M. .K. March 1994.. Vol. No. pp. ‘Railway-Bridge Impact: Simplified Train and Bridge Model’. ST9.. pp. 1123-1128.. 116-121.S .S . Proc.1036-1051. Van Hoove J. Petrangeli M. ‘Finite Element Analysis of Structures Under Moving Loads’. San Francisco. J.K. ...P. of Sound and Vibration. of Sound and Vibration. Cheli F. Vol. [17] Chu K. Datta T.. [16] Chu K. Coffey H. Struct..K.and Structures’. 105. Vol. ‘Train Structure Interaction in Railway Cable Stayed Bridge Dynamics’. No... 125(2). . pp.K. No. [11] Burdet O. [20] Filho F. 120.K. Eng. [15] Chatterjee P.. [8] Brancaleoni F. Sniady P. 27-42.K. [18] Coussy O. 175(5). ASCE. No. Eng.1823-1844... pp. Corthay S.

. Song J. Pasco. pp. Feb.. Vibration of Solids and Structures under Moving Loads.. Vol. 641-658. 27-35. No. Serie R. Earthquake Eng. Dec. . Res. of Denmark.. No. Eng.. . Vol. Cable-Stayed Bridges. [25] Garg V. 1984. Groningen.. Mechanics Division. Academic Press. ‘Dynamic Response of Cable-Stayed Bridge Structures’. [23] Frýba L. [30] Green M. Eng. 1997. . 1978. [21] Fleming J. Eng. Ting E.K.H.. 38(1). London. [32] Hayashikawa T.C. Vol. 355-370. . 2. 121. 1995.K.D.A. pp.. of Struct. 1976. ‘A Complete Formulation of Inertial Effects in the Guideway-Vehicle Interaction Problem’. . Cebon D. 146168... 1980. Egeseli E. Ginsberg J. Chichester. Dynamics. 8. pp. and Struct. Toronto.J. pp..A. Traill-Nash R. 272-282. EM4. Eng. of Sound and Vibration. Techn. Dept.. . Aug. No. [24] Frýba L. Watanabe N. Dukkipati R. 1977. Feb. [26] Genin J. [27] Gimsing N.. 1996. ‘ Effects of Vehicle Suspension Design on Dynamics of Highway Bridges’.. 32. J. Lockwood J. [28] Gimsing N. 10. Symposium Cable-Stayed Bridges. J. – 88 – 88 [22] Fleming J..V. pp. Vol.. No.. Struct. ASCE. Dynamics of Railway Bridges.J. 106. Wiley. Egeseli E.W. [29] Ginsberg J. No.C.H. Appl. Sci..F. ‘ Dynamic Behavior of Continuous Beams with Moving Loads’. Vol. a compilation of papers presented at Eng. Cable Supported Bridges. 247.F. ‘Parametric Study of the Interaction of Bridges and Moving Vehicles’. pp. 8. 1975. pp.. EM1. ‘ Vehicle Braking on Highway Bridges’. J. [31] Gupta R. 107. Noordhoff International Publishing. 1980.. second edition.F. ‘Analysis of Erection Procedures for CableStayed Bridges’. 1989. 1972. ASCE. Lyngby. Thomas Telford. Washington. J.J. ASCE.. 15-26. Mechanics Division.Vibration Digest. Ting E. 1-16. Dynamics of Railway Vehicle Systems. Genin J. Univ. J. Cole D. . ‘Dynamic Behaviour of a Cable-Stayed Bridge’. .

– 90 – 90 [44] Inbanathan M. 1967. Royal Institute of Technology. ASCE.. ‘ Dynamic Factor of Highway Steel Girder Bridges’. ‘A Finite Element Method Prediction of the Vibration of a Bridge Subjected to a Moving Vehicle Load’. Sep. pp. . [43] Hwang E. 57-75.H. pp.. No. Wang T. of Struct. J. Vol. Jan. J. Vol. 121.. Doctoral Thesis. Vol. ‘ Dynamic Response Analysis of Slab-Type Bridges’. [38] Honda H.. Dec. 9. 3427-3443. Wieland M. 1051. 1..L. ‘ Simulation of Dynamic Load for Bridges’.. No. 45-53. ‘ Dynamic Influences of Smoothly Running Loads on Simply Supported Girders’. Struct. Kajikawa Y. J.. J.L. ASCE. EM6. 57. Stockholm.. Konishi K.. Sep.. ASCE. ‘ Spectra of Road Surface Roughness on Bridges’. 108. Kobori T.. [40] Huang D. Eng. Eng. pp. pp. 118. Division.... Struct. Eng.. of Sound and Vibration. Wang T. 5. J. Shahawy M. pp. 121. Computer & Structures. 108.. Ito M. pp..L. [41] Huang D.. pp.. Wang T.. Eng. ASCE. [36] Hirai A. 1951. 1995. 1956-1966. ‘ Impact Analysis of Continuous Multigirder Bridges due to Moving Vehicles’.J. 1982.. Periodica 2/1986. Yoshimura T. Vol. 229-246. No. Eng... 1. 48-62. of the Faculty of Eng. 117. pp. Shahawy M. . J. 1982. Vol. [35] Hino J. ST9.. Vol.. [37] Honda H. Nowak A.. pp. 1413-1434. 1330-1337. Shahawy M. ASCE. University of Tokyo (B). Struct. [34] Hillerborg A.. No. May 1991. No. 1995. No. 19942008. 1992. Watanabe N. Struct. Eng. Eng. Vol. XXIX. Struct. J. Yamada Y. 4.S. ‘ Response of Suspension Bridges to Moving Vehicles’. 1984. Dept. – 89 – 89 [33] Hayashikawa T. No. J. 1987. Ananthanarayana N. ‘ Dynamic Behavior of Horizontally Curved I-Girder Bridges’..1066.. 9. ‘ Vibration of Thin-Walled Box-Girder Bridges Excited by Vehicles’. Kashif A. Struct. 1995. J. [42] Humar J.L. ASCE.. No.1981. Vol. Sep. Mechanics Division. 703-714. ‘ Suspension Bridge Response to Moving Loads’. ‘ Bridge Vibrations due to Vehicle Moving Over Rough Surface’. 96(1). ASCE. Vol. IABSE Proceedings P-98/86... pp. 12. Kobori T.. Dec. 113. pp.. [39] Huang D.S. No...

(in Swedish) [48] Karoumi R. ‘ A Numerical Solution for Dynamic Interaction Between Rigid Wheel and Flexible Beam’.. 1994. ‘ Dynamic Response of a Multi-Span Beam on One-Sided Point Constraints Subjected to a Moving Load’. The MathWorks. ‘Mechanical vibration. [47] Karoumi R. 3-4. IABSE 15th. Congress.L. Master Thesis. 22..A. ‘ Vibrations of a Cable-Stayed Bridge due to Vehicle Moving over Rough Surface’. [49] Khalifa M. University of Southern California.. . ‘Dynamic Response of Cable-Stayed Bridges Subjected to Moving Vehicles’. version 4. 55. [46] ISO 8608:1995(E). 1995. [54] Menn C.H. . Vehicle System Dynamics. Boston.. Vol... Stockholm. Faculty of the Graduate School.. Denmark. Communications in Numerical Methods in Eng. March 1989. Natick. 1995. ‘ Analysis of Traffic Induced Vibration and Fatigue of CableStayed Bridges’. Inc. Birkhäuser Verlag. – 91 – . ‘Aerodynamic Stability. Dept. of Struct. 209-262. June 1996. 267-279. No. 1994.Road surface profiles. [51] Lee H. Doctoral Thesis.Reporting of measured data’. pp. c1990. 11. Mass. Vol.P.. Eng. pp. Vol. Prestressed Concrete Bridges. pp.. 1991. Wind Forces on Suspension and CableStayed Bridges’. Royal Institute of Technology. TRITA-BKN Report No. Asian Institute of Technology. [53] Matlab Reference Guide.2. 1993. 4. Bangkok. Computer & Structures. .. 615-623. Grassie S. 11. 87-92.[45] Indrawan B. [52] Lee K. No. [50] Knothe KL. pp. Aug. ‘Modelling of Railway Track and Vehicle/Track Interaction at High Frequencies’.

649-665. [56] Mulcahy N. Division of Struct. ‘Finite Element. and Struct. Sweden. 4th Int. of Technology. C. Conference on Speedup Technology for Railway and Maglev Vehicles. Div. Earthquake Eng.. . Mechanics. . ‘On the Fundamental Moving Load Problem’. of Solid Mechanics. O. Coupling of Moving and Stationary Dynamic Systems . 421-430. pp. of Sound and Vibration. Lund Institute of Technology.. 599-605. Sundara K.. Nov. [64] Olsson M. Doctoral Thesis. Sweden.91 [55] Miyazaki S. ‘ Train/Track Interaction. . Mechanics. [57] Nagaraju N. . Lund.. 1985. Los Angeles.Nonlinearities and Weakly Coupled Modes of Vibration’.. IABSE Publications 33-II. pp. Kanamori M. Lund. 1973. 1987. ‘Bridge Response with Tractor-Trailer Vehicle Loading’. ‘Modal Testing of the Tjörn Bridge’. ‘ Dynamic Behaviour of Cantilever Bridges Under Moving Loads’. Int. 1983.L. Sweden. pp. Sogabe M. Chalmers Univ. J. 299-307. pp. [61] Olsson M. pp.. [59] Ohlsson S. ‘Analytical Study on Dynamic Response and Riding Comfort of PC Multicable-Stayed Railway Bridge’. 1991. pp.Theoretical and Experimental Analysis of Railway Structures Considering Wheel and Track Imperfections’. Report TVSM-3004. 424-429. ‘Finite Element Analysis of Structures Subjected to Moving Loads’. Doctoral Thesis. Bangkok. [62] Olsson M. Conference on CableStayed Bridges. 99(1). 145(2). Modal Co-ordinate Analysis of Structures Subjected to Moving Loads’.T. ‘ Analysis of Structures Subjected to Moving Loads’. of Sound and Vibration. pp. Proc. [60] Ohlsson S. .. Feb. Dynamics. [58] Nielsen J. Division of Struct. 1993. [63] Olsson M. Wakui H. 149-171. Report TVSM-1003.S. . 1-12. 1993. Jagadish K. Vol. 1983. Matsumoto N. 1986. . J... Lund Institute of Technology. Yokohama. Int. 11. Göteborg.. ‘ Dynamic Characteristics of Cable-Stayed Bridges . Nov. 1986. Modal Analysis Conference.

BSP Professional Books. Tanabe M. . ASCE. Vol. Vol. 2. [68] Podolny W. 49.a Review of Analytical and Experimental Findings’. 749755. Vol. 1992. of Sound and Vibration. pp.. 5-6. Guidelines for the Design of Cable-Stayed Bridges. Aug.. pp. ‘Effects of Surface Irregularities Upon the Dynamic Response of Bridges Under Suspended Moving Loads’. 96.. New York.P. J. Structural Mechanics Archives. Issue 3.S. Bamberger Y. 1974. ‘ A Study on Dynamic Interaction . Dec. pp. 1983.C.. . ‘A General Algorithm for Moving Mass Problems’.C. 2. Computer & Structures. Elf H. 1983. Proulx J... 217-252.. Talbot M. 1993. [67] Paultre P. Proulx J. 1985. Ginsberg J. [75] Troitsky M. 362-376. 1986. J. [76] Ulstrup C.. Struct. Genin J. Chaallal O. Feb. Construction and Design of Cable-Stayed Bridges.. . pp. [71] Schneider H. of Civil Eng. Eng. . New York..C.. Shock and Vibration Digest.B. J. NY. 17.[65] Palamas J.. 260-278. 49-58. 2nd ed. No. [77] Veletsos A.A. No. Purdue University. Vol. ‘ Analysis of Dynamic Response of Highway Bridges’. Canadian J. . ASCE. – 92 – 92 [66] Paultre P..S. 5. 1995. pp. 1980. Eng.. Cable-Stayed Bridges... No.. 33(1). [69] Rasoul M. 5.. 3-9. [72] Ting E. Vol. 99(2). 121. Matsumoto N. EM5. ‘ Finite Element Analysis of Highway Bridges Subjected to Moving Loads’. ‘ Dynamic Analysis of Cable-Stayed Bridges’. pp. Scalzi J. No. 1981. 2nd ed. Reported by the ASCE Committee on Cable-Stayed Bridges. 837-842. London. 19 (2). ‘Dynamic Testing Procedures for Highway Bridges Using Traffic Loads’. 593-620... Computer & Structures. J.. ‘Bridge Dynamics and Dynamic Amplification Factors . pp. 1970..J. [74] Ting E.H. 1992. . Genin J. 15.. [70] Saadeghvaziri M. Wiley & Sons. Vol. . Kölle P... Yener M. NY.. Mechanics Division. pp. – 93 – 93 [78] Wakui H. Huang T. Jr. ‘ Dynamics of Bridge Structures’. Oct. No. Doctoral Thesis. [73] Ting E. ‘ Vehicle-Structure Interaction in Bridge Dynamics’. pp. Coussy O. 1988. ASCE.C. et al. 235-245. of Sound and Vibration.. ‘ Modeling of Travelling-Loads and TimeDependent Masses with Adina’.

1990.11. Huang D. ‘ Railway Bridge/Vehicle Interaction Studies with New Vehicle Model’. ‘ Vehicle-Bridge Interaction Analysis by Dynamic Condensation Method’. 118. [90] Yang Y. [87] Wang T. Isler W.Analysis for Railway Vehicles and Structures . Struct. 81.. 35. pp. of Transp. 8.F. – 94 – 94 [89] Wilson J.. 1994. Computer & Structures. ..L. Shahawy M.. 1995..IBAP... of Technology . Liao S.B. Struct. 1055-1067. ASCE. 6.B. 2222-2238. J. No. [82] Wang T. Vol. Timoshenko S.. ‘ Impact Formulas for Vehicles Moving Over Simple and Continuous Beams’. 49. J. Struct. Dec. Shahawy M. Lausanne.. Cable Stayed Bridges. 20992116.Mechanical Model and Practical Analysis Method’.. [86] Wang T. Sept.P. Huang D. Huang D. Computer & Structures. J. 116. QR of Railway Tech.L.H. 5. Thomas Telford. 1992. Eng.. Report No. [79] Walther R. 1644-1650. 121.K.H. No. Struct. ASCE. Vibration Problems in Engineering. ASCE.. Vol.. ‘Responses of Continuous Elastically Supported Beam Guideways to Transit Loads’. [80] Walther R..L. Eng. pp. 247-254. [81] Wang T. 6. London. and Control. Vol.L. Barbas S. 1988. Vol. Florida Dept.L. J.. Moïa P... Aug.... ASCE.. pp.T. . [83] Wang T. 1993. No. Nov.. 11. .. Vol. Vol. 1648-1659. 118. 1045-1054. Huang D.. ‘ Impact in a Railway Truss Bridge’. pp. No. 1992..L... pp. ‘ Ramp/Bridge Interface of Railway Prestressed Concrete Bridge’. No. Wiley & Sons. Tallahasee. 11. Vol.. Lin B. 7. J. FL/DOT/RMC/0542-3394. 121. Dynamic System. 1988. Lin B. pp. ASCE. Eng.. Eng. Young D.. ‘ Cable Stayed Bridges with Slender Deck’. Garg V. Test Report No. 2. 117. No. pp. 49. Nov. . 102. ‘ Computer Modeling Analysis in Bridge Evaluation’. [88] Weaver W. ASCE. 5th edition. Vol. May 1992. [85] Wang T. July 1991. pp. J.1980. ‘ Dynamic Response of Multigirder Bridges’.S. Eng. pp.03.Z. Chu K. Struct.H.. Measurement. June 1990. No. Vol.. No.L. [84] Wang T. 1993. May. Research Inst. Tokyo.H.. Vol.. New York. ‘ Cable-Stayed Bridge Vibration due to Road Surface Roughness’. 6.‘ Dynamic Response of Highway Trucks due to Road Surface Roughness’. Struct. Swiss Federal Inst. [91] Yang Y. Houriet B. Florida. Eng. J. 13541374. No. 1995. 96-104.

pp. However. Various studies of the dynamic response due to moving vehicles have been conducted on ordinary bridges.. Computer & Structures. Assoc. 179-195. [93] Yoshida D. little is known about their dynamic behavior under the action of moving traffic loads. Bridge Struct. [92] Yener M. 31. 1. see the state-of-the-art review in Part A of this thesis. ‘Numerical Method of Lines for Analysis of Vehicle-Bridge Dynamic Interaction’. Eng. Weaver W. Chompooming K. 1994. a detailed structural analysis is required. ‘Finite-Element Analysis of Beams and Plates with Moving Loads’.1 General Although several long span cable-stayed and suspension bridges are being build or proposed for future bridges..pp... – 95 – 95 – 96 – 96 Part B Refined Analysis Utilizing the Nonlinear Finite Element Method – 97 – – 98 – Chapter ______________________________________________________________________ Introduction ______________________________________________________________________ 6.. Publ. Thus. 709-726. Vol. 3. Cable supported bridges are complex structures consisting of various structural components with different properties. pp. 1971. Vol. 53. 1636-1643. No. Int.M. to take account of the complex structural response and to more realistically predict their response due to traffic loading. due to . No.

is presented. as the ratio between traffic load and dead load can be larger for these modern 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. which includes geometrically nonlinear effects and is derived using a consistent mass formulation. a two-node cable element derived using “exact” analytical expressions for the elastic catenary. is adopted for modeling the cables. Moreover. At this stage the bridge structure can readily be modified. Vibration effects should play a much more dominating role for the design and not be underestimated. it is not always easy to find a location with significant relative movements and enough space to accommodate these devices.reasons stated above. the response is evaluated using the mode superposition technique utilizing the deformed dead load tangent stiffness matrix. this is not always the most effective and economic solution. In the first approach. rather than having to make costly modifications later on. the nonlinear finite element method is utilized considering all sources of geometric nonlinearity. The second . Two approaches for evaluating the dynamic response are adopted. 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. This is a linear dynamic procedure based on results from a nonlinear static analysis. Such dampers can be found today on many existing cable supported bridges [33. In addition. vibrations induced by traffic can be a serious problem. is adopted for modeling the girder and the pylons. dynamic forces of heavy vehicles can lead to bridge deterioration and eventually increasing maintenance costs and decreasing service life of the bridge structure. lightweight and high flexibility of modern long span cablestayed and suspension bridges. 58]. they cannot be directly applied to modern cable supported bridges. one possible option is to incorporate external dampers1 into the system. This approach has won considerable popularity in spite of its limitations. Due to the low damping. For cable supported bridges and in particular long span cable-stayed bridges. Costly repairs and modifications have been undertaken on relatively new suspension and cable-stayed bridges. as one usually only need to consider the first dominant modes of vibration to obtain sufficiently accurate results. The application of such devices for bridge structures is discussed later on in Chapter 8. because the possibility of fatigue caused by traffic-induced vibrations had not been sufficiently investigated at the design stage. 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. However. 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. Whereas. Since the cable supported bridge structures to be analyzed in this part of the thesis are flexible and can undergo large displacements. A beam element.

very few [2. conventional linear dead load analysis which assumes small displacements is often not applicable [49. a refined nonlinear dead load analysis procedure should be adopted. 22. trial-and-error search procedures have been proposed for the nonlinear computer analysis of simple cable problems. Still. 40. 39]. 55] studied the nonlinear behavior of cable supported bridges. the so-called NewtonNewmark algorithm is adopted. the trend today is to use more shallow and slender stiffening girders combined with increasing span lengths. 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. For cable roof structures. 13. 7] tackled the problem of using cable elements for modeling the cables. i. 32. as well as a more recent history in [39]. the idea of replacing each cable by a bar element with equivalent cable stiffness or by several – 101 – . 7. In [31. Methods of static and dynamic analysis and the behavior of cable structures are thoroughly presented in [20. A tuned mass damper (TMD) is a vibration absorber tuned to a particular mode of the bridge and consists of a mass. 55].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. a viscous damper and a linear spring. Cable structures exhibit geometrically nonlinear behavior. Although several investigators [2. 6. 34. Here. 6. 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. the natural frequencies will vary with the amplitude of response and linear dynamic analysis will consequently be inadequate. except in special cases in the feasibility design stage. due to this inherently nonlinear behavior. 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. analyses methods and several very illustrative design details. – 100 – behave linearly during the application of traffic loads. 12. For that reason. suspension and cable-stayed bridges.e. If this is the case. As an example. As the cable represents a flexible member with virtually no resistance to applied moments. 41.approach evaluates the nonlinear dynamic response using a direct time integration method combined with a nonlinear solution procedure. In the final design stages however. This is a much more CPU time-consuming approach. For cable supported bridges. 49]. A brief early history of the research on the behavior of cables has been published in [34]. details of connections. e.g. they are very flexible and undergo large displacements before attaining their equilibrium configuration. are presented in [14]. 27.

57].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.1. In [25]. Samples of these Maple procedures are given in Appendix A. The presented element can be used for modeling large sag cables such as suspension bridge main cables. is similar to that described in [2. The cable element used in this thesis is derived using the “exact” analytical expressions for the elastic catenary given in [35]. 34]. at least to the author’s knowledge. using cubic polynomial interpolation functions. In [60]. another 2-node curved finite element was developed using Lagrangian functions for the interpolation of element geometry. etc. – 102 – 6. 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]. and used for the static and dynamic analysis of 3-D prestressed cable nets. However. can be found in the literature.2. Despite the fact that the cable modeling technique based on the expressions given in [35] has been available for many years it has. In [11.. where straight elements are not readily applicable. the analytical expressions for the elastic catenary adopted here are somewhat simpler and therefore easier to handle. This approach was later adopted by other investigators. 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 procedure presented later in section 7. cables in large cable roofs. . very seldom been used for analysis of cable supported bridges. cables in long span cablestayed bridges. a 2-node curved finite element was developed. non-uniform cross-sections. than those studied in Part A. nonlinear geometric effects. For this reason a more straightforward and general approach. was suggested in [56. and variable material properties. is developed to handle such analysis difficulties and allow a thorough study of the moving load problem of cable supported bridges. derivations of isoparametric cable elements which includes the element curvature are presented. An iterative analysis procedure for cables. to derive the element matrices. based on using “exact” analytical expressions for the elastic catenary. developed and used for the analysis of very simple cable structures [35.beam elements with negligible moment of inertia has found wide acceptance and has been adopted by many investigators and designers using commercial codes. In [34]. two-dimensional bridge models considering bridge damping. the search for more efficient methods has intensified and today various other cable modeling techniques. exact cable behavior. based on the nonlinear finite element method. and in [7] a four-node isoparametric cable element is presented and used for modeling cables in cable-stayed bridges. than the crude modeling with a bar element mentioned above. As the popularity of cable structures has increased. 63] and of power transmission lines [64]. 49].

are implemented using the MATLAB language [53]. and frequency analysis of cable supported bridges. The developed codes have been tested by comparing results against those obtained using the commercial finite element code ABAQUS [1]. in this study. have been included in the paper ‘Some Modeling Aspects in the Nonlinear Finite Element Analysis of Cable Supported Bridges’. This model is an improved version of the one used in Part A as it here includes both primary and secondary vehicle suspension systems. road surface roughness. Parts of the work presented in the following chapters. Since the main aim is not to develop design formulas for calculating the dynamic amplification factors. 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. number of vehicles on the bridge. 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. fully capable of handling the above stated important factors and parameters.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. and bridge damping • to investigate the influence of the bridge girder supporting condition on the response of the bridge. static dead load analysis. or finer bridge models (more elements for discretizing the bridge girder and pylons) have not been investigated in this study. This paper has been accepted for publication in the journal Computer & Structures. In addition. For the purpose of this study. The parameters that are believed to significantly influence the dynamic response and therefore considered in this study are: bridge-vehicle interaction. Since. concerning nonlinear finite element modeling. vehicle speed. computer codes. – 104 – Chapter . The effect of using more complex vehicle models. a very simple vehicle model is adopted. than the one mentioned above. – 103 – only hypothetical trains of moving traffic are simulated and used for analysis. 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.

1 General The cable supported structures considered in this study are cable-stayed and suspension bridges. often adopted when . For simplicity the present study focuses on two-dimensional problems. This is because the behavior of cables is inherently nonlinear and also because large displacements introduce nonlinearities in the geometric sense. change of cable geometry under different tension load levels (cable sag effect). there are mainly three approaches used today to consider the nonlinear behavior of the cables. 41]. Such bridges consist of cables. In the first approach each cable is replaced by one bar element with equivalent cable stiffness. Consequently.______________________________________________________________________ Nonlinear Finite Elements ______________________________________________________________________ 7.g. global mass matrix M. change of the bridge geometry due to large displacements. and earthquakes [4. 7. are considered in the present study. All sources of geometric nonlinearity. simplified two-dimensional bridge models are still very useful for bridge designers in the preliminary design stage. The matrices will be given in the element local coordinate system. 28]. The superscript e. As modern cable supported structures are flexible three-dimensional structures. an alternative approach is presented where accurate and efficient cable and beam elements are used for the modeling. The structure matrices (i. For cable supported bridges. i. These are included for the sake of clarity and also for the purpose of having a thesis as self-contained as possible. 61].2 Modeling of cables The problem of analyzing cables under different configurations and loading conditions is very complex.g. the element length Le or the element nodal displacement vector q e. accurate three-dimensional cable and beam elements can be found in [35. two-dimensional models are of course not adequate when studying the response of such structures under the action of environmental loads like wind.e.δ effect). torsional effects and torsional modes of vibration are disregarded. traffic. and axial force-bending moment interaction in the bridge deck and pylons (P. For each individual element in the model. used later to denote e. This approach. In the following. the evaluated element – 105 – matrices in the local coordinate system are transformed to global coordinate system by the usual coordinate transformation technique [11]. The formulation of cable and beam element matrices will be described in the following sections. pylons and girders (bridge decks) and are usually modeled using beam and bar elements for the analysis of the global structural response [28. for investigating the feasibility of alternative structural solutions. For the interested reader. However. the global tangent stiffness matrix Kt . 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]. e. is omitted in this chapter for notational convenience.e.

those elements are stiffer and require numerical integration to formulate the element stiffness matrix [49]. 49]. In [7. at those nodes. making this method very attractive for static response calculations. equation (2. especially in the feasibility design stage. for some cases. are presented. 68] and also in Part A of this thesis. – 106 – The third approach to model cables is to use isoparametric elements. and the simplicity of including the effect of pretension of the cable by simply giving the unstressed cable length. and its tangent stiffness matrix from a given unstressed cable length and given positions of the ends of the cable. long span cable-stayed bridges built today or proposed for future bridges are very flexible. 13. which include the element curvature.modeling cables in cable-stayed bridges. The second approach is to divide each cable into several straight elements. 22. the author still believes that this approach is more efficient to adopt. Derivation of the equivalent tangent modulus of elasticity for the bar element. This is mainly because fewer internal nodes need to be defined for each cable in the model. In contrast to other modeling techniques mentioned above. The alternative approach presented in this thesis is based on “exact” analytical expressions for the elastic catenary. 42. derivations of isoparametric cable elements. However. . 7. for short span cable-stayed bridges. they undergo large displacements. the cable element internal force vector. This would introduce many added degrees of freedom with a consequent increase in computer storage requirements and computational cost. 12. 41].g. the exact treatment of cable weight as it is included in the equations used for element formulation. In addition.2. is referred to as the equivalent modulus approach and has been used by several investigators [6. linear analysis utilizing the equivalent modulus approach is often sufficient [27. each cable may be represented by a single 2-node finite element. Even if each cable must be divided into several catenary cable elements. Still. are not fulfilled. is presented in [27]. the exact treatment of cable sag. e. 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]. numerical problems can occur and spurious results can be obtained if equilibrium conditions. as in [3]. in order to adequately model the curved geometry of the cable.10). 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. 11. The main advantages of the proposed cable element are the reduction of degrees of freedom. which accurately consider the curved geometry of the cable. to include cable modes of vibration in the dynamic analysis or external loads acting between cable ends. Whereas. the simplicity of finding the dead load geometry of the cable system.1 Cable element formulation The procedure presented in this section determines the complete geometry of the cable. and should therefore be analyzed taking into account all sources of geometric nonlinearity. 55.

Consider an elastic cable element.P2 u x 1. modulus of elasticity E.1. with an unstressed length u L .E. cross section area A.w u2.P4 u3. the “exact” relations between the element projections and cable force components at the ends of the element are [35]: – 107 – u4.P1 node i Lx Figure 7. and weight per unit length w (uniformly distributed along the unstressed length). For the elastic catenary. stretched in the vertical plane as shown in Figure 7.A.P3 node j y Ly Lu.1 Catenary cable element L 1 P+T u 4 j L=− .

For the above expressions it is assumed that the cable is perfectly flexible and Hooke’s law is applicable to the cable material. . y L=y L(1 P. P . P .2 P ).2) 2 EAw w where i T and T j are the cable tension forces at the two nodes of the element. The expressions for x L and y L in equations (7. in terms of the end forces 1 P and 2 P only.2) may be written.1) and (7.1)  EA w T− i 2 P 1 − L= y (T2−T2 j i) T T j i + (7.x 1 P  + ln  (7.3) because P . as: x L=x L(1 P.2 P ) (7. P .

2 Tj= 3 P+4 P (7.3 P=−1 P (7.4) – 108 – 2 2 T 2 i= 1 P+2 P.3) and rewriting the results using matrix notation gives: L ∂ L ∂ L ∂ L ∂ x x d y .1 2 3 4 i T T and j are related by the following equations: 4 P= w u L−2 P .5) Differentiating equation (7.

dL= dP+ d P (7.7) d L L L .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.y x L= d1 P+ d2 P.

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 . The stiffness matrix K is given by the inverse of F as: 1 1 k k 2 K=− F=   (7.y ∂ ∂ y y  d 2 P d 2 P ∂P ∂P 1 2 where F is the flexibility matrix.

may be used to evaluate the matrices K.9)  −  k −k P  1 2 3  sym. u ∆ }T 4 . However. u ∆ 2. and p:  1 L 1P P   k=− u +  4 + 2  .∆2 P. The Maple software package for symbolic computations [52] was used to perform the above mentioned operations and produce the necessary Fortran code. Kt . −k P 4 4 The element tangent stiffness matrix Kt relates the incremental element nodal force vector { ∆ 1 P.2 4 t .∆ u 3. if this package is not available the following expressions.∆3 P.3).P ∆ }T 4 to the incremental element nodal displacement vector { ∆ 1 u. This Maple procedure is listed in Appendix A. obtained by derivation of equation (7.p= (7.

10b) det F  w  T T  j i   1 L 1P P   k= x +  4 + 2  4 (7.1 (7.10c) det F    P wT T 1 j i  2 .10a) det F    EA wT T  j i  – 109 –   1 1 1 k=k=− P 1  −  2 3 (7.

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 .

P= 2 − L + y u L  (7. 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 λ x L w cosh  1 P=− . from given positions of cable end nodes.11) λ 2 2 sinh λ  where 2 L−2 L  u y  λ=3 −  . the end forces 1 P and 2 P must be determined first. Those forces are adopted as the redundant forces and are determined.1   j i   j i  To evaluate the tangent stiffness matrix Kt . using an iterative stiffness procedure.

a conservative value of 0.2) and the misclosure vector {∆ L x .5).1) and (7. Using equations (7. In that case an arbitrary large value of 106 for λ is used. 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 .13) ∆ 2 P ∆ y L 2 P 2 P  ∆ 2 .12) for vertical cables. ∆ L y}T is evaluated as the positions of the end nodes are given. =+  (7. Another difficulty arises in equation (7.(7.12) cannot be used because the unstressed cable length is less than the chord length.12) 2  1  x L  In cases where equation (7.4) and (7. new cable projections corresponding to the assumed end forces 1 P and 2 P are now determined directly from equations (7.2 for λ is assumed [35].

coordinates for a number of points along the cable must be computed. As will be demonstrated later. equal to the cable chord length.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. a similar iteration procedure can be adopted. this iterative procedure converges very rapidly.P – 110 – where the stiffness matrix K is given in equation (7. u L . For the dynamic analysis.1) and (7.5). so equations (7.8) and i is the iteration number. mass discretization is simply done by static lumping of the element mass at both ends giving the following lumped mass matrix (ρ is the mass density of the cable element): 1 0 0  0   ρAL 0100 M= u  (7. Using equation (7. If the complete geometry of the cable is to be determined. This is very simple because 1 P and 2 P are now known. e. this iteration process continued until ∆ L x and ∆ L y are less than 5 1 10− ⋅ . for cases where the initial cable tension is known instead.g. This is then compared with the given initial tension to obtain a better approximation for u L for the next iteration step. A starting value for the unstressed cable length is assumed.14) 2 0 0 1  0   . cable tension can now be computed. For the present study. To determine the unstressed cable length. and cable end forces 1 P and 2 P are computed using the iterative procedure described above.

3. a horizontal force of To = 1.1.5 20 beam elements.0 span cable sag of 30. To To 304. This problem was earlier studied in [7] using isoparametric cable elements and the published results can now be compared to results obtained here.8 m Figure 7.0 the longitudinal displacement along 0 .2 Analytical verification A cable hanging under its own weight and subjected to a tensile force at both ends along its chord.2 Cable under its own weight subjected to tensile force at both ends To span the distance of 304.11 N/m. using a cable with the above given 1.7794·104 N was needed at both ends. =10-4 Using the two models.2. For the two models: the cable was replaced by one catenary cable element.31·1011 N/m2. modulus of elasticity E = 1.2. The beam element used is described in the next section.7 m. This force gave a mid1. I =10-5 0.48 m and was 1 catenary cable element adopted as the initial force when 20 beam elements. A cable with an unstressed length u L = 312. the sag and 0. I calculating the curves in Figure 7. – 111 – was studied using two different models. and by twenty beam elements with negligible moment of inertia. and weight per unit length w = 46. cross section area A = 5. as shown in Figure 7.0 0 0 1 7.2. was studied to verify the cable element and the analysis procedure described in section 7.8 m.5 (a) properties.48·10-4 m2.

.3. I =10-5 ment is observed when comparing 20 beam elements. Good agree20 beam elements.2 – 112 – beam elements increase and the moment of inertia decreases. As the number of Figure 7.0 parametric cable element presented in 2.0 several beam elements with negligi0 5 10 15 20 ble flexural stiffness can give Tension / Initial tension acceptable results. 6.0 1 catenary cable element plotted in Figure 7. the results should converge to those of the catenary cable element model.3 Response of the cable defined in Figure 7. Figure 7.0 [7].0 I =10-4 the curves for the adopted catenary cable element with those for the iso4.5 10 15 20 (Displacement / Horizontal length) x100 Tension / Initial tension the chord of the cable were deter10.3 shows also that the replacement of the curved cable by (Sag / Horizontal length) x100 0. 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.0 mined for different values of the (b) tensile force T and the results are 8.

4. This element is chosen also because it can handle large displacements and shear deformations and because it is simple to formulate the element matrices. have shown that the element is efficient and accurate. u( x). the simplest 2D beam element introduced in [61] is adopted and the treatment given there will be followed below when deriving the element matrices. Previous studies.15) where the abscissa x∈[0. w u6 u4 b( x) s a( x) o. b( x) = −sinθ i + cosθ j (7. x . 7. w( x) represent the axial and transversal displacement components and i and j are unit axis vectors. γ . For the present study. This finite element is developed following the total Lagrangian approach and using a linear interpolation scheme for the displacement components. L] is measured on the straight reference configuration of the beam. By introducing the angle θ ( x) as the rotation of the cross sections (S') in the deformed configuration.16) Further. 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. shear. defining the deformation measures ε. the unit vectors orthogonal and parallel to the cross sections for each point on the deformed beam are obtained as: a( x) = cosθ i + sinθ j. reported in [61]. Referring to Figure 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.κ according to: – 113 – u5 y.The results from this simple numerical experiment provide confidence in the application of the catenary cable element for modeling cables in cable supported bridges. and axial forces.

= = x (1+ ε)a +γ b .17) dx x x . κ = = θ (7. u x L Figure 7. .deformed beam u2 θ ( x) u3 u1 S' s s( x) w( x) o( x) u( x) j S i undeformed beam x.4 Deformed and undeformed configuration of the beam element ds dθ s o o.

u 2 .16).22) where T u={u . ux)sin θ (7. .wθ . ux)cos θ + w sinθ −1 . N ( x) = x L j / are the interpolation functions. For u.15) and (7. and N ( x) = 1 x / L i − . w and θ a linear interpolation scheme is used according to: u = Ni ( x q )i+Nj(xq ) j (7.d and using equations (7.18) γ = w cosθ . the strain energy can be written as: 1 L Π (u) = i ∫( 2 2 2 EAε + GAγ + EIκ )d x (7. GA and EI represent the axial. T q={ T 1 u . shear and flexural rigidities. the following expressions are obtained: ε = (1+ .x − (1+ .20) If the constitutive relations are assumed as linear.21) 20 – 114 – where EA.19) κ = θ. u 5 .x (7. }. u 6} contain the corresponding i j values of the displacements at the two nodes of the element. u 3} and q = { u 4 . x (7.

The Maple procedure which performs the above mentioned operations and produces the necessary Fortran code for p and Kt .23) 2 q ∂ q ∂ q ∂ where q is the nodal displacement vector { 1 u .4.u4.Finally.u2. The position of this point is defined by the vector s( x) as shown in Figure 7.u3. the final expression for the kinetic energy becomes [23]: ρL 2 ρL 2 ρL .24) V 2 where ρ is the mass density and s&( x) represents the velocity in a general point of the beam. is listed in Appendix A.5 u . 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 Π ∂i p ∂ ∂ Πi p= . Kt= = (7. The kinetic energy is expressed as the integral over the volume V : Π=1 k ∫ ρs&( x)Ts&( x) V d (7. u }T 6 . For this element.

From the resulting expression for the kinetic energy.26) . 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. the consistent element mass matrix is evaluated as: – 115 – 2 A 0 0 A 0 0   0 2A 0 0 A 0 ∂2Π ρL  0 0 2I 0 0 I M= k=   (7.Π= Au&( x) dx + A w&( x) dx + I θ&( x 2 ) dx k ∫ ∫ ∫ (7.25) 2 2 2 0 0 0 Using the interpolation functions.

in order to improve the earlier model adopted in Part A. an additional suspension system is introduced. 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. – 116 – Chapter ______________________________________________________________________ Vehicle and Structure Modeling ______________________________________________________________________ 8. the improved model includes not only the body-bounce motion. can be seen as filters. Consequently. the reader is referred to [23]. The two suspension systems.1 Vehicle models Vehicles have at least two suspension systems. as the one in Part A. but also the wheel-hop motion. each consisting of a spring and a damper element.∂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 first suspension system (primary suspension) reduces the road input into the vehicle structure and creates isolation for frequencies higher than about 15 Hz [9]. This system should also give sufficient damping to the axle and tire system to prevent the tire leaving the ground on very rough road . Thus.

it is assumed that the vehicle never loses contact with the bridge.1a. (b) typical modes of vibration 3 . Referring to Figure 8.1 (a) vehicle model on a bridge element. 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. However. Despite that. as was discussed in Part A. detailed suspension nonlinearities and complexities of vehicle body motion that are typical of heavy vehicles. 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. the simple vehicle model adopted does not contain all suspension systems.g. for the purpose of this study. For the current study. and denoting the contact force between the bridge and the vehicle by F( t). and the contact between the bridge and the moving vehicle is assumed to be a point contact. the spring and the viscous damper have linear characteristics.5-4 Hz 8-15 Hz Figure 8.surfaces. e. it is believed that this model is sufficiently realistic. Other suspension systems.

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 . A dot superscript denotes differentiation with respect to time.2 w and 3 w are the vertical displacements of the masses measured from the static equilibrium position. c and c the damping coefficients of the viscous dampers.1a) 2 m 2 w& + ks( 2 w−3 w ) + cs( &2 w − &3 w ) + kp( 2 w−1 w ) + cp( &2 w − w& ) = 0 1 (8. kp and ks the stiffness of the linear springs connecting the masses.m 3 w& + ks( 3 w−2 w ) + cs( &3 w − w& ) = 0 2 (8. Using the equations above. and g the p s acceleration of gravity.1c) Where 1 w.1b) −1 m 1 w& − k p ( 1 w−2 w ) − c p ( &1 w − w&2 )− ( 1 m + m 2 + m ) g + F( t) = 0 3 (8.

xx c.3b) w e &=N q v2 e + N q& v e 2+N q a e +N q& + r v2+ra 1 c. 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. is always set to zero in this study. The moving force and the moving mass vehicle models can now be obtained. x (8.w& (8. x c c. x c. by modifying the contact force expression.3a) w e &=Nq v e +N q& + r v 1 c. x c . 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. When modeling suspended roadway vehicles. as in Part A.2). 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. the unsprung mass. as shown in Part A.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. In this way this degree of freedom is eliminated and. m 1. – 118 – With the assumption that the road profile cannot be rough enough to make the vehicle jump or leave the road surface. equation (8.

1a. Equations (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 . As shown in Figure 8.c.1b) may be written using matrix notation as: 3 m 0  &3 wc −c s s  & 3 wk −k s s  3 w 0     +    +    =   (8. xc is the distance from left node of the element to the contact point.1a) and (8. x (8. 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.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. and rc the surface irregularity evaluated at the contact point. The subscript x denotes derivation with respect to x. xx c.

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. The undamped natural frequencies for this simple sprung mass vehicle model is determined from the eigenvalue equation: det( 2 K−ω . where I is a 2x2 identity matrix.5) In order to solve this second order equation of motion. This will give the extended dynamic equilibrium equation: – 119 – I 0 w&  0 − Iw  0     +    =   (8.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. 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.6) 0 M v w& K v C v w& f v  which can be rewritten as: w&  0 I w  0  = − or & = + (8. it is transformed into a system of first order equations by complementing it with the equality Iw& − Iw& = 0 .

)=0 v Mv (8.9) 2 + 1+ ±    + 1+  − m m m m m m mm 2 3 2 2 .8) giving the eigenvalues:  2  2 1  kp k m k s p k m  4kk 3 s 3 ps ω=− 1.2 (8.

the traffic part can be expressed as: a Le e f live = − ∑ i F ( t) a T N δ( x − x . as mentioned earlier in Part A of this thesis: body-bounce and pitch motions at 1. is time-dependent. m 2. the bodybounce mode for a vehicle traveling at 70 km/h is excited by irregularities with wavelengths of 4. se Figure 8. i.1b. – 120 – 8. Heavy roadway vehicles generate most of their dynamic wheel loads in two distinct frequency ranges.3-2. while the second mode (wheel-hop) almost only excites the wheel mass. Using.e.1b.4 m. m 3. as the external force vector.3 2    2 3  For typical vehicle mass and stiffness values. the first vibration mode (body-bounce) excites mainly the vehicle body mass. a surface irregularity of a certain wavelength may be effective in both frequency ranges. containing the interaction forces existing at the contact points between the vehicles and the bridge. as before (see Part A). e e e f = f dead + f live . se Figure 8. As an example. 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.9-13 m and the wheel-hop mode by irregularity wavelengths of 1. depending on the vehicle speed. Moreover. 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.5-4 Hz and wheel-hop motion at 8-15 Hz. 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.2 Vehicle load modeling and the moving load algorithm The moving load problem is more complicated than other problems in structural dynamics. the Dirac function δ( x − ci x ) to characterize the action of a unit force concentrated in point x = ci x .

illustrated in Figure 8.10) i 1 = 0 i 1 = where a is the total number of vehicles on the element. for each vehicle on the bridge. Le the element length (see Figure 8.e. and fully considers the dynamic interaction between the vehicles and the bridge.∫ ci ) d x = − ∑ i F ( t) T N ci (8. 0 . For consistency reasons. 0 . 0 1− . i. and N ci the row vector:  x x  N=. and prepares the global external force vector caused by the moving traffic.2. the same linear interpolation functions are adopted here as those used in section 7. Fi( t) the interaction force between the bridge and the i th vehicle wheel. – 121 – START . calculates the bridge-vehicle contact force. all having the same velocity (independent on the deflected shape of the bridge deck). The code developed is capable of handling unlimited number of vehicles.3 for deriving the beam element matrices. 0 (8. x is the distance measured from left node of the element.11)  e e L L  containing the element interpolation functions evaluated at the contact point of vehicle i. for x = ci x .1a). The moving load algorithm. .

the vectors N ci. w 1 & i. f =0 new vehicle positions.xx Loop through all vehicles on the bridge For vehicle i. N ci.Call algorithm for each iteration.x. w . N ci. live and surface roughness parameters rc. rc.xx and determine the loaded element’s dof No Interaction problem Yes Determine e e e q . rc. C vi and K vi Identify the loaded element. w&0 i . q& and calculate w i. M vi.x. traffic load vector. Input: initial vehicle vectors w Initialize global external 0 and w &0. initialize w0 i. Evaluate xci. q& .

3 Bridge structure The bridge structures are discretized. w& i .2 The moving load algorithm – 122 – 8. Furthermore. the torsional behavior caused by eccentric loading of the bridge deck is disregarded.1 & i 1 form equations 8.1 Modeling of damping in cable supported bridges . As discussed earlier. Fi( t).3 Solve differential equation 8. 8. Only twodimensional bridge models are considered in this study. this cable element is adopted in order to simplify cable modeling and to more accurately predict the response of cable supported bridge structures. w& i. for the nonlinear finite element analysis. the nonlinearity considered is of the geometric type. The modeling technique and derivation of the elements are described in detail in Chapter 7.3. Evaluate contact i = ( m + m + m 1 2 3) g force. as linear material behavior is assumed. consequently. 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 8.7 to Moving force model F calculate w i. shear and axial forces. using a catenary cable element and a simple beam element able to resist bending.

parallel wire strands have been favored in resent years due to economic and durability reasons. The different forces that contribute to the damping of a structure may vary with vibration amplitude. theoretical evaluation of damping is extremely difficult and hence at present we have to rely upon the empirical approach. loss of energy from foundation to ground. and stress intensity. 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. The most commonly employed damping model in the field of structural engineering is the viscous one. 37. one frequently used technique is to assume the damping matrix proportional to mass and stiffness matrices giving the so-called Rayleigh damping [8. – 123 – In the case of cable-stayed bridges. the damping ratio depends on the type of cables and on the cable configuration used. in which the damping force is proportional to the velocity. Using spiral or locked coil strands instead of parallel wire strands (PWS). structural damping such as friction at movable bearings and in the joints of the structure. In [19]. To derive the damping matrix. increases the total damping capacity of the bridge as such strand exhibits significant hysteresis in its longitudinal load-displacement behavior. C = α K + β M . For practical use however. is very popular as it leads to the simplest mathematical treatment and generally gives the most satisfactory results [15]. 43]. 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. 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. Energy dissipation is generally developed by material nonlinearity. As pointed out in [65]. For practical problems the parameters α and β are often chosen based on the knowledge of the damping property of a similar structure. 15]. also adopted for the present study. velocity. However. acceleration. A limited amount of information is available on damping of cable supported bridges. One disadvantage of the Rayleigh damping is the fact that damping behavior of the . This model. and aerodynamic damping by friction with air. the dissipation of energy in cables is potentially a major contribution to the overall damping of the bridge.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. opening and closing of hair cracks (in reinforced concrete structures). Attempts on theoretical approach to damping evaluation of cable-stayed bridges have been made in [43. fundamental data on the damping ratio is increasing drastically. It was shown that the proposed method predicted the dependency of the damping ratio on oscillation amplitude and cable type with reasonable accuracy. damping models that represent more or less satisfactory approximations have been introduced. as more and more forced-excitation and ambient-vibration tests are conducted. Despite this. There are various factors causing energy dissipation in bridge structures. Thus.

– 124 – 2 1 ξ1 ω 0 0   [  ξω  2ξω] 0 222 0 = = ZT CZ   (8. On the other hand.complete bridge structure is described only by the two parameters α and β .12)  O  0 0 2   ξ nω n Due to the earlier mentioned limitation of the Rayleigh damping. To . of equation (8. by premultiplying and postmultiplying with the mode shape matrix.1. the damping matrix derived gives a poor simulation of the real damping characteristics of the bridge structure. Thus. this method gives damping matrices that have the same orthogonal properties with respect to the eigenvectors of the undamped system. which are taken as constant values over the entire bridge model.12). as the mass and stiffness matrices. the damping matrix can be diagonalized giving the matrix of generalized (or modal) damping values3 (see also section 9.2): 3 In the present study the eigenvectors are normalized such that the generalized (or modal) mass is set to unity.1. An improvement may be achieved by ignoring the proportionality condition and instead establish directly the diagonalized matrix. [ ξω 2 ].

is that only the damping ratios of the required modes can be accounted for. Another advantage. In practice.15)  2π  . the i th mode damping ratio in equation (8. ωi ξ = 0.14) For the present study.derive a damping matrix with orthogonal properties. 0 − . the inversion of the mode shape matrix requires a large computational effort. Thus. by using the expression3 ZTM Z = I to express −T 1 Z and − Z in equation (8. this method is much more flexible than the Rayleigh damping and is preferred if modal damping ratios are available. as: −T C=Z [2ξω] 1− Z (8. This can be avoided by taking advantage of the orthogonality properties of the mode shapes relative to the mass matrix. one can obtain: C= M Z[ ] ZT 2ξω M (8. but this is of little significance.0042 i   (for suspension bridges [19]) (8. when compared to Rayleigh damping.13) The damping matrix obtained is full regardless of whether the stiffness and mass matrices are banded or not.13).12) is assumed as: −0 84 . The actual expression for the damping matrix may be deduced from the assumed matrix [ ξω 2 ]. 0 645 1 ξ = ξ2 = ξ L 237 .

is generated to simulate roughness of the local type. for x ≤x≤(x +L ) . The first type is a random profile described using a power spectral density function. For the second roughness profile.17) bump   2  bump L  where xbump is the position where the bump starts. a bump. The generation of such a profile is described in detail in Part A.16) where L is the main span length in meter. It should be noted that here x and xbump are measured starting from the left end support of the bridge. Chapter 2. The above two approximations are based on field forced-excitation tests. assumed to vary harmonically as shown in Figure 8. hbump the maximum height. – 125 – 8. as: bump bump bump 1  2π(x−x ) r( x) = h 1− bump cos  (8.3.2 Bridge deck surface roughness Two types of roughness profiles are considered in this study. .3. The profile is evaluated.= L n (for cable-stayed bridges [43]) (8. and Lbump the length of the bump.

– 126 – 8. The primary intention. as modern bridges become longer and more flexible and as energy dissipation in these bridges is very low (see section 8. Usually such a device consists of an additional mass connected by means of an elastic and a damping element to the structure needed protection. vibration absorbers are nowadays widely used mainly to reduce vibrations caused by wind or earthquakes. such devices are still not that common. it is realized that additional vibration control measures are needed.1) and often not enough on its own to suppress vibrations.3. 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. and high-rise buildings. some investigations have been carried out to find if such devices are also efficient for controlling traffic-induced vibrations [10. For civil engineering structures like chimneys. see for example [50. For bridges. 54. TV towers.h r( x) bump x ° xbump Lbump Figure 8. the developed algorithm temporary reduce the time step as a vehicle reaches the bump. In [45] the efficiency of a tuned mass damper . This is done automatically. 66]. There are innumerable examples of vibration absorbers being applied to control vibrations in various engineering structures. The invention of the vibration absorber is usually associated with the name of Frahm. As such bridges not only are vulnerable to vibrations caused by wind and earthquakes but also to those caused by traffic.4 Tuned vibration absorbers A vibration absorber is a device that reduces the vibration level of a protected structure. 33. 38. 45].3 Bump dimensions To obtain a realistic simulation of vehicles moving over a bump. was to control wind-induced vibrations. in the few examples found where vibration absorbers are practically applied on long span bridges. However. 18. who in 1909 first patented a vibration absorber design [21].

In the present study. In [33] the effect of a TMD on the dynamic response of the Rama IX cable-stayed bridge5 in Bangkok. the abbreviation TMD is used hereafter for such vibration absorbers. among other things. – 127 – . was studied. jumping and running tests. 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. subjected to a moving vehicle. the bridge-vehicle interaction was not fully considered in [33].6. the efficiency of a TMD is also investigated but this is done more correctly here as. For convenience.(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. 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].

.

the attached dashpot increases the overall damping of the structure by providing an additional source of energy dissipation. by using TMDs. the dynamic response due to wind could be reduced by approximately 35 % [16]. The TMDs were designed so their eigenfrequency could be adjusted from 0. in the two approach bridges for the Great Belt suspension bridge in Denmark.bridge structures. and (c) stockbridge dampers are mounted on the long hangers of the Humber suspension bridge in England [58]. Figure 8. More information can be found in [51]. 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. As construction progressed.2 Hz and the relative internal damping from 3 % to 18 %. For example. and secondly. two TMDs each weighting about 40 tonnes. each with a mass of 8 tonnes (which approximately equals 0.4 Design measures for suppression of cable vibrations [58] TMDs are also used temporarily during erection of long span bridges.5b.4 illustrate how cable vibrations in bridges can be controlled. As an example Figure 8. auxiliary ropes are found on the Farø cable-stayed bridge in Denmark and the Normandie (b) bridge in France. as free-standing pylons and cantilevered spans are very sensitive to wind and often need vibration control.5a.5 % of the modal mass) [59].1 Hz to 0. it reduces the resonance response of the main structure. Such design (a) measures for suppression of cable vibrations can be found on existing bridges. For example. 32 TMDs have been installed. these TMDs were relocated and tuned to maximum efficiency. There are two important roles of a TMD: firstly. The efficiency of the TMD depends on the correct tuning of its parameters (eigenfrequency) relative to – 128 – . 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. see Figure 8. Most recently. Wind tunnel tests and calculations showed that. see Figure 8.

.

.

.

often dictate the final choice of TMD parameters. ξ tmd the damping ratio of the TMD. However.18a) 1+ µ 6 These equations are used in the present study to determine the TMD parameters without considering available standard components and dimensions. Thus. are adopted6: ω ω = i tmd (8.18b) 8(+µ)3 1 where ω tmd ω and i are the circular frequencies of the TMD and the dominant bridge mode to be tuned to. These equations are derived for undamped structures affected . In the literature. µ = tmd m /i m .5 TMD of the (a) Normandie Bridge during erection [16]. TMDs can still be advantageous as they work without requiring any connection to the ground or external power supply. derived in [21] for a structure with no damping. On the other hand. a great deal of attention is given to the optimization of parameters and evaluation of the efficiency of the TMD. For the present study. – 129 – 3µ ξ = tmd (8. and µ is the mass ratio which relates the TMD mass to the modal mass of the dominant bridge mode to be tuned to. (b) Great Belt Bridge [59] the eigenfrequancy of the mode of the structure it is designed to suppress. the following most often used optimum tuning parameters. 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. in practice dimensions of available standard springs etc.(a) (b) Figure 8.

However. so that the damper mass can also respond to the bridge movement at low levels of excitation.6 Cross section of bridge girder with a tuned mass damper. ktmd ctmd wtmd mtmd Figure 8.by a stationary harmonic load. Another engineering problem is to provide a very low friction bearing surface for the TMD mass. 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 . But as expected. 45]. TMD – 130 – Now referring to Figure 8. Consequently.6. very important. using equations (8. perform a trail and error computer study to find the optimal TMD parameters. The mass ratio µ . Therefore. the response reduction was not as great as for the case with a stationary harmonic load. the damping needs only be assumed approximately as its efficiency is relatively less sensitive for variations in ξ tmd [10]. whereas. which is usually small and lies between 0. 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 µ lead to large relative movement of the TMD mass).01-0. 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.18a-b) as starting values. they may not be valid for a damped bridge structure subjected to moving traffic loads and one maybe should. in [38] it was found that the difference is small and for practical design purpose these optimization formulas are suitable. The exact tuning of the TMD frequency is. as mentioned earlier. as far as the bridge is lightly damped.04 for bridges [10.

q e the nodal displacement vector for the element on which the TMD is attached. wtmd the vertical displacement of the TMD mass measured from the static equilibrium position. Based on the element property matrices and the external nodal force vector derived in the foregoing chapters. equation (8. the general equation of motion for a bridge under the action of moving traffic loads is formulated.1 Dynamic Analysis In the previous chapter.11).20) It is interesting to note that when a vehicle is modeled as a sprung mass system instead of a constant moving force (i. for such cases the dynamic response. ctmd the damping coefficient of the viscous dampers. calculated by ignoring interaction. In the following.5). To include the force applied to this beam element by the TMD.− e tmd N tmdq )= 0 (8.19) where mtmd is the mass of the TMD. The spring stiffness and damping coefficient are evaluated using equations (8. – 131 – – 132 – Chapter ______________________________________________________________________ Response Analysis ______________________________________________________________________ 9.18a-b). ktmd the spring stiffness. is conservative. 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]. one may obtain the following general equation of motion for the entire . Thus. 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. and an algorithm for assembling the global external force vector are presented.e. and N tmd the interpolation vector given in equation (8.10) is made as: e e T f live = f live − tmd m tmd w& N tmd (8. an addition to the external force vector given in equation (8. the vehicle equation of motion.

Consequently. for the contact points. 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.1 Linear dynamic analysis If the structure is assumed to respond linearly during the application of traffic loads. q& . a new iteration is preformed recalculating the external force vector and resolving the bridge equation.1) q . as in Part A equation (3. q&. The type of convergence criterion used in Part A is also adopted here to check the errors in not only the displacement vector. q&. the moving traffic. In the present study.bridge by means of conventional FEM assemblage: M q& + Cq& + p(q) = f (q. 9. To consider the bridge-vehicle interaction. but also the velocity and acceleration vectors. Then the bridge equation of motion.14). Accordingly. of the contact points are assumed and the vehicle equations are solved to obtain the interaction forces and the external force vector. is solved to obtain improved values of displacements etc. the external force vector is not only time dependent but is also dependent on the bridge displacements. it is possible to evaluate the dynamic response using the mode superposition technique starting from the deformed dead load state. the external force vector is dependent on the motion of both the bridge structure and the vehicles. for each numerical example studied here. q. q & bridge node displacement.1) in the time domain: one for evaluating the linear dynamic response and one for the nonlinear dynamic response. an iterative – 133 – procedure is adopted. equation (9. First the displacements etc. t) external force vector resulting from the dead load.& t) (9. These are described in detail in the following sections. q&. and the tuned mass dampers As indicated. The time step and the number of considered eigenmodes (for the mode superposition procedure) are problem-dependent. velocities and accelerations. velocity. and acceleration vectors M bridge mass matrix C bridge damping matrix p(q) vector of internal elastic forces f( q.1.1). If the convergence criteria are not fulfilled. Thus. this linear dynamic procedure . 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. two approaches are adopted for solving equation (9.

The eigenmode extraction and the mode superposition technique are well described in many textbooks on structural dynamics [11].( 2 ω. which may be expressed as [11]: q = z sin(ω t −ψ ) (9.f. Thus equation (9. the undamped free vibration is considered. 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. this eigenvalue problem is solved using the Matlab algorithm eig [53] giving n (= number of d. we obtain the following eigenvalue problem: (K−2 ω M) z 0 t = (9.L. Furthermore.) eigensolutions ( 2 ω .1. and ψ is a phase angle. 9. as usually one only need to consider the first dominant modes of vibration.2) where Kt is obtained from a nonlinear static dead load analysis. as the effect of damping on the natural frequencies of most real structures is small.5) In the present study. z ).3). For this.1) is reduced to: M q& + K q 0 t = (9.4) where z is a vector with components independent of time. this approach requires frequency analysis and eigenmode extraction to start with.( 2 ω . The internal force vector in equation (9.1) is therefore evaluated as: pq ()=Kq t (9. 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. On the other hand.3) in which 0 is a zero vector. as the required number of eigenmodes for satisfactory convergence is difficult to estimate beforehand.o. the – 134 – effectiveness of mode superposition technique is reduced. it is well known that this approach give sufficiently accurate results with minimum consumption of CPU time.utilizes the dead load tangent stiffness matrix.1. Assuming harmonic motion. ω is a circular frequency.) 1 . Kt . and substituting into equation (9. Moreover. which can be expensive and time consuming for large systems. Nevertheless. z ).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.

e.5) as: K Z − M Z Ω2 0 t = (9. we can now write the n solutions to equation (9.: 2 ω  1  2 ω Z=[1 z . i. where z i is called the i th mode shape vector (eigenvector) and ω i is the corresponding circular frequency of vibration ( 2 ω i is the eigenvalue).7)    O   2 ω n  The obtained eigenvectors are then normalized such that the modal mass is set to unity. T =zMz=1 i .e.6) – 135 – where the matrix Z contains the eigenvectors z i in columns and the matrix 2 Ω is a diagonal matrix. which stores the eigenvalues on its diagonal. z n]   2 and Ω =  2  (9.L.1 2 2 n z n . Storing the obtained eigensolutions in two matrices Z and 2 Ω . z2. i.

1. The vector of nodal displacement. q. As noted.6) equation from and ZKZΩ t = (9.1. giving: s q ≈ ∑ z i φ i ( t) = Z s φ s (9. with s much less than n. one of the advantages of this normalization method is that the modal stiffnesses will be equal to the eigenvalues.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. can be approximated by a linear combination of s eigenvectors.9) i=1 where φ i in vector φ s are the modal amplitudes and state the proportion of each eigenvector in the transformation.1) and postmultiplication of each term by T Zs yields: ZT T T T & sMZ sφ & s+Z sCZ sφs+Z sKZ t s φ s = Z s f live (9. From the – 136 – .8) where I is identity or unit matrix.m i i . the equation of motion can be solved. From this M-normalization and the orthogonality properties of eigenvectors [11] it follows that: T T 2 ZMZ=I (9. Substitution of the above expression into equation (9. 9.2 Mode superposition technique When the natural frequencies and mode shapes are determined.

L.12) where c .L.L.ξ 1 2 s .L. To obtain a system of s uncoupled differential equations the matrix C must have the same orthogonal properties. c . With the introduction of this matrix we can obtain a system of s independent equations which can be written as: 2 φ& + 2ξ ω φ& +ω φ = zT f 1 111 . equation (9.8). 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.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. c 1 2 s are the modal damping values and are here expressed using the damping ratios7 ξ . the above equation is reduced to: I T 2 T φ& & s + Z s C Z s φ s + Ω φ s = Z s f live (9. c . such as the mass and stiffness matrices. When this is the case. as the mass and stiffness matrices.ξ .properties of M-orthonormalized eigenvectors.ξ 2 ω . c 1 2 s}= { diag ξ 2ω. ξ 2 11 22 sω s } (9. the damping matrix will be diagonalized by premultiplying and postmultiplying with the mode shape matrix giving: T Z s CZ s = { diag c . with respect to the eigenvectors of the undamped structure. The obtained matrix is referred to as the matrix of generalized (or modal) damping values.

i – 137 – T T φ= & o zMq i i o ........8) giving: 7 ξ is the fraction of critical damping in mode i.13) are obtained......... ... .. ........9) with zTΜ i and taking note of equation (9............ ... Initial values for equations (9..... by premultiplying both sides of equation (9. a program code for the exact solution of the above equations was generated...... . ......... from known initial values of q and q& ........ This code is incorporated in the developed computer program for linear dynamic analysis....... 2 T φ& & s + 2ξ sω s φ s +ω s φ s = z s f live Using the Maple software package for symbolic computations [52].... . ....... (9..13) ..1 1 1 live 2 φ& + 2ξ ω φ& +ω φ = zT f 2 222 2 2 2 live ....

9. thus theoretically such damping matrix ought to be updated for each time step. 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. frequencies and damping ratios will also change as the structure deforms. an implicit procedure based on Newmark’s modified average acceleration method combined with a full Newton-Raphson solution procedure is adopted for solving the bridge equation of motion.14) Finally. In the following.2 Nonlinear dynamic analysis A procedure for evaluating the nonlinear dynamic response using a direct time integration method combined with a nonlinear solution procedure. – 138 – Denoting the time step by ∆ t. 23]. CPU time-consuming) than the mode superposition method which is usually chosen for linear studies. but also the damping matrix. for the Newmark method the following expressions hold for the displacements and for the velocities: 21  q . For this procedure. however. because the damping ratios used will in most cases be only approximate values evaluated from field tests. A method for modeling the damping in matrix form in terms of modal damping ratios is presented in Chapter 8. the adopted procedure will be briefly described. If this is the case. this is usually not necessary.1. having determined the elements in vector φ s by solving the above s equations. the total response of the bridge is obtained. the natural frequencies and mode shapes will vary with the amplitude of response and the earlier described mode superposition technique will consequently be inadequate. is described in the following. In practice.e. traffic and TMD) is found from equation (9. can be found in [26]. taken from codes of practice or the literature. determined using nonlinear static analysis. As the stiffness is a function of the response amplitude. more details concerning the derivation etc.e. Among engineers and researchers in the field of structural engineering.and φo = z Mq i i & o (9. For this study. By adding the dead load response. Still. This method is significantly more expensive (i. 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. it is necessary to assemble not only the stiffness and mass matrices. For the interested reader.9). the response due to live loads (i.

the residual after k+1 iterations is approximated by linearization around the previous iterated solution as: err r ( k+1 + q 1 t+ t . For the present study. The second order accurate average acceleration method is obtained by setting α = 0 (no numerical energy loss). q&.15b) where γ and β are the Newmark’s integration parameters and are evaluated for the modified average acceleration method as γ = 5 . 0 + α and 2 β = 0. Denoting k q t+ t ∆ as the approximate value of q t+∆ t resulting from iteration k.e. a procedure using full8 Newton-Raphson iterations is adopted for solving the residual equation. by means of an iterative solution technique. with α > 0 . the residual (error in nodal forces or dynamical out-of-balance forces) can be specified by rewriting equation (9. i. r err (q t+∆ t ) = 0 .15a) 2  q& t+ t ∆ = q& t + t ∆ (1− γ )q& t + t ∆ γ q& t+ t∆ (9.& t) (9. The parameter α introduces numerical damping to the solution but reduces the accuracy to the first order.16) These should vanish at the end of each incremental time step.25 1 ( + α ) . In analogy to the static nonlinear analysis.2 t+ t ∆=qt+t ∆ q& t + t ∆  − β q& t + t ∆ β q& t+ t∆ (9.1) to give (note that f also includes the dead load): err r (q) = Mq& + Cq& + p(q)− f(q. q.

the tangent stiffness and iteration matrices are updated in each iteration.∆) err =r(k q t+ t ∆ )+ S( k q t+ t ∆ )( k k q t+∆ t − q t+ t ∆ ) (9.18)  ∂q  ∂q ∂q ∂q k k    q q t+ t t+ t ∆ . – 139 – & & S( ∂r   ∂q ∂q ∂f  q t+ ) err kt= ∆   = M +C +K− 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.

15) and setting q& t+ t = q ∆ & t .15). In each iteration. velocities. This is done here by using equation (9. Making now use of equation (9. giving: k −1 err ∆q = −S r ( k q t+ t ∆ ) (9. see [26]. velocities and accelerations. the iteration has to be started from a prediction of the displacements. and accelerations are then computed as: k+1 k k q t+∆ t = q t+∆ t + q ∆ (9.∆ The last term f ∂/q ∂ reproduces the effect of displacement-dependent external loads and is.19) β∆t 2 β∆t At each time step. the displacement corrections are computed using equation (9. for simplicity.20) The corrected displacements.21a) k+ k q ∂& k k γ k q& 1 t+ t = q ∆ & t+ t + q ∆ =q ∆ & t+ t + . the iteration matrix can be specified in detail as: 1 γ S(q) = M+ C + K t (9.17) and setting the residual at iteration k+1 equals zero. omitted in the present study.

the residual vector err r . i.21b) q ∂ ∆ βt ∆ k+1 k q ∂& k k 1 k q& t+∆ t = q& t+ t + q ∆ =q ∆ & t+ t + q ∆ (9. a convergence study was carried out for each numerical example to find the optimum number of increment. as this not only drives the accuracy of the integration but also governs the stability of the iteration process. the error in nodal forces. choosing an adequate time step is a critical issue. As mentioned earlier. 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. .q ∆ (9. For the described procedure.e. since the time step is – 140 – problem-dependent.21c) q ∂ ∆ βt ∆2 By iteratively updating the solution using the corrections k q ∆ . At each time step. is reduced when proceeding form iteration k to k+1.

Frequency analysis was also conducted for all three examples.22) and/or nonlin lin nonlin q 1 − static.: nonlin q static. 17. the dead load response is always evaluated using a nonlinear procedure. under the action of moving vehicles. This procedure is generally expected to give quadratic convergence [17]. In the present study. tot = q dead + q live = q dead + K f t live (9.23) To evaluate the nonlinear static response.e. 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 . and a medium span cable-stayed bridge. an incremental-iterative procedure using full Newton-Raphson iterations is adopted. the applicability of the catenary cable element. tot = q dead+ live (9.9. The nonlinear behavior of cable supported bridges during erection and application of the dead load is also presented. In the last two examples. 55]. while the live load static response is evaluated either linearly based on the dead load tangent stiffness matrix or nonlinearly. – 141 – – 142 – Chapter ______________________________________________________________________ Numerical Examples ______________________________________________________________________ In this chapter. the stiffness matrix is reformulated as the bridge deforms. are presented. is demonstrated.2 Static analysis In this study. i. a long span suspension bridge (the Great Belt suspension bridge). To verify and illustrate the efficiency of the developed finite element code. There are several procedures for evaluating the nonlinear static response and the most frequently used are the Newton-Raphson iteration schemes [11. for modeling cables in cable supported bridges. static and dynamic responses of a simply supported bridge.

the analysis was repeated using a smaller time step (more increments) until good agreement is shown between the last two analysis results. . For the linear and nonlinear dynamic analysis results presented in the following study. and 55.91 Hz. 10. For the nonlinear dynamic procedure. when evaluating the bridge-vehicle contact forces. 15. due to considering shear effects and the utilization of the deformed dead load tangent stiffness matrix. used to control the convergence of the bridge-vehicle dynamic interaction (similar to equation (3. the first four bending natural frequencies obtained utilizing the dead load tangent stiffness matrix are: 3. The bridge was discretized. 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).14) in Part A. However.2) was set to 10-5. the bridge initial velocity and acceleration vectors at time t = 0 were assumed to be zero.1. These tolerance values were chosen from the practical engineering point of view. As expected. see also section 9. To find the optimum number of eigenmodes. As the main aim of this study is not to develop formulas to be used in design specifications for calculating the dynamic amplification factors.6 . For the present bridge model.01. The dead load displacements obtained from the nonlinear static analysis were used as the bridge initial displacements. the shear modulus of all beam elements in the models was evaluated as G = E / 2. to be considered in the mode superposition procedure. these are lower than the frequencies given in Table 4. Zero initial values were also assumed for the moving vehicles vertical displacements and vertical velocities. the results were also checked after repeating the analysis including more eigenmodes. if nothing else is mentioned. the tolerance for the residual based convergence criterion (see section 9. 33.19. Throughout the dynamic investigation.1 Simply supported bridge The simply supported bridge studied earlier in Part A of this thesis was also adopted for this investigation. for convenience. 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. The response of this simple bridge model was mainly studied in order to verify the developed algorithms. for both the ABAQUS and the present model. the bridge deck initial dead load vertical displacements were assumed to be zero. This was done by comparing the results with those obtained using the commercial finite element code ABAQUS. were set to 0. the tolerances for the three criteria.bridge structures are given in Chapter 4. For the convergence study.1.41.1).97. Further. 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. using 70 beam elements.

The modeled vehicle. 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.05.2 it was assumed that. Thus. as this is usually the case. 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. except for the mid-point vertical acceleration. to obtain a stable mid-point vertical acceleration curve with good agreement to the curves in Figure 10. No bridge damping was included. If no compensation is made during construction.2b-d.1b-d and 10. 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.The properties for the bridge and the vehicle models are given in Figure 10. The present linear and nonlinear static traffic load responses were found to be identical. For the results in Figures 10. this vehicle was modeled as two separate sprung mass systems one for the truck and one for the trailer. Only 4 modes were considered when producing the present solution curves. no initial dead load vertical displacements in the bridge deck were present.35 Hz. The body-bounce and wheel-hop frequencies for the truck and the trailer models were chosen as 1. with the corresponding mode shapes shown in Figure 8. permitted in Sweden but not within the European community. As seen. are plotted on the same diagrams and the results are found to be in very good agreement with each other. It is normally attempted to compensate by precambering the bridge deck during construction so the dead load displacements are cancelled by the specified camber. For the present direct integration solution. due to only traffic load.2a. About 30 % increase in the vehicle weight is observed in Figure 10. which required 10 modes to converge.1 and 10. it was essential to introduce numerical damping by setting the parameter α to be equal to 0. In Figure 10. For the three solutions. the initial displaced geometry of the 9 For a certain point P. except for the bending moment.2b as a result of having about 9 mm higher surface at the bridge entrance than the approach pavement.1b. except the automatically introduced numerical (artificial) damping in the ABAQUS solution. In Table 10. This numerical damping is believed to cause the difference in Figure 10.1. the interaction between the truck and trailer was – 144 – disregarded and only vertical modes of vibration of the vehicle were considered.1a.89 Hz and 11. the ABAQUS and the present mode superposition solutions. 300 increments were chosen with a time step of about 6 ms. as a result of this simplification.1d. 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. is a 24 m long truck and trailer with a total weight of 60 ton.1d for the mid-point vertical acceleration. ξ = 0 . – 145 – . The vehicle model was assumed to move.

k = 14. 6 N/m s k c k /2 c /2 -3.6 10 .0 10 . 4 Ns/m s -2.0 ss s s Bridge m m 2 2 /2 . 4 Ns/m p c = 12.0 106 N/m p m m /2 3 3 k = 8.0 (b) m = 0 kg ABAQUS 1 m = 4400 kg Present 2 m = 35600 kg -1.0 10 .0 Static (present) 3 24 m c = 1.(a) v = 25 m/s Truck and Trailer 0.0 .

50 1.75 1.25 1.50 0.75 Time (s) 0.00 1.00 0.E = 3.0 10 .0 m = 11400 kg/m 34 m 0.75 m2 -5.0 m m 1 1 /2 Vertical displacement (mm) 12 m A = 4.5 (d) ABAQUS )2 -1.0 0.0 Present (10 modes) ABAQUS (c) Static (present) 0.25 0.3 . 10 N/m2 k c /2 k c pp p /2 p I = 3.307 m4 -4.

0 Bending moment (MNm) Vertical acceleration (m/s -5.1 Mid-point vertical displacement (b).0 Present-trailer 1.00 1.75 1.1 .0 1.50 1.0 -0.25 0.4 (a) (b) ABAQUS-truck 1.00 0.0 0.25 0.3 ABAQUS-trailer 10.25 1.00 1.0 -3.75 Time (s) Time (s) Figure 10.75 1.50 1.3 -4.2a.2 Present-truck 5.50 0.0 1.00 0.5 0. The road surface roughness profile is shown in Figure 10.Present -2.75 0. bending moment (c).0 -0.25 1. and acceleration histories (d) for the problem defined in (a).50 0. The dashed vertical lines indicate when the trailer enters and the truck leaves the bridge 15.

9 Surface roughness (mm) -5.0 0.0 .0 Vertical displacement (mm) Vertical acceleration (m/s -15.0 5.0 0.25 1.0 -1.75 1.00 1.50 1.0 0.0 Normalized contact force 0.00 0.25 0.50 0.1.0 (d) ABAQUS-trailer ) ABAQUS-trailer 2 Present-trailer 15.0 Present-trailer Present-truck Present-truck 1.0 -5.0 ABAQUS-truck ABAQUS-truck (c) 2.75 Distance along the bridge (m) Time (s) 25.8 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 0.

The dashed vertical lines indicate when the trailer enters and the truck leaves the bridge 0.0 0.50 1.-2.00 1.00 0.25 1.25 0. truck and trailer body vertical displacement (c).0 With initial displacement (nonlinear) 10.75 Time (s) Time (s) Figure 10.0 0.50 0.0 -20.25 0. when travelling over a bridge having the road surface profile shown in (a).75 0.0 -3.0 (a) No initial displacement (linear) 20.50 0.00 0.0 -2.0 .0 Vertical displacement (mm) Vertical displacement (mm) -5.75 1.75 1.25 1.00 1.50 1.2 Normalized bridge-vehicle contact force (b). and acceleration histories (d).0 -10.0 (b) No initial displacement (linear) With initial displacement (linear) With initial displacement (linear) With initial displacement (nonlinear) -1.0 -4.

5 -4.05 (nonlinear) ξ = 0.25 Time (s) Time (s) Figure 10.25 0.50 1.75 0.75 1.3 Effect of initial dead load displacements on mid-point vertical bridge displacement (a) and on vertical displacement of the truck body (b) 0.50 0.05 (linear) = ξ 0.5 = ξ 0 (linear) ξ = 0.0 = 0.0 Bendi .5 -3.00 0.25 1.0 ng moment -3.5 -2.05 (nonlinear) (MNm) -1.00 0.0 -2.50 0.00 1.75 1.0 (a) ξ = 0 (linear) -0.00 1.05 (linear) ξ (b) -1.0 0.-30.25 0.

00 1. are plotted.151 1.00 0. as this will influence the bridge-vehicle contact force and thereby the bridge and vehicle responses.145 1. 3 .25 0.75 0. in Figure 10. and truck body displacements.4 Effect of damping on mid-point vertical bridge displacement (a) and bending moment (b) bridge deck must be considered.00 0. assumed constant for all modes. due to traffic load only.00 1. using the two present analysis methods.75 1.50 1.50 1.50 0.0 -4.Vertical displacement (mm) -5.25 1. Mode Superposition (present) Direct Integration 4 modes 10 modes (present) DAF d 1.5 0. iterations/inc.25 1.25 0.50 0. are plotted.095 1. This is shown.3 where mid-point vertical bridge displacements.75 Time (s) Time (s) Figure 10.154 CPU time (s) 60 85 830 Max.4.75 1. where the undamped solution and the solution with 5 % damping ratio.142 1. The effect of damping is shown in Figure 10.143 DAF m 1.

180 77. Both the Great Belt bridge and the Akashi Kaikyo bridge were opened for traffic in 1998. only analysis results obtained using these two algorithms will be presented and no comparison will be made to results from commercial codes like ABAQUS. 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. The two main cables. without expansion joints at the pylons. the girder section is streamlined (almost wing shaped) to resist strong wind action. i. Like many other modern suspension bridges. A frequency analysis for the completed bridge was also conducted. – 149 – girder as it is called – is a 31 m wide and 4 m deep steel box.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. This gives higher pylons than usual and therefore a more flexible structure. 10. the main cables are fixed to the stiffening girder through rigid clamps at mid-span. To minimize deflections under asymmetric traffic load. including cable and cable saddle. were erected in the autumn of 1996 by the air-spinning method11 [27]. The Great Belt (Storebælt) suspension bridge in Denmark has.3 2 Table 10. The total height of the concrete pylons. each include a total of 18648 high strength galvanized steel wires 5.2 The Great Belt suspension bridge In this numerical example. the second largest span in the world10. Therefore. made using this simple numerical example. in the following sections.85 m. the behavior of the Great Belt suspension bridge during girder erection and under the action of moving vehicles is presented.6 535 m 1624 m 535 m . This girder is continuous over the full cable supported length of 2694 m.e.38 mm in diameter and have a final diameter of 0. is approximately 258 m. The ratio of cable sag to main span length was chosen to be 1/9 as this was found economically favorable. with its 1624 m main span. provide some confidence in the application of the implemented direct integration and mode superposition algorithms.

the girder was assumed to be simply supported at the ends. for the Brooklyn bridge in New York dating from 1883. and every third hanger from the original bridge was included and modeled using catenary cable elements.5 m Pylons 0.803 Hz.00 3. To simplify the data input process.147 Hz and 0. 200. E (N/m2) A (m2) I (m4) w (t/m) Girder 2.g. 60) 75. 30.41 2 × 3.2 were taken from [28.Figure 10.5. which are acceptable when compared to the measured pylon frequencies reported in [46]. – 150 – positioned at the same level as the girder. e.78 (31 m × 4 m) Pylons 0.45 side spans Cable . after few iterations.4·1011 2 × 37.5 Geometry of the Great Belt suspension bridge For the results presented here. during analysis. a simple bridge model consisting of 126 elements.1·1011 1.5 2 × 750 2 × 90 0 Æ 75.5 and the properties given in Table 10. The main cables were assumed to be fixed at the pylon tops and the pylons were rigidly fixed to the piers. The bridge geometry shown in Figure 10. 47]. 150) 2 × (78. was used. The section properties for the pylons were assumed by the author to give the first and second pylon frequencies 0. and the configuration of the main cable under dead load was determined accurately.6 m Cable 2.32 14.1·1011 2 × 0.4·1011 2 × (32. 72.5 Æ 257. 89 nodes and a total of 215 active degrees of freedom. 25) 2 × (275. In the bridge model. all internal cable nodes were vertically 11 Much the same method was used for suspension bridges more than 100 years ago.

cable sag at mid-span and pylon top displacements. i.1 Static response during erection and natural frequency analysis Erection of a suspension bridge involves many challenging problems. the initial side span and main span cable lengths were determined using a trial and error procedure. different erection stages 185 were analyzed and the results are .2.1·1011 2 × 0. When the initial cable lengths were known.5. 48]. Table 10. and the pylon tops horizontal displacement). The estimated values were then improved and the calculation was repeated until the required dead load profile was obtained.09 m and 0.1·1011 2 × 0. For the analysis presented in this thesis it was assumed that the erection proceeded simultaneously from mid-span and anchor blocks towards the pylons.e.2.025 † † Mass of hangers and clamps are considered distributed uniformly along the main cable and included in the cable mass. 47.e.2 Parameters for the model of the Great Belt suspension bridge 10.36 main span Hangers 2.8 m and 1672. The erection of the girder may proceed in a number of different ways [36. It was found that if the initial lengths of the side span and main span cables were chosen as 564. were determined and compared with the desired dead load profile shown in Figure 10. To study the erection procedure of the Great Belt suspension bridge and calculate the initial profile for the free hanging cables (i. thus various structural components do not receive or render the kind of support intended in the complete structure.7 m. the initial cable lengths. which relate to the fact that the bridge structure is incomplete. To start the analysis. the simple 2D model described earlier is used.40 2 × 3. Those initial cable lengths were therefore considered to be good enough and were used for all the following results presented for this example. the cable sag at mid-span. Thus the initial cable lengths were estimated and the – 151 – final dead load profile. the calculated mid-span cable sag and pylon tops horizontal displacements would be 180. especially aerodynamic stability problems.04 m.

(a) plotted in Figure 10. Studying this figure one can notice that. This is done either by displacing the saddles in relation to the pylon tops 0. Consequently.0 or by pulling back the pylons with 0 20 40 60 80 100 .5 cable during the subsequent erection (b) of the girder. 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 relation to the vertical pylon axis. the dead load =180 m pylon tops must be displaced about 0. be horizontally displaced in 0.0 the girder the saddles at the pylon tops must therefore.85 m outwards and the cable sag 175 should be about 173 m. to compensate for the later erection of 1. prior to cable erection.6. to arrive 180 specified final cable sag under at the desired dead load profile.

i.7 within brackets and as can be seen the agreement is very good when comparing with the result obtained from the present analysis. the remaining displacement before erecting the girder was about 1 m. The CPU time used by the MATLAB process. 48]. When the main cables were erected and the tie-back cables were dismantled.135 Hz) Figure 10.6 Cable sag variation (a) and [36].Pylon tops horizontal displacement (m) so called tie-back cables or a % of bridge deck erected combination of the two methods Figure 10.115 Hz) mode 3: 0. when comparing this value with the one obtained from the present analysis.e.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.130 Hz (0. analytical frequency results from a 3D finite element model and results from a 1:200 scale aeroelastic bridge model. 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. 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. 0. These results are also given in Figure 10. This is believed to be the major explanation for the differences in the results. Acceptable agreement is found.112 Hz (0. Frequency analysis was also conducted for the completed bridge and the first three natural frequencies and mode shapes are given in Figure 10.100 Hz) mode 2: 0. according to the author’s opinion.20 m [28]. 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.099 Hz (0.24 m [28] prior to main cable erection. for each bridge member. Thus. the pylon tops moved back about 0. tie-back cables running from each pylon top to the nearest anchor block were used pulling back each pylon 1. During the constructing of the Great Belt suspension bridge.7 Natural frequencies and mode shapes for the lowest three vertical bending modes of vibration. are presented.7. made for wind tunnel testing. mode 1: 0. . was about 200 seconds. Values inside brackets are reported in [48] – 153 – In [47.

1b and the model properties are given in Figure 10.2. was assumed to have a total weight of 44 ton and a length of 18.2 104 Ns/m .10. moving on a smooth road surface from the left to the right at the constant speed of 25 m/s (90 km/h). The free distance between the two trucks was assumed to be 40 m. The body-bounce and wheel-hop frequencies.8. The corresponding mode shapes are shown in Figure 8.35 Hz.89 and 11.75 m. For the c = 1.2 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.76 104 Ns/m k c p k c pp pp c = 13. were chosen as 1. 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 each truck model.

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 .10. 0 ξ= .9 and 10. reasonably converged 1 1 k = 15.8 Model of the two trucks step of about 0. – 154 – Some interesting diagrams are shown in Figure 10. Bridge damping ratios were evaluated according to equation (8. acceleration. No numerical damping was introduced in the direct integration procedure. these are insignificant from the engineering point of view.18 s.s m m two procedures.4 106 N/m p k = 8.9 and 10. and contact force curves in Figures 10.8 106 N/m reliable solutions were obtained using 58.75 m s 600 increments corresponding to a time Figure 10. 026 . In Table 10. 0 1 2 ξ3 = L 023 .10 where the linear and nonlinear solutions are plotted. the first 25 modes were found to be sufficient for calculating the dynamic response of the bridge and the trucks. 029 .3. For the mode superposition procedure.and 30-mode solutions did not differ significantly. Although some differences can be observed between the linear and nonlinear dynamic moment. as 25.15) giving: ξ = . 0 etc. These figures do not show very significant difference between the two solutions.

250 1.274 CPU time (s) 845 5930 Max. the maximum dynamic moment for the fixed end of the left pylon was found to be nearly 30 % larger than the static one. as this will greatly affect the dynamic response of the bridge-vehicle system. linear static and linear dynamic traffic load response analysis of long span suspension bridges is adequate. as road surface with no roughness was assumed. utilizing the dead load tangent stiffness matrix. see Figure 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. As an example. Even though the excitation of the dynamic system was only caused by the elastic displacement of the bridge itself.10a. if bridge damping is not considered. 4 3 Table 10.22. the dynamic amplification factor for the deck vertical displacement at the center of the bridge will increase from about 1.corresponding CPU time and maximal iterations per increment required for solving the problem.9b) to about 1.10 (Figure 10. The response was also evaluated neglecting bridge damping and it was found that correct estimation of bridge damping is very important. Linear Dynamic Nonlinear Dynamic (mode superposition) (direct integration) DAF d (pylon top) 1.127 1. – 155 – 30 20 (a) 10 0 -10 -20 Dynamic (nonlinear) . iterations/inc.123 DAF m (pylon fixed end) 1.

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 Static (linear & nonlinear) Vertical displacement (mm) -300 0 10 20 30 40 50 60 70 80 90 100 110 Time (s) 0.06 ) (c) .

06 0 10 20 30 40 50 60 70 80 90 100 110 Time (s) Figure 10.00 -0. The curves are not easy to distinguish.04 0.2 0. as the responses are almost identical – 156 – 40 Dynamic (nonlinear) (a) 20 Dynamic (linear) Static (linear & nonlinear) 0 -20 -40 Bending moment (MNm) -60 0 10 20 30 40 50 60 .02 0.9 Horizontal displacement of the left pylon top (a). bridge deck vertical displacement (b) and vertical acceleration (c) at the center of the bridge.02 -0.04 Nonlinear Vertical acceleration (m/s Linear -0.

94 Normalized contact force 0.98 0.70 80 90 100 110 Time (s) 15 Dynamic (nonlinear) (b) 10 Dynamic (linear) Static (linear & nonlinear) 5 0 -5 -10 Bending moment (MNm) -15 0 10 20 30 40 50 60 70 80 90 100 110 Time (s) 1.08 (c) Nonlinear 1.04 1.00 0.92 .02 1.96 0.06 Linear 1.

This bridge is similar in configuration to an existing bridge in Japan (The MeikoNishi Bridge in Nagoya) with some modifications in dimension.65 above deck level Pylons 2.38 deck to pylons .56 0.3 Medium span cable-stayed bridge A 2D model of the cable-stayed bridge described in [55] was adopted for this investigation.01 34.0·1011 0. E (N/m2) A (m2) I (m4) w (t/m) Girder 2.10 4.8·1010 18.8·1010 13.52 30. 55]. 40.26 19.64 † central part Pylons 2.0·1011 0.4.93 0. The static and dynamic behavior of this bridge has been studied earlier by several other investigators [4.11 1.78 below deck level Links 2.10 Bending moment at the fixed end of the left pylon (a) and for the bridge deck at the left pylon (b).29 19.11 and the properties are given in Table 10.64 † Girder 2.0 10 20 30 40 50 60 70 80 90 100 110 Time (s) Figure 10. The bridge geometry is shown in Figure 10.31 43.0·1011 1. normalized bridge-vehicle contact force for the first truck (c) – 157 – 10.58 86.

0·1011 0. Cable no. 13 2.409 Table 10.0·1011 0. .0204 111.0·1011 0. . 17. 30.0·1011 0.† Including weight of cross beams. 11.125 12. 22 2.4 Parameters for the cable-stayed bridge model defined in Figure 10.0·1011 0.12 0.0139 68.0232 134. 20 2.194 5.7 cable12 cable13 .0362 158. 21 2. 24 2. 8.0·1011 0.0372 158.80 0.69 0. 23 2. 9.43 0. 14.153 6.11 – 158 – node 43 .5 cable 6 node 7 elem. 36 node 14 element 52 node 36 node 13 node 18 . E (N/m2) A (m2) u L (m) w (t/m) 1.255 3. 18.0176 89.13 0. 5x3 element 61 cable16 45.64 0.225 4. 7.398 2.0·1011 0. 19 2. 15.10.0113 51.66 0. 16..

a hardening characteristic with respect to the applied load is apparent. and all cables were assumed fixed to the pylons and to the girder at their joints of attachment. The cables are numbered from the left to the right starting with cable 1 For the model. natural frequencies were also determined replacing each stay cable by 3. it was assumed that the girder was pinned at the ends. i. Examining this figure. only rotations were allowed. 5.e. a nonlinear static analysis under dead load is essential to arrive at the deformed dead load tangent stiffness matrix.3. It is also evident that at the start there is a significant nonlinear behavior during the static application of the dead load. element 30 146. to include cable motions. and 7 catenary cable elements. This means that influence lines and superposition technique can be used in the design process. described in terms of the vertical displacement of the girder at the center of the bridge and the tension in cable 12 and 13. For the frequency analysis. starting from this dead load deformed state.e. as the span length increases this nonlinearity will get more pronounced [55] and linear live load analysis might no longer be adequate. Thus. i. 10. The simplest model analyzed in this example.1 Static response and natural frequency analysis Figure 10.3 m Figure 10..3 m 335. was composed of 66 elements and 43 nodal points.3 m 146. However. it is believed that this bridge will behave as a linear system. For this cablestayed bridge with modest main span length. when affected by live static and dynamic loads.11 Geometry of the cable-stayed bridge. as the nonlinearity is not so strong above this dead load equilibrium point. the model with one element per cable.12 shows the nonlinear behavior of the model under static dead load. The pylons were assumed to be rigidly fixed to the piers. This was – 159 – 5 5 (a) (b) 4 4 3 3 tangent tangent . and elastically connected to the pylons by vertical links.

i.703 Hz (0. it is evident that cable motions are associated with every . the model with one element per cable.e. (b) cable tension mode 1: 0. the resulting modes of vibration only include the vibrating girder and pylons. For the finer models.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. were obtained between those basic bending modes. 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.13 Natural frequencies and mode shapes for the first three vertical bending modes of vibration. i.334 Hz (0. cable modes and the dynamic interaction between the vibrating cables and the bridge were disregarded. Moreover. For the simplest model. Thus.12 Nonlinear behavior of the cable-stayed bridge defined in Figure 10. pure cable modes.311 Hz) mode 2: 0.e.437 Hz (0. additional new mode shapes characterized only by vibrating cables.650 Hz) Figure 10.11: (a) vertical displacement.411 Hz) mode 3: 0.

741 (8) 0.436 (2) 0.742 (8) 5 0.044 (6) 1.218 (27) .334 (1) 2 0.692 (3) 0.703 (7) 4 0.053 (18) 1. as can be noticed in Figure 10.874 (13) 0.739 (8) 0.703 (7) 0. For the four alternative models.437 (2) 3 0.212 (7) 1.868 (5) 0. Table 10.334 (1) 0.437 (2) 0.5 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.875 (13) 6 1.051 (18) 1.053 (18) 7 1.437 (2) 0.734 (4) 0.13. element/cable elements/cable elements/cable elements/cable 1 0.875 (13) 0.332 (1) 0.mode of vibration.334 (1) 0.700 (7) 0.

388 (29) 10 1.214 (8) 1.1. For the simplest model with a total of 119 active degrees of freedom.388 (33) 1.690 (42) 1. 55]. surface irregularities at the bridge entrance.220 (28) 1. the girder and pylons were modeled in [4. This indicates high efficiency of the presented elements. and girder supporting conditions. the CPU time used by the MATLAB process. tuned vibration absorbers.379 (9) 1. 3. Each cable was modeled using 1.11.222 (24) 9 1.13 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. 5 and 7 catenary cable elements. 55]. 55] using conventional beam elements modified by the stability functions and a diagonal lumped mass matrix was adopted for all elements. was about 15 seconds. bridge damping. number of vehicles on the bridge. Moreover.220 (23) 8 1. vehicle speed.680 (38) 1.220 (23) 1.387 (29) 1. 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 [4.3.2 Dynamic response due to moving vehicles – parametric study For the cable-stayed bridge shown in Figure 10. cable modeling.222 (24) 1.5 Comparison of the first ten natural frequencies for vertical bending modes of vibration. 40.671 (10) 1.687 (38) Table 10. 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. The disagreement in the frequency results shown in Figure 10. 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. . 10. a study was conducted to assess the importance of different factors that influence the dynamic response due to moving vehicles.

The corresponding dynamic amplification factors (DAF) for the different studied cases are given in Table 10. these static solutions corresponding to the linear and nonlinear dynamic solutions were used when 12 This can be concluded when comparing Figure 7.35 Hz. including all 119 modes of vibration and using 1000 increments.0056 0 1 2 30 . .1 Response due to a single moving vehicle In this study. one should observe that the dead load response (always evaluated using a nonlinear procedure) was subtracted from the solutions.e. • Vehicles moving from the left to the right at the constant speed of 25 m/s (90 km/h) on a smooth road surface.6 (original configuration). Also. • No numerical damping was introduced in the direct integration procedure. • The first 30 modes were considered for the mode superposition solution. was also performed for this case and the results are also given in Table 10. In Table 10. A linear mode superposition analysis. The locations of these nodes and elements are shown in Figure 10. thus only the responses due to traffic loads are plotted. the model with one element per cable.11. 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. – 162 – calculating the DAF. • 1000 increments corresponding to a time step of 0.16) giving: ξ =ξ = ξ L = . 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. 10.89 and 11.Some interesting figures are presented at the end of this numerical example. and 1500 increments are given.8 for the model properties.025 s. The following was adopted for the parametric study unless otherwise specified: • The simplest bridge model. See Figure 10. The body-bounce and wheel-hop frequencies for the truck model were chosen as 1. linear and nonlinear dynamic responses due to a 44 ton moving truck are compared. The plotted cable axial forces are the average of the axial forces at the two ends of the cable. Also. • Bridge damping ratios were evaluated according to equation (8.3a with the one published in [7] showing the response for the one bar element with an equivalent modulus of elasticity. • The same 44 ton vehicle model as for the Great Belt suspension bridge example. i.2. some results for different time steps corresponding to 500.6 at the end of this numerical example.3.6. where element and node numbers are referred to. 1000.

a maximum of 3 and 2 iterations per increment were required and the corresponding CPU time was 470 s and 1890 s.and 35-mode solutions did not differ significantly. The response curves for the tension in this cable. Based upon the presented results.6). 1000 increments and 30 modes were adopted for this investigation. show a fluctuation with a frequency of about 1. For the linear and nonlinear dynamic solutions. 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. the natural frequencies for the first three vertical bending modes of vibration will decrease to 0.19 Hz.159 when considering all 119 modes of vibration. Consequently. respectively.15b and 10. for the case studied here. it was found that the 30. unless otherwise specified.16a. a significant increase in the DAF was obtained for the tension in cable 13. the DAF for the tension in cable 6 and for the vertical displacement at node 18 are lower for this case. are presented in Figures 10.14 – 10. which also lead to differences in the DAF (see Table 10. However. 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. As can be seen. the highest DAF of all cables (1. As mention earlier. as seen from the results in Table 10. there are higher modes than 35 that are affecting the response.275). which is the traditional way of supporting such a medium long bridge girder.676 Hz. for this bridge model it was assumed that the girder was pinned at the ends. for the mode superposition procedure. and 0. and 1500 increment solutions it is clear that the 500solution is not good enough and the 1000-solution does not give perfectly converged results. The DAF for this simply supported girder case are also included in Table 10. was caused by vibration modes 7 and 8 having a frequency of about 1. Figure 10.413. The first is that linear dynamic analysis give sufficiently accurate results as there is no apparent nonlinear . 0. 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. the horizontal displacement of the right pylon top. 1000. from the practical engineering point of view. for this single moving vehicle case.21 Hz. Although there are some minor differences at the peak regions in. for the bending moment at the right pylon fixed end. also the dominant frequency for the vertical acceleration of the corresponding cable-girder attachment point. for example. Comparing the 500.16.180 to 1. and for the fixed end moment of the same pylon. three general conclusions can be made.6. For example. If a simply supported girder at the ends was assumed instead.During the convergence study. This fluctuation in cable tension. the DAF decreases from 1. However. By examining those figures. – 163 – Some linear and nonlinear response curves.6.16a. However.314. Figures 10. analyses with 1000 increments and 30 modes give reasonably converged reliable solutions. Analysis results indicated that cable 6 has.

17 shows some response curves obtained from a nonlinear dynamic and nonlinear static traffic load analysis. where the moving vehicles were modeled either as constant moving forces (i.4. it was found that the moving force model. gives for this flexible bridge significant differences especially for the tension in cable 6.39 for the axial force in the deck at the left pylon) even though a road surface with no roughness was assumed. Figure 10. For the study of bridge-vehicle interaction effect. cable motions and . The distance between truck models and model properties are shown in Figure 10. ignoring interaction) or as sprung masses moving on a smooth road surface. 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. Therefore.e. 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.021 s was required for this analysis. 1500 increments corresponding to a time step of 0.6.2 Response due to a train of moving vehicles. It is believed that this difference in response occurs. an investigation of the bridge-vehicle interaction effect was also conducted for this case. On the other hand. The DAF values are given in Table 10. no particular trend was found when comparing the DAF of the two cases. 10. According to the study conducted in Part A. which usually gives negligible differences in results compared to the sprung mass results when the road surface roughness is ignored.behavior of the bridge during application of the moving traffic load. The second conclusion is that high DAF values can be reached (1. – 164 – The third conclusion is that the supporting condition of the bridge girder has a significant influence on the dynamic response. as mentioned earlier in section 8.8. 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 was found that the maximal dynamic part of the contact force increases by 87 %. Consequently.2. Comparing the nonlinear dynamic results of the one truck case (original configuration) with the case of four trucks modeled as sprung masses. as the DAF – 165 – increases significantly for some of the studied elements and nodes but decreases for others.6. one of the important factors affecting the bridge response is the dynamic interaction between the vehicles and the bridge deck.3. Examining this figure and also Table 10. In addition. and the absolute maximal vertical acceleration of the girder at the bridge center by 125 %. as the result of having the vehicles acting as vibration absorbers when modeled as sprung mass systems. The case of four moving trucks modeled as sprung masses was also solved for the bridge model with 5 catenary cable elements per cable.

and 130 km/h.333 to 1.18d.19c and 10. Only – 166 – nonlinear dynamic analysis was performed and the results are presented in Figure 10. 0.6.3 Speed and bridge damping effect To investigate the effect of vehicle speed and bridge damping on the dynamic response. as the tension in all cables were effected regardless of their length or position (side or main span).19 and in Table 10. was also found for the bending moment in pylon element 61. even though the DAF tends to reduce at certain speeds. to avoid an underestimation of the dynamic response.400 to 1. The original configuration ( v = 90 km/h) was also analyzed here for new bridge damping ratios.388 to 1. Mixed results were obtained when the new solution was compared with the 1 element per cable solution. where the DAF increased from 1. Furthermore. no particular trend could be found for the cables. As seen. 0. it is evident that the response generally increases with the increase in vehicle speed. 1500 increments were used for the solution of the 50 and the 70 km/h cases. The effect of damping on the reduction of the bridge dynamic response was found to be considerable.18c). As a conclusion. 0.19b for different speeds and damping ratios. with respect to different speeds and damping ratios are shown in Figures 10. This is the same case as the socalled original configuration but is solved here for the new speeds of 50. For the cable tension. The fluctuation of tension in this cable is shown in Figure 10. the largest difference was found for cable 16 where the DAF increased from 1.02.422.19a and 10.596). The variation of the DAF.01.658. 110.6. while the influence of damping on the axial . see Figure 10. namely: 0. and 1000 increments for the rest.cable vibration modes were considered in this analysis.19d. when the 5 element per cable model was adopted. some response curves obtained from a nonlinear dynamic and nonlinear static traffic load analysis are presented. Among the selected elements and nodes. 0.03. As an example.2. 70. for some selected elements and nodes. As expected. 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. In Figure 10. Moreover. damping reduces the DAF. the cables should always be modeled so their motion and modes of vibration are considered in the analysis. the amount of reduction is largest for the axial force in pylon element 52 and the vertical displacement at node 18. 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.18b. A significant increase in DAF (from 1. 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. a single 44 ton moving truck was adopted.3.18. The vertical displacement of the girder at the center of the bridge is shown in Figures 10. 10.015. see Table 10.

6. Some results from the nonlinear dynamic analysis are presented in Table 10. The results for this case presented in Table 10. The optimum tuning parameters given in equation (8. The TMD was positioned at the center of the bridge (node 14) and tuned to the first bending mode of vibration (0. which is not unusual. .39 to 1. Results from the nonlinear dynamic analysis are presented in Figure 10.2.3. Figures 10.g. where the mass ratio was set to 0. the response of nodes and elements that are located close to the bridge entrance were affected the most by this bump.6 and in Figures 10. it is noted in Figure 10.18) were adopted.force in deck element 30 and the tension in cable 6 are comparatively small. In addition.21-10.20d some curves are plotted to show the variation of the DAF with the vehicle speed. 10.. Moreover this bump affected the response of this cable more than the response of other cables.2. is considerable. 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. for as little bump height as 3 cm.3. From those results we can infer that the effect of this type of irregularity. and in Figure 10. on the bridgevehicle response.60.2. Of course. a 3 cm high and 2.33 Hz). increasing the DAF from 1. the bridge deck vertical displacement and acceleration at the center of the bridge were not affected at all.20a-c show response curves for a vehicle speed of 90 km/h. was generated as described in section 8. a significant increase in the axial force in deck element 30 was – 167 – found. The truck was assumed to move on a smooth road surface at the constant speed of 110 km/h.3.4 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.20 and in Table 10. e.6 should be compared with those corresponding to v =110 km/h. 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. a 45 % larger contact force than the static one was obtained.6.5 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. whereas. Again the highest DAF for all cables was found for cable 6.19d that the relationship between bridge damping ratio and the DAF is not always linear. As noted from Table 10. Also. 10.5 m long bump located at the left end of the bridge.005 giving a TMD mass of approximately 15.6 ton.23.

This should be taken into consideration when conducting field measurements to estimate the vibration frequencies of a bridge of this type.14 20 (a) . As an example. and the DAF was reduced from 1. despite the fact that it is not always effective in reducing the maximum dynamic response during the forced vibration period. the tension due to traffic load is shown in Figure 10.348 to 1. see Table 10. This is due to the increase of the overall damping of the bridge by the TMD. these antisymmetric modes do contribute for the pylon tops horizontal response. 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).e. In addition. For cable 24. As expected and observed in those figures. it was found that the two anchorage cables 1 and 24 had the largest reduction in the DAF value.4. e. tuned to the third mode and installed at the top of the two pylons. the DAF for cable 14 increased from 1. when the TMD was considered.13. with and without TMD.23b and 10. see mode 2 in Figure 10.049 to 1.22c.g.6. The DAF for cable 1 (highest DAF among all cables) was reduced by the TMD from 1. as seen in Figure 10. The corresponding bending modes of vibration are also indicated in those figures.13. It is clear from those figures that several modes were excited. However. due to the interaction between the bridge-vehicle-TMD systems. The results also revealed that the TMD not always is very effective in reducing the maximum dynamic response during the forced vibration period (i. In fact. However.130 to 1.g.23c. – 169 – 0.063. 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. the response and the DAF for certain elements and nodes can even increase due to the TMD. even for the case of cable 14. This is one of the disadvantages of a TMD. On the other hand. when the vehicle is on the bridge). but the first mode has the largest contribution to the response of the two chosen nodes. it is evident that the TMD is very effective in reducing the – 168 – vibration level in the free vibration period. a TMD tuned to the first mode is only effective in reducing the contribution of this mode to the dynamic response.Examining the results.306. increases the overall damping of the bridge by working as an additional energy dissipater. e. A considerable reduction was also found for the horizontal displacement of node 43 and the moment in pylon element 52. From this investigation it is concluded that a TMD. as an accelerometer installed at the pylon top will identify not only the symmetric modes but also the antisymmetric ones. To further reduce the vibration level of the bridge one might install extra TMDs.079.23c and also Figure 10. are presented in Figures 10. see the discussion in section 8.

02 -10 -20 -0.06 0 0.02 -30 -0.50 Dynamic (nonlinear) (c) (d) Dynamic (linear) 0.Dynamic (nonlinear) ) Nonlinear (b) Dynamic (linear) 2 0.14 0 5 10 15 20 25 0 5 10 15 20 25 Time (s) Time (s) 20 0.10 Linear 10 Static (linear & nonlinear) 0.10 Vertical displacement (mm) Vertical acceleration (m/s -50 -0.06 -40 -0.25 .

00 0 -0.00 0 5 10 15 20 25 0 5 10 15 20 25 Time (s) Time (s) Figure 10.50 Dynamic (nonlinear) Axial force (MN) Dynamic (linear) -20 -0. The dashed vertical lines indicate when the vehicle is at the left and right pylon 15 10 (a) Dynamic (nonlinear) (b) 10 Dynamic (linear) Static (linear & nonlinear) 5 5 (MNm) . axial force in deck element 30 at node 7 (d).25 -10 -0.75 Static (linear & nonlinear) Vertical displacement (mm) -30 -1.10 Static (linear & nonlinear) 0. vertical displacement at node 18 (c).14 Vertical displacement (a) and acceleration histories (b) at node 14.

80 0.40 Dynamic (nonlinear) Shear force (MN) -0.40 Static (linear & nonlinear) 0.80 (c) (d) Dynamic (nonlinear) 0.60 Dynamic (linear) 0.40 0.20 Dynamic (linear) .00 Axial force (MN) -0.0 0 -5 ng moment -5 Dynamic (nonlinear) -10 Dynamic (linear) Bendi Static (linear & nonlinear) Horizontal displacement (mm) -15 -10 0 5 10 15 20 25 0 5 10 15 20 25 Time (s) Time (s) 0.00 0.20 0.

bending moment (b).45 0. The dashed vertical lines indicate when the vehicle is at the left and right pylon 0. axial force (c).08 -0.20 Dynamic (nonlinear) Dynamic (nonlinear) (a) (b) Dynamic (linear) Dynamic (linear) 0.05 -0.Static (linear & nonlinear) -0.15 Horizontal displacement of node 43 (a).35 0. and shear force (d) in element 52 at node 36.15 0.16 Static (linear & nonlinear) Static (linear & nonlinear) 0.80 0 5 10 15 20 25 0 5 10 15 20 25 Time (s) Time (s) Figure 10.12 0.05 0.00 Axial force (MN) Axial force (MN) -0.25 0.04 -0.04 0.40 -0.15 0 .08 0.

50 0 5 10 15 20 25 0 5 10 15 .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 0.25 -40 Vertical displacement (mm) Vertical acceleration (m/s -50 -0.00 -30 -0.

16 Axial force in cable 6 (a).20 25 Time (s) Time (s) Figure 10. and acceleration histories (d). axial force in cable 13 (b). truck body vertical displacement (c). 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 Dynamic (sprung mass model) Dynamic (moving force model) -50 Vertical displacement (mm) Bendi -10 Static -60 -15 0 5 10 15 20 25 30 0 .

0 Dynamic (moving force model) 0.25 (d) Dynamic (sprung mass model) 2.05 0.5 Axial force (MN) 0.20 Static 1.5 10 15 20 25 30 Time (s) Time (s) 2.5 -0.5 0.0 Dynamic (moving force model) -0.0 0.30 (c) 0.5 0.00 Axial force (MN) Dynamic (sprung mass model) 0.10 0.15 1.10 0 5 10 15 20 25 30 0 5 10 .05 Static -0.

5 0 0.5 Dynamic (1 element/cable) -80 Dynamic (5 elements/cable) Vertical displacement (mm) Bendi -5.5 Static -100 -6.5 -60 ng moment -3.5 -40 -2.5 (MNm) -20 -0.5 -1.5 0 5 10 15 20 25 .5 -4.5 (b) 20 Dynamic (5 elements/cable) Static 1. and axial force (c) in element 52 at node 36.5 Dynamic (1 element/cable) (a) 2. bending moment (b).15 20 25 30 Time (s) Time (s) Figure 10. axial force in cable 6 (d) 40 3.17 Vertical displacement at node 18 (a).

0 Axial force (MN) -0.5 0.2 0.4 Dynamic (5 elements/cable) Static Static 0.5 Dynamic (1 element/cable) Dynamic (1 element/cable) (c) Dynamic (5 elements/cable) 0.30 0 5 10 15 20 25 30 Time (s) Time (s) 0.1 0.0 0.2 -0.1 0.3 (d) 0.2 0 5 10 15 20 25 30 .4 0.2 0.1 Axial force (MN) -0.3 0.6 0.3 -0.1 -0.

axial force in cable 16 (d) 25 25 (a) (b) 15 15 5 5 -5 -5 -15 -15 -25 -25 = 50 v km/h = ξ0 -35 = 90 v km/h -35 = ξ 0.18 Vertical displacement at node 14 (a). axial force in element 36 at node 13 (c).0 5 10 15 20 25 30 Time (s) Time (s) Figure 10. bending moment in pylon element 61 at the attachment point of cable 17.0056 = 110 v km/h -45 ξ Vertical displacement (mm) -45 . (b).

40 1.40 1.01 .005 0.10 1.50 (c) (d) 1.00 50 70 90 110 130 0 0.10 1.20 1.30 1.03 Vertical displacement (mm) -55 -55 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 Time (s) Time (s) 1.00 Dynamic amplification factor Dynamic amplification factor 1.20 1.= 0.30 1.

85 -0.03 Vehicle speed (km/h) Bridge damping ratio.35 0.16 Dynamic (30 mm bump) 0.19 Vertical displacement at node 14 calculated for different vehicle speeds (a) and damping ratios (b).60 Axial force (MN) Dynamic (30 mm bump) -0.02 0.08 -0.0.08 0.00 Dynamic (0 mm bump) -0.015 0.10 0 5 10 .15 Static 0.12 -0.20 0.12 -1.025 0.04 -0.40 (a) Dynamic (0 mm bump) (b) 0.04 Axial force (MN) Static -0. the influence of vehicle speed (c) and damping ratio (d) on some dynamic amplification factors 0.10 0. ξ 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 Axial force in cable 13 Figure 10.

5 Dynamic (30 mm bump) 0 1.3 -20 1.4 -10 1.1 -40 Bending moment in element 52 at node 36 Vertical displacement (mm) Axial force in cable 6 -50 Dynamic amplification factor 1.0 0 5 10 15 20 25 50 70 90 110 .6 (c) Dynamic (0 mm bump) (d) 10 1.2 -30 Axial force in deck element 30 at node 7 1.15 20 25 0 5 10 15 20 25 Time (s) Time (s) 20 1.

truck body vertical displacement (c).20 Axial force in cable 6 (a). 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 Vertical displacement (mm) without TMD -55 0 10 20 30 40 50 60 70 Time (s) 15 (b) 9 3 -3 -9 -15 -21 with TMD -27 Vertical displacement (mm) without TMD -33 0 10 20 .130 Time (s) Vehicle speed (km/h) Figure 10. axial force in deck element 30 at node 7 (b).

21 Vertical displacement at node 14 (a) and node 18 (b). horizontal displacement at node 43 (c).30 40 50 60 70 Time (s) 10 (c) 6 2 -2 -6 with TMD -10 without TMD Horizontal displacement (mm) -14 0 10 20 30 40 50 60 70 Time (s) Figure 10. 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 Vertical displacement (mm) -120 0 10 20 30 40 50 .

20 0.40 (b) )2 0. axial force in cable 24 (c).60 (c) with TMD 0. The dashed vertical line indicate when the truck leaves the bridge – 178 – Effect of tuned vibration absorbers 12 (a) .20 0.20 Vertical acceleration (m/s -0.20 -0.22 Vertical displacement (a) and acceleration (b) of the TMD mass.40 0 10 20 30 40 50 60 70 Time (s) 0.40 without TMD 0.00 -0.00 Axial force (MN) -0.60 70 Time (s) 0.40 0 10 20 30 40 50 60 70 Time (s) Figure 10.

9 2.E-6 1.5 0.E-2 without TMD 2nd 10th 4th 9th .0 0.E-8 0.E+0 1st with TMD (c) 3rd 6th 1.with TMD 8 without TMD 4 0 -4 Bending moment (MNm) -8 0 10 20 30 40 50 60 70 Time (s) 1.3 0.1 0.E+0 1st 3rd (b) with TMD without TMD 5th 9th 1.5 1.4 1.E-2 1.8 1.6 1.3 1.0 Frequency (Hz) 1.0 1.E-4 PSD 1.4 0.6 0.2 1.8 0.7 1.9 1.1 1.2 0.7 0.

180 1.6 0.4 1.1 1.060 1.186 1.9 1.270 .2 1.9 2.131 1.8 0.048 Vertical displacement at node 18 1.1 0.5 1.30 modes all modes nonlinear dynamic linear nonlin.3 1.186 1.7 1.E-8 0.3 0. linear nonlin.187 1.2 0.5th 7th.23 Bending moment in element 52 at node 36 (a).4 0.0 0.0 1.048 1.207 1.7 0.5 0.convergence study simply supported 4 trucks 4 trucks moving number of increments 500 1000 1500 1000 500 1000 1500 girder force model linear dynamic .056 1.E-4 PSD 1. 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 .6 1.208 1.0 Frequency (Hz) Figure 10.E-6 1.8 1.179 1.8th 1. linear nonlin. Vertical displacement at node 14 1.130 1.

119 1.243 1.237 1.112 1.205 .246 1.086 1.065 1.394 1.235 1.265 1.240 1.057 1.160 1.387 1.214 Shear force in element 52 at node 36 1.234 1.199 Axial force in deck element 30 at node 7 1.393 1.286 1.279 1.174 1.212 1.391 1.390 1.360 1.243 1.274 1.094 1.374 1.162 Horizontal displacement of node 43 1.404 1.250 1.364 1.205 1.247 1.1.065 1.230 1.202 1.053 1.164 1.060 1.171 1.

301 1.181 Axial force in element 52 at node 36 1.198 1.271 1.268 1.173 Factors (DAF) Bending moment in element 52 at node 36 1.149 1.087 1.1.284 1.284 1.240 1.238 .222 1.222 1.240 1.154 1.180 1.285 1.442 1.229 1.275 1.202 1.150 1.284 1.431 1.275 1.159 1.166 1.152 1.270 1.290 1.294 1.316 1.207 1.279 1.270 1.188 Axial force in cable 6 1.268 1.268 1.251 1.276 1.210 1.262 1.

211 1.036 1.123 0.071 1.133 1.266 Maximum normalized bridge-vehicle contact force 1.358 1.049 1.041 1.160 0.158 0.232 1.125 0.133 1.039 1.1.242 1.136 0.036 1.121 0.157 1.135 1.035 1.049 1.239 1.345 Dynamic Amplification Axial force in cable 13 1.086 1.039 1.164 1.124 0.234 1.000 .275 0.125 0.073 1.125 1.140 0.126 1.213 Absolute maximum vertical accel.041 1.316 0.336 0.039 1. at node 14 (m/s2) 0.

057 1.03 30 mm v = 110 nonlinear nonlin. nonlin.02 0.003 1.289 1.188 1. nonlin.124 1. nonlin.282 1.01 0.220 1.1. nonlin.204 1.015 0.170 .245 Vertical displacement at node 18 1.087 1. nonlin.060 1.285 1.093 1.000 4 trucks v = 50 v = 70 v = 110 v = 130 ξ= ξ= ξ= ξ= ξ= bump TMD 5 elem/cable km/h km/h km/h km/h 0 0. nonlin.202 1. nonlin.132 1.041 1. Vertical displacement at node 14 1. nonlin.164 1. nonlin.427 1. nonlin.143 1.

086 1.087 1.266 Axial force in element 52 at node 36 1.127 1.272 1.112 1.032 1.176 1.331 1.211 1.205 1.102 1.233 1.329 1.172 1.984 1.131 1.427 1.097 1.1.272 1.189 .414 1.992 0.143 1.065 Axial force in deck element 30 at node 7 1.237 1.010 0.059 1.195 1.413 Shear force in element 52 at node 36 1.102 1.381 1.401 1.100 1.178 1.106 1.230 1.601 1.272 1.378 Horizontal displacement of node 43 1.362 1.372 1.

331 1.300 1.107 1.211 1.205 1.395 1.087 1.082 1.241 1.258 1.461 1.408 1.329 1.168 1.193 1.511 1.008 1.226 1.178 1.463 Factors (DAF) Bending moment in element 52 at node 36 1.024 1.104 1.1.101 1.260 Axial force in cable 6 1.239 1.103 1.323 1.205 Dynamic Amplification Axial force in cable 13 1.222 1.244 1.101 1.241 1.097 1.082 .258 1.202 1.206 1.180 1.104 1.123 1.208 1.268 1.

6 Dynamic amplification factors (DAF).137 Maximum normalized bridge-vehicle contact force 1.316 0. and in the second subsection.039 1. .060 1.099 0.040 1. conclusions concerning the nonlinear finite element modeling of cable supported bridges are presented.145 0.1. 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 ______________________________________________________________________ 11.195 0.1 Conclusions of Part B The conclusions from the study conducted in Part B of this thesis are presented in the following two subsections. at node 14 (m/s2) 0.086 0.141 0. absolute maximum vertical acceleration at node 14.118 0.039 1.446 1.067 0. and maximum normalized bridge-vehicle contact force. In the first subsection.083 Absolute maximum vertical accel.059 1.202 0.091 0.059 Table 10.042 1.039 1.075 0.039 1.063 1.038 1.138 1. conclusions are presented concerning the response due to moving vehicles.024 1.078 1.

However. However. To simplify the data input process when utilizing the multi-element cable discretization. It was found that the catenary cable element is simple to formulate. ABAQUS. as they are highly flexible. and can correctly model the geometric change of the cable at any tension level. Replacing each cable by several catenary cable elements has demonstrated that. This study has shown that the adopted elements are accurate and efficient for nonlinear analysis of cable-stayed and suspension bridges.g. It has been demonstrated that cable supported bridges have a hardening characteristic with respect to the applied load. 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. Furthermore. Moreover.1. guyed masts. 61]. this work has only focused on two-dimensional modeling of cable supported bridges.11. Moreover. linear analysis utilizing the traditional equivalent modulus approach.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. as only few commercial codes. cable motions are also associated with every bending mode of vibration. the iterative process adopted. one drawback is when using commercial finite element codes for analysis. three-dimensional catenary cable and beam elements can be found in [35. a nonlinear static analysis is required to arrive at the deformed dead load tangent stiffness matrix. accurate. is not satisfactory for modern cable-stayed bridges. 64]. This makes the element very attractive. especially for static response calculations. due to the highly nonlinear behavior during the static application of the dead load. 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. with some minor modifications of the cable element matrices this element can also be used for modeling cables for three-dimensional analysis. 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]. e. electric transmission lines. 63. According to the author’s opinion. such as: suspended roofs. in addition to obtaining new pure cable modes of vibration. For such analysis. Modern cable-stayed bridges built today or proposed for future bridges are. . This disadvantage applies also to the one bar element equivalent modulus approach. was found to converge very rapidly. and the author strongly recommends the use of this element. to find the internal force vector and tangent stiffness matrix for the cable element. moored floating bridges. enable the users to define their own elements. etc. the catenary cable element used in this study is also applicable for modeling cables in other types of cable structures [35.

such a device can even increase some of the dynamic amplification factors. when the vehicle is on the bridge). . and tuned vibration absorbers. Moreover. Thus. the following conclusions can be made: • utilizing the tangent stiffness matrix (obtained from a nonlinear static analysis under dead load). 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. 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. 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. Some dynamic amplification factors are very sensitive to bridge damping ratio and the relationship is not always linear. it is recommended to consider the cables motion and modes of vibration in the dynamic analysis by utilizing the multielement cable discretization. “exact” cable behavior. 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. Bridge damping ratios should be carefully estimated to insure more correct and accurate representation of the true dynamic response. such estimation should be based on results from tests on similar bridges. results from many studies of the dynamic response of cable-stayed – 183 – bridges found in the literature are not useful. 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. bridge damping.1. nonlinear geometric effects. To obtain realistic damping ratios. linear static and linear dynamic traffic load analysis of cable supported bridges give sufficiently accurate results from the engineering point of view. two approaches were implemented: one for evaluating the linear dynamic response and one for the nonlinear dynamic response. road surface roughness. Based on this investigation of the traffic load response of cable-stayed and suspension bridges. Unfortunately. vehicle speed. Further.e.11. cable modeling.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. In fact. This investigation has mainly focused on comparing linear and nonlinear traffic load dynamic responses and also on the effect of bridgevehicle interaction. For the analysis of the dynamic response. were considered. 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. and realistically – 182 – estimated bridge damping.

the designer should consider installation of cable dampers especially for the shortest cables to increase the fatigue life of the cables.e. • 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. 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. the performance of a tuned mass damper (TMD.2 Suggestions for further research Based on the performed investigation. modify the TMD properties to improve its efficiency. For this bridge. 11. 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. . Moreover. are different from the original characteristics assumed during the optimal design of the TMD). a superior solution can be obtained by using a so-called active tuned mass damper (ATMD).e.However. when needed. For the studied cable-stayed bridge. even if maintenance of the road surface is made regularly. 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. sensors attached to the bridge deck. should be investigated. • Extensive testing on a cable supported bridge should be performed to assess the validity of the analysis methods and the theoretical findings.g. As discussed earlier in section 8. The computer continuously monitors the dynamic characteristics of the bridge using e.4. – 184 – • Further work is needed to study the effect of using finer models (i. 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. Thus. for future research it is suggested to use simulated trains of moving traffic based on collected statistical traffic data. as this could not be accomplished in this study due to time limitation. 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. Such a damper comprises computer controlled servo-hydraulic actuators that can.30. This situation should be considered in the design practice of such bridges. passive device) significantly deteriorates when the dynamic characteristics of the bridge changes (i. • 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.

# the catenary cable element.g. used to generate the Fortran code for the elements presented in Chapter 7. Such bridges can be referred to as smart bridges as they have the capability of modifying their behavior during the dynamic loading. 62.g. Furthermore. – 186 – Appendix ______________________________________________________________________ Maple Procedures ______________________________________________________________________ The Maple procedures. computer controlling the tension in some cables. 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. For cable-stayed bridges real time vibration control can be achieved by e. 67]. in high-speed trains like the Swedish train X2000 to improve riding comfort.g. whereas e. the pretensioning force in the cables is changed. since some people are perhaps not ready to rely on computers when crossing a bridge. Each comment line starts with the symbol #. with(linalg). .Active control of structures using cables was proposed by Freyssinet as early as 1960 [67]. e. excellent literature review and state-of-the-art review on control and monitoring of civil engineering structures are found in [30. they can only be economical for long span bridges where they can induce big saving in construction material. in order to counteract – 185 – traffic loads at any time. active control is applied in advanced airplanes for suppression of aerodynamic instability. are also considered for new cable supported bridges with very long spans such as for the Messina crossing and the Gibraltar crossing.1 Cable element # Tangent stiffness matrix Kt for. It is believed that. are given below. readlib(fortran). active modification of bridge deck edge shape to enhance resistance to aerodynamic instabilities like flutter. A. For the interested reader. the stability of the bridge have to rely entirely on the bridge structure itself. Till now analysis and application of active vibration control of structures excited by moving loads have attracted limited research efforts. Ly:=1/(2*E*A*w)*(Tj^2-Ti^2)+(Tj-Ti)/w. and in modern cars like the Mercedes A-class to improve stability. 69]. Today. Recent studies describing active control of bridges are presented in [5. active control should as a first step only be used to improve serviceability aspects such as riding comfort. Active controls. as the cost of such active systems is high. Such control system is based on the idea of constantly monitoring movements of the bridge using attached sensors and via computer controlled tensioning jacks. so-called active cables.

-k4]): fortran(Kt.P2): f:=matrix(2. wx:=(u5-u2)/L.u4.1]: k2:=k[1.2 Beam element # Internal force vector p and.u6]).[f11.-k2.u6]).P1): f12:=diff(Lx. – 188 – ______________________________________________________________________ .k2.optimized): A.f12.u3. fortran(p.-k2.k1. Ti:=sqrt(P1^2+P2^2).-k1.k2. PIe:=1/2*L*E*A*e^2.k4.optimized). P4:=w*Lu-P2. Tj:=sqrt(P3^2+P4^2). PI:=PIe+PIg+PIk.2]: k4:=k[2.-k2.P1): f22:=diff(Ly. # the beam element. PIg:=1/2*L*G*A*g^2.-k4.k2.[-k1. ux:=(u4-u1)/L.2.u5.-k2. f11:=diff(Lx.u3.f22]): k:=inverse(f): k1:=k[1. p:=grad(PI.u2. k:=tx.k1. readlib(fortran). # tangent stiffness matrix Kt for. fortran(Kt.Lx:=-P1*(Lu/E/A+1/w*ln((P4+Tj)/(Ti-P2))).[u1.k4. tx:=(u6-u3)/L. e:=(1+ux)*cos(t)+wx*sin(t)-1. g:=wx*cos(t)-(1+ux)*sin(t).u2.2]: – 187 – Kt:=matrix(4. P3:=-P1. PIk:=1/2*L*E*I*k^2.[u1.4.u4.k2.u5. t:=(u3+u6)/2.f21. with(linalg).P2): f21:=diff(Ly.optimized). Kt:=hessian(PI.

Bibliography of Part B ______________________________________________________________________ [1] ABAQUS User’s Manual... 1985.S. . ‘ Modeling the Nonlinear Seismic Behavior of Cable-Stayed Bridges with Passive Control Bearings’.J. Conference on Cable-Stayed and Suspension Bridges. Thomas Telford. Amsterdam. 1994..A. Prentice Hall. Struct. pp. J.. 1985. J. France. 1994. 1996. 218-226.. Inc. Eng. 195-210. Cambridge..P.A. 1991.. 2. pp.. 1991. 303-313. 54. pp. pp. 123. 1997. 20. Saleh A. ASCE. Computers and Structures. Karlsson & Sorensen. 117.. . Providence. Dynamics of Structures. 267-277. 486506. Mlejnek H. J. North-Holland. ‘ Nonlinear Analysis of Cable-Stayed Bridges’..M. [8] Argyris J. Computers and Structures . pp. 2571-2589. Vol. Meccanica.. Khalifa M.. [3] Abdel-Ghaffar A. 1997. 20-35. pp. Structural Dynamics for Engineers. ASCE. [10] Bachmann H. 5. J.. – 189 – [11] Bathe K. No. Int. 31-36. [13] Bruno D. 461-492. pp.. Mechanics.. ASCE. 1994. Weber B. ‘ Fully Nonlinear Analysis of Composite Girder CableStayed Bridges’. ‘ Optimal Control of Adaptive/Smart Bridge Structures’. Eng... Hibbitt. 1. Struct. Proc. [14] Buchholdt H. 1995. 120. [2] Abbas S.. Cambridge University Press. 54. Int... ‘ Importance of Cable Vibration in Dynamics of Cable-Stayed Bridges’.. Zhang J.A. 3456-3476...M. Miyata T. Grimaldi A.. Scordelis A. Deauville... . ‘ Nonlinear Behaviour of Long-Span Cable-Stayed Bridges’. ‘ Wind-Induced Nonlinear LateralTorsional Buckling of Cable-Stayed Bridges’. 1995. Eng. Struct. pp. New Jersey. [12] Boonyapinyo V. 1995. [15] Buchholdt H. London. [5] Adeli H.. An Introduction to Cable Roof Structures. Rhode Island.M. ‘ Vehicle Dyamics of Commercial Vehicles’. [7] Ali H. Finite Element Procedures. pp. 1991. Yamada H. Eng.. ‘Tuned Vibration Absorbers for “Lively” Structures’. [9] Aurell J. Eng. pp. Oct. Struct. Volvo Technology Report. [6] Adeli H. [4] Abdel-Ghaffar A. 117. Abdel-Ghaffar A. Vol.M. ASCE. Edlund S. Nazmy A.. 1990. ‘ 3-D Nonlinear Seismic Behavior of CableStayed Bridges’.

No.P. Dey S..F. . [27] Gimsing N. [24] Frýba L. [29] Green M. 1977.. Methods in Eng. – 190 – [23] Forsell K. Vol. ‘ Static and Dynamic Analysis of Guy Cables’. Mechanics. pp. Chassiakos A. Sept.. 1-21.K. [28] Gimsing N. 1997.T. Stockholm. and Future’. Struct. [17] Crisfield M. 745-750.. pp. Proc. ‘ Dynamic Behaviour of a Cable-Stayed Bridge’. TRITA-BKN Bulletin 34. Yao J.A. Royal Institute of Technology. 1992.. Oct. Conference on Cable-Stayed and Suspension Bridges. ASCE. ‘ Dynamic Interaction Between Heavy Vehicles and Highway Bridges’. Eng. ‘ Wind Effects on the Normandie CableStayed Bridge: Comparison Between Full Aeroelastic Model Tests and Quasi-Steady Analytical Approach’. 62. J.L.O. Mechanical Vibrations. Gaughey T.. 1699-1718. Chichester. [20] Dean D. pp. Vol. COWIconsult A/S. 1997. [18] Das A.W. Cebon D. Batchelor B. [22] Fleming J. .E.J. ‘ The Structural Damping of Long Span Bridges: An Interpretation of Observations’. No.T.... 2. J.. Tanaka H . Eng. [21] Den Hartog J. Masri S. Dept. Division.. Jacob J. Cable Supported Bridges. 43. Spencer B.P. .. of Struct.S.. Deauville. 2. . Dynamics.L.. Computer & Structures. 253-264... Claus R. .. Licentiate Thesis. ‘ Effect of Tuned Mass Dampers on Random Response of Bridges’. Soong T.. and Struct. Skelton R.G.A. Earthquake Eng. Egeseli E. McGraw-Hill. Rixen D.. 1989. Int. ‘Dynamic Analyses of Static Instability Phenomena’. 1994. Groningen. 1956. Bergman L.. 1997. 1991. Grillaud G. pp. 4th edition. Present. Int. J. 87. ‘ Structural Control: Past. and Esdahl S. [25] Gambhir M. Personal communication. France. Mechanical Vibrations.. . [26] Géradin M. Computer & Structures. Vol. Canada-Japan Workshop on Aerodynamics. second edition. pp. Wiley. 1997.. ‘ A Finite Element for 3-D Prestressed Cablenets’. Wiley.. Cohen N.. pp.F. [30] Housner G. 1980..F... pp. 11. Technical University of Denmark. . 1972. Danish Maritime Institute. 1997. Non-linear Finite Element Analysis of Solids and Structures. 81-90. Chichester. 1-16. Numer. 1994. [19] Davenport A. ASCE ..K. Wiley. New York.[16] Conti E. Chichester. Ottawa. Vibration of Solids and Structures under Moving Loads. Larose G. 123.. Noordhoff International Publishing. 1961.F. 4. 8..A.J..

. J. ASCE. . Proc. Bangkok. Eng. ‘ Field Measurements of the WindInduced Response of a 254 m High Free-Standing Bridge Pylon’. pp.J. Larose G. [45] Kwon H.. Proc. June 1997.. 87-92. . Eng.W. pp. Vol.. Public Works Res.. ‘ Vibration Absorbers and Bridges’. Yiu P. Petersen A. Eng. Jacobsen A. ‘ Vibration Control of Bridges Under Moving Loads’.. Struct. Knudson W. pp.V.. Pretlove A. [39] Judd B.. 4.. [42] Karoumi R. Computer & Structures. Struct. [41] Kanok-Nukulchai W. Vol. Oct. 1992. 1994.K. pp. pp.C. [46] Larose G.. 119. 27-37. No.. Zasso A. Tsunomoto M. Cambridge. 1996. Vol. 1992. 1. Cable Structures. Mechanics Div.. ‘ A Curved Element for the Analysis of Cable Structures’.. 325-333. Ham H. No.M.. 351-362. Melelli S.. Inst. . ‘ Mathamatical Modelling of Cable-Stayed Bridges’. Int.. 1981.. [44] Kawashima K. J.. Eng.. Hong G.S. Lee I. 26. Mechanics Div. Vol. 1981. 26. – 191 – [36] Jensen G. Int. ‘ Dynamic Response of Cable-Stayed Bridges Subjected to Moving Vehicles’. 1993. Struct.897-971. J.J. Conference on Cable-Stayed and Suspension Bridges. Stockholm. 14. 1998. Italy.. [43] Kawashima K... Royal Institute of Technology. 27. MIT Press. pp.. [33] Indrawan B. 108-113.. TRITA-BKN Report 49. April 1993. Eng. ‘ Aerodynamic Design of the Great Belt East Bridge’. 1979.. ‘ Damping Characteristics of Cable Stayed Bridges for Seismic Design’. Unjoh S.. Wheen R. March 1989. 567-575.J.. ASCE. Dec. 2-9. No.. ‘ Erection of Suspension Bridges’.C.. pp. pp. Brotton D.M. 4. [31] Huddleston J. 473-480. Construct. of the 2nd European African Conference on Wind Engineering. J. 1978.B. 2. [38] Jones R.. ‘ Computer Analysis of Extensible Cables’..T. 249-266. Deauville. 1015-1031. ‘ Poisson Effect in Extensible Cables with Both Ends Fixed’. pp. 1590-1595. 104. 66.A. J. ‘ Nonlinear Cable Behavior’. France. Master Thesis... Division . ‘ Field Measurements of the Dynamic Response of the Höga Kusten Bridge During Construction’. – 192 – [47] Larsen A. Kim M. ‘ Nonlinear Modelling of Cable-Stayed Bridges’. 120. pp. [37] Johnson R.J. 1994. pp... J. Res. Genova. [40] Kanok-Nukulchai W. ‘ Vibrations of a Cable-Stayed Bridge due to Vehicle Moving over Rough Surface’. Dept. ‘ Estimation of Damping Ratio of Cable-Stayed Bridges for Seismic Design’. 1-8. 2. Casanova D. The Highway Engineer. of Struct.. Unjoh S. Asian Institute of Technology. IABSE 15th Congress. 1991. 107.V. [35] Jayaraman H.L. ASCE .C.. Steel Research . Tsunomoto M. Computers and Structures. ASCE .. [34] Irvine H. pp. 1998. Denmark. [32] Huddleston J...

pp. 1977. ‘ Manufacture and Erection of the Steel Main Span’. pp.. 1992. pp. [62] Patten W.. Division. 15. – 193 – [60] Ozdemir H. 144. Vol. [59] Ostenfeld K. 1997. [61] Pacoste C. [64] Peyrot A. Solids Structures.T. ‘ General Solution of Suspended Cable Problems’. [51] Madsen B. [65] Raoof . 93. 122.. 2. Eng. 1st Int. Computers and Structures. ASCE. Methods Appl.W. 691-697. pp. 1994. Deauville. No. J. 1988.N. Comput.H. Symposium on Aerodynamics of Large Bridges. 103.L. ‘ Beam Elements in Instability Problems’. pp. J. Symposium on Aerodynamics of Large Bridges. J. Division. and Industrial Aerodynamics. Struct. Denmark. Division.. pp. Int. 1979. ASCE. 6·97. 90. Int. [55] Nazmy A. 2766-2772. Mech. pp. J. 26-31.. Division. ‘ Tuned Mass Dampers for Buildings’.... . Oct. Sorensen L. [56] O'Brien T.. Goulois A. ASCE.. ‘ A Finite Element Approach for Cable Problems’.. Struct. 1967.. 1964. 763-779. 261-285. [52] MAPLE V Language Reference Manual. Eng. 34. 257-271. New York. Proc. 3-22. ‘ Controlled Semiactive Hydraulic Vibration Absorber for Bridges’. J. [57] O'Brien T.. Wind Eng. Tension Structures. 1st Int. [50] Luft R.. [53] MATLAB Using Matlab Version 5. Copenhagen. McGraw-Hill. [54] McNamara R. 105. Vol. 1979. pp. J. Vol. Springer-Verlag.W. The MathWorks Inc. Dec... 1785-1798. ST9. ASCE. ASCE. Division. Abdel-Ghaffar A. 427-437. Proc. 48. Struct. 1. He Q. pp.J. J.S.S. Denmark. 1979. ‘ Aerodynamic Aspects of the Final Design of the 1624 m Suspension Bridge Across the Great Belt’. Computers and Structures.M. 163-197. 1993.M. France. 187-192.. ST12. Sack R. [58] Ostenfeld K.. ‘ Cable Movements Under Two-Dimensional Loads’. Elsevier. [63] Peyrot A.H. No. 1978.. Struct. Struct. New York. 805-813. pp.. 104. 1996. Struct. Feb. Swedish Society of Civil Eng. Vol.H. ASCE. Väg. 10. J.Proc. Copenhagen.. Feb. Francis A. 1990. J. No. [48] Larsen A. 89-123. 1-26.. pp.. Natick.J. pp. [49] Leonard J. 1991. ‘ Optimal Tuned Mass Dampers for Buildings’.och Vatten-byggaren. ‘ Analysis of Cable Structures’.H. Goulois A. 1992. Feb. ‘ Bridge Engineering and Aerodynamics’. 1996.. ‘ Analysis of Flexible Transmission Lines’. pp. Conference on Cable-Stayed and Suspension Bridges. pp. Eriksson A. Feb. 269-283.. ‘ Long Span Bridges: State-of-the-art’.M.. pp.. Larsen A. ‘ Three-Dimensional Nonlinear Static Analysis of Cable-Stayed Bridges’.

pp.M.D. 74-84. ‘ Tuned Mass Dampers to Control Floor Vibration from Humans’. 1994. . London Docklands’. – 194 – . [69] Wells M.. Moïa P..T. 118. [66] Setareh M.. No. 2548-2550.. pp. Thomas Telford.. ASCE. ASCE.. Houriet B. March 1992. Constructional Steel Research. 10. Eng. 1998. Vol. 741762. ‘ Estimation of Damping Ratio of Cable-Stayed Bridges for Seismic Design’. Struct. pp. ‘ Active Structural Control: an Appraisal for a Practical Application. Vol. J.. 59. ‘ State-of-the-art Review: Active Structural Control in Civil Engineering ’. Aug. The Royal Victoria Footbridge. J. Hanson R. 8. paper no. Struct. [68] Walther R. Eng. Isler W. Struct. London... 1988. [67] Soong T. 46:1-3. J. No.. 3. Eng. Cable Stayed Bridges. 120. 1988..