Sie sind auf Seite 1von 203

Arbeitsbereich Baustatik und Stahlbau

Berechnungsstrategien
zur sicheren und formtreuen
Errichtung von Schrägseilbrücken
In englischer Sprache

Diplomarbeit
zur Erlangung des Grades einer Diplom-Ingenieurin
der Fachrichtung Bauingenieurwesen und Umwelttechnik
von
Svenja Mueller-Haagen
Matrikel-Nr. 14104

Betreuer: Prof. Dr.-Ing. Uwe Starossek und Dipl.-Ing. Thomas Löhning

Eingereicht am 24. Mai 2005


TUHH, AB 3-08, D-21071 Hamburg Univ.-Prof. Dr.-Ing. Uwe Starossek

Arbeitsbereich 3-08
Baustatik und Stahlbau
Sta, Löh 02. März 2005

Diplomarbeit Frau Svenja Mueller-Haagen


Berechnungsstrategien zur sicheren und formtreuen Errichtung
von Schrägseilbrücken
Schrägseilbrücken gelten für Spannweiten von 100m bis 1000m als eine wirtschaftliche
und ästhetische Brückenform. Sie weisen eine komplexe und hochgradig statisch unbe-
stimmte Tragstruktur auf, bei deren Entwurf und Bemessung insbesondere die auftreten-
den Bauzustände zu beachten sind. Die Abmessungen und die verwendeten Materialien
des Tragwerks können eine Berücksichtigung geometrischer und physikalischer Nichtli-
nearitäten erforderlich machen.
Aufgrund dieser Komplexität sind beim Bau der Brücke Abweichungen gegenüber der
Berechnung zu erwarten. Abweichungen außerhalb von Toleranzgrenzen führen zu er-
heblichen Änderungen der Geometrie und Schnittgrößen, die nachfolgend zu korrigieren
sind.
Im Rahmen dieser Arbeit soll an einem einfachen Beispiel die oben angedeutete Problem-
stellung bei der Berechnung von Schrägseilbrücken untersucht werden. Details werden
nach Absprache mit den Betreuern festgelegt.

Aufgabenstellung

• Darstellung der Grundlagen zur Berechnung von Schrägseilbrücken unter besonde-


rer Berücksichtigung der Optimierung der Spannstrategie und des zeitabhängigen
Materialverhaltens

• Berechnungen und Vergleich der Ergebnisse für ein sinnvoll gewähltes System unter
Anwendung verschiedener Programme

• Entwicklung und Darstellung von Korrekturstrategien zum Ausgleich von Abwei-


chungen während des Bauablaufs und Anwendung auf das oben gewählte System

Betreuer:
Prof. Dr.-Ing. Uwe Starossek
Dipl.-Ing. Thomas Löhning

Postanschrift Telefon E-mail


TUHH, AB 3-08 (+49-40) 428 78-3223 sekretariat bs@tuhh.de
D-21071 Hamburg

Dipl.-Ing. Thomas Löhning (+49-40) 428 78-2435 loehning@tu-harburg.de

Besucheranschrift Fax Home Page


Denickestraße 17 (+49-40) 428 78-2585 www.sh.tu-harburg.de
D-21073 Hamburg
Erklärung
Ich versichere, dass ich die vorliegende Arbeit selbstständig verfasst und keine anderen als die
angegebenen Quellen und Hilfsmittel benutzt habe. Die Arbeit hat in gleicher oder ähnlicher
Form noch keiner anderen Prüfungsbehörde vorgelegen. Ich stimme der Weitergabe meiner
Arbeit zu wissenschaftlichen Zwecken zu.

Hamburg, den 24.05.2005 .................................................


Svenja Mueller-Haagen
Abstract
Due to their aesthetic appearance and structural efficiency, cable-stayed bridges are increasingly
designed in recent decades. They are usually an economical choice for spans between 200m
and about 1000m. Being mostly built by the cantilevering method, cable-stayed bridges have
to be analyzed separately in every construction stage, taking into account many load cases. In
addition, the forces in the cable stays have to be determined and possibly changed during con-
struction so as to obtain the desired deck and pylon condition at the time of bridge completion.

The determination of the initial cable forces for a given final dead load condition is an important
but difficult task in the design of cable-stayed bridges. Not only the construction sequence has
to be taken into account, but also time-dependent effects, such as creep and shrinkage play an
important role in the structural behavior of cable-stayed bridges. Moreover, in case of longer
spans, nonlinear effects can no longer be neglected and should be included in the analysis.

A second major issue during the construction of cable-stayed bridges is the continuous control
of geometry and internal forces. Due to inevitable errors between the structural design parame-
ters and the actual ones, unexpected discrepancies between the predicted and the actual state of
the structure in a given construction stage may occur. These discrepancies need to be monitored
and adjusted so as to achieve the desired final condition.

The present work concerns the analysis and control of cable-stayed bridges during construc-
tion by the cantilevering method. A procedure to determine an optimum tensioning strategy
is presented that allows for the consideration of several effects that are relevant for the design
of cable-stayed bridges, including the construction sequence, second-order theory, large dis-
placements, cable sag and time-dependent effects such as creep and shrinkage. Furthermore, an
economically efficient technique to control discrepancies between the predicted and actual state
of a cable-stayed bridge by the adjustment of cable tension forces and deck elevations during
the erection process is formulated. Both methods are demonstrated in a sample calculation of
a simple cable-stayed bridge, which is performed by the dint of standard structural engineering
software programs developed for the analysis and design of bridge structures. The results of
this sample calculation show that both methods may well be used to ensure safety and geometry
accuracy during the erection of cable-stayed bridges.
Contents

1 Introduction and Overview 1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Objective and scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.3 Thesis overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Cable-Stayed Bridges 6

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.2 Short historical review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.3 Erection of cable-stayed bridges . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.3.1 Construction of cable-stayed bridges by the cantilevering method . . . 9

2.3.1.1 Cast-in-place construction . . . . . . . . . . . . . . . . . . . 11

2.3.1.2 Precast Construction . . . . . . . . . . . . . . . . . . . . . . 12

2.3.1.3 Construction of steel and composite decks . . . . . . . . . . 12

3 Structural Analysis of Cable-Stayed Bridges 13

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3.2 Review of existing methods to determine the initial cable forces in cable-stayed
bridges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3.2.1 Optimization method . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.2.2 Zero displacement method . . . . . . . . . . . . . . . . . . . . . . . . 17

3.2.3 Force equilibrium method . . . . . . . . . . . . . . . . . . . . . . . . 18

3.2.4 Summary of existing methods . . . . . . . . . . . . . . . . . . . . . . 20


CONTENTS II

3.2.4.1 Consideration of non-linearity effects . . . . . . . . . . . . . 20

3.2.4.2 Consideration of construction process and time-dependent ef-


fects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3.3 Expanded unit load method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3.3.1 Construction process . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3.3.2 Second-order theory and large deflections . . . . . . . . . . . . . . . . 27

3.3.2.1 Second order theory . . . . . . . . . . . . . . . . . . . . . . 27

3.3.2.2 Large displacement effects . . . . . . . . . . . . . . . . . . 29

3.3.2.3 Consideration in the ”expanded unit load method” . . . . . . 30

3.3.3 Cable sag effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3.3.4 Time-dependent material behavior . . . . . . . . . . . . . . . . . . . . 37

3.3.4.1 Creep of concrete . . . . . . . . . . . . . . . . . . . . . . . 37

3.3.4.2 Shrinkage . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

3.3.4.3 Relaxation of prestressed steel . . . . . . . . . . . . . . . . 43

3.3.4.4 Concrete aging . . . . . . . . . . . . . . . . . . . . . . . . . 44

3.3.4.5 Mathematical modeling of time dependent effects . . . . . . 45

3.3.4.6 Prediction of creep and shrinkage by the CEB-FIB 1990 model


code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

3.3.4.7 Simulation of creep and shrinkage in the expanded unit load


method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

3.3.5 Implementation of the expanded unit load method into finite element
software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

3.3.5.1 Optimization for non-linear problems . . . . . . . . . . . . . 59

3.4 Camber control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

4 Comparison of Structural Analysis Programs 64

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

4.2 General description of Larsa2000 and MIDAS/Civil . . . . . . . . . . . . . . . 65

4.2.1 Larsa2000 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
CONTENTS III

4.2.2 MIDAS/Civil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

4.3 Construction stage analysis by Larsa2000 and MIDAS . . . . . . . . . . . . . 67

4.3.1 Definition of construction stages . . . . . . . . . . . . . . . . . . . . . 68

4.3.1.1 Larsa2000 . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

4.3.1.2 MIDAS/Civil . . . . . . . . . . . . . . . . . . . . . . . . . 69

4.3.2 Accounting for deformation . . . . . . . . . . . . . . . . . . . . . . . 71

4.3.3 Consideration of nonlinearities . . . . . . . . . . . . . . . . . . . . . . 72

4.3.3.1 Nonlinear construction stage analysis by Larsa2000 . . . . . 73

4.3.3.2 Nonlinear construction stage analysis by MIDAS/Civil . . . 74

4.3.4 Consideration of time-dependent material behavior . . . . . . . . . . . 74

4.3.4.1 Time-dependent construction stage analysis by Larsa2000 . . 74

4.3.4.2 Time-dependent construction stage analysis by MIDAS/Civil 76

4.3.5 Running a construction stage analysis . . . . . . . . . . . . . . . . . . 77

4.3.5.1 Running a construction stage analysis by Larsa2000 . . . . . 77

4.3.5.2 Running a construction stage analysis by MIDAS/Civil . . . 78

4.3.6 Solution for unknown loads using optimization technique . . . . . . . . 79

4.3.6.1 The unknown load factor function in MIDAS/Civil . . . . . . 79

4.3.6.2 Calculation of unknown load factors in Larsa2000 . . . . . . 81

4.3.6.3 Consideration of time-dependent effects in the optimization


process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

4.4 Comparison of a time-dependent construction stage analysis by Larsa2000 and


MIDAS/Civil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

4.4.1 Linear construction stage calculation . . . . . . . . . . . . . . . . . . . 84

4.4.2 Construction stage calculation considering creep effects . . . . . . . . 85

4.4.2.1 Analysis by Larsa2000 and MIDAS/Civil . . . . . . . . . . . 85

4.4.2.2 Hand Calculations . . . . . . . . . . . . . . . . . . . . . . . 85

4.4.3 Construction stage calculation considering shrinkage strain . . . . . . . 90

4.4.4 Evaluation of the time-dependent construction stage analysis results . . 90


CONTENTS IV

5 Construction Control and Monitoring 93

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

5.2 Causes of discrepancies between the predicted and actual state of a cable-stayed
bridge during construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

5.3 Supplementary measurement of structural parameters on site . . . . . . . . . . 95

5.4 Construction control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

5.4.1 Simulation analysis of the construction process . . . . . . . . . . . . . 96

5.4.2 Field measurements during erection . . . . . . . . . . . . . . . . . . . 96

5.4.3 Comparison of the theoretical and actual results and forecast of the fu-
ture evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

5.4.4 Adjustment of deck elevations and cable forces during construction . . 101

5.4.4.1 Procedure for optimum cable force and deck elevation adjust-
ment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

6 Example Calculation of a Cable-Stayed Bridge 106

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

6.2 Example model dimensions and properties . . . . . . . . . . . . . . . . . . . . 107

6.3 Restrictions for the solution for unknown load factors . . . . . . . . . . . . . . 108

6.4 Calculation of the initial cable forces in a linear static analysis . . . . . . . . . 111

6.4.1 Calculation by MIDAS/Civil . . . . . . . . . . . . . . . . . . . . . . . 111

6.4.2 Calculation by Larsa2000 . . . . . . . . . . . . . . . . . . . . . . . . 114

6.4.3 Summary of the determination of the initial cable forces in a linear static
analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

6.5 Calculation of the initial cable forces and the jacking distance in a construction
stage analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

6.5.1 Calculation of the cable forces at the time of installation and the jacking
distance in a time-independent construction stage analysis . . . . . . . 122

6.5.1.1 Calculation by MIDAS/Civil . . . . . . . . . . . . . . . . . 122

6.5.1.2 Calculation by Larsa2000 . . . . . . . . . . . . . . . . . . . 125


CONTENTS V

6.5.1.3 Summary of the determination of the cable forces and jacking


distance in a time-independent construction stage analysis . . 130

6.5.2 Calculation of the cable forces at the time of installation and the jacking
distance in a time-dependent construction stage analysis . . . . . . . . 131

6.5.2.1 Calculation by MIDAS/Civil . . . . . . . . . . . . . . . . . 132

6.5.2.2 Calculation by Larsa2000 . . . . . . . . . . . . . . . . . . . 138

6.5.2.3 Summary of the determination of the cable forces and jacking


distance in a time-dependent construction stage analysis . . . 144

6.5.3 Summary of the calculation of the initial cable forces and jacking dis-
tance in a construction stage analyis . . . . . . . . . . . . . . . . . . . 146

6.6 Performance of a nonlinear time-dependent construction stage analysis . . . . . 148

6.7 Amendment of discrepancies during the erection process . . . . . . . . . . . . 149

6.7.1 Simulation analysis of the construction process . . . . . . . . . . . . . 149

6.7.2 Field measurements during erection . . . . . . . . . . . . . . . . . . . 152

6.7.3 Stage by stage construction control . . . . . . . . . . . . . . . . . . . 152

6.7.3.1 Construction stage CS01 and CS02 . . . . . . . . . . . . . . 153

6.7.3.2 Construction stage CS03 . . . . . . . . . . . . . . . . . . . 159

6.7.3.3 Construction stages CS04 and CS05 . . . . . . . . . . . . . 163

6.7.3.4 Construction stages CS06 to CS09 . . . . . . . . . . . . . . 166

6.7.3.5 Construction stages CS10 to CS14 . . . . . . . . . . . . . . 168

6.7.4 Results of the construction control . . . . . . . . . . . . . . . . . . . . 169

6.7.5 Summary of construction control . . . . . . . . . . . . . . . . . . . . . 172

7 Summary and Conclusions 176

7.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176

7.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178


List of Figures

2.1 Construction by the cantilevering method using precast segments . . . . . . . . 10

3.1 Discontinuity between two segments [37] . . . . . . . . . . . . . . . . . . . . 23

3.2 Unit load cases and desired moment distribution [21] . . . . . . . . . . . . . . 24

3.3 Unit load cases considering the construction process [4] . . . . . . . . . . . . . 26

3.4 Column behaviors due to P-Delta effects [30] . . . . . . . . . . . . . . . . . . 28

3.5 Cable element with sag [52] . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.6 Force deflection curve for a cable [15] . . . . . . . . . . . . . . . . . . . . . . 31

3.7 Definition of the tangent and secant modulus of elasticity of a cable [15] . . . . 32

3.8 Catenary cable element [22] . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

3.9 Creep of concrete under the effect of sustained stress [14] . . . . . . . . . . . . 39

3.10 Definition of mechanical and aging strain . . . . . . . . . . . . . . . . . . . . 44

3.11 Time-dependent concrete deformations . . . . . . . . . . . . . . . . . . . . . . 46

3.12 Creep isochrones [2] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

3.13 Typical creep curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

3.14 Time-dependent concrete deformations [2] . . . . . . . . . . . . . . . . . . . . 48

3.15 Linear variation of the elastic strain during a time interval . . . . . . . . . . . . 57

3.16 Flowchart of the iterative solution for nonlinear optimization (AddCon-Method) 62

4.1 Different segmental construction methods in Larsa2000 . . . . . . . . . . . . . 72

4.2 Flowchart for the calculation of initial cable forces in a linear analysis . . . . . 80

4.3 Flowchart for the calculation of initial cable forces in a construction stage analysis 81

4.4 Construction sequence of the verification model for creep and shrinkage . . . . 83
LIST OF FIGURES VII

4.5 Elastic vertical deflection and rotation of a cantilever . . . . . . . . . . . . . . 86

4.6 Curvature diagram of a cantilever under a uniform unit load . . . . . . . . . . . 91

4.7 Creep curvature diagram of a cantilever under a uniform unit load . . . . . . . 91

6.1 Configuration of the final stage of an asymmetrical cable-stayed bridge . . . . . 107

6.2 Bending moment distribution of an equivalent continuous beam [kNm] . . . . . 110

6.3 Bending moment distribution under dead load and unit cable forces [kNm] . . . 111

6.4 Vertical deformation under dead load and unit cable forces [mm] . . . . . . . . 111

6.5 Moment distribution when restricting bending moments [kNm], Case I . . . . . 112

6.6 Vertical deflection when restricting bending moments [mm], Case I . . . . . . . 112

6.7 Moment distribution when restricting deformations [kNm], Case II . . . . . . . 113

6.8 Vertical deflection when restricting deformations [mm], Case II . . . . . . . . . 113

6.9 Bending moment distribution under dead load and unit cable forces [kNm] . . . 114

6.10 Vertical deformation under dead load and unit cable forces [mm] . . . . . . . . 114

6.11 Moment distribution when restricting bending moments [kNm], Case I . . . . . 116

6.12 Vertical deflection when restricting bending moments [mm], Case I . . . . . . . 117

6.13 Moment distribution when restricting deformations [kNm], Case II . . . . . . . 117

6.14 Vertical deflection when restricting deformations [mm], Case II . . . . . . . . . 117

6.15 Construction sequence (construction stages 1+2) . . . . . . . . . . . . . . . . 118

6.16 Construction sequence (construction stages 3-9) . . . . . . . . . . . . . . . . . 119

6.17 Construction sequence (construction stages 10-13) . . . . . . . . . . . . . . . 120

6.18 Construction sequence (construction stage 14) . . . . . . . . . . . . . . . . . . 121

6.19 Bending moment distribution under dead load and unit forces [kNm], CS14 . . 122

6.20 Vertical deformation under dead load and unit forces [mm], CS14 . . . . . . . 123

6.21 Moment distribution in CS14 when restricting bending moments [kNm], Case I 123

6.22 Vertical deflection in CS14 when restricting bending moments [mm], Case I . . 124

6.23 Moment distribution in CS14 when restricting deformations [kNm], Case II . . 124

6.24 Vertical deflection in CS14 when restricting deformations [mm], Case II . . . . 125

6.25 Bending moment distribution under dead load and unit forces [kNm], CS14 . . 125
LIST OF FIGURES VIII

6.26 Vertical deformation under dead load and unit forces [mm], CS14 . . . . . . . 126

6.27 Moment distribution in CS14 when restricting bending moments [kNm], Case I 128

6.28 Vertical deflection in CS14 when restricting bending moments [mm], Case I . . 129

6.29 Moment distribution in CS14 when restricting deformations [kNm], Case II . . 129

6.30 Vertical deformation in CS14 when restricting deformations [mm], Case II . . . 129

6.31 Vertical deformation in CS14 when the initial tangent displacement for erected
structures option is not activated [mm], Case II (MIDAS) . . . . . . . . . . . . 131

6.32 Bending moment distribution under dead load and unit forces at day 89 [kNm] . 132

6.33 Vertical deformation under dead load and unit forces at day 89 [mm] . . . . . . 132

6.34 Moment distribution at day 89 when restricting bending moments [kNm], Case I 134

6.35 Vertical deflection at day 89 when restricting bending moments [mm], Case I . 135

6.36 Moment distribution at day 89 when restricting deformations [kNm], Case II . . 135

6.37 Vertical deformation at day 89 when restricting deformations [mm], Case II . . 135

6.38 Redistribution of the girder bending moments after bridge completion, Case I . 136

6.39 Changes of the vertical deflection of the girder after bridge completion, Case I . 136

6.40 Redistribution of the girder bending moments after bridge completion, Case II . 137

6.41 Changes of the vertical deflection of the girder after bridge completion, Case II 137

6.42 Bending moment distribution under dead load and unit forces at day 89 [kNm] . 138

6.43 Vertical deformation under dead load and unit forces at day 89 [mm] . . . . . . 138

6.44 Moment distribution at day 89 when restricting bending moments [kNm], Case I 141

6.45 Vertical deflection at day 89 when restricting bending moments [mm], Case I . 141

6.46 Moment distribution at day 89 when restricting deformations [kNm], Case II . . 141

6.47 Vertical deformation at day 89 when restricting deformations [mm], Case II . . 142

6.48 Redistribution of the girder bending moments after bridge completion, Case I . 142

6.49 Changes of the vertical deflection of the girder after bridge completion, Case I . 143

6.50 Redistribution of the girder bending moments after bridge completion, Case II . 143

6.51 Changes of the vertical deflection of the girder after bridge completion, Case II 143

6.52 Changes of the vertical deflection of the girder after bridge completion, Case II 145

6.53 Changes of the vertical deflection of the girder after bridge completion, Case II 146
LIST OF FIGURES IX

6.54 Final bending moment distribution along the bridge girder, Case I (MIDAS/Civil)147

6.55 Final vertical deformations of the bridge girder, Case I (MIDAS/Civil) . . . . . 147

6.56 Bending moment distribution along the bridge girder at day 89, Case I (Larsa2000)148

6.57 Vertical displacements of the bridge girder at day 89, Case I (Larsa2000) . . . . 148

6.58 Fabrication camber (Case I) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

6.59 Forecast of the final bending moment distribution after adjustment in CS02 . . 157

6.60 Forecast of the final vertical deflection after adjustment in CS02 . . . . . . . . 158

6.61 Actual fabrication camber after adjustment in CS02 . . . . . . . . . . . . . . . 158

6.62 Forecast of the final bending moment distribution after adjustment in CS03 . . 162

6.63 Forecast of the final vertical deflection after adjustment in CS03 . . . . . . . . 162

6.64 Actual fabrication camber after adjustment in CS03 . . . . . . . . . . . . . . . 163

6.65 Actual fabrication camber after adjustment in CS05 . . . . . . . . . . . . . . . 166

6.66 Actual fabrication camber after adjustment in CS09 . . . . . . . . . . . . . . . 168

6.67 Final fabrication camber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

6.68 Final bending moment distribution of the real structure after adjustment . . . . 170

6.69 Final vertical displacements of the real structure after adjustment . . . . . . . . 170

6.70 Final vertical displacements after adjustment including precamber . . . . . . . 171

6.71 Vertical displacements after adjustment including precamber at day 6000 . . . . 171

6.72 Bending moment distribution in the final construction stage . . . . . . . . . . . 174

6.73 Vertical displacements in the final construction stage . . . . . . . . . . . . . . 174


List of Tables

3.1 Summary of factors accounted for by different prediction methods [12] . . . . . 41

4.1 Material data for the verification model for creep and shrinkage . . . . . . . . . 84

4.2 Section data for the verification model for creep and shrinkage . . . . . . . . . 84

4.3 Loading data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

4.4 Linear construction stage calculation - Vertical deflections . . . . . . . . . . . 84

4.5 Time-dependent construction stage calculation - Vertical deflections . . . . . . 85

4.6 Elastic deformation of joint 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

4.7 Total elastic deformation of joint 2 and initial elastic deformation of joint 3 . . 87

4.8 Total elastic deformation of joint 3 . . . . . . . . . . . . . . . . . . . . . . . . 88

4.9 Creep coefficients φ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

4.10 Sum of elastic and creep deformation of joint 2 . . . . . . . . . . . . . . . . . 88

4.11 Total deformation of joint 2 and initial deformation of joint 3 . . . . . . . . . . 89

4.12 Total deformation of joint 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

4.13 Total vertical deflections of joint 2 and 3 . . . . . . . . . . . . . . . . . . . . . 89

4.14 Axial deformation due to shrinkage . . . . . . . . . . . . . . . . . . . . . . . 90

4.15 Time-dependent construction stage calculation - Vertical deflections . . . . . . 91

4.16 Total vertical deflections using 10 elements per segement . . . . . . . . . . . . 92

6.1 Material data of the example model . . . . . . . . . . . . . . . . . . . . . . . 108

6.2 Section data of the example model . . . . . . . . . . . . . . . . . . . . . . . . 108

6.3 Loading data of the example model . . . . . . . . . . . . . . . . . . . . . . . . 108

6.4 Constraints for the unknown load factor calculation . . . . . . . . . . . . . . . 110


LIST OF TABLES XI

6.5 Ideal cable forces for restricted bending moments, Case I . . . . . . . . . . . . 112

6.6 Ideal cable forces for restricted deformations, Case II . . . . . . . . . . . . . . 113

6.7 Construction activities in the construction stages 1 to 14 . . . . . . . . . . . . . 121

6.8 Ideal cable forces at time of installation and ideal support jacking for restricted
bending moments, Case I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

6.9 Ideal cable forces at time of installation and ideal support jacking for restricted
deformations, Case II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

6.10 Comparison of the ideal cable forces at time of installation and the ideal support
jacking determined by MIDAS/Civil and Larsa2000 . . . . . . . . . . . . . . . 130

6.11 Comparison of the ideal cable forces at time of installation and the ideal support
jacking determined in a time-dependent analysis by MIDAS/Civil and Larsa2000 144

6.12 Comparison of the cable forces at time of installation and the ideal support
jacking determined in a time-independent and time-dependent analysis by MI-
DAS/Civil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

6.13 Comparison of the cable forces at time of installation and the ideal support jack-
ing determined in a time-independent and time-dependent analysis by Larsa2000 147

6.14 Cable forces and jacking distance applied in the construction stage analysis . . 149

6.15 Theoretical cable forces during construction . . . . . . . . . . . . . . . . . . . 150

6.16 Theoretical deformations during construction . . . . . . . . . . . . . . . . . . 151

6.17 Error factors considered in the simulation of the actual erection process . . . . 152

6.18 Comparison of theoretical and actual deformations in CS01 and CS02 . . . . . 153

6.19 Changes of input data of the theoretical analysis model after error identification
in CS01 and CS02 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

6.20 Cable and jacking force adjustments after error identification in CS01 and CS02 157

6.21 Comparison of theoretical and actual deformations in CS03 . . . . . . . . . . . 159

6.22 Cable and jacking force adjustments after error identification in CS03 . . . . . 161

6.23 Comparison of theoretical and actual deformations in CS04 and CS05 . . . . . 163

6.24 Comparison of theoretical and actual cable forces in CS05 . . . . . . . . . . . 164

6.25 Changes of input data of the theoretical analysis model after error identification
in CS04 and CS05 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

6.26 Cable and jacking force adjustments after error identification in CS05 . . . . . 166
LIST OF TABLES XII

6.27 Comparison of theoretical and actual deformations in CS09 . . . . . . . . . . . 167

6.28 Comparison of theoretical and actual cable forces in CS09 . . . . . . . . . . . 167

6.29 Changes of input data of the theoretical analysis model after error identification
in CS09 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

6.30 Cable and jacking force adjustments after error identification in CS09 . . . . . 168

6.31 Comparison of the cable forces in the original analysis and the actual erection . 172

6.32 Input data of the real structure and the theoretical analysis model . . . . . . . . 173
List of Symbols
Matrices and Vectors:
A Influence matrix
Adispl Displacement influence matrix
Aerr Error influence matrix
Amoment Bending moment influence matrix
B Vector of strain-displacement-relationships
D Elastic stiffness matrix
D Vector of actual errors
Df inal Vector of discrepancies between the desired condition and actual final state
E Vector of errors
F Incremental flexibility matrix
K Stiffness matrix
Kt Tangent stiffness matrix
K e
Stiffness matrix of element e
M 0
Vector of target bending moments
M d
Vector of dead load bending moments
M n
Vector of updated deck bending moment in step n
P Load Vector
T Vector of stay cable forces
T n
Vector of updated stay cable forces in step n
∆T Vector of cable force adjustments
∆T n
Vector of cable force adjustments in step n
W Weight matrix
X Vector of multiplication factors
Z Vector of bending moments or displacements of the ideal state
a Vector of nodal displacements
a e
Vector of nodal displacements of element e
f Vector of external loads
f e
Vector of nodal element forces due to creep and shrinkage
q e
Vector of nodal forces due to element e
Ω Vector of square errors
Ψ Sum of internal and external generalized forces
LIST OF SYMBOLS XIV

Scalars:
A Cross-section
Ac Concrete cross-section
C Product specific constant
E Modulus of elasticity
Eci Modulus of elasticity at the age of 28 days
Eeq Equivalent modulus of elasticity
ER Relaxation function
Eµ0 Coefficient related to the initial shapes of specific creep curves at the loading
application time
I Moment of inertia
J Creep compliance function
L Length
Lx Length in x direction
Ly Length in y direction
Lu Unstressed length of a cable
M Bending moment
MPK Bending moment due to permanent load in point K
MTKm Bending moment due to the action Tm in point K
Pn Force number n
RH Relative humidity
T Tension force of a cable
Vx Horizontal deformation
Vy Vertical deformation
Xn Multiplication factor n

ai,j Response at target i by applying a unit force at cable j


fck Charcteristic compressive strength of concrete
fcm Mean compressive strength of concrete at age of 28 days
fpy Yield stress of steel
h Notional size
m Number
n Number
s Coefficient that depends on the type of cement
t Age, time
t0 Age of concrete at the beginning of loading
ts Age of concrete at the beginning of shrinkage
u Perimeter
LIST OF SYMBOLS XV

un Deformation, degree of freedom number n


w Weight

Γµ Retardation time
∆0 Correction factor
Ψ Curvature
α Power
βc Coefficient to describe the development of creep with time after loading
βcs Coefficient that depends on the type of cement
βs Function of time which depends upon the size and shape of an element
δc Creep deformation
δel Elastic deformation
 Strain
0 Initial strain
0 Stress independent elastic strain
a Aging strain
c Concrete strain
cσ Total stress dependent strain
cc Creep strain
cs Total shrinkage strain
cs0 Notional shrinkage coefficient
el Elastic strain
sh Shrinkage strain
sh,0 Total shrinkage strain that occurs after concrete hardening
T h Thermal dilatation
θc Creep rotation
θel Elastic rotation
σ Stress
σ0 Initial stress
σc Concrete stress
σp0 Initial stress
σpr Stress after loading for a period of time
τ Time of loading
φ Creep coefficient
φ0 Notional creep coefficient
ϕz Rotaion
Chapter 1

Introduction and Overview

1.1 Introduction

The cable-stayed bridge is a modern form of bridge which is both economical and aesthetic. It
has been extensively employed in the construction of long-span bridges in the past few decades.
Bridges of this type are now entering a new era with main span lengths reaching 1000m. How-
ever, cable-stayed bridges with multiple stays are highly statically indeterminate structures that
require a computationally intensive design.

For cable-stayed bridges the trend today is to use more slender stiffening girders combined with
increasing span lengths which makes it imperative that the internal forces in the bridge girder
remain within tight limits throughout the construction process. Moreover, for the complete
structure, it is generally aspired to minimize the deformations and internal forces in the bridge
deck and tower. This desired optimum final condition mainly relies on the balancing of loads by
the post-tensioning forces of the stay cables. Due to the high flexibility of cable-stayed bridges,
slight changes in the cable forces may already have a significant influence on the geometry and
the internal forces of the girder and the pylon. Theoretically it is possible to calculate a set of
initial cable forces that exactly cause the desired ”ideal state” in the complete structure. How-
ever, the member forces and deformations at the time of completion of the bridge are generated
during the construction and are dependent on the specifics of the sequence of construction.

Cable-stayed bridges can be built using different erection techniques, but the cantilevering
method is certainly the most efficient construction method for this type of bridges. Especially
long-span cable-stayed bridges are generally erected using this method. During the construc-
tion by cantilever method new girders are installed and then supported by new cables in each
erection stage. In order to balance the weight of the bridge deck, the stays are stressed at the
CHAPTER 1. INTRODUCTION AND OVERVIEW 2

time of their erection. These initial cable forces at the time of installation drastically change
during the erection process and adjustments may be necessary to achieve the desired final con-
dition. However, adjusting the tension force in the stay cables is expensive and thus, tensioning
strategies must be optimized to achieve an economic design.

In case of concrete or composite cable-stayed bridges, the change of the member forces and de-
formations during the erection process is significantly influenced by time-dependent processes,
such as creep and shrinkage. Furthermore, due to the high redundancy of cable-stayed bridges,
the dead load bending moments and cable forces achieved in the final construction stage will
still redistribute due to time-dependent deformations after continuity of the structure is reached.
Therefore, the consideration of time effects on materials is unavoidable for designing concrete
or composite structures.
In addition to time effects on materials, also nonlinear effects influence the structural behavior
of cable-stayed bridges. Generally, high tensile forces exist in the cable stays which induce high
compression forces in the towers and the girders that increase as the main span length increases.
These high normal forces cause a risk to the stability of these components and thus, especially
for long-span cable-stayed bridges, second-order theory and large displacement effects should
be taken into account in the structural analysis. Moreover, the nonlinear behavior of the cable
system as a result of the changes in sag and corresponding axial tension may significantly influ-
ence the optimum initial cable tension forces.

Another problem during the erection of cable-stayed bridges by the cantilevering method is that
even if appropriate initial cable forces have been determined, they may not necessarily cause
the desired final condition. Due to inevitable errors between the structural design parameters
and the actual ones, unexpected discrepancies between the predicted and the actual state of
the structure in a given construction stage may occur. In order to avoid the accumulation of
these discrepancies and to ensure a safe erection process, it is necessary to carry out a detailed
simulation analysis and a continuous monitoring throughout the erection process. This way
the discrepancies can be detected and the erection can be controlled by certain construction
adjustments.

1.2 Objective and scope

The present thesis concerns the analysis and control of cable-stayed bridges during construction
by the cantilevering method. The first main objective is the calculation of suitable stay cable
forces so that the stresses in the stays and the structure remain allowable during the construction
process and a desired final condition is achieved at the time of completion of the structure.
CHAPTER 1. INTRODUCTION AND OVERVIEW 3

In literature, there exists a great variety of approaches to determine optimum stay cable forces,
but most of them are based on the configuration in the final structure and do not take into ac-
count the actual construction process. Furthermore, time effects on materials and nonlinear
behavior are rarely considered. The method presented in this thesis allows the definition of a
desired bending moment distribution and/or a desired geometry in the final structure under dead
load and then computes the tensioning strategy that will exactly achieve the desired condition
taking into account the changes of the structural system during the construction process and
time-dependent effects such as creep and shrinkage. The nonlinear behavior of cable-stayed
bridges may also be considered, but is neglected when adopting the method in the determina-
tion of the optimum tension forces in an example calculation. In this regard it is shown that
in case of small cable-stayed bridges nonlinear effects have minor influence on the structural
behavior.

Because the main concern of this thesis is to achieve an economic design, no provision is made
for expensive cable force adjustments during the erection process. The goal is to determine the
initial cable forces that need to be applied at the time of their installation to achieve a desired
final dead load condition at the time of completion of a structure without restressing any of the
stay cables.

The second major objective of this study is the development and demonstration of a method to
control unexpected discrepancies between the theoretical predictions and the actual structural
responses of a cable-stayed bridge. As before the main concern lies in the economic efficiency
of cable-stayed bridges and thus, the method that is mainly presented in this thesis does not im-
ply the retensioning of already installed cables. The goal is to adjust the cable forces and deck
elevations of the remaining cables and girder segments to achieve the desired final condition
within reasonable tolerance.

Both the determination of the initial cable forces and the control of the geometry during con-
struction are shown on the basis of an example calculation of a simple model of a cable-stayed
bridge. This calculation is performed by the dint of two standard structural analysis programs,
namely Larsa2000/4th Dimension and MIDAS/Civil. Both programs have a particular emphasis
on bridge structures, but they cannot automatically determine the post-tensioning forces needed
in the cable stays.
From the analysis of the example cable-stayed bridge by Larsa2000 and MIDAS/Civil a sec-
ondary objective of the present thesis arises that includes the comparison of the functionality of
these structural engineering software programs with respect to the structural analysis of cable-
stayed bridges.
CHAPTER 1. INTRODUCTION AND OVERVIEW 4

1.3 Thesis overview

Chapter 2 introduces the reader to the general background of cable-stayed bridges. Different
advantages of this type of bridges are pointed out and a short historical review is given. Further-
more, the erection techniques of cable-stayed bridges are briefly outlined and the cantilevering
method is explained in more detail.

Chapter 3 deals with the description of the methods of structural analysis of cable-stayed bridges
erected by the cantilevering method, whereby the focus is on the determination of an optimum
tensioning strategy. First, three different existing methods to calculate the initial cable forces
are described and compared. Then, based on the disadvantages of these methods regarding the
consideration of the actual construction process, time-dependent effects and nonlinear behavior
in the cable force calculation, a new method, namely the ”expanded unit load” method, utilizing
the idea of unit forces in the cables is presented in detail. Within the scope of this section, the
consideration of the change of structural systems during construction, second-order theory and
large displacement effects, cable sag effects and time-dependent effects, such as creep, shrink-
age and steel relaxation, in the optimization process is explained in detail. The main attention in
this explanation focuses on the time-dependent material behavior of concrete. In addition to the
description of the influence of time-dependent material behavior, different methods to predict
the effects of creep and shrinkage in a cable-stayed bridge are presented. Closing the section
about the ”expanded unit load” method, the implementation of this method into a finite element
software is demonstrated. Finally, the last part of this chapter gives an insight of the geometry
control of cable-stayed bridges by an appropriate precamber of the bridge deck.

In Chapter 4 different structural engineering software programs with special emphasis on bridge
structures are introduced and the functionality of the analysis programs Larsa2000/4th Dimen-
sion and MIDAS/Civil concerning the construction-stage analysis of cable-stayed bridges is
described. As the main focus of this thesis is the consideration of time-dependent material ef-
fects, on the basis of a simple construction stage analysis, the application of the functions to
include creep and shrinkage in Larsa2000 and MIDAS/Civil is proved at the end of Chapter 4.

Chapter 5 concerns itself with the construction control of cable-stayed bridges during the erec-
tion process. Possible causes of discrepancies between the predicted and actual state of a struc-
ture are revealed and a method to adjust these discrepancies without restressing already installed
cables is presented.
CHAPTER 1. INTRODUCTION AND OVERVIEW 5

In Chapter 6 the procedure to determine the initial cable forces, presented in Chapter 3, and the
method to control the geometry and internal forces during the cantilever construction, explained
in Chapter 5, are adopted in the analysis of a simple cable stayed bridge. The calculations are
performed by the dint of the analysis programs Larsa2000 and MIDAS/Civil. The first part
of this chapter shows the dimensions, structural properties and loading data of the example
model. Then, the restrictions for the optimization process are explained and the initial cable
forces and the ideal jacking distance of the right support are determined in a linear static, a
time-independent and a time-dependent construction stage analysis. The results determined
by the use of Larsa2000 and MIDAS/Civil are compared and the influence of time-dependent
effects on the optimum tensioning strategy is emphasized. In order to clarify the impact of
second-order theory and large-displacement effects, on the behavior of the example structure,
subsequently also a nonlinear time-dependent construction stage analysis is performed and the
results are confronted with those of the linear analysis. Finally, various error factors are incor-
porated into the analysis model and the amendment of discrepancies between the predicted and
actual state during the construction process is exemplary shown. On the basis of the results of
the adjusted analysis, the method to control the construction of cable-stayed bridges presented
in Chapter 5 is verified.

Chapter 7 summarizes the work and provides a brief conclusion to this thesis.
Chapter 2

Cable-Stayed Bridges

2.1 Introduction

Due to their aesthetic appearance, efficient utilization of structural materials, relatively small
size of the bridge elements and other notable advantages, cable-stayed bridges have gained
much popularity in recent decades. They are usually the most economical choice for spans be-
tween 200m and about 1000m.
A cable-stayed bridge consists of three principal components, namely girders, towers and in-
clined stays. The girder is supported elastically at points along its length by numerous cable
stays so that the girder can span a much longer distance without intermediate piers. The dead
and live loads of the girder are transmitted to the pylon by the inclined stay cables that are
attached to it above the bridge deck. The pylons of a cable-stayed bridge can be shaped in a
great number of ways, including A, H, X, and inverted V and Y-shapes, or combinations and
variations of these. For the cable layout a multitude of arrangements exists. The bridge can be
designed with one central or two lateral planes of stay cables that can even be inclined toward
each other. The cable system itself can be arranged as a fan system, i.e. the cables run radiating
from one point at the pylon, as a harp system, i.e. the cables run parallel from equal spacing at
the pylon or as a mixture of both. The stay cables can be anchored both on the deck and at the
pylon or can run continuously over a saddle at the top of the pylon.

The contemporary cable-stayed bridge is becoming more and more popular and is being used
where previously suspension bridge might have been chosen. Some of the advantages of cable-
stayed bridges when comparing them to suspension bridges are:

• they are self-anchored even during construction and therefore do not require the costly
counterweight required to anchor suspension bridges

• they react to live loads with only small deformations


CHAPTER 2. CABLE-STAYED BRIDGES 7

• they can be easily constructed (balanced cantilevering method)

• they provide an increased stiffness over suspension bridges

Due to the tension forces in the stay cables, cable-stayed bridges develop high compressive
stresses in the girder. For spans up to about 350m, the concrete decks are economical because
they can utilize these axial forces as cost-free prestress. Beyond these span lengths, the concrete
decks become too heavy and costly, but a steel-only deck would also be to expensive. There-
fore, from 350m to about 700m, the composite deck is advantageous. The concrete slab of
the composite deck receives a cost-free prestress from the horizontal components of the cable
forces and is concurrently stiffened by the steel girder to resist bending. At spans beyond about
700m, the composite deck becomes too heavy, and the steel girder with an orthotropic deck
remains the only economical choice.

In the following a short historical review of cable-stayed bridges is given and then different
erection methods are introduced. Because the focus of this thesis is on the analysis of cable-
stayed bridges constructed by the cantilevering method, the main concern lies in the description
of this erection method. For further information about cable-stayed bridges in general and their
methods of construction it is referred to the literature (e.g. [15, 40, 42, 47, 49, 50]).

2.2 Short historical review

The concept of supporting a bridge deck by inclined tension stays and achieving a more favor-
able distribution of bending moments in the bridge deck can be traced back to the the 17th and
18th century. As at that time neither precise calculation methods nor adequate tension members
were available, some bridges collapsed and the system disappeared for about two centuries.
The reappearance of cable-stayed bridges started with Franz Dischinger, a German engineer, in
1938. He discovered the effect of stays upon the deflection characteristics of a bridge whilst
designing a suspension bridge across the Elbe River near Hamburg. From this study and from
the reconstruction of the bridges destroyed in Second World War it was found that cable-stayed
bridges had a part to play in spans between girder bridges and suspension bridges.
The 1953 completed Strömsund Bridge in Sweden, which was designed by Franz Dischinger,
is generally regarded to be the first modern cable-stayed bridge. After the Str ömsund bridge
numerous true cable-stayed bridges were constructed in Germany. The Theodor Heuss Bridge
across the Rhine River in Duesseldorf with a main span of 260m was opened to traffic in 1957,
the Severins Bridge in Cologne spanning 302m was erected between 1956 and 1961 and the
third and fourth German cable-stayed bridges, the Norderelbe Bridge in Hamburg (main span
length 172m) and the Leverkusen Bridge across the Rhine River (main span length 280m), were
CHAPTER 2. CABLE-STAYED BRIDGES 8

completed in 1962 and 1964. The Theodor Heuss bridge introduced a harp-shaped cable system
with parallel stays and free-standing pylon, the Severins Bridge was the first application of an
A-shaped pylon with transversally inclined cable planes and the last two bridges were built wit a
central cable plane with pylons and stay cables positioned in the central reserve of the motorway.

However, the bridge decks of the early cable stayed-bridges built at the end of the 1950s and the
beginning of the 1960s were supported by a limited number of cables that were generally com-
posed of several prefabricated strands to achieve the required large cross-section. The Friedrich
Ebert Bridge across the Rhine River in Bonn, completed in 1964 (Homberg), was the first cable-
stayed bridge designed with a multi-strand arrangement. It contains a central cable plane with
two pylons, each supporting 2x20 stays. The spacing between the cable anchorages is only
2.24m. Since then, multistay systems have become common in modern cable-stayed bridges
where the high number of cables allows for more slender main girders that require less flexural
stiffness.
Whereas in early cable-stayed bridges steel was dominating the structural material of deck and
towers, today towers are normally of concrete; the choice of material of the bridge deck depends
on the span length as it was explained above. Nevertheless, prestressed concrete and composite
decks have increasingly been used in more recent projects.

Until today many cable-stayed bridges have been built using increasingly slender girders with
larger and larger spans. Currently, the Stonecutter Bridge in Hong Kong is under construction.
It will be the first cable-stayed bridge spanning over 1000m, but the trend goes to even bigger
spans and more slender girders with an increasing span-to depth ratio.

2.3 Erection of cable-stayed bridges

Cable-stayed bridges can be built using different erection techniques, which are chosen accord-
ing to local conditions and bridge characteristics. The construction methods that can be used
are:

• Construction of the whole bridge deck on scaffoldings or on temporary supports

• Construction by rotation

• Construction of the bridge deck by the incremental launching method on temporary sup-
ports in the main span

• Construction of the bridge deck by the cantilevering method


CHAPTER 2. CABLE-STAYED BRIDGES 9

In the relatively simple first construction method, the entire bridge girder is erected on tempo-
rary supports or on scaffoldings that can be adjusted in order to achieve the correct position.
Thereafter, the mounting cable forces are precisely evaluated to balance the vertical deck reac-
tions on the temporary supports, leading to the pretended geometry and stress distribution. The
advantage of this erection procedure is that the deck geometry and the cable tension forces can
be controlled easily. However, the use of temporary supports fails when clearance of the main
span is required during construction and is not economical when the main span of the bridge
crosses deep water.
The construction of cable-stayed bridges by rotation is the preferred method when building over
a waterway, for example a river. The bridge deck is erected on temporary supports at the shore
parallel to the bank and after the tensioning of the stay cables it is rotated around its pylon.
When erecting a cable-stayed bridge by the incremental launching method, the superstructure is
cast-in-situ at a stationary location behind one of the abutments and is then jacked horizontally
into place. The procedure has the advantage that, in contrast to the first two methods, it does not
require falsework to cast the girder. However, all these three construction methods are generally
limited to small and medium sized cable-stayed bridges.
Nowadays, most of the cable-stayed bridges - and all long span cable-stayed bridges - are built
by the last method, the cantilevering method, which is described in more detail in the following.

2.3.1 Construction of cable-stayed bridges by the cantilevering method

The cantilevering method is a construction method where segments, either precast or cast-in-
place, are assembled and stressed together subsequently to form the self-supporting superstruc-
ture. When using the cantilevering method to erect cable-stayed bridges the procedure includes
the following stages:

1. The pylons and the girder units above the main piers are erected and fixed to the piers

2. A cantilevering is initiated by sequentially adding new segments. Normally the structure


advances from a short stub on top of a pier in segments of about 3m to 5m length to the
mid span or to an abutment

3. As the cantilevers grow, the stay cables are installed and post-tensioned to prevent exces-
sive deflections and to relieve bending moments in the bridge girder

4. The bridge is closed at mid span at the second abutment and the additional loading from
wearing surface is placed

An example of the erection of a three-span cable-stayed bridge constructed by the cantilevering


method using precast segments is illustrated in Figure 2.1.
CHAPTER 2. CABLE-STAYED BRIDGES 10

Figure 2.1: Construction by the cantilevering method using precast segments

Depending on the specific segment configuration and erection sequence chosen for the can-
tilevering method, the cantilever may never be exactly balanced so that the superstructure needs
to be balanced to ensure stability. It is possible to fix the supports at the piers of cantilevering
superstructures and install vertical prestressing tendons. Furthermore, it is very common to
make use of an additional temporary pier with vertical prestressing that is located close to the
permanent one. This pier helps withstanding overturning moments from unbalanced load cases
on the bridge superstructure. Additionally, the lateral bending stiffness of the girder must be
sufficient, to ensure the stability of the cantilever arm during erection.

Compared to the first three construction methods described above, the cantilevering method
provides several advantages. Certainly the most important one is that no falsework or centering
is required, leaving traffic under the spans widely unobstructed during construction. Access
from the ground is only necessary for construction of the piers and abutments and in preparation
for the start of cantilevering, which starts from these locations. In contrast to the incremental
launching method, this way not only small and medium sized, but also long-span cable-stayed
bridges can be built.
Only relatively little formwork is required due to the segmental nature of the superstructure.
Cantilevering is a very feasible method if the bridge spans are too high above ground for e.g.
economical use of falsework, and if the terrain under the spans is otherwise inaccessible or
unfeasible, being e.g. a deep gorge with danger of flood events. Especially in these cases rapid
construction can be achieved with cantilevering.
CHAPTER 2. CABLE-STAYED BRIDGES 11

Nevertheless, the cantilevering method also has disadvantages regarding the control of the de-
sired geometry. The erection procedure produces deflection and stress histories, which must be
carefully evaluated. Even if great flexibility and slender decks, which are now favored, gener-
ally allow for easy adjustments when geometry has not been perfectly controlled, the control of
the desired geometry becomes a major problem when constructing cable-stayed bridges by the
cantilever method.
The methods of analysis and construction control of cable-stayed bridges constructed by the
cantilevering method are explained in detail in Chapter 3 to 5 and are therefore not discussed in
this part of the report.

Following the initial choice of material for a cable-stayed bridge superstructure, the next deci-
sion to be made by the designer is the method of cantilever construction. Assuming concrete
has been selected, there are two main choices for construction:

1. Cast-in-place with travelling forms

2. Precast segmental construction

If steel or composite sections are chosen, this choice does not exist. At least the steel girder is
always prefabricated, transported to the site and installed there.

2.3.1.1 Cast-in-place construction

When the cast-in-place method is used, so-called form travelers made of steel framework are
attached to the cantilever tip where they carry the formwork in which new segments are cast. A
typical casting cycle starts with the detachment of the form traveler from the previous position
after finishing all work on the last segment and the forward movement to its new position. In
order to remain balanced during advancement, the form traveler may be equipped with a coun-
terweight. Then the form traveler is adjusted and anchored to the existing superstructure at its
rear to be able to withstand overturning moments that will occur from the weight of the new
concrete. Subsequently, the external formwork is aligned to the required geometry of the next
segment, also incorporating the desired camber, the reinforcement and tendon ducts are installed
and connected with the previous ones and the concrete is placed. Before the casting cycle starts
all over again, the cast segment needs to have developed at least the specified strength to be
prestressed to the previous elements and support the subsequent one.
The maximum segment length achieved with form travelers is about 5.00 m, but there are possi-
bilities to increase the length of segments when keeping in mind the increasing weight and cost
of bigger form travelers. Moreover, a stepwise construction of the segments or a combination
of cast-in-place and precast segment sections is also possible.
CHAPTER 2. CABLE-STAYED BRIDGES 12

2.3.1.2 Precast Construction

Precast construction means that bridge members or segments are prefabricated at a location
different than the site, transported to the site, and installed there. The lifting of the precast
segments can be carried out by floating cranes or by using beam-and-winch system. The floating
crane is preferable for low-level bridges and for high level bridges it is easier to use a beam and
winch system where the segments are lifted off a barge with a winch on the deck.
The construction with precast segments has several advantages in comparison with cast-in-
place erection method. First, the casting of the segments can be performed under controlled
conditions at the precasting yard. This industrialized process allows easy quality control of
segments prior to placement in the superstructure and saves costs. Surface finishing works,
such as texturing, sandblasting, painting, and coating can be performed on the ground level
without scaffolding when the segments are still accessible from all sides prior to installation in
the superstructure. Second, the complete casting of the superstructure can be removed from the
critical path of the overall construction schedule, since superstructure segments can be precast
during construction of the substructure. Assembly of the bridge superstructure takes much less
time than cast-in-place construction, as precast segments do not need to cure on site before
being prestressed together. As segments are usually stored at the precasting yard or on site for a
while the concrete will have gained more strength until installation than cast-in-place elements
have when being loaded.
Furthermore, the time-dependent effects of concrete will be reduced due to the increased age
of the concrete segments and will thus cause smaller deflections of the superstructure than with
cast-in-place construction. However, cost for the precasting yard, storage, transportation, and
installation of precast segments needs to be evaluated in comparison with cost for the form
travelers for cast-in-place construction to achieve an economical solution.

2.3.1.3 Construction of steel and composite decks

Similar to the construction of precast concrete bridges, the segments of steel or composite
bridges can be lifted into place by derrick cranes or floating cranes. In case of pure steel
cable-stayed bridges, after lifting a new segment, it is welded to the previous one while it is
still supported by the crane. Then the cables are installed and post-tensioned and the casting
cycle starts all over again.
Composite cable-stayed bridges consist of a steel girder and a concrete slab. In this case the
steel elements are prefabricated and the girder is built by the cantilever method segment after
segment and suspended to the cables. Then the concrete slab is cast or installed (precast slab)
on the existing bridge girder. However, the concrete slab can also be constructed segment by
segment during the cantilever construction, but this may result in a slow-down of the construc-
tion process. Generally, a distance of about three segments is required to allow an installation
of the concrete slab that is independent from the construction of the steel girder.
Chapter 3

Structural Analysis of Cable-Stayed


Bridges

3.1 Introduction
In cable-stayed bridges the girders are supported at several locations, namely abutments, piers
and cable points with the cables connected to the towers. The abutments and piers are usually
considered as fixed supports and the points of cable attachments are elastic supports as the ca-
bles change length under load and because the towers are also flexible and can move.
Modern bridges use closely spaced multiple stays, which leads to ease of erection and allows for
more slender main girders that require less flexural stiffness. Cable-stayed bridges are, there-
fore, highly statically indeterminate structures.
During the construction by cantilever method new girders are installed and then supported by
new cables in each erection stage. In order to balance the weight of the deck without initial
deformation due to slack and elongation of steel all stays are stressed at the time of erection. If
no provision is made for retensioning, for the subsequent construction phases, the stays act as
passive members like the pylon and the deck.

The slenderness of bridge girders in modern cable-stayed bridges requires to minimize the bend-
ing moments throughout the whole construction process and to achieve an optimal moment dis-
tribution in the finished structure under dead load. Especially in case of prestressed concrete
cable-stayed bridges it is important to choose an appropriate set of initial cable forces in or-
der to reduce time-dependent displacements and the consequential redistribution of the internal
forces due to creep and shrinkage. Theoretically it is possible to search for a scheme of initial
cable forces which minimizes those effects, but it is difficult to take the many factors affecting
the subsequent time-dependent deformations into account. Factors complicating the effects of
creep and shrinkage are, among other things, the consideration of the age of concrete, especially
when using cast-in-situ methods, and the presence of longitudinal prestressing. In these cases it
is inevitably to make some simplifications to consider the long term behavior of the structure.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 14

In cable-stayed bridges an almost stress-free condition of the girder and pylon can be achieved
by an adequate tensioning of the cables in the particular erection stages. The intention is that
the bending moments in the finished structure do not exceed tight limits. To reach this condi-
tion the designer needs to determine the optimum final moment distribution and then the initial
cable forces can be established. But it has to be considered that during erection this final dead
load condition will mainly be influenced during tensioning of the cables. The problem is that
the initial cable forces at installation and tensioning are usually quiet different from those of
the final dead load condition. As the erection proceeds the cable forces drastically change,
which has led to the practice of adjusting the tension of the stay cables during the individual
construction stages. However, adjusting stay tension forces has disadvantages that are not to be
scoffed at. Due to the high redundancy of the structural system restressing one single cable to
correct discrepancies also effects the forces in all other cables. In some cases even compressive
cable forces might develop by stressing an adjacent cable, which those tension-only members
can not withstand. Furthermore, stressing of stay cables is an expensive procedure causing an
overwhelming amount of work, especially in multi-cable-systems. Hence, the ideal solution
would be to stress each cable only once immediately after its installation or at least attempt to
reduce the number of stressing operations during the erection of cable-stayed bridges as much
as possible.
In the following, different existing methods to determine the initial cable forces are presented
and compared. Then an improved method called the ”expanded unit load” method is described
in detail and the advantages over the conventional methods are outlined.

3.2 Review of existing methods to determine the initial cable


forces in cable-stayed bridges

To analyze cable-stayed bridges a number of different techniques can be used. Indeed, because
of the large degree of indeterminateness, the analysis is relatively complicated and an exact
calculation by manual procedures is virtually an impossible task. Hence, the intention of this
report is not to go into these ’classical methods’ to estimate the cable forces in more detail.
In the literature three main categories of methods to adjust the stress distribution and the geome-
try of steel, composite or concrete cable-stayed bridges have been proposed. These methods are:

1. The ”optimization method”

2. The ”zero displacement method”

3. The ”force equilibrium method”


CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 15

3.2.1 Optimization method

In the ”optimization method” [33, 45, 54, 56], the cable forces are determined based on certain
functions that are related to structural efficiency or economy. In order to achieve an economic
and safe solution for a cable-stayed bridge there is a multiplicity of objective functions to be
minimized. Usually it is not the intention of the optimization to consider all factors influencing
the economics of a design. Objective functions to be minimized by the optimization method
are often the volume/cost of the starting trial design, the allowable stress of materials and the
tolerance deflection for geometry control. Assuming the geometry of the bridge including the
cross sectional areas of the cables, the girder and the pylon is known, the optimum solution of a
cable-stayed bridge is, as mentioned before, the solution in which the bending moment and/or
the displacement distribution along the girder and the pylon of the finished structure reaches the
”ideal state”. For pure concrete and composite decks, the ideal bending moment distribution
after the completion of the bridge deck usually equates the moment distribution of the contin-
uous beam on rigid supports under dead load. The reason is that here the sum of the positive
and negative bending moments, and therefore the spanwise overall curvature, is zero. Hence,
moment variations due to creep and shrinkage eliminate each other and theoretically the mo-
ment distribution does not change with time. If the bending moments differ from this state, they
would change towards these values in long term anyway. The bending moment in the pylon
base is normally optimized when reaching zero or small values at the end of construction.
For the ideal displacement state, the deflections at certain locations (i.e. anchorage points of the
cable and top of the pylon) are set to zero or limited to small values.

Considering only one objective function, the moment or displacement of the ideal state, Z, can
be written as:
h iT
Z= z1 z2 ... zn (3.1)

where n is the total number of targets that need to be satisfied and T stands for the transformation
of a matrix or a vector. To approach the ideal state, Equation (3.1) needs to be as close to the
designated value as possible. The cable stresses, T, resulting from the optimal Z can be written
as:
h iT
T= t1 t2 ... tm (3.2)

where m is the total number of cables to be tuned.


CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 16

By analyzing the response of a unit prestress applied at each tuning cable, the influence value
of all the targets can be obtained. When m rounds of analysis are done, the influence matrix A
can be written as:
 
a11 a12 ... a1m
 a
 21 a22 ... a2m 

A=  (3.3)
 ... ... ... ... 
an1 an2 ... anm

where aij is the response at target i by prestressing the unit stress at cable j. Thus, their rela-
tionship can be written as:

AxT = Z (3.4)

If the number of cables to be tuned is the same as the number of targets, the cable stresses can
be obtained by setting Z to the designated target values and solving the linear Equation (3.4).
Indeed, engineering experience is required in selecting the cables and specifying the targets. A
bad or contradictory tuning of cables and targets may cause Equation (3.3) not to be a diagonal
dominant matrix or Equation (3.4) to be in ill condition. Using this method the number of
columns, m, of matrix A cannot be greater than the number of rows, n. If m is less than n, then
the cable stresses can be optimized so that the error of the target value, Z, and the designated
ideal state, D, is minimum. The error, E, can be written as:

E=D − Z (3.5)

One of the most effective ways to obtain the optimal Z is to minimize the square error Ω, which
can be written as:

Ω = (D − Z)2 (3.6)

The condition to have Ω minimized is to equate the matrix differential of Ω with 0.

∂Ω
= 0, i = 1, 2, 3, ..., m (3.7)
∂Si
By using the minimum of Ω and considering Equations (3.6) and (3.4), the following equation
can be obtained.

AT A x T = A T D (3.8)

After solving for T from the linear equation group in Equation (3.8), the optimized target value
will be known from Equation (3.4).
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 17

In practice, the following procedures are common in approaching the ideal state at service stage.

(a) Select all the cables to be tuned

(b) Perform the static analysis under structural weight and the superimposed dead load

(c) Select the negative displacement of the girder at each anchor in (b) as D

(d) Evaluate T above

(e) Similar to prestressing loads, reapply T on the structure and perform the analysis

(f) The sum of (b) and (e) is the ideal state of the structure at complete stage

Using the ”optimization method” it is possible to calculate the optimum initial cable forces, but
neither nonlinearity effects nor time-dependent material behavior are taken into account in the
optimization process. Furthermore, it is necessary to impose the constraints for optimization
very carefully, or else the resulting schemes may not remain within practical limits.

3.2.2 Zero displacement method

The ”zero displacement method” described by Wang et al. [51, 52, 53] determines the tension-
ing forces of stay cables by an iterative ”shape finding” procedure, which reduces the deflections
at the cable anchorages in each iteration step and finally makes them vanish. The idea is that
in case of a straight and horizontal bridge deck, the horizontal components of the cable forces
have little effects on the bending moments of the deck, and hence only the vertical components
influence the bending moment distribution. In this case the resulting bending moments and de-
flections in the deck are essentially those of an equivalent continuous beam with all the supports
from the cables and towers considered as rigid supports. In other words, if the deformation of
the bridge after construction looks like the deformation of an equivalent continuous beam the
”ideal state” is reached and the initial cable-forces are determined.
The computations for shape finding are performed by using the two loop iteration method,
which is started with estimated (usually zero or very small) tension forces in inclined cables.
Based on a reference configuration, having no deflection and zero prestress in girders and tow-
ers, the equilibrium of the cable-stayed bridge under dead load is first determined iteratively
(Newton-Raphson method). Although this first determined configuration satisfies the equilib-
rium and boundary conditions, in general, because of the small pretension in inclined cables,
large deflections and bending moments may appear in girders and towers.
Hence, the shape iteration has to be carried out in order to reduce the deflection and at the same
time smooth the bending moments of the girder. In doing so, the cable axial forces determined
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 18

in the previous step will be taken as the initial element forces for inclined cables for the next
iteration and the the equilibrium configuration will be determined anew.
In each iteration step the ratio of the vertical displacement to the main span length at certain
control points (usually nodes intersected by the girder and the cable) is calculated and then
compared with the convergence tolerance.

Vertical displacement at control points


Main span length ≤ s

The shape iteration is repeated until the defined convergence tolerance is achieved.

In the described procedure, nonlinearities, such as P-Delta effects, large displacement effects
and the cable sag effect can be taken into account, but the method does not consider the stress
distribution in the pylon or the girder resulting from the tensioning procedure. Furthermore,
when the vertical profile of the bridge deck is significant, the horizontal components of the ca-
ble forces cause an additional bending moment in the girder. Even if the displacements at the
anchorage points of the cables are zero, these higher bending moments will negatively affect
the long term behavior of the bridge. Hence, the main intention of adjusting the cable stresses
should rather be to achieve a designated moment distribution than to minimize displacements.
The final geometry of the bridge can always be controlled by defining an appropriate precamber.

3.2.3 Force equilibrium method

In contrast to the first two methods to determine the applied cable tensions, the ”force equi-
librium method” is an approach to influence the moment distribution in the bridge girder by
adjusting the tension in stay cables. The basic concept of this method was already described
1972 by Lazar et al. and repeated in 1988 and 1997 by Troitsky [47] and Gimsing [15], but it
was first proposed as the ”force equilibrium method” by Chen et al. [6, 7]. The method searches
for a ”stable” scheme of initial cable forces which will give rise to desirable bending moments
at selected locations of the structure. As this method only deals with the equilibrium of forces
rather than deformation, nonlinearities caused by cable sag and other effects do not need to
be considered during the determination of the cable forces. However, in order to determine the
corresponding final geometry of the bridge and to define a appropriate precamber it is necessary
to take these nonlinearities into account.
The first step of the ”force equilibrium method” is to choose certain sections of the bridge girder
where the bending moments are adjusted by varying the cable forces. These sections could for
example be located at the points where the stays are anchored and at the tower base. The more
control points are chosen, the better the results.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 19

To establish the target bending moments, the deck of the cable-stayed bridge is considered as
an equivalent continuous beam with all the supports from cables and towers considered as rigid
simple supports. These target bending moments are adopted, so that the effects of creep and
shrinkage can be minimized (see Section 3.2.1). If the initial bending moments in the towers
can be controlled at the same time, the scheme of initial cable forces is reasonably stable.
In the second phase, all the cables are taken away from the bridge and replaced by the internal
forces. These forces are assumed as independent variables for the adjustment of bending mo-
ments at the control sections. Then approximate influence coefficients are evaluated, which are
the bending moments at the control sections caused by a unit load in a certain cable. In order
to simplify the calculation, the weight of the cables is neglected (forces at the cable ends are
roughly equal) and it is assumed that the cable forces acting on the tower do not influence the
bending moments in the deck and similarly the other way around. Considering the equilibrium
of this phase, the following equation can be written:

M0 = A x T + M d (3.9)

where M0 is a vector containing the target bending moments, A is a matrix of the approximate
influence coefficients, T is the vector of the cable forces and Md is the vector containing the
bending moments caused by dead load and prestress only.
If M0 contains the bending moments of the equivalent continuous beam on rigid simple sup-
ports, and the control sections are well chosen so that A is nonsingular, an initial estimation of
the cable forces T can be calculated as:

T0 = A−1 x (M0 − Md ) (3.10)

However, the cable forces obtained above are only rough estimates as the interaction of the
tower, the cables and the deck has not been taken into account yet. Thus, a third stage is per-
formed in which these errors are almost eliminated. Here the cable forces at the deck anchor-
ages are taken as independent variables and the self weight of the cables can also be introduced.
Using the initial estimate of the cable forces, T0 , as well as the bending moments, Md , the
updated deck bending moments, M1 , which are normally different from the target moments,
can be calculated. Now an iteration process starts. To equate the updated moments, M 1 , with
the target moments, M0 , the following cable force adjustment, ∆T1 , is introduced:

∆T1 = A−1 x (M1 − M0 ) (3.11)

Using the updated cable force, T1 is given by:

T1 = T0 + ∆T1 (3.12)
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 20

The updated deck bending moments, M2 , can then be calculated. Since this moment still differs
from the target value further cable force adjustments, ∆T2 , ∆T3 , ..., ∆Tn , are necessary. The
above described adjustment procedure is repeated until the updated deck bending moments,
Mn , converge to M0 .
This method to obtain the initial cable forces is simpler than the traditional ”zero displacement
method”, but it is obvious that non-linear effects need to be included in the algorithm in order
to get more precise results. Furthermore, this method also does not consider time-dependent
material behavior.

3.2.4 Summary of existing methods

3.2.4.1 Consideration of non-linearity effects

Concerning the nonlinear behavior of cable-stayed bridges both the ”optimization method” and
the ”force equilibrium method” do not consider any non-linearity effects in the determination
of the initial cable forces of the complete structure. All nonlinearities of cable-stayed bridges,
such as 2nd order theory (P-Delta and large displacement) and the nonlinearity of the cables
are completely neglected and only linear-elastic structural behavior is assumed. In the ”zero
displacement method”, in contrast, a linear or a nonlinear computation procedure is possible.
In the latter the nonlinear cable sag effects, the P-Delta effects and large displacement effects
are taken into consideration. However, studies on the nonlinear analysis of cable-stayed bridges
using the ”zero displacement method” [52] showed that there are only small differences in ge-
ometry and prestress distribution between the results determined by linear and nonlinear com-
putation procedure. Therefore a reasonable optimum initial design of cable-stayed bridges can
also be achieved by a linear analysis and thus, any of the above mentioned analysis methods is
suitable. In general, the first determination of the optimum initial cable forces should always be
a simple linear analysis, because that way a lot of computation efforts can be saved.

3.2.4.2 Consideration of construction process and time-dependent effects

All solutions described in Sections 3.2.1-3.2.3 are based on the determination of the initial cable
forces of the final structure and do not take into account the actual construction process. This
is problematic, because the internal forces as well as the geometry of a completed cable-stayed
bridge highly depend on the construction sequence of the structure. As already mentioned
before the initial cable forces at the time of installation clearly differ from those of the final
dead load condition. Moreover, the geometric profile of the girder during the construction
greatly changes and it is important to ensure that the cantilever ends finally meet in the center
of the span.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 21

Another reason for considering the erection process is that sometimes the structural behavior
during construction might be much more critical than that of the final stage. In order to com-
plete the design of the bridge, the stresses in the cables, the girder and the pylon need to be
checked not only in the final but also in intermediate erection stages. Therefore, the internal
forces in members of the bridge structure at each erection stage have to be known.
From all this it becomes clear that for a feasible optimum design and correct modeling of a
cable-stayed bridge a detailed reference to the erection process is needed.

Backward analysis

Assuming that the ideal state determined by one of the above mentioned methods is really what
is reached at the end of construction, then there is a very popular approach to take the con-
struction sequence into account, called the backward solution. First, the final erection stage
is analyzed and the optimum complete design of the bridge is defined by any of the described
methods. Then the structure is virtually disassembled stage by stage in reverse direction to the
sequence of erection stages in the real bridge construction. It is assumed that the sequence of
events during disassembly is exactly the opposite of that which occurs during assembly. After
releasing girder segments or stay cable the internal forces of the members are determined in
each erection stage of the backward analysis. The tension in a particular cable just before its
removal can be taken as the initial stressing force of exactly this cable at the time of installation
in the real bridge construction. Hence, this reverse analysis enables the determination of initial
cable tensions that are the guidance for the forward erection in order to finally reach the ideal
state at the end of construction.

Forward analysis

The forward process analysis is performed exactly by following the sequence of erection stages
in construction. In the first construction stage, only the pylon is activated and than in the follow-
ing stages the segments and stay cables are erected and the corresponding loads are applied. The
analysis is carried out stage by stage until the bridge girder is completely erected and the results
are continuously accumulated. In order to achieve the optimum final dead load condition in the
complete structure the cable forces calculated from the backward analysis are applied as initial
tension forces at installation of the cables. In doing so, the last stage in the forward analysis
should theoretically correspond to the first stage of the backward analysis (ideal state) whereby
the construction process of the bridge is considered in the analysis.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 22

Final comments on the backward and forward analysis

The idea of performing a backward and forward analysis is that both - if correctly applied -
give identical results for all construction stages. However, in reality this is hardly ever the case.
Changes of the structural system by removing certain elements or modifying the boundary
conditions might cause gaps between the backward and forward analysis. The reason is that
the backward method requires a stress-free situation in an element before its removal. An
example is that in the forward construction the key segment is applied in zero stress condition,
whereas in the backward analysis the closure segment is not in zero stress state after removing
the superimposed dead load. In order to also achieve a stress free situation in the backward
analysis there are the two following options:

1. Applying a detension load at the cable nearest to the closure, which makes the bending
moment at the tip of the key segment vanish

2. Applying a vertical movement at the bridge closure until the bending moment at the tip
of the key segment becomes zero

However, this way only the bending stresses in the key segment can be eliminated but stresses
caused by possible axial forces remain in the structure. Furthermore, it is a complicating proce-
dure which may require an additional optimization of the structure.
Another problem causing discrepancies between the backward and forward analysis are dis-
continuities at the joints between two segments. If elements are applied straight forward, the
forward method automatically introduces a displacement (V x , Vy , ϕz ) in the structure when
activating a new segment (see Figure 3.1), while in the backward method no discontinuity is ap-
plied when inactivating a segment. Hence, discrepancies in the deflected shape of the structure
arise.

In order to get similar results in both analyses there are two options:

1. Compensate Vx , Vy , and ϕz by element end deflection in the ”forward procedure”

2. Add Vx , Vy , and ϕz by element end deflection to ”backward procedure”

The easiest option to compensate for the difference in the deformed shape is to rather apply the
new segments in the forward analysis tangentially than straight forward (Option 1). Most analy-
sis programs providing staged construction analysis features offer a function, which calculates
the real displacement/rotation for the elements installed in the following stage and thus, make
it possible to avoid these discontinuities.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 23

X1

X1

2
1 1 2

Vy

Vx

Figure 3.1: Discontinuity between two segments [37]

Concluding it can be stated that, as far as time-dependent effects are not considered in the anal-
ysis, it may be possible to achieve almost identical results in the backward and forward analysis
if all these remarks are taken into account. Even nonlinearities, such as P-Delta effects, large
displacement effects and the cable sag effect can be included in both methods.
However, the concept fails when time-dependent effects must be included. Because the de-
termination of concrete creep and shrinkage in modern concrete material models is forward
orientated, by performing a backward analysis the real time-factors can not be taken into con-
sideration. A backward analysis can, therefore, only roughly approximate these effects. Only a
correctly modeled forward analysis can exactly predict the state when the bridge closes in the
middle span and the condition a few years after the completion of the bridge.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 24

3.3 Expanded unit load method

The ”unit load method”, suggested by Bruer et al. [4], Janjic et al. [21] and Pircher [37], is a
novel solution for the derivation of an optimal tensioning sequence of the stay cables of cable-
stayed bridges. At first a desired moment distribution in the final structure under dead load is
defined and then tensioning strategies to exactly achieve this moment distribution are computed.
In contrast to the previous described methods the ”unit load method” takes changes in the struc-
tural system during erection, time-dependent effects and geometrically nonlinear behavior into
account.

The basic idea of the ”unit load method” resembles that of the ”optimization method”. In the
first step a ”unit load case” and the ”ideal moment diagram” for the final structure are defined
(see Figure 3.2(a)). Commonly used degrees of freedom (unit loads) are:

• A unit shortening or a unit tensioning of the cables (see ”force equilibrium method”)

• A unit translation of a rigid support (jacking of a support)

In the displayed example (Figure 3.2(a)) the selected degrees of freedom are tensioning of eight
cables (X1 to X8 ) and jacking of the support (X9 ).

(a)

(b)

Figure 3.2: Unit load cases and desired moment distribution [21]

The ideal dead load bending moment diagram along the girder usually equates that of an equiv-
alent continuous beam (see Section 3.2.1). Certain control sections (i.e. anchorage points of
cables) for the desired moment distribution are defined in the deck and the pylon. In Fig-
ure 3.2(b) the target bending moments (M A , M B , M C , ..., M 1 ) are given in the control points
(A, B, C, ..., 1) along the main girder.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 25

In this example the ”ideal dead load bending moment diagram” is defined for the deck girder in
nine points and also nine degrees of freedom (8 x cable tensioning + 1 x support jacking) are
selected. Setting up the same number of unit load cases as ”fixed moment” points is necessary
to directly solve the problem for the unknown factors of the defined ”unit loads”.

Now, in order to set up the ”unit force equations” , the structure is analyzed for the specific dead
load and a unit load case for each degree of freedom. The results are stored for the bending
moments in the defined control points. Thus, the following system of linear equations can be
established:

M A = MPA + MTA1 · X1 +MTA2 · X2 + ... +MTA9 · X9


M B = MPB + MTB1 · X1 +MTB2 · X2 + ... +MTB9 · X9
:
:
M I = MPI + MTI1 · X1 +MTI2 · X2 + ... +MTI9 · X9

or more compact:

n
X
M K
= MPK + MTKm · Xm (3.13)
m=1

where MPK and MTKm are the moments in point K caused by dead load and by action T m , Tm
signifies each single unit loading case with m ranging from 1 to n (9) and X m is the unknown
multiplication factor for the particular unit load. Solving this system of equations gives the
exact multiplication factors, Xm , for the chosen degrees of freedom, Tm , to achieve the desired
moment distribution.

This example basically corresponds to the procedure of the ”optimization method”, which is
described in Section 3.2.1. The system of Equations 3.13 almost resembles Equation 3.4 in
matrix-notation, where the target values, zk , equate M k − MPk . The reason is that in this simple
example no nonlinearities are included, time-dependent effects are not considered and it is as-
sumed that the tensioning of the cables takes place in the final system. Hence, so far there is no
improvement compared to any of the analysis methods mentioned in Section 1.2.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 26

In order to apply the ”unit load method” in practical design situations, it is expanded to include
the following effects in the determination of the optimum design:

• The construction process

• Second-order theory and large deflections

• Cable sag effects

• Time-dependent material behavior

3.3.1 Construction process

As already mentioned above the continuous change of structural systems during construction
of a cable stayed bridge considerably affects the distribution of internal forces in the complete
structure. Different construction techniques require different tensioning strategies. For exam-
ple, where temporary supports are used, the boundary conditions during erection change and
the final dead load condition may be strongly influenced.
These changes in the structural system during construction of a cable-stayed bridge can easily
be taken into account by analyzing different structural systems according to the various con-
struction stages. This must be done for the dead load as well as for the unit load cases to allow
for the tensioning to take place at the time of installation of each particular cable. When taking
into account the construction process, the unit tensioning of the cables as well as the dead load
of the segments are not applied to the complete structure, but to the different structural systems
which exist at the individual construction stages (see Figure 3.3). In this manner also the influ-
ences that might occur after installation and tensioning of each cable are considered.

Figure 3.3: Unit load cases considering the construction process [4]

The resulting equations and their solution for the required multiplication factors are similar
to the above described simple example. The only difference is that in the calculation of the
moments in point K caused by dead load, MPK , and by the unit tensioning of the cables, MTKm ,
the influence of the different stages is taken into account.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 27

3.3.2 Second-order theory and large deflections

Geometric nonlinearity effects such as P-Delta and large displacement effects are present in
the girder and the tower when subjected to compressive loads and moments simultaneously.
In cable-stayed bridges, most of the permanent load is carried by axial force in the particular
members. Whereas the cable stays are subjected to high pretension forces, the pylon and the
girder receive large compression forces. The relative magnitude of these compression forces
compared to the critical Euler value may cause significant nonlinear effects even if the bending
moments of cable-stayed bridges are relatively small.
There are no specified rules under which circumstances second order theory and large dis-
placement need to be taken into account. In general it can be stated that, because of the huge
axial compression forces accumulating along the girder and the pylon after the completion of
the bridge, P-delta effects should be considered in most cable-stayed bridges. Large displace-
ment effects, in contrast, usually only need to be considered for long-span cable-stayed bridges
erected by cantilever method. This is due to the fact that for small structures the deformations
compared to the overall structural dimensions are small [42], and large displacement usually
only develop in large structures having spans of more than 600m.
However, especially in the case of large structures these nonlinear effects may significantly in-
fluence the behavior of cable-stayed bridges and should therefore be included in the analysis.

3.3.2.1 Second order theory

The second order theory considers the nonlinear axial force- and bending moment-deformation
relationships for the towers and longitudinal girder elements of cable-stayed bridges under com-
bined bending and axial forces. When the deflections in a structural system are assumed to be
small, the axial and flexural stiffnesses of bending members are usually considered to be uncou-
pled. However, when deformations are no longer small, there is an interaction between axial
and flexural deformations in such members.
The P-Delta analysis of a structure accounts for this secondary structural behavior when axial
and transverse (bending) loads are applied simultaneously. The lateral deflection and the axial
force are interrelated so that the bending stiffness depends on the element axial force, and the
presence of bending moments will affect the axial stiffness. In case of element tension forces,
the bending moment of the member due to second order effects reduces (Figure 3.4(a), whereas
compression forces increase the member forces (Figure 3.4(b)). The given effects of a com-
pression or tension force are tantamount to an increase or decrease in the bending stiffness of
the member.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 28

P-Delta effect
ignored

Free body diagram


P-Delta effect
considered

Before deflection After deflection

(a) Column subjected to tension and lateral forces simultaneously

P-Delta effect
Free body diagram ignored

P-Delta effect
considered
Before deflection After deflection

(b) Column subjected to compression and lateral forces simultaneously

Figure 3.4: Column behaviors due to P-Delta effects [30]

As already mentioned above the members of the girder and the tower of cable-stayed bridges
are usually subjected to large compression forces. Hence, when P-Delta effects are taken into
account the bending stiffness of the members of the bridge decreases, and the relationship be-
tween the axial force and the bending moment is nonlinear.
However, because the stiffness matrix depends on the unknown displacements, the decrease
in stiffness of the girder and the tower of cable-stayed bridges can only be determined by an
iterative process. The static equilibrium equations need to be formulated with respect to the
deformed geometry of the structure, which is not known in advance and will change in each
step of iteration until certain convergence requirements are satisfied.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 29

3.3.2.2 Large displacement effects

In a linear structural analysis, it is assumed that geometric changes in the structure during the
application of the design loads, are small, so that the original geometry can be used to compute
the member lengths, member slopes, and load moment arms. In this case the overall stiffness
of the structure in the deformed shape is considered to be equal the stiffness of the undeformed
structure. However, in cable-stayed bridges, as very flexible structures, displacements of several
meters may occur under normal design loads, and accordingly, significant changes in the bridge
geometry may occur. In such a case, the stiffness of the bridge in the deformed shape should be
computed from the new geometry of the structure.

When large displacements of a structural system are considered, the equilibrium conditions
between internal and external forces still have to be satisfied as [54]:
Z
Ψ(a) = B̄T · σ dV − f = 0 (3.14)
V

where Ψ is the sum of internal and external generalized forces, a is the nodal displacement, σ
the stress, and B̄ is the strain-displacement relationship:

 = B̄ · a (3.15)

The bar suffix shows that the relationship is no longer linear since large deformations are con-
sidered. The strain now depends on the displacement as:

B̄ = B0 + BL (a) (3.16)

If strains are reasonably small, then the general elastic relation can be written as:

σ = D · ( − 0 ) + σ0 (3.17)

where D is the elastic stiffness matrix, 0 is the initial strain and σ0 is the initial stress.
In order to solve for a) with the given external load f , the Newton-Raphson method can be
adopted. The procedure of this method is explained in detail by Zienkiewics et al. [57].
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 30

3.3.2.3 Consideration in the ”expanded unit load method”

As a consequence of considering the P-Delta and Large displacement effects by performing a


nonlinear analysis, superposition of the results of the different construction stages is no longer
possible. Rather than accumulating the resulting internal forces, loads must be accumulated and
the influence of previous construction stages can be taken into account as initial displacements
[21]. Thus the Equation system (3.13) also becomes nonlinear.
In order to obtain a solution in nonlinear analysis it is necessary to employ an iterative technique.
A process based on the already mentioned Newton-Raphson has to be implemented to solve
this problem. Generally, as a first approximation, an initial linear analysis of the system can
be performed and the resulting tension forces of the cables can then be applied as starting
values for the iterative procedure. In every iteration step, the tangent stiffness matrix for the ”P-
Delta” effects is updated. The estimated initial tension forces are then improved by transforming
the equation system for the tension forces to define the iterative correction. The procedure is
repeated until convergence is reached.

3.3.3 Cable sag effects

A fundamental problem encountered in cable structures of all types is the nonlinear behavior of
the cable system as a result of the changes in sag and corresponding axial tension. Nonlinearity
in the cables occurs as the load increases and the cable sag decreases producing an increase in
cable chord length with an associated elongation of the cable.
The cables are assumed to be perfectly flexible and possess only tension stiffness; they are inca-
pable of resiting compression, shear and bending forces. When the weight is neglected, a cable
can be considered as a straight member. However, this is not the case anymore when the self
weight of the cable is considered. The only case where an axially loaded cable is without a sag
is when the cable is vertical. As soon as a cable supporting its own uniformly distributed weight
is inclined and the ends are connected to stable anchorages, the cable will sag into a catenary
shape, as illustrated in Figure 3.5.

Since the mathematical expression for a parabolic curve is simple when compared with the
equation for a catenary curve, for stay cables it is usually assumed that the cable has a parabolic
shape. According to Podolny [40] the substitution of the parabolic curve for the catenary curve
of an inclined cable chord in cable-stayed bridges is a good approximation when:
sag
• the sag to span ratio n = l
is less than 0.15,

• the ratio of the horizontal cable component to the total weight of the cable exceeds unity,
H0
W
≥1

• and the inclination of the cable chord is less than 70 degrees.


CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 31

Fig. 1. Plane cable element with sag.

Figure 3.5: Cable element with sag [52]

The general theory of cables is reviewed in various textbooks on structural analysis. Extensive
information on the structural behavior of cables is given in Peterson [36], and a detailed study
on cables and cable systems is provided by Gimsing [15]. In this work only a summary of the
theory and the application to cable-stayed bridges is given.

The cable sag becomes an important factor during construction when the tension in the cables
is temporary reduced. With changing sag, the axial stiffness of a cable will change, and thus,
the stress/strain relationship of a cable is highly nonlinear. A typical load deflection curve of a
cable is shown in Figure 3.6. As illustrated it is convenient to measure elongations from a given
initial condition (i.e. initial dead load condition of the cable). In Figure 3.6 T 0 defines the cable
force in the initial condition and c0 is the corresponding chord length. For higher cable forces
the chord length will increase and for cable forces of less than T 0 the chord length will decrease.

Figure 3.6: Force deflection curve for a cable [15]

In addition to the nonlinear force-displacement relationship of cables nonlinearities in the geo-


metric sense are introduced by large displacements.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 32

In the literature several methods to consider the nonlinear behavior of cables elements have
been proposed by various authors. In this report two of these approaches are demonstrated and
briefly explained.
The first approach assumes that the cable is a straight, inclined member, in which the sag ef-
fect has to be taken into account. For this purpose several investigators suggest the use of an
equivalent straight cable element with an equivalent modulus of elasticity. However, each of
the approaches results in the solution introduced by Ernst [11], who was the first presenting a
concept of a cable equivalent modulus of elasticity, which is based on the parabolic cable curve
but can well describe the catenary action of the cable.
If the change in tension force for a cable during a load increment is not large, the axial stiff-
ness of the cable will not change significantly and the equivalent modulus of elasticity of the
cable can be considered constant during the load increment. The equivalent tangent modulus of
elasticity (see Figure 3.7(a)) is given by the following expression:

E
Eeq,tan = (w·L)2
(3.18)
1+ 12·T 3
· AE
where Eeq = equivalent modulus of elasticity
E = modulus of elasticity of the cable
w = specific weight of the cable, weight per unit volume
A = the cross-sectional area of the cable
L = horizontal projected length of the cable
T = tension force in the cable
If the tension in the cable varies considerably during a load increment, the axial stiffness of the
cable will significantly change and the equivalent modulus of elasticity over the load increment
is given by the secant modulus of elasticity (see Figure 3.7(b)).

E
Eeq,sec = (w·L)2 ·(T1 +T2 )
(3.19)
1+ 24·T12 ·T22
· AE

where the subscripts i and j denote the initial and final value of tension over the load increment.

Figure 3.7: Definition of the tangent and secant modulus of elasticity of a cable [15]
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 33

The cable equivalent modulus of elasticity combines both the effects of material and geometric
deformation. It can be noticed that the value of the equivalent modulus is dependent upon the
weight and the tension in a cable. With an increasing tensile stress or decreasing weight of the
cable, the sag decreases and results in an increase in the apparent axial stiffness of the cable.
Hence, the axial stiffness of the actual cable can be expressed by the axial stiffness of the equiv-
alent element combining cable sag and cable tension determined by the above equations.
Indeed, according to Karoumi [22] the equivalent modulus approach results in softer cable re-
sponse as it accounts for the sag effect but does not account for the stiffening effect due to
large displacements. Thus the linear analysis utilizing the equivalent modulus approach, is not
satisfactory for modern cable-stayed bridges. As those long-span bridges are generally highly
flexible structures, they are subjected to large displacements, and should therefore be analyzed
taking all sources of geometric nonlinearity into account. Furthermore, in the analysis of the
construction process of a cable-stayed bridge, the high stress variations in the cables due to
the lifting of new segments and the installation and pretensioning of new cables, will make it
difficult to calculate a proper equivalent modulus of elasticity.

The procedure presented in Karoumi [22] considers a catenary element which can correctly
model the geometric change of the stay cable at any tension level. In contrast to the first ap-
proach, it is possible to include the effect of pretension of the cable and the exact treatment of
cable sag and cable weight. The complete geometry of the cable, the cable element internal
force vector and the tangent stiffness matrix are determined.

An elastic two-node catenary cable element as illustrated in Figure 3.8 is considered.

Figure 3.8: Catenary cable element [22]


CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 34

The notations of Figure 3.8 are defined as follows:

Lu = the unstressed length of the cable


E = modulus of elasticity of the cable
A = cross sectional area of the cable
w = the weight per unit length

Assuming that the cable is perfectly flexible and Hooke’s law is applicable to the cable material,
the exact relations between the element projections and cable force components at the ends of
the element are:
 
Lu 1 P4 + T j
Lx = −P1 · + · ln (3.20)
EA w Ti − P 2

1 Tj − T i
Ly = · (Tj2 − Ti2 ) + (3.21)
2EAw w
where Ti and Tj are the cable tensions at the two nodes of the element.

By using the following relationships

P3 = −P1 ; P4 = w · Lu − P2 (3.22)

q q
Ti = P12 + P22 ; Tj = P32 + P42 (3.23)

the Expressions (3.20) and (3.21) can be rewritten in terms of the end forces P 1 and P2 only, as:

Lx = Lx (P1 , P2 ); Ly = Ly (P1 , P2 ) (3.24)

Differentiating Equation (3.24) and rewriting the results using matrix notation gives:

∂Lx ∂Lx ∂Ly ∂Ly


dLx = dP1 + dP2 ; dLy = dP1 + dP2 (3.25)
∂P1 ∂P2 ∂P1 ∂P2

 
( ) ∂Lx ∂Lx
 ( ) ( )
dLx ∂P1 ∂P2 dP 1 dP 1
(3.26)
 
=   = F
dLy 
 ∂Ly ∂Ly 
 dP2 dP2
∂P1 ∂P2

where F is the incremental flexibility matrix.


CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 35

The stiffness matrix K is given by the inverse of F as:


" #
k1 k2
K = F−1 = (3.27)
k3 k4

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
k2 = k3 ):
   
−k1 −k2 k1 k2 
 P 1 

 P 

 −k4 k2 k4 
2

Kt =  ; P = (3.28)
 
 −k1 −k2  

 P3 

−k4 P4
 

where
  
1 Lu 1 P4 P2
k1 = − · + · + (3.29)
detF EA w Tj Ti

  
1 P1 1 1
k1 = k 3 = · · − (3.30)
detF w Tj Ti

  
1 Lx 1 P4 P2
k4 = · + · + (3.31)
detF P1 w Tj Ti

     
Lu 1 P4 P2 Lx 1 P4 P2
detF = − − · + · + · +
EA w Tj Ti P1 w Tj Ti
  2
P1 1 1
− · − (3.32)
w Tj Ti

The element tangent stiffness matrix, Kt , relates the incremental element nodal force vector
{∆P1 , ∆P2 , ∆P3 , ∆P4 }T to the incremental element nodal displacement vector {∆u1 , ∆u2 ,
∆u3 , ∆u4 }. To evaluate the tangent stiffness matrix, Kt , the end forces P1 and P2 must be de-
termined first. Those forces are adopted as the redundant forces and are determined, from given
positions at cable end nodes, using an iterative stiffness procedure. This procedure requires
starting values for the redundant forces.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 36

Based on the catenary relationships the following expressions will be used for starting values:
 
w Lx w cosh λ
P1 = − and P2 = · −Ly · + Lu (3.33)
2λ 2 sinh λ

s
L2u − L2y
 
where λ= 3· −1 (3.34)
L2x

In cases where Equation (3.34) cannot be used because the unstressed cable length is less than
the chord length, a conservative value of 0.2 for λ is assumed.
Using Equations (3.22) and (3.23), new cable projections corresponding to the assumed end
forces P1 and P2 are now determined directly from Equations (3.20) and (3.21) and the misclo-
sure vector {∆Lx , ∆Ly } is evaluated as the positions of the end nodes are given. Corrections
to the assumed end forces can now be made using the computed misclosure vectors as:
( ) ( ) ( )i+1 ( )i ( )
∆P1 ∆Lx P1 P1 ∆P1
= K ; = + (3.35)
∆P2 ∆Ly P2 P2 ∆P2

A similar iteration procedure can also be performed to determine the unstressed cable length,
Lu , if the initial cable tension is known instead.

It becomes clear that the adopted catenary elements are efficient for a nonlinear analysis of
cable-stayed bridges. The tangential stiffness matrix, K, can be obtained by an iterative pro-
cess, which takes the effect of pretension of the cable as well as the exact cable sag and cable
weight into account. However, all equations used in described procedure are based on a two-
dimensional catenary element. In most modern analysis programs providing nonlinear analysis
options, the cable stiffness is calculated by similar procedures which are extended for three-
dimensional modeling.

Taking into account the nonlinear behavior of the cable elements in the determination of the ini-
tial tension forces using the ”unit load method” has similar consequences on the basic system
of Equations (3.13) as the consideration of the second order theory and large displacement ef-
fects. Equation system (3.13) becomes nonlinear and can only be solved by an iterative method.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 37

3.3.4 Time-dependent material behavior

Stress and strain in reinforced or prestressed concrete or composite cable-stayed bridges are sub-
jected to change over a long period of time. The effects causing these time-dependent changes
in the structural include:

• Concrete creep strain, which is the time-dependent change in strain under sustained load

• Concrete shrinkage, which is the time-dependent change in strain under constant temper-
ature

• Relaxation of tension cables, which is the loss of prestress in elements when subjected to
constant strain

• Concrete aging - as concrete ages the modulus of elasticity increases, quickly at first and
more slowly as curing slows down
Creep and shrinkage as well as concrete aging effect correspond to a time-dependent change of
the modulus of elasticity of the concrete. The relaxation of the cable stays basically describes
the creep effect of prestressed steel, which causes a loss in tension of the steel cables over time.
However, it has been shown in several investigations that these time-dependent processes can
be treated as nonlinear problems in close analogy to the consideration of second order theory,
large displacements and cable sag effects. In order to take these effects into account during the
determination of the initial cable forces, time functions for the materials need to be employed
and included in the analysis method.

3.3.4.1 Creep of concrete

In concrete or composite cable-stayed bridges creep has a significant influence on the girder
deflection and the cable forces during and after construction of the bridge. In order to reach the
ideal final state, not only the time of completion, but also the final stress state after termination
of the influence of creep has to be included in the analysis. Furthermore, the deformed shape of
the structure in all construction stages, including the stage just before closure, highly depends
on the magnitude of creep deformations as they can be 1.5-3.0 times more than elastic defor-
mations. Because about 50% of the creep deformations take place in the first few months, the
amount of creep deflection can already be significant at the time of completion of the structure.
Considering this fact is particulary relevant when computing the casting curves for the structure.
In order to achieve reliable results for all erection stages, the load history and the casting se-
quence must be taken into account very accurately. Particularly, for cast-in-place concrete can-
tilever bridges, due to the difference of concrete age between the girder segments, creep effects
of concrete are highly influenced by the casting and loading time of the particular segments.
Thus, creep must be considered in detail through the whole construction process and thereafter.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 38

Creep is the time dependent increase in strain of under constant applied loads. The creep de-
formations in a member are, therefore, a function of sustained stress, but there are many other
parameters that influence the effect of creep, including:

• the quality of concrete

• the water-cement ratio of the concrete

• the curing method of the structure

• the ambient temperature and relative humidity

• the aggregates

• the section properties (e.g. average depth etc.)

• the concrete age at load application

• etc.

All these parameters allow predicting the creep phenomenon and, therefore the behavior of a
structure. They are included in most creep formulation to predict creep effects. Brief descrip-
tions of some creep guidelines and specifications are given later in this chapter.
The total creep of the girder of a cable-stayed bridge can be separated in creep effects due to
bending stresses resulting from external loads and creep effects due to axial compression intro-
duced by the stay cables and post-tensioning. In this work the effect of post-tensioning is not
investigated in more detail. However, losses in prestress due to creep are different for tendons
with pre- and post-tensioning.
At each cable anchorage, high compression forces are induced to the bridge deck of a cable-
stayed bridge. Because of these high compression forces compared to relatively low bending
moments the girder of a cable-stayed bridge can usually be assumed to be uncracked during all
construction stages. Hence, creep deformations of cracked sections are also not considered in
this report.

As already mentioned, creep deformations in concrete are a function of sustained stress. As-
suming that the stress is proportional to the strain, the instantaneous strain that occurs during
the application of stress is expressed as follows:

σc (t0 )
c (t0 ) = (3.36)
Ec (t0 )
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 39

where: σc (t0 ) = the concrete stress


Ec (t0 ) = the modulus of elasticity of the concrete at age t0

Under sustained stress , the strain increases with time due to creep and the total strain at time t
can be expressed as follows (see Figure 3.9):

σc (t0 )
c (t0 ) = [1 + φ(t, t0 )] (3.37)
Ec (t0 )

where φ(t, t0 ) = dimensionless creep coefficient


t = the age at loading
t0 = the age for which strain is calculated

The ratio of creep to instantaneous strain, φ, is dependent on the stress level in the material.
It’s value increases with the decrease of age at loading, t0 , and the increase of length of period,
(t − t0 ) during which the stress is sustained. The relationship between the instantaneous elastic
strain and the plastic strain due to creep is basically linear, but the rules to determine the creep
factor, ϕ, are generally very complex.

Figure 3.9: Creep of concrete under the effect of sustained stress [14]

The magnitude of creep can be estimated at various levels. The choice of level usually depends
on demand of accuracy and the quality of the data available for the design. In cases where
only a rough estimate of the creep is required an estimate can be made on the basis of a few
parameters such as relative humidity, age of concrete and member dimensions. On the other
extreme, in the case of deformation-sensitive structures, estimates are based on comprehensive
laboratory testing and mathematical and computer analyses. Ideally, a compromise has to be
sought between the simplicity of the prediction procedure and the accuracy of results obtained.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 40

Several material models for concrete creep and shrinkage have been proposed both in literature
and in national and international codes. Different models for predicting the magnitude of creep
require different input parameters. A few of these parameters influencing the creep deforma-
tion have already been mentioned above. However, there are a lot more variables which can be
separated in intrinsic and extrinsic variables. Intrinsic factors (e.g. aggregate type and water
cement ratio) are fixed once concrete is cast, extrinsic variables (e.g. temperature and relative
humidity) can change through the life span of the structure.
The following eight empirically based models are commonly uses to predict creep strains with-
out the need for creep tests:

• SABS 0100 (1992)

• BS 8110 (1985)

• ACI 209 (1992)

• AS 3600 (1988)

• CEB-FIB (1970)

• CEB-FIB (1978)

• CEB-FIB (1990)

• RILEM Model B3 (1995)

With the exception of the RILEM Model B3 (1995), the models considered derive from struc-
tural design codes of practice and express creep strain as the product of the elastic deformation
of the concrete and the creep coefficient, which has already been demonstrated in Equation
(3.1).
Since the purpose of since work is not the evaluation of different models for predicting creep,
it is not the intention to go into the degree of accuracy of these different creep models in more
detail. In fact, none of these models predict creep effects perfectly, although some formulas are
definitely more accurate than others. They all require a different number of certain parameters
and thus, they are complex or accurate to a greater or lesser extent.

In order to get an overview of the parameters that are relevant for the determination of creep
deformations, Table 3.1 gives a summary of the factors accounted for by the different prediction
methods.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 41

TableTable
3.1: 1Summary
Summary of
of factors
Factors accounted for by
Accounted for byDifferent
differentPrediction
predictionMethods
methods [12]

B3 (1995)
SABS 0100 (1992)

(1978)
CEB – FIP (1970)

CEB – FIP (1990)


AS 3600 (1988)
ACI 209 (1992)
BS 8110 (1985)

RILEM Model
CEB - FIP
METHOD

Aggregate Type X

A/C Ratio X

Air Content X

Cement Content X X
Intrinsic Factors

Cement Type X X X X

Concrete Density X X

Fine/Total Aggregate Ratio (Mass) X

Slump X

W/C Ratio X X

Water Content X

Age at First Loading X X X X X X X X

Age of Sample X

Applied Stress X X X X X X X X

Characteristic Strength at Loading X X

Cross-section Shape X
Extrinsic Factors

Curing Conditions X

Compressive Strength at 28 Days X X X X X X

Duration of Load X X X X X X

Effective Thickness X X X X X X X X

Elastic Modulus at Age of Loading X X X

Elastic Modulus at 28 Days X X X X X X X X

Relative Humidity X X X X X X X X

Temperature X X

Time Drying Commences X


CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 42

Nowadays the most commonly used models for predicting creep and shrinkage are those sug-
gested by the ACI Committee 209 (1992) and the CEB-FIB model code (1990). The details of
the assumptions and expressions of the recommendations are different and will not be discussed
here. The ACI model is simpler but the CEB-FIP (1990) is based on a product approach, which
considers latest researches and includes a certain nonlinearity of creep in case of high stresses.
Many new design codes, such as Eurocode, DIN, etc. are based on the rules of this model code.
Therefore, this concept is also adopted for considering creep effects during the determination
of initial cable forces of cable-stayed bridges. The details of the procedure are explained in
Section 1.3.4.6.

3.3.4.2 Shrinkage

Shrinkage as well as creep highly influences the the girder deflection and the internal forces
during and after construction of a cable-stayed bridge. The axial shortening and bending de-
formations of the girder and the pylon caused by the evolution of shrinkage with time must be
taken into account in order to predict the correct deck geometry and girder stresses at every
construction stage.

Shrinkage is defined as the deformation of concrete in the absence of applied load. There
are four main types of shrinkage in concrete; plastic, autogeneous, carbonation, and drying
shrinkage. Plastic shrinkage is due to moisture loss from the concrete before the concrete sets.
Autogeneous shrinkage is associated with the loss of water from the capillary pores due to the
hydration of the cement. This type of shrinkage tends to increase at higher temperatures and at
higher cement contents. In general, it is relatively small and is not distinguished from shrinkage
caused by drying of concrete. Carbonation shrinkage is caused by the carbonation of hydrated
cement products. This type of shrinkage is usually limited to the surface of the concrete. Drying
shrinkage can be defined as the volumetric change due to drying of the concrete.
The factors that affect the magnitude of shrinkage in concrete are basically the same as the pa-
rameters influencing the effects of creep mentioned above. The only difference is that shrinkage
does not depend on the applied loads; even unloaded elements will shrink.
The major part of axial shortening due to shrinkage is caused by drying shrinkage. When the
change in volume by drying shrinkage is restrained stresses develop. In reinforced concrete
structures, the restraint may be caused by the reinforcing steel, by the supports or by the differ-
ence in volume change in various parts of the structure. Due to tensile stresses caused by the
restraint conditions drying shrinkage can cause cracking in concrete. Thus, drying shrinkage is
related to not only the amount of shrinkage, but also the modulus of elasticity, creep, and tensile
strength of the concrete. All of these properties vary with time, so it is difficult to determine the
cracking tendency of the concrete based solely on shrinkage. Therefore, the modeling of creep
and shrinkage of concrete has to take into account many different variables.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 43

Stresses caused by shrinkage are generally reduced by the effect of creep of concrete. Thus the
effects of these two simultaneous phenomena must be considered in stress analysis. For this
purpose, the amount of free shrinkage and an expression for its variation with time are needed.
Shrinkage starts to develop at time ts when moist curing stops. The strain that develops due to
free shrinkage between ts and a later time t may be expressed as follows:

sh (t, ts ) = sh0 · βs (t − ts ) (3.38)

where: sh0 = the total shrinkage that occurs after concrete


hardening up to time infinity
βs (t0 ) = a function of time which depends upon the size
and shape of the element

As creep and shrinkage cannot be separated from each other, for the prediction of the amount
of shrinkage in concrete, the same prediction models as for creep may be used. In this case
the shrinkage coefficient will also be determined by using the CEB-FIB (1990) model, which is
explained in Section 1.3.4.6.

3.3.4.3 Relaxation of prestressed steel

As already mentioned, the creep effect on prestressed steel is called relaxation. The stress
relaxation in steel is the loss of tension of a member (tendon/cable) when it is prestressed and
maintained at a constant strain and temperature for a long period of time. Prestress loss due
to relaxation varies with the magnitude of initial stress, elapsed time in which the stress is
applied and material properties. An equation widely used for the relaxation at any time, τ , of
stress-relieved wires or strands is:
 
σpr log(τ − t0 ) σp0 σp0
= 1− · − 0.55 , where, ≥ 0.55 (3.39)
σp0 C fpy fpy

where: fpy = the ’yield’ stress, defined as the stress at a strain of 0.01
σp0 = the initial stress
σpr = the stress after loading for a period of time (τ − t0 )
(τ − t0 ) = the period in hours for which for which the member is stressed
C = a product-specific constant, C = 10 for general steel and
C=45 for low relaxation steel are typically used
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 44

3.3.4.4 Concrete aging

Material properties that are influenced by time include the strength f ck0
(t) and the stiffness Ec (t)
of concrete. The strength and stiffness of concrete increase significantly during the first month
after casting; then they increase more slowly over the remainder of life of the structure. Aging
strains are fictitious used in an analysis to allow for time dependent stiffness.
Aging strain, a (t), can be defined as the decrease in elastic strain over time due to the increase
in elastic modulus of concrete (see Figure 3.10).

Figure 3.10: Definition of mechanical and aging strain

The elastic strain, el (t), is a stress originated strain and is the independent variable in the
concrete stress strain relationship:

σ(t) = Ec (t) · el (t) (3.40)

where: σ(t) = the concrete stress at time t


Ec (t) = the instantaneous modulus of elasticity at time t

Aging strain does not represent an actual physical deformation of the concrete and should in-
stead be considered as a correction factor for the calculation of the current total stress as a
function of the current total strain. Under constant stress, the aging strain increment, δ a , oc-
curring between time tn−1 and tn may be expressed as follows:
 
1 1
δa = σ · − (3.41)
Ec (tn−1 ) Ec (tn )
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 45

3.3.4.5 Mathematical modeling of time dependent effects

The total strain, (t), of a uniaxially loaded concrete specimen at the time t after the casting of
the concrete can be found by the superposition of elastic, creep, shrinkage and thermal strains
as follows:

(t) = el (t) + cc (t) + sh (t) + a (t) + T (t) (3.42)

where: (t) = the total strain at time t


el (t) = the instantaneous elastic strain at time t
cc (t) = the creep strain at time t
sh (t) = the shrinkage strain at time t
a (t) = the aging strain at time t
RT
T (t) = T0 α dT , the thermal dilatation at time t
(Assuming a constant temperature, T (t) will not be of interest)

An important assumption made here is that the total strain in the concrete may be considered as
a superposition of all these independent components caused by different phenomena. According
to Bažant [2] this principal of superposition is considered valid if the following conditions are
satisfied:

• the stresses are less than about 45% of the concrete strength

• appreciable reductions in strain magnitude due to unloading do not occur

• no significant change in moisture content distribution during creep occurs

• no large, sudden, stress increase long after the initial loading occurs

Even though in practice all of these conditions may be violated to some extent, it has been exper-
imentally verified that these assumptions are acceptable. Under good design and construction
practices, the first and third condition, which are the most important, are generally true and only
the other two, least important conditions suffer more substantial violations.

The meaning of each of the above listed strain components is illustrated for the case of a struc-
ture under constant stress and no temperature change in Figure 3.11. Even though creep and
shrinkage occur simultaneously in real structures, for practical analysis and design purposes,
elastic shortening, creep and shrinkage are separately considered. Furthermore, in most cases
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 46

Figure 3.11: Time-dependent concrete deformations

the apparent elastic strain, el (t0 ), instead of true elastic strain, el (t), is considered in the anal-
ysis.

Creep strain is dependent on stress and may be graphically depicted by creep isochrones (see
Figure 3.12), which are the lines connecting the values of strain produced by various constant
stresses during a given time interval. Figure 3.12 shows that for stresses up to about 50% of the
concrete strength, the creep is approximately proportional to stress.

Figure 3.12: Creep isochrones [2] Figure 3.13: Typical creep curves

For constant uniaxial stress, σ, the strain may be written as:

(t) = σ · J(t, t0 ) + 0 (t) (3.43)


CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 47

where: (t) = the uniaxial strain


t = the time (normally age of concrete)
J(t, t0 ) = the compliance (creep) function, which represents the strain at time t
produced by a unit constant stress that has been acting since time t 0
0
 (t) = the stress-independent inelastic strain(sh + T )

Due to the proportionality off creep strain to stress, the creep is fully represented by J(t, t 0 ),
the typical shape of which is sketched in Figure 3.13. As shown here, the compliance function,
J(t, t0 ), can be presented by the sum of initial elastic strain and creep strain as follows:

1 1 + φ(t, t0 )
J(t, t0 ) = + C(t, t0 ) = (3.44)
E(t0 ) E(t0 )

where: E(t0 ) = the modulus of elasticity at time of load application


C(t, t0 ) = the creep compliance (specific creep)
φ(t, t0 ) = the ratio of creep deformation to the elastic deformation
(dimensionless creep coefficient, see Section 3.3.4.1)

As already mentioned in Section 3.3.4.1 there are different methods in order to specify the creep
coefficient and the compliance function, J(t, t0 ), resulting out of it. Here the CEB-FIP 1990
model shall be used, which is explained in Section 3.3.4.6.

The stresses and deformations of a cable-stayed bridge continually change during both the con-
struction phase and the service life of the structure. These variations are caused by the structure
modification as well as creep and shrinkage. In order to accurately account for time dependent
variables, a time history of stresses in a member and creep coefficients for numerous loading
ages are required. As a result, the above expressions, which are based on constant stress, must
be adopted to situations of varying stress.
The following outlines a method of calculating creep in which specific functions are numeri-
cally expressed, and stresses are integrated over time. Using superposition, the strain, due to
any stress history σ(t), may be obtained regarding the history as the sum of increments dσ(τ )
applied at increments of time, τ . Equation (3.14) may then be written as:
σ(t) t
∂σ
Z Z
(t) = 0
J(t, τ ) dσ(τ ) +  (t) = σ(t0 ) J(t, t0 ) + J(t, τ ) dτ + 0 (t) (3.45)
σ(t0 ) t0 ∂τ

For the evaluation of this integral-type stress-strain-time relationship, different numerical meth-
ods can be used. In this study, two methods for the time-dependent analysis of structures are
presented, the general step-by-step method and the age adjusted modulus method.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 48

Step-by-step method

In the step-by-step method, a numerical solution technique is employed using the incremental
form of Equation (3.45). The total time is divided into a number of time steps the length of
which should increase with time. Assuming the number of time intervals is i, the discrete
intervals are t0 , t1 , ..., tj , ...ti , where the time ti+1 = t is the time at the end of the last interval.
Figure 3.14 illustrates the notation used for the step by step numerical analysis. The stress
increment which occurs during the j-th time interval, ∆σ j , is assumed to be applied at the
middle of the interval (time tj ). The elastic strain component is calculated at that time with
E(tj ), and creep is determined from time tj on. At the end of the j-th time interval the strain
increment is:

∆j (tj+1 ) = ∆σj · J(tj+1 , tj ) (3.46)

and similarly at the end of the i-th interval the strain increment is:

∆i (ti+1 ) = ∆σi · J(ti+1 , ti ) (3.47)

Figure 3.14: Time-dependent concrete deformations [2]

The total concrete strain at the end of the i-th time interval is obtained by superposition of
the strain increments caused by stress changes in all previous time intervals and by shrinkage.
Equation (3.45) can now be simplified as follows:

i
X
(t) = (ti+1 ) = σ(t0 ) J(t, t0 ) + ∆σj · J(ti+1 , tj ) + 0 (t) (3.48)
j=1
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 49

1+φ(t,τ )
Using J(t, τ ) = E(τ )
Equation (3.48) can also be rewritten as:

i
σ(t0 ) X ∆σj
(t) = · [1 + φ(t, t0 )] + · [1 + φ(t, tj )] + 0 (t) (3.49)
E(t0 ) j=1
E(tj )

If the strain, (t), is known at the end of all intervals in the step-by-step analysis, the stress at
the end of the i-th interval can be calculated by Equation (3.48), provided the stress at the end
of the preceding interval has already been determined.
" i−1  #
E(ti ) X 1 + φ(ti+1 , tj
σi+1 = σi−1 + 0
(ti+1 ) −  (ti+1 ) − ∆σj (3.50)
1 + φ(ti+1 , ti ) j=1
E(tj )

To implement Equation (3.45) using a step-by-step numerical technique, it is required to store


the entire stress history for all elements in the analysis and to retrieve all these stresses for
each new time interval in which the strain increment is to be calculated. This requires an enor-
mous amount of storage space and significant retrieval times. If the creep compliance function,
J(t, τ ), in Equation (3.45) is expressed in a rate-type creep law (i.e. creep law represented by a
system of first-order differential equations), the need for saving the entire stress time history can
be eliminated. Converting the creep compliance function, J(t , τ ), can be done by approximating
it by the degenerate kernel as follows:

n
1 X 1 −(t−τ )/Γµ
(3.51)
 
J(t, τ ) = + · 1 − e
E(τ ) µ=1
Eµ0 (τ )

where: t = the time in days when the creep strain is desired


τ = the time of loading
Eµ (τ ) = coefficients related to the initial shapes of specific creep curves
0

at the loading application time t0 (creep compliance coefficients)


Γµ = constants called retardation times
n = the number of retardation times in the Dirichlet series

When this function is introduced into the superposition integral (Equation(3.45)) the integrand
degenerates into the product of a function of τ and a function of t. The latter function does not
involve the variable of integration and can be extracted from the integral, leaving only an inte-
gration of functions that are independent of t. Thus, at each new time step, it is only necessary
to compute the change in value of the integral from the last time step rather than from the time
of initial loading.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 50

The presented method is greatly affected by the analysis time interval. While creep is consid-
ered exactly with Equation (3.49) for systems having a constant strain, for systems having a
varying strain an error is introduced. The contributions to the strain from the nodal displace-
ments due to creep between subsequent stages are ignored. However, this approximation has
proved to be good if the time steps of an analysis are short enough.
Time intervals for construction stages in general cases are relatively short and hence do not
present problems. However, if a long time interval is specified for a stage, it is necessary to
divide this interval into sub-time intervals to closely reflect the creep effects.

Age adjusted modulus method

Due to aging effect, a reduced creep coefficient must be used to calculate creep strain if stress
is gradually applied. Thus, the creep coefficient, φ(t, t0 ), is replaced by a reduced value which
equals χ(t, t0 ) · φ(t, t0 ), where χ(t, t0 ) is a dimensionless multiplier (smaller than 1) which is
referred to as the aging coefficient.
To explain the aging coefficient concept, the total strain resulting from an initial stress applied
at age t0 , and the subsequent continuously varying stress, Equation (3.45) is expressed in terms
of the creep coefficient, φ(t, τ ).

1 + φ(t, τ )
J(t, τ ) =
E(τ )
Z t
σ(t0 ) 1 + φ(t, τ ) ∂σ(τ )
⇒ (t) = · [1 + φ(t, t0 )] + dτ + 0 (t) (3.52)
E(t0 ) t0 E(τ ) ∂τ

Evaluating the integral of this equation for an assumed variation of stress with time and express-
ing the change of stress as follows:

t
∂σ(τ )
Z
∆σ(t) = dτ = σ(t) − σ(t0 ) (3.53)
t0 ∂τ

where σ(t) is the total stress at time t, we can rewrite Equation (3.45/3.52):

σ(t0 ) σ(t) − σ(t0 )


(t) = · [1 + φ(t, t0 )] + · [1 + χ(t, t0 )φ(t, t0 )] + 0 (t) (3.54)
E(t0 ) E(t0 )

Numerical values for the aging coefficient χ(t, t0 ) can be obtained from the following equation:

E(t0 ) 1
χ(t, t0 ) = − (3.55)
E(t0 ) − ER (t0 ) φ(t, t0 )
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 51

where ER (t0 ) is the relaxation function which can be calculated from the following approximate
relationship between the relaxation function and the creep compliance function (Bazant and
Kim 1979):
 
1 − ∆0 0.115 J(t0 + θ, t0 )
ER (t, t0 ) = − · −1 (3.56)
J(t, t0 ) J(t, t − 1) J(t, t − θ)

where: t0 = age at loading (days)


t = time since casting of the concrete
∆0 = correction factor (∆0 ≈ 0.008)
θ = 21 (t − t0 )

The advantage of both methods for the time-dependent analysis of structures is that they can
deal with any creep function and any prescribed history of stress.

Shrinkage

Shrinkage strain is a function of time, which is independent from the stress in the concrete
member. It is usually about one half of that for creep strain and, hence, can contribute signifi-
cantly to deflection and stress redistribution. Shrinkage strain can be expressed as follows (see
Section 3.3.4.2):

sh (t, t0 ) = sh0 · βs (t, t0 ) (3.57)

The values sh0 and βs (t, t0 ) can be obtained from the CEB-FIB 1990 model code (see Section
3.3.4.6).

Relaxation

Stress relaxation in prestressed elements is the loss of stress under conditions of constant strain.
The magnitude of this loss is dependent on both the duration of the sustained prestressing and
the ratio of the initial prestress to the yield strength of the steel. To calculate relaxation Equation
(3.39) is commonly used.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 52

3.3.4.6 Prediction of creep and shrinkage by the CEB-FIB 1990 model code

The CEB-FIP 1990 prediction model for creep and shrinkage predicts the mean behavior of a
concrete cross-section. It is valid for ordinary structural concrete (12M P a < f ck < 80M P a)
subjected to a compressive stress and exposed to mean relative humidities in the range of 40 to
100% at mean temperatures from 5◦ C to 30◦ C.

Creep strain
Creep strain is by far the most difficult to predict because it is stress originated and depends on
the entire stress history of the concrete as well as other factors.

According to Section 2.1.6.4.3 of the model code, within the range of service stresses |σ c | <
0.4 fcm (t0 ), creep is assumed to be linearly related to stress.
For a constant stress applied at t0 , the creep, cc (t, t0 ), strain at any time t is calculated as
follows:
σc (t0 )
cc (t, t0 ) = · φ(t, t0 ) (3.58)
Eci

where: φ(t, t0 ) = the creep coefficient


σc (t0 ) = sustained stress
Eci = the modulus of elasticity at the age of 28 days

The total stress dependent strain, cσ (t, t0 ), may be expressed as:
 
1 φ(t, t0 )
cσ (t, t0 ) = σc (t0 ) · + = σc (t0 ) J(t, t0 ) (3.59)
Ec (t0 ) Eci

where: J(t, t0 ) = the creep function or creep compliance, representing the total
stress dependent strain per unit stress
Ec (t0 ) = the modulus of elasticity at the time of loading t0
(1/Ec (t0 ) represents the initial strain per unit stress t loading)

According to Section 2.1.6.3 and 2.1.6.1the modulus of elasticity of concrete at an age t 6= 28


days may be calculated as:

Eci (t) = βE (t)Eci = [βcc (t)]0.5 (3.60)


 
1/2
s· 1−( t/1day
28
)
where: βcc (t) = e · Eci
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 53

where s is a coefficient which depends on the type of cement:

s=0.2 for rapid hardening high strength cements RS

s=0.25 for normal and rapid hardening cements N and R

s=0.38 for slowly hardening cements SL

In the CEB-FIP 1990 model code the principle of superposition, as mentioned above, is as-
sumed to be valid. Therefore, the constitutive equation for concrete may be expressed just like
in Equation (3.45).

The creep coefficient, φ(t, t0 ), may be calculated from:

φ(t, t0 ) = φ0 · βc (t − t0 ) (3.61)

where: φ0 = the notional creep coefficient (see Equation (3.62))


βc = the coefficient to describe the development of creep
with time after loading
t0 = the age of concrete at loading (days), adjusted according
to Equation (3.68)

The notional creep coefficient may be estimated from:

φ0 = φRH · β(fcm ) · β(t0 ) (3.62)


with:
1 − RH/100%
φRH = 1+ (3.63)
0.46 · (h/100mm)1/3
5.3
β(fcm ) = (3.64)
(fcm /10M P a)0.5
1
β(t0 ) = (3.65)
0.1 + (t0 /1day)0.2

where: fcm = the mean compressive strength of concrete at age of 28 days


RH = the relative humidity
h = 2Ac /u; the notional size of member (mm), where Ac is the cross
section and u is the perimeter of the member in contact with the
atmosphere
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 54

The development of creep with time is given by:


 0.3
(t − t0 )/1day
βc (t − t0 ) = (3.66)
βH + (t − t0 )/1day

with:
"  18 #
RH h
βH = 150 · 1 + 1.2 · · + 250 ≤ 1500 (3.67)
100% 100mm

The effect of type of cement on the creep coefficient of concrete may be taken into account by
modifying the age of loading, t0 , according to Equation (3.68).
 α
9
t0 = t0,T · + 1 ≥ 0.5 days (3.68)
2 + (t0,T /1day)1.2

where: t0,T = the age of concrete at loading (days) when T = 20◦ C


(for T 6= 20◦ C, t0,T needs to be adjusted according
to Equation (2.1-87) of the CEB-FIP 1990 model code)
α = the power which depends on the type of cement
α = −1; for slowly hardening cement SL
α = 0; for normal or rapid hardening cement N and R
α = 1; for rapid hardening high strength cement RS

Shrinkage strain
In accordance with section 2.1.6.4.4 the total shrinkage strain,  cs (t, ts ), may be calculated from:

cs (t, ts ) = cs0 · βs (t − ts ) (3.69)

where: cs0 = the notional shrinkage coefficient (Equation (3.70))


βs = the coefficient to describe the development of shrinkage with time
(Equation (3.73))
t = the age of concrete (days)
ts = the age of concrete (days) at the beginning of shrinkage
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 55

The notional shrinkage coefficient may be obtained from:

cs0 = s (fcm ) · βRH (3.70)


with:
 
fcm
s (fcm ) = 160 + 10 · βsc · (9 − ) · 10−6 (3.71)
10M P a
"  3 #
RH
βRH = − 1.55 · 1 − , for 40% ≤ RH < 99% (3.72)
100%
βRH = + 0.25, for RH ≥ 99%

where: fcm = the mean compressive strength of concrete at age of 28 days


βsc = coefficient which depends on the type of cement
βsc = 4; for slowly hardening cements SL
βsc = 5; for normal or rapid hardening cements N and R
βsc = 8; for rapid hardening high strength cements RS
RH = the relative humidity of the ambient atmosphere

The development of shrinkage with time is given by:


 0.5
(t − ts )/1day
βs (t − ts ) = (3.73)
350 · (h/100mm)2 + (t − ts )/1day

where h is defined as the notional size of the member as described above.

3.3.4.7 Simulation of creep and shrinkage in the expanded unit load method

According to Janjic et al. [21] time-dependent processes can be treated as nonlinear problems
in close analogy to the procedure outlined under second-order theory and large displacements.
Thus, during the calculation of the structural system under permanent loads creep and shrinkage
can be taken into account and they can be ignored for the unit loading cases. This way it is
possible to consider creep and shrinkage together with the nonlinear behavior of the structure
without increasing the number of iterations necessary to solve the problem.

However, in spite of the above statement it can be shown that the effects of creep and shrinkage
can be treated in a linear manner. In order to predict time dependent internal forces and displace-
ments due to creep and shrinkage Pircher [39] presents a very simple finite-difference-scheme
in the time domain. Through a series of mathematical equations he shows a linear relationship
between the final internal forces and displacements and the elastic strain which initially caused
the creep.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 56

The elastic behavior of a structure is described by the following standard finite element formu-
lation:

qe = K e · a e + f e (3.74)
Z
K =e
[B]T · D · B dV (3.75)
VZ
e
f =− [B]T · D · 0 dV (3.76)
V

where: qe = the vector of nodal forces due to element e


Ke = the stiffness matrix of element e
ae = the vector of nodal displacements
fe = the vector of nodal element forces due to creep
and shrinkage
B = the strain shape function
D = the elasticity matrix
0 = the initial strain vector due to creep and
shrinkage within the time step

The initial strain, 0 , consists of creep, shrinkage and creep based on stress increments arising
during a time step as follows:
X
0 = c,1 + sh + c,3 = φ · el + sh + c,3 = el,j · φj + sh + c,3 (3.77)

The creep factor φ and the shrinkage strain sh are defined by the CEB-FIB 1990 model code
(see above).

Determination of c,3 :
As already mentioned in Section 3.3.4.5 the stresses and corresponding elastic strain generally
do not remain constant during a time step. Assuming a linear variation of the elastic strain
(Figure 3.15) during the time step, a strain increment can be written as:

t t
el (t) = el · (1 − ) + (el + ∆el ) · (3.78)
∆t ∆t
Equation (3.78) can be rewritten using weighting factors w 1 and w2 that define the influence of
the values at the beginning and end of the time step on the average value:

c,el (t) = el · w1 + (el + ∆el ) · w2 (3.79)


CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 57

Figure 3.15: Linear variation of the elastic strain during a time interval

As w1 + w2 = 1 Equation (3.79) can be simplified:

c,el (t) = el + ∆el · w2 (3.80)

The weighting factor, w2 , corresponds to a chosen time stepping strategy - e.g 0.5 for Crank-
Nicolson or 2/3 for Galerkin-scheme.

The contribution of the third part to the initial strain,  c,3 is:

c,3 = ∆el · w2 · φ (3.81)

where φ is the creep factor (CEB-FIB 1990) corresponding to the age of concrete in the actual
time step.

Thus, the total initial strain is:


X
0 = el,j · φj + sh + ∆el · w2 · φ (3.82)
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 58

According to the basic finite element formulation the nodal displacements are related to the total
strain as follows:

ae · B = ∆el + 0 (3.83)

Substituting Equation (3.82) in Equation (3.83) results in:


X
∆el = ae · B − ( el,j · φj + sh + ∆el · w2 · φ)
X
∆el · (1 + w2 · φ) = ae · B − ( el,j · φj + sh ) (3.84)

By substituting Equation (3.83) and Equation (3.84) into the basic finite element formulation
(Equation (3.74)) the recursive formula in Equation (3.85) can be derived, which is described
in detail in [39].

1 1
Z X
e
q = · K · ae − · [B] · D · ( j · φj + sh ) dV (3.85)
1 + w2 · φ 1 + w2 · φ

Equation (3.85) shows that in order to get the solution for a single time step the same finite
element formulation as for normal ”static” analysis can be used with the exception that the elas-
ticity matrix D is replaced by 1/(1 + w2 · φ) · D. Hence, all creep and shrinkage influences
on the final distribution of internal forces and displacements are related to the elastic strain in a
linear manner.
To set up the stiffness matrix of a member it is only necessary to multiply the elastic modulus
by the factor 1/(1 + w2 · φ). To set up the load vector, one can multiply the strain values re-
lated to the individual load increments by the corresponding creep factors, add the strain due to
shrinkage and calculate the load vector.

The consequence of this linear relationship between the final internal forces and displacements
and the elastic strain is that the principles of linear superposition may be applied. Thus, the
total creep occurring during a single time step may be decomposed into single contributions.
For example the creep moment at one of the ideal moment positions (Equation (3.13)) can be
written as:

Mcreep = MP + Mct=1 · X1 + Mct=2 · X3 + Mct=13 · X3 + ... (3.86)

Mcreep at any location therefore consists of one part which is related to the permanent load and
the other parts are related to the unit loads. The unit loading cases are linearly coupled to the
same unknown factors X1 .....X9 as described in Section 3.3.1.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 59

As for the simple example described in Section 3.3 the basic concept of solving Equation (3.13)
for the multiplication factors, X1 , X2 ..., X9 , can now be applied. The effects of creep are sep-
arated into contributions of each time interval and then summed up. The system of Equations
(3.13) therefore remains linear when including time dependent effects. The only approximation
is that the behavior within a single time step is linear (see Figure 3.15).

3.3.5 Implementation of the expanded unit load method into finite ele-
ment software

The ”expanded unit load method”, as described above has been incorporated into the computer
program RM2004. The function to optimize the initial cable forces is called the AddCon-
method. It uses an iterative process to appropriately factor the user-defined unit loading cases
such that in combination with a fixed loading case certain constraints are achieved. The pur-
pose is to find the best solution for some given design criteria. Generally the constraints can
be applied to any ”calculation results” (deformations, stresses, forces, etc. at certain locations),
but also other constraints such as geometric parameters, material properties, etc. can be used
as design criteria (constraints). However, one has to pay attention that the assortment of the
degrees of freedom and the final target condition does not result in singularity. The design cri-
teria specified must be numerically achievable i.e. if a moment is specified as the constraint
then there must be at least one variable force on the relevant side of it such that when this is
appropriately factored and combined with the ”fixed loading case” the defined constraint can be
achieved. Furthermore, there must be as many user defined design criteria (constraints) as there
are unknown variable loading cases to solve the problem.

3.3.5.1 Optimization for non-linear problems

As the solution for the linear optimization problem is obvious in this section the iterative so-
lution for nonlinear optimization is derived. The procedure is valid for nonlinear effects that
produce a solution not to far from the linear solution [37].

The linear expression to determine the initial cable forces introduced at the beginning of Section
3.3 can be written as:

Einput = A · X + Econst (3.87)


CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 60

where: A = the influence matrix


Einput = the vector of input results (i.e. desired constraints)
Econst = the vector of constant results
(i.e. deformations, stresses, forces caused by dead load)
X = the vector of unknown multiplication factors

In order to cover the nonlinear effects in Equation (3.87), the matrix A is split into a linear and
a nonlinear part:

A = ALIN + ∆A (3.88)

Equation (3.87) changes to:

Einput = [ALIN + ∆A] · X + Econst (3.89)

The nonlinear matrix ∆A may include structural nonlinearities as well as nonlinearities result-
ing from time effects and the change of the structural system over time.

The nonlinear part of matrix A multiplied by the multiplication factors can now be added to the
constant results. Thus, the constant results get quasi-constant results marked with an asterisk:

∗Econst = ALIN · X + Econst (3.90)

Equation (3.89) changes to:

Einput = ALIN · X + ∗ Econst (3.91)

The difference of the constant parts of the results is now a measure for the nonlinear part of A.

Eerr = ∗ Econst − Econst (3.92)

Linearization of the nonlinear part of A:

The matrix A generally depends on the multiplication factors X i , but this dependency is not
given directly. One way of describing it mathematically is to produce changes in X and to
watch the corresponding changes in E.
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 61

 
∂A11 ∂A12 ... ∂A1n
∂X1 ∂X2 ... ∂Xn
 
 
     
δE1 
 ∂A21 ∂A22 ... ∂A2n  
  δX1 

 δE2   ∂X1 ∂f X2 ... ∂Xn    δX2


= · (3.93)
   
   
 ...  
 ... ... ... ...  
   ... 


δEn   δXn
 
 
 ∂An1 ∂An2 ... ∂Ann
 

∂X1 ∂X2 ... ∂Xn

Each design step results in one set of E and X. After n steps, there are (n-1) sets of δE and δX.
The linear change of matrix A can be estimated by the solution of the following Equation:
 
∂A11 ∂A12 ... ∂A1n
∂X1 ∂X2 ... ∂Xn
 
 
  
δE11 δE12 ... δE1n δX11 δX12 ... δX1n
  

 ∂A21 ∂A22 ... ∂A2n 

 δE21 δE22 ... δE2n   ∂X1 ∂f X2 ... ∂Xn   δX21 δX22 ... δX2n 
= · (3.94)
     

 ... ... ... ...  
 ... ... ... ...  
 ... ... ... ... 
δEn δEn2
1
... δEnn 


 δXn δXn2
1
... δXnn
 ∂An1 ∂An2 ... ∂Ann
 

∂X1 ∂X2 ... ∂Xn

The unknown nonlinear part of A can be computed by solving the linear system:
 
∂A11 ∂A12 ... ∂A1n
∂X1 ∂X2 ... ∂Xn
 
 
   −1
δE11 δE12 ... δE1n δX11 δX12 ... δX1n
 

 ∂A21 ∂A22 ... ∂A2n 

 ∂X1 ∂f X2 ... ∂Xn   δE21 δE22 ... δE2n 
  δX21 δX22
 ... δX2n 
= ·  (3.95)
   


 ... ... ... ...  
 ... ... ... ...   ... ... ... ... 



 δEn δEn2
1
... δEnn
δXn δXn2
1
... δXnn
 ∂An1 ∂An2 ... ∂Ann
 

∂X1 ∂X2 ... ∂Xn

Iterative search for solution

In order to start an iterative process to solve the nonlinear problem, at first approximate val-
ues for the multiplication factors Xi are determined from a linear calculation. Based on these
starting values the calculation of construction process is repeated and the starting values are
improved . Each full calculation loop provides one additional piece of numerical information
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 62

about the nonlinear behavior. As the number of iterations grows, more and more data is avail-
able to find the proper nonlinear part of A.

Figure 3.16 shows a flowchart of the iteration procedure in the AddCon-Method. More infor-
mation about the iteration procedure can be found in [37].

Figure 3.16: Flowchart of the iterative solution for nonlinear optimization (AddCon-Method)
CHAPTER 3. STRUCTURAL ANALYSIS OF CABLE-STAYED BRIDGES 63

3.4 Camber control

As already mentioned above the purpose of tensioning the stay cables with their optimum ten-
sion forces can either be to achieve a desired final geometry or a desired final moment/stress
distribution in the bridge girder and pylon. However, the optimum would be to achieve both
objectives simultaneously.
In general, especially when considering creep effects in the analysis of cable-satayed bridges,
the main intention of adjusting the cable stresses should rather be to achieve a designated mo-
ment distribution than to minimize displacements. In this case the final geometry of the bridge
can still be achieved by defining an appropriate precamber.
A structure is said to be cambered when so constructed that it assumes a desired theoretical
form for a defined condition of load. Concerning a cable-stayed bridge the defined load con-
dition is the dead load and the stay cable forces determined from the optimization of the cable
tensioning for an optimum final moment distribution. The desired theoretical form is to achieve
zero displacements at certain control points of the girder and the pylon under this condition of
load.
The precamber that ensures that the final bridge profile will coincide with the desired profile is
called a fabrication camber. It gives information about the angle between two segments that is
needed to balance the displacements of the structure.
The fabrication camber of a cable-stayed bridge can easily be determined when the deflected
shape of the structure under dead load, construction load and stay forces is known after all
deflections (elastic + time-dependent) have occurred. When zero displacement is desired, it
simply results as the curve between the up-side-down final displacements of all joints between
two segments.
Chapter 4

Comparison of Structural Analysis


Programs

4.1 Introduction

The analysis of cable-stayed bridges can be performed by different finite element analysis soft-
ware. In this chapter the following three structural engineering software programs with a par-
ticular emphasis on bridge structures are introduced.:

• RM2004 (Austria)

• Larsa2000 (USA)

• MIDAS/Civil (Korea)

They all provide many features for the analysis of cable-stayed bridge structures, such as the
construction stage analysis feature which allows for changes to the structure over time, the
possibility to take the effects of geometric and material nonlinearities into account, the con-
sideration of time dependent material properties in the analysis and so on. Larsa2000 and
MIDAS/Civil are used for an example calculation of a simple cable stayed bridge in Chapter 6.
The functionality of these programs is described in detail in Section 4.2 and 4.3. Unfortunately,
the structural engineering software program RM2004 was not available for this work. However,
a brief introduction of RM2004 is given in the following.

RM2004 is a structural engineering software developed by TDV (Technische Datenverarbeitung


Gesellschaft), an Austrian software development company. It is capable for the analysis and
optimization of cable tensioning of cable stayed bridges. The ”expanded unit load method”,
as described in Chapter 3.3, has been developed by TDV and fully incorporated in the analysis
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 65

program RM2004. As already mentioned in Chapter 3.3.5, the feature to optimize design pa-
rameters with RM2004 is called the AddCon-Method (Additional Constraint Method). It allows
for the consideration of the construction sequence, nonlinearities and time-dependent effects in
the optimization process.
In order to reach the ”ideal state” the AddCon-Method uses an iterative process to appropriately
factor one or several user-defined unit loading cases such that in combination with a fixed load-
ing case a set of user-defined design criteria are achieved. The user-defined constraints can be a
set of forces/moments, stresses or displacements or a combination of these at defined locations
on the bridge [37].
However, because RM2004 was not available during the process of performance of this thesis,
it is not possible to comment or judge on the reliability of this analysis feature. The analysis
programs Larsa2000 and MIDAS/Civil, in contrast, were present during the period of this work
and are therefore described and commented in the following.

4.2 General description of Larsa2000 and MIDAS/Civil

Larsa2000 and MIDAS/Civil are finite-element-based analysis and design software for struc-
tural engineering that offer special features for the analysis of any bridge structure. They pro-
vide 3D analysis engines allowing the simulation of any erection procedure linking it with the
time axis in the calculation (4th dimension). Thus, for cable-stayed bridges erected by the can-
tilever construction the changes to the structure over time can be considered in the analysis.
Furthermore, Larsa2000 as well as MIDAS/Civil offer a function to consider geometric and
material nonlinear effects. Indeed, concerning the optimization of the cable tensioning strategy
both Larsa2000 and MIDAS/Civil are less developed analysis programs than RM2004.

4.2.1 Larsa2000

Larsa2000 is a structural engineering software program developed by the American software


company Larsa, Inc. In this thesis the version 6.09.03 of Larsa2000/4th Dimension is used,
which is a version of Larsa2000 that has not yet been released officially. When the actual ver-
sion 6.08.76 of Larsa2000 was used with a German operating system a localization conflict
between the Larsa modules occurred. The problem was because of multiple-language support
of Larsa2000. Somehow the localization on a German computer had an effect on the input file
(time-dependent options). Instead of English, the engine replaced some of the time-dependent
options in the input file with a German equivalent. Then the Larsa engine looked for ”true”, did
not see it and ignored the time effects.
Since the main concern of this thesis is the time-dependent analysis of cable stayed bridges,
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 66

using a version of Larsa2000 which is not capable for time-dependent effects is not useful. For-
tunately, promptly after the problem appeared, the support team of Larsa, Inc. provided the
newer version 6.09.03 of Larsa 2000/4th Dimension where some major improvements in terms
of localization have been included.

Larsa was originally developed to perform nonlinear static analysis of structures that have large
displacements such as suspension and cable-stayed bridges. However, Larsa2000 has been
improved and Larsa2000/4th Dimension includes new analytical features. The key features for
the analysis and design of bridges are:

• Staged construction analysis

• Time-dependent material properties

• 3D tendons for post-tensioning

• Influence line generator for moving load

The first two features are described in detail in Section 4.3 and are used in the analysis of a
simple cable-stayed bridge in Chapter 6. The latter two are very advantageous for modeling
a real bridge structure, but they are not examined in this work. Detailed information about
Larsa2000/4th Dimension is given in the Larsa2000 Reference Manual [25] and User’s Guide
[26].
Finally it can be stated that the Larsa engine is powerful, using both tangent stiffness and the
full Newton-Raphson method with iterations.

4.2.2 MIDAS/Civil

The analysis program MIDAS/Civil has been developed by the Korean structural software de-
velopment company MIDASSoft, Inc. As well as Larsa2000 it has been used over years, and
the reliability has been established through applying them to thousands of real projects. In this
thesis the version 6.3.3 of MIDAS/Civil is used, which has been released in 2004.
MIDAS offers many special features for the analysis and design of bridges. These include:

• Construction stage analysis

• Time-dependent analysis feature

• Solution for unknown loads using optimization technique

• Analysis of prestressed concrete box girders


CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 67

• Moving load analysis for bridge structures

• Hydration heat analysis

• Composite steel bridge analysis considering section properties of pre- and post-combined
sections

• ILM/MSS/FCM bridge wizard (automatic generation of the model data of a prestressed


or post-tensioned box bridge constructed by: incremental launching method (ILM), a
movable scaffolding system (MSS) or by free cantilever method (FCM))

• Cable-stayed bridge wizard (automatic generation of two-dimensional cable-stayed bridges)

Again, it is not the intention to dwell on all these features. In Section 4.3 the first three special-
ities for bridge analysis are explained and judged. For further information on the other features
please refer to MIDAS/Civil Analysis Reference [30] or the Online Manual [31].

4.3 Construction stage analysis by Larsa2000 and MIDAS

As mentioned above, both Larsa2000 and MIDAS/Civil offer a construction stage analysis fea-
ture, which allows for the consideration of the continuous change of structural systems during
construction of a cable stayed bridge.
In both analysis programs changes to a structure over time are defined in a series of construc-
tion stages. Construction stages can include construction and deconstruction activities, such as
constructing or removing parts of the structure, applying loads, modifying support conditions
and stressing or slackening tendons or stay cables.

Using Larsa2000 or MIDAS/Civil a linear or nonlinear construction stage analysis can be per-
formed, which is described in detail in Section 4.3.3.

In reality the structural behaviors such as deflection and stress redistribution continue to change
during and after the construction due to varying time-dependent properties. Larsa2000 as well
as MIDAS/Civil have the ability to include time-effects on material, such as concrete creep,
shrinkage, elastic modulus variation (aging) and steel relaxation. The way of including time-
dependent effects in a construction stage analysis by Larsa2000 and MIDAS/Civil is explained
in Section 4.3.4.
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 68

4.3.1 Definition of construction stages

4.3.1.1 Larsa2000

Before defining construction activities in Larsa, all elements in the structure that will ever be
assembled must be modeled and restraints must be set to initial conditions. When modeling
a cable-stayed bridge the elements will mainly be beam elements for the bridge deck, cable
elements for the stay cables and tendon pseudoelelements for post-tensioning of the bridge.
The beam element has three translational and three rotational degrees of freedom at each end
joint. It includes axial, shear, twisting and bending deformations. The beam element is capa-
ble of exactly representing constant axial deformation along the beam with constant torsional
shear deformation and linear bending deformations within the element. This is sufficient for
analyzing structures with loads applied at joint points. However, when subjected to axial loads,
torsional loads, lateral loads, or moments along the element, modeling these deformations re-
quires a higher order representation. The use of the fixed end forces allows these higher-order
deformations without the need of additional degrees of freedom for the beam elements. The
beam element in Larsa2000 has geometric nonlinearity and and stress-stiffening in all nonlin-
ear analysis types and includes P-Delta properties.
The cable element has only translational displacements in X, Y and Z directions at each end
and it has no rotational stiffness. Cable elements can carry axial tension force only. The cable
element can only be used in a nonlinear analysis. If during the analysis, the axial force in a cable
element becomes compressive, then the cable element is assumed to have no axial stiffness and
cannot carry any load. If a cable element is used in a linear analysis it is considered as a truss
element which can carry axial tension and compression forces. As well as the beam element,
this element has geometric nonlinearity and stress-stiffening in all nonlinear analysis types.
Tendons are used for post-tensioning members. In Larsa, they can be modeled with both short-
term losses due to friction and anchorage slip and long-term losses, such as relaxation, elastic
shortening, creep and shrinkage, etc. Tendons are pseudoelements that are modeled along side
the other geometric elements of a model. They take any arbitrary path in three dimensions. The
path is determined through control points at which a location and curvature type are specified.
The stressing of the tendons takes place during the construction process. However, in this work
only simple cable-stayed bridge structures are considered and the presence of post-tensioning
is neglected. For more information about post-tensioning with tendon elements in Larsa2000,
refer to the Larsa2000 Reference [25] and the LARSA 2000 User’s Guide [26].

After preparing the structural model, groups of elements that will be constructed or decon-
structed together need to be put into structure groups. Then, in addition to the model data, all
loads that will be applied during the construction process of the structure have to be defined.
Finally, in order to run a construction stage analysis by Larsa2000, the construction activities
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 69

during the erection of the structure need to be composed. When a staged construction analysis
begins, Larsa assumes no elements have been activated. Only elements explicitly constructed
contribute to the stiffness of the structure.

Construction activities in Larsa2000 are arranged in stages and steps. A stage represents one day
of construction, which can consist of one or more steps. Stages are labeled with a day number,
a temperature and a humidity value, which apply to all steps within the stage. A construction
step represents one set of construction activities that can be:

• Construction and deconstruction of element groups

• Support changes (Joints are restrained or released)

• Changes in slave/master connectivity

• Load application

• Tendon application

When a support is added to the structure it can be activated at the actual (deformed) location
of the particular joint (fixed) or at the original position defined in the input data (hoist). When
choosing the hoist command for adding an additional support a forced displacement in the di-
rection opposite to the displacement of the preceding construction stage is applied to the node.
Thus, additional internal stresses resulting from this forced displacement are introduced to the
structure.

After all activities during the construction of a cable-stayed bridge are set up in construction
stages/steps, a construction stage analysis of the structure can be performed.

4.3.1.2 MIDAS/Civil

Similar to the definition of construction stages by Larsa2000, before defining construction ac-
tivities a structural model including all elements, boundary conditions and loads that will ever
be applied need to be prepared. The main element types used in the staged construction analysis
of a cable stayed bridge are the same as explained above, but MIDAS/Civil does not consider
tendons as independent elements. However, MIDAS/Civil also offers a feature for pre- or post-
tensioning members that allows for the consideration of prestress losses, etc. More information
on the prestressed concrete analysis can be found in the MIDAS/Civil Analysis Reference [30]
or in the Online Manual [31].
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 70

Each node of a MIDAS/Civil beam element retains six degrees of freedom (3 translational + 3
rotational). The formulation of the beam element is founded on the Timoshenko Beam Theory
taking into account the stiffness effects of tension/compression, shear, bending and torsional
deformations.
The cable element in MIDAS/Civil is a tension-only, three dimensional line element, which is
capable of transmitting axial tension force only. As already mentioned in Chapter 3.3.3 a ca-
ble element reflects the change in stiffness varying with internal tension forces. In a geometric
nonlinear analysis, the tangent stiffness of a cable element is calculated considering a catenary
element which can correctly model the geometric change of the cable at any tension level. This
procedure is explained in detail in Chapter 3.3.3. When performing a linear analysis, the cable
element is automatically transformed into an equivalent truss element with the stiffness result-
ing from cable sag calculated by the ”Ernst formula” given in Equation (3.18).

After the preparation of the model and loading data Structure, Boundary and Load Groups that
will be constructed or applied together need to be defined. Each construction stage in MI-
DAS/Civil is then composed of the activation or deactivation of one or more of those groups.
In contrast to construction stages in Larsa2000, a construction stage in MIDAS/Civil represents
a number of days that needs to be specified by the user. The construction or deconstruction
of geometry and boundary conditions always takes place at the first day of the particular con-
struction stage and then remains unchanged, whereas loadings may be applied at the first, last
or any intermediate day and may change within the stage. However, additional construction
steps within the stage need to be defined if loadings shall be applied at specific days during the
runtime of the stage.
The actual date of a particular construction stage in MIDAS/Civil is determined by the sum
of the duration of the preceding stages. The temperature and relative humidity is not defined
for the construction stages, but for the entire construction stage analysis which applies to all
construction stages. Hence, changes in temperature and relative humidity during a construction
process can not be considered in a construction stage analysis by MIDAS/Civil.

The contents included in each construction stage in MIDAS/Civil are:

• Activation (creation) and deactivation (deletion) of elements

• Activation and deactivation of loadings at certain points in time

• Changes in boundary conditions

When a boundary is added in MIDAS/Civil similar to Larsa2000 the user has the choice of
activating this boundary condition at the original (undeformed) location or at the deformed
location of the particular node. The influence on the internal forces are the same as explained
above.
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 71

4.3.2 Accounting for deformation

The problem of a discontinuity at the joints between two segments (construction stages) in a
forward analysis has already been briefly discussed in Chapter 3.2.4.2. Generally, when new
segments are activated the joints enter at the location that was initially given to them in the input
data. This way a displacement (δz , δx and θz ) is automatically introduced in the structure. As
already mentioned in Chapter 3.2.4.2, one way to avoid these discontinuities is to rather apply
the new segments tangentially than straight forward. Larsa2000 as well as MIDAS/Civil offer a
function, which calculates the real displacement as well as the rotational angle for the elements
installed in the following stage.

In MIDAS/Civil using the option initial tangent displacement for erected structures for the con-
struction stage analysis the real displacement may be calculated for all members or for specific
groups only. If this option is not chosen all segments will be applied straight forward instead of
tangentially and a discontinuity is introduced at the joints between the segments.

Larsa2000 offers a choice of the method of construction for each construction activity, i.e. for
each construction step. It allows for three different methods of segmental construction updating
the locations of all joints becoming active in the step. In addition to the straight forward and
tangential activation of segments it also offer the option to install the next segment by shifting
the end node down to match the translational displacement of the start node. This way the
segment remains horizontal as it was initially drawn.
The example below shows the difference between the special segmental construction methods
used in Larsa2000. It is a simple segmental bridge built in stages. In the first stage, only the
left segment is activated. After the application of loads, the segment deforms. Then the second
segment is installed. Figure 4.1 shows the behavior of the two segments when using standard,
hinged or matched casting of segments.

• Standard cast
The center node is displaced, but the right node is left unchanged because inactive joints
are ignored in the analysis. As a result, when the next segment is constructed, the seg-
ments form an angle, even though there was no angle when the segments were initially
modeled.

• Hinged cast
The second segment is constructed by shifting the end node down to match the transla-
tional displacement of the center node.
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 72

• Matched cast
The second segment is constructed by shifting the end node down to match the angular
displacement of the center node, i.e. the segment is installed tangentially.

Figure 4.1: Different segmental construction methods in Larsa2000

In addition to these three methods joints sometimes need to become active in a location relative
to the deformed structure of the model, rather than in an exact position known ahead of time. In
Larsa2000 displacement initializations may be defined that specify how to place a joint relative
to the deformed location of other joints. They are explained in more detail in the Larsa2000
Staged Construction Analysis Manual [27].

4.3.3 Consideration of nonlinearities

Larsa2000 as well as MIDAS/Civil offer to take into account structural nonlinear behavior in the
construction stage analysis. The nonlinearity considered in a nonlinear construction stage anal-
ysis can be classified as material nonlinearity or as geometric nonlinearity. Material nonlinearity
is associated with changes in material properties: inelastic behavior. Geometric nonlinearity is
associated with changes in configuration, such as P-Delta effects and large deflections of cable-
stayed bridges.
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 73

4.3.3.1 Nonlinear construction stage analysis by Larsa2000

In principle, the staged construction analysis by Larsa2000 is a form of nonlinear static analysis
that retains the state of the structure from step to step. Performing a linear construction stage
analysis is not possible when using Larsa2000. This nonlinear analysis automatically takes
into account material nonlinearity (if elements with nonlinear properties are used) and stress-
stiffening effects such as that encountered when a cable is stressed in tension or in compression
and offers the choice whether geometric nonlinearity shall be included in the analysis. The
nonlinear construction stage analysis is performed by iteratively solving for the displacements
of a structure, stopping when convergence criteria have been met.

Deformations can significantly alter the location or distribution of loads. Therefore, equilibrium
equations in Larsa2000 are written with respect to the deformed geometry, which is not known
in advance. One seeks a displacement state in which the deformed structure is in equilibrium
with the loads applied to it.

Geometric nonlinearity in Larsa2000 includes both second-order theory and large displacement
effects. The material nonlinearities that can be considered in Larsa2000 include the inelastic
beam element, the nonlinear elastic and nonlinear inelastic springs, gap and hook elements,
foundation springs with nonlinear elastic or inelastic material behavior, etc.

When geometric nonlinearity is included in the nonlinear construction stage analysis of a cable-
stayed bridge, Larsa2000 takes into consideration:

• Large displacements (The stiffness matrices of the elements are based on deformed ge-
ometry)

• Second-order theory

• Stress stiffening due to the internal axial forces in cable elements

However, Larsa2000 does not consider the sag effect of cable elements. The only possibility to
capture the sag effects is to divide the cables into a number of elements. If only one element
per cable is used, its tension stiffness is similar to the equivalent truss element.

In a nonlinear analysis, the equilibrium equations are formulated with respect to the deformed
geometry of the structure, which is not known in advance and will change with the applied
loads. The stiffness matrix depends on the unknown displacements. The result is a set of
nonlinear simultaneous equations. It is necessary to employ an iterative technique to obtain a
solution in nonlinear analysis. An iterative scheme based on Newton-Raphson Method has been
implemented in Larsa2000.
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 74

4.3.3.2 Nonlinear construction stage analysis by MIDAS/Civil

The nonlinear analysis by MIDAS/Civil can be classified in three main categories. First, mate-
rial nonlinear behaviors are encountered when relatively big loadings are applied to a structure
thereby resulting in high stresses in the range of nonlinear stress-strain relationship. Second, a
geometric nonlinear analysis including second-order theory and large displacement effects and
third, boundary nonlinearity of a load-displacement relationship that may occur when boundary
conditions change with structural deformations due to external loads.

In principle, using MIDAS/Civil a construction stage analysis is a linear static analysis. How-
ever, to include geometric nonlinear behavior the analysis option include non-linear analysis
must be activated. If this option is checked in a construction stage analysis of a cable-stayed
bridge, MIDAS/Civil considers large displacement effects as well as the nonlinear behavior of
cable elements including cable sag, but P-delta analysis has not been added to the program yet.

Similar to Larsa2000 the iterative technique to obtain a solution in a geometric nonlinear anal-
ysis is based on Newton-Raphson Method, which is explained in detail in the MIDAS/Civil
Analysis Reference [30].

4.3.4 Consideration of time-dependent material behavior

Deflections and stress redistributions continue to change during and after the construction of
cable-stayed bridges due to varying time-dependent properties, such as concrete creep, shrink-
age, modulus of elasticity (aging) and tendon relaxation. The importance of the consideration of
time-dependent material behavior in the analysis and design of cable-stayed bridges has already
been explained in Chapter 3. Both Larsa2000 and MIDAS/Civil allow for performing a time-
dependent construction stage analysis taking into account different time effects on materials.
The procedure in the different analysis programs is explained in detail in the following.

4.3.4.1 Time-dependent construction stage analysis by Larsa2000

Concrete is subject to creep and shrinkage. Steel is subject to relaxation. These changes and
elastic modulus variation are accounted for in Larsa’s time-dependent staged construction anal-
ysis.
Time-dependent staged construction analysis in Larsa2000 requires time, temperature, and hu-
midity conditions, that are defined for all construction stages (see section 4.3.1.1). The day of
a particular construction stage is used for computing time-effects on materials. Concrete creep
and shrinkage, tendon relaxation and the time-effect on elastic modulus in Larsa2000 are based
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 75

on the prediction models and equations from CEB-FIP 1990 model code and CEB-FIP 1978
model code. They can be included in the analysis separately or in combination. Which effects
to include is specified by the user.

When the CEB-FIP 1990 model code is chosen for a time-dependent staged construction anal-
ysis, the following input values must be specified to allow Larsa2000 to determine creep and
shrinkage behavior and the time-effect on the elastic modulus of concrete:

• Environmental conditions (relative humidity and temperature which is defined for each
construction stage)

• Compressive strength of concrete at day 28 (Larsa requires the mean compressive strength,
fcm , instead of the characteristic compressive strength, fck )

• Cement hardening type (slow, normal, rapid or rapid high strength)

• Cross-sectional area and perimeter of the member (→ notional size of member)

• Casting day of all members (the age of a member when constructed is determined by
subtracting the casting day from the day of construction set for the construction stage
being analyzed)

Based on these input parameters together with the loading age (day of load application sub-
tracted by casting day) and actual time defined in the construction stages, Larsa2000 automati-
cally computes the creep and shrinkage coefficients.
In addition to those input parameters that allow for the computation of creep, shrinkage and
variation of elastic modulus with time by built-in equations of the CEB-FIB 1990 model code,
Larsa2000 requires the following time-dependent material curves in order to take steel relax-
ation into account :

• Stress/GUTS vs. Relaxation Curve (Relaxation losses for different stress levels in the
tendon. Stress/GUTS is the stress in the tendon divided by the guaranteed ultimate tensile
strength)

• Time vs. Relaxation Curve (Time versus Relaxation is a function corresponding to how
relaxation varies over time)

Relaxation is derived from the product of the coefficients from the Stress/GUTS vs. Relaxation
and Time vs. Relaxation curves. The coefficients are the values of the curves with time being
the number of days since the tendon was stressed.
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 76

4.3.4.2 Time-dependent construction stage analysis by MIDAS/Civil

MIDAS/Civil offers the option of a time-dependent construction stage analysis considering the
following time effects on materials:

• Creep in concrete members having different maturities

• Shrinkage in concrete members having different maturities

• Compressive strength gains of concrete members as a function of time

• Relaxation of pre-stressing tendons

The creep and shrinkage effects as well as the compressive strength gain properties of concrete
in MIDAS/Civil can be defined by choosing one of the following model codes for the prediction
of creep and shrinkage:

• CEB-FIB 1990

• ACI 209 (1992)

Alternatively, test data may also be directly entered into the program.

When the CEB-FIB 1990 model code is selected for the determination of creep and shrinkage
the following input parameters are required:

• Relative humidity of ambient environment

• Compressive strength of concrete at the age of 28 days (MIDAS requires the characteristic
compressive strength, fck )

• Type of cement (slow hardening, normal or rapid hardening, rapid hardening high strength)

• Notional size of the members (h = 2 · Ac /u)

• Age of a structural group when activated

MIDAS/Civil does not require the input of a temperature for a time-dependent construction
stage analysis so that the effect of elevated or reduced temperatures on the maturity of concrete
can not be taken into account in the calculation of the creep coefficient φ. This may lead to
incorrect results when the temperature is very low or extremely high. However, based on these
input parameters MIDAS/Civil automatically calculates the creep coefficients.
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 77

The increase of compressive strength with time is calculated based on the concrete compressive
strength at 28 days and the type of cement.

In order to consider prestress losses due to tendon relaxation MIDAS/Civil adopts the widely
used formula for the determination of relaxation at any time t given in Equation (3.39).

4.3.5 Running a construction stage analysis

4.3.5.1 Running a construction stage analysis by Larsa2000

When running a staged construction analysis by Larsa2000 different analysis options may be
selected. First the user needs to choose whether a standard or time-dependent analysis shall
be performed. In a standard analysis time effects on materials are ignored and in the time-
dependent type they are included.
Then the user can choose which construction stages are to be analyzed. The analysis can only
pick up from any stage up to where it last left off, or from the start. A starting and ending
construction stage are defined.
Because the staged construction analysis in Larsa2000 is a derivative of the nonlinear static
analysis, convergence criteria for the iterative solution are specified by the user. Equilibrium
iterations at a given load level can cease when the result is ”close enough” according to one or
more criteria. Two criteria used in Larsa are that the unbalanced force be a small fraction of
the total applied force in the current load level and that the current displacement increment be
a small fraction of the displacement increment. The convergence criteria are specified by the
user as the displacement tolerance, force tolerance and maximum number of iterations. The
iterative analysis continues until all degrees of freedom in the model satisfy displacement and
force tolerance criteria within the maximum number of iterations.
If the solution is not convergent for a load step, the analysis will continue if there are additional
load steps and the structure is not unstable. Larsa carries the unbalanced forces into the next
load step and a convergent solution may be obtained in the next load step. This problem can
usually be avoided by either using less restrictive error ratios or by increasing the maximum
number of iterations.

Finally, the user needs to specify whether geometric nonlinearity (large displacement and P-
Delta effects) will be included in the analysis.

When a time-dependent staged construction analysis is chosen, in addition to these options, the
time-dependent analysis options need to be chosen. It has to be selected which prediction code
for creep and shrinkage shall be used in the analysis and which time effects on materials shall
be considered.
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 78

4.3.5.2 Running a construction stage analysis by MIDAS/Civil

As well as in Larsa2000 different analysis options may be selected when running a staged
construction analysis by MIDAS/Civil. In contrast to Larsa, the construction stage analysis in
MIDAS is generally performed as a linear analysis. Hence, no convergence criteria have to be
defined when time-dependent effects and nonlinearities are not considered.
MIDAS/Civil offers to include time-dependent effects or nonlinear analysis in a construction
stage analysis. However, in the current version of MIDAS/Civil, nonlinear and time-dependent
effects cannot be considered simultaneously.
If the option include time-dependent effects is checked, the user may select the time-dependent
material properties that shall be included in the analysis. Furthermore, convergence criteria for
creep iteration need to be defined when creep is taken into account. MIDAS requires a maxi-
mum number of iterations and a tolerance for convergence. Additionally, a number of internal
time steps and a number of automatic time step generation for large time gap may be specified.
The first number is used to divide a construction stage to create internal steps for considering
creep and the second one is used to divide a construction stage to create internal steps when the
duration of the construction stage is too long.
If include nonlinear analysis is checked on the analysis option, MIDAS/Civil also requires
information for the iteration procedure. In order to perform a Newton-Raphson iteration the
maximum number of iterations as well as convergence criteria need to be defined. In MIDAS
up to three convergence criteria can be used that are energy, displacement and force tolerance.
The solution procedure is similar to the method explained above.

As a second option, the user can choose which construction stages are to be analyzed. In contrast
to Larsa2000, the analysis always starts from the first stage and only the ending construction
stage is defined.

Two additional options in MIDAS/Civil are the definition of the method of applying pretension
forces of cable elements and the choice whether the initial tangent displacement for erected
structures option shall be included in the analysis.
Regarding the pretension forces of cable elements, the choice is to either apply the pretension
forces as internal or as external forces. If the initial pretension forces are applied as internal
forces, the forces in the cable elements become reduced due to the deformation of the support
structure based on its stiffness. If the initial pretension forces are applied as external forces,
the forces are treated as external loads to the support structure at the construction stage of pre-
tensioning; hence the forces in the cable elements remain unchanged as the initial pretension
force at the corresponding construction stage.
When choosing to apply the pretension forces of cable elements as external loads, an additional
choice has to be made; MIDAS/Civil offers to add external forces to already existing pretension
forces of cable elements or to replace them with the applied external force.
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 79

In Larsa such an option is not needed. The problem is avoided by modeling the load of cables
as pre-tensioning (internal force) or post-tensioning (external force).

Concerning the initial tangent displacement for erected structures function, one is referred to
Section 4.3.2, where this option is explained in detail.

In addition to these basic options, there are some more options regarding the output of the
results that are not be discussed here. For further information it is referred to the MIDAS/Civil
Online Manual [31].

4.3.6 Solution for unknown loads using optimization technique

In contrast to RM2004, both Larsa2000 and MIDAS/Civil do not offer a feature which allows
for the optimization of stay cable forces considering the construction sequence, time-dependent
effects and nonlinearities. However, MIDAS/Civil provides a function called the unknown load
factor function, which is capable to calculate unknown loading conditions necessary to satisfy
a given design requirement in a linear analysis.

4.3.6.1 The unknown load factor function in MIDAS/Civil

The unknown load factor function likewise the AddCon-Method in RM2004 can be used to
calculate optimal load factors that satisfy specific constraints of a structure. The function de-
termines superposition factors for previously calculated load cases to obtain a prescribed state
in the structure by combining these load cases. It can be used when performing a simple linear
analysis, but also for construction stage analyses.
In case of a linear analysis, a load combination containing the applied external loads and user
defined unit loads is defined. The load cases of this combination for which the unknown load
factors are to be obtained are then selected and certain constraints (displacement, internal forces,
etc.) are defined. Based on these input data the program computes the influence of all load case
on the objectives and calculates unknown load factors for the unit loads that satisfy the require-
ments. The output of the program are the calculated results of the unknown load factors and the
corresponding influence matrix.
In case of a construction stage analysis, the definition of load combinations is not necessary.
All ”unit loads” that are applied during the construction process can be used as load cases for
which the unknown load factors are to be obtained. They can be chosen by simply selecting the
construction step at which the unit loads have been activated.
Similar to the linear analysis, constraint conditions that are to be satisfied are specified. They
can be defined for an arbitrary construction stage. In most cases this will be the final construc-
tion stage, but it can also be any other stage. When performing a construction stage analysis,
MIDAS/Civil computes the influence of the activities in the different construction steps (load
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 80

application, support jacking, etc.) on the objectives in the ”final” construction stage and deter-
mines the unknown load factors of the ”unit activities” that satisfy the constraint conditions. In
this case at least the construction sequence is considered in the optimization procedure. How-
ever, time-dependent effects and nonlinearities can not be taken into account in the unknown
load factor calculation by MIDAS/Civil.
For constraint conditions to be satisfied, in both cases equality and inequality conditions are
permitted. In the first case MIDAS searches for a condition where the optimized constraint is
equal to an entered constraint and in the latter a upper and lower bound for the constraint is
specified by the user.
The equality conditions are solved using linear algebraic equations. If the numbers of the un-
known loads and equations are equal, the solution can be directly obtained by solving a linear
equation system.
If a solution satisfying inequality conditions is obtained, numerous solutions to the unknown
loads exist depending on the constraints imposed to the inequality conditions. MIDAS/Civil
finds a solution to inequality conditions, which uses variables that minimizes the given object
functions. MIDAS/Civil allows to select the sum of the absolute values, the sum of the squares
and the maximum of the absolute values of variables for the object functions. Weight factors
can be assigned to specific variables to control their relative importance, and the effective ranges
of the variables can be specified.

The general procedures of the determination of initial cable forces by the unknown load factor
function in a linear and a construction stage analysis are illustrated in Figures 4.2 and 4.3.

Figure 4.2: Flowchart for the calculation of initial cable forces in a linear analysis
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 81

Figure 4.3: Flowchart for the calculation of initial cable forces in a construction stage analysis

However, comprehensive understanding of a structure is required to use the above optimization


technique. Since the desired conditions may not have a solution depending on the constraints,
selection of appropriate design conditions and object functions are very important.

4.3.6.2 Calculation of unknown load factors in Larsa2000

In contrast to MIDAS/Civil, Larsa2000 does not offer a function to automatically calculate un-
known loading conditions that satisfy user defined constraints. However, the influence matrix,
which was briefly introduced, can also be determined from an analysis by Larsa2000 and then
a linear algebraic equation can be solved to determine the unknown load factors.

The influence matrix in Larsa2000 can be assembled by simply applying each unit pretension
force separately to the structure, and compute the influence on the deformations or bending
moments, respectively, at selected control points. These control points and the type of influence
(deformation, bending moment, etc.) are the locations and type of design criteria that are re-
stricted in the optimization process. The influence matrix may contain influences on deflections
or bending moments only, but also a combination of both. Thus, for instance, restrictions can
be made to the bending moments along the girder and to the deflection at the top of the pylon at
the same time.
Computing the influence of all unit pretension forces to all objectives a n x m influence matrix
results, where n is the number of unit pretension forces and m is the number of constraints. If
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 82

the number of unit pretension force and the number of objectives is equal, the cable tension
forces can be calculated by solving a linear equation.
The described method to determine the influence matrix can be applied to a linear, but also
to a construction stage analysis. In case of a staged construction analysis the unit pretension
forces are applied at the time of installation and the influences on the objectives in the ”final”
construction stage are computed.

4.3.6.3 Consideration of time-dependent effects in the optimization process

There is a general belief that the consideration of creep is a nonlinear problem. However, it
was shown in Chapter 3.3.4.7 that the effects of creep and shrinkage can be treated in a linear
manner. As a consequence, the principles of linear superposition may still be applied when
considering time-dependent material behavior in the optimization process.
It was shown that the total creep occurring during a single time step may be decomposed into
single contributions. The creep deformation or creep moment at any location consists of one
part which is related to the permanent load and the other parts are related to the unit loads. The
unit loading cases are linearly coupled to the same unknown factors as for the linear static or
linear construction stage analysis, respectively.
Due to this fact an influence matrix containing the effects of creep and shrinkage can be ob-
tained in a similar manner as described above. Instead of a linear construction stage analysis, a
time-dependent construction stage analysis is performed. The unit pretension forces are applied
at the time of installation and the influences on the objectives in the final construction stage
are computed. Needless to say, not only the the influence matrix but also the moments and
displacements due to dead load change compared to the linear construction stage analysis.
Because in the unknown load factor function in MIDAS/Civil time-dependent effects can not
be considered, a similar procedure as described for Larsa2000 need to be used in MIDAS/Civil
when considering time-dependent material behavior.

The above described method to compute an optimum tensioning strategy considering the con-
struction sequence as well as time-dependent material behavior is adopted in the analysis of a
simple model of a cable stayed bridge in Chapter 6. On the basis of this example the procedure
is described in detail and the results determined by Larsa2000 and MIDAS/Civi are compared
and evaluated.
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 83

4.4 Comparison of a time-dependent construction stage anal-


ysis by Larsa2000 and MIDAS/Civil

As already mentioned, both Larsa2000 and MIDAS/Civil offer a function to consider creep and
shrinkage in a construction stage analysis. In the following part of this thesis the application
of these functions in the calculation of a simple model of a bridge structure shall be verified.
Therefore, in both analysis programs the CEB-FIB 1990 model code for prediction of creep and
shrinkage is chosen to consider the time effects on materials. In order to control the computer
calculations the results are then compared to hand calculations that are also based on the CEB-
FIB 1990 model code and should, therefore, yield similar results.

Figure 4.4 illustrates the construction sequence of a simplified model of a two span bridge struc-
ture erected by one-sided free cantilevering. In the first stage the first part of the bridge deck
is erected and the self weight is applied. Then the derrick crane is installed and the load of
segment 2 is applied. In stage 3 the next segment is placed and loaded. In the final fourth
construction stage the derrick crane is removed from the structure. The material, section and
loading data for the verification model are given in Table 4.1 to 4.3.

Figure 4.4: Construction sequence of the verification model for creep and shrinkage
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 84

Table 4.1: Material data for the verification model for creep and shrinkage
Classification Modulus of Elasticity Poisson’s Ratio Compressive Strength at
E (kN/m2 ) age of 28 days fck (kN/m2 )
Concrete 3.5034 · 107 0.3 35, 000

Table 4.2: Section data for the verification model for creep and shrinkage
Classification Cross-sectional Area Moment of Inertia Perimeter
(m2 ) (m4 ) (m)
Deck 0.4000 0.1 4.00

Table 4.3: Loading data


Classification Load
Self Weight (Deck) 10 kN/m
Derrick crane 50 kN

4.4.1 Linear construction stage calculation

A linear construction stage analysis is performed by Larsa2000 and MIDAS/Civil. The resulting
joint deflections are almost identical in both analysis programs. Table 4.4 shows the resulting
deflections in all construction stages. The corresponding internal forces and moments are also
similar for both analyses, but they are not illustrated here.

Table 4.4: Linear construction stage calculation - Vertical deflections


Vertical deflection [mm]
Construction Stage Larsa2000 MIDAS/Civil
Joint 2 Joint 3 Joint 2 Joint 3
CS01 −3.5680 - −3.5680 -
CS02 −24.9757 - −24.9757 -
CS03 −24.9757 −68.9805 −24.9757 −68.9806
CS04 −20.2184 −57.0873 −20.2185 −57.0874
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 85

4.4.2 Construction stage calculation considering creep effects

4.4.2.1 Analysis by Larsa2000 and MIDAS/Civil

A time-dependent construction stage analysis is performed by Larsa2000 and MIDAS/Civil. As


already mentioned above in both cases the CEB-FIB 1990 model code is chosen as the basis
for the definition of the time dependent parameters. In addition to the section and material
data given in Table 4.1 and 4.2 a relative humidity of 70% and a normal hardening cement are
assumed.
In this section only time-dependent effects due to creep are considered. Shrinkage strain as well
as the increase of the modulus of elasticity of concrete with time (aging) are neglected.
Since the structure is statically determinate the internal forces and moments do not change due
to the effects of creep. They stay the same as for the linear analysis. The vertical deformation
at joint 2 and 3 at the end of construction stage CS01 to CS04 are given in Table 4.5.
In contrast to the linear calculation there are small differences in the results from the analysis by
Larsa2000 and MIDAS/Civil. These deviations are summarized and evaluated in comparison
with a hand calculation in Section 4.4.4.

Table 4.5: Time-dependent construction stage calculation - Vertical deflections


Vertical deflection [mm]
Construction Stage Larsa2000 MIDAS/Civil
Joint 2 Joint 3 Joint 2 Joint 3
CS01 −3.5680 - −3.5680 -
CS02 −28.2272 - −27.4685 -
CS03 −34.7837 −94.5777 −34.0452 −92.5005
CS04 −31.6030 −88.7385 −30.5879 −85.0733

4.4.2.2 Hand Calculations

In order to check the results of the construction stage analyses by Larsa2000 and MIDAS/Civil
a simple hand calculation of the structure is performed.
The structural system in all four construction stages is a cantilever under different load combi-
nations. In the first construction stage the cantilever is loaded by a uniform load in the amount
of the self weight of the bridge deck. In the second phase an additional external concentrated
load and an external moment are applied (concentrated load = weight of segment 2 + derrick
crane, moment = moment caused by weight of segment 2). In the third stage no changes occur
to the first part of the bridge deck. The only difference is, that the second segment is installed
and loaded by a uniform load. This uniform load produces the same concentrated load and
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 86

bending moment at joint 2 as in stage 2. The second half of the structure can be considered as
a cantilever under a uniform load rigidly fixed to the first part. In the final fourth construction
stage the derrick crane is removed, which corresponds to an upward load at joint 2.
Concerning the vertical deflections and joint rotations the structural systems and corresponding
expressions for the deflection and rotation of Figure 4.5 can be taken to determine the theoretical
deformation [44].

Figure 4.5: Elastic vertical deflection and rotation of a cantilever

wL3 P L2
Figure 4.5 a): θel (w) = Figure 4.5 b): θel (P ) =
6EI 2EI
wL4 P L3
δel,z (w) = δel,z (P ) =
8EI 3EI

ML
Figure 4.5 c): θel (M ) =
EI
M L2
δel,z (M ) = (4.1)
2EI

Using these expressions, the elastic vertical deflections and rotations of the structure due to the
activities in the construction stages CS01 to CS04 can be calculated. First only element 1 is
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 87

considered and the elastic deformations of joint two are determined. Table 4.6 shows the values
separated into deflections and rotations due to the different activities. In Table 4.7 the total
elastic deformation of joint 2 and the corresponding initial deformation of joint 3 are given.

Table 4.6: Elastic deformation of joint 2


+Self Weight Seg 1 + Load of Segment 2 +/- Derrick crane
δel [mm] θel [10−4 ] δel [mm] θel [10−4 ] δel [mm] θel [10−4 ]
CS01: −3.5680 −4.7573 - - - -
CS02: - - −16.6505 −28.5437 −4.7573 −7.1359
CS03: - - - - - -
CS04: - - - - +4.7573 +7.1359

The total elastic deformation of joint 2 in a given construction stage is the sum of the defor-
mations of the actual and all previous construction stages. Through the deformation of the first
segment (Joint 2) also the second segment (Joint 3) experiences an initial deformation. This
initial deformation can be calculated by Equation (4.2) and is, in addition to the total elastic
deformation of joint 2, given in Table 4.7.

δel (Joint 3) = δel (Joint 2) + θel (Joint 2) · l2


θel (Joint 3) = θel (Joint 2) (4.2)

Table 4.7: Total elastic deformation of joint 2 and initial elastic deformation of joint 3
Total elastic deformation of joint 2 Initial elastic deformation of joint 3
δel [mm] θel [10−4 ] δel [mm] θel [10−4 ]
CS01: −3.5680 −4.757 - -
CS02: −24.9757 −40.4369 - -
CS03: −24.9757 −40.4369 −65.4126 −40.4369
CS04: −20.2185 −33.3010 −53.5194 −33.3010

Considering the second element of the structure, the total vertical deflections and rotations of
joint 3 can be determined. The only load that is to be applied to this element is the uniformly
distributed self weight of segment 2. The deformation due to all other activities is considered
in the initial deformation of joint 3. The total deflections and rotations of joint 3 are the sum of
the initial deformations and the deformations due to the construction of segment 2.
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 88

Table 4.8: Total elastic deformation of joint 3


Total elastic deformation Initial elastic deformation + Self Weight Seg 2
δel [mm] θel [10−4 ] δel [mm] θel [10−4 ] δel [mm] θel [10−4 ]
CS01: - - - - - -
CS02: - - - - - -
CS03: −68.9806 −45.1942 −65.4126 −40.4369 −3.5680 −4.7573
CS04: −57.0874 −38.0583 −53.5194 −33.3010 - -

Creep deformation

The creep deformation is determined by multiplying the elastic deformation by the creep co-
efficient, φ. The creep coefficients for the different loading ages, t 0i , are calculated from the
formulas given in the CEB-FIB 1990 model code using the same input parameters as for the
computer calculation. The detailed procedure is described in Chapter 3.3.4.6. The resulting
creep factors, φ, are given in Table 4.9.

Table 4.9: Creep coefficients φ


t0,1 = 10 days t0,2 = 20 days t0,3 = 21-16 days
+ SW Segment 1 +Load of Seg 2/+Derrick + SW Segment 2
CS01: φ(t0,i , 10) 0 0 0
CS02: φ(t0,i , 20) 0.6834 0 0
CS03: φ(t0,i , 21) 0.7029 0.3019 0
CS04: φ(t0,i , 22) 0.7211 0.3715 φ(t0,3 , 22 − 16) = 0.2991

The sum of the elastic and creep deformation of joint 2 due to the activities in the construction
stages CS01 to CS04 can be calculated by multiplying the different elastic deformation given
in Table 4.6 by the factor (1 + φ). The results are shown in the following Table:

Table 4.10: Sum of elastic and creep deformation of joint 2


+Self Weight Seg 1 + Load of Segment 2 +/- Derrick crane
δel+c [mm] θel+c [10−4 ] δel+c [mm] θel+c [10−4 ] δel+c [mm] θel+c [10−4 ]
CS01: −3.5680 −4.7573 - - - -
CS02: −6.0064 −8.0085 −16.6505 −28.5437 −4.7573 −7.1359
CS03: −6.0758 −8.1011 −21.6776 −37.1617 −6.1936 −9.2904
CS04: −6.1408 −8.1877 −22.8363 −39.1480 −1.7674 −2.6511
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 89

From these values the total deformation of joint 2 in a given construction stage can be calculated
by summing up the values of the particular row. Then initial deformation of joint 3 (including
elastic and creep deformation) can be calculated by the same procedure as explained above. The
results are given in Table 4.11.

Table 4.11: Total deformation of joint 2 and initial deformation of joint 3


Total deformation of joint 2 Initial deformation of joint 3
δel+c [mm] θel+c [10−4 ] δel+c [mm] θel+c [10−4 ]
CS01: −3.5680 −4.757 - -
CS02: −27.4142 −43.6881 - -
CS03: −33.9470 −54.5531 −88.5002 −54.5531
CS04: −30.7445 −49.9868 −80.7313 −49.9868

As explained above the total deformation of joint 3 is determined by adding the initial deforma-
tion of joint 3 and the deformation due to the self weight of segment 2.

Table 4.12: Total deformation of joint 3


Total elastic deformation Initial elastic deformation + Self Weight Seg 2
δel+c [mm] θel+c [10−4 ] δel+c [mm] θel+c [10−4 ] δel+c [mm] θel+c [10−4 ]
CS01: - - - - - -
CS02: - - - - - -
CS03: −92.0681 −59.3104 −88.5002 −54.5531 −3.5680 −4.7573
CS04: −85.6974 −56.6083 −80.7313 −49.9868 −4.6353 −6.1804

Thus, the vertical deflections at joint 2 and 3 including elastic and creep deformation result to:

Table 4.13: Total vertical deflections of joint 2 and 3


Vertical deflection [mm]
Construction Stage Hand Calculations
Joint 2 Joint 3
CS01 −3.5680 -
CS02 −27.4142 -
CS03 −33.9470 −92.0681
CS04 −30.7445 −85.6974
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 90

4.4.3 Construction stage calculation considering shrinkage strain

An additional time-dependent construction stage analysis considering concrete shrinkage is per-


formed by Larsa2000 and MIDAS/Civil. Shrinkage strain likewise creep strain is defined by
choosing the CEB-FIB 1990 model code. The input parameter equate those used in section
4.4.2. Since the considered verification example is statically determinate, shrinkage strain only
affects the axial deformation of the structure. The axial deflection due to shrinkage of joint 2
and 3 is given in Table 4.14.

Table 4.14: Axial deformation due to shrinkage


Horizontal deformation [mm]
Construction Stage Larsa2000 MIDAS/Civil
Joint 2 Joint 3 Joint 2 Joint 3
CS01 0 - 0 -
CS02 −0.1569 - −0.1569 -
CS03 −0.1695 −0.1695 −0.1695 −0.1695
CS04 −0.1817 −0.2158 −0.1817 −0.2158

Because the horizontal deflection due to shrinkage of joint 2 and 3 determined by Larsa2000
and MIDAS/Civil has the same magnitude in all construction stages, in this case no further
hand calculation is performed. The feature of including shrinkage effects in a construction
stage analysis is considered to be fully functional in both analysis programs.

4.4.4 Evaluation of the time-dependent construction stage analysis results

From the results in Sections 4.4.2 and 4.4.3 it can be seen that the structural deformations due
to shrinkage are absolutely the same for the analysis by Larsa2000 and MIDAS/Civil, whereas
for the creep deformation discrepancies occurred in both computer simulations. In order to dis-
cover the reason for these variations, Table 4.15 once again shows the sum of the elastic and
creep deformation of joint 2 and 3 determined by Larsa2000 and MIDAS/Civil in comparison
with the results from the hand calculation in Section 4.4.2.2.

From Table 4.15 it is visible that the MIDAS results are very close to the results from the
hand calculation. Only very small deviations of less than 1% occur, which is tolerable for
the analysis. The difference between the theoretical deformations and the values calculated by
Larsa2000, on the other hand, show larger discrepancies. The reason for this inaccurateness
can be explained considering a simple cantilever under a uniform unit load as an example. The
elastic deformation is given by Equation (4.3).
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 91

Table 4.15: Time-dependent construction stage calculation - Vertical deflections


Vertical deflection [mm]
Larsa2000 MIDAS/Civil Hand Calculations
Joint 2 Joint 3 Joint 2 Joint 3 Joint 2 Joint 3
CS01 −3.5680 - −3.5680 - −3.5680 -
CS02 −28.2272 - −27.4685 - −27.4142 -
CS03 −34.7837 −94.5777 −34.0452 −92.5005 −33.9470 −92.0681
CS04 −31.6030 −88.7385 −30.5879 −85.0733 −30.7445 −85.6974

Figure 4.6 a) shows the qualitative curvature/bending moment diagram of the cantilever. If the
structure is modeled by one single element it seems that Larsa2000 takes the curvature at both
ends of the element and assumes a linear distribution over the element (Figure 4.6 b)).

Figure 4.6: Curvature diagram of a cantilever under a uniform unit load

In general, creep of concrete is taken into account by multiplying the elastic strain by the creep
coefficient φ, which is similar to multiplying the curvature, Ψ, by the creep coefficient. Do-
ing so the theoretical curvature of the element due to creep is shown in Figure 4.7 a) and the
creep curvature calculated by Larsa2000 is shown in Figure 4.7 b). The additional deflections
resulting from these creep curvature distributions are given in Equations (4.4) and (4.5).

Figure 4.7: Creep curvature diagram of a cantilever under a uniform unit load
CHAPTER 4. COMPARISON OF STRUCTURAL ANALYSIS PROGRAMS 92

L4
Elastic deflection: δel = (4.3)
8EI
L4
Theoretical creep deflection (Figure 4.7 a)): δc = φ· (4.4)
8EI
L4
Creep deflection by Larsa2000 (Figure 4.7 b)): δc = φ· (4.5)
6EI
The sum of the elastic and creep deflection results to:

a) Theoretical Calculation:
L4 L4 L4
δel + δc = +φ· = (1 + φ) · = (1 + φ) · δel (4.6)
8EI 8EI 8EI

b) Analysis by Larsa2000: (4.7)


4 4 4
L L L
δel + δc = +φ· = (1 + 4/3 · φ) · = (1 + 4/3 · φ) · δel (4.8)
8EI 6EI 8EI

Equations (4.6) and (4.8) show a clear difference between the theoretical result and the Larsa2000
result when the cantilever is modeled by one element. For different structural systems and load
cases the magnitude of this error in the determination of creep effects varies depending on the
theoretical distribution of the curvature along the element. The closer the theoretical curvature
distribution to a linear distribution, the more exact are the Larsa results. However, this error
can be reduced by refining the element size. This way the distribution of the curvature in the
Larsa2000 analysis converges to the theoretical curvature diagram and the creep deformation
gets more exact.
When modeling the two segments of the verification example with 10 instead of 1 element, the
creep deflections change and the sum of the elastic and creep deformation results to the values
shown in Table 4.16. Now they are very close to the deflections determined by the hand calcu-
lations. Consequently, in order to perform a proper time-dependent analysis by Larsa2000 the
element sizes of a structure need to be sufficiently small.

Table 4.16: Total vertical deflections using 10 elements per segement


Vertical deflection [mm]
Larsa2000 Hand Calculations
Joint 2 Joint 3 Joint 2 Joint 3
CS01 −3.5680 - −3.5680 -
CS02 −27.4225 - −27.4142 -
CS03 −33.9561 −92.0947 −33.9470 −92.0681
CS04 −30.7539 −85.7298 −30.7445 −85.6974
Chapter 5

Construction Control and Monitoring

5.1 Introduction
In addition to the construction sequence a multiplicity of parameters influence the erection pro-
cess of cable-stayed bridges. In the simulation of the construction sequence by structural anal-
ysis programs, usually assumptions are made for these parameters that are experience values,
measured values or guidelines from design codes. Due to the high flexibility of cable-stayed
bridges, slight variations in the parameter values may already cause significant differences be-
tween the predicted and the actual state of the structure in a given construction stage. Because
errors between the design values and actual ones are inevitable, it is necessary to carry out a
continuous monitoring of the deck elevations and cable forces throughout the erection process
and to compare the actual values with the theoretical predictions. If the difference is small it
may be tolerated, but generally adjustments during the construction process are necessary to
reduce discrepancies in the final dead load condition. Should such differences be not corrected
in a timely manner and thus be allowed to accumulate, the geometry and internal forces of the
bridge may be out of control.
Generally, in order to achieve the desired geometry and internal forces in the bridge at the end
of construction, the deformation of the structure and the cable forces are continuously observed
and then compared with the results of the theoretical calculation. On the basis of the discrepan-
cies, a forecast of the future evolution of structure can be made and the cable-forces and camber
data can be adjusted so as to minimize the error between the forecast and the theoretical final
dead load condition.
In the following possible uncertainties in cable-stayed bridges are presented and methods to
avoid or at least reduce discrepancies between the actual and theoretical state of a cable-stayed
bridge are described. Within the scope of the description of construction control of cable-stayed
bridges a method to adjust discrepancies during the construction process is explained. In order
to safe time and to reduce the construction costs, the aim is to only adjust the cables forces and
the camber data of subsequently installed stay cables and segments and avoid retensioning of
already installed cables.
CHAPTER 5. CONSTRUCTION CONTROL AND MONITORING 94

5.2 Causes of discrepancies between the predicted and actual


state of a cable-stayed bridge during construction

The discrepancies between the theoretical predictions and the actual structural responses can be
attributed to the following factors:

• Differences between the assumed structural parameters for the design and the actual val-
ues achieved on site that include:

– Error in material properties such as modulus of elasticity of concrete or stay cables


– Error in section data of the girder, the pylon or the cables
– Incorrect estimation of dead load and construction loads
– Incorrect prediction of time effects on materials such as creep and shrinkage
– Incorrect assumption of the temperature distribution in the bridge deck and tower
– etc.

• Construction errors that include:

– Defective installation of segments


– Error in the magnitude of stay cable tension forces
– Welding or bolting errors
– etc.

• Fabrication errors such as incorrect member dimensions

• Measuring errors due to defect control instrumentation

The sensitivity of of a cable-stayed bridge to variations of these structural parameters is dif-


ferent. In order to detect the influence of variations of a particular parameter, calculations
considering the errors separately can be performed. This way the structural characteristics can
be cleared and it becomes much easier to identify the structural parameters that are responsi-
ble for the discrepancies between the predicted and actual state of a cable-stayed bridge during
construction. Generally, errors of girder segment weights have the most important effects on
the girder deflection and the cable forces of cable-stayed bridges.
CHAPTER 5. CONSTRUCTION CONTROL AND MONITORING 95

5.3 Supplementary measurement of structural parameters on


site

In order to avoid high discrepancies between the actual and theoretical deformations and internal
forces and moments, it is current practice to determine the actual loads, material characteristics
and environmental conditions on site before starting the erection of the bridge. If variations
of the input parameters occur they can be implemented in the input file of the analysis model
for the construction stage analysis and the initial cable forces and fabrication camber data that
cause the desired final condition can be recalculated.
In addition to the corrections of the input parameters before starting the erection process, the
computational model should also be continuously updated with input values measured on site
during the construction process. Segment weights should be monitored by checking member
dimensions and concrete unit weight or weighting segments (precast) and the environmental
conditions such as temperature and relative humidity need to be repeatedly measured. When
the computer simulation is frequently calibrated with the actual values it is relatively easy to
consistently recalculate the stay tensions which will produce the same final state of the tower
and the deck geometry that was previously anticipated. Furthermore, possible error factors are
diminished and by precisely knowing individual parameters it is much easier to identify the
reason for discrepancies between the actual structural responses and the theoretical predictions.

5.4 Construction control

The construction control of cable-stayed bridges is important to guarantee safety during the
construction process and to achieve the designed geometry and internal forces of the bridge
within reasonable tolerance. The construction control consists of the following four tasks:

1. Simulation analysis of the construction process using the input data (e.g. material and
loading data) updated with the actual values determined on site

2. Field measurements during erection such as:

• Elevations of the deck and horizontal displacements of the pylon


• Tensions of the cables
• Stresses of the concerned sections in the deck and the pylon
• Temperatures and gradients in the deck, pylon and cables

3. Comparison of the theoretical and actual results and forecast of the future evolution

4. Adjustment of deck elevations and cable forces during construction


CHAPTER 5. CONSTRUCTION CONTROL AND MONITORING 96

5.4.1 Simulation analysis of the construction process

The determination of the initial cable forces that need to be applied at the time of their installa-
tion to achieve a designed geometry or bending moment distribution at the time of completion
of a cable-stayed bridge was already explained in detail in Chapter 3 and 4. If the cable forces
are known, a detailed stage by stage construction procedure of the considered structure can
be simulated that takes into account the exact construction process, creep and shrinkage and
nonlinear effects. The construction stage analysis can be performed by any analysis program,
but the program should be sufficiently flexible to allow any stay force, dead load or material
adjustment in a very short time. In this thesis the structural engineering software programs
MIDAS/Civil and Larsa2000 are used for the analysis of cable-stayed bridges (see Chapter 4).
They provide detailed information about the condition in every construction stage of an analysis
(deformations, bending moments, stresses, etc.). By performing a construction stage analysis
by one of these programs a theoretical reference for each construction stage can be established
that includes information about elevations, cable tensions and internal forces at any time during
erection. Furthermore, input parameters as well as casting and installation times of segments
can easily be changed to predict the real behavior (including time-dependent effects) of a struc-
ture during erection. When there is a need for slight modifications of the input parameter or the
construction times, the simulation of the actual construction process provides information on
the adjustment of cable tensions and precamber.

5.4.2 Field measurements during erection

Field measurements during the erection of cable-stayed bridges should be done in each con-
struction phase. The higher the frequency of surveys, the easier it is to get a continuous picture
of the conditions of the structure and to detect abnormal deviations. Moreover, stopping the
erection process to be able to conduct a detailed survey is not necessary.
In order to avoid problems caused by temperature variations, all field measurements should be
performed between 11 p.m. and 6 a.m. when the temperature of the structure remains constant.
Thus, corrections due to thermal gradients in deck and due to difference in temperature between
the stays and the rest of the structure are small.

Elevations of the deck and horizontal displacements of the pylon:

Usually the deck elevations at a particular construction stage are measured at the ends of the
previous three to five segments to monitor the configuration of the bridge deck; the horizontal
displacements of the pylon are measured at different levels. Instant movements of the deck and
the pylon should be measured during loading and unloading operations. When measurements
CHAPTER 5. CONSTRUCTION CONTROL AND MONITORING 97

are systematically carried out before and after each tensioning operation and segment lifting an
interpretation of the results will give indications on the actual stiffness of the structure.
The movements of the pylon can be monitored by means of bench marks fixed to the soil and
installed perpendicularly to the pylon so as to allow measurements at different levels of the
pylon (footing, deck level, struts). The vertical deck displacements can be followed through
topographical marks placed on the longitudinal bridge girders.

Tensions of the cables:

During the erection of a cable-stayed bridge, the profile of the main girder and the structural
internal force state are strongly related to the cable forces. It is therefore required that cable
stay tensions are measured at high accuracy and efficiency. However, stay stressing is one of
the most critical operations performed during the construction of a cable-stayed bridge.
Different methods for the measurement of stay cable forces exist, but the frequency domain
method is presently the most commonly used method. The method basically consists of two
steps. First, the random vibration signals of cables under ambient excitation are picked up by
accelerometers attached to them and the signals are analyzed in the frequency domain with the
natural frequencies of the cables being identified. Second, the cable forces are deduced ac-
cording to the relationships between tensions and frequencies of the cables, which should be
determined in advance by theoretical analysis and field calibration. The frequency method is
more exact than other methods to measure cable forces, but it is evident that its effectiveness
depends on the accuracy in both the signal-pickup techniques and tension-frequency relation-
ship of the cables.
During the erection each cable is calibrated after it is just installed and stressed. Additionally,
sensors should be installed to a few other previously installed cables to establish whether the
actual stay tensions precisely correspond to those predicted by the design model. Before and
after the bridge closure the cable tensions of all installed cables should be measured once again.

An important aspect regarding the initial stay forces at the time of installation is the effect of
thermal gradients on stay forces. The nominal stay force calculated in the design does not
account for thermal effects. If a stay has to be stressed at day time, one solution is to verify
and adjust the stay force in early morning; another possibility is to compute a correction that
accounts for the loss due to gradients. Generally, during daytime, thermal gradients create a
downward deflection of the cantilever. When the gradient dissipates, the cantilever deflects
back up and the result is a loss of stay force. Therefore, when stressing a cable during daytime,
an increase of stay force is required.
CHAPTER 5. CONSTRUCTION CONTROL AND MONITORING 98

Stresses in the girder and the pylon:

At each construction stage the concrete stresses at critical sections along the girder and pylons
must be monitored and reviewed so as to ensure the structural safety as construction goes for-
ward. This can be done by embedding strain gauges at the concerned sections of the girder and
pylons. However, precautions must be taken to obtain valid results from the gauges:

• Gauges must have the proper accuracy to detect long-term strains

• Gauges must be installed properly and must not disturb during concrete pours

• Gauges have to be distributed across the width and the depth of the section so that average
axial stresses can be estimated

• A ”zero” reading must be taken just after erecting or pouring a segment

It should be noted that based on the use of strain gauges only concrete strains can be measured
directly. The strains must then be transformed into equivalent stresses by stress-strain relation-
ships, whereby the concrete modulus of elasticity has to be estimated. The critical issue in
determining the concrete stresses is to separate the non-stress strains due to creep and shrinkage
effects from the total strain in concrete. This problem can be solved on the basis of the fact that
the stresses at the neutral axis of the girder are only related to cable forces. First, the stresses
along the neutral axis of the girder are calculated by use of measured cable forces. Second,
those stresses are then used to calibrate the corresponding stresses by field measurements, from
which the creep and shrinkage coefficients of concrete can be identified. Finally, the above
results can further be used to modify the measured stresses at upper and lower edges of the
sections.

Temperature:

Grasping the regularity of temperature effects is of great concern to construction control of


cable-stayed bridges during erection. There are three different kinds of temperature effects,
namely the effect of overall temperature difference, the effect of temperature difference between
the girder and cables and the effect of temperature gradient of the girder. Overall temperature
difference has only little influence while the latter two factors have a large influence on the
bridge.
The average deck temperature, the temperature gradient of the bridge deck and the average cable
temperature are monitored by placing thermocouples in different parts of segments, distributing
them over the full height of the pylon and inserting thermocouples in the strands of a reference
cable.
CHAPTER 5. CONSTRUCTION CONTROL AND MONITORING 99

5.4.3 Comparison of the theoretical and actual results and forecast of the
future evolution

The comparison of the results involves holding the actual data up against the theoretical results
determined by the simulation analysis of the erection process in each construction phase. This
way the accuracy of the computer simulation can be checked, particularly with respect to the
assumed structural parameters.
If the deviations between the measured values and the values expected from design in a particu-
lar construction stage are lower than prescribed limits, the construction process can be continued
with the next stage without any changes. However, the erection hardly ever exactly follows the
predicted sequence and thus, in most cases significant differences between the actual structural
responses and the theoretical predictions will appear. In order to eliminate or at least reduce
these errors, a serious analysis must be performed to understand the reasons for the deviations,
which is of major importance in order to select the adapted amendment.
If stay force or camber adjustments are made without searching for causes of the discrepancies,
unacceptable conditions of the structure may result in the following construction stages. A sim-
ple example is:

If the differences in cantilever deflection come from a deviation between the actual and theo-
retical stiffness of the bridge girder, unacceptable stresses may be caused in the subsequent
stages when the bridge is erected following the theoretical elevations only.
The deck deflects up and down under stay stressing and segment lifting or pouring, whereby the
amplitude of the deflections increases with the length of the cantilever. If the bridge is stiffer
than assumed, the cantilever tip elevation will seem low after stressing a cable. In this case, it
would be a mistake to correct the elevation by overstressing the stay until the theoretical eleva-
tion is reached. Instead of an improvement of the erection process this would create extremely
high stresses in the bridge deck and the desired final condition will not be reached.

It becomes clear that the identification of the errors causing the discrepancies between the actual
and theoretical state of a cable-stayed bridge is an important task to obtain sufficient structural
amendments. However, the error identification is a difficult procedure. It is obvious that the
complex distribution of actual structural errors can not be estimated from the limited data ob-
tained from the measurement on site.
A relatively simple error identification method is to assume several patterns of typical structural
errors and approximate the actual error by their linear combination. The response of a cable-
stayed bridge to the assumed error can be expressed by Aerr x X, where X is a vector of which
the components are the multipliers for respective error patterns and A err is the error influence
matrix. The columns of Aerr represent the responses to the unit errors X1 , X2 , ..., Xn . The
multipliers X can be calculated by minimizing the difference between the actual errors and the
CHAPTER 5. CONSTRUCTION CONTROL AND MONITORING 100

approximated errors:

Minimize E = D − Aerr x X (5.1)

As already explained in Chapter 3.2.1 one of the most effective ways to obtain the minimum of
the difference between two vectors is to minimize the square error, which can be written as:

Minimize Ω = E2 = (D − Aerr x X)T (D − Aerr x X) (5.2)

In order to vary the weight between respective errors in the objective value, an additional weight
matrix W can be included:

Minimize Ω = E2 = (D − Aerr x X)T WT W (D − Aerr x X) (5.3)

where D is the vector of the actual errors.

Equation (5.3) can be solved as follows:

X = (ATerr WT W Aerr )−1 ATerr WT W D (5.4)

Even if a solution for the multiplication factors X is found, it may happen that the identified error
does not coincide with the real error. Generally, an extremely inaccurate error identification may
result due to the fact that proper error patterns were not used in the identification process. This
may be interpreted that the identification of the actual error is almost impossible, since the
correct patterns of the actual error are unknown.
However, according to Fujisawa et al. [13] this does not mean that the error identification is
useless. The objective of an economic cable adjustment is not to know the accurate structural
error, but to predict the response of the structure in future. Thus, even if the identified errors
are wrong, the calculated error factors may be used as an equivalent error to the actual one for
the prediction of the final state. A good approximation of the final condition may be reached if
it can be assumed that the change in the structural characteristics up to the final stage does not
significantly affect the response to the error identified in the current stage. However, satisfying
estimations of the condition of the structure at the time of completion may be achieved by
repeating the error identification procedure in subsequent construction stages.
CHAPTER 5. CONSTRUCTION CONTROL AND MONITORING 101

The validity of the assumption that a satisfactorily prediction of the final state can be reached,
even if the actual error patterns are not identified, is proved in the example calculation of a
cable-stayed bridge in Chapter 6.

If the choice of error patterns does absolutely not correspond to the actual combination or if
field measurements contain errors and the future evolution of the structure does not follow the
prediction when the identified error is introduced in the analysis, a complete check-up of the
structure becomes necessary to find explanations for the discrepancies. In this case the erection
has to be stopped until the designer is able to understand what happens. The complete check-up
includes:

• the weight of all equipments must be precisely checked

• tension forces must be seriously controlled

• segment weights and construction loads must be controlled

• all geometrical measurements must be questioned

• all methods used to evaluate construction data must be questioned and checked

• parasitic phenomena must be foreseen, such as concrete hardening effects, shrinkage, etc.

Without much doubt from such a detailed and complete checking an explanation of the differ-
ences between the actual and theoretical response of the structure will be found.

5.4.4 Adjustment of deck elevations and cable forces during construction

During the construction of most cable-stayed bridges, at least some discrepancies occur be-
tween the actual state and the state of design expectation. Such discrepancies arise from the
errors of material and/or loading parameters as stated in the previous sections. Therefore, it is
generally required that certain construction adjustments are performed to control the discrep-
ancies within allowable tolerance. Since cables are installed in turn in the cantilever erection,
these adjustments can be made in two ways:

1. Adjustment of all cable forces simultaneously in the final stage of construction

2. Adjustment of the cable forces and deck elevations timely during erection, i.e. limited
cables (usually cables installed in the current stage) are adjusted in arbitrary intermediate
stages
CHAPTER 5. CONSTRUCTION CONTROL AND MONITORING 102

1. Adjustment of the cable forces in the final construction stage

When the desired final dead load condition of a cable stayed bridge is achieved by adjusting the
cable forces in the final construction stage, the construction is conducted carefully in relative
geometry; each new segment is carefully referenced to the previous one. After completion of
the cantilevers they are adjusted to the desired final geometry by an adjustment of all cables.
In this case, geometrical control and intermediate computations are only used to detect large
errors.
This procedure has two big advantages. First, forecasting the future evolution of the bridge is
not needed. The discrepancies between the actual and the desired condition can be exactly mea-
sured in the final construction stage just before adjusting the cable forces. Moreover, because
all stay cables are adjusted before the completion of the bridge, the cable tension forces at the
time of installation do not need to be precisely known. Second, in the final construction stage,
the error in the global geometry is much easier to correct than in previous stages. The longer the
cantilever, the more flexible it is and the smaller the change in cable forces required to correct
discrepancies in geometry.
However, adjusting the cable forces in the final construction stage implicates that each cable
will be tensioned twice. Once at the time of installation and again in the final construction stage
just before the closure of the bridge. These cable force adjustments of already installed cables
are not preferable, because it takes time and increases the construction costs. From the econom-
ical point of view it would be much more reasonable to adjust tension forces of cables that have
not been installed yet. Furthermore, a small error at the beginning of the construction of a cable
stayed bridge generates a larger error later on. If the cable tension forces are not adjusted until
the final construction stage, unacceptable stresses and deformations due to the error may occur
during the erection of the bridge.

2. Adjustment of the cable forces and deck elevations timely during erection

The adjustment of the cable forces and deck elevations timely during erection is inferior to the
adjustment of the cable forces in the final stage from the standpoint of final residual errors, but
it is frequently employed because of economy. Especially in case of long-span cable-stayed
bridges the stressing of the stay cables is an expensive procedure and the aim should be to
reduce the number of stressing operations during the erection as much as possible.
During the construction of cable-stayed bridges, discrepancies in geometry can not only be
reduced by adjusting cable forces, but also by adjusting the elevation of a segment which can be
done by including an extra angle at the joint between two adjacent segments. In this case, only
the geometrical position of the girder is modified and the internal forces remain unchanged.
Thus, eliminating discrepancies between the actual and theoretical deformation by adjusting
deck elevations does not reduce possible discrepancies in internal forces and moments.
CHAPTER 5. CONSTRUCTION CONTROL AND MONITORING 103

In case of adjusting the cable forces and deck elevations timely during erection, the error iden-
tification takes place in earlier stages. After detecting the error in the current situation, the final
condition of the structure is extrapolated and becomes a reference for the adjustment. The fore-
cast of the final state of the bridge is independent from the construction phase being examined.
Based on the predicted final condition, adjustments of the segment elevations and cables forces
of subsequent construction phases can be determined. This way retensioning of stay cables is
avoided, which saves time and decreases the construction costs.

5.4.4.1 Procedure for optimum cable force and deck elevation adjustment

The goal of construction adjustment is to achieve both the desired deck profile or the desired
bending moment distribution, respectively, and the the cable forces calculated in the original
stage analysis. But, in fact, the requirements for deck-profile geometry and cable-stay tensions
are frequently incompatible; it seems impossible to achieve both requirements simultaneously
due to the influence of other parameter errors. In view of these considerations, in the following
the deck elevations or the bending moment distribution along the girder and the pylon, respec-
tively, are controlled, while the cable forces can be turned over a suitable range to eliminate or
reduce the effects of the error factors on the deck profile and internal force state of the girder
and pylons.
The desired geometry and/or the desired bending moment distribution can practically not ex-
actly be achieved by construction adjustments. In reality there always exist errors between the
ideal and actual values; the adjustment is necessary to reduce these errors.
The adjustment strategy presented in this thesis aims the minimization of the residual errors of
the structure. As it was found in Chapter 3, the relationship between the initial cable forces at
the time of installation and the response of the structure in the final construction stage is linear
as long as nonlinear effects are not considered. Thus, if the error in structural input parameters
is identified and introduced into the analysis model, a modified influence matrix, A, can be ob-
tained that describes the response of the structure to unit tension forces in the cables when using
revised input data (error included). If all cable forces shall be adjusted in the final construction
stage, the unit cable forces also need to be applied in the final construction stage; for an adjust-
ment of the cable forces and deck elevations timely during erection, the unit forces are applied
at the time of cable installation. In order to avoid retensioning of already installed cables, in the
latter case, only the influence of those cable forces is considered that are not yet installed at the
time of examination. Using the influence matrix the residual error of a structure is given as:

Eres = Df inal − A x X (5.5)


CHAPTER 5. CONSTRUCTION CONTROL AND MONITORING 104

where: Eres = the vector of residual errors


Df inal = the vector of discrepancies between the desired condition and
the forecasted (adjustment during construction) or actual final
state (adjustment in the final stage), respectively, when the
identified error is included in the analysis
A = the influence matrix of which columns represent the response of
the structure to the unit cable forces
X = the vector multiplication factors for the additional forces

Similar to the procedure of error identification, the vector of residual errors, E res , can be mini-
mized in the sense of least square errors. If a weight matrix is included (see Equation (5.3)) the
expression that needs to be minimized is:

Minimize Ω = E2res = (Df inal − A x X)T WT W (Df inal − A x X) (5.6)

which can be solved as follows:

X = (AT WT W A)−1 AT WT W Df inal (5.7)

Unless the discrepancies definitely result from incorrect installed segments the error in the final
construction stage is tried to be reduced by cable force adjustments. If the desired final condi-
tion is a restricted deformation, no further deck elevation adjustments are needed. Apart from
the residual, error the desired geometry will be achieved by adjusting the cable tension forces.
If the desired final condition is a restricted bending moment distribution (e.g. bending moment
distribution of an equivalent continuous beam), the desired geometry is generally reached by
an appropriate precamber. To make sure that the desired bending moment distribution and the
desired geometry in the final construction stage can still be achieved simultaneously when the
error is included in the analysis, the fabrication camber also has to be modified during erection.
As already explained in Chapter 3.4, when zero displacement is desired, the fabrication camber
is the curve between the up-side-down final displacements of all joints between two segments.
In order to determine the additional angle that has to be introduced at the joints between the
segments, the final displacements of the original stage analysis need to be compared with the
final displacements of the construction stage analysis including the revised input data. The ad-
ditional angle at a particular joint is the difference between the angle in the original fabrication
camber and the angle in the fabrication camber resulting from the modified analysis. However,
since the deck elevation of the segments that have already been installed can not be changed,
slight modifications of the updated camber data need to be made in these regions.
CHAPTER 5. CONSTRUCTION CONTROL AND MONITORING 105

The procedure of cable forces and deck elevation adjustments timely during erection to re-
duce discrepancies between the actual and theoretical state of a cable-stayed bridge is shown
in the analysis of a simple model of a cable-stayed bridge in Chapter 6. Different errors in
the structural parameters are incorporated in the construction stage analysis and, on the basis
of the difference in geometry and cable forces between the modified and original analysis, the
error is identified and construction adjustments are performed. Finally, the results in the final
construction stage of the adjusted model are compared to the theoretical results of the original
construction stage analysis which proves the functionality of the above described adjustment
procedure.
Chapter 6

Example Calculation of a Cable-Stayed


Bridge

6.1 Introduction

In the following part of this thesis the procedure to determine the initial cable forces presented
in Chapter 3 and the method to control the geometry and internal forces during the cantilever
construction explained in Chapter 5, are adopted in the analysis of a simple cable stayed bridge.
The calculations are performed by the dint of the structural engineering software programs
Larsa2000 and MIDAS/Civil that were introduced in Chapter 4. The dimensions and structural
properties of the example model that is used for the calculations are given in Section 6.1.
Primarily, only the final construction stage is considered in the calculation of the unknown
load factors for the applied unit loads. In a second step the initial cable force are determined
based on a time-independent construction stage analysis and finally not only the construction
process, but also time effects on materials are considered in the analysis. The pre-tension forces
resulting from the different calculations are compared and evaluated with the main concern on
the influence of effect of creep and shrinkage on the optimum stay cable forces. Furthermore,
based on the resulting cable forces and the corresponding bending moments and deformations
of the structure, the functionality of the construction stage analysis features of the used analysis
programs is proved.
In Section 6.6 a nonlinear time-dependent construction stage analysis using the initial cable
forces determined in the linear analysis is performed to demonstrate the influence of second-
order and large displacement effects on the erection of cable-stayed bridges and the last part
of this chapter deals with the construction control during the bridge erection. Different error
factors are included in the analysis model and are controlled during the construction process by
adjusting the cable forces and the deck elevations timely during erection.
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 107

6.2 Example model dimensions and properties

The structural system that is used for the example calculations is shown in Figure 6.1. It is a
simple asymmetrical cable-stayed bridge consisting of 5 girder segments and a key segment.
The bridge deck is supported by 5 cables that are arranged in a fan system, i.e. all cables are
anchored at the top of the pylon. The total height of the pylon is 30m; the distance from the
pylon base to the bridge deck is 10m.

Figure 6.1: Configuration of the final stage of an asymmetrical cable-stayed bridge

The anchorage points of the cables at the bridge deck are located at intervals of 12m and 4x16m
which corresponds to the lengths of the girder segments (girder 1 to girder 5). The key segment
at the end of the bridge has a length of 8m. However, in reality the length of concrete segments
would be much shorter (usually 3m to 5m), but in in order to limit the number of erection stages
of the example model, one segment is assumed to have the length between two cables.

The bridge deck is modeled of 21 elements, each having a length of 4m and the pylon consists
of 5 elements that have a length of 2x4,75m and 3x6,83m. Node number 1 is fixed, node num-
ber 101 is rigidly fixed and node number 22 is supported in z-direction.
The bridge girder is supported vertically on the pylon, but is free to move laterally, which is
modeled by rigidly linking the degree of freedom in vertical direction of joint 103 to the ver-
tical displacement of joint 8. The joints will move independently in horizontal direction while
they will have the same vertical displacement.

For the bridge girder a simple T-beam concrete cross-section is assumed. A Eurocode standard
concrete C35/45 is used. The material, section and loading data of the example model are given
in Tables 6.1 to 6.3.
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 108

Table 6.1: Material data of the example model


Classification Modulus of Elasticity 2 Poisson’s Ratio 3 Compressive strength at age
(kN/m2 ) - of 28 days fck (kN/m2 )
Deck 3.3282 · 107 0.3 35.000
Pylon 3.3282 · 107 0.3 35.000
Cable 2.1000 · 108 0.3 -

Table 6.2: Section data of the example model


Classification Cross-sectional Area Moment of Inertia Perimeter
(m2 ) (m4 ) (m)
Deck 4.35 0.92 25.00
Pylon 1.00 2.76 10.00
Cable 1 0.0208 - -
Cable 2-5 0.0062 - -

Table 6.3: Loading data of the example model


Classification Load-Value
Dead load (self weight) 25.0 kN/m3
Additional dead load 10.0 kN/m
Unit pre-/post-tensioning load (cables) 10.0 kN
Derrick crane 750 kN
Segment load Vertical load: A · 25 kN/m · LSeg
Eccentric moment: A · 25 kN/m · L2Seg /2
Unit support movement 1.0 mm

The self weight of the cables is neglected in the analysis of the example model. Thus, the cables
are treated as truss elements with the real modulus of elasticity of the cable and cable sag effects
are not considered.

6.3 Restrictions for the solution for unknown load factors

For the solution for unknown load factors constraints need to be defined that are to be satisfied
in the final dead load condition. These constraints can be a set of forces/moments, stresses or
displacements or a combination of these at defined locations on the bridge. The purpose of
an optimum tensioning strategy is to achieve a desired final geometry and a desired final mo-
ment/stress distribution simultaneously. As already mentioned, the geometry of the bridge is
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 109

influenced by an appropriate fabrication camber and thus the main objective is to minimize the
bending moments along the bridge girder and the pylon. However, dead load bending moments
and cable forces will redistribute due to concrete creep and minimizing the bending moments
at the time of completion of the bridge does not necessarily mean that they will stay the same
during service life.
There is a general belief that creep effects due to bending can be minimized if the deck is erected
so that the internal forces for the dead load configuration are those of the continuous beam on
rigid supports. In this case the spanwise overall curvature is zero and moment variations due to
creep eliminate each other. However, this may be the case when all segments of the bridge have
the same age, the same loading age and creep effects of the pylon are neglected which is hardly
ever the case. Generally, the bending moments will still redistribute, even if they equate those
of an equivalent continuous beam at the end of construction.

For the example model described above it was found that if the bending moment distribution of
an equivalent continuous beam is achieved in the final construction stage, considerable changes
of the internal forces and moments occur during the service live. The magnitude of the nega-
tive bending moment at the connection of the bridge girder to the pylon clearly increases and
the magnitude of the negative bending moments at the anchorage points of the last two cables
change towards positive bending due to creep.
It is also possible to aim for a desired moment distribution at a time when the majority of creep
effects have occurred. However, this causes an unacceptable condition at the time of comple-
tion of the bridge. For instance, if the bending moment distribution of an equivalent continuous
beam is desired for day 6000 by stressing each cable only once at the time of installation, ex-
tremely high bending moments and deformations occur in the final stage of construction. The
bending moments may be acceptable, but adjusting vertical deflections of more than 500mm by
a fabrication camber is not reasonable.

In the following calculations the unknown load factors of the cable tension forces are calculated
for a restricted moment distribution (Case I) and for restricted displacements (Case II). In both
cases the restricted condition is requested for the time of completion of the bridge. The desired
bending moment distribution is that of an equivalent continuous beam along the girder and zero
at the pylon base. In case of the restricted deformation a zero vertical displacement is aspired
at the anchorage points of cable 1-4 and at the end of the bridge girder and a zero horizontal
displacement is desired at the top of the pylon.
This way the general procedure of optimizing the cable tension forces for a desired condition at
a specific date can be demonstrated. However, it becomes clear that it is almost impossible to
determine optimum initial tension forces that reduce bending stresses and deformations during
construction, at the end of construction and after years of operation simultaneously. The de-
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 110

signer of a cable stayed bridge always needs to find an ideal solution that minimizes the stresses
and deformations in all construction phases and the stressing operations at the same time. A
good solution may be to aim for a desired condition at the time of completion of the bridge and,
if needed, restress the cable forces after a few years of bridge operation, in order to compensate
for time-dependent effects occurring during service life of the bridge, which is done in the fol-
lowing.

The bending moment distribution of the equivalent continuous beam of the example model is
illustrated in Figure 6.2. The target values and locations that are used for the optimization
process are shown in Table 6.4.

Figure 6.2: Bending moment distribution of an equivalent continuous beam [kNm]

Table 6.4: Constraints for the unknown load factor calculation


Case I Case II
Location Target bending moment Target vertical displacement
(kN m) (mm)
Node 4 -2350 0
Node 8 -2600 -
Node12 -2500 0
Node 16 -2700 0
Node 20 -2000 0
Node 22 - 0
Node 101 0 -
Node 106 - 0 (horizontal)
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 111

6.4 Calculation of the initial cable forces in a linear static


analysis

6.4.1 Calculation by MIDAS/Civil

As already described in Chapter 4, in MIDAS/Civil the initial cables forces can be calculated
using the unknown load factor function. Figures 6.3 and 6.4 illustrate the bending moment
distribution and the deflected shape, respectively, in the final stage under self weight, additional
dead load and unit pretension forces in the cables. Both diagrams are the result of a linear static
analysis of the complete structure.

Figure 6.3: Bending moment distribution under dead load and unit cable forces [kNm]

(dx=+28.23)

Figure 6.4: Vertical deformation under dead load and unit cable forces [mm]
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 112

Case I: Restricted bending moments


In order to calculate the initial cable forces for restricted bending moments, constraints are
defined for the bending moment at node number 4, 12, 16, 20 and 101. The values for the
constraints are given in Table 6.4. The cables forces calculated by the unknown load factor
function in MIDAS/Civil are shown Table 6.5 and the resulting bending moment distribution
and vertical deformation of the structure when these cable forces are applied are illustrated in
Figures 6.5 and 6.6.

Table 6.5: Ideal cable forces for restricted bending moments, Case I
Cable: Cable 1 Cable 2 Cable 3 Cable 4 Cable 5
(kN ) (kN ) (kN ) (kN ) (kN )
Tension force: 9064.95 2790.89 2885.18 3992.78 4460.56

Figure 6.5: Moment distribution when restricting bending moments [kNm], Case I

(dx=0.00)

Figure 6.6: Vertical deflection when restricting bending moments [mm], Case I
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 113

Case II: Restricted deformations


In order to calculate the initial cable forces for restricted deformations, constraints are defined
for the vertical deflection at node number 4, 12, 16, 20 and and for the horizontal deflection at
node number 106. They are all set to be zero. The cables forces calculated by the unknown
load factor function in MIDAS/Civil are given in Table 6.6 and the resulting bending moment
distribution and vertical deformation of the structure when these cable forces are applied are
illustrated in Figures 6.7 and 6.8.

Table 6.6: Ideal cable forces for restricted deformations, Case II


Cable: Cable 1 Cable 2 Cable 3 Cable 4 Cable 5
(kN ) (kN ) (kN ) (kN ) (kN )
Tension force: 9483.69 3318.52 3439.01 3765.49 4578.80

Figure 6.7: Moment distribution when restricting deformations [kNm], Case II

(dx=0.00)

Figure 6.8: Vertical deflection when restricting deformations [mm], Case II


CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 114

6.4.2 Calculation by Larsa2000

The procedure of calculating unknown load factors by Larsa2000 was explained in Chapter
4.3.6.2. The bending moment distribution and the deflected shape of the example model under
self weight, additional dead load and unit pretension forces in the cables determined by a linear
static analysis in Larsa2000 are shown in Figures 6.9 and 6.10. Both diagrams equate those
determined by MIDAS/Civil.

-17644

2239

17663
8644

Figure 6.9: Bending moment distribution under dead load and unit cable forces [kNm]

(dx=+28.23)

+1.71
-82.35 -75.51
-140.75

Figure 6.10: Vertical deformation under dead load and unit cable forces [mm]
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 115

By applying each unit tension force (10 kN ) separately to the structure, the influence on the
objectives can be obtained. The resulting influence matrices for the moment (Case I) and dis-
placement (Case II) restriction are shown below.

Case I: Restricted bending moments

 
+7.8069 −30.0183 +3.8952 −1.0193 −6.3109
−3.1736 +0.5621 −25.0806 +9.2918 +7.3318
 
 
 
Amoment = 
 −6.4576 −3.0379 −3.8192 −27.2154 +2.8456 

−4.3352 −2.0774 −0.1913 −6.5033 −16.2726
 
 
−18.1750 +8.3804 −2.1709 −7.4557 −17.0789

Case II: Restricted deformations

 
−0.01828 +0.05306 −0.02065 −0.00264 +0.01342
+0.03957 −0.00535 +0.10789 +0.05480 −0.02394
 
 
 
Adispl. =
 +0.05704 +0.01080 +0.09986 +0.13458 −0.00359 

+0.02805 +0.00790 +0.03533 +0.06656 +0.02016
 
 
−0.05936 −0.02737 +0.00709 +0.02435 +0.05578

The first to the fifth row in the moment influence matrix, Amoment , describe the bending mo-
ments [kNm] at node 4, 12, 16, 20 and 101 due to a unit pre-tensioning of 10kN in cable 1 to
5. The first to the fifth row in the displacement influence matrix, A displ , describe the vertical
deflection [mm] at node 4, 12, 16 and 20 and the horizontal deflection [mm] at node 106 due to
a unit tension forces of 10kN in Cable 1 to 5.

The bending moments [kNm] and displacements [mm] at the selected nodes due to the self
weight and the additional dead load are the following:
   
+597.8020 +1.68327
 +659.2632   −82.52467 
   
   
MP =   +15073.9935 
 δP =  −141.05189 

 +12670.5173   −75.67107 
   
−8643.5826 +28.22907
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 116

The constraint vectors for a restricted moment distribution and restricted deformations are given
below (see Table 6.4):
   
−2350 0
 −2500   0 
   
   
Mdesired =  −2700 
  δdesired = 
 0 

 −2000   0 
   
0 0

The initial cable forces can now be calculated by solving the following linear equations:

moment · [Mdesired − MP ] · 10 kN or 1mm = Tmoment


A−1

displ · [δdesired − δP ] · 10 kN or 1mm


A−1 = Tdispl

The resulting tension forces are:

Case I: Case II:


   
+9642.955 +9483.695
+2790.892 +3318.517
   
   
   
Tmoment = 
 +2885.179 
 Tdispl =
 +3439.013 

+3992.780 +3765.486
   
   
+4460.557 +4578.799

Figures 6.11 to 6.14 illustrate the adjusted moment distribution and vertical deflections when
the above calculated cable forces are applied to the structure.

-2700
-2350 -2500
-2097 -2000

1136
1225 1482
1601 1528

Figure 6.11: Moment distribution when restricting bending moments [kNm], Case I
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 117

dz=-12.14,
dx= 0.00

-2.17 -3.53 -2.09 -0.65

Figure 6.12: Vertical deflection when restricting bending moments [mm], Case I

-3894 -3933

-2317 -2096

-327

645 742
1618
1922 1906

Figure 6.13: Moment distribution when restricting deformations [kNm], Case II

dz=-12.42,
dx= 0.00

0.00 0.00 0.00 0.00

Figure 6.14: Vertical deflection when restricting deformations [mm], Case II


CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 118

6.4.3 Summary of the determination of the initial cable forces in a linear


static analysis

From the above calculations it is visible that the results determined by MIDAS/Civil and Larsa2000
equate each other. Furthermore, the bending moment and deformation diagrams that arise when
the calculated forces are applied show that the constraints are satisfied in all cases. However,
the calculated initial cable forces are those of the final erection stage. In order to determine
the cable forces that need to be applied at the time of installation, a backward analysis starting
from the final stage can be performed. This method is most likely the fastest to achieve a result
for the initial cable forces, but, as already explained in detail in Chapter 3.2.4.2, many different
problems may occur during the backward and subsequent forward analysis. However, such a
backward analysis is not performed in this thesis. The initial cable forces at the time of instal-
lation are determined in a forward construction stage analysis in the following section.

6.5 Calculation of the initial cable forces and the jacking dis-
tance in a construction stage analysis

The continuous change of structural systems during construction of a cable-stayed bridge con-
siderably affects the distribution of internal forces in the complete structure. Calculating the
initial cable forces in a construction stage analysis allows for the tensioning to take place at the
time of installation of each particular cable. Thus, the initial cable forces at the time of installa-
tion are directly determined and a backward analysis of the structure is not needed anymore.
When taking into account the construction process, the unit tensioning of the cables as well
as the dead load of the segments are not applied to the complete structure, but to the different
structural systems which exist at the individual construction stages. The construction sequence
of the example bridge is illustrated below:

Figure 6.15: Construction sequence (construction stages 1+2)


CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 119

Figure 6.16: Construction sequence (construction stages 3-9)


CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 120

Figure 6.17: Construction sequence (construction stages 10-13)


CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 121

Figure 6.18: Construction sequence (construction stage 14)

The construction activities in the construction stages 1 to 14 are summarized in Table 6.7.

Table 6.7: Construction activities in the construction stages 1 to 14


Structure Boundary Load
Stage
Activation Deactivation Activation Deactivation Activation Deactivation

Node 1
Girder 1-3 SW Girder 1-3
CS01: - Node 101 - -
Pylon Rigid Link SW Pylon

CS02: Cable 2 - - - Tension 2 -


CS03: Cable 3 - - - Tension 3 -
D/C-12
CS04: - - - - -
Seg-12
CS05: Girder 4 - - - SW Girder 4 Seg-12
CS06: Cable 1 - - - Tension 1 -
CS07: Cable 4 - - - Tension 4 -
D/C-16
CS08: - - - - D/C-12
Seg-16
CS09: Girder 5 - - - SW Girder 5 Seg-16
CS10: Cable 5 - - - Tension 5 -
D/C-20
CS11: - - - - D/C-16
Seg-20
CS12: D/C-20
Key - - -
Seg-20
CS13: - - Node 22 - - -
CS14: - - - - Jack Up -
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 122

6.5.1 Calculation of the cable forces at the time of installation and the
jacking distance in a time-independent construction stage analysis

In order to calculate the initial cable forces at the time of installation, a time-independent con-
struction stage analysis is performed by MIDAS/Civil and Larsa2000. The unit tension force
of 10 kN of a cable is applied when the particular cable is installed. In contrast to the linear
static analysis the cable forces are not applied as internal, but as external loads. The method to
apply the cable forces as external loads in MIDAS/Civil and Larsa2000 was already explained
in Chapter 4.3.5.2.
The girder segments 4 and 5 as well as the key segment are installed tangentially, i.e a new acti-
vated segment is constructed by shifting the end node down to match the angular displacement
of the last node of the previously installed segment. The procedure to tangentially apply new
members in both analysis programs is described in Chapter 4.3.2.

6.5.1.1 Calculation by MIDAS/Civil

The procedure of calculating unknown load factors in a construction stage analysis by MI-
DAS/Civil is illustrated in the flow chart in Figure 4.3. The bending moment distribution and
the deflected shape in the final construction stage under self weight, additional dead load, unit
tension forces in the cables and unit support jacking are shown in Figures 6.19 and 6.20. The
self weight of the deck segments and the unit tensioning of the cables are applied to the struc-
tural systems that are present at the time of their installation and the additional dead load is
applied to the complete structure in construction stage 14. In addition to the permanent loads
also the construction loads (derrick crane, segment lifting, etc.) are taken into account in the
construction process. Just like the permanent loads they have an influence on the final dead load
condition.

Figure 6.19: Bending moment distribution under dead load and unit forces [kNm], CS14
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 123

(dx=+153)

Figure 6.20: Vertical deformation under dead load and unit forces [mm], CS14

Case I: Restricted bending moments


In order to calculate the initial cable forces at the time of installation and the support movement
of node 22 for restricted bending moments, constraints are defined for the bending moment
in the final construction stage at node number 4, 8, 12, 16, 20 and 101. The values for the
constraints are given in Table 6.4. The cables forces and the support jacking calculated by the
unknown load factor function in MIDAS/Civil are shown in Table 6.8 and the resulting bending
moment distribution and vertical deformation of the structure when the calculated forces are
applied are illustrated in Figures 6.21 and 6.22.

Table 6.8: Ideal cable forces at time of installation and ideal support
jacking for restricted bending moments, Case I
Load: Cable 1 Cable 2 Cable 3 Cable 4 Cable 5 Jack Up
(kN ) (kN ) (kN ) (kN ) (kN ) (mm)
Ideal force: 3647.10 880.52 1471.47 2425.77 3281.85 21.45

Figure 6.21: Moment distribution in CS14 when restricting bending moments [kNm], Case I
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 124

(dx=0.00)

Figure 6.22: Vertical deflection in CS14 when restricting bending moments [mm], Case I

Case II: Restricted deformations


In order to calculate the initial cable forces at the time of installation and the support movement
of node 22 for restricted deformations, constraints are defined in the final construction stage for
the vertical deflection at node number 4, 12, 16, 20 and 22 and for the horizontal deflection at
node number 106. For all these deformations a value of zero is desired. The cables forces and
the support jacking calculated by the unknown load factor function in MIDAS/Civil are given
in in Table 6.9 and the resulting bending moment distribution and vertical deformation of the
structure when the calculated forces are applied are illustrated in Figures 6.23 and 6.24.

Table 6.9: Ideal cable forces at time of installation and ideal support
jacking for restricted deformations, Case II
Load: Cable 1 Cable 2 Cable 3 Cable 4 Cable 5 Jack Up
(kN ) (kN ) (kN ) (kN ) (kN ) (mm)
Ideal force: 3750.53 1277.39 2364.82 1065.36 2456.73 −205.22

Figure 6.23: Moment distribution in CS14 when restricting deformations [kNm], Case II
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 125

(dx=0.00)

Figure 6.24: Vertical deflection in CS14 when restricting deformations [mm], Case II

6.5.1.2 Calculation by Larsa2000

The procedure of calculating unknown load factors is similar to the linear static analysis. The
only difference is that the unit tensioning of the cables is applied at the time of the cable installa-
tion and the influence on the structure is computed in the final construction stage. The bending
moment distribution and the deflected shape of the example model under dead load and unit
forces in the final construction stage are shown in Figure 6.25 and 6.26. The diagrams almost
equate those determined by MIDAS/Civil. Only small deviations in the bending moments and
deflections occur that result from the fact that a construction stage analysis in Larsa2000 is prin-
cipally a nonlinear analysis. However, as long as P-delta and large displacement effects are not
included in the analysis the difference is small (<1%) and is therefore neglected in the follow-
ing. The Larsa2000 results are treated as if they came from a linear construction stage analysis.
Only small adjustments of the unit forces are necessary to compensate for the nonlinear effects.

-38991
-32286
-23504
-15682

-46736

Figure 6.25: Bending moment distribution under dead load and unit forces [kNm], CS14
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 126

dz= -11.13,
dx=+152.64

43.84
-286.71
-843.14
-1566.85
-1944.02

Figure 6.26: Vertical deformation under dead load and unit forces [mm], CS14

By applying each unit tension force (10 kN ) and the unit support movement separately to the
structure, the influence on the objectives in the final construction stage can be obtained. The
resulting influence matrices for the moment (Case I) and displacement (Case II) restriction are
shown below.

Case I: Restricted bending moments

 
+40.6836 −52.9580 +11.3965 +4.4473 −0.0186 +8.0449
+13.8203 0.0000 +122.2637 +24.9434 +0.5137 +2.2793
 
 
 
 +0.0078 0.0000 0.0000 +84.1484 +47.4180 +17.3828 
Amoment =  

 0.0000 0.0000 0.0000 0.0000 +61.3555 +30.9688 

0.0000 0.0000 0.0000 0.0000 0.0000 +20.7705
 
 
+97.9180 185.3516 −39.8867 −1.3867 +0.3555 +22.0000

Case II: Restricted deformations

 
−0.09358 +0.11015 −0.09672 −0.02419 −0.00027 −0.01729
+0.22596 −0.18644 +0.79462 +0.28670 +0.06917 +0.06956
 
 
 
 +0.47189 −0.37163 +1.75923 +1.07712 +0.48882 +0.28026 
Adispl. = 

 +0.71752 −0.55695 +2.72310 +1.98412 +1.31667 +0.71669 

+0.84019 −0.64945 +3.20458 +2.43735 +1.77264 +1.00005
 
 
−0.31979 −0.60534 +0.13027 +0.00453 −0.00115 −0.07185
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 127

The first to the sixth row in the moment influence matrix, A moment , describe the bending mo-
ments [kNm] at node 4, 8, 12, 16, 20 and 101 in CS14 due to a unit post-tensioning of 10kN in
cable 1 to 5 and a unit upward movement of 1mm of joint 22. The first to the sixth row in the
displacement influence matrix, Adispl , describe the vertical deflection [mm] at node 4, 12, 16,
20 and 22 and the horizontal deflection at node 106 [mm] due to the unit loads.

The bending moments [kNm] and displacements [mm] at the selected nodes in CS14 due to the
permanent and construction loads are the following:

   
−15685.08105 +43.94349
−32447.93359 −287.90420
   
   
   
 −39123.04297   −846.56829 
MP =   δP = 

 −23565.32422 


 −1573.03834 

−2424.57568 −1951.62916
   
   
−46978.55859 +153.42693

The constraint vectors for a restricted moment distribution and restricted deformations are given
below (see Table 6.4):

   
−2350 0
−2600 0
   
   
   
 −2500   0 
Mdesired =   δdesired = 

 −2700 


 0 

−2000 0
   
   
0 0

The initial cable forces and the upward movement of node 22 can be calculated by solving the
same linear equation as for the linear static analysis. The resulting tension forces and support
movement are:

Case I: Case II:


   
+3705.806 +3489.802
+888.388 +1053.356
   
   
   
 +1504.592   +1756.306 
Tmoment =   Tdispl = 

 +2451.442 


 +2246.385 

+3297.552 +3354.321
   
   
+20.441 +21.882
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 128

Because of the slight nonlinearity of the Larsa2000 calculation, the above calculated cable
forces and support movement do not exactly cause the desired final condition. Small deviations
to the constraints occur that can be adjusted by adding a ∆T to the calculated values. The ad-
justment load vector, ∆T, can be calculated by multiplying the error by the particular influence
matrix. The resulting cable forces and the upward movement of node 22 differ only little from
the initial values. The adjusted loads are given below:

Case I: Case II:


   
+3649.067 +3447.845
+878.008 +1042.967
   
   
   
+1473.010 +1721.642
Tnew Tnew
   
m = Tm + ∆Tm =  ; d = Td + ∆Td =  

 +2436.594 


 +2227.100 

+3287.779 +3375.160
   
   
+21.439 +16.643

Figures 6.27 to 6.30 illustrate the adjusted moment distribution and vertical deflections when
the above calculated adjusted cable forces and the adjusted upward movement of joint 22 are
applied.

-2600 -2500 -2700


-2350 -2000

1136 1349 1247 1224 1482

Figure 6.27: Moment distribution in CS14 when restricting bending moments [kNm], Case I
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 129

dz=-12.69,
dx= 0.00

-1.88 -6.69 -9.06 -11.45 -12.71

Figure 6.28: Vertical deflection in CS14 when restricting bending moments [mm], Case I

-3906 -3946

-2311 -2100

-308

641 739
1913 1619
1928

Figure 6.29: Moment distribution in CS14 when restricting deformations [kNm], Case II

dz=-13.01,
dx= 0.00

0.00 0.00 0.00 0.01 0.01

Figure 6.30: Vertical deformation in CS14 when restricting deformations [mm], Case II
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 130

6.5.1.3 Summary of the determination of the cable forces and jacking distance in a time-
independent construction stage analysis

The initial cable forces at the time of installation and the jacking of the right support deter-
mined from the time-independent construction stage analyses by MIDAS/Civil and Larsa2000
are summarized in Table 6.10. It is visible that in case of restricting the bending moments the
results determined by MIDAS/Civil and Larsa2000 almost equate each other. Only small devi-
ations of less than 1% occur that can be ascribed to the fact that a construction stage analysis by
Larsa2000 is principally a nonlinear analysis. Even if second-order theory and large displace-
ment effects are not included in the analysis and no elements with nonlinear material properties
are used, Larsa still considers the stress-stiffening effect of the members. The bending moment
diagrams in Figure 6.21 and 6.27 show that the moment constraints are satisfied when the cal-
culated loads are applied.
This nonlinear behavior of the structure in the analysis by Larsa2000 not only results in devi-
ations compared to the results from a linear construction stage analysis, but also requires an
iteration process to determine the initial cable forces and the movement of the right support that
cause the desired condition in the final construction stage. However, the nonlinear effects are
small and only one adjustment step is needed to obtain satisfactorily results.

Table 6.10: Comparison of the ideal cable forces at time of installation and the ideal support
jacking determined by MIDAS/Civil and Larsa2000
Restriction Program Cable 1 Cable 2 Cable 3 Cable 4 Cable 5 Jack Up
(kN ) (kN ) (kN ) (kN ) (kN ) (mm)
Moments MIDAS 3647.10 880.52 1471.47 2425.77 3281.85 21.45
(Case I) Larsa 3649.07 878.01 1473.01 2436.59 3287.78 21.44
Displacements MIDAS 3750.53 1277.39 2364.82 1065.36 2456.73 −205.22
(Case II) Larsa 3447.85 1042.97 1721.64 2227.10 3375.16 16.64

Regarding the results for the adjustment of the displacements much larger differences between
the two analysis programs occurred. As visible from Figure 6.24 the resulting cable forces
and the support movement determined by the unknown load factor function in MIDAS/Civil
do not cause the desired deflected shape of the structure. The reason is that the unknown load
factor function does not consider the real displacements of the structure, but only the nodal
displacements (net displacement). The fact that the new segments are activated tangentially
is neglected in the unknown load factor calculation. When using the calculated values in a
construction stage analysis where the initial tangent displacement for erected structures option
is not activated, the desired constraints are satisfied (see Figure 6.31). However, the cable forces
and the support movement calculated by the unknown load factor function in MIDAS/Civil are
not useful for the real bridge construction. Using this function in a construction stage analysis
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 131

(dx=0.00)

Figure 6.31: Vertical deformation in CS14 when the initial tangent displacement
for erected structures option is not activated [mm], Case II (MIDAS)

only yields practical results for moment constraints as the bending moments do not change
when the segments are activated tangentially.
The deflected shape diagram in Figure 6.30 shows that when applying the initial cable forces
and support movement calculated by Larsa2000 to the structure, the desired constraints are
satisfied.

6.5.2 Calculation of the cable forces at the time of installation and the
jacking distance in a time-dependent construction stage analysis

Time effects on materials such as creep and shrinkage considerably influence the deformation
and the bending moment distribution during and after construction of a cable-stayed bridge. In
order to achieve a desired geometry and a desired stress condition at the time of completion of
the structure, time-dependent effects need to be taken into account in the determination of the
unknown load factors. In the following, the unknown load factors for the cable forces at the
time of installation and the jacking of the right support are computed in a time-dependent con-
struction stage analysis by MIDAS/Civil and Larsa2000. The restrictions are applied to the final
construction stage and the changes over time after completion of the bridge are presented. In
order to compensate for the time-dependent effects occurring during service life of the bridge,
the stay cables can be retensioned after a few years of bridge operation. The required retension
forces can be calculated in a similar manner as the initial cable forces. However, the determi-
nation of restressing forces of the stay cables is not the objective of this thesis and is therefore
not presented in this report.
The basis for the determination of the effects of creep and shrinkage is the CEB-FIB 1990 model
code. In addition to the material and section data of the example model, a relative humidity of
70% and a normal hardening cement are assumed. The construction time is shown in Figure
6.15 to 6.18.
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 132

6.5.2.1 Calculation by MIDAS/Civil

As already explained in Chapter 3.3.6.3 the unknown load factor function in MIDAS/Civil does
not allow for the consideration of time-dependent effects. Therefore, in order to determine the
initial cable forces and the support movement considering time effects on materials a similar
procedure as for the determination of unknown load factors by Larsa2000 is used. Because of
the linear relationship between the final internal forces and displacements and the elastic strain,
influence matrices containing the effects of creep and shrinkage can be obtained in a similar
manner as described for the time-independent construction stage analysis. The unit pretension
forces are applied at the time of installation, a time-dependent construction stage analysis in-
cluding creep and shrinkage is performed and then the influences on the objectives in the final
construction stage are computed.
The bending moment distribution and the deflected shape of the example model under perma-
nent and unit loads in the final construction stage of a time-dependent analysis by MIDAS/Civil
are shown in Figure 6.32 and 6.33.

Figure 6.32: Bending moment distribution under dead load and unit forces at day 89 [kNm]

(dx=+199)

Figure 6.33: Vertical deformation under dead load and unit forces at day 89 [mm]
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 133

By applying all unit loads separately to the structure, the influence on the objectives in the final
construction stage can be obtained. The resulting influence matrices for the moment (Case I)
and displacement (Case II) restriction are shown below.

Case I: Restricted bending moments

 
+36.0873 −36.3393 +3.0681 +1.1252 −0.5642 +8.0447
+7.0653 +4.0183 +88.3451 +11.1211 −3.1112 +2.2783
 
 
 
 −0.5960 +0.0124 −1.5306 +70.2103 +34.8914 +17.3832 
Amoment =  

 +0.0291 −0.0023 −0.0116 −0.8571 +54.1535 +30.9689 

+0.0055 +0.0063 +0.0114 +0.0158 +0.0170 +20.7702
 
 
−64.3278 −120.1794 +16.4919 −0.1673 −0.4807 −21.9989

Case II: Restricted deformations

 
−0.12226 +0.11965 −0.09779 −0.01458 +0.00408 −0.01729
+0.25904 −0.17118 +0.90636 +0.26574 +0.04203 +0.06957
 
 
 
 +0.52966 −0.33238 +2.00167 +1.17775 +0.47926 +0.28028 
Adispl. = 

 +0.79926 −0.49358 +3.09397 +2.24598 +1.44877 +0.71677 

+0.93409 −0.57417 +3.64014 +2.77940 +1.99177 +1.00000
 
 
−0.34656 −0.69098 +0.09614 −0.00058 −0.00233 −0.07185

The first to the sixth row in the moment influence matrix, A moment , describe the bending mo-
ments [kNm] at node 4, 8, 12, 16, 20 and 101 at day 89 due to the unit forces in cable 1 to 5 at
time of installation and the unit upward movement of joint 22. The first to the sixth row in the
displacement influence matrix, Adispl , describe the vertical deflection [mm] at node 4, 12, 16,
20 and 22 and the horizontal deflection [mm] at node 106 at day 89 due to the six unit loads.

The bending moments [kNm] and displacements [mm] at the selected nodes at day 89 due to
the permanent and construction loads are the following:

   
−14619.5649 +56.53704
−23015.1107 −350.43678
   
   
   
 −32414.0725   −1055.76985 
MP =   δP = 

 −22022.0781 


 −2007.93937 

−2464.0948 −2508.31960
   
   
+36488.1468 +200.28026
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 134

The constraint vectors for a restricted moment distribution and restricted deformations are given
below (see Table 6.4):

   
−2350 0
−2600 0
   
   
   
 −2500   0 
Mdesired =   δdesired = 

 −2700 


 0 

−2000 0
   
   
0 0

The initial cable forces and the upward movement of node 22 can be calculated by solving the
same linear equation as for the linear static analysis by Larsa2000. The resulting tension forces
and support movement are:

Case I: Case II:


   
+4170.551 +3912.032
+983.941 +1171.994
   
   
   
 +1728.723   +2063.412 
Tmoment =   Tdispl = 

 +2549.188 


 +2397.176 

+3482.837 +3803.304
   
   
+21.631 +35.280

Figures 6.34 to 6.37 illustrate the adjusted moment distribution and vertical deflections when
the above calculated cable forces and the upward movement of joint 22 are applied.

Figure 6.34: Moment distribution at day 89 when restricting bending moments [kNm], Case I
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 135

(dx+1.86)

Figure 6.35: Vertical deflection at day 89 when restricting bending moments [mm], Case I

Figure 6.36: Moment distribution at day 89 when restricting deformations [kNm], Case II

(dx=+0.00)

Figure 6.37: Vertical deformation at day 89 when restricting deformations [mm], Case II
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 136

After the completion of the bridge at day 89 the deflections and the bending moments continue
to change. Figure 6.38 and 6.39 show the bending moment distribution and the vertical defor-
mation of the bridge girder at day 90, 100, 1000 and 6000 when the cable forces and support
movement for restricted bending moments in the final construction stage are applied and the
cables are not retensioned. Regarding the deflected shape a fabrication camber that makes the
displacements at node 12, 16, 20 and 22 in the final construction stage to zero is assumed.
In Figure 6.40 and 6.41 the forces for restricted deformations are used. In this case a fabrication
camber is not needed, because the structure has been constrained to achieve a desired geometry.

-5000
-4000
Bending Moment [kNm]

-3000
-2000
-1000 0 10 20 30 40 50 60 70 80
0
1000
2000
3000
4000
5000

Day 90 Day 100 Day 1000 Day 6000

Figure 6.38: Redistribution of the girder bending moments after bridge completion, Case I

20

10
Vertical Deflection [mm]

0
0 10 20 30 40 50 60 70 80
-10

-20

-30

-40

-50

Day 90 Day 100 Day 1000 Day 6000

Figure 6.39: Changes of the vertical deflection of the girder after bridge completion, Case I
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 137

-5000
Bending Moment [kNm] -4000
-3000
-2000
-1000 0 10 20 30 40 50 60 70 80
0
1000
2000
3000
4000
5000

Day 90 Day 100 Day 1000 Day 6000

Figure 6.40: Redistribution of the girder bending moments after bridge completion, Case II

20

10
Vertical Deflection [mm]

0
0 10 20 30 40 50 60 70 80
-10

-20

-30

-40

-50

Day 90 Day 100 Day 1000 Day 6000

Figure 6.41: Changes of the vertical deflection of the girder after bridge completion, Case II
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 138

6.5.2.2 Calculation by Larsa2000

The procedure of calculating the unknown load factors considering time effects on materials by
Larsa2000 is similar to the determination by MIDAS/Civil. However, also the time-dependent
construction stage analysis is principally a nonlinear analysis and thus, an adjustment of the ini-
tial cable forces and support jacking is needed to achieve the desired condition. The adjustment
process is identical with the adjustment in the time-independent analysis by Larsa2000.
The bending moment distribution and the deflected shape of the example model under perma-
nent and unit loads in the final construction stage of a time-dependent analysis by Larsa2000
are shown in Figure 6.42 and 6.43.

-31307

-20989 -21656
-14224

-34131

Figure 6.42: Bending moment distribution under dead load and unit forces at day 89 [kNm]

dz= -20.28,
dx=+202.92

57.08
-354.95
-1071.73

-2039.57
-2547.93

Figure 6.43: Vertical deformation under dead load and unit forces at day 89 [mm]
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 139

By applying all unit loads separately to the structure, the influence on the objectives in the final
construction stage is obtained. The resulting influence matrices for the moment (Case I) and
displacement (Case II) restriction are shown below.

Case I: Restricted bending moments

 
+34.9961 −33.2070 +0.9160 +0.2529 −0.8271 +8.0449
+5.7461 +4.8516 +80.2246 +8.1895 −4.0605 +2.2793
 
 
 
 −0.7520 +0.0547 −2.0664 +67.5547 +32.9805 +17.3848 
Amoment =  

 −0.0195 −0.0176 −0.1113 −1.2383 +52.9727 +30.9688 

−0.0103 −0.0049 −0.0027 −0.0310 −0.0930 +20.7705
 
 
+56.5859 +105.4570 −10.8438 +0.6250 +0.4570 +22.0000

Case II: Restricted deformations

 
−0.12306 +0.12063 −0.09553 −0.01341 +0.00484 −0.01729
+0.25955 −0.17205 +0.89923 +0.26247 +0.03859 +0.06956
 
 
 
 +0.53024 −0.33355 +1.98460 +1.17922 +0.47708 +0.28026 
Adispl. = 

 +0.79894 −0.49496 +3.06511 +2.25210 +1.45578 +0.71692 

+0.93317 −0.57578 +3.60489 +2.78711 +2.00295 +1.00017
 
 
−0.34451 −0.69436 +0.09128 −0.00149 −0.00225 −0.07185

The meaning of the values of these matrices is similar to the definition of the data content of the
influences matrices determined by MIDAS/Civil.

The bending moments [kNm] and displacements [mm] at the selected nodes at day 89 due to
the permanent and construction loads are the following:

   
−14233.9902 +57.21223
−21085.8262 −356.31371
   
   
   
 −31422.4980   −1075.84572 
MP =   δP = 

 −21738.3574 


 −2047.36423 

−2435.6084 −2557.68442
   
   
−34305.7031 +203.93158
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 140

The constraint vectors for a restricted moment distribution and restricted deformations are given
below (see Table 6.4):

   
−2350 0
−2600 0
   
   
   
 −2500   0 
Mdesired =   δdesired = 

 −2700 


 0 

−2000 0
   
   
0 0

The initial cable forces and the upward movement of node 22 are calculated by solving the lin-
ear equation. The resulting tension forces and support movement are:

Case I: Case II:


   
+4340.053 +3967.158
+1034.033 +1181.011
   
   
   
 +1836.839   +2131.797 
Tmoment =   Tdispl = 

 +2604.358 


 +2436.725 

+3525.036 +3790.335
   
   
+23.202 +48.655

Just like in the time-independent analysis these values need to be slightly adjusted to account
for the nonlinear behavior of the structure. The adjusted values are:

Case I: Case II:


   
+4276.289 +3917.606
+1023.271 +1171.305
   
   
   
+1802.405 +2090.098
Tnew Tnew
   
m = Tm + ∆Tm =  ; d = Td + ∆Td =  

 +2589.618 


 +2425.669 

+3514.455 +3799.826
   
   
+23.199 +40.432

Figures 6.44 to 6.47 illustrate the moment distribution and vertical deflections when the above
calculated adjusted cable forces and the adjusted upward movement of joint 22 are applied.
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 141

-2600 -2500 -2700


-2350 -2000

1136 1349 1247 1224 1482

Figure 6.44: Moment distribution at day 89 when restricting bending moments [kNm], Case I

dz=-22.30,
dx= -3.46

-3.53 -12.54 -33.45 -71.04 -92.67

Figure 6.45: Vertical deflection at day 89 when restricting bending moments [mm], Case I

-3983

-2403
-1644
-578 -634

620
1731
2387 2356
2701

-401

Figure 6.46: Moment distribution at day 89 when restricting deformations [kNm], Case II
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 142

dz=-22.54,
dx= 0.00

0.00 0.00 0.00 0.01 0.01

Figure 6.47: Vertical deformation at day 89 when restricting deformations [mm], Case II

After the completion of the bridge at day 89 the deflections and the bending moments continue
to change. Figure 6.48 and 6.49 show the bending moment distribution and the vertical defor-
mation of the bridge girder at day 90, 100, 1000 and 6000 when the cable forces and support
movement for restricted bending moments in the final construction stage are applied. Again a
fabrication camber that makes the displacements at node 12, 16, 20 and 22 in the final construc-
tion stage to zero is assumed.
In Figure 6.50 and 6.51 the forces for restricted deformations are used. Because of the geometry
restriction, a fabrication camber is not needed.

-5000
-4000
Bending Moment [kNm]

-3000
-2000
-1000 0 10 20 30 40 50 60 70 80
0
1000
2000
3000
4000
5000

Day 90 Day 100 Day 1000 Day 6000

Figure 6.48: Redistribution of the girder bending moments after bridge completion, Case I
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 143

20.0000

10.0000
Vertical Deflection [mm]
0.0000
0 10 20 30 40 50 60 70 80
-10.0000

-20.0000

-30.0000

-40.0000

-50.0000

Day 90 Day 100 Day 1000 Day 6000

Figure 6.49: Changes of the vertical deflection of the girder after bridge completion, Case I

-5000
-4000
Bending Moment [kNm]

-3000
-2000
-1000 0 10 20 30 40 50 60 70 80
0
1000
2000
3000
4000
5000

Day 90 Day 100 Day 1000 Day 6000

Figure 6.50: Redistribution of the girder bending moments after bridge completion, Case II

20

10
Vertical Deflection [mm]

0
0 10 20 30 40 50 60 70 80
-10

-20

-30

-40

-50

Day 90 Day 100 Day 1000 Day 6000

Figure 6.51: Changes of the vertical deflection of the girder after bridge completion, Case II
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 144

6.5.2.3 Summary of the determination of the cable forces and jacking distance in a time-
dependent construction stage analysis

The initial cable forces at the time of installation and the jacking of the right support determined
from the time-dependent construction stage analyses by MIDAS/Civil and Larsa2000 are sum-
marized in Table 6.11. It is visible that the resulting forces determined by the different structural
engineering software programs are of similar magnitude. For both Case I and Case II only small
differences occur that result from the performance of a linear analysis in MIDAS/Civil and a
nonlinear analysis in Larsa2000 and from slight inaccuracies in the determination of creep and
shrinkage effects in both analysis programs as it was found in Chapter 4.4. However, in case
of restricting the bending moments as well as for the restricted deformations the desired final
condition was achieved when the calculated values for the cable forces and the jacking distance
were applied to the structure (Figures 6.34 to 6.37 and 6.44 to 6.47).

Table 6.11: Comparison of the ideal cable forces at time of installation and the ideal support
jacking determined in a time-dependent analysis by MIDAS/Civil and Larsa2000
Restriction Program Cable 1 Cable 2 Cable 3 Cable 4 Cable 5 Jack Up
(kN ) (kN ) (kN ) (kN ) (kN ) (mm)
Moments MIDAS 4170.49 983.72 1726.82 2549.15 3482.84 21.63
(Case I) Larsa 4276.29 1023.27 1802.41 2589.62 3514.46 23.20
Displacements MIDAS 3912.03 1171.99 2063.41 2397.18 3803.30 35.28
(Case II) Larsa 3917.61 1171.31 2090.10 2425.67 3799.83 40.43

Because the unknown load factor function can not be used in a time-dependent analysis, the
solution for the unknown load factors by MIDAS/Civil, as well as the solution by Larsa2000,
was done by manually determining influence matrices and solving linear equations. In MIDAS
the cable forces and support displacement resulting from the solution of the linear equation
exactly caused the target condition in the final construction stage. No iterative adjustment of
the forces was necessary to achieve the desired bending moment distribution or deflected shape,
which verifies the assumption that creep and shrinkage can be treated in a linear manner.
In order to achieve the desired condition using the cable forces and support jacking determined
by Larsa2000, similar to the time-independent analysis, a slight iterative adjustment of the
forces was necessary to compensate for the nonlinear effects. However, including time effects
on materials does not change the level of nonlinearity and still only one adjustment step is
needed to obtain satisfactorily results.
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 145

Time-dependent material effects not only need to be considered during the construction pro-
cess, but also after the completion of the bridge. By continuing the construction stage analyses
in MIDAS/Civil and Larsa2000 until day 6000, it was shown that the bending moments and
deformations due to creep and shrinkage considerably continue to change during service life of
the bridge. In all cases a high increase of the negative moment value at the pylon and a shifting
towards positive bending near the key segment was visible. Concerning the displacements, in
the middle of the main span an increase of up to 40mm occurred, which was mainly caused by
the time-dependent vertical deflection of the top of the pylon. When neglecting the effects of
creep and shrinkage in the pylon, the maximum increase of the girder deflections reduced to
10mm.
However, the behavior of the example model during and after construction in MIDAS/Civil and
Larsa2000 was almost identical. Figure 6.52 and 6.53 show the bending moments and defor-
mations along the bridge girder (Case I) at the end of construction and at day 6000 determined
by MIDAS/Civil and Larsa2000 by comparison. It is visible that the diagrams resulting from
the different analysis programs almost coincide, which verifies the results. The general belief
that creep has the tendency to change the internal forces into the direction of the continuous
beam condition does not apply to the example structure as it will also be the case for most other
cable-stayed bridges. All concrete or composite cable-stayed bridges constructed and loaded in
stages will not behave according to this belief. Different segment ages, different loading ages
and creep and shrinkage effects in the pylon lead to deviant behavior of a structure over time.

-5000
-4000
Bending Moment [kNm]

-3000
-2000
-1000 0 10 20 30 40 50 60 70 80
0
1000
2000
3000
4000
5000

MIDAS/Civil - day 90 MIDAS/Civil - day 6000


Larsa2000 - day 90 Larsa2000 - day 6000

Figure 6.52: Changes of the vertical deflection of the girder after bridge completion, Case II
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 146

20

10
Vertical Deflection [mm]
0
0 10 20 30 40 50 60 70 80
-10

-20

-30

-40

-50

MIDAS/Civil - day 90 MIDAS/Civil - day 6000


Larsa2000 - day 90 Larsa2000 - day 6000

Figure 6.53: Changes of the vertical deflection of the girder after bridge completion, Case II

6.5.3 Summary of the calculation of the initial cable forces and jacking
distance in a construction stage analyis

The above calculations show that time dependent effects have a significant influence on the
structural behavior of cable-stayed bridges during and after construction. In order to clarify the
concern to take time-dependent effects into account when calculating the initial cable forces,
Tables 6.12 and 6.13 summarize the cable forces that need to be applied at the time of cable
installation and the jacking distance determined in the time-independent and time-dependent
analysis. It is visible that in all cases the values determined in a time-dependent analysis were
higher than those calculated in a time-independent analysis. Generally, the magnification was
less than 25%, but in case of restricting deformations even an increase of 143% occurred for the
jacking distance (Larsa2000).

Table 6.12: Comparison of the cable forces at time of installation and the ideal support jacking
determined in a time-independent and time-dependent analysis by MIDAS/Civil
Case Time-independent/ Cable 1 Cable 2 Cable 3 Cable 4 Cable 5 Jack Up
time-dependent (kN ) (kN ) (kN ) (kN ) (kN ) (mm)
Time-independent 3647.10 880.52 1471.47 2425.77 3281.85 21.45
Case I Time-dependent 4170.49 983.72 1726.82 2549.15 3482.84 21.63
Deviation +14.4% +11.7% +17.4% +5.1% +6.1% +0.8%
Time-independent 3750.53 1277.39 2364.82 1065.36 2456.73 −205.22
Case II Time-dependent 3912.03 1171.99 2063.41 2397.18 3803.30 35.28
Time-independent results are wrong (see above) → No comparison
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 147

Table 6.13: Comparison of the cable forces at time of installation and the ideal support jacking
determined in a time-independent and time-dependent analysis by Larsa2000
Case Time-independent/ Cable 1 Cable 2 Cable 3 Cable 4 Cable 5 Jack Up
time-dependent (kN ) (kN ) (kN ) (kN ) (kN ) (mm)
Time-independent 3649.07 878.01 1473.01 2436.59 3287.78 21.44
Case I Time-dependent 4276.29 1023.27 1802.41 2589.62 3514.46 23.20
Deviation +17.2% +16.5% +22.4% +6.3% +6.9% +8.2%
Time-independent 3447.85 1042.97 1721.64 2227.10 3375.16 16.64
Case II Time-dependent 3917.61 1171.31 2090.10 2425.67 3799.83 40.43
Deviation +13.6% +12.3% +21.4% +8.9% +12.6% +143%

The deviations in geometry (no precamber) and bending moments of the final structure when
the forces calculated in a time-independent analysis are applied in a time-dependent analysis are
exemplary shown for the Case I-model in MIDAS/Cvil in Figures 6.54 and 6.55. The graphs
clearly show the importance of considering time effects on material when calculating the cable
tension forces.

-6000
Bending Moment [kNm]

-5000
-4000
Cable forces determined in
a time-dependent analysis
-3000
-2000 Cable forces determined in
-1000 0 10 20 30 40 50 60 70 80 a time-independent analysis
0
1000
2000

Figure 6.54: Final bending moment distribution along the bridge girder, Case I (MIDAS/Civil)

50
Vertical Deflection [mm]

0
Cable forces determined in
-50 0 10 20 30 40 50 60 70 80
a time-dependent analysis
-100
-150 Cable forces determined in
-200 a time-independent analysis
-250
-300
-350

Figure 6.55: Final vertical deformations of the bridge girder, Case I (MIDAS/Civil)
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 148

6.6 Performance of a nonlinear time-dependent construction


stage analysis

In order to demonstrate the effects of second-order theory and large displacements, a time-
dependent construction stage analysis of the example cable-stayed bridge including geometric
nonlinearity is performed. Because in MIDAS/Civil it is not possible to consider nonlinear and
time-dependent effects simultaneously, this is only done by the analysis program Larsa2000.
During the construction stage analysis the same initial cable and jacking forces as determined
in the ”linear” construction stage analysis are applied. The resulting final bending moments and
deformations of the bridge girder determined in the analysis including and not including second-
order and large displacement effects are exemplified for the Case I-model in Figure 6.56 and
6.57. It is visible that the values from both analyses differ only marginal and thus the effects of
second-order theory and large displacements can be neglected in the analysis of the considered
structure. This finding verifies the statement in Chapter 3.3.2 that at least large displacement
effects generally only need to be considered in large cable-stayed bridges having spans of more
than 600m. For small cable stayed bridges, such as the considered example bridge, these effects
are of minor importance and do not need to be taken into account in the design process.
Because cable elements in Larsa2000 do not consider sag effects unless they are divided into
a number of elements, the nonlinear behavior of the cables due to sag was not included in the
presented nonlinear analysis.

-3000
Bending Moment [kNm]

-2000
Neglecting geometric
-1000 nonlinearity
0 10 20 30 40 50 60 70 80
Including geometric
0 nonlinearity

1000

2000

Figure 6.56: Bending moment distribution along the bridge girder at day 89, Case I (Larsa2000)

25
Vertical Deflection [mm]

0 Neglecting geometric
nonlinearity
0 10 20 30 40 50 60 70 80
-25 Including geometric
nonlinearity
-50

-75

-100

Figure 6.57: Vertical displacements of the bridge girder at day 89, Case I (Larsa2000)
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 149

6.7 Amendment of discrepancies during the erection process

As already described in detail in Chapter 5, unavoidable errors in cable tension forces, geometry
of girder and pylon, dead load, stiffness of members, etc. may possibly cause unexpected dis-
crepancies between the actual and theoretical state of a cable-stayed bridge during construction.
It is hardly ever the case that the erection process exactly follows the predicted construction
sequence.
In the following, different error factors are introduced into the analysis model of the example
cable-stayed bridge and the construction stage analysis of the modified model is taken as a sim-
ulation of the actual construction process on site. Based on the discrepancies in geometry and
cable tension forces between the theoretical predictions (original model) and the actual mea-
surements (modified model), cable force and deck elevation adjustments are computed. The
goal is to achieve the desired final dead load condition without retensioning already installed
cables. The method that is used to obtain sufficient structural amendments follows the construc-
tion control procedure described in Chapter 5.4.
Because the functionality of the structural engineering programs Larsa2000 and MIDAS/Civil
has already been tested and proved during the calculation of the initial cable forces and jack-
ing distance in Section 6.4 and 6.5, in order to limit the length of this report, the following
calculations are only performed by MIDAS/Civil. They could similarly be performed by the
analysis program Larsa2000, but due to the fact that a construction stage analysis in Larsa2000
is principally a nonlinear analysis the application of influence matrices is easier when using
MIDAS/Civil.

6.7.1 Simulation analysis of the construction process

For the example model it is aspired to achieve a desired final geometry and bending moment
distribution simultaneously. Therefore, the initial cable forces and support jacking determined
for a restricted bending moments (Case I) are used for the construction stage analysis and the
segments are cambered so as to achieve a zero vertical displacement at the anchorage points
of cable 3 to 5 and at the end of the bridge. According to Section 6.5.2.1 the cable forces and
jacking distance that need to be applied to achieve the desired condition are:

Table 6.14: Cable forces and jacking distance applied in the construction stage analysis
Load: Cable 1 Cable 2 Cable 3 Cable 4 Cable 5 Jack Up
(kN ) (kN ) (kN ) (kN ) (kN ) (mm)
Ideal force: 4170.55 983.94 1728.72 2549.19 3482.84 21.63
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 150

The angles that have to be introduced at the joint between two segments in order to achieve a
zero vertical displacement at the anchorage points of cable 3 to 5 and at the end of the bridge can
be evaluated from the fabrication camber graph. Because a zero vertical displacement is desired
at the control nodes (12, 16, 20 and 22), the fabrication camber results as the curve between
the up-side-down final displacements of these joints between segments (see Figure 6.35). The
final displacements of the girder segments 1 and 2 are small and thus, these segments are not
precambered. The resulting fabrication camber is given below:

140
120 122.12

100
Camber [mm]

95.68
80
60
48.34
40
20 18.68
0
8 12 16 20
Node number

Figure 6.58: Fabrication camber (Case I)

The initial cable forces and the precamber data are applied to the analysis model and a time
dependent construction stage analysis is performed by MIDAS/Civil. The resulting cable forces
and joint deformations of the structure in the construction stages CS01 to CS14 are given in
Table 6.15 and 6.16. These values provide a reference for the real construction sequence.

Table 6.15: Theoretical cable forces during construction


Stage Cable 1 Cable 2 Cable 3 Cable 4 Cable 5
(kN ) (kN ) (kN ) (kN ) (kN )
CS01: - - - - -
CS02: - 983.94 - - -
CS03: - 2202.55 1728.72 - -
CS04: - 3605.38 4691.88 - -
CS05: - 3777.87 4786.54 - -
CS06: 4170.55 1179.37 5334.45 - -
CS07: 3939.30 1514.18 2288.31 2549.19 -
CS08: 7693.53 2159.75 3416.46 8302.48 -
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 151

Stage Cable 1 Cable 2 Cable 3 Cable 4 Cable 5


(kN ) (kN ) (kN ) (kN ) (kN )
CS09: 7693.53 2159.75 3416.46 8302.48 -
CS10: 7640.55 2172.52 2324.64 3311.41 3482.84
CS11: 10433.84 2684.41 2340.40 4405.38 5542.86
CS12: 8742.40 2380.94 2180.53 3652.39 4450.47
CS13: 8878.13 2395.58 2419.29 3763.66 4377.18
CS14: 8747.13 2373.26 2424.54 3713.69 4264.65

Table 6.16: Theoretical deformations during construction


Vertical deflection Horizontal
deflection
Stage Node 4 Node 8 Node 12 Node 16 Node 20 Node 22 Node 106
(mm) (mm) (mm) (mm) (mm) (mm) (mm)
CS01: −7.54 −1.26 −28.62 - - - 0.00
CS02: +0.16 −2.09 −67.54 - - - −60.22
CS03: −16.66 −2.07 +71.60 - - - −41.00
CS04: +21.61 −3.16 −84.07 - - - +55.32
CS05: +27.56 −3.41 −98.75 −260.94 - - +69.37
CS06: −4.18 −4.21 −20.70 −120.81 - - −49.19
CS07: −14.22 −4.25 +58.29 +165.55 - - −51.68
CS08: −3.55 −5.81 −9.24 −113.06 - - −11.21
CS09: −3.55 −5.81 −10.41 −113.06 −289.49 - −11.28
CS10: −4.75 −6.43 +16.13 +47.07 +118.44 - −10.83
CS11: +2.71 +7.43 −12.17 −65.75 −174.13 - +18.68
CS12: −2.05 −6.89 +8.32 +6.75 −14.38 −28.77 +0.97
CS13: −1.79 −7.37 −0.86 −4.74 −14.71 −21.63 +3.12
CS14: −2.25 −7.40 0.00 0.00 0.00 0.00 +1.86
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 152

6.7.2 Field measurements during erection

Since field measurements of the real construction process of the considered cable-stayed bridge
were not available, the actual erection of the structure is simulated in a construction stage analy-
sis by MIDAS/Civil that contains different error factors. The error factors that are incorporated
in the analysis model are shown in below:

Table 6.17: Error factors considered in the simulation of the actual erection process
Type of error Deviation from original input data
Moment of inertia of girder −0.12 m4
Modulus of elasticity of cables −3.0 · 107 kN/m2
Unit weight of concrete +1 kN/m3
Weight of derrick crane +200 kN
Relative humidity −10 %
Length of girder segment 3 −10 mm
Girder segment 5 is erected with an error pro- −20 mm
ducing an extra vertical displacement

During the ”real” erection process the deck elevations and the horizontal displacements of the
pylon as well as the tension forces of the cables are systematically measured in each erection
stage. The deck elevations are measured at the ends of the previously installed segments and
the horizontal displacement of the pylon is measured at the top end.
The temperature is assumed to remain constant during the erection process and monitoring of
the stresses in the girder and the pylon is not performed. The field measurements as well as the
comparison with the theoretical values, the forecast of future evolution of the structure and the
adjustment of deck elevations and cable forces during construction are performed in a stage by
stage construction control in the following section.

6.7.3 Stage by stage construction control

As explained in chapter 5.4 the construction control consists of four tasks. The simulation of
the construction process has already been performed and the theoretical deformations and cable
tension forces are known. The other three steps of construction control are conducted in a stage
by stage consideration in the following.
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 153

6.7.3.1 Construction stage CS01 and CS02

In order to control discrepancies within an allowable tolerance, certain construction adjustments


can be performed timely during erection. In the following the error identification and the sub-
sequent construction adjustment are performed on the basis of discrepancies between the actual
and theoretical geometry of the structure in a given construction stage, while the tension forces
of the stay cables are observed for the references only but not for the controlling.

1. Field measurement and comparison with the theoretical values:

The discrepancies between the actual and theoretical deformations in construction stage CS01
and CS02 are given below:

Table 6.18: Comparison of theoretical and actual deformations in CS01 and CS02
CS01 CS02
Node Prediction Actual Error Prediction Actual Error
(mm) (mm) (mm) (mm) (mm) (mm)
Node 4 -7.54 -8.96 -1.42 +0.16 -0.64 -0.80
Node 8 -1.26 -1.31 -0.05 -2.09 -2.26 -0.17
Node 12 -28.62 -37.38 -8.76 -67.54 -85.42 -17.88
Node 106 (dx) 0.00 0.00 0.00 -60.22 -60.22 0.00

The initial tension force of cable 1 in construction stage 2 is equal in both the theoretical and
actual construction stage analysis. Thus, an error in cable force can be ruled out as being re-
sponsible for the discrepancies in geometry.

2. Error identification:

As explained in Chapter 5.4.3 the error identification is performed by assuming several pat-
terns of typical structural errors and approximating the actual error by their linear combination.
However, it was stated that the aim of an economic cable adjustment is not to know the accurate
structural error, but to predict the response of the structure in future. According to Fujisawa [13]
a good approximation of the final condition may also be reach if the assumed structural error
patterns are not components of the actual error. Therefore, the error patterns that are assumed
in the following are slightly different from the real error factors.
The error patterns that are assumed to be responsible for the discrepancies in the first two con-
struction stages are the following:

• Error 1: Error in stiffness of the bridge girder (+10%)


CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 154

• Error 2: Error in stiffness of the pylon (+10%)

• Error 3: Error in dead load of the bridge girder (+10%)

• Error 4: Error in compressive strength of concrete, fck (+10%)


(affects creep and shrinkage of concrete)

• Error 5: Error in elevation of girder 3, additional displacement at end of segment (+10mm)

In a linear analysis an error in the stiffness of the stay cables does not affect the geometry of
the structure in the first two construction stages. This error pattern is checked separately in
construction stage 3.

The error influence matrix, Aerr , of which the columns represent the responses to the errors 1
to 5 is given below:

 
+0.64 +0.05 −0.74 0.00 0.00
0.00 +0.11 −0.11 0.00 0.00
 
 
 

 +4.12 +0.18 −4.70 0.00 +10.00 

 0.00 0.00 0.00 0.00 0.00 
Aerr = 
 


 −0.36 +0.06 −1.06 +0.21 0.00 


 0.00 +0.13 −0.16 +0.04 0.00 

+5.78 +0.19 −6.72 +1.28 +10.00
 
 
0.00 +5.47 0.00 0.00 0.00

In this matrix the first four rows represent the changes in deformation of node 4, 8, 12 and 106
in CS01 [mm] and the last four rows the changes in deformation of node 4, 8, 12 and 106 in
CS02 [mm] due to the error patterns of unit.

According to Chapter 5.4.3 the vector of the multipliers for respective error patterns can be
calculated as:

X = (ATerr WT W Aerr )−1 ATerr WT W D (6.1)

where D is the vector of errors as given in Table 6.18:


 
DT = −1.42 −0.05 −8.76 0.00 −0.80 −0.17 −17.88 0.00
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 155

Neglecting the weight of respective errors in the objective value, the multiplication factors that
minimize the difference between the actual and approximated errors are:
 
−1.664
 0.000 
 

X= 
 
 +0.477 

 −4.213 
 
+0.034

Hence, the following changes have to be made to the input data of the theoretical analysis to
predict the future evolution of the bridge:

Table 6.19: Changes of input data of the theoretical analysis model after error identification
in CS01 and CS02
Error Type of error Original Change of Modified
value value value
1 Stiffness of the girder (Egirder ) [kN/m2 ] 3.3282 · 107 −16.64% 2.7744 · 107
2 Stiffness of the pylon (Epylon ) [kN/m2 ] 3.3282 · 107 0.00% 3.3282 · 107
3 Dead load of the bridge girder [kN/m] 109.50 +4.77% 114.72
4 Compressive strength of concrete fck 35.00 −42.13% 20.25
[M N/m2 ]
5 Error in elevation of girder 3 [mm] 18.68 +0.34 19.02

3. Prediction of the future evolution of the bridge and adjustment of the cable forces,
jacking distance and deck elevations:

When the above calculated modifications are incorporated in the analysis model a forecast of the
final condition of the structure can be made. For the considered cable-stayed bridge the bending
moment distribution of an equivalent continuous beam is desired in the final construction stage.
A zero displacement can be achieved simultaneously by adjusting the precamber of the bridge.
Thus, a prediction of the bending moment distribution in the final construction stage is made
and then the cable forces and the support jacking are adjusted so as to reduce the residual error
between the final bending moment distribution and the desired bending moment distribution.
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 156

If the original cable forces are used the bending moments at node 4,8,12,16,20 and 101 result
to:
 
−3381.36
 −3563.71 
 
 
 −3434.25 
Mf inal =  −3706.05 

 
 −2227.92 
 

+2758.98

Thus, the vector of discrepancies between the desired condition and the forecast is:
     
−2350.00 −3381.36 +1031.36
 −2600.00   −3563.71   +963.71
     

     
 −2500.00   −3434.25 
Df inal = Mdesired − Mf inal =   =  +934.25
 
− 
 −2700.00   −3706.05   +1006.05 
     
 −2000.00   −2227.92   +227.92
     

0.00 +2758.98 −2758.98

By separately increasing the tension forces of cable 1, 3, 4, and 5 by 10 kN and the jacking
distance by 1mm a modified influence matrix can be determined of which the columns represent
the changes in bending moments at the control points due to a unit change of the cable forces
and a unit upward movement of joint 22. Because cable 2 has already been installed and a
retensioning shall be avoided a unit variation of the tension force of cable 2 is not considered.
The resulting influence matrix is:
 
+33.88 +1.35 +0.25 −0.54 +6.99
 +4.25 +82.38 +6.16 −4.29 +0.89 
 
 
 −0.43 −1.49 +67.16 +29.47 +13.93 
Amoment =   +0.05 +0.02 −0.92 +52.53 +27.74 

 
 +0.01 +0.01 +0.03 +0.05 +19.62 
 

−58.15 +10.83 −0.44 −0.46 −21.01

The multiplication factors that minimize the residual discrepancies between the desired condi-
tion and the actual final state can be calculated according to Equation (6.2).

X = (AT WT W A)−1 AT WT W Df inal (6.2)


CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 157

When neglecting the weight matrices, the resulting multiplication factors are:
 
+38.847
 +9.117 
 

X= 
 
 +6.377 

+10.339
 
 
+16.871

Hence, the following cable and jacking force adjustments need to be made to achieve a bending
moment distribution in the final construction stage that differs as less as possible from the
desired bending moment distribution:

Table 6.20: Cable and jacking force adjustments after error identification
in CS01 and CS02
Force Original value Additional value Modified value
Cable 1 [kN] 4170.55 +388.47 4559.02
Cable 2 [kN] 983.94 - -
Cable 3 [kN] 1728.72 +91.17 1819.89
Cable 4 [kN] 2549.19 +63.77 2612.96
Cable 5 [kN] 3482.84 +103.39 3586.23
Jacking distance [mm] 21.63 +16.87 38.50

The bending moment distribution and vertical displacements that are reached in the final con-
struction stage when using the above calculated cable forces in the modified analysis model are
illustrated in Figure 6.59 and 6.60.

Figure 6.59: Forecast of the final bending moment distribution after adjustment in CS02
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 158

Figure 6.60: Forecast of the final vertical deflection after adjustment in CS02

Figure 6.60 shows the vertical deflection of the bridge girder when no precamber is applied to
the structure. In order to achieve a zero vertical displacement at the anchorage points of cable
3 to 5 and at the end of the bridge the original fabrication camber illustrated in Figure 6.58 has
to be modified. Figure 6.61 shows the original designed camber data and the required changes
due to the assumed error factors.
150 144.03

125 122.12
110.99
Camber [mm]

100 95.68 22

75
52.15
50 Original fabrication camber
48.34
19.02
25 Required fabrication camber
18.28
0 0 Actual fabrication camber after
adjustment in CS02
8 12 16 20
Node number

Figure 6.61: Actual fabrication camber after adjustment in CS02

The required fabrication camber is the curve between the up-side-down final displacements of
joint 12, 16, 20 and 22 as visible in Figure 6.60. The original fabrication camber basically
equates the design camber data shown in Figure 6.58. The only difference is that the real angle
of girder 3 is slightly bigger than in the design camber data. It was found from the error identi-
fication that girder 3 was installed with an angle producing an additional vertical displacement
of 0.34mm (see Table 6.19). Thus, the ordinate at joint 12 of the original fabrication camber is
0.34mm higher than in the design fabrication camber.
Girder 3 is already installed and therefore, no changes in elevation can be applied to this seg-
ment. Only the camber of the remaining girder segments can be adjusted to achieve a zero
displacement in the final structure. The adjusted fabrication camber equates the original cam-
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 159

ber data at node 12 and equates the required fabrication camber at node 16, 20 and 22.

The adjusted cable forces and jacking distance as well as the installation angles of the segments
that have not been installed yet are not only applied to the theoretical analysis, but also to the
real construction (analysis containing the real errors). Then the construction of the structure
proceeds to the next stage.

6.7.3.2 Construction stage CS03

In construction stage 3 the second cable (cable 3) is installed and tensioned to its adjusted initial
cable force. In a linear construction stage analysis the cable stiffness does not affect the behav-
ior of this cable, because the initial pretension force is applied as an external force, i.e. the
tension force of cable 3 remains unchanged as the initial pretension force during construction
stage 3. However, cable 2 was installed in the previous stage and thus, the initial pretension
force in this cable may reduce or increase due to the deformation of the support structure based
on its stiffness. Hence, in contrast to the first two construction stages the structural behavior of
the bridge in construction stage 3 is affected by the stiffness of the stay cables. Assuming the
error factors determined above are correct, based on the actual and theoretical deformations in
construction stage 3 the error in cable stiffness is investigated separately.

1. Field measurement and comparison with the theoretical values:

The discrepancies between the actual and theoretical deformations in construction stage CS03
are given below:

Table 6.21: Comparison of theoretical and actual deformations in CS03


Node Prediction Actual Error
(mm) (mm) (mm)
Node 4 -22.05 -21.37 +0.68
Node 8 -2.46 -2.24 +0.22
Node 12 +82.70 +83.71 +1.01
Node 106 (dx) -44.18 -38.48 +5.70

The theoretical values in Table 6.21 are the result from the analysis of the original model con-
taining the error factors calculated in CS01 and CS02. The actual values come from the analysis
model containing the ”real” errors. In both models the above calculated adjusted cable forces
are used.
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 160

Regarding the cable forces, the tension force in cable 3 is identical in both models and the ac-
tual force in cable 2 (2228.1 kN) is lower than the theoretical force in cable 2 (2263.4 kN). The
difference is small and therefore tolerable, but it may be an indication for a variation in cable
stiffness.

2. Error identification:

Because an error in stiffness of the stay cables could not be considered in CS01 and CS02, this
error pattern is considered separately in this construction stage. The influence of a change in
cable stiffness by 10% on the deformation of node 4, 8, 12 and 106 is:
 
+0.45
 +0.00 
Aerr = 
 

 −0.77 
−2.09

The procedure to determine the multiplication factor is similar to the procedure in CS01 and
CS02. However, in this case the calculation is much easier, because only one error pattern is
taken into account. The resulting multiplication factor is:

X = − 2.399

Hence, the stiffness of the stay cables has to be reduced by 23,99% which leads to a modulus
of elasticity of the cable of 1.5962 · 108 kN/m2 instead of 2.1000 · 108 kN/m2 .

3. Prediction of the future evolution of the bridge and adjustment of the cable forces,
jacking distance and deck elevations:

When the above calculated reduction in stiffness of the stay cables is incorporated in the analysis
model, the predicted bending moments at node 4, 8, 12, 16, 20 and 101 in the final construction
stage are:
 
−2151.99
 −3222.59 
 
 
 −3553.46 
Mf inal =  
 −3155.95 

 −1964.75 
 

+2610.25
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 161

Thus, the vector of discrepancies between the desired condition and the forecast is:
     
−2350.00 −2151.99 −198.01
 −2600.00   −3222.59   +622.59
     

     
 −2500.00   −3553.46   +1053.46
Df inal = Mdesired − Mf inal = 

  −  =  
 −2700.00   −3155.95 
   +455.95 
 
 −2000.00   −1964.75   −35.25
     

0.00 +2610.25 −2610.25

By separately increasing the tension forces of cable 1, 4 and, 5 by 10 kN and the jacking
distance by 1mm (cable 2 and 3 have already been installed) the modified influence matrix can
be determined as follows:
 
+32.58 +1.20 −0.54 +6.48
 +6.59 +11.31 −2.93 +2.18 
 
 
 −0.63 +68.60 +34.90 +14.90 
Amoment =  +0.05 −0.90 +53.31 +25.18 

 
 +0.01 +0.03 +0.04 +16.45 
 

−64.11 −0.14 −0.52 −21.75

The multiplication factors that minimize the residual discrepancies between the desired condi-
tion and the actual final state can be calculated according to Equation (6.2). The results are:
 
+25.928
 +12.078 
X= 
 

 +0.190 
+17.487

Hence, the following adjustments need to be made to minimize the residual error between the
bending moment distribution in the final construction stage and the desired condition:

Table 6.22: Cable and jacking force adjustments after error identification in CS03
Force Original value Additional value Modified value
(after CS02) (after CS03)
Cable 1 [kN] 4559.02 +259.28 4818.30
Cable 2 [kN] 983.94 - -
Cable 3 [kN] 1819.89 - -
Cable 4 [kN] 2612.96 +120.78 2733.74
Cable 5 [kN] 3586.23 +1.90 3588.13
Jacking distance [mm] 38.50 +17.49 55.99
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 162

The bending moment distribution and vertical displacements (no precamber) that are reached in
the final construction stage when using the above calculated cable forces in the modified analy-
sis model are illustrated in Figure 6.62 and 6.63.

Figure 6.62: Forecast of the final bending moment distribution after adjustment in CS03

Figure 6.63: Forecast of the final vertical deflection after adjustment in CS03

In order to achieve a zero or at least small vertical displacement at the anchorage points of cable
3 to 5 and at the end of the bridge, the fabrication camber of CS02 has to be modified. Figure
6.64 shows the camber data determined in CS02 and the required changes due to the reduction
in stiffness of the stay cables in CS03. Because girder segment 3 has already been installed, no
changes in camber can be applied to this segment. The actual camber curve equates the original
curve at node 12 and the required curve at node 16, 20 and 22.
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 163

156.63
150
144.03
125 120.87
110.99
Camber [mm]

100 22

75 Original fabrication camber


55.95
(after CS02)
50 52.15 Required fabrication camber
19.02
25
Actual fabrication camber after
18.27 adjustment in CS03
0 0
8 12 16 20
Node number

Figure 6.64: Actual fabrication camber after adjustment in CS03

6.7.3.3 Construction stages CS04 and CS05

1. Field measurement and comparison with the theoretical values:

The discrepancies between the actual and theoretical deformations in construction stage CS04
and CS05 are given below:

Table 6.23: Comparison of theoretical and actual deformations in CS04 and CS05
CS01 CS02
Node Prediction Actual Error Prediction Actual Error
(mm) (mm) (mm) (mm) (mm) (mm)
Node 4 +20.47 +23.48 +3.01 +30.68 +32.55 +1.87
Node 8 -3.52 -3.40 +0.13 -3.90 -3.71 +0.19
Node 12 -99.98 -100.17 -0.19 -129.85 -125.26 +4.59
Node 16 - - - -338.39 -323.62 +14.77
Node 106 (dx) +69.83 +71.66 +1.83 +96.09 +93.63 -2.46

The theoretical and actual cable forces construction stage 5 are given in Table 6.24. The actual
values of both installed cables are about 10% higher than the theoretical prediction (analysis
model including identified errors), which is an indication that the assumed structural parameters
do not correspond with the actual structural parameters.
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 164

Table 6.24: Comparison of theoretical and actual cable forces in CS05


Cable 1 Cable 2 Cable 3 Cable 4 Cable 5
(kN ) (kN ) (kN ) (kN ) (kN )
Prediction: - 3324.74 4666.47 - -
Actual value: - 3625.59 5150.21 - -
Deviation: - +9.05% +10.37% - -

2. Error identification:

Because the deviations between the theoretical and actual deformations and cable forces in
the construction stages 4 and 5 are significant, all error factors that have been detected in the
previous construction stages are again taken as potential error patterns in the following. Only
the stiffness of the pylon which was found to be correct is not checked again. In addition to
the previously assumed error factors, the variation of the weight of the derrick crane is also
assumed to be a component of the actual error. Thus, the error patterns that are assumed to be
responsible for the discrepancies in the fourth and fifth construction stage are the following:

• Error 1: Error in stiffness of the stay cables (+10%)

• Error 2: Error in stiffness of the bridge girder (+10%)

• Error 3: Error in dead load of the bridge girder (+10%)

• Error 4: Error in weight of the derrick crane (+10%)

• Error 5: Error in compressive strength of concrete, fck (+10%)

• Error 6: Error in elevation of girder 4, additional displacement at end of segment (+10mm)

The error influence matrix, Aerr , of which the columns represent the additional deformation at
node 4, 8, 12 and 106 in CS04 and at node 4, 8, 12, 16 and 106 in CS05 to the unit errors 1 to 6
is given below:
 
+0.03 −0.88 −1.44 +1.02 +0.27 0.00
 
 −0.01 0.00 −0.20 −0.03 +0.06 0.00 
 
 +6.30 −1.02 −8.98 −4.17 +1.23 0.00 
 
 −3.68 −3.24 0.00 +2.43 +0.55 0.00 
 

Aerr = 
 
 +0.03 −1.06 +2.11 +1.23 −0.21 0.00 

0.00 0.00 −0.27 −0.03 +0.08 0.00
 
 
 
 +7.57 −1.53 −25.09 −4.75 +2.55 0.00 
 
 +17.21 +2.64 −62.38 −10.16 +5.82 +10.00 
 
−4.50 −3.41 +9.15 +2.96 −0.70 0.00
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 165

The multiplication factors of the error patterns 1 to 6 are calculated as described above. The
resulting values are:
 
+1.147
 +0.701 
 
 
 −0.062 
X=  +2.684 

 
 +3.203 
 

−0.208

Hence, the following changes have to be made to the input data of the theoretical analysis to
predict the future evolution of the bridge:

Table 6.25: Changes of input data of the theoretical analysis model after error identification
in CS04 and CS05
Original Change of Modified
Error Type of error value value value
(after CS03) (after CS05)
1 Stiffness of the cables (Ecable ) [kN/m2 ] 1.5962 · 108 +11.47% 1.7793 · 108
2 Stiffness of the girder (Egirder ) [kN/m2 ] 2.7744 · 107 +7.01% 2.9689 · 107
3 Dead load of the bridge girder [kN/m] 114.72 −0.62% 114.01
4 Weight of derrick crane [kN ] 750.00 +26.84% 951.30
5 Compressive strength of concrete fck 20.25 +32.03% 26.73
[M N/m2 ]
6 Error in elevation of girder 4 [mm] 55.95 −2.08 53.87

3. Prediction of the future evolution of the bridge and adjustment of the cable forces,
jacking distance and deck elevations:

When the above calculated modifications are incorporated in the analysis model, a prediction
of the future evolution can be made. Then, similar to the amendments in the construction stage
CS02 and CS03, appropriate adjustments of the stay cable forces that have not been installed
yet can be calculated and the camber data can be adjusted to achieve the desired geometry of
the structure in the final construction stage. As the procedure to determine the adjustments
was already shown in detail in the previous adjustment calculation, in the following only the
resulting cable forces, jacking distance and camber data are presented.
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 166

Table 6.26: Cable and jacking force adjustments after error identification in CS05
Force Original value Additional value Modified value
(after CS05) (after CS09)
Cable 1 [kN] 4818.30 −54.08 4764.22
Cable 2 [kN] 983.94 - -
Cable 3 [kN] 1819.89 - -
Cable 4 [kN] 2733.74 +9.13 2742.87
Cable 5 [kN] 3588.13 +124.10 3712.23
Jacking distance [mm] 55.99 −15.00 40.99

The actual camber curve is illustrated in Figure 6.65. It is visible that the required elevation
of girder segment 3 and 4 differs from the actual elevation. However, because these segments
have already been installed their precamber can not be changed anymore. Only the remaining
segments can be cambered so as to achieve the desired geometry.

156.63
150
127.27
125 120.87
Camber [mm]

100 98.23

75 Original fabrication camber


53.87 (after CS03)
50 45.94 Required fabrication camber
19.02
25 Actual fabrication camber after
15.45 adjustment in CS05
0 0
8 12 16 20
Node number

Figure 6.65: Actual fabrication camber after adjustment in CS05

6.7.3.4 Construction stages CS06 to CS09

1. Field measurement and comparison with the theoretical values:

The discrepancies between the actual and theoretical deformations and cable forces in construc-
tion stage CS06 to CS08 were within tolerable limits. They are not presented here. The differ-
ences between the actual deformations and cable forces and those expected from the theoretical
analysis in construction stage 9 are given in Tables 6.27 and ??.
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 167

Table 6.27: Comparison of theoretical and actual deformations in CS09


Node Prediction Actual Error
(mm) (mm) (mm)
Node 4 -5.23 -5.88 -0.65
Node 8 -6.54 -6.42 +0.12
Node 12 -17.06 -13.84 +3.22
Node 16 -152.19 -144.98 +7.21
Node 20 -385.60 -399.04 -13.44
Node 106 (dx) -4.58 -5.83 -1.25

Table 6.28: Comparison of theoretical and actual cable forces in CS09


Cable 1 Cable 2 Cable 3 Cable 4 Cable 5
(kN ) (kN ) (kN ) (kN ) (kN )
Prediction: 8371.63 2138.89 3722.46 6745.43 -
Actual value: 8383.70 2142.27 3686.54 6777.96 -
Deviation: +0.14% +0.16% -0.96% +0.48% -

2. Error identification:

In construction stage 9, five different error patterns were assumed to be responsible for the
discrepancies in geometry. The multiplication factors have been calculated by the procedure
explained above. The changes to the input data of the theoretical analysis model are shown
below:

Table 6.29: Changes of input data of the theoretical analysis model after error identification
in CS09
Original Change of Modified
Error Type of error value value value
(after CS05) (after CS09)
1 Stiffness of the cables (Ecable ) [kN/m2 ] 1.7793 · 108 −2.62% 1.7327 · 108
2 Stiffness of the girder (Egirder ) [kN/m2 ] 2.9689 · 107 −1.48% 2.9250 · 107
3 Weight of derrick crane [kN ] 951.3 −0.93% 942.45
4 Compressive strength of concrete fck 26.73 +9.58% 29.29
[M N/m2 ]
5 Error in elevation of girder 4 [mm] 98.23 −24.15 74.08
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 168

3. Prediction of the future evolution of the bridge and adjustment of the cable forces,
jacking distance and deck elevations:

When the above calculated modifications are incorporated in the analysis model, a prediction of
the future evolution is made and appropriate adjustments of the tension force in the remaining
cable 5 and the jacking distance are calculated. Based on the predicted deflected shape in the
final construction stage the camber data is adjusted. The resulting cable forces are presented in
Table 6.30 and the actual fabrication camber is illustrated in Figure 6.66.

Table 6.30: Cable and jacking force adjustments after error identification in CS09
Force Original value Additional value Modified value
(after CS02) (after CS03)
Cable 1 [kN] 4764.22 - -
Cable 2 [kN] 983.94 - -
Cable 3 [kN] 1819.89 - -
Cable 4 [kN] 2742.87 - -
Cable 5 [kN] 3712.23 −14.89 3697.34
Jacking distance [mm] 40.99 0.00 40.99

150

125 127.27 22

106.03
Camber [mm]

100
80.82
75
74.08 Original fabrication camber
53.87 (after CS05)
50
Required fabrication camber
35.86
25 19.02
Actual fabrication camber
11.40 after adjustment in CS09
0 0
8 12 16 20
Node number

Figure 6.66: Actual fabrication camber after adjustment in CS09

6.7.3.5 Construction stages CS10 to CS14

In the construction stages CS10 and CS14 no further adjustments are made. The discrepancies
of the deformations and cable forces between the theoretical predictions and the actual values
are in tolerable limits. Only the vertical deflection of the end of the bridge needs to be slightly
adjusted which is done by an additional jacking force in the final construction stage. The exact
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 169

upward movement of joint 22 that is necessary to achieve a zero vertical displacement at the
end of the bridge is measured (taken from the simulation of the real construction) just before
completion of the bridge.
In the final construction stage before jacking the right support, the vertical deflection of node
22 in the real structure is −50.55 mm. From the theoretical analysis a jacking distance of
+40.99 mm was predicted. Thus, in order to achieve a zero vertical displacement at the end
of the bridge the support movement needs to be increased by 9.56 mm. The resulting bending
moments and deformations of the example cable-stayed bridge after completion are presented
in the following section.

6.7.4 Results of the construction control

During the real construction process the girder segments are installed with the angle given in
the graph of the actual fabrication camber. In the error identification it was found that among
other patterns in some construction stages an error in elevation of the installed segment was re-
sponsible for the discrepancies in geometry. For the theoretical analysis this error was assumed
to be true and was included in the camber data. However, in the real construction apart from
the additional vertical deflection of girder 5, no installation inaccuracy occurred. The segments
were installed with the angle given in the camber graph which was valid at the time of segment
installation and the elevation was not modified later on. Thus, the final theoretical camber data
slightly differs from the final real precamber of the bridge. Both curves are shown in Figure
6.67.

150
125 111.92
Camber [mm]

100 106.03
79.97
75 74.08
55.61
50 53.87
25 19.02 Real precamber
18.68 Theoretical precamber
0
8 12 16 20
Node number

Figure 6.67: Final fabrication camber

The final moment distribution and the deflected shape of the real structure after all adjustments
in cable and jacking forces are applied are illustrated in Figure 6.68 and 6.69. The precamber of
the segments was not applied in the analysis by MIDAS/Civil. Thus, the camber data given in
Figure 6.67 need to be added to the vertical deflections of the bridge girder to achieve the real
final girder deformation.
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 170

Figure 6.68: Final bending moment distribution of the real structure after adjustment

(dx=+4.53)

Figure 6.69: Final vertical displacements of the real structure after adjustment

From these figures it is visible that the bending moment distribution as well as the deflected
shape of the real structure is close to the desired final condition. An almost zero bending mo-
ment at the base of the pylon and evenly distributed bending moments along the bridge are
reached. The vertical displacement at the right end of the bridge equates the value given in the
fabrication camber and thus, a zero displacement is achieved at the end of the cantilever. At the
other nodes of the structure the vertical deflections do not vanish, but they are small enough to
be tolerated. In order to get an idea about the deflected shape of the whole bridge, Figure 6.70
shows the vertical deformation of the bridge girder when the precamber is applied. They are
compared to the results from the original analysis.
Figure 6.70 shows that the maximum vertical displacement in the main span of the girder is
less than 20mm. This maximum deflection is an upward deformation at node 16 which was
most probably caused by an inappropriate precamber of girder segment 4. Apparently, at the
time of installing girder 4 the prediction of the final vertical displacements was not accurate
enough. One reason may be the fact that, for example, the estimated modulus of elasticity
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 171

of the cables in CS03 was 1.5962 · 108 kN/m2 which is 11.32% lower than the actual cable
stiffness (Ereal = 1.8 · 108 kN/m2 ). Furthermore, it was found in CS02 that the compressive
strength of concrete was 42.13% less than in the original model, which most probably caused
an incorrect prediction of the effects of creep and shrinkage. The installation angle of girder 4
was determined based on these inaccurate input data which may be the reason for the higher
discrepancies in geometry in this region. Before installing the subsequent segments the input
data of theoretical model were updated again and better predictions of the future evolution of
the structure could be made. Thus, the camber data of these segments could be improved before
their installation.
However, this inaccuracy in the early construction stages caused an additional upward move-
ment in the middle of the main span, which will reduce due to the effects of creep and shrinkage
during service life of the bridge anyway. Figure 6.71 shows the vertical deflection of the original
analysis and the real construction at day 6000. At this time almost all time-dependent deforma-
tions have occurred and the final shape of the structure is reached. It is visible that the deformed
shape of both structures looks similar at this time.

20
Vertical deformation [mm]

10
0
-10 0 10 20 30 40 50 60 70 80

-20
-30
-40
-50

Real construction Original analysis

Figure 6.70: Final vertical displacements after adjustment including precamber

20
Vertical deformation [mm]

10

0
0 10 20 30 40 50 60 70 80
-10

-20

-30

-40

-50

Real construction Original analysis

Figure 6.71: Vertical displacements after adjustment including precamber at day 6000
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 172

During the construction of the example cable-stayed bridge, adjustments of the cable forces
have been made to reduce the residual discrepancies between the actual final dead load condition
and the desired state. However, to ensure the safety of the structure not only the internal forces
and deformations in the bridge girder and the pylon need to be controlled, but also the cable
tension forces.
It was explained in Chapter 5.4.4.1 that it is generally impossible to achieve the desired deck
profile/bending moment distribution and the cable forces calculated in the original stage analysis
simultaneously. The presented solution was to turn the cable forces over a suitable range to
reduce the effects of error factors on the bridge deck and the pylon. Anyhow, the stay cable
forces need to remain within tolerable limits to ensure that the cable properties are sufficient to
withstand the applied loads.
The initial cable tension forces at the time of installation and the cables forces in the complete
structure in the real construction and the original analysis are given in Table 6.31. It can be seen
that the differences between the initial cable forces in the original analysis and the real erection
are within ±15% and the deviations of the final cable forces are less than 5% which is tolerable.

Table 6.31: Comparison of the cable forces in the original analysis and the actual erection
Classification Cable 1 Cable 2 Cable 3 Cable 4 Cable 5
(kN) (kN) (kN) (kN) (kN)
Initial Original analysis 983.94 1728.72 4170.55 2549.19 3482.84
cable Real erection 983.94 1819.89 4764.22 2742.87 3697.34
forces Deviation ±0.00% +5.27% +14.23% +7.60% 6.16%
Final Original analysis 8747.13 2373.26 2424.54 3713.69 4264.65
cable Real erection 9036.10 2302.65 2525.48 3849.58 4278.67
forces Deviation +3.30% -2.98% +4.16% +3.66% +0.33%

6.7.5 Summary of construction control

During the construction of the example cable-stayed bridge, different errors in the input data
of the theoretical analysis model were detected. These errors do not necessarily correspond to
the real errors. In Table 6.32 the input data containing the error factors that were introduced in
the analysis model simulating the real construction process are compared with the input data
containing the error patterns that were identified in the respective construction stages.
It is visible that the identified error patterns in some cases correspond to the components of the
real error. The detected values of the girder stiffness (E · I), the stiffness of the stay cables, the
dead load of the girder and the weight of the derrick crane are within ±5% of the real values.
However, there are additional error factors in the real structure that have not been identified
such as the difference in the weight of the pylon (unit weight of concrete influences the weight
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 173

of all concrete members), the relative humidity and the discrepancy in length of segment 3.
Instead, error patterns such as the compressive strength of concrete and installation inaccuracy
of segment 3 and 4, have been identified that were not responsible for the discrepancies.
Furthermore, some of the variations in structural input parameters have not been identified
correctly (±5%) until the ninth construction stage or even until the end of construction. The
additional displacement that was introduced in the structure due to an inaccurate installation of
girder 5 was detected to be 21% higher than in reality. Hence, the error identification during the
construction of the example model pointed out some error patterns, but the approximated error
did not exactly coincide with the real error.

Table 6.32: Input data of the real structure and the theoretical analysis model
Real Theoretical analysis model
structure CS01 CS02 CS03 CS05 CS09
Moment of inertia of girder [m4 ] 0.80 0.92 0.92 0.92 0.92 0.92
Modulus of elasticity of the girder 3.33 3.33 2.77 2.77 2.97 2.93
[107 m4 ]
Modulus of elasticity of cables 1.80 2.1 2.1 1.59 1.78 1.73
[108 kN/m2 ]
Unit weight of concrete [kN/m3 ] 26.0 25.0 25.0 25.0 25.0 25.0
Dead load of the girder [kN/m] 113.88 109.50 114.72 114.72 114.01 114.01
Weight of derrick crane [kN] 950.0 750.0 750.0 750.0 951.3 942.5
Relative humidity 60% 70% 70% 70% 70% 70%
Compressive strength of concrete 35.00 35.00 20.25 20.25 26.73 29.29
[M N/m2 ]
Length of girder segment 3 [m] 15.99 16.00 16.00 16.00 16.00 16.00
Installation inaccuracy [mm]:
Girder 3 - +0.34 +0.34 +0.34 +0.34 +0.34
Girder 4 - - - - -2.08 -2.08
Girder 5 -20.00 - - - - -24.15

However, even if the identified error patterns did not exactly coincide with the actual error, it
was possible to satisfactorily predict the response of the structure in the final construction stage
and determine appropriate adjustments. It was shown above that the final bending moments and
deformations of the bridge girder of the real structure were within tolerable limits so that no
further retensioning of the stay cables was needed after the completion of the bridge.
CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 174

In order to clarify the improvement of the final dead load condition due to the cable force,
jacking and deck elevation adjustments during construction and to point out what would have
happened if no adjustments would have been made, Figures 6.72 and 6.73 show the final bend-
ing moment distribution and deflected shape of the bridge resulting from the the real analysis
including the errors with and without adjustments during erection. Additionally, the values
from the original analysis (desired condition) and the results from the continuously adjusted
theoretical analysis (prediction) are included.
It is visible that the resulting bending moments and particularly the vertical deflections of the
cantilever in the real structure are not acceptable when no adjustments are made during con-
struction. In this case a complete adjustment in the final construction stage is necessary to
achieve an appropriate deck profile. However, when adjusting the cable forces in the final con-
struction stage to achieve a desired geometry the bending moments need to be controlled as
well to avoid the development of excessively high stresses in the bridge girder.

-5000
-4000
Bending Moment [kNm]

-3000
-2000
-1000 0 10 20 30 40 50 60 70 80
0
1000
2000
3000
4000
5000

Original analysis Real construction, adjusted Prediction, adjusted Real construction, not adjusted

Figure 6.72: Bending moment distribution in the final construction stage

20

-20 0
Vertical deformation [mm]

10 20 30 40 50 60 70 80

-60

-100

-140

-180

-220

-260

original analysis Real construction, adjusted Prediction, adjusted Real construction, not adjusted

Figure 6.73: Vertical displacements in the final construction stage


CHAPTER 6. EXAMPLE CALCULATION OF A CABLE-STAYED BRIDGE 175

Moreover, Figure 6.72 and 6.73 show that, even if the error patterns were partly incorrect, the
analysis model including the identified errors gave a good approximation of the real behavior of
the structure. The bending moment distribution resulting from the prediction almost coincides
with the real bending moments and the differences between the vertical deflections of the pre-
diction and real analysis are zero in the side span and increase up to 9mm at the end of the main
span.

Recapitulating, it can be stated that the presented method to adjust discrepancies between the
theoretical and actual condition of a cable stayed bridge during construction leads to satisfac-
torily results. In the example calculation the residual error between the actual and desired final
condition is sufficiently small and thus, the validity of the assumption that error patterns may be
used for the prediction of the final state even if the identified errors do not absolutely coincide
with the real error factors is proved. Nevertheless, from the vertical deflections of almost 20mm
in the middle of the main span it becomes clear that the adjustment in early stages may lead to
incorrect adjustments. From the standpoint of final residual errors the adjustment in the final
construction stage is preferable. In this case a forecast of the future evolution is not needed,
which leads to clearly better results.
However, in the presented calculation the residual errors were sufficiently small which shows
that, if field measurements and adjustment calculations are performed carefully, fruitful results
can also be achieved in the construction control timely during erection. In contrast to the ad-
justment in the final construction stage, using this method, no expensive and work-consuming
restressing operations are required. Hence, it is much more economical and, especially for long
span cable-stayed bridges having a large number of cables, the adjustment during construction
is proposed.
Chapter 7

Summary and Conclusions

7.1 Summary

The present work concerns the analysis and control of cable-stayed bridges during construction
by the cantilevering method. A new method to determine an optimum tensioning strategy is
presented that allows for the consideration of several effects that are relevant for the design of
cable-stayed bridges, including the construction sequence, second-order theory, large displace-
ments, cable sag and time-dependent effects such as creep and shrinkage. In contrast to other
existing methods, the ”expanded unit load” method automatically computes the initial cable
forces that need to be applied at the time of erection of the cables to achieve a desired bending
moment distribution and/or desired deformation in the complete structure. The performance
of a traditional backward and subsequent forward analysis, which is usually accompanied by
problems in particular construction stages, is not necessary.

As a matter of fact the issue to control the stress distribution and the final geometry of a cable-
stayed bridge not only concerns the initial design of an optimum tensioning strategy, but also
requires a detailed construction control during the erection process. Due to inevitable errors be-
tween the structural design parameters an the actual ones, unexpected discrepancies may occur
between the predicted and actual state of the structure in a given construction stage. This thesis
presents an economically efficient method to control these discrepancies by the adjustment of
cable tension forces and deck elevations during the erection process. The method is based on
the identification of structural errors by the least square minimization, the forecast of the fu-
ture evolution of the structure and the determination of construction adjustments that reduce the
residual error in the complete structure.
CHAPTER 7. SUMMARY AND CONCLUSIONS 177

For the analysis of cable-stayed bridges there exist different standard structural analysis pro-
grams. Two of them, namely Larsa2000/4th Dimension and MIDAS/Civil, are demonstrated
in this thesis. Both programs have a particular emphasis on bridge structures and offer a con-
struction stage analysis feature, but their functionality in a construction stage analysis of a
cable-stayed bridge differs.
Larsa2000 is an American software program that allows for the consideration of material non-
linearity, second-order theory, large displacement effects and time effects on materials including
creep, shrinkage, concrete aging and relaxation of steel in a construction stage analysis, but the
sag effect of cables can not be taken into account. In a construction stage analysis by MI-
DAS/Civil, a Korean analysis program, almost the same relevant effects can be considered.
However, in contrast to Larsa/2000, MIDAS/Civil can include the nonlinear behavior of ca-
bles due to sag, but P-Delta effects can not be taken into account. An additional function in
MIDAS/Civil is the unknown load factor function that automatically computes the ideal post-
tensioning forces in the stay cables for a desired final condition. The basic idea of this function
corresponds to the ”expanded unit load” method, but it only works for time-independent, linear
construction stage analyses.

Concluding this thesis, the above mentioned methods to determine an optimum tensioning strat-
egy and to control the geometry and internal forces during the erection process are verified on
the basis of an example calculation of a simple cable-stayed bridge.
The initial cable forces are determined in a linear static, a time-independent and a time-dependent
construction stage analysis. The results show that the calculated cable forces in almost all cases
exactly lead to the predetermined moment distribution or deformation within the deck and the
pylon. Comparing the results determined by Larsa2000 and MIDAS/Civil they are almost iden-
tical. Only when using the unknown load factor function in MIDAS/Civil in a construction
stage analysis for restricted deformations problems occur. Apparently, this function neglects
the fact that segments are applied tangentially and, when a designated geometry is aspired, the
calculated cable forces do not yield the desired condition. Hence, cable forces determined by
the unknown load factor function in MIDAS/Civil for a restricted geometry are generally use-
less for the real bridge construction.
Comparing the resulting cable forces determined in the time-independent and time-dependent
construction stage analysis, it is visible that time effects on materials have a considerable influ-
ence on the structural behavior of concrete or composite cable-stayed bridges during construc-
tion. The post-tensioning forces that are needed to achieve the desired condition significantly
increase when considering creep and shrinkage.
The performance of a nonlinear, time-dependent construction stage analysis shows that geomet-
ric nonlinear effects are of minor importance for small cable-stayed bridges like the example
model.
CHAPTER 7. SUMMARY AND CONCLUSIONS 178

Regarding the construction control during the erection process, the example calculation shows
that the presented method to adjust discrepancies between the theoretical and actual condition
of a cable-stayed bridge leads to satisfactorily results. A considerable improvement of the
resulting bending moment distribution and deformed shape of the adjusted analysis compared
to the uncorrected construction analysis is visible. Even if the desired condition is not exactly
reached, the deviations are within tolerable limits.

7.2 Conclusions

The present work shows that the determination of an optimum tensioning strategy can be carried
out efficiently by the proposed method. In the example calculations, the construction process
and time-dependent effects are sufficiently taken into account in the computation of the initial
cable forces. In order to keep the calculations linear and to avoid iterative solutions, the consid-
eration of nonlinear effects in the ”expanded unit load” method is explained, but not exemplary
demonstrated. From the performance of a nonlinear construction stage analysis using the cable
forces determined in a linear analysis, it is shown that the approach to neglect the nonlinear be-
havior is sufficient as long as the considered cable-stayed bridge is small. However, as the main
span length increases, the geometric nonlinear effects can no longer be ignored. This obviates
the need for performing an iterative solution to determine appropriate initial cable forces.
Concerning time-dependent effects, it is found that creep and shrinkage have a considerable
influence on the structural behavior of cable-stayed bridges during the construction process.
Clearly higher cable forces are needed to control the geometry and internal forces during erec-
tion when these effects are taken into account. Thus, for concrete or composite cable-stayed
bridges, it is generally recommended to include time effects on materials in the determination
of the optimum tensioning strategy, even if a small cable-stayed bridge is investigated.

Furthermore, it is shown in this thesis that in cable-stayed bridges continuous geometrical mon-
itoring is absolutely necessary in order to obtain acceptable geometry and tension conditions
for the structure. Error factors in the structural parameters are inevitable and continuous mon-
itoring enables the contractor to adapt to any changes and to avoid founding any decision on a
single measurement; a suitable decision for the adjustment on site can be made.
In the example calculation, the adjustment of cable forces and deck elevations during the erec-
tion process is demonstrated. The achieved final condition is acceptable, but there are still
deviations to the desired state. Construction adjustments in the final construction stage are gen-
erally easier and yield better results. However, as the residual error in the example calculation is
sufficiently small, from the standpoint of economic efficiency the presented method is proposed.
Moreover, using this method the accumulation of the discrepancies of the structural response
during erection is avoided.
CHAPTER 7. SUMMARY AND CONCLUSIONS 179

The analysis programs Larsa2000 and MIDAS/Civil are proved to be powerful tools for the
structural analysis of cable-stayed bridges. Both offer many features for the analysis and design
of these structures, whereby MIDAS/Civil offers even more than Larsa2000. However, these
additional features in MIDAS/Civil do not necessarily work in the program. For example, MI-
DAS/Civil offers the function to include time-dependent and nonlinear effects in a construction
stage analysis, but does not allow to consider them both simultaneously. When including non-
linear behavior in an analysis, MIDAS considers large displacement effects and the nonlinear
behavior of the cables, but P-Delta has not been added to the program yet. Concerning the
unknown load factor function, MIDAS offers a tool to automatically solve for unknown load
factors. This function works properly in a linear static analysis and is to some extend incorrect
in a linear construction stage analysis; in time-dependent or nonlinear analyses it does not work
at all. MIDAS/Civil claims all these features as special tools for the analysis of cable-stayed
bridges, but does generally not inform about their limitations. The program should therefore
always be handled with special care.
Larsa2000 is simpler and offers less features, but they generally work in the program. The only
limitation is that cable sag effects are not considered when performing a nonlinear analysis.
Apart from this, all features used during this work seem to operate properly.

Nevertheless, by the dint of both Larsa2000 and MIDAS/Civil satisfactorily results could be
achieved during the development of this thesis. Using these programs in the example calcula-
tion of a simple cable-stayed bridge, it was possible to exemplary demonstrate the procedure of
computing an optimum tensioning strategy and to adjust discrepancies during erection of cable-
stayed bridges. Unfortunately, Larsa2000 and MIDAS/Civil could not automatically manage
these calculations. It was necessary to manually determine the data for the influence matrices
and to solve the respective equations, which is a time-consuming process. For bigger struc-
tures including a multiplicity of cables, the calculation effort gets enormous and the presented
procedures become uneconomically. Especially when nonlinear effects shall be included and
iterative solutions are required, it would safe a considerable amount of work to fully implement
the presented optimizaion procedures in the structural analysis programs, as it has already been
done in RM2004. Alternatively, a special computer program that handles these problems could
be developed. However, the implementation into software programs has not been the topic of
this thesis, but would be the next step to apply the presented analysis and construction control
methods to real projects.
Bibliography

[1] Adeline, R., Monitoring and adjustment of the geometry and cable tension of cable-stayed
bridges built using the balanced cantilever method - The case of Pont de Normandie,
International Conference A.I.P.C.-F.I.P.- Cable-stayed and suspension bridges, Vol. 2,
Deauville, 1994, pp. 379-386

[2] Bažant, Z.P. and Wittmann, F.H., Creep and shrinkage in concrete structures, John Wiley
& Sons, New York, 1982

[3] Bishara, A.G. and Papakonstantinou, N.G., Analysis of cast-in-place concrete segmental
cantilever bridges, Journal of Structural Engineering, ASCE, Vol. 116, No. 5 (1990), pp.
1247-1268

[4] Bruer, A., Pircher, H. and Bokan, H., Computer based optimizing of the tensioning of
cable-stayed bridges, IABSE Conference on cable-stayed bridges - Past, present and fu-
ture, Malmö, Sweden, 1999

[5] Chauvin, A.H. and Alvares Solis, H., The El Canon & El Zapote bridges, Mexico - Control
of geometry during construction, International Conference A.I.P.C.-F.I.P. - Cable-stayed
and suspension bridges, Vol. 2, Deauville, 1994, pp. 405-412

[6] Chen, D., A new method to assign initial cable forces for prestressed cable-stayed bridges,
IABSE Conference on cable-stayed bridges - Past, present and future, Malm ö, Sweden,
1999

[7] Chen, D.W., Au, F.T.K., Tham, L.G. and Lee, P.K.K., Determination of initial cable forces
in prestressed concrete cable-stayed bridges for given design deck profiles using the force
equilibrium method, Computers and Structures, Vol. 74 (2000), pp. 1-9

[8] Cluley, N.C. and Shepherd, R., Analysis of concrete cable-stayed bridges for creep, shrink-
age and relaxation effects, Computers and Structures, Vol. 58, No.2 (1996), pp. 337-350

[9] Cruz, J. and Almeida, J.F., A new model for cable-stayed bridge control and adjustment,
IABSE Conference on cable-stayed bridges - Past, present and future, Malm ö, Sweden,
1999
BIBLIOGRAPHY 181

[10] Cruz, P.J.S., Mari, A.R. and Roca, P., Nonlinear time-dependent analysis of segmentally
constructed structures, Journal of Structural Engineering, ASCE, Vol. 124, No. 3 (1998),
pp. 278-287

[11] Ernst, J.H., Der E-Modul von Seilen unter Berücksichtigung des Durchhanges, Der Bauin-
genieur, Vol. 40, No. 2 (1965), pp. 52-55

[12] Fanourakis, G.C. and Ballim, Y., Predicting creep deformation of concrete - Comparison
of results from different investigations, Proceedings, 11th FIG Symposium on Deforma-
tion Measurements, Santorini, Greece, 2003

[13] Fujisawa, N. and Nakamura, N., Computer sytem for cable adjustment of cable-stayed
bridges during erection, International Conference A.I.P.C.-F.I.P.- Cable-stayed and sus-
pension bridges, Vol. 2, Deauville 1994, pp. 387-394

[14] Ghali, A., Favre, R. and Elbardy, M., Concrete structures - Stresses and deformation, 3rd
Ed., E & FN Spon, New York, 2002

[15] Gimsing, N.J., Cable-supported bridges - Concept & Design, 2nd Ed., John Wiley & Sons,
West Sussex, United Kingdom., 1997

[16] Gimsing, N.J., History of cable-stayed bridges, IABSE Conference on cable-stayed


bridges - Past, present and future, Malmö, Sweden, 1999

[17] Grabow, M.J., Construction stage analysis of cable-stayed bridges, Diplomarbeit, Arbeits-
bereich Baustatik und Stahlbau, Technische Universität Hamburg-Harburg, 2004

[18] Han, D. and Yan, Q., Construction control practice for Panyu cable-stayed bridge, IABSE
Conference on cable-stayed bridges - Past, present and future, Malm ö, Sweden, 1999

[19] Han, D. and Su, C., Construction control of the Yamen cable-stayed bridge, Department
of Civil Engineering, South China University of Technology, Guangzhou

[20] Hansvold C., Faller P., Nilsson H. and Svahn P.-O., Erection of the Uddevalla Bridge,
IABSE Conference on cable-stayed bridges - Past, present and future, Malm ö, Sweden,
1999

[21] Janjic, D., Pircher, M. and Pircher H., Optimization of cable tensioning in cable-stayed
bridges, Journal of Bridge Engineering, ASCE, Vol. 8, No. 3 (2003), pp. 131-137

[22] Karoumi, R., Some modeling aspects in the nonlinear finite element analysis of cable
supported bridges, Computers and Structures, Vol. 71 (1999), pp. 397-412

[23] Kwak, H.-G. and Seo, Y.-J., Numerical analysis of time-dependent behavior of pre-cast
pre-stressed concrete girder bridges, Construction and building materials, Vol. 16 (2002),
pp. 49-63
BIBLIOGRAPHY 182

[24] Kwak, H.-G. and Son, J.-K., Determination of design moments in bridges constructed by
balanced cantilever method, Engineering Structures, Vol. 24 (2002), pp. 639-648

[25] Larsa, Inc., LARSA 2000: Reference for LARSA 2000, Finite element analysis and design
software, Melville, New York, USA, August 2004

[26] Larsa, Inc., LARSA 2000: User’s Guide for LARSA 2000, Finite element analysis and
design software, Melville, New York, USA, August 2004

[27] Larsa, Inc., LARSA 2000/4th Dimension: Staged construction analysis for LARSA 2000,
Finite element analysis and design software, Melville, New York, USA, August 2004

[28] Manabe, Y., Mukasa, N., Hirahara, N. and Yabuno, M., Accuracy control on the con-
struction of Tatara Bridge, IABSE Conference on cable-stayed bridges - Past, present and
future, Malmö, Sweden, 1999

[29] Marchetti, M. and Lencinq, B., Stay adjustment: From design perspective to on site prac-
tice, IABSE Conference on cable-stayed bridges - Past, present and future, Malm ö, Swe-
den, 1999

[30] MIDAS/Civil, Analysis for civil structures, Rev.Date: 2005-04-01

[31] MIDAS/Civil, Online Manual - Version 6.3.0, Rev.Date: 2005-04-01

[32] Nazmy A.S., Abdel-Ghaffar, A.M., Three-dimensional nonlinear static analysis of cable-
stayed bridges, Computers and Structures, Vol. 34 (1990), pp. 257-271.

[33] Negrao, J.H.O. and Simoes, L.M.C., Optimization of cable-stayed bridges with three-
dimensional modelling, Computers and Structures, Vol. 64 (1997), No. 1-4, pp. 741-758

[34] Paik, J., Song, S., Cho, S. and Kang, H., Geometry control of cable-stayed bridge in
SeoHae Bridge, IABSE Conference on cable-stayed bridges, Seoul, Korea, 2001

[35] Paulik, L. and Deschamps, D., Construction and geometry control of the Mezcala cable-
stayed bridge Mexico, International Conference A.I.P.C.-F.I.P. - Cable-stayed and suspen-
sion bridges, Vol. 2, Deauville, 1994, pp. 395-404

[36] Peterson, C., Stahlbau - Grundlagen der Berecnung und baulichen Ausf ühung von
Stahlbauten, 3rd Ed., Vieweg, Wiesbaden, 1993

[37] Pircher, H., Cable-stayed bridges, TDV-Seminar 2003, TDV-Technische Datenverar-


beitung G.m.b.H, Graz, Austria, 2003

[38] Pircher, H., Die computerunterstützte Berechnung für eine große Schrägseilbrücke (Udde-
valla Brücke) in Schweden, Festschrift zum 60. Geburtstag von Professor Greiner, Tech-
nical University Graz, Austria, 2001
BIBLIOGRAPHY 183

[39] Pircher, H., Finite differences to simulate creep and shrinkage in prestressed concrete and
composite structures, International Conference on Computation Modelling of Concrete
Structures, H. Mang, N. Bicanic and R. De Borst, eds, Innsbruck, 1994, pp. 579-588

[40] Podolny, W. and Scalzi, J.B., Construction and design of cable-stayed bridges, 2nd Ed.,
John Wiley & Sons, New York, 1986

[41] Quast, J., Zeitabhängige Spannungsumlagerungen bei seilabgespannten Massivbr ücken,


Werner-Verlag, Düsseldorf, 1980

[42] Roik, K., Albrecht, G. and Weyer, U., Schrägseilbrücken, Ernst & Sohn, Berlin, 1986

[43] Schlaich, M., Erection of cable-stayed bridges having composite decks with precast con-
crete slabs, Journal of Bridge Engineering, ASCE, Vol. 6, No. 5 (2001), pp. 333-339

[44] Schneider, K.-J., Bautabellen für Ingenieure, 14. Auflage, Werner Verlag, Düsseldorf,
2001

[45] Simoes, L.M.C. and Negrao, J.H.O., Optimization of cable-stayed bridges with box-girder
decks, Advances in Engineering Software, Vol. 31 (2000), pp. 417-423

[46] Starossek, U., Cable-stayed bridge concept for longer spans, Journal of Bridge Engineer-
ing, ASCE, Vol. 1, No. 3 (1996), pp. 99-103

[47] Troitsky, M.S., Cable-stayed bridges - Theory and design, 2nd Ed., BSB Professional
Books, Oxford, U.K., 1988

[48] Virlogeux, M., Erection of cable-stayed bridges - The control of the designed geome-
try, International Conference A.I.P.C.-F.I.P.- Cable-stayed and suspension bridges, Vol. 1,
Deauville, 1994, pp. 321-350

[49] Walther R., Schrägseilbrücken, Beton-Verlag, Düsseldorf, 1994

[50] Walther R., Houriet, B., Isler, W., Moia, P. and Klein, J.F., Cable-stayed bridges, Thomas
Telford, London, 1999

[51] Wang, P.H., Tseng, T.C. and Yang, C.G., Initial shape of cable-stayed bridges, Computers
and Structures, Vol. 46(6) (1993), pp. 1095-1106

[52] Wang, P.H., Lin, H.T. and Tang, T.Y., Study on nonlinear analysis of a highly redundant
cable-stayed bridge, Computers and Structures, Vol. 80(2) (2002), pp. 165-182

[53] Wang, P.H., Tang, T.Y. and Zheng, H.N., Analysis of cable-stayed bridges during con-
struction by cantilever methods, Computers and Structures, Vol. 82 (2004), pp. 329-346
BIBLIOGRAPHY 184

[54] Wang, S. and Fu, C.C., Static and stability analysis of long-span cable-stayed steel bridges,
University of Maryland, 2003

[55] Xanthakos, P.P., Theory and design of bridges, New York, 1994

[56] Xiao, R., Xiang, H., Jia, L. and Song, X., Influence matrix method of cable tension
optimization for long-span cable-stayed bridges, IABSE Conference on cable-supported
bridges, Seoul, Korea 2001

[57] Zienkiewics, O.C. and Taylor, R.L., The finite element method, 5th Ed., Volume 2,
Butterworth-Heinemann, 2000

Das könnte Ihnen auch gefallen