You are on page 1of 198

)

CHARACTERIZATION OF POLYESTER-ROPE

SUSPENDED FOOTBRIDGES

Edward Matthew Segal

A DISSERTATION

PRESENTED TO THE FACULTY

OF PRINCETON UNIVERSITY

IN CANDIDACY FOR THE DEGREE

OF DOCTOR OF PHILOSOPHY

RECOMMENDED FOR ACCEPTANCE

BY THE DEPARTMENT OF

CIVIL AND ENVIRONMENTAL ENGINEERING

Adviser: Sigrid Adriaenssens

June 2015
ProQuest Number: 3728851

All rights reserved

INFORMATION TO ALL USERS


The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.

ProQuest 3728851

Published by ProQuest LLC (2015). Copyright of the Dissertation is held by the Author.

All rights reserved.


This work is protected against unauthorized copying under Title 17, United States Code
Microform Edition © ProQuest LLC.

ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346
)

© Copyright by Edward Matthew Segal, 2015. All rights reserved.


)

Abstract
In rural parts of the world, lack of access to roads that are useable year-round

significantly contributes to poverty. Suspended footbridges can improve access at

locations that require medium span crossings (15 m to 64 m). This dissertation challenges

the idea that modern bridges of this type must use steel rope, a well-established material

for this application. Polyester rope, an unconventional bridge material, is investigated as

an alternative to steel rope for rural suspended footbridges. The specific goals of this

research are to: (i) characterize the static and dynamic behavior of polyester-rope bridges

and (ii) determine which design criteria and system parameters will influence future

design guidelines for these structures.

Numerical and analytical, static, natural frequency, and pedestrian excitation

computations are performed to investigate the influence of polyester rope’s material

stiffness on the static and dynamic response of polyester-rope suspended footbridges.

Polyester rope’s low stiffness leads to larger static bridge deflections than occur for steel-

rope structures. These deflections are accompanied by a nonlinear increase in a bridge’s

geometric stiffness and lead to high levels of safety against overloading. Polyester rope’s

low stiffness also requires that these bridges be prestressed to meet static and dynamic

serviceability (pedestrian comfort) limits specified in footbridge guidelines. The damping

ratios that are utilized in the pedestrian excitation analyses follow from the first set of

full-scale physical tests that have been performed on a medium-span polyester-rope

bridge.

iii
)

Multi-objective optimization is utilized to find minimum volume polyester-rope and

steel-rope suspended footbridge designs across the medium span range when subject to

in-plane static and dynamic strength and serviceability constraints. The optimization

problems are evaluated with a novel methodology that combines a genetic algorithm with

static, natural frequency, and pedestrian excitation analyses. The impact of cross-

sectional area, material stiffness, prestress, damping, mass, and stiffening stay elements

on rope volume requirements for these bridges are investigated. Minimum volume results

are presented graphically as functions of span to provide visual design aids that can be

included in future bridge guidelines to facilitate comparisons between different systems

under a range of constraint combinations.

iv
)

Acknowledgements
I am very thankful to my adviser, Professor Sigrid Adriaenssens, as well as Theodore P.

Zoli (HNTB, USA), Landolf Rhode-Barbarigos, Professor Rajan Filomeno Coelho

(Université libre de Bruxelles), Professor Maria Garlock and Professor Branko Glisic.

Professor Adriaenssens allowed me to pursue a project of my own interest and then

provided continual support and guidance. She challenged my ideas and helped me to shift

my focus from that of a designer to that of a researcher. Zoli had the inspired idea to

study polyester-rope suspended footbridges, brought me onboard with the project, lent his

technical advice, and funded materials and equipment. Landolf was a constant source of

optimism, energy, and support. Professor Filomeno Coelho fielded many questions

related to structural optimization. Professor Garlock and Professor Glisic were always

generous with their time and offered much appreciated advice and assistance.

A set of Princeton fellowships provided me with financial support: the Sherrerd

Fellowship, the Norman J. Sollenberger Fellowship, and the Upton Graduate Fellowship

in Engineering. Additionally, the Department of Civil and Environmental Engineering

generously funded my travel to conferences to present my research.

John Flory (Tension Technology International, USA), Maria Gibbs (University of

Notre Dame, USA), Amin Karbassi (Basler and Hofmann AG, Switzerland), Mitchell A.

Nahmias, Professor Paul Prucnal, Bhavin J. Shastri, Justin Steinhouse (HNTB, USA),

Joseph Vocaturo, and Ryan Woodward (HNTB, USA) offered technical advice on a

range of topics. Colin Barrett (Carollo Engineers, USA) and a group of Columbia

v
)

University Engineers Without Borders students and U.S. Peace Corps volunteers assisted

with setting up and performing the dynamic tests described in this dissertation.

My colleagues at Princeton, Islam Elnaggar, Minmin Fan, Kendall Schmidt, and Jane

Soohoo provided crucial technical and administrative assistance. My fellow graduate

students, Negar Elhami Khorasani, Jonathan Glassman, Julio Herrera Estrada, Daniel

Reynolds, and Dorotea Sigurdardottir were always available to talk things through.

Finally, I have my family, especially my loves, K and M, to thank. M’s constant cheer

never failed to lift my spirits. K’s endless and selfless support made this all possible.

vi
)

Table of Contents
Abstract iii

Acknowledgements v

1 Introduction 1 

1.1  Motivation ............................................................................................................ 1 

1.2  Research Goals and Objectives ............................................................................ 3 

1.3  Dissertation Layout .............................................................................................. 4 

1.4  Prior Publications and Presentations Contributing to this Dissertation ............... 6 

2 Literature Review 8 

2.1  Introduction .......................................................................................................... 8 

2.2  Suspended Bridge Systems .................................................................................. 8 

2.3  Synthetic Rope Material Selection..................................................................... 14 

2.4  Serviceability Design Criteria ............................................................................ 16 

2.5  Summary ............................................................................................................ 19 

3 Characterization of Geometric Nonlinear Static Behavior 20 

3.1  Introduction ........................................................................................................ 20 

3.2  Detailed Model: Numerical Static Analysis ...................................................... 21 

3.2.1 Case Study Bridge Description .............................................................. 21 

3.2.2 Numerical Model ................................................................................... 24 

3.2.2.1 Model Components and Boundary Conditions ....................... 24 

vii
)

3.2.2.2 Material Properties .................................................................. 25 

3.2.2.3 Load Definitions ..................................................................... 26 

3.2.3 Numerical Calculations .......................................................................... 28 

3.2.4 Numerical Results and Discussion......................................................... 28 

3.2.4.1 Lower Bound Material Stiffness Case, Results and

Discussion............................................................................... 28 

3.2.4.2 Comparison of Lower and Upper Bound Material Stiffness

Cases, Results and Discussion ................................................ 30 

3.3  Single Suspended Rope Model: Analytical Evaluation ..................................... 35 

3.3.1 Analytical Model ................................................................................... 35 

3.3.2 Analytical Calculations .......................................................................... 36 

3.3.3 Analytical Results and Discussion ......................................................... 36 

3.4  Conclusions ........................................................................................................ 37 

4 Characterization of Modal Parameters 39 

4.1  Introduction ........................................................................................................ 39 

4.2  Natural Frequency Analyses .............................................................................. 40 

4.2.1 Detailed Model: Numerical Natural Frequency Analysis ...................... 41 

4.2.1.1 Numerical Model Properties ................................................... 41 

4.2.1.2 Numerical Calculations ........................................................... 42 

4.2.1.3 Numerical Results and Discussion.......................................... 43 

4.2.2 Taut String Model: Analytical Natural Frequency Evaluation .............. 44 

4.2.2.1 Analytical Model .................................................................... 44 

4.2.2.2 Analytical Calculations ........................................................... 45 

viii
)

4.2.2.3 Analytical Results and Discussion .......................................... 45

4.3  Experimentally Found Natural Frequencies ...................................................... 47 

4.3.1 Accelerometer Details ............................................................................ 48 

4.3.2 Free Vibration Tests ............................................................................... 49 

4.3.3 Ambient Vibration Tests ........................................................................ 52 

4.3.4 Discussion of Natural Frequencies ........................................................ 53 

4.4  Damping Estimation .......................................................................................... 56 

4.4.1 Damping Estimation Procedure ............................................................. 57 

4.4.2 Damping Estimation Results and Discussion ........................................ 60 

4.5  Conclusions ........................................................................................................ 63 

5 Bridge Response to Pedestrian Excitation 65 

5.1  Introduction ........................................................................................................ 65 

5.2  Load Definitions ................................................................................................ 66 

5.3  Detailed Model: Numerical Dynamic Analysis ................................................. 68 

5.3.1 Numerical Model Properties .................................................................. 68 

5.3.2 Numerical Calculations .......................................................................... 69 

5.3.3 Numerical Results and Discussion......................................................... 72 

5.3.3.1 Lower Bound Material Stiffness Case, Results and

Discussion ............................................................................... 72 

5.3.3.2 Comparison of Lower and Upper Bound Material Stiffness

Cases, Results and Discussion ................................................ 76 

5.4  Taut String Model: Analytical Dynamic Evaluation ......................................... 77 

5.4.1 Analytical Model ................................................................................... 78 

ix
)

5.4.2 Analytical Calculations .......................................................................... 78 

5.4.3 Analytical Results and Discussion ......................................................... 78 

5.5  Conclusions ........................................................................................................ 80 

6 Multi-Objective Optimization with Stress, Slope, and Natural Frequency

Constraints 82 

6.1  Introduction ........................................................................................................ 82 

6.2  Structural Model Features .................................................................................. 84 

6.2.1 Model Configurations ............................................................................ 84 

6.2.2 Material Properties ................................................................................. 86 

6.2.3 Load Definitions .................................................................................... 87 

6.3  Optimization Problems ...................................................................................... 88 

6.3.1 Problem Formulation ............................................................................. 88 

6.3.2 Objective Functions ............................................................................... 89 

6.3.3 Static Constraints ................................................................................... 89 

6.3.4 Natural Frequency Constraint ................................................................ 90 

6.3.5 Design Variables .................................................................................... 90 

6.4  Optimization Methodology ................................................................................ 94 

6.4.1 Optimization Algorithm ......................................................................... 94 

6.4.2 Structural Analysis Algorithms.............................................................. 96 

6.4.3 Objective Function and Constraint Calculations ................................... 97 

6.4.3.1 Objective Function Calculations .............................................. 97 

6.4.3.2 Constraint Calculations for the Problem with Static

Constraints Only ...................................................................... 98 

x
)

6.4.3.3 Constraint Calculations for the Unstayed Configurations in

the Problem with Static and Natual Frequency Constraints .... 98 

6.4.3.4 Constraint Calculations for the Stayed Configurations in the

Problem with Static and Natural Frequency Constraints ......... 98 

6.4.4 Performance Measurement Algorithm ................................................... 99 

6.5  Results and Discussion .................................................................................... 100 

6.5.1 Results and Discussion for the Problem with Static Constraints

Only...................................................................................................... 101 

6.5.2 Results and Discussion for the Problem with Static and Natural

Freqeuncy Constraints ......................................................................... 103 

6.5.3 Comparison of Polyester-Rope and Steel-Rope Live Load Overload

Factors .................................................................................................. 106 

6.6  Conclusions ...................................................................................................... 107 

7 Multi-Objective Optimization with Stress, Slope, Natural Frequency, and

Acceleration Constraints 110 

7.1  Introduction ...................................................................................................... 110 

7.2  Structural Model Features ................................................................................ 111 

7.2.1 Model Configuration ............................................................................ 111 

7.2.2 Material and System Properties ........................................................... 112 

7.2.3 Load Definitions .................................................................................. 113 

7.3  Optimization Problems .................................................................................... 114 

7.3.1 Problem Formulations.......................................................................... 115 

7.3.2 Objective Functions ............................................................................. 116 

xi
)

7.3.3 Static Constraints ................................................................................. 117 

7.3.4 Dynamic Constraints ............................................................................ 117 

7.3.5 Design Variables .................................................................................. 119 

7.4  Optimization Methodology .............................................................................. 119 

7.4.1 Algorithm Summary ............................................................................ 120 

7.4.2 Objective Function Calculations .......................................................... 121 

7.4.3 Constraint Calculations ........................................................................ 121 

7.5  Results and Discussion .................................................................................... 122 

7.6  Conclusions ...................................................................................................... 127 

8 Conclusions and Future Work 129 

8.1  Research Contributions .................................................................................... 129 

8.1.1 Static and Dynamic Bridge Behavior .................................................. 129 

8.1.2 Damping Characteristics ...................................................................... 131 

8.1.3 Key Design Criteria and System Parameters ....................................... 132 

8.2  Areas for Future Research ............................................................................... 134 

8.2.1 Polyester-Rope Material Characterization for Suspended

Footbridges .......................................................................................... 134 

8.2.2 Extensions to the Presented Optimization Methodology ..................... 135 

8.2.2.1 Alternative Objective Function ............................................. 135 

8.2.2.2 Additional Constraints .......................................................... 136 

8.2.2.3 Additional Configurations .................................................... 137 

8.2.2.4 Robust Optimization ............................................................. 137 

8.2.3 Large-Scale Physical Testing ............................................................... 138 

xii
)

8.3  Summary .......................................................................................................... 140 

A Comprehensive Analytical Static Calculations for the Single Suspended

Rope Model Presented in Chapter 3 141 

A.1 Introduction ...................................................................................................... 141

A.2 Analytical Model.............................................................................................. 142 

A.3 Analytical Calculations .................................................................................... 143 

A.3.1 Calculations for Load Combination 2 ................................................. 143 

A.3.2 Calculations for Load Combination 1 ................................................. 144 

A.3.3 Calculations for Load Combination 3 ................................................. 145 

A.4 Summary of Analytical Results ....................................................................... 147 

B Accelerometer Alignment 148 

B.1 Introduction ...................................................................................................... 148 

B.2. Discussion of Small Angle Misalignments ...................................................... 148 

C Equations for the Dynamic Analysis Presented in Chapter 5 151 

C.1 Introduction ...................................................................................................... 151 

C.2 Dynamic Load Calculations ............................................................................. 151 

C.3 Steady-State Displacement and Acceleration Equations ................................. 153 

C.4 Rope Force Equations ...................................................................................... 155 

C.5 Taut String Acceleration Equation ................................................................... 156 

D Initial Pedestrian Excitation Tests Performed on the Ait Bayoud Bridge 158 

D.1 Introduction ...................................................................................................... 158 

xiii
)

D.2 Accelerometer Details ...................................................................................... 159 

D.3 Single Pedestrian Walking Tests...................................................................... 160 

D.3.1 Procedure for Single Pedestrian Walking Tests .................................. 160

D.3.2 Results and Discussion for Single Pedestrian Walking Tests ............. 161 

D.4 Group Walking Tests ....................................................................................... 164 

D.4.1 Procedure for Group Walking Tests ................................................... 164

D.4.2 Results and Discussion for Group Walking Tests .............................. 164

D.5 Conclusions ...................................................................................................... 167 

Figure Credits 168 

Works Cited 172 

xiv
)

Chapter 1

Introduction

1.1 Motivation
Footbridges provide essential connections in a range of environments and are especially

crucial for those living in rural parts of the world (Bang 2009). The World Bank (2007)

estimates that over 1 billion people living in these areas lack access to roads that are

useable year-round. This lack of access significantly contributes to poverty (Roberts et al.

2006). For a poor community, the only financially viable option for improving its

infrastructure may be the enhancement of non-motorized access within existing transport

networks using structures such as footbridges (Lebo and Schelling 2001). Building

suspended bridges can improve access at sites that require spans longer than those

achievable with simple beams (Bridges to Prosperity 2011) or trusses.

Steel-rope (i.e., steel-cable) suspended bridges are currently built worldwide by

humanitarian groups such as Helvetas (HMG, TBSSP/Helvetas 2003), Bridges to

Prosperity (2011), and Bridging the Gap Africa (Bloss et al. 2014). While there are

historic examples of alternative materials such as iron chains (Gerner 2007) and natural

fibers (Ochsendorf 2004) being utilized in bridges, there are few contemporary suspended

1
)

footbridges that have been built with rope material other than steel. This dissertation

investigates the use of polyester rope, a material that is novel to bridge applications, as an

alternative to steel rope for medium-span rural suspended footbridges. In this dissertation,

15 m to 64 m bridge spans are classified as medium spans.

Polyester rope offers a combination of low cost, low creep, low weight, and high

durability (McKenna et al. 2004) that makes it a better alternative than other synthetic

rope materials such as aramid, carbon fiber reinforced polymer (CFRP), high-modulus

polyethylene (HMPE), and nylon, for this bridge application. Low cost is essential if

those living in rural areas are to afford these bridges. A low creep rate ensures that the

ropes are not susceptible to excessive strain under sustained prestress. Low weight is

advantageous because transportation of materials and construction equipment to rural

sites is challenging. High material durability is crucial to ensure that bridges do not

deteriorate quickly and can remain in service.

One key property that distinguishes polyester rope from steel rope is material

stiffness. Polyester rope has low material stiffness compared to steel rope. Consequently,

polyester-rope suspended footbridges are inherently more flexible than steel-rope bridges.

This dissertation examines the potential for this more flexible polyester-rope system to

serve as an alternative to stiffer steel-rope systems by addressing the following research

questions:

i. How does polyester rope’s material stiffness influence the static and dynamic

behavior of suspended footbridges built with this material?

ii. What advantages does the high flexibility of the polyester-rope system provide

over less flexible systems?

2
)

iii. Which static and dynamic strength and serviceability limits will be key design

criteria for polyester-rope structures?

iv. What are the important system parameters for polyester-rope suspended

footbridges?

1.2 Research Goals and Objectives


The research goals associated with the research questions posed in Section 1.1 are to:

i. Characterize the in-plane vertical static and dynamic behavior of polyester-rope

suspended footbridges.

ii. Establish the key static and dynamic criteria as well as important system

parameters for polyester-rope suspended footbridges.

By meeting these goals this dissertation will contribute to the limited design guidelines

available for polyester-rope suspended footbridges.

To achieve the research goals, a combination of structural analysis, physical testing,

and multi-objective optimization techniques is utilized. The five specific objectives of

this research are to:

i. Characterize the effect of polyester rope’s material stiffness on the geometric

nonlinear static behavior of polyester-rope suspended footbridges.

ii. Investigate the influence of polyester rope’s material stiffness on the natural

frequencies of polyester-rope suspended footbridges.

iii. Determine the approximate damping characteristics of a polyester-rope suspended

footbridge.

3
)

iv. Examine the impact of mass, damping, and polyester-rope’s material stiffness on

the accelerations and rope tensions of polyester-rope suspended footbridges.

v. Identify the key parameters that lead to minimum volume designs for medium-

span polyester-rope suspended footbridges by applying a novel multi-objective

optimization methodology to a set of problems involving various bridge

configurations subject to static and dynamic strength and serviceability

constraints.

1.3 Dissertation Layout


The remainder of this dissertation is organized as follows. Chapter 2 establishes the

current state of knowledge of polyester-rope suspended footbridges and this dissertation’s

contribution in this area. This chapter includes a review of suspended bridge systems,

synthetic rope materials, and design criteria.

Chapter 3 addresses research objective i. This chapter investigates the effect of

polyester rope’s material stiffness on the geometric nonlinear static behavior of polyester-

rope suspended footbridges. This task is accomplished by performing two-dimensional

numerical analyses on a model of a 64 m case study bridge. Analytical calculations are

also performed on a simplified, single suspended rope model. These analytical

calculations are conducted to demonstrate the ability of the simplified model to serve as a

surrogate for the detailed numerical model.

Chapter 4 addresses research objectives ii and iii. This chapter examines the impact of

polyester rope’s material stiffness on natural frequencies for polyester-rope suspended

footbridges. This task is carried out with a combination of two-dimensional numerical

4
)

and analytical computations. Additionally, this chapter presents approximate damping

ratios determined from full-scale testing on the case study bridge introduced in Chapter 3.

Chapter 5 addresses research objective iv. This chapter investigates how mass,

damping, and polyester rope’s material stiffness influences the dynamic response, namely

the accelerations and rope tensions of polyester-rope suspended footbridges. This task is

completed using two-dimensional numerical and analytical calculations.

Chapters 6 and 7 address research objective v. These chapters identify parameters

that aid in minimizing rope volume across the medium span range. This task is

accomplished by evaluating a set of optimization problems. Chapter 6 presents a novel

methodology that combines a non-dominated sorting genetic algorithm with nonlinear

static and natural frequency computations that utilize dynamic relaxation and

eigenanalysis algorithms. The methodology is used to evaluate problems with static

stress, slope, and natural frequency constraints. Chapter 7 presents an extension to the

optimization methodology. The modified methodology includes modal calculations to

evaluate cases that are subject to acceleration and dynamic stress constraints in addition

to static stress, slope, and natural frequency constraints. A range of systems including

two-dimensional polyester-rope and steel-rope bridges with or without prestress and with

or without below-deck stays are optimized.

Chapter 8 summarizes this dissertation’s contributions and suggests potential areas

for future numerical and experimental research related to polyester-rope suspended

footbridges.

5
)

1.4 Prior Publications and Presentations Contributing

to this Dissertation
Portions of this dissertation have appeared previously in or are under review for inclusion

in a series of journal and conference papers. With the permission of the respective

publishers, these sections, with edits and additions to provide continuity, have been

included in the dissertation. The citations for the papers along with a reference to the

primary dissertation chapter(s) in which sections of the papers appear, are listed below.

Additional text from these papers may also appear in other locations within the

dissertation. This section serves as the only explicit reference in the dissertation to these

publications.

Refereed journal publications:

i. Segal, E. M., Rhode-Barbarigos, L., Adriaenssens, S., and Filomeno Coelho, R.

D. “Multiobjective Optimization of Polyester-Rope and Steel-Rope Suspended

Footbridges.” Engineering Structures. [accepted for publication]. (Sections of this

manuscript appear primarily in Chapters 2 and 6.)

Conference publications (underline indicates the presenter and * indicates the paper

was displayed as a poster, not presented):

ii. *Segal, E. M., Adriaenssens, S., Zoli, T. P., and Flory, J. F. (2013). “Polyester

Rope, an Alternative to Steel Cable for Pedestrian Suspended Bridges.” Structures

Congress 2013: Bridging Your Passion with Your Profession – Proceedings of the

2013 Structures Congress, American Society of Civil Engineers (ASCE), Reston,

VA, 2837-2847. (Sections of this publication appear primarily in Chapter 2.)

6
)

iii. Segal, E. M., Rhode-Barbarigos, L., Filomeno Coelho, R. D., and Adriaenssens,

S. (2014). "An Automated Robust Design Methodology for Suspended

Structures." Proceedings of the IASS-SLTE 2014 Symposiuim - Shells,

Membranes, and Spatial Structures: Footprints, International Association for

Shell and Spatial Structures (IASS), Brasília, Brazil. (Sections of this publication

appear primarily in Chapter 8.)

iv. Segal, E. M., Adriaenssens, S., and Rhode-Barbarigos, L. (2015). “Strategies for

Improving the Dynamic Response of Suspended Footbridges in the Developing

World.” IABSE Geneva Conference 2015: Structural Engineering – Providing

Solutions to Global Challenges, International Association for Bridge and

Structural Engineering, Geneva, Switzerland. [abstract accepted]. (Sections of this

manuscript appear primarily in Chapter 1.)

Conference presentation without publication:

v. Segal, E. M., Rhode-Barbarigos, L., Adriaenssens, S. and Zoli, T. (2015). “Design

and Development of Polyester-Rope Suspended Footbridges.” Structures

Congress 2015, Structural Engineering Institute (SEI) and ASCE, Portland, OR.

(Sections of the presentation are elaborated on in many of the chapters.)

The author of this dissertation was the first and primary author of the listed

publications and presentations. The co-authors for these works served primarily as

advisers and edited the manuscripts.

7
)

Chapter 2

Literature Review

2.1 Introduction
The objective of this chapter is to provide the context for this dissertation’s contributions

to future design guidelines for medium-span (15 m to 64 m) polyester-rope suspended

footbridges built in rural areas. First, this chapter describes what a suspended footbridge

is, what configurations these structures have taken, and what materials these structures

have been built with. Next, the reasons why this dissertation investigates polyester rope

as opposed to other synthetic rope materials are discussed. Then, the serviceability

criteria likely to influence the designs for polyester-rope bridges are presented.

2.2 Suspended Bridge Systems


This section describes various suspended bridge systems that have been built and shows

that there are few polyester-rope suspended footbridge precedents. The terms

“suspended” and “suspension” are often used interchangeably when describing bridge

systems. In this dissertation, a suspended bridge (Figure 2.1) is defined as having a deck

that sits directly on draped below-deck ropes and may have additional draped ropes that

8
)

also serve as handrails (hand ropes). In contrast, a suspension bridge has a deck that is

hung from one or more draped ropes that pass over towers. A suspended bridge’s hand

ropes also span between towers. These towers are shorter than those of suspension

bridges because of the low sag-to-span ratios of the hand ropes.

Figure 2.1. Polyester-rope suspended footbridge in Ait Bayoud, Morocco showing the
below-deck and hand draped ropes, deck, and towers.

Schematic elevations for three suspended bridge configurations that have been built

are presented in Figure 2.2.

a. Suspended ropes only. b. Suspended ropes with c. Suspended ropes with below-
underslung ropes forming a deck stays.
truss.
Figure 2.2. Elevations showing geometry and components for three suspended bridge
configurations.

The configurations in Figure 2.2 are characterized by: (a) suspended ropes only, (b)

suspended ropes with underslung ropes forming a truss, and (c) suspended ropes with

9
)

below-deck stays. For each configuration in Figure 2.2 only a single suspended rope at

the below-deck level is shown. Suspended hand ropes, decking, etc. have been excluded

for clarity.

For all three configurations the primary features are the suspended ropes. The

mechanics of suspended ropes have been well studied. Irvine (1981) provided a historic

discussion on the development of the static catenary equations. He also presented an

approximate theory for static calculations performed on flat sag ropes (sag-to-span ratios

less than 1:8) and described the linear theory of free vibration of ropes. Gimsing and

Georgakis (2012) presented static and dynamic equations for a single rope and described

how to analyze suspension bridges. Conventional suspension bridges are similar to

suspended bridges whose ropes are not prestressed. Figure 2.3a presents a steel-rope

suspended bridge of this kind. Suspended bridges whose ropes are prestressed may be

referred to as stress ribbons. These structures could include a prestressed concrete deck

whose composite stiffness is greater than the ropes on their own (Strasky 2005). Figure

2.3b presents a stress ribbon with steel ropes and a concrete deck.

10
)

a. Collpatomaico Bridge with nonprestressed b. David Kreitzer Lake Hodges stress ribbon
ropes. Photograph courtesy of Bridges to with concrete deck. Photograph courtesy of
Prosperity (Wikimedia Commons).  Susan Williams (Flikr).  

c. Suspended bridge with below-deck truss. d. Nepalese bridge with below-deck stays.
Photgraph courtesy of Lev Yakupov (Flikr). Photgraph courtesy of John Pavelka (Flikr).
Figure 2.3. Built examples of steel-rope suspended footbridges.

An alternative way to stiffen a suspended rope in a suspended or suspension bridge is

to provide a prestressed supplemental rope with opposite curvature that acts with the

suspended rope to form a rope truss (configuration b). The trusses can be biconcave or

biconvex where the upper and lower truss chords are connected by elements subject to

tension in the former and compression in the latter (Irvine 1981). Various biconcave truss

footbridge configurations have been proposed and evaluated statically (Huang et al.

2005a; Chen et al. 2014) and dynamically (Huang et al. 2005b; Huang et al. 2005c). As

shown in Figure 2.3c, these trusses may be inclined from vertical to increase a bridge’s

lateral stiffness in addition to its vertical stiffness.

11
)

While discrete stays are most commonly provided as the primary load carrying

members in a cable-stayed system (Billington and Nazmy 1990), these elements can also

be used instead of a rope truss to stiffen suspension and suspended (configuration c)

bridges. A prominent example of a hybrid cable-stayed/suspension system (Gimsing and

Georgakis 2012) in which both suspension ropes and above-deck stays in addition to a

stiffening truss are utilized is John A. Roebling’s Brooklyn Bridge (Buonopane 2006). In

another Roebling structure, the Niagara Railroad Bridge, the suspension ropes, above-

deck stays, and stiffening truss are supplemented with below-deck stays to stiffen against

wind effects (Buonopane 2012). Figure 2.3d presents an example of a suspended

footbridge with below-deck stays.

Historic examples of suspended footbridges date back to at least the 15th century and

include Tibetan and Bhutanese iron chain (Gerner 2007) and Incan natural fiber-rope

(Ochsendorf 2004) bridges (Figures 2.4a and 2.4b, respectively). Modern suspended

bridges such as the ones shown in Figures 2.3 are typically constructed with steel ropes.

There are few contemporary examples of suspended bridge built with rope material other

than steel. Two temporary, laboratory-scale prototypes that do not use steel rope include

a 13 m span stress ribbon (Figure 2.5a) that has carbon fiber reinforced polymer (CFRP)

ropes and a concrete deck (Bleicher et al. 2011) and a 13.7 m polyester-rope truss (Figure

2.5b) (Sarner et al. 2010). One larger scale example is the 64 m span, 1.02 m wide

polyester-rope suspended footbridge (Ait Bayoud, Morocco, 2013) shown in Figure 2.1.

12
)

a. Iron chains from the Chakzam Bridge b. Keshwa-chaka natural fiber-rope bridge near
crossing the Yarlung Tsangpo River in Huinchiri, Peru. Photograph courtesy of Rutahsa
Tibet. Photograph courtesy of Edmund Adventures (Wikimedia Commons).
Candler (Wikimedia Commons).
Figure 2.4. Historic examples of suspended footbridges.

a. 13 m span CFRP and concrete stress ribbon. b. 13.7 m span polyester-rope truss.
Photograph courtesy of William Plunkett.
Figure 2.5. Laboratory-scale suspended footbridges.

While the 64 m span structure suggests that polyester rope has potential as an

alternative to steel rope for medium-span footbridges, its design was limited to meeting a

strength criterion only (Woodward 2012a). The smaller-scale polyester rope truss

demonstration bridge does not provide additional insight into the behavior of polyester-

rope bridges. Sarner et al. (2010) in their design report for this truss bridge described the

13
)

selection of the truss arrangement and the construction process, but did not include any

analysis. There has not been research on how meeting designated serviceability criteria

such as deflections, natural frequencies, and accelerations, influences the required

parameters for polyester-rope suspended footbridges. These parameters include prestress

and cross-sectional area for suspended ropes and stiffening elements such as truss or stay

ropes, as well as supplemental mass.

2.3 Synthetic Rope Material Selection


This section compares cost and material properties for various synthetic rope materials

such as aramid, carbon fiber reinforced polymer (CFRP), high-modulus polyethylene

(HMPE), nylon, and polyester. This comparison is performed to show that polyester rope

has greater potential than the other synthetic ropes for use in bridge applications in rural

areas. Qualitative cost and material property descriptions for the synthetic ropes are listed

in Table 2.1. Qualitative data are provided in Table 2.1 because the variety of grades

available for each material makes reporting exact quantitative data difficult.

Table 2.1. Qualitative cost and material property descriptions for aramid, CFRP, HMPE,
nylon, and polyester ropes.
Strength-to- Creep Abrasion Performance when
Material Cost Stiffness
weight performance resistance wet vs. dry i
Aramid ii High High High Very good Poor Decreased
CFRP iii High High High -- iv -- iv -- iv
HMPE ii High High High Poor Fair Similar
Nylon ii Low Low Low Fair Excellent Decreased
Polyester ii Low Low Low Very good Very good Similar
i. Performance qualifiers refer to changes to some or all of the other properties listed
in the table.
ii. Qualifiers follow from McKenna et al. (2004) directly or through interpretation of
their test results and discussions.
iii. Qualifiers follow from Burgoyne (2001).
iv. Cost is so high for CFRP (Burgoyne 2001), that it was eliminated for this
application (suspended footbridges in rural areas) before compiling these properties.

14
)

For pedestrian bridges in rural areas, minimizing cost is critical (Lebo and Schelling

2001). While rope materials such as aramid (Burgoyne 1993; Burgoyne and Head 1993)

and CFRP (Ding et al. 2011; Meier 2012) have been implemented as stays and

prestressing tendons, the high costs of these materials precludes their use in rural bridge

applications. HMPE has not been used in bridge applications, but Gupte et al. (2010) in

their study of potential synthetic ropes for rural footbridges concluded that HMPE,

primarily because of its high strength, was the best alternative to steel rope. However,

this study did not consider creep performance; HMPE is susceptible to creep failure

(McKenna et al. 2004). Since the ropes of a suspended bridge are under constant tensile

stress due to dead loads and possibly prestress, creep can occur. The combination of high

cost and potential for creep failure indicates that HMPE is not an appropriate rope

material for this application.

Polyester and nylon are both low cost options. Table 2.1 indicates that polyester has

better creep performance (lower creep rates) and performance when wet (its properties do

not change significantly when the rope is wet) than nylon. Additionally, most polyester

grades have strength and stiffness values that while low compared to aramid or CFRP are

greater than those of most nylon grades (McKenna et al. 2004). Polyester rope’s

combination of low cost, low creep, and high durability (very good resistance to abrasion

and little degradation of material properties when wet) makes it the most suitable

synthetic material for rural suspended footbridges.

Polyester rope’s frequent use in other applications previously dominated by steel

rope, such as deep water mooring (Flory et al. 2007), suggests that polyester-rope bridges

may also be competitive with steel-rope bridges. Direct comparisons between steel-rope

15
)

and polyester-rope bridges are required to show the advantages that polyester rope offers

over steel rope. A comprehensive comparison of all material properties and the influence

of these properties on the performance of polyester-rope and steel-rope bridges is beyond

the scope of this dissertation. This dissertation examines material stiffness and its impact

on the static and dynamic behavior of suspended footbridges.

2.4 Serviceability Design Criteria


In Section 2.2 it was noted that polyester-rope suspended footbridges have not been

designed to meet serviceability criteria such as deflections, natural frequencies, and

accelerations. It is anticipated that serviceability criteria will influence design

requirements for these structures because suspended footbridges have low stiffness

compared to other bridge types like trusses and arches and polyester rope’s material

stiffness is low relative to that of steel rope. This section presents static and dynamic live

load serviceability criteria for pedestrian bridges. Since it is anticipated that polyester-

rope suspended footbridges will be constructed in rural areas, these criteria are discussed

in the context of rural bridges.

Walking slope rather than deflection limits are more directly linked to pedestrian

comfort on a footbridge. For the rural bridges constructed by Bridges to Prosperity

(2011), the maximum allowable sag under full loading for bridges up to 80 m in span

results in a maximum walking slope of 20%.

The American Association of State Highway and Transportation Officials

(AASHTO) guidelines (2009) include natural frequency limits that follow from the range

susceptible to pedestrian vibrations (less than 1.3 Hz and 3 Hz in the lateral and vertical

16
)

directions, respectively). The AASHTO guidelines (2009) require that either the natural

frequencies are shifted outside the range of vulnerability or a dynamic pedestrian

excitation analysis is performed. AASHTO (2009) cites the Service d’Études Techniques

des Routes et Autoroutes (Sétra) technical guide (2006) as providing a potential method

for the dynamic analysis. Per Sétra’s methodology, a bridge can be considered

serviceable from a pedestrian comfort perspective if accelerations found during the

dynamic analysis are below certain limits (Sétra 2006). Table 2.2 lists lateral and vertical

accelerations corresponding to ranges of pedestrian comfort presented by Sétra.

Acceleration limits for minimum comfort are 0.082 g and 0.255 g in the lateral and

vertical directions, respectively (Sétra 2006).

Table 2.2. Sétra’s (2006) lateral and vertical acceleration pedestrian comfort limits.
Pedestrian comfort ranges defined by acceleration (g)
Acceleration direction
Maximum Mean Minimum Not acceptable
Lateral 0 – 0.015 0.015 – 0.031 0.031 – 0.082 > 0.082
Vertical 0 – 0.051 0.051 – 0.102 0.102 – 0.255 > 0.255

Information about a structure’s damping characteristics is required when performing a

dynamic analysis to evaluate accelerations. A footbridge’s damping is dependent on

multiple factors including the materials, structural system, bearing conditions (fib 2005),

and connections (Sétra 2006). A few of the aspects that have potential to contribute to the

damping of polyester-rope suspended footbridges have been studied previously.

McKenna et al. (2004) described how polyester loses heat due to hysteresis during load

cycling. Hennessey et al. (2005) examined the snap load response of synthetic ropes,

including polyester ropes, to determine the potential for these ropes to dissipate energy

when going from slack to taut. Ramberg and Griffin (1977) showed that slack cables

subject to an excitation followed by free vibration have greater damping ratios than taut

17
)

cables. To estimate the total damping of a structure requires full-scale testing (Sétra

2006). Testing of polyester-rope suspended footbridges has not been done previously.

Test results presented in this dissertation are the first available damping values for

polyester-rope suspended footbridges.

Sétra (2006), in addition to providing acceleration comfort limits, presents footbridge

classes based on anticipated traffic levels on a structure. This classification recognizes

that for rural bridges with limited traffic, considering pedestrian vibrations and meeting

even minimum comfort limits may not be necessary for the communities that the bridge

serves. While polyester-rope suspended footbridges are likely to be built in rural rather

than urban areas, this dissertation still considers natural frequency and minimum

acceleration comfort criteria to establish how requirements for bridge parameters such as

rope prestress, cross-sectional area, and superimposed mass change with the inclusion of

these limits. These analyses are restricted to vertical in-plane behavior and two-

dimensional analytical and numerical models are used throughout the dissertation to

examine this behavior. A future study can use the optimization methodology presented in

this dissertation to consider three-dimensional structures and their associated torsional

and lateral (i.e., out-of-plane) behavior. For these cases, additional pedestrian-induced

lateral vibration issues such as crowd synchronization (“lock-in”) with the bridge

response (Fujino et al. 1993; Dallard et al. 2001; Roberts T. M. 2005; Fujino and

Siringoringo 2014) could be examined.

18
)

2.5 Summary
This chapter established that there are few polyester-rope suspended footbridge

precedents and that limited analysis of these structures has been completed.

Consequently, there are many areas of potential research. This dissertation focuses on

characterizing the in-plane vertical static and dynamic behavior of these polyester-rope

structures and identifying the key design criteria and system parameters that will inform

future design guidelines for medium-span bridges.

19
)

Chapter 3

Characterization of Geometric

Nonlinear Static Behavior

3.1 Introduction
Chapter 2 showed that there has been limited analysis performed on polyester-rope

suspended footbridges. This chapter characterizes the static behavior while Chapters 4

and 5 examine the dynamic behavior of these bridges.

Suspended bridges are more flexible than other bridge types like trusses and arches.

The flexibility of the suspended bridge system combined with polyester rope’s low

material stiffness suggests that polyester-rope suspended footbridges will undergo large

deflections. A geometric nonlinear static analysis should be performed for structures that

undergo large deflections to account for the influence of geometry changes on the system.

The primary objective of this chapter is to characterize how polyester rope’s material

stiffness impacts the geometric nonlinear static behavior of polyester-rope suspended

footbridges. This task is carried out by performing a set of numerical computations on a

20
)

detailed two-dimensional model of a case study bridge. The numerical static calculations

introduced in this chapter will also be incorporated into the analyses appearing in

Chapters 4 – 7.

A secondary objective is to identify the single suspended rope as a surrogate model

that approximates the maximum rope tensions and deflections of the numerical model.

Simplified configurations such as this single rope, rather than their more computationally

expensive detailed counterparts are optimized in Chapters 6 and 7.

3.2 Detailed Model: Numerical Static Analysis


This section presents the numerical static analysis performed on a detailed two-

dimensional model of a case study bridge to evaluate the influence of material stiffness

on the geometric nonlinear behavior of the structure. First, the case study bridge is

described. Then, the numerical model, calculations and results are presented.

3.2.1 Case Study Bridge Description

This section describes the case study bridge that is investigated in this chapter as well as

in Chapters 4 and 5. The case study structure is a 64 m span, 1.02 m wide polyester-rope

suspended footbridge built in Ait Bayoud, Morocco in 2013. This structure spans

between two reinforced concrete towers that differ in elevation by 2.51 m (Figures 3.1

and 3.2).

21
)

Figure 3.1. Realized 64 m span Ait Bayoud footbridge.

Figure 3.2. South elevation of the Ait Bayoud bridge showing the hand and below-deck
suspended ropes, backstays, suspenders (typical layout), towers, and grade beams. The
nonstructural mesh fencing is omitted for clarity. Drawing courtesy of Ryan Woodward.

The suspended rope system consists of ten 18 mm (nominal) diameter (Ø) polyester

ropes, each prestressed to 24.5 kN and with midspan sags of approximately 1.15 m. Four

of these ropes are located below the deck and the remaining six ropes are bundled in two

sets of three to form hand ropes, located 1.22 m above the lower ropes and 0.40 m outside

the deck (Figure 3.3). All suspended ropes pass over the tower in saddles to form

backstays that are anchored to steel reinforcing bars, which extend through reinforced

concrete grade beams into solid rock. The anchorage points are at the same elevations as

the below-deck tower saddles. The resulting back spans are 6.1 m and 3.05 m at the low

(west) and high (east) ends of the bridge, respectively.

The deck consists of 6 cm x 20 cm timber planks attached to 10 cm x 10 cm timber

cross beams, typically spaced at 1.22 m. These cross beams connect directly to the four

22
)

below-deck suspended ropes. Additionally, inclined 9 mm (nominal) diameter (Ø)

polyester-rope suspenders attach to the ends of each cross beam and transfer a portion of

the load to the bundled hand ropes above the deck. Each suspender forms a “V-shape” in

elevation; the suspender begins at the hand ropes, wraps around a cross beam, and then

ends at the hand ropes. Also, attached to the suspended ropes is a non-structural mesh

fencing.

Figure 3.3. Typical Ait Bayoud bridge section showing the bundled hand ropes, below-
deck ropes, suspenders, deck planks, and a cross beam. Drawing courtesy of Ryan
Woodward.

The rope prestress values and all dimensions presented for the case study bridge,

except for the midspan sag, follow from the bridge designer’s structural drawings

(Woodward 2012b). The midspan sag value shown in Figure 3.2 (1.15 m) is determined

during numerical analysis of the bridge under the combination of prestress and dead load

while assuming the rope stiffness is equivalent to the lower bound “modulus of elasticity”

value defined in Section 3.2.2.2.

23
)

3.2.2 Numerical Model


This section describes the two-dimensional representation of the realized bridge that is

analyzed numerically. Model components, boundary conditions, material properties, and

design loads are presented.

3.2.2.1 Model Components and Boundary Conditions

To transform from the three-dimensional realized structure to the two-dimensional model

that is analyzed numerically, the ropes (e.g., hand, below-deck, suspender, and backstay

ropes) at the same elevations are grouped together into a single element (Figure 3.4). The

rope cross-sectional areas (A) used in the detailed model are summarized in Table 3.1.

Since the deck is not designed to influence the global behavior of the system, it is not

included in the detailed model, aside from its contribution to the system dead load.

Figure 3.4. Elevation of the detailed model of the case study.

Table 3.1. Rope areas used in the detailed model.


Rope A (cm2)
Hand rope and adjacent backstays 12.43
Below-deck rope and adjacent backstays 8.28
Suspenders 0.93

The model includes rope clusters which are groups of adjacent rope elements that

form sliding (i.e., continuous) ropes (Moored and Bart-Smith 2009). All elements in a

24
)

cluster have the same internal force. In the model, the suspenders consist of two element

clusters forming a “V-shape” in elevation. Both suspended ropes are segmented into

elements. The lengths for these elements are typically 0.61 m which is equivalent to half

of the typical cross beam spacing in the realized design. Each of the four backstays is a

single element clustered with its adjacent suspended rope element.

Pin supports are used at the backstay anchorages. Towers are not modeled directly.

Instead, pin supports are used at the locations of the hand rope and below-deck saddles.

In the model, the ropes are free to slide over these supports. Replacing the towers with

pin supports overestimates the system’s stiffness, which leads to lower deflections and

consequently, higher rope tensions than what will be present in the realized design.

Therefore, this modeling approach provides conservative force results.

3.2.2.2 Material Properties

The ropes are commercially available, 12-strand, single braid polyester-rope products

(Figure 3.5).

Figure 3.5. 12-strand single braid polyester rope.

Polyester is a viscoelastic material and the rope exhibits nonlinear stress-strain

behavior (Flory et al. 2004). While considering a single complete stress-strain curve in

the structural analyses would account for the varying stiffness, this approach has its

limitations. Due to polyester’s viscoelastic properties, the rope’s stiffness is dependent on

load duration, magnitude, and history (ABS 2011). Therefore, a single stress-strain curve

25
)

does not capture the rope’s behavior. Rather than utilizing a more complex stiffness

model that considers all these factors, the approach in this dissertation is to use lower

bound and upper bound “modulus of elasticity” values and then investigate the influence

of these values on the bridge response (i.e., rope forces, deflections, etc.). Material

stiffness cases c1 and c2 evaluated in the calculations performed in Chapter 3 - 5 use the

lower bound and upper bound values of 2.86 GPa (El) and 5.96 GPa (Eu), respectively

across the entire stress range. These modulus of elasticity values are found by drawing

lower bound and upper bound secant lines on a stress-strain curve generated from the

manufacturer’s data. The polyester-rope stiffness model that considers two bounding

stiffness values is known as the upper-lower bound model (ABS 2011).

Polyester rope’s strength limits are taken as the manufacturer’s listed minimum break

forces (noted as splice strengths). The 9 mm suspenders and the 18 mm suspended ropes

have strength limits of 24.5 kN and 99.2 kN, respectively.

3.2.2.3 Load Definitions

The analyses consider prestress, dead (DL), and service live (LL) loads. In section 3.2.1

it was noted that each of the ten suspended ropes is prestressed to 24.5 kN. The

equivalent strains (εp) for cases c1 and c2 are 0.0413 and 0.0198, respectively. In the

numerical model, holding the prestress forces in the suspended ropes constant, but

varying the prestress strain is done because the prestress forces and not the strains were

measured directly during bridge construction. While the suspenders were not prestressed

during bridge construction, in this analysis a small amount of prestress is applied to avoid

numerical errors. The strain in the suspender is set to 0.001 (which is 2.4% and 5.1% of

26
)

the suspended rope strains in cases c1 and c2). Table 3.2 summarizes the rope prestress

strains used in the numerical model.

Table 3.2. Prestress strains used in the detailed model.


εp
Rope
Case c1 Case c2
Hand rope and adjacent backstays 0.0413 0.0198
Below-deck rope and adjacent backstays 0.0413 0.0198
Suspenders 0.0010 0.0010

Both DL and LL are uniformly distributed gravity loads applied across the entire

span. The DL includes the contributions from the suspended ropes, suspenders, deck,

cross beams, and mesh fencing (Woodward 2012a). This value is 0.56 kN/m (qDL). The

service LL is the area load, 3.78 kN/m2 multiplied by the bridge width, 1.02 m. This

value is 3.84 kN/m (qLL). The American Association of State Highway and

Transportation Officials’ (AASHTO) reduction method based on the loaded deck area

(AASHTO 2002), provides the area load in the LL calculation (Woodward 2012a).

Concentrated loads translated from the uniformly distributed DL and LL are applied

vertically downward to below-deck suspended rope element nodes spaced at locations

where cross beams are positioned in the realized design. The typical cross beam spacing

is 1.22 m.

The following working stress load combinations (LCs) are used to investigate the

geometric nonlinear behavior of the bridge under service (unfactored) loads:

LC1: prestress

LC2: prestress + DL

LC3: prestress + DL + LL

27
)

3.2.3 Numerical Calculations


The numerical calculations performed on the detailed model utilize a dynamic relaxation

algorithm (Barnes 1999; Bel Hadj Ali et al. 2011). In this algorithm, suddenly applying

the loads sets the structure into motion. The structure’s geometry is tracked, until due to

fictitious damping, static equilibrium is reached (Day 1965).

3.2.4 Numerical Results and Discussion

This section first presents the tensions and deflections for case c1 (El = 2.86 GPa) under

LC3 to provide a representative set of results in detail. The LC3 results are presented

because by inspection, the forces and deflections under this load combination are greater

than the corresponding values under LC1 and LC2. Then, the minimum factor of safety,

maximum LL deflections, and maximum slope results from the two stiffness cases c1 (El

= 2.86 GPa) and c2 (Eu = 5.96 GPa) under all three load combinations are compared to

show how material stiffness influences the structure’s geometric nonlinear behavior.

3.2.4.1 Lower Bound Material Stiffness Case, Results and Discussion

Rope tensions and the deflected shape for case c1 (lower bound material stiffness case, El

= 2.86 GPa) under LC3 are shown in Figure 3.6.

Figure 3.6. Rope tensions and deflected shape for case c1 under LC3.

28
)

The minimum tensions in the hand and below-deck ropes are 220.3 kN and 145.1 kN,

respectively. Figure 3.6 shows that these minimum tensions occur near midspan where

the bridge slope is approximately zero. The maximum tensions (T) are 235.7 kN and

161.3 kN in the hand and below-deck ropes, respectively. Figure 3.6 shows that these

maximum tensions occur at the right side of the model where the maximum bridge slope,

42.8%, is located. As expected, based on the clustering, the backstays have the same

tension forces as their adjacent suspended rope elements. Suspender tensions range from

1.5 kN to 1.8 kN (Tmax). Figure 3.6 shows the minimum suspender tension occurs at the

right side of the model and the maximum suspender tension occurs near midspan.

Maximum tension results for the ropes are summarized in Table 3.3.

Factors of safety for the ropes are calculated using the equation:

Tult
FS = (3.1)
Tmax

where Tult is the lumped rope strength limit (i.e., combined strength limit for all ropes

grouped together). These strength limits and factors of safety are summarized in Table

3.3. Among all of the ropes, the hand rope and its adjacent backstay on the right side of

the model have the overall minimum factor of safety (FSmin), namely 2.46.

Table 3.3. Rope tension and factor of safety numerical results for case c1 under LC3.
Rope Tmax (kN) Tult (kN) FS
Hand rope and adjacent backstays 235.7 595.2 2.53
Below-deck rope and adjacent backstays 161.3 396.8 2.46
Suspenders 1.8 49.0 27.22

The maximum vertical LL deflections (δLL) in the hand and below-deck ropes shown

in Figure 3.5 occur near midspan and are both 5.05 m. The maximum walking slope

corresponds to the 42.8% bridge slope at the right side of the model. This slope is 2.14

29
)

times larger than the 20% rural walking slope criterion (Bridges to Prosperity 2011)

presented in Chapter 2. The large LL deflections and corresponding walking slopes for

the case study bridge may deter crowds from forming on the bridge making it unlikely

that the bridge will ever experience the full service LL.

3.2.4.2 Comparison of Lower and Upper Bound Material Stiffness Cases, Results

and Discussion

The results presented in Section 3.2.4.1 are now supplemented with a comprehensive

summary of the results for static numerical analyses performed for the two material

stiffness cases (c1: El = 2.86 GPa and c2: Eu = 5.96 GPa). Comparing the factor of safety

and deflection results for these two cases provides insight into how polyester rope’s

material stiffness is influencing the bridge’s static behavior.

Table 3.4 presents minimum factors of safety (FSmin) among all ropes (i.e., hand,

below-deck, backstay, and suspender ropes) as well as maximum LL deflections (δLL),

and maximum slopes (smax) in the below-deck rope for the three load combinations (LCs)

defined in Section 3.2.2.3. Load combinations that include additional LL beyond the

service LL of 3.84 kN/m are also presented in Table 3.4 (labeled LC4 – LC7) to show

how the structure behaves up until the strength limit is reached (i.e., FSmin = 1). The final

load combination in Table 3.4 for each material stiffness case corresponds to when the

bridge has reached its strength limit. These load combinations are LC7 for case c1 and

LC6 for case c2. The coefficient that appears before the LL term in each load

combination in Table 3.4 represents the LL overload factor (γ) applied. For example, for

LC4, an overload factor of 2 indicates that 200% of the service LL or 7.68 kN is included

in the analyses performed with that load combination.

30
)

Table 3.4. Minimum factors of safety, maximum live load deflections, and maximum
slopes from the numerical calculations for cases c1 and c2.
Case c1 Case c2
LC
FSmin δLL (m) smax (%) FSmin δLL (m) smax (%)
1: prestress 4.06 0 4.0 4.06 0 4.0
2: prestress + DL 3.96 0 11.0 3.89 0 10.8
3: prestress + DL + 1.00(LL) 2.46 5.05 42.8 2.16 4.09 36.5
4: prestress + DL + 2.00(LL) 1.75 7.62 59.7 1.50 6.05 49.3
5: prestress + DL + 3.00(LL) 1.37 9.48 72.2 1.17 7.44 58.4
6: prestress + DL + 3.84(LL) 1.23 10.77 81.0 1.00 8.39 64.8
7: prestress + DL + 4.75(LL) 1.00 11.99 89.3 - - -

Table 3.4 shows that under LC3 where the service LL (3.84 kN/m) is applied, cases

c1 and c2 have minimum factors of safety of 2.46 and 2.16, respectively. No factor of

safety for polyester-rope bridges exists, but these values are close to the typical value for

steel rope, 2.2 (Gimsing and Georgakis 2012). The factors of safety for cases c1 and c2

are 11.8% greater and 1.8% less than the value for steel rope, respectively.

The plot in Figure 3.7a illustrates how the factor of safety results listed in Table 3.4

vary with the applied LL. For both cases c1 and c2, there is a nonlinear decrease in factor

of safety that occurs as LL is increased (i.e., increasing the overload factor, γ). The

decrease in factor of safety is not linear with the load applied because of the change in

geometry that occurs during loading. As LL on the bridge increases, the suspended ropes

deflect and the geometric stiffness of the system increases (Irvine 1981). The plot in

Figure 3.7b shows for both cases the LL deflections that lead to this geometric nonlinear

stiffening.

31
)

4.5 12
4 c1 (El = 2.86 GPa)
c2 (E = 5.96 GPa) 10
3.5 u
3 8

(m)
min

2.5
6
FS

LL
2

δ
1.5 4
1 c1 (El = 2.86 GPa)
2
0.5 c2 (Eu = 5.96 GPa)
0 0
0 1 2 3 4 5 0 1 2 3 4 5
γ γ

a. Minimum factors of safety versus b. Maximum LL deflections versus


overload factors. overload factors.
Figure 3.7. Minimum factor of safety and maximum LL deflection results for cases c1
and c2.

The concept of geometric stiffening can be illustrated by examining a single

suspended rope with level pin supports subject to a uniformly distributed load, q (Figure

3.8).

a. Elevation.

b. Free body diagram.


Figure 3.8. Single suspended rope subject to a uniformly distributed load.

32
)

The equations for the horizontal (H) and vertical (V) components of the maximum

rope tension (Tmax) are:

qL2
H= (3.2)
8d

qL
V= (3.3)
2

where L is the rope span, d is the rope sag under the applied loads, and q is the uniformly

distributed load. The maximum rope tension (T) is found with the expression:

2 2 2 2
⎛ qL2 ⎞ ⎛ qL ⎞ ⎛ L2 ⎞ ⎛L⎞
T = H + V = ⎜⎜
2 2
⎟⎟ + ⎜ ⎟ =q ⎜⎜ ⎟⎟ + ⎜ ⎟ (3.4)
⎝ 8d ⎠ ⎝ 2 ⎠ ⎝ 8d ⎠ ⎝2⎠

Equation 3.4 indicates that the maximum rope tension is inversely related to the rope

sag (d). In a linear analysis, rope sag is held constant and there is a linear relationship

between the applied load (q) and the maximum rope tension (T). However, in a nonlinear

analysis, such as the one used in the numerical calculations, the rope sag and

consequently, the geometric stiffness of the system increases as the applied load increases

and the system deflects. This change in sag creates a nonlinear relationship between the

applied load and the maximum rope tension.

The amount of overload that a structure can support is an alternative measure of

safety. Table 3.4 indicates that at their strength limits (FSmin = 1), cases c1 and c2 have

overload factors of 4.75 (LC7) and 3.84 (LC6), respectively. These overload factors are

equivalent to LL of 18.24 kN/m and 14.74 kN/m, respectively. Therefore, case c1 can

carry 23.7% more LL than case c2 before the bridge reaches its strength limit. Case c1

has a higher overload factor because the case’s lower material stiffness (El = 2.86 GPa

versus Eu = 5.96 GPa) results in greater deflections and greater geometric nonlinear

33
)

stiffening. At their strength limits case c1’s maximum LL deflection (δLL = 11.99 m) is

43% larger than case c2’s maximum LL deflection (δLL = 8.39 m). The concept of lower

material stiffness leading to higher overload factors is examined further in Chapter 6

where a comparison of overload factors for polyester-rope and steel-rope bridges is

included. Steel rope has a modulus of elastic of 196.5 GPa, which is approximately 68.7

times greater than the lower bound polyester-rope stiffness value (El = 2.86 GPa).

The high overload factors developed by the polyester-rope bridge, suggests that it will

be difficult to load this structure to its strength limit. Additionally, at the strength limit

the resulting walking slopes are 89.3% and 64.8% for case c1 and c2, respectively. These

slopes are more than 3.2 times larger than the 20% rural walking slope criterion (Bridges

to Prosperity 2011) and further suggest that the bridge’s strength limit is not likely to be

reached. As was discussed in section 3.2.4.1, steep slopes are likely to deter additional

people from walking onto the structure until others have finished crossing. Consequently,

the deflections and associated walking slopes may limit the LL on the bridge.

To aid in visualizing the magnitudes of the rope deflections, Figure 3.9 presents the

deflected shapes for case c1 under LC2, LC3, and LC7 (load combination at the strength

limit).

Figure 3.9. Case c1 deflected shapes under LC2, LC3, and LC7.

34
)

3.3 Single Suspended Rope Model: Analytical

Evaluation
This section presents a nonlinear static analysis performed on a single suspended rope

representation of the case study bridge to demonstrate the ability of this model to rapidly

provide results that approximate the global behavior of the more computationally

expensive model investigated in Section 3.2.

Woodward (2012a) performed similar computations to the analytical calculations

presented in this dissertation in addition to utilizing elastic catenary equations (Irvine

1981) when designing the Ait Bayoud footbridge. In his design, Woodward did not

complete any numerical calculations. The comparison of the analytical calculations

performed on the simple model to the numerical calculations performed on the detailed

model is unique to this dissertation.

3.3.1 Analytical Model

This section describes the model evaluated analytically. The simplified model includes a

single suspended rope spanning 64 m (L) between pinned end supports (Figure 3.10).

Figure 3.10. Elevation of the single suspended rope model. The geometry is for case c1
under LC2.

Under LC2 (prestress + DL) the rope has a 1.15 m sag (d). All suspended ropes in the

three-dimensional realized bridge are grouped together in this single suspended rope. The

35
)

assigned area is 20.71 cm2 (ten 18 mm diameter ropes). Only one representative case,

namely the lower bound stiffness case c1 (El = 2.86 GPa), is evaluated. The three load

combinations (LC1 – LC3) defined in Section 3.2.2.3 are considered.

3.3.2 Analytical Calculations

The calculations utilize Irvine’s approximate theory flat cable theory (Irvine 1981) which

is intended for cables (ropes) with shallow profiles (sag-to-span ratios less than 1:8). In

this theory, the prestress is not directly applied; instead the desired sag under the applied

DL (qDL) is set, which indirectly accounts for the rope prestress. This theory includes

equations for calculating rope tensions and deflections under uniformly distributed loads.

The comprehensive calculations performed using these equations are provided in

Appendix A.

3.3.3 Analytical Results and Discussion

This section compares the case c1 (El = 2.86 GPa) analytical results for the single

suspended rope model to the numerical results for the detailed model. Minimum factor of

safety results for the two models under LC1 – LC3 are presented in Table 3.5. The

numerical results appearing in Table 3.5 were previously presented in Table 3.4. For all

three load combinations, the analytical and numerical factors of safety differ by less than

1%.

Table 3.5. Minimum factor of safety results from the analytical and numerical
calculations.
FSmin
LC
Numerical Analytical
1: prestress 4.06 4.07
2: prestress + DL 3.96 3.97
3: prestress + DL + LL 2.46 2.45

36
)

The maximum LL defection (δLL) of the single suspended rope evaluated analytically

(4.79 m) is 5.1% less than the deflection of the detailed model evaluated numerically

(5.05 m). Neglecting the back stays that are present in the detailed model creates a stiffer

system that consequently has lower deflections. The small differences in factor of safety

and deflection results indicate that the single suspended rope model can serve as a

surrogate for approximating the detailed model’s static behavior.

3.4 Conclusions
This chapter characterized the impact of polyester rope’s material stiffness on the

geometric nonlinear static behavior of polyester-rope suspended footbridges. Numerical

calculations were performed on a detailed model of a 64 m case study bridge using lower

bound (El = 2.86 GPa) and upper bound (Eu = 5.96 GPa) elastic moduli for the polyester

ropes. The calculations demonstrated that for the low and high material stiffness values,

the bridge deflected significantly under service loads (deflections of 5.05 m and 4.09 m,

respectively) and at the bridge’s strength limits (deflections of 11.9 m and 8.39 m,

respectively). These deflections corresponded to increased rope sags which led to the

systems having increased geometric stiffness. Due to the nonlinear relationship that arises

between the applied load and the rope tensions because of the change in sag, the bridge is

able to carry high levels of LL beyond the service LL. At the bridge’s strength limit the

LL overload factors (ratio of applied LL-to-service LL) were 4.75 and 3.84 for the low

and high stiffness material cases. Since deflections and consequently the geometric

nonlinear stiffening is inversely related to material stiffness, the upper bound modulus of

37
)

elasticity value for polyester-rope provides a conservative estimate for the structure’s

overload capacity.

For the case study, the walking slopes found under service loads and at the bridge’s

strength limit were significant. For example, the slopes for the low and high stiffness

material cases under service loads were 42.8% and 36.5% compared to the 20% criterion

used in rural areas (Bridges to Prosperity 2011). It will be difficult for the strength limit

to be reached because groups of pedestrians will not likely attempt to traverse the bridge

in this condition. Establishing that these structures have high LL overload factors and/or

can be designed to use deflections to control the amount of LL on the bridge indicates

that even without prescribed factors of safety for polyester rope in this application,

designers can make informed decisions about the safety of these structures.

This chapter also demonstrated that the rope system for the case study bridge can be

reduced to a single suspended rope model that can be evaluated analytically to

approximate the structure’s global static behavior. Neglecting the influence of the

suspenders, backstays, and minor elevation difference at the ends of the span (the bridge

has a 4% slope in elevation) in the simple model had little impact on the results. The

numerical and analytical minimum factor of safety and maximum LL deflection results

differed by less than 1% and 6%, respectively. Designers can utilize the presented

surrogate model with little computational effort during the conceptual design of a more

complex suspended bridge.

38
)

Chapter 4

Characterization of Modal Parameters

4.1 Introduction
Chapter 3 investigated the impact of polyester rope’s low material stiffness on the static

behavior of polyester-rope suspended footbridges. This low material stiffness in

combination with the flexibility of the suspended bridge system suggests that polyester-

rope suspended footbridges will have low natural frequencies. The American Association

of State Highway and Transportation Officials (AASHTO) pedestrian bridge guidelines

(2009) indicate that bridges with natural frequencies less than 3 Hz in the vertical

direction may be susceptible, from a serviceability standpoint, to pedestrian vibrations.

Bridges with vertical natural frequencies less than 3 Hz may still be acceptable for

pedestrian comfort (serviceability) if their accelerations are less than prescribed limits

(Sétra 2006). In order to compute a polyester-rope suspended footbridge’s accelerations,

as is done in Chapters 5 and 7, the structure’s damping characteristics must be

established. Currently, there are no documented damping values for polyester-rope

suspended footbridges

39
)

This chapter characterizes the modal parameters (i.e., natural frequencies and

damping ratios) of polyester-rope suspended footbridges. The first objective of this

chapter is to examine the influence of material stiffness on the vertical natural

frequencies for polyester-rope suspended footbridges. This task is completed by

performing numerical and analytical natural frequency calculations on the Ait Bayoud

case study bridge (introduced in Chapter 3). The numerical computations utilize an

eigenanalysis that will also be used in the analyses performed in Chapters 5 – 7. The

analytical evaluation is performed using taut string equations.

The second objective of this chapter is to present damping ratios for the case study

bridge. These damping ratios are determined from full-scale free vibration tests

performed on the structure. The bridge’s natural frequencies have to be established

experimentally prior to finding the structure’s damping ratios. The natural frequencies

found in the free vibration tests are compared to results from ambient vibration tests to

verify that the natural frequencies have been determined.

4.2 Natural Frequency Analyses


This section first describes the numerical analysis performed on the case study bridge to

show the impact of material stiffness on natural frequencies. Then, an analytical

evaluation of a taut string model representation of the case study bridge is presented to

show the potential for this simple model to serve as a surrogate for the detailed model

analyzed numerically.

40
)

4.2.1 Detailed Model: Numerical Natural Frequency Analysis

This section presents the calculations and results from the numerical analysis of the case

study bridge.

4.2.1.1 Numerical Model Properties

This section describes the model evaluated and the methodology used to perform the

numerical calculations. The detailed model analyzed in Chapter 3 is used in this analysis

and the model’s geometry and boundary conditions are presented again in Figure 4.1.The

two material stiffness cases (c1 and c2) introduced in Chapter 3 are considered in the

calculations. Cases c1 and c2 use polyester-rope’s lower-bound (El) and upper-bound

(Eu) modulus of elasticity values, namely 2.86 GPa and 5.96 GPa, respectively.

Figure 4.1. Elevation of the detailed model of the case study. The geometry is for case c1
subject to LC1.

Load combination LC1: prestress + dead load (DL) is considered in this analysis.

Concentrated DL masses are applied along the below-deck suspended rope at nodes

typically spaced at 1.22 m (i.e., the cross beam spacing in the realized design). These

masses are derived from the concentrated DLs used in the static analysis presented in

Chapter 3. Those concentrated loads are calculated from a uniformly distributed DL of

0.56 kN/m that represents the combined weight of the ropes, deck, cross beams, and mesh

fencing (Woodward 2012a).

41
)

4.2.1.2 Numerical Calculations

The calculations are performed by combining dynamic relaxation (Day 1965; Barnes

1999; Bel Hadj Ali et al. 2011) and eigenanalysis algorithms. First, dynamic relaxation is

utilized as a nonlinear geometric solver to find the structure’s static equilibrium under the

combination of prestress and DL. Next, the system’s lumped mass and stiffness matrices

are assembled. To limit the results to natural frequencies that are predominately in the

vertical direction, mass contributions in the in-plane horizontal (i.e., longitudinal

direction) degrees of freedom are set to a small value (0.0001 kN). Small, rather than

nonzero mass contributions are used to avoid numerical errors. In Chapter 5, the same

mass matrix will be used while Chapters 6 and 7 introduce variations to this mass matrix.

The stiffness matrix used in this chapter, as well as in Chapter 5 – 7, is generated for the

static equilibrium position and therefore, accounts for prestress and dead load impacts on

rope stresses and bridge geometry.

Then, the system’s stiffness and lumped mass matrices are used in the eigenanalysis

performed with MATLAB’s (2012) “eig” function to find the system’s eigenvalues

{λ1, λ2, . . .} and eigenvectors {φ1, φ2, . . .}. The stiffness matrix in these calculations is

assembled using a formulation that accounts for element clusters (Moored and Bart-

Smith 2009). The natural frequencies (fn) are then calculated from the eigenvalues:

λn
fn = , for n = 1, 2, . . . (4.1)

where n is the mode number.

42
)

4.2.1.3 Numerical Results and Discussion

This section presents the numerical results for the detailed model. For case c1 (El = 2.86

GPa), the natural frequencies ( f ) for the first three in-plane modes are 0.52 Hz, 1.03 Hz,

and 1.54 Hz, respectively. The second and third natural frequencies are approximate

harmonics of the fundamental (i.e., 1st) frequency. Figure 4.2 presents the mode shapes

that follow from the eigenvectors that accompany these natural frequencies. Minor

longitudinal displacements between the mode shapes and the undeformed shape indicate

that the eigenvectors have some in-plane longitudinal components in addition to vertical

components. While the modes are not purely vertical and are instead some combination

of longitudinal and vertical, it is apparent from the mode shapes in Figure 4.2 that these

in-plane modes are predominately vertical. Therefore, these in-plane modes are referred

to as “vertical modes” throughout this chapter.

Figure 4.2. Mode shapes for the first three vertical modes from the numerical
calculations.

The natural frequencies ( f ) for case c2’s (Eu = 5.96 GPa) first three vertical mode are

0.54 Hz, 1.04 Hz, and 1.56. These values are within 4% of the case c1 results. The

43
)

similarities in the results indicate that the difference in elastic moduli values for the

lower-bound (c1) and upper-bound (c2) material stiffness cases is not having a significant

impact on the natural frequencies. In both cases, the fundamental frequency as well as the

subsequent two frequencies fall below the AASHTO (2009) limit of 3 Hz in the vertical

direction. Table 4.1 summarizes the natural frequency results for cases c1 and c2.

Table 4.1. Numerical natural frequency results.


f (Hz)
Mode
Case c1 Case c2
1st 0.52 0.54
2nd 1.03 1.04
3rd 1.54 1.56

4.2.2 Taut String Model: Analytical Natural Frequency Evaluation

This section presents a set of analytical calculations performed on a model that represents

the case study bridge as a single, flat taut string. The results both establish the key

variables that are influencing the structure’s global behavior and indicate that this model

can be used as a surrogate for the detailed model analyzed numerically in Section 4.2.1.

4.2.2.1 Analytical Model

The two-dimensional model used in the numerical calculations is transformed into a taut

string model by: (1) lumping all the hand rope and below-deck ropes together, (2)

omitting the back spans, and (3) assuming the rope sag and incline are negligible. Figure

4.3 shows the geometry and boundary conditions for the taut string model.

Figure 4.3. Elevation of the taut string model.

44
)

4.2.2.2 Analytical Calculations

Natural frequencies ( fn) for a taut string are defined by (McConnell and Varoto 2008):

n T
fn = , for n = 1, 2, . . . (4.2)
2L m

where L is the rope span, T is the rope tension, m is the uniformly distributed mass, and n

is the mode number. Span and rope tension are the two variables in this expression that

influence the structure’s stiffness.

4.2.2.3 Analytical Results and Discussion

This section compares the analytical results for the simple model to the numerical

material stiffness case c1 (El = 2.86 GPa) results for the detailed model. Table 4.2

presents the analytical natural frequency results assuming the tension in the lumped ropes

is equivalent to the maximum rope tension found numerically for case c1 under (i) just

prestress or (ii) LC1: prestress + DL. The numerical results presented in Section 4.2.1.3

are also summarized in the table.

Table 4.2. Natural frequency results from the analytical and numerical calculations.
Calculation Applied loads T (kN) f1 (Hz) f2 (Hz) f3 (Hz)
Analytical Prestress 244.7 0.51 1.02 1.53
Analytical LC1: Prestress + DL 250.4 0.52 1.03 1.55
Numerical (case c1) LC1: Prestress + DL 250.4 0.52 1.03 1.54

Table 4.2 indicates that the analytical and numerical natural frequency results for LC1

are nearly identical. Additionally, the analytical results with and without the contribution

of DL to the rope tension are within 2% of each other. This occurs because the

contribution of prestress to the total rope tension dominates the contribution from DL

under LC1. The tension due to prestress makes up 97.7% of the rope tension.

Consequently, prestress is the primary feature that provides the system with its dynamic

45
)

stiffness. Rope area and material stiffness do not influence the results. These variables do

not appear in the taut string expression or influence the prestress (which is defined

directly in the analyses as a force to correspond with how prestress was measured during

construction). Using the taut string equation with the prestress force set as the rope

tension reveals that the prestress force has to increase by 34.6 times to meet AASHTO’s

(2009) 3 Hz limit. This significant increase in force will require that the rope’s cross-

sectional area increase to meet other criteria such as stress limits. Optimization studies in

Chapters 6 and 7 look at the impact of meeting natural frequency constraints on the

required rope volumes across the range corresponding to medium-span bridges (15 m to

64 m).

Investigating the more general case of a suspended rope that may or may not be taut

provides insight into the applicability of the taut string equation to bridges other than the

case study evaluated. Irvine (1981) presents a set of natural frequency equations for a

suspend rope that follow from the linear theory of free vibrations. These equations

simplify to those of a flat taut string when Irvine’s stiffness parameter (λ2) approaches

zero. This parameter is defined by the equation (Irvine 1981):

2
⎛ qDL L ⎞
⎜ ⎟ L
⎝ H ⎠
λ =
2
(4.3)
⎛ HLe ⎞
⎜ ⎟
⎝ EA ⎠

where qDL = is the uniformly distributed DL, L is the rope span, H is the horizontal rope

reaction under the uniformly distributed DL, E is the modulus of elasticity, A is the rope

46
)

cross-sectional area, and Le is the length parameter. The equation for the length

parameter is:

⎛ ⎛d ⎞ ⎞
2

Le ≅ L⎜1 + 8⎜ ⎟ ⎟ (4.4)
⎜ ⎝ L ⎠ ⎟⎠

where d is the rope sag.

In Appendix A, λ2 is found to be 0.49 for the case study bridge modeled as a single

suspended rope when the case c1 modulus of elasticity (2.86 GPa) is used. The

similarities in the analytical and numerical results presented above for this case study

indicate that for a stiffness parameter of 0.49 the taut string model is applicable. Equation

4.3 shows that the system stiffness parameter (λ2) is inversely proportional to the

horizontal rope reaction cubed (H3). Consequently, for more highly prestressed bridges

such as those attempting to meet AASHTO’s (2009) 3 Hz limit, this stiffness parameter

will decrease further and the taut string equation will still be applicable.

4.3 Experimentally Found Natural Frequencies


Before determining the damping ratios for the case study bridge, the structure’s natural

frequencies need to be established experimentally. This section describes the full-scale

free and ambient vibration tests that were performed to find these natural frequencies.

Free and ambient vibration tests have been frequently used to characterize the modal

parameters of pedestrian suspension bridges (Meng et al. 2007; Gentile and Gallino

2008) and stress ribbons (Caetano and Cunha 2004; Magalhães et al. 2007).

In this section, the accelerometers used to measure the bridge response during all of

the tests are described first. Then, the free vibration and ambient vibration test procedures

47
)

and results are presented. The free vibration results are also used when determining the

damping ratios in Section 4.4. The ambient vibration test results are presented to verify

that the bridge’s natural frequencies have been identified. This section concludes with a

discussion of the experimental and computed natural frequency results.

4.3.1 Accelerometer Details


Six tri-axial micro-electro mechanical system (MEMS) accelerometers (SENSR 2014)

recording with a sampling rate of 100 Hz were used in the physical tests. These

accelerometers collected data independently of each other; there was no time

synchronization between sensors. The accelerometers were fastened to the top of the deck

at positions along the suspended span to coincide approximately with the peak locations

in the Figure 4.2 mode shapes. Figure 4.4 shows in plan, the accelerometer locations on

the bridge.

Figure 4.4. Plan showing the accelerometer locations on the bridge. Note that the
drawing is not to scale.

Typically, the accelerometers were located along the longitudinal axis (centerline) of

the bridge. At midspan, one accelerometer was placed near each deck edge (Figure 4.5).

Placement of the accelerometers near the edges was done to aid in distinguishing between

vertical and torsional vibration modes.

48
)

Figure 4.5. Accelerometers fastened to the top of the bridge near the deck edges at
midspan.

While small angle misalignments between the accelerometer’s axes with the global

vertical, lateral, and longitudinal directions may have occurred, the impact on the results

should not significantly impact the results presented in this dissertation. Appendix B

provides additional discussion regarding the effect of misalignments on the results. To

account for any initial accelerometer deviation from zero, the recorded acceleration time

series are shifted based on average acceleration readings taken over a 30 second interval

that occurred before performing the tests.

4.3.2 Free Vibration Tests

This section describes the free vibration test procedure and results. Five tests (fv1 – fv5),

ranging in duration from 22 to 45 seconds were performed. In each free vibration test, a

vertical impulse was applied near the midspan of the bridge by having two people pull

and release a hanging rope attached to the deck. The bridge then vibrated freely until

coming to rest. The damping characteristics that dictate how quickly the bridge stops

vibrating are determined from the free vibration test results in Section 4.4. Figure 4.6

shows a similar setup to the one used during the tests.

49
)

a. Hanging rope attached to the bridge deck. b. Two people pulling on the rope prior to
releasing.
Figure 4.6. Setup similar to the one employed in the free vibration tests.

Figure 4.7 presents the acceleration time series for accelerometer a3b during test fv4

to demonstrate the typical trend in the results. The spike early in the results represents the

applied impulse. After this spike, the acceleration decays back to the ambient vibration

level. Due to the shape of the time series, it is apparent that more than one mode was

excited during the test.

0.3
a3b
0.2

0.1
Acceleration (g)

−0.1

−0.2

−0.3

−0.4
0 5 10 15 20
Time (s)

Figure 4.7. Acceleration time series for accelerometer a3b during free vibration test fv4.

The acceleration time series for every accelerometer in the five tests are transformed

from the time domain to the frequency domain to allow for the identification of the

50
)

structure’s natural frequencies. Specifically, the power spectral density (PSD) is

calculated for every accelerometer in all of the tests. The PSD indicates how the signal’s

power is distributed across the frequency range (Press et al. 1992). MATLAB’s (2012)

“periodogram” function is used to calculate the PSD for each data series. This function

performs a single discrete Fourier transform on the entire data series (Press et al. 1992).

The typical frequency resolution is 0.0244 Hz which is sufficient relative to the spacing

of the bridge’s frequencies.

The Peak Picking method (Zimmerman et al. 2008) is used to determine the natural

frequencies. On each of the plots, the peak PSD amplitudes among all accelerometers are

identified. Frequencies that correspond to the peak amplitudes represent the bridge’s

natural frequencies. Visual inspection of the results from all data series indicates that

there are peak amplitudes in the frequency ranges of [0.65, 0.72] Hz, [1.27, 1.34] Hz, and

[2.03, 2.27] Hz. The presence of multiple peaks confirms that more than one mode was

excited during the test. Figure 4.8 shows how the PSD varies with frequency for test fv4,

the test providing the overall (from all tests and accelerometers) maximum peak

amplitudes within these ranges.

−3
x 10
3
a1
2.5 1.32 Hz a2
0.68 Hz
a3a
2 a3b
PSD (g /Hz)

a4
2

1.5 a5

1
2.12 Hz
0.5

0
0 0.5 1 1.5 2 2.5 3
Frequency (Hz)

Figure 4.8. PSD versus frequency for the six accelerometers in free vibration test fv4.

51
)

The peak values in the Figure 4.8 plot occur at 0.68 Hz, 1.32 Hz, and 2.12 Hz. The

second and third frequencies are approximate harmonics of the fundamental frequency.

4.3.3 Ambient Vibration Tests

This section describes the ambient vibration test procedure and results. Two tests (av1

and av2) were performed. The test durations were 28 and 15 minutes in the first and

second tests, respectively. During these tests, pedestrians were not allowed on the bridge

and wind loading was the only excitation. These tests complement the free vibration tests

by providing acceleration results obtained under lower levels of excitation.

As was done for the free vibration tests, the ambient vibration acceleration time series

are transformed into the frequency domain to identify the natural frequencies.

Specifically, the average PSD is determined for every data series. The average PSD

determined for the ambient vibration tests is found using Welch’s method (Welch 1967).

With Welch’s method the acceleration time series is first divided into sections. Then, a

modified periodogram of each section is calculated. The periodograms are “modified”

because a window function is applied to the data. Finally, the periodograms are averaged

to provide the average PSD. MATLAB’s (2012) “pwelch” function with a Hamming

window (Press et al. 1992) is used to perform the calculations. Due to the difference in

test durations, the frequency resolution in the results following the application of this

method differs. Both values, 0.0015 Hz and 0.0031 Hz for tests av1 and av2, respectively,

provide sufficient resolution relative to the spacing of the bridge’s frequencies. Note that

Welch’s method is not used to segment the data and find the average PSD for the free

52
)

vibration tests because the technique leads to poor frequency resolution for short test

durations.

Figure 4.9 shows how the PSD varies with frequency for the six accelerometers in the

two tests. The Peak Picking method (Zimmerman et al. 2008) is again used to determine

the natural frequencies. While test av1 has one more peak than test av2, the three

remaining peaks occur at similar natural frequencies in the two tests. The frequencies

from the three peaks in test av2 (where the peak amplitudes are greater than those in test

av1) are 0.67 Hz, 0.83 Hz, and 1.33 Hz.

−5 −4
x 10 x 10
4 6
0.86 Hz a1 a1
a2 5 0.83 Hz a2
3 0.73 Hz a3a a3a
a3b 4 a3b
PSD (g2/Hz)

PSD (g2/Hz)

a4 a4
2 0.95 Hz a5 3 0.67 Hz a5

2 1.33 Hz
1
1.30 Hz 1

0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
Frequency (Hz) Frequency (Hz)
a. Ambient vibration test av1. b. Ambient vibration test av2.
Figure 4.9. PSD versus frequency for the six accelerometers in the ambient vibration
tests.

4.3.4 Discussion of Natural Frequencies

This section discusses the natural frequency results from the physical tests and also

compares these physical test results to the results from the numerical and analytical

computations. Two of the natural frequencies from the ambient vibration testing, 0.67 Hz

and 1.33 Hz (Section 4.3.3), are approximately equal to the 0.68 Hz and 1.32 Hz

frequencies, found during the free vibration tests (Section 4.3.2). The similarity in these

53
)

results suggests that at least two of the structure’s natural frequencies have been

determined. The additional natural frequency found during the free vibration tests,

namely 2.12 Hz, while not confirmed by the ambient vibration tests, also seems likely to

be a vertical natural frequency because it is a harmonic of the fundamental frequency. In

the numerical and analytical computations, natural frequencies for higher modes are

harmonics of the fundamental frequency (Sections 4.2.1.3 and 4.2.2.3).

The peak found at a frequency of 0.83 Hz during ambient vibration testing is more

difficult to interpret because it is not confirmed by the free vibration tests and does not

follow a pattern (such as being a harmonic) observed in the computations. Lack of

synchronization between the accelerometers precluded the use of more advanced

identification techniques such as stochastic subspace identification methods and

prevented the determination of the mode shapes (Gibbs et al. 2014) which could have

helped in the interpretation of the 0.83 Hz frequency.

In general, without information regarding the physical mode shapes it is not possible

to differentiate with certainty between torsional and vertical modes. Consequently, some

of the physically determined natural frequencies may be torsional and not vertical.

However, there’s strong evidence to suggest that at least one if not more of the physically

observed frequencies is in the vertical direction. First, the impulse applied during free

vibration was oriented primarily in the vertical direction near the centerline of the bridge.

The predominately vertical impulse should have produced excitation primarily in the

vertical direction which would lead to high levels of power in vertical natural

frequencies. Second, as shown in Figure 4.8, during free vibration testing, the four

accelerometers placed along the bridge’s centerline and the two accelerometers near the

54
)

edges of the bridge deck provided similar natural frequency results. Since little torsional

response is anticipated along the bridge’s centerline, these centerline accelerometers

should have identified vertical natural frequencies. Finally, as previously mentioned, the

second and third natural frequencies are harmonics of the fundamental frequencies which

corresponds with the trend observed in the computations.

It was anticipated that the natural frequencies determined experimentally for the

realized bridge would be lower than predicted computationally for two reasons. First, in

the numerical calculations (Section 4.2.1), representing the towers with pin supports

rather than modeling them directly overestimates the system stiffness. Second, in the

realized bridge, the tower saddle connections allow for some upward translation and it

seemed possible that under large vibration amplitudes the effective bridge span would

increase and lead to a system with less stiffness. In actuality, the natural frequencies of

the realized bridge are greater than predicted. For example, the fundamental natural

frequency, 0.68 Hz, found during free vibration (Section 4.3.3) is 30.8% greater than the

numerical prediction, 0.52 Hz (Section 4.2.1) for case c1 (El = 2.86 GPa).

To resolve the difference between computational and physical test results, a model’s

parameters can be updated until the modal characteristics match. In this case, the

comparison between the analytical and numerical results indicated that the bridge acts as

a taut string where span, rope tension, and mass are the only variables that are influencing

the results. As discussed in Chapter 3, the deck was not detailed to act as part of the

global structural system so neglecting it in the analyses should not be contributing to the

discrepancies. Since the span has been confirmed, uncertainties in the rope tension and

mass are causing the differences between the computations and the physical test results.

55
)

The numerical/analytical models can be updated to modify the mass and/or rope tension

to match the natural frequencies. For example, assuming that (i) the physically

determined fundamental frequency is in the vertical and not the torsional direction and

(ii) the mass is correct, then the tension force in the calculations needs to increase by

71.0%. If it is assumed that the tension force is correct, then the mass in the calculations

needs to decrease by 41.5%. It is likely that a combination of differences in the assumed

and actual rope tensions and bridge mass that led to discrepancies in natural frequency

values, but without additional results such as measured tensions or deflections the exact

combination of tension and mass that has actually been achieved in the realized bridge

cannot be determined.

Despite the discrepancies in the calculations and physical results and the uncertainty

in whether all of the natural frequencies determined from physical tests are vertical rather

than torsional, it is clear that frequencies are low compared to AASHTO’s (2009) 3 Hz

limit.

4.4 Damping Estimation


This section presents a set of approximate damping ratios determined from results of the

free vibration tests described in Section 4.3.2. Without damping ratios available in the

literature for this bridge type, these estimations need to be established prior to performing

a dynamic analysis. These ratios are found by filtering the acceleration time series to

isolate single mode response that can then be fit with an exponential decay curve

(Magalhães et al. 2010). As was the case with the physically determined natural

frequencies, the lack of time synchronization between the accelerometers does not allow

56
)

for the use of higher accuracy identification techniques such as covariance-driven

subspace stochastic identification (Magalhães et al. 2010). However, for the purposes of

this dissertation, a range of approximate values is sufficient; these values are utilized in

dynamic analyses performed in Chapters 5 and 7 where the influence of varying the

damping magnitude on bridge behavior is evaluated.

4.4.1 Damping Estimation Procedure

This section presents the method used to estimate the damping ratios from the free

vibration results. The first step is to isolate the contributions of each of the three

identified modes of vibration to the total response during free vibration. This is

accomplished by applying band-pass filters to the acceleration time series (Magalhães et

al. 2010). The filtering is completed with an infinite impulse response (IIR) Butterworth

filter (Hamming 1989) in MATLAB (2012).

The filter parameters are the lower (Fc1) and upper (Fc2) cutoff frequencies for the

pass band and the order (N). The cutoff frequencies vary in each data series and

correspond to frequencies with PSD amplitudes that are half of the peak PSD amplitude

(equivalent to a reduction of 3 dB). Linear interpolation between frequencies in the data

series is used to determine these cutoff frequencies. The range bound between the two

cutoff frequencies is referred to as a 3-dB bandwidth, a commonly used filter bandwidth

(McConnell and Varoto 2008). The exact values of the natural frequencies correspond to

the peak PSD amplitudes in the data series under consideration and may differ between

data series. Additionally, the natural frequencies may not be centered between the cutoff

frequencies.

57
)

An eighth-order filter is used. To avoid phase distortion the data is filtered in both the

forward and reverse directions (Smith 2007) using MATLAB’s (2012) “filtfilt” function.

Filtering in both directions doubles the filter order. Eight represents the equivalent order

after filtering in both directions. Eighth-order IIR filters have been utilized in similar

applications such as reducing the noise in Global Positioning System (GPS) data taken

during excitation of a steel truss footbridge (Moschas and Stiros 2011) and determining

the dynamic characteristics for a suspension footbridge (Meng et al. 2007). In both of

those studies, a Type-I Chebyshev filter, which has one more parameter than a

Butterworth filter, was used. The additional parameter is the pass band ripple (Ap).

Filtering in both directions amplifies the pass band ripple by 3 dB (Smith 2007). Since

the cutoff frequencies in this analysis are located where the PSD is 3 dB less than the

peak PSD, a pass band ripple of 3 dB is too large to be used in this analysis. Therefore, it

was preferable to use an IIR filter without a pass band ripple.

For each acceleration time series, once filtering is performed, an exponential decay

curve is fit to the peaks bound between 20% and 90% of the maximum acceleration

(Magalhães et al. 2010). The exponential fit is defined as:

a y = Ce −bt (4.5)

where ay is acceleration, t is time, and C and b are coefficients. The damping ratio (ξ) for

a mode associated with a natural frequency (fn) is calculated from the coefficient b with

the following expression:

b
ξ= (4.6)
2π fn

58
)

Only data for accelerometers located near the anticipated extremes in the respective

mode shapes (Figure 4.2) are included in the damping ratio calculations. For the third

mode, since the peak PSD amplitudes for accelerometers a1 and a5 are typically low

(Figure 4.8 presents a representative plot), calculations are not performed for these

accelerometers and only results for accelerometers a3a and a3b are found. All free

vibration tests are used to calculate the mode 1 damping ratios. For the second mode, free

vibration test fv1 is excluded because the PSD peaks near the second mode frequency are

not as well defined as in the other tests. For the third mode, free vibration test fv1 is

excluded because there are no PSD peaks near the third mode frequency. Figure 4.10

shows how the PSD varies with frequency for test fv1. Table 4.3 summarizes the

accelerometers, free vibration tests, and the resulting total number of data series used to

calculate the damping ratios in the three modes.

−3
x 10
3
a1
2.5 a2
a3a
2 a3b
PSD (g /Hz)

a4
2

1.5 a5

0.5

0
0 0.5 1 1.5 2 2.5 3
Frequency (Hz)

Figure 4.10. PSD versus frequency for the six accelerometers in free vibration test fv1.

Table 4.3. Data series used to calculate the damping ratios.


Mode Accelerometers Free vibration tests Number of data series
1st a3a, a3b fv1 – fv5 10
2nd a2, a4 fv2 – fv5 8
3rd a3a, a3b fv2 – fv5 8

59
)

4.4.2 Damping Estimation Results and Discussion

This section first presents the damping results for one of the data series in detail. Then,

the damping ratio ranges for the case study bridge’s three modes are compared to the

damping values for other existing bridges to show that the case study bridge has high

levels of damping. Finally, potential reasons for the case study bridge’s high levels of

damping and the observed relationships between the case study bridge’s amplitude of

oscillation and damping ratios are discussed.

The results for one representative data series analyzed are presented in Figure 4.11.

These results correspond to accelerometer a3b during free vibration test fv4. In these

plots, the first mode response is isolated.

−3
x 10
3 0.04
a3b, unfiltered
0.03 ay = 0.125e
−0.183t
2.5 a3b, filtered
0.68 Hz 0.02 ξ1 = 4.3%
Acceleration (g)

2
PSD (g2/Hz)

0.01
1.5 0
−0.01
1
−0.02
0.5 a3b, filtered
−0.03
exponential fit
0 −0.04
0 0.5 1 1.5 2 2.5 3 0 5 10 15 20
Frequency (Hz) Time (s)

a. PSD versus natural frequency for the b. Acceleration time series for the filtered data
unfiltered and filtered data (isolating first (isolating mode 1) with exponential fit.
mode).
Figure 4.11. Results for accelerometer a3b for free vibration test fv4.

Figure 4.11a shows the PSD versus natural frequency for both the unfiltered and

filtered data. The filtered results closely match the unfiltered results near the pass band

and are zero elsewhere. Figure 4.11b shows the acceleration time series for the filtered

60
)

data and the exponential decay curve fit to the peak accelerations. Compared to the

unfiltered acceleration time series presented in Figure 4.7, the filtered results show clear

single mode response.

For modes 1 – 3 the damping ratio ranges are [2.6, 5.2] %, [3.0, 4.0] %, and

[2.8, 6.1] % respectively. The lowest damping ratio found for the case study bridge is

greater than average damping ratios at service vibration levels for reinforced concrete (ξ

= 1.3%), steel (ξ = 0.4%), timber (ξ = 1.5%) and stress ribbon (ξ = 1.0%) footbridges

(Heinemeyer et al. 2009).

Due to the range of factors that can influence a footbridge’s damping characteristics,

these average damping values can deviate greatly from what is observed for a specific

bridge. Existing suspension and stress ribbon bridges provide the closest comparisons to

the case study bridge. Reported damping ratios for steel-rope suspension footbridges

include 0.5% for the Kochenhofsteg in Stuttgart, Germany, 0.3% for the Enzsteg II in

Pforzheim, Germany, and 0.6% to 0.8% for the Deutsche Museum Bridge in Munich,

Germany (fib 2005). Damping ratios for stress ribbon footbridges include 0.3% for Glacis

Bridge in Ingolstadt, Germany, 0.1% for Enzsteg III in Pforzheim, Germany (fib 2005),

and 1.7% to 2.6% for a bridge on the Campus of the Faculty of Engineering of the

University of Porto in Porto, Portugal (Caetano and Cunha 2004).

The lowest damping ratio for the case study bridge is greater than the values for all,

but the higher end of the damping ratios for the University of Porto stress ribbon

footbridge. Hysteresis of the polyester ropes (McKenna et al. 2004) may be contributing

to the footbridge’s high levels of damping. Since the inclined suspenders are not

prestressed, they have the potential to undergo cyclic loading and unloading that relates

61
)

not only to hysteresis (McKenna et al. 2004), but also to snap loading when going from

slack to taut (Hennessey et al. 2005). Additionally, it is likely that when the suspenders

are slack they will provide higher levels of damping than when they are taut (Ramberg

and Griffin 1977). Finally, since damping relates to the entire system (fib 2005; Sétra

2006) and not just the polyester ropes, it is likely that the timber deck and the

nonstructural mesh fencing are also contributing to the damping.

Damping is also dependent on the amplitude of oscillation (Sétra 2006; Heinemeyer

et al. 2009). Figure 4.12 presents the damping ratio versus the amplitude of oscillation for

each data series analyzed, where the amplitude of oscillation is defined as the average of

the peak accelerations used to generate the exponential fit (Heinemeyer et al. 2009).

4
ξ (%)

2 1st mode
2nd mode
1
rd
3 mode
0
0 0.005 0.01 0.015 0.02 0.025 0.03
Amplitude of Oscillation (g)

Figure 4.12. Damping ratio vs. amplitude of oscillation plot for the three modes.

Figure 4.12 shows that for the first mode, there is an overall positive trend between

amplitude of oscillation and damping. However, the amplitude of oscillation values for

most of the data series are similar with scatter in the damping ratio values. The third

mode also shows a positive trend. The maximum amplitude of all data points in the first

and third modes is 0.025 g, which is lower than typical serviceability criteria such as the

62
)

minimum comfort limit, 0.25 g, listed in the Service d’Études Techniques des Routes et

Autoroutes (Sétra) technical guide (2006) for bridge classes with dynamic criteria

(Classes I to III). Consequently, based on the observed positive trends, even higher

levels of damping are expected when accelerations in the first or third modes approach

this comfort limit. Figure 4.12 also shows that damping ratios in the second mode

demonstrate a slight negative trend.

4.5 Conclusions
This chapter investigated the modal parameters (i.e., natural frequencies and damping

ratios) of polyester-ropes suspended footbridges. Numerical calculations were performed

on a detailed two-dimensional model of a case study footbridge to examine the influence

of material stiffness on the structure’s vertical natural frequencies. The results found

using polyester rope’s lower bound (El = 2.86 GPa) and upper bound (Eu = 5.96 GPa)

elastic moduli values were similar (within 4% of each other) which indicates that a single

modulus of elasticity value can be used instead of modeling polyester rope’s nonlinear

stress-strain behavior in a natural frequency analysis.

An analytical evaluation of a taut string representation of the case study bridge

produced natural frequencies that were within 2% of the natural frequencies found in the

numerical analysis of the detailed model. Consequently, the taut string model can

potentially be used as a surrogate for the detailed model to rapidly analyze the global

response for these suspended footbridges. In the taut string model, span and rope tension

are the variables that contribute to the structure’s stiffness. For the case study bridge, the

63
)

prestress was the main source of rope tension and consequently, prestress was the key

parameter influencing the system’s dynamic stiffness.

Damping ratios for the case study bridge were estimated for the first three modes

identified in the free vibration tests. The damping ratio ranges found for modes 1 – 3

were [2.6, 5.2] %, [3.0, 4.0] %, and [2.8, 6.1] %, respectively. These values are typically

larger than values for other built footbridges including suspension and stress ribbon

structures. The damping ratios presented in this chapter are the first set of values found

for a polyester-rope suspended footbridge. Designers can utilize these values as a starting

point when performing a dynamic analysis on one of these structures. Dynamic analyses

using these damping ratios are presented in Chapters 5 and 7.

64
)

Chapter 5

Bridge Response to Pedestrian

Excitation

5.1 Introduction
Chapter 4 demonstrated that vertical natural frequencies for the case study polyester-rope

suspended footbridge are significantly lower than the 3 Hz limit specified in the

American Association of State and Highway Transportation Officials (AASHTO)

guidelines (2009) for avoiding pedestrian vibration issues. If it is not possible to shift a

bridge’s natural frequencies outside the range of vulnerability by modifying the

structure’s mass or stiffness, then the bridge may still be acceptable from a pedestrian

serviceability perspective if its accelerations are lower than acceptable comfort limits

(Sétra 2006). Consequently, it is important to establish a bridge’s dynamic response to

pedestrian excitation to determine whether the structure can be considered serviceable.

The primary objective of this chapter is to characterize the influence of parameters

such as mass, damping, and polyester rope’s material stiffness on the accelerations and

65
)

rope tensions for polyester-rope suspended footbridges. This task is accomplished by

performing numerical modal calculations on the Ait Bayoud case study bridge

(introduced in Chapter 3) using the dynamic analysis methodology presented in the

Service d’Études Techniques des Routes et Autoroutes (Sétra) technical guide (2006).

The modal calculations that are introduced in this chapter will also be implemented in the

optimization methodology described in Chapter 7.

A secondary objective is to identify the taut string model as a surrogate for the

numerical model that is capable of approximating the accelerations found numerically

with less computational expense.

5.2 Load Definitions


This section presents the loads used in the pedestrian excitation analysis. The AASHTO

guidelines (2009) do not specify the dynamic loads that should be applied. However, the

guidelines do refer directly to Sétra’s technical guide (2006), by stating that this guide

provides a potential method for evaluating a bridge’s dynamic response (e.g.,

accelerations, stresses, etc.) under pedestrian excitation.

In this dissertation the dynamic loads used follow from Sétra’s dynamic analysis

methodology for a Class III footbridge (Sétra 2006). As discussed in Chapter 2, Sétra’s

footbridge classes correspond to expected traffic on a structure (Sétra 2006). Class III is

the most appropriate classification for a rural footbridge that is expected to meet dynamic

comfort limits. Class I and II footbridges have higher traffic levels than Class III

footbridges. Class IV footbridges are located in rural areas that do not need to meet

comfort requirements.

66
)

The uniformly distributed, but time varying dynamic load for the case study bridge is

defined as:

wv = 269.9 cos(2π ft) ξ N/m (5.1)

where f is natural frequency, ξ is the damping ratio, and t is time. The calculations used to

solve for the coefficient in equation 5.1 (269.9 N/m) are presented in Appendix C. A

dynamic analysis is performed for each of the bridge’s modes with natural frequencies

between 1.7 Hz and 2.1 Hz. The dynamic load in each analysis will depend on the natural

frequency of the mode under investigation. The 1.7 Hz to 2.1 Hz range corresponds to

the set of frequencies at which a bridge is at the maximum risk of experiencing resonance

(Sétra 2006).

Sétra (2006) in its methodology specifies that natural frequencies should be

calculated for two different assumed masses: (i) Dead load (DL) only and (ii) DL + 71.4

kg/m2 live load (LL). For the 1.02 m wide case study bridge the dead load and live loads

are equivalent to 57.1 kg/m (refer to Chapter 3 for a description of the components

contributing to the DL) and 72.8 kg/m, respectively. The following load combinations

(LCs) that accompany these masses are:

LC1: prestress + DL

LC2: prestress + DL + LL

The distributed dynamic load (wv) is applied along the bridge span and its direction

corresponds with the concavity of the mode shape associated with the mode under

investigation. Where the mode shape is concave down, the load is applied in the gravity

direction. The loading is reversed when the mode shape is concave up. See Figures 5.2 –

67
)

5.4 in Section 5.3.2 for examples of how the loading direction changes as the mode shape

concavity changes.

The distributed load represents a fictitious number of pedestrians (nf) walking at one

of the bridge’s natural frequencies ( f ) and causing resonance. The response of this

fictitious set of pedestrians is equivalent to the response of a larger number of actual

pedestrians (n) not walking in-sync with one of the bridge’s natural frequencies. The

calculations used to determine the actual number of pedestrians on the bridge (32.6

pedestrians), are presented in Appendix C. The relationship between the fictitious and the

actual number of pedestrians is defined by the following expression (Sétra 2006):

n f = 10.8 ξn (5.2)

The damping ratios considered in the analysis and the corresponding synchronized group

sizes are presented in Section 5.3.2.

5.3 Detailed Model: Numerical Dynamic Analysis


This section presents the numerical dynamic analysis performed on the case study bridge

to show the impact of variables such as mass, damping, and polyester rope’s material

stiffness on the bridge’s dynamic response.

5.3.1 Numerical Model Properties

The detailed model presented in Chapter 3 is utilized in this analysis. The model’s

geometry and boundary conditions are presented again in Figure 5.1. The calculations are

performed for the two material stiffness cases (c1 and c2) introduced in Chapter 3. Case

68
)

c1 uses polyester rope’s lower-bound modulus of elasticity (El), 2.86 GPa. Case c2 uses

polyester rope’s upper-bound modulus of elasticity (Eu), 5.96 GPa.

Figure 5.1. Elevation of the detailed model of the case study. The geometry is for case c1
subject to LC1.

5.3.2 Numerical Calculations

Numerical natural frequency computations like the ones presented in Chapter 4 are

performed for LC1 and LC2 to determine which modes have natural frequencies that fall

between 1.7 Hz and 2.1 Hz, the range corresponding to a high risk of resonance. For case

c1 (El = 2.86 GPa) under LC1, the fourth mode has a natural frequency of 2.05 Hz and

under LC2 the fifth and sixth modes have natural frequencies of 1.74 Hz and 2.09 Hz,

respectively. Therefore, three dynamic analyses, one performed using each of these three

modes with its associated dynamic load, wv are required for case c1. Figures 5.2a, 5.3a,

and 5.4b show the three relevant mode shapes. Figures 5.2b, 5.3b, and 5.4b show the

dynamic load arrangements that correspond to each of the mode shapes. These loads are

applied to the static equilibrium position that results from the application of LC1 or LC2.

69
)

a. Vertical mode shape.

b. Corresponding dynamic load arrangement applied to the deformed (under LC1) bridge.
Figure 5.2. Case c1 LC1 fourth mode.

a. Vertical mode shape.

b. Corresponding dynamic load arrangement applied to the deformed (under LC2) bridge.
Figure 5.3. Case c1 LC2 fifth mode.

a. Vertical mode shape.

b. Corresponding dynamic load arrangement applied to the deformed (under LC2) bridge.
Figure 5.4. Case c2 LC2 sixth mode.

70
)

For case c2 (Eu = 5.96 GPa) under LC1 the fourth mode has a natural frequency of

2.08 and under LC2 the fifth mode has a natural frequency of 1.82. Therefore, only two

dynamic analyses are required for case c2. The mode shapes and dynamic load

arrangements for modes 4 and 5 will be similar to those presented in Figures 5.2 and 5.3,

respectively. Table 5.1 summarizes for the two cases (c1 and c2) the natural frequencies

that fall within the range corresponding to a high risk of resonance.

Table 5.1. Natural frequencies in the range corresponding to a high risk of resonance.
f (Hz), mode
Load combination
Case c1 Case c2
LC1 2.05, 4th 2.08, 4th
1.75, 5th 1.82, 5th
LC2
2.10, 6th -

The five dynamic analyses are performed using modal calculations. In a modal

analysis the global response is found by combining the responses from each of the

individual modes (Chopra 2007). Since each dynamic load is harmonic and acting in

resonance with one of the bridge modes, the global response is dominated by and

therefore, can be approximated by just the response in the single mode. The response

amplitudes at resonance are the values of interest (Sétra 2006). Therefore, it is only

necessary to calculate the maximum (across time) steady-state response for the mode

under consideration. The transient response can be neglected because it does not

contribute to the resonant response (Sétra 2006).

Appendix C presents the equations for steady-state displacements and accelerations.

While these values vary along the length of the bridge, the values at a specific point are

the maxima across time. Appendix C also shows how the dynamic displacements are

71
)

superimposed with the static displacements from the associated static load combination

(LC1 or LC2) and then converted into rope force results.

Since damping ratios for the modes of interest (LC1 fourth mode and LC2 fifth and

sixth modes) have not been determined experimentally, the calculations are performed for

a range of damping ratios to see how damping influences the results. This range

corresponds approximately with the values found in Chapter 4 for the first three vertical

modes under LC1 (2.6% to 6.1%) and includes damping ratios between 2% and 6%.

Substituting these damping ratios into equation 5.2 indicates that the dynamic loads on

the bridge in the analyses correspond to groups that range in size from approximately 9

(at 2% damping) to 15 (at 6% damping) synchronized pedestrians.

5.3.3 Numerical Results and Discussion

This section first discusses the maximum acceleration and rope tension results for case c1

(El = 2.86 GPa). Then, the results for cases c1 and c2 are compared to show how material

stiffness impacts accelerations and rope tensions.

5.3.3.1 Lower Bound Material Stiffness Case, Results and Discussion

The maximum accelerations (amax) and rope tensions (Tmax) along the bridge for case c1’s

(lower bound material stiffness case, El = 2.86 GPa) three modes across the damping

ratio range of 2% to 6% are included in Table 5.2. The LC1 mode has larger accelerations

than the LC2 modes. Accelerations for the two LC2 modes for a given damping ratio are

approximately the same in magnitude; these values never differ by more than 2%. In all

instances, the maximum acceleration exceeds Sétra’s (2006) 0.25 g minimum comfort

criterion for a Class III bridge.

72
)

Table 5.2. Maximum acceleration and rope tension results from the numerical
calculations for case c1 and variable damping.
ξ Tmax (kN)
Load combination f (Hz), mode amax (g)
(%) Hand ropes Below-deck ropes Suspenders
LC1 2.05, 4th 2.15 151.1 100.8 0.7
2 1.75, 5th 0.95 162.5 109.0 1.0
LC2
2.10, 6th 0.94 162.1 108.7 1.0
LC1 2.05, 4th 1.76 150.7 100.5 0.6
3 1.75, 5th 0.77 162.2 108.7 0.9
LC2
2.10, 6th 0.77 161.9 108.5 0.9
LC1 2.05, 4th 1.52 150.3 100.4 0.6
4 1.75, 5th 0.67 162.0 108.6 0.9
LC2
2.10, 6th 0.67 161.8 108.4 0.9
LC1 2.05, 4th 1.36 150.5 100.3 0.5
5 1.75, 5th 0.60 161.9 108.5 0.8
LC2
2.10, 6th 0.60 161.7 108.4 0.8
LC1 2.05, 4th 1.24 150.4 100.3 0.5
6 1.75, 5th 0.55 161.8 108.5 0.8
LC2
2.10, 6th 0.54 161.7 108.3 0.8

The primary factor influencing the difference between the LC1 and LC2 results is that

the LC1 mass (mLC1) is lower than the LC2 mass (mLC2). The ratio (mr) of the LC2 mass-

to-LC1 mass is 2.25:

mLC2 (57.1 kg/m + 71.4 kg/m )(64 m span )


mr = = = 2.25 (5.3)
mLC1 (57.1 kg/m)(64 m span )
For a given damping value this mass ratio is approximately equal to the inverse of the

ratio (ar) of the LC2 maximum acceleration (amaxLC2) to the LC1 maximum acceleration

(amaxLC1). As an example, the inverse of the acceleration ratio is calculated using the fifth

mode results for 2% damping:

−1 −1
⎛ amaxLC2_mode5 ⎞ ⎛ 0.95 g ⎞
a = ⎜⎜
−1
r
⎟⎟ = ⎜⎜ ⎟⎟ = 2.26 (5.4)
⎝ amaxLC1 ⎠ ⎝ 2.15 g ⎠

The inverse of the acceleration ratio is 2.26 which is 1.0% larger than the mass ratio,

2.25. There is not an exact match because as shown in equation C.6 in Appendix C

73
)

accelerations are also dependent on the eigenvector components which differ in all the

modes under investigation.

Comparing results for different damping ratios indicates that as damping increases,

acceleration decreases by a factor equivalent to ξ . Equation C.14 in Appendix C shows

this relationship by combining contributions from the equations for the dynamic load and

the steady-state accelerations.

The overall maximum rope tensions (Tmax) in the hand ropes, below-deck ropes, and

suspenders from all instances presented in Table 5.2 are 162.5 kN, 109.0 kN, and 1.0 kN,

respectively. The lumped strength limits (Tult) for these ropes are 595.2 kN (six ropes

each with a limit of 99.2 kN), 396.8 kN (four ropes each with a limit of 99.2 kN), and 49

kN (2 ropes each with a limit of 24.5 kN), respectively. The corresponding minimum

factors of safety (FS) are determined using the equation:

Tult
FSmin = (5.5)
Tmax

Table 5.3 summarizes these tension and factor of safety results.

Table 5.3. Maximum rope tension and minimum factor of safety results for the numerical
calculations for case c1.
Rope Tult (kN) Tmax (kN) FSmin
Hand rope 595.2 162.5 3.66
Below-deck rope 396.8 109.0 3.64
Suspenders 49.0 1.0 49.00

Rope tensions are larger for the LC2 cases than the LC1 case because the static

contributions in the former are greater. Between the LC2 cases, the higher mode case

(sixth mode, f = 2.10 Hz) has lower rope tensions. This occurs because in the higher

mode the dynamic displacements and consequently the equivalent dynamic axial forces

74
)

are less than in the lower mode. Equation C.5 in Appendix C shows that the dynamic

displacements are inversely proportional to the modal frequency squared. As was found

for accelerations, the dynamic displacements decrease by a factor equivalent to ξ . The

decreased dynamic displacements again translate into lower rope tensions.

In the LC1 cases, some of the suspenders are found to be loaded in compression. The

static analysis is nonlinear and ensures that in the final iteration the ropes either carry

tension or have zero force to indicate they have gone slack. However, the natural

frequency and dynamic analyses are linear and no adjustments are made to ensure that the

ropes remain only in tension. Since, the dynamic analysis results are simply

superimposed on the static analysis results it is possible to have members loaded in

compression in the final configuration. Under LC2, suspenders are never loaded in

compression because the additional LL mass provides high enough static tensions that are

not overcome by the dynamic contributions. For the LC1 results to be feasible for the

case study bridge, the rope suspenders would have to be replaced by members capable of

taking compression or prestressed to avoid going into compression. In general, the

dynamic analysis results are valid either for: (i) suspenders that are capable of taking both

tension and compression or (ii) tension-only suspenders when the static tensions are high

enough that no members are put into compression following the addition of any dynamic

load.

75
)

5.3.3.2 Comparison of Lower and Upper Bound Material Stiffness Cases, Results

and Discussion

The maximum accelerations (amax) and rope tensions (Tmax) along the bridge for all modes

investigated in cases c1 (El = 2.86 GPa) and c2 (Eu = 5.96 GPa) when a damping ratio of

2% is used are presented in Table 5.4. The results for the 2% damping ratio are

compared, because as shown in Section 5.3.3.1 for case c1, this damping ratio out of all

the damping ratios considered, provides the overall maximum accelerations and rope

tensions.

Table 5.4. Maximum acceleration and rope tension results from the numerical
calculations for cases c1 and c2 and 2% damping.
Load Tmax (kN)
Case f (Hz), mode amax (g)
combination Hand ropes Below-deck ropes Suspenders
LC1 2.05, 4th 2.15 151.1 100.8 0.7
c1 1.75, 5th 0.95 162.5 109.0 1.0
LC2
2.10, 6th 0.94 162.1 108.7 1.0
LC1 2.08, 4th 2.15 154.8 103.3 0.7
c2
LC2 1.82, 5th 0.95 171.8 114.8 1.0

For a given load combination and mode, the two cases have the same accelerations.

For example, under LC1 the fourth mode maximum accelerations for both cases are 2.15

g. The modulus of elasticity does not factor directly into acceleration equation C.6 in

Appendix C and the only remaining variable in equation C.6 that the modulus of

elasticity can influence is the eigenvector for the mode under investigation. The

similarities in the acceleration results indicate that the difference in elastic moduli values

for the lower-bound (c1) and upper-bound (c2) material stiffness cases is not resulting in

significant differences in the eigenvectors.

For a given load combination and mode, the two cases have different maximum rope

tensions. These force differences occur because of the influence of the modulus of

76
)

elasticity on the static contributions to the rope tensions. As shown in Chapter 3, the

higher modulus case deflects less than the lower modulus case under a given static load

which results in the higher modulus case having higher tensions. These tension

differences lead to different factors of safety for the two cases. Table 5.5 provides case c2

minimum factors of safety (FSmin) that correspond to the overall maximum rope tensions

(Tmax) presented in Table 5.4. Also included in Table 5.5 are the case c1 results first

presented in Table 5.3. The case c2 minimum factors of safety are within 6% of the case

c1 minimum factors of safety.

Table 5.5. Maximum rope tension and minimum factor of safety results for the numerical
calculations for cases c1 and c2.
Case c1 Case c2
Rope  Tult (kN) 
Tmax (kN) FS Tmax (kN) FS
Hand rope 595.2 162.5 3.66 171.8 3.46
Below-deck rope 396.8 109.0 3.64 114.8 3.46
Suspenders 49.0 1.0 49.00 1.0 49.00

In Chapter 3, where a greater LL was used in the static analysis (Chapter 3: LL = 3.84

kN/m versus Chapter 5: LL = 0.71 kN/m), the minimum factor of safety for the upper-

bound modulus of elasticity case was 2.16. Since the minimum factor of safety found in

the dynamic analysis, namely 3.46, is 60% larger than the value from the static analysis,

system strength under dynamic loading is not likely to be a key design criterion for

polyester-rope suspended footbridges.

5.4 Taut String Model: Analytical Dynamic Evaluation


This section presents the analytical dynamic calculations performed on a taut string

model that is used to represent the case study bridge. A comparison between the

77
)

analytical and numerical acceleration results indicates that this simple model closely

approximates the detailed model (Figure 5.1) analyzed in Section 5.3.

5.4.1 Analytical Model


In Chapter 4 it was shown that the case study bridge can be represented by a single flat

rope whose natural frequencies can be approximated with the equation for a taut string.

This taut string model is now used to calculate accelerations for the case study. The

geometry and boundary conditions for the taut string model first presented in Chapter 4

are shown again in Figure 5.5.

Figure 5.5. Elevation of the taut string model.

5.4.2 Analytical Calculations


Appendix C provides the derivation for the taut string equation used to determine the

maximum accelerations (amax) across time and along the structure. This equation is:

539.8 N/m
amax = (5.6)
ξ jπ m

where j is the mode under investigation, ξ j is the modal damping ratio, and m is the

linear mass for the load combination under consideration.

5.4.3 Analytical Results and Discussion

This section compares the acceleration results calculated analytically with equation 5.6 to

the numerical results presented in Section 5.3.3.1 for case c1 (El = 2.86 GPa). The case c2

78
)

(Eu = 5.96 GPa) results are not included because as discussed in Section 5.3.3.2 the

difference between the case c1 and case c2 acceleration results is minor. Table 5.6

presents these analytical and numerical (previously presented in Table 5.2) results for the

three modes of interest across the damping ratio range of 2% to 6%. Equation 5.6

indicates that for a taut string, the modal damping ratio (ξ) is the only mode specific

variable that influences the maximum acceleration (amax). Otherwise, the maximum

acceleration is independent of the mode under consideration. In this analysis the same

damping ratio (ξ) is assumed for all modes, so only a single acceleration value is found

for each load combination. The percent differences listed in Table 5.6 are calculated with

the equation:

amax_numerical − amax_analytical
amax % difference = x 100 (5.7)
amax_numerical

Table 5.6. Maximum acceleration results from the analytical and numerical calculations.
amax (g)
ξ (%) Load combination f (Hz), mode amax % difference
Numerical Analytical
LC1 2.05, 4th 2.15 2.17 0.9
2 1.75, 5th 0.95 1.1
LC2 0.96
2.10, 6th 0.94 2.1
LC1 2.05, 4th 1.76 1.77 0.6
3 1.75, 5th 0.77 1.3
LC2 0.78
2.10, 6th 0.77 1.3
LC1 2.05, 4th 1.52 1.53 0.7
4 1.75, 5th 0.67 1.5
LC2 0.68
2.10, 6th 0.67 1.5
LC1 2.05, 4th 1.36 1.37 0.7
5 1.75, 5th 0.60 0
LC2 0.60
2.10, 6th 0.60 0
LC1 2.05, 4th 1.24 1.25 0.8
6 1.75, 5th 0.55 0
LC2 0.55
2.10, 6th 0.54 1.9

The maximum percent difference from all the cases is 2.1%. The small percent

differences between the numerical and analytical accelerations results indicate that the

79
)

taut string model closely approximates the case study bridge for the purposes of

calculating maximum accelerations.

5.5 Conclusions
This chapter utilized numerical pedestrian excitation calculations performed on a detailed

two-dimensional model of a case study polyester-rope suspended footbridge to establish

the influence of mass, damping ratios, and material stiffness on system accelerations and

rope tensions. These excitations were equivalent to groups of approximately 9 to 15

synchronized pedestrians (low levels of traffic) walking at bridge natural frequencies in

the range susceptible to resonance, namely 1.7 Hz to 2.1 Hz (Sétra 2006).

The calculations were conducted for damping ratios in the range of 2% to 6%. Lower

accelerations were found with higher damping ratios, but for the case study evaluated,

even using the upper end of the damping range, the minimum pedestrian comfort limit of

0.25 g (Sétra 2006) was exceeded by a large factor, 5 times for the controlling mode (the

LC1 fourth mode). The numerical calculations demonstrated the inverse relationship

between mass and acceleration. Optimization studies presented in Chapter 7 will examine

using supplemental mass as a strategy to reduce bridge accelerations to acceptable

comfort levels.

The calculations considered both lower bound (El = 2.86 GPa) and upper bound (Eu =

5.96 GPa) elastic moduli. For the two stiffness cases, maximum accelerations were the

same while the maximum rope tensions were within 6% of each other. The minor

differences in acceleration and tensions for the two stiffness cases suggests that

80
)

neglecting polyester rope’s nonlinear stress-strain behavior and using a single modulus of

elasticity value in the dynamic analysis does not significantly impact the results.

Comparing the minimum factors of safety found in the Chapter 3 static analysis and

this chapter’s pedestrian excitation analysis indicates that strength limits under static

rather than dynamic loads are likely to control the design of rural (Class III) polyester-

rope suspended footbridges. The impact of static and dynamic strength limits on the

required rope prestress, cross-sectional area, etc. for medium-span (15 m to 64 m)

polyester-rope suspended footbridges is further investigated in the optimization studies

presented in Chapter 7.

Results from the analytical calculations indicate that just as the taut string model

showed potential as a surrogate for the numerical model for determining natural

frequencies in Chapter 4, the taut string is also able to provide acceleration results

approaching those of the numerical analysis (within 2.1%) at low computational expense.

81
)

Chapter 6

Multi-Objective Optimization with

Stress, Slope, and Natural Frequency

Constraints

6.1 Introduction
Chapter 3 demonstrated that the large static deflections that can occur for polyester-rope

suspended footbridges lead to high live load (LL) overload factors (ratio of applied LL-

to-service LL) at a bridge’s strength limit. While these large deflections also have

potential to control the LL on the bridge by deterring pedestrians from walking onto the

structure until those on the structure have finished crossing, this practice may not always

be acceptable. Instead, the bridge may be required to meet a serviceability criterion such

as a maximum walking slope that necessitates limiting deflections under service loads.

This chapter presents multi-objective optimization problems in which a walking slope

criterion is used as a constraint that must be met for a design to be considered feasible.

82
)

Other constraints include static stress and natural frequency. Accelerations are not

considered in this study, because the natural frequency constraint is set such that no

dynamic analysis and consequently no acceleration check are required. Chapter 7

presents an extension to this study that includes an acceleration constraint.

The first objective of this chapter is to present the novel methodology developed for

the optimization process. This methodology combines a non-dominated sorting genetic

algorithm with geometric nonlinear static and natural frequency computations that utilize

dynamic relaxation and eigenanalysis algorithms. The second objective of this chapter is

to use the optimization problems to establish the important parameters that produce

minimum volume designs for two-dimensional systems (with or without prestress and

unstayed or stayed from below the deck) with medium spans (15 m to 64 m). Both

polyester-rope and steel-rope footbridges are investigated to see how their minimum

volume designs compare. In each optimization problem, two objective functions are

considered. The first objective function, the span, is maximized while the second

objective function, the volume of the unstressed rope system, is minimized. Maximizing

span in a multi-objective optimization provides solutions across a range of spans. This

enables conclusions to be drawn about what criteria and/or parameters may be critical at

different spans. Similar results could be obtained by repeating for each span of interest, a

single objective optimization, where volume is minimized. Minimizing volume is

important to reduce the quantity and the cost of rope. Since polyester and steel have

different densities, self-weight results are scaled from the volume results to provide an

alternative measure of rope quantity.

83
)

6.2 Structural Model Features


This section describes the features for the structural models evaluated in the optimization

problems. These features include model configurations, material properties, and design

loads.

6.2.1 Model Configurations

In this study, 1 m wide, three-dimensional suspended rope systems with and without

below-deck stays are reduced to two-dimensional models. As was done in Chapter 3, this

reduction is achieved by grouping together all elements occurring at the same elevation.

The stiffness of the deck is not included in these models because it is assumed that the

deck’s detailing is such that it does not contribute to the global stiffness of the system.

Figures 6.1a – 6.1c show detailed two-dimensional configurations that consist of

components such as suspended hand and below-deck ropes, suspenders, backstays, and

below-deck stays (Figures 6.1b and 6.1c). Simplifications to the configurations in Figures

6.1a – 6.1c appear in Figures 6.1d – 6.1f. The configurations in Figures 6.1d – 6.1f have a

single suspended rope and no suspenders or backstays. The extremes among these

configurations, the low-stiffness and high-stiffness configurations in Figures 6.1d and

6.1f, respectively, are the configurations optimized in this study to observe how much

stays can potentially reduce a system’s volume. The stayed, high-stiffness configuration

consists of a single vertical stay whose support is located at a vertical distance L/2 below

the suspended rope supports (Figure 6.1f). A configuration with a vertical stay has greater

stiffness than a configuration with stays rotated from the vertical by an angle ϕ (Figure

6.1e) when both configurations have equal cross-sectional area, equal rope prestress, and

84
)

stay supports located along or outside of a semi-circle with a radius of L/2 that is centered

about the mid-point between the suspended rope supports. The unstayed configuration

(Figure 6.1d) will have an even lower stiffness.

a. b. c.

d. e. f.
Figure 6.1. Elevations showing the detailed configurations a – c and simplified
configurations d – f. The low and high stiffness simplified configurations shown in d and
f, respectively, are optimized in this study.

While analytical methods such as those used in Chapters 3 and 4 could be used to

evaluate the unstayed configuration (Figure 6.1d), numerical calculations are necessary

for the stayed configuration (Figure 6.1f). In order to provide a general methodology that

can be used to evaluate both of these configurations in the present study, as well as more

complex configurations in the future, all the structural analyses are performed

numerically. For the analysis of each model, the suspended rope is discretized into twelve

elements and where applicable, the below-deck stay is represented with a single element.

85
)

The suspended rope discretization follows from a preliminary study performed on a 64 m

span configuration like the one shown in Figure 6.1d. Models with twelve and twenty-

four elements had stress, slope, and fundamental vertical natural frequency results that

differed by less than 5%. A discretization of twelve elements is used to balance finding a

solution that reasonably approximates the structural behavior under investigation with the

required computational time. This balance also informs the decision to optimize the

simplified (i.e., surrogate) configurations, rather than their more detailed counterparts.

Chapters 3 and 4 demonstrated that surrogate configurations such as the ones under

investigation in this chapter can approximate the static and natural frequency results of

the more detailed configurations.

6.2.2 Material Properties

Two rope materials are considered for the models: polyester and steel. Table 6.1 lists the

moduli of elasticity (E), ultimate stresses (σult), and densities (ρ) for these rope materials.

Table 6.1. Polyester-rope and steel-rope properties.


Rope material E (GPa) σult (MPa) ρ (kg/m3)
Polyester 2.86 471.3 1381
Steel 196.50 1862.0 7849

The polyester rope is the same commercially available, 12-strand, single braid

product used in the case study bridge. For this polyester rope product, the ultimate stress

varies based on the nominal diameter of the rope. The value used in this study is the

minimum ultimate stress for ropes up to 24 mm (nominal) in diameter. The nonlinear

stress-strain behavior of polyester rope (Flory et al. 2004) is neglected and the lower-

bound “modulus of elasticity” of 2.86 GPa is used. By comparison, the steel rope (Grade

86
)

270) has a modulus of elasticity of 196.5 GPa, a value nearly 70 times greater than that of

the polyester rope.

6.2.3 Load Definitions

Prestress, dead (DL), and live (LL) loads are considered in this study. Prestress is applied

to a rope as a strain and varies in the analysis (see Table 6.4 in Section 6.3.5 for the strain

ranges). The DL includes the rope self-weight in addition to a uniformly distributed load,

0.53 kN/m, representative of a wood deck and cross beams, suspenders and mesh

guardrails. This distributed DL is similar to that of the case study bridge introduced in

Chapter 3. The LL is a uniformly distributed load, 4.3 kN/m derived from the 4.3 kN/m2

unreduced area load prescribed in the American Association of State Highway and

Transportation Officials (AASHTO) guidelines (2009) applied across the 1 m wide deck.

Both the uniformly distributed DL and LL are applied along the entire span as equivalent

concentrated loads located at element nodes.

The following three working stress load combinations (LCs) are considered in the

analysis:

LC1: prestress

LC2: prestress + DL

LC3: prestress + DL + LL

87
)

6.3 Optimization Problems


This section presents the multi-objective optimization problems under investigation.

To capture the performance of a range of systems subject to varying constraints, eight

cases are analyzed, four in each of two multi-objective optimization problems (i and ii).

Every system has an assigned rope material (polyester or steel) and a configuration

(unstayed or stayed). Additionally, the ropes may or may not be prestressed. First, the

general problem formulation is defined and then the objective functions, constraints, and

design variables for the different problems and cases are discussed.

6.3.1 Problem Formulation

The general formulation for the optimization problems is defined as:

minx f(x) (6.1)

subject to: g(x) ≤ 0

where x is the design variable vector, f = [ f1, . . ., fq]T is a vector of q objective functions,

and g = [g1, . . . gr] T is a vector of r constraints. Table 6.2 lists the objective functions and

constraints considered in each of the optimization problems. The variables in Table 6.2

are defined and described in Sections 6.3.2 - 6.3.5.

Table 6.2. Objective functions and constraints for optimization problems i and ii.
Objective
Constraints, g(x)
functions, f(x)
Problem
g3(x) (for stayed
f1(x) f2(x) g1(x) g2(x) g4(x)
configurations only)
max {σ i } − σ all max {s j } − s all
i -L V i =1, ... , n + m j =1, ... , n
- -
max {σ i } − σ all max {s j } − s all if σn+m > 0: -σn+m
ii -L V λmin – λall
i =1, ... , n + m j =1, ... , n else: σn+m+ 0.1

88
)

6.3.2 Objective Functions

In both problems, the first objective function ( f1(x)) is defined as f1(x) = -L, where L is

span. The negative sign appears in the function because the aim is to maximize span, but

the general problem formulation is defined as a minimization. The second objective

function ( f2(x)) is defined as f2(x) = V, where V is rope volume. The aim is to minimize

the second objective function.

6.3.3 Static Constraints

All cases are subjected to two static constraints g1(x) and g2(x). The first constraint is that

the maximum service stress among all the ropes (max{σi} for i =1, … , n+m) does not

exceed the allowable stress (σall):

g1 (x) = max{σ i } − σ all ≤ 0 (6.2)


i =1,..., n + m

where n and m are the number of elements comprising the suspended rope and stay,

respectively. The allowable stress is defined as σult/FS, where σult is the rope’s ultimate

stress and FS is the factor of safety. The polyester-rope and steel-rope ultimate stress

values are given in Table 6.1. No factor of safety has been established in design

guidelines for polyester rope, so a factor of safety typically used for steel ropes, 2.2

(Gimsing and Georgakis 2012), is assumed for ropes constructed of both materials.

The second constraint is that the suspended rope’s maximum service slope

(max{sj} for j =1,…,n) does not exceed the allowable slope (sall):

g 2 ( x) = max{s j } − sall ≤ 0 (6.3)


j =1,..., n

89
)

The allowable slope used in this study is 20%. This value follows from the Bridges to

Prosperity (2011) criterion for rural, steel-rope suspended footbridges.

Cases with stayed configurations in problems ii are also subjected to a slack stay

constraint (g3(x)):

if σn+m > 0: g3(x) = -σn+m ≤ 0 (6.4)

else: the constraint is violated and g3(x) is set to σn+m+ 0.1

where σn+m is the normal stress in a stay element. A slack stay is one that is no longer

under tension and has zero stress. Since the stay is intended to remain tensioned to stiffen

the suspended rope system, only designs in which the stay does not go slack meet the

slack stay constraint.

6.3.4 Natural Frequency Constraint

All cases in problem ii have a natural frequency constraint (g4(x)) in addition to the static

constraints:

g4(x) = λmin – λall ≤ 0 (6.5)

where λmin is the fundamental vertical natural frequency and λall is the limit on the natural

frequency, which in this study is 3 Hz. This value corresponds to the AASHTO (2009)

criterion and avoids the bridge being vulnerable to pedestrian vibrations in the vertical

direction.

6.3.5 Design Variables

Each case’s combination of system features (material, configuration, and whether or not

the ropes are prestressed) dictate that case’s design variables (x). All design variables are

90
)

continuous. In all cases, span (L) is a design variable with L ∈ [15, 64] m. The lower limit

is set at 15 m because other structures, such as beams and trusses, may be more likely

than suspended bridges to be built for shorter spans. The upper limit is taken as the span

of the case study bridge, 64 m. Suspended rope area (A1) is also a design variable in all

cases and A1 ∈ [0.01, 100] cm2. The lower limit is set to a value close to zero. The upper

limit for the suspended rope area is equivalent to approximately forty-eight, 18 mm

(nominal) diameter ropes for the polyester product considered in this study. By

comparison, the case study bridge, which based on the results presented in Chapter 3 and

4 would not have met the slope and natural frequency constraints set in this study, has ten

ropes (six hand and four below-deck ropes) of this size.

For cases with stayed configurations, the stay rope area (A2) is a design variable with

A2 ∈ [0.01, 15] cm2. The lower limit is again set to a value close to zero while the upper

limit is equivalent to approximately four, 24 mm (nominal) diameter polyester ropes.

This upper limit reflects that individual stays will be used in up to four planes parallel to

the span. Stays do not contribute significantly to the static performance of a system

because the stay prestress decreases under increasing DL and LL. However, stays do

influence the natural frequency of a stayed configuration because these components aid in

resisting upward vibrational motions. Therefore, stayed cases are only included in

problem ii, where there is a natural frequency constraint.

For cases with prestress, prestress strains for the suspended and stay (where

applicable) ropes (ε1 and ε2, respectively) are design variables. For polyester-rope cases,

ε 1 and ε 2 ∈ [10 −8 ,7.49x10 −2 ] while for steel-rope cases, ε 1 and ε 2 ∈ [10−8 ,4.31x10 −3 ] . The

lower limits in both ranges are set to a value close to zero to approximate systems

91
)

approaching zero prestress. In order to avoid numerical error, the lower limits are not set

to zero. The suspended rope in a nearly zero prestress system has an almost horizontal

profile prior to the application of DL and/or LL. The upper limits in the ranges are set to

the strain values that occur at the allowable stress levels.

For cases without prestress, the suspended rope prior to the application of DL and/or

LL is assumed to have a parabolic profile. While a free-hanging rope’s true profile is a

catenary, for flat ropes (sag-to-span ratio less than 1:8), a parabola closely approximates a

catenary (Irvine 1975). The relationship between a parabola’s maximum slope (smax), sag

(d), and span (L) is defined by the equation:

4d
smax = (6.6)
L

For a bridge that is at the 20% slope constraint, the sag-to-span ratio is 1:20.

Therefore, all designs that meet this slope constraint can be classified as flat ropes and the

designs can be represented geometrically by parabolas. Assuming the midpoint of the

rope is taken as the origin, the parabola is defined by the equation:

β = bα2 (6.7)

where α and β are the horizontal and vertical coordinates of a point on the parabola

relative to the origin and b is a coefficient. This coefficient is a design variable with

b ∈ [10 −8 , 1.33x10 −4 ] cm2. The lower limit is set close to zero. The upper limit is set such

that if the minimum span value is selected during the optimization the resulting slope is at

the 20% constraint. If longer span values are selected, the upper limit on this coefficient

will result in slopes that exceed 20%. These designs are filtered out during the

optimization process because the slope constraint is not met. Results for the prestressed,

92
)

unstayed cases in problem ii (Section 6.5.2) indicate that prestress is required (i.e., the

prestress strain design variable is not near its lower limit of approximately zero) to meet

the natural frequency constraint. Consequently, nonprestressed cases are only included

for problem i where there is no natural frequency constraint. Figure 6.2 shows the

elevations of the unstayed configurations with and without prestress and the stayed

configuration with prestress prior to the application of DL and/or LL.

a. Unstayed configuration b. Unstayed configuration c. Stayed configuration with


with prestress. without prestress. prestress.
Figure 6.2. Elevations showing the configurations prior to the application of DL and/or
LL.

Table 6.3 summarizes the cases analyzed in the optimization problems while Table

6.4 summarizes the design variable ranges.

Table 6.3. Cases analyzed in the optimization problems.


Problem Case Rope material Configuration Prestressed Design variables, x
d1 yes [L A1 ε1]
polyester Unstayed
d2 no [L A1 b]
i
d3 yes [L A1 ε1]
steel Unstayed
d4 no [L A1 b]
d5 Unstayed [L A1 ε1]
polyester yes
d6 Stayed [L A1 A2 ε1 ε2]
ii
d7 Unstayed [L A1 ε1]
steel yes
d8 Stayed [L A1 A2 ε1 ε2]

93
)

Table 6.4. Design variable ranges for the optimization problems.


Design variable Range
L [15, 64] m
A1 [0.01, 100] cm2
A2 [0.01, 15] cm2
ε1 and ε2 (polyester-rope) [10-8, 7.49x10-2]
ε1 and ε2 (steel-rope) [10-8, 4.31x10-3]
b [10-8, 1.33x10-4]

6.4 Optimization Methodology


This section describes the novel methodology used to evaluate the multi-objective

optimization problems in this study.

6.4.1 Optimization Algorithm

The multi-objective optimization utilizes a non-dominated sorting genetic algorithm,

NSGA-II (Deb et al. 2002) to allow for the identification of feasible solutions in a large

complex design space that encompasses two competing objective functions, static and

natural frequency constraints, and continuous design variables. Zavala et al. (2014) in a

detailed review, show that NSGA-II is one of the most commonly used multi-objective

metaheuristics applied to structural optimization problems. Among the literature cited in

this review are studies demonstrating how NSGA-II has been used when optimizing a

variety of structural systems including trusses (Deb 2001), frames (Greiner et al. 2007),

and grids (Winslow et al. 2010). The NSGA-II algorithm is flexible enough that the

general methodology presented in this study can be applied to other suspended systems

constructed of different materials, and/or having alternative/additional configurations and

constraints. Genetic algorithms can widely explore these design spaces, even if the

problems contain features that may impede gradient-based search techniques such as,

94
)

nonlinear constraints, a combination of continuous and discrete variables, and local

minima.

Genetic algorithms are stochastic search methods in which the main assumption is

that the fittest (closest to optimal) solution has the greatest likelihood of surviving

(Raphael and Smith 2003). With this method, an initial population is generated randomly.

Next, all individuals in the population are subjected to a structural analysis. This analysis

provides the results from which the objective functions and constraints can be calculated.

Then, individuals in a population undergo crossover and mutation operations (i.e.,

individuals are modified using analogous genetic methods) to produce a new population

in the next generation. Individuals with high fitness are selected for these operations at a

greater frequency than individuals with comparatively low fitness. Some low fitness

individuals do not survive. Consequently, on average, each successive generation’s

population improves (Raphael and Smith 2003). In NSGA-II, the constraints are used as

criteria during the determination of fitness (Deb et al. 2002). There are three possible

outcomes. First, when two individuals are feasible (i.e., the constraints are met), the

dominant one is assigned a greater fitness. Second, when two individuals are infeasible,

the one with the lowest constraint violation is assigned a greater fitness. Third, when one

individual is feasible and the other individual is infeasible, the former is assigned a

greater fitness.

In the multi-objective problems considered in this study, the two objective functions

considered are in competition. Therefore, a Pareto-optimal or non-dominated solution set

(Raphael and Smith 2003), instead of a single solution is found for every run performed.

NSGA-II finds a discretized non-dominated solution set that is a representative subset of

95
)

a continuous solution set and convergence of the NSGA-II algorithm to the true Pareto-

optimal solution is not guaranteed (Deb et al. 2002). The genetic parameters used can

influence the resulting non-dominated sets. Preliminary parametric studies indicated that

values for the genetic parameters listed in Table 6.5 produced non-dominated sets that

were comparable to or better (based on a visual inspection of plotted results) than non-

dominated sets found using other values.

Table 6.5. Genetic parameters for all of the optimization problems.


Parameter Value
Number of generations 45
Population size 100
Simulated binary crossover probability 0.75
Simulated binary mutation probability 0.2
Simulated binary crossover distribution index 10
Simulated binary mutation distribution index 20

6.4.2 Structural Analysis Algorithms

The structural analysis used to evaluate the designs generated in the optimization routine

includes geometrically nonlinear static and natural frequency calculations. The static

calculations utilize the dynamic relaxation algorithm introduced in Chapter 3 while the

natural frequency calculations combine the dynamic relaxation and eigenanalysis

algorithms as was done in Chapter 4. In contrast to the lumped mass matrices used in the

eigenalyses performed in Chapters 4, 5, and 7, the mass matrix in this chapter’s study

includes full mass contributions in the vertical and longitudinal directions. Preliminary

studies and checks on the designs found in the optimization indicated that the

fundamental in-plane natural frequencies for these systems were in the vertical and not

the longitudinal direction. Therefore, retaining the full longitudinal mass contributions in

the matrix did not appreciably impact the results.

96
)

6.4.3 Objective Function and Constraint Calculations

Table 6.6 shows which load combinations (LCs) are used to determine the objective

function and constraint values for the different cases. In some instances, the design

variable vector rather than a load combination is used. These instances are indicated in

Table 6.6 with an x. In Sections 6.4.3.1 – 6.4.3.4, the entries in this table are described in

more detail.

Table 6.6. Load combinations (LCs) used in the structural analysis to determine the
objective function and constraint values.
Natural frequency
Static analysis
Problem Case Configuration analysis
f1(x) f2(x) g1(x) g2(x) g3(x) g4(x)

i d1 – d4 Unstayed x x LC3 LC3 - -

d5, d7 Unstayed x x LC3 LC3 - LC2


ii LC1,
d6, d8 Stayed x LC3 LC3 LC3 LC2
LC3

6.4.3.1 Objective Function Calculations

For all cases (d1 – d8), the span objective function ( f1(x)) is determined directly from the

design variable vector. For cases d1 – d4 (unstayed configurations with and without

prestress) in problem i, the unstressed volume objective function ( f2(x)) is calculated

using the design variable vector. Similarly, for cases d5 and d7 (unstayed configurations

with prestress) in problem ii, the volume objective function is calculated with the design

variable vector. For cases d6 and d8 (stayed configurations with prestress) in problem ii,

the volume objective function is calculated using the strain results of LC3. This

computation is required because the combination of prestress strains selected from the

design variable vector may result in a structure that is not initially in equilibrium. During

the dynamic relaxation process, the prestress and consequently the initial lengths and

97
)

volume of the system are updated until equilibrium is reached. The same procedure is not

required for the unstayed configurations with and without prestress because these systems

are initially in equilibrium, regardless of their combination of design variables.

6.4.3.2 Constraint Calculations for the Problem with Static Constraints Only

Cases d1-d4 (unstayed configurations with and without prestress) included in

optimization problem i are only subjected to stress (g1(x)) and slope (g2(x)) constraints,

so only static calculations are required. These calculations are performed for LC3. By

inspection, the LC3 stresses and slopes are greater than the corresponding LC1 and LC2

values.

6.4.3.3 Constraint Calculations for the Unstayed Configurations in the Problem

with Static and Natural Frequency Constraints

Cases d5 and d7 (unstayed configurations with prestress) included in optimization

problem ii are subjected to stress (g1(x)) and slope (g2(x)) constraints whose calculations

are performed for LC3. Additionally, these cases must meet a natural frequency

constraint (g4(x)). The natural frequency calculations are executed with LC2 because

AASHTO (2009) states that LL should not be included in the analysis.

6.4.3.4 Constraint Calculations for Stayed Configurations in the Problem with

Static and Natural Frequency Constraints

Cases d6 and d8 (stayed configurations with prestress) included in optimization problem

ii are also subjected to stress (g1(x)) and slope (g2(x)) constraints. The slope calculations

are performed for LC3. The stress calculations are executed with LC3 to check the

maximum stress in the suspended rope and with LC1 to check the maximum stay stress.

98
)

Stay stresses decrease with the addition of DL and LL. As a result, LC1, rather than LC2

or LC3 yields the maximum stay stress. Stayed cases in this problem have another static

constraint, i.e., the slack stay constraint (g3(x)). This constraint is checked under the

combination most likely to result in zero stress in the stay, LC3. Finally, these cases are

subjected to a natural frequency constraint (g4(x)) whose calculated are performed for

LC2.

6.4.4 Performance Measurement Algorithm

For each case, the optimization algorithm is run ten times because the stochastic

optimization method produces different non-dominated solution sets every time it is

implemented. The choice for the number of runs needs to balance finding a solution set

that reasonably approaches the true Pareto-optimal set with the computational time

required for the analysis. Generating multiple non-dominated sets increases the likelihood

of finding solutions that fall closely to the true Pareto-optimal set. The performance

measure used to compare solution sets, the S-metric (Zitzler 1999), relates to the size of

the space dominated by a set. After the S-metric is found for all runs, the mean value (μs),

standard deviation (σs), and maximum value (ms) of the S-metric for the case are

calculated. In this study, the run with the maximum S-metric value is considered to have

the non-dominated set closest to the true Pareto-optimal set. Figure 6.3 shows the S-

metric values for two general non-dominated sets. In Figure 6.3, the S-metric value for

non-dominated set 1 is larger than the S-metric value for non-dominated set 2. The reader

is referred to Zitzler et al. (2002) for a comparative analysis of performance measures

including the S-metric.

99
)

Figure 6.3. S-metric values for two general non-dominated sets.

6.5 Results and Discussion


This section presents and discusses the results for the cases evaluated in optimization

problems i and ii. Mean value (μs), standard deviation (σs), and maximum value (ms)

results for the S-metric computations are presented in Table 6.7.

Table 6.7. S-metric results for the cases included in optimization problems i and ii.
Problem Case μs (cm4) σs/μs (%) ms/μs – 1 (%)
d1 1.23x109 0.14 0.19
i d2 4.14x107 2.46 2.27
d3 4.31x108 0.59 0.64
d4 1.52x109 0.09 0.10
d5 2.90x108 0.87 0.96
d6 7.09x108 1.82 1.92
ii
d7 8.97x108 0.84 0.90
d8 1.38x108 7.21 6.71

In Table 6.7, σs/μs ranges from 0.09% to 7.21% and ms/μs – 1 ranges from 0.10% to

6.71%. While these ranges depend on the reference point selections used in the S-metric

computations, the results suggest that the spread of solutions is typically low. For each

case in Table 6.7, the non-dominated set for the run with the maximum S-metric value is

plotted as a function of span. These plots (Figures 6.4 and 6.6) are presented in Sections

100
)

6.5.1 and 6.5.2. Comparisons between the various cases are made by inspecting the plots

and the underlying objective function, constraint, and design variable ranges. Although

the S-metric is used to differentiate between runs of a single case, this metric is not used

to compare cases because the reference points vary between cases. In all of the plots

presented in this section the spans are negative values. As described in Section 6.3.2,

spans are negative because the general optimization problem formulation is defined as a

minimization, while the aim is to maximize span.

6.5.1 Results and Discussion for the Problem with Static

Constraints Only

The volume versus span results for optimization problem i are presented in the Figure 6.4

plot. This plot shows that the steel-rope case without prestress (d4) dominates the case

with prestress (d3). This phenomenon occurs because the prestressed solutions do not

deflect as much (slope constraint is not active) as the nonprestressed solutions (slope

constraint is active). For a suspended rope system, increased deflections lead to decreased

horizontal forces, which result in decreased rope volumes.

In contrast to the steel-rope cases, the polyester-rope case with prestress (d1)

dominates the case without prestress (d2). Due to polyester-rope’s low material stiffness,

nonprestressed systems can deflect too much, causing the slope limit to be exceeded

unless large areas are provided. Figure 6.4 shows that the results for the nonprestressed

systems (case d2) do not extend across the span range. The area required for the

suspended rope is approaching the upper area limit (i.e., 100 cm2) specified in Section

6.3.5, for systems with spans less than 64 m. As described in Section 6.3.5, the upper

101
)

limit is already set at a value that is 4.8 times larger than what was used in the Ait

Bayoud bridge. Consequently, prestress is required to limit deflections such that the slope

constraint is met with less volume. For the plotted case, the prestress strains range from

0.063 to 0.067. The maximum strain is 89.5% of the 0.0749 strain upper limit specified in

Section 6.3.5. Since this upper strain limit corresponds to the allowable stress, the results

indicate that prestress is the main contributor to the stress in the polyester-rope case.

5
x 10
4
d1 − polyester, unstayed with prestress
d2 − polyester, unstayed without prestress
3 d3 − steel, unstayed with prestress
d4 − steel, unstayed without prestress
f2 (cm3)

0
−65 −60 −55 −50 −45 −40 −35 −30 −25 −20 −15
f1 (m)

Figure 6.4. Volume versus span for optimization problem i. The plot presents the non-
dominated sets for the four cases (d1 – d4) included in the problem.

In both the dominant steel-rope (d4) and polyester-rope (d1) cases, the stress

constraint in addition to the slope constraint is active. The steel-rope case (d4) is able to

meet these constraints with a lower volume than the polyester-rope case (d1) across the

span range. When the volume results for problem i are transformed into self-weight

results, the dominant polyester-rope case (d1) has lower self-weight values than the

dominant steel-rope case (d4) across the span range (Figure 6.5).

102
)

30
d1 − polyester, unstayed with prestress
25 d2 − polyester, unstayed without prestress
d3 − steel, unstayed with prestress
20
d4 − steel, unstayed without prestress
f2 (kN)

15

10

0
−65 −60 −55 −50 −45 −40 −35 −30 −25 −20 −15
f1 (m)

Figure 6.5. Self-weight versus span for optimization problem i. These results are scaled
from the volume versus span results presented in Figure 6.4

Polyester-rope and steel-rope volume (V) and self-weight (Wself) values for the

dominant cases at one example span, 64 m (the maximum span in the design variable

range) are presented in Table 6.8. The polyester-rope case (d1) has a volume that is 3.8

times greater and a self-weight that is 1.5 times less than the corresponding steel-rope

(d4) values.

Table 6.8. Volume and self-weight results for the 64 m span solutions in the dominant
polyester-rope (d1) and steel-rope cases (d4) in optimization problem i.
Case V (cm3) Wself (kN)
5
d1 – polyester, unstatyed with prestress 2.11x10 2.86
d4 – steel, unstayed without prestress 5.57x104 4.29

6.5.2 Results and Discussion for the Problem with Static and

Natural Frequency Constraints

The volume versus span results for optimization problem ii are presented in the Figure

6.6 plot. This plot shows that the stayed polyester-rope case (d6) dominates the unstayed

polyester-rope case (d5). For these cases, the natural frequency constraint is typically

103
)

active. The stay stiffens the structural system against vibrations, rather than additional

area for and/or prestress in the suspended rope stiffening the system. This phenomenon

leads to the natural frequency constraint being met with less volume in the stayed case

(d6) than in the unstayed case (d5). For example, at a span of 29.4 m (the maximum span

attainable for the unstayed case given the design variable ranges) the unstayed case (d5)

has a volume of 2.69x105 cm3 which is approximately 2 times greater than the stayed

case’s (d6) 1.32x105 cm3 volume.

5
x 10
4

3
f2 (cm3)

2
d5 − polyester, unstayed with prestress
d6 − polyester, stayed with prestress
1
d7 − steel, unstayed with prestress
d8 − steel, stayed with prestress
0
−65 −60 −55 −50 −45 −40 −35 −30 −25 −20 −15
f1 (m)

Figure 6.6. Volume versus span for optimization problem ii. The plot presents the non-
dominated sets for the four cases (d5 – d8) included in the problem.

In contrast, the unstayed steel-rope case (d7) dominates the stayed steel-rope case

(d8). This occurs because the slack stay constraint is difficult for the steel-rope system to

meet. In order to satisfy this constraint the suspended rope in the stayed case (d8) can

only undergo a limited amount of deflection and thus, requires greater prestress and area

which leads to greater volume than its unstayed counterpart (d7). It is more difficult for

the stay in the stayed polyester-rope case (d6) to go slack because it can be prestressed to

a significantly higher strain than the steel-rope stay. The ratio of the maximum strains

104
)

(polyester rope-to-steel rope) from the strain ranges specified in Section 6.3.5 is 17.4. For

other stayed configurations, such as where the stays are inclined in plane as shown in

Figure 6.1b and Figure 6.1e, it may be more difficult for the steel stays to slacken.

In problem ii, high prestress forces are required to meet the natural frequency

constraint. For the plotted cases, these prestress strains can nearly reach the upper limits

in the strain ranges specified in Section 6.3.5 (0.0749 and 0.00431 for polyester rope and

steel rope, respectively). Therefore, for these designs, the effects of dead and live load on

the results are minimal compared to the contribution from prestress. The resulting flat

geometries easily meet the slope constraint. However, the prestress forces require high

areas (complementing the high prestress strains) to ensure that the stress constraints are

met. Consequently, systems that must meet a natural frequency constraint in addition to

static constraints (cases in problem ii) require significantly more rope material than

systems that are only required to meet static constraints (cases in problem i). For all cases

in problem ii, the area needed for the suspended rope is approaching the 100 cm2 upper

limit for systems with spans less than 64. In contrast, the dominant polyester-rope (d1)

and steel-rope cases (d4) in problem i could achieve 64 m spans with areas of 35.21 cm2

and 8.67 cm2, respectively.

The self-weight results for problem ii are presented in Figure 6.7. In problem ii (as

was seen in problem i) the dominant steel-rope case (d7) has a lower rope volume, but a

higher self-weight than the dominant polyester rope case (d6) across the span range.

105
)

30

25

20
f2 (kN)

15
d5 − polyester, unstayed with prestress
10 d6 − polyester, stayed with prestress
d7 − steel, unstayed with prestress
5
d8 − steel, stayed with prestress
0
−65 −60 −55 −50 −45 −40 −35 −30 −25 −20 −15
f1 (m)

Figure 6.7. Self-weight versus span for optimization problem ii. These results are scaled
from the volume versus span results presented in Figure 6.6.

6.5.3 Comparison of Polyester-Rope and Steel-Rope Live Load

Overload Factors

This section evaluates the LL overload factors (ratio of applied LL-to-service LL) for

optimal polyester-rope and steel-rope designs at their ultimate stresses (i.e, strength limit

or FS = 1). This overload analysis assumes that both materials remain elastic up to their

designated strength limits. While this investigation is not the primary aim of the study

performed in this chapter, comparing LL overload factors provides insight into the

differences in overload capacity for suspended footbridges built with ropes having low

(polyester rope) and high (steel rope) material stiffness. Chapter 3 first introduced the

concept of a LL overload factor and compared results for lower bound (El = 2.86 GPa)

and upper bound (Eu = 5.96 GPa) material stiffness values for polyester rope. In the

optimization, the material stiffness values (E) used for polyester and steel differ by nearly

70 times. These values are 2.86 GPa for polyester rope and 196.5 GPa for steel rope.

106
)

The 64 m span designs from the unstayed dominant cases d1 and d4 in problem i are

used as examples. The optimal polyester-rope and steel-rope designs under LC3 both

have stress within 1% of their respective allowable stresses. Therefore, the stress

constraint is active and the factor of safety is approximately 2.2 in both instances.

However, at their strength limits the overload factors are 7.1 and 3.05, for the polyester-

rope and steel-rope designs, respectively. As discussed in Chapter 3, polyester-rope

bridges have potential to develop high overload factors because their low material

stiffness results in significant LL deflections and geometric stiffening. In this example,

the polyester-rope bridge deflected 10.78 m at its strength limit which is 5 times greater

than the corresponding deflection (2.15 m) of the steel-rope design.

6.6 Conclusions
This chapter presented a novel optimization methodology that utilized a non-dominated

sorting genetic algorithm with dynamic relaxation and eigenanalysis algorithms to

generate and then evaluate the feasibility of suspended footbridge designs. The

methodology was applied to a set of multi-objective optimization problems to identify the

key parameters that produce minimum volume polyester-rope and steel-rope bridges with

spans ranging from 15 m to 64 m. Polyester-rope suspended bridges, optimized for

minimum rope volume, always required prestress while steel-rope cases only required

prestress when subjected to the vertical natural frequency constraint. For both polyester-

rope and steel-ropes suspended bridges, high prestress levels were required when the

natural frequency constraint was present. The prestress strains needed were approaching

the values (0.0749 and 0.00431 for polyester rope and steel rope, respectively) that occur

107
)

at the ropes’ allowable stress limits. Consequently, the required cross-sectional areas

were also high and for the rope area ranges used in this study, it was not possible to span

64 m with either a polyester-rope or steel-rope suspended footbridge. For the polyester-

rope cases subject to the natural frequency constraint, systems with vertical stays had

lower rope volume and self-weight quantities than their unstayed counterparts.

In addition to providing insight into critical parameters across the span range, the

results were provided graphically as functions of span to enable direct volume or self-

weight comparisons between different footbridge systems given a site requiring a specific

span. This study showed that for the optimization problems and cases considered,

polyester-rope suspended footbridges bridges have higher volumes but lower self-weights

than steel-rope bridges for medium spans.

This section also included a comparison between the LL overload factors (ratio of

applied LL-to-service LL) found at the strength limits for optimal polyester-rope and

steel-rope designs. For the optimal 64 m span designs found in problem i, the LL

overload factors were 7.1 and 3.05 for the polyester-rope and steel-rope bridges,

respectively. The polyester-rope bridge developed the greater overload factor because of

the structure’s lower material stiffness (2.86 GPa for polyester rope versus 196.5 GPa for

steel rope). Due to this lower material stiffness the polyester-rope bridge was able to

deflect more than the steel-rope bridge (10.78 m for the polyester-rope bridge versus 2.15

m for the steel-rope bridge at their respective strength limits) and consequently undergo

greater nonlinear stiffening. As discussed in Chapter 3, without a prescribed factor of

safety for polyester rope in this application (in Section 6.3.3, the factor of safety was

selected to match the typical factor of safety for steel rope), establishing that polyester-

108
)

rope bridges can carry significant LL beyond the required service LL provides an

alternative measure of safety.

109
)

Chapter 7

Multi-Objective Optimization with

Stress, Slope, Natural Frequency, and

Acceleration Constraints

7.1 Introduction
Chapter 6 presented suspended rope bridge designs with minimum volume for medium

spans (15 m to 64 m) that were subject to static stress, slope, and vertical natural

frequency constraints. As discussed in Chapters 4 and 5, bridges that do not meet vertical

natural frequency limits (AASHTO 2009) may be considered serviceable (comfortable to

walk across) if their accelerations are below pedestrian comfort limits (Sétra 2006).

Consequently, accelerations in addition to natural frequencies are key dynamic criteria

that can influence the required cross-sectional area and rope prestress for a polyester-rope

suspended footbridge to be considered serviceable. This chapter extends the multi-

objective optimization study performed in Chapter 6 by considering the same objective

110
)

functions, volume and span (which are minimized and maximized, respectively) while

including acceleration and dynamic stress constraints in addition to static and natural

frequency constraints. The objective of this chapter is to establish how the inclusion of

these additional dynamic constraints alters the minimum volume designs found in

Chapter 6 for two-dimensional unstayed polyester-rope bridges with prestress. The

optimization methodology presented in Chapter 6 is modified to include modal

calculations to perform the dynamic pedestrian excitation analysis used to evaluate the

additional dynamic constraints.

7.2 Structural Model Features


This section summarizes the model configuration, material properties, and design loads

for the structural models evaluated in the optimization problem.

7.2.1 Model Configuration

The study only includes the two-dimensional unstayed configuration evaluated in Chapter

6. This bridge is assumed to have a 1 m wide deck that contributes to the mass, but not

the stiffness of the system. Figure 7.1 shows a typical detailed unstayed configuration and

the simplified configuration that serves as its surrogate in the optimization problem.

Chapters 3 – 5 demonstrated the ability of similar surrogate models to approximate the

static and dynamic results of a more detailed configuration. For the numerical analysis

of the model, the suspended rope is discretized into forty-eight elements. This increase

from the twelve element discretization in Chapter 6 is done to improve the smoothness of

111
)

the mode shapes utilized in the modal analysis while still balancing the accuracy of the

approximation with computational time.

a. Detailed configuration. b. Simplified configuration evaluated in


the optimization.
Figure 7.1. Elevations showing the two-dimensional unstayed bridge.

7.2.2 Material and System Properties

This study only considers polyester rope systems. The material properties follow from

those used in Chapter 6 for polyester rope. These properties are used to enable

comparisons between the results in this chapter and Chapter 6. The “modulus of

elasticity” is taken to be the lower-bound value (El), namely 2.86 GPa. The rope’s

ultimate stress (σult) is 471.3 MPa.

Damping is a required system parameter in the pedestrian excitation analysis that is

used to calculate accelerations and dynamic stresses. Damping ratios (ξ) of 2% and 6%

are considered to evaluate the impact of damping on the minimum volume results. These

damping values correspond approximately with the lower and upper bounds for the

damping ratio range found in Chapter 4 for the case study bridge (2.6% to 6.1%). Since a

system’s damping can only be determined experimentally after its construction (Sétra

2006), it is not guaranteed that when built, the bridge designs found in the optimization

will have the amount of damping assumed. The deck, suspender arrangement, and mesh

112
)

configuration are among the components that are not included in the bridge systems

optimized, but that can influence the bridge’s damping.

7.2.3 Load Definitions

This study considers prestress, dead (DL), and live (LL) loads. Prestress strains vary in

the analysis. Refer to Table 7.1 in Section 7.3.5 for the strain ranges. The DL includes the

rope’s self-weight, a 0.53 kN/m uniformly distributed load representative of bridge

components other than the suspended ropes, and potentially additional superimposed DL.

Refer to Table 7.1 in Section 7.3.5 for the superimposed DL range. Superimposed DL is

considered because of the potential for supplemental mass to reduce bridge accelerations

(Sétra 2006).

Two static uniformly distributed LLs are considered. The first load (LL1) is 4.3 kN/m.

This is the LL value used in Chapter 6 and follows from the unreduced area load of 4.3

kN/m2 presented in the American Association of State and Highway Transportation

Officials (AASTHO) guidelines (2009) applied to the 1 m wide deck. LL1 is used when

checking the static stress and slope constraints described in Section 7.3.3. The second

load (LL2) is 0.7 kN/m and comes from 0.7 kN/m2 presented in the Service d’Études

Techniques des Routes et Autoroutes (Sétra) technical guide (2006) applied to the 1 m

wide deck. LL2 is used when checking the dynamic constraints described in Section

7.3.4.

113
)

The working stress load combinations (LCs) considered in the static and/or natural

frequency analyses are:

LC1: prestress + DL

LC2: prestress + DL + LL1

LC3: prestress + DL + LL2

In addition to static loads, a dynamic live load similar to one used in Chapter 5 is

included in this optimization study. The dynamic live load follows from Sétra’s dynamic

analysis methodology for a Class III footbridge (Sétra 2006) and is defined as:


wv = 1512cos(2πft ) N/m (7.1)
L

where f is natural frequency for the mode under consideration, ξ is the damping ratio, t is

time, and L is span (with unit m). Equation C.1 in Appendix C presents the original

expression as presented in Sétra (2006) from which this load was derived.

The working stress load combinations (LCs) used to examine the combined effect of

the static and dynamic loads in the dynamic analysis are:

LC4: prestress + DL + wv

LC5: prestress + DL + LL2 + wv

7.3 Optimization Problems


This section presents the two multi-objective problems (iii and iv) under investigation.

The problem and case numbering in this chapter continues from the numbering used in

Chapter 6. Multi-objective optimization problems i and ii and cases d1 – d8 were

considered in Chapter 6.

114
)

Problem iii includes a single case (d9) subject to static and natural frequency

constraints. Problem iv differs from problem iii in that dynamic acceleration and stress

constraints are also considered and consequently, damping must be incorporated into the

analysis. Two cases (d10 and d11) are included in problem iv. The only difference

between these two cases is the damping ratio used. Cases d10 and d11 are assumed to

have 2% and 6% damping (ξ), respectively, in all of their vertical modes. In the

remainder of this section, the problem formulations are described and then the objective

functions, constraints, and design variables are discussed.

7.3.1 Problem Formulations

The problem formulation for problem iii is defined as:

f (x) = − L
minx ⎧⎨ 1 (7.2)
⎩ f 2 (x) = V

⎧g1 (x) = max{σ i } − σ all ≤ 0


⎪ i =1, ... , n
⎪⎪g 2 (x) = max{si }− sall ≤ 0
subject to: ⎨ i =1, ... , n
⎪g (x) = λ − λ ≤ 0
⎪ 3 min all

⎪⎩x = [L A1 ε 1 ]

115
)

The problem formulation for problem iv is defined as:

f (x) = − L
minx ⎧⎨ 1 (7.3)
⎩ f 2 (x) = V


⎪ g 1 (x) = max{σ i } − σ all ≤ 0
⎪ i =1, ... , n
⎪ g 2 (x) = max{si }− sall ≤ 0
⎪ i =1, ... , n

⎪ g 3 (x) or the following three constraints : g 4 (x), g 5 (x), and g 6 (x)
⎪where :

subject to: ⎨ g 3 (x) = λmin − λall ≤ 0
⎪ g (x) = max{a } − a ≤ 0
⎪ 4
j =1, ... , n +1
j all


⎪ g 5 (x) = max{σ sdi } − σ all ≤ 0
⎪ i =1, ... , n

⎪ g (x) = 10 −10 − min{σ } ≤ 0


⎪ 6
i =1, ... , n
sdi


⎩x = [L A1 ε 1 qSDL ]

In problem formulations 7.2and 7.3, x is the design variable vector, f1(x) and f2(x) are

the objective functions, and g1(x), g2(x), … are the constraints. The remaining variables

in equations 7.2 and 7.3 are defined and described in Sections 7.3.2 – 7.3.5.

7.3.2 Objective Functions

In both problems, the two objective functions are the same as those considered in Chapter

6. The first objective function is span (L) and is defined as: f1(x) = -L. The negative sign

is included in the function because span is being maximized, but the general problem

formulations are defined as minimizations. The second objective function is rope volume

(V) and is defined as f2(x) = V. The aim is to minimize volume.

116
)

7.3.3 Static Constraints

In both problems, there are two static constraints g1(x) and g2(x). These are the same two

static constraints considered in Chapter 6. The first constraint is that the maximum

service stress along the suspended rope (max{σi} for i =1, … , n) does not exceed the

allowable stress (σall):

g1 (x) = max{σ i } − σ all ≤ 0 (7.4)


i =1,..., n

where n is the number of elements comprising the suspended rope and the allowable

stress is σult/FS. σult is the ultimate stress defined in Section 7.2.2, namely 471.3 MPa and

FS is the assumed factor of safety, 2.2, introduced in Chapter 6. The second constraint is

that the maximum suspended rope slope (max{si} for i =1, …, n) does not exceed the

allowable walking slope (sall) of 20% introduced in Chapter 6:

g2 (x) = max{si } − sall ≤ 0 (7.5)


i =1,..., n

7.3.4 Dynamic Constraints

For problem iii, the only dynamic constraint is g3(x):

g3(x) = λmin – λall ≤ 0 (7.6)

where λmin is the fundamental vertical natural frequency and λall is the limit on the natural

frequency. In this study, the limit is 2.1 Hz. This value corresponds to the criterion

presented in the Sétra’s technical guide (2006) as the upper limit for greatest risk of

achieving resonance in the vertical direction due to pedestrian excitation. This differs

from the natural frequency constraint presented in Chapter 6 which followed from the

117
)

3 Hz criterion specified in the American Association of State Highway and

Transportation Official (AASHTO) guidelines (2009).

For problem iv the natural frequency constraint, g3(x), is also included. However, if

this constraint is violated, the design under evaluation is still considered feasible if a set

of additional dynamic constraints (g4(x), g5(x), and g6(x)) evaluated in a pedestrian

excitation analysis are satisfied.

The first of these additional dynamic constraints (g4(x)) is that the suspended rope’s

maximum acceleration (max{aj} for j =1, …, n+1) does not exceed the allowable

acceleration (aall):

g4 (x) = max{a j } − aall ≤ 0 (7.7)


j =1, ... , n+1

The maximum allowable acceleration used in this study is 0.25 g. This value corresponds

to Sétra’s criterion for minimum pedestrian comfort limit in the vertical direction (Sétra

2006).

The next dynamic constraint (g5(x)) is that the maximum suspended rope stress under

the superposition of dynamic and static loads (max{σsdi} for i =1, …, n) does not exceed

the allowable stress (σall) defined in Section 7.3.3:

g 5 (x) = max{σ sdi } − σ all ≤ 0 (7.8)


i =1, ... , n

Section 5.3.3.1 in Chapter 5 described how there is no explicit check that the ropes

remain in tension during the superposition of dynamic and static loads. Therefore, an

additional dynamic constraint (g6(x)) is included to ensure that the suspended rope has

not gone slack (i.e., zero stress) or into compression. This constraint is defined such that

118
)

that the minimum suspended rope stress (min{σsdi} for i =1, …, n) is not less than 10-10

(i.e., a value approaching zero):

g6 (x) = 10−10 − min{σ sdi} ≤ 0 (7.9)


i=1, ..., n

7.3.5 Design Variables

In both problems, the design variables (x) are continuous. The common design variables

in the problems are span (L), suspended rope area (A1), and suspended rope prestress

strain (ε1). The ranges for these variables follow from those used in Chapter 6 and are

presented in Table 7.1. Additionally, in problem iv, superimposed DL (qSDL) is a design

variable with qSDL ∈[0, 10] kN/m. Superimposed DL is included to evaluate the impact of

using supplemental mass to reduce bridge accelerations below acceptable limits and

decrease the required rope volume. The lower superimposed DL limit is set at 0, which

indicates that no superimposed DL is applied to the bridge. The upper superimposed DL

limit is set to a value that is not reached during the optimization and therefore will not

inhibit the discovery of the optimal designs.

Table 7.1. Design variable ranges for the optimization problems.


Design variable Range
L [15, 64] m
A1 [0.01, 100] cm2
ε1 [10-8, 7.49x10-2]
qs [0, 10] kN/m

7.4 Optimization Methodology


This section describes the optimization methodology used to evaluate multi-objective

problems iii and iv.

119
)

7.4.1 Algorithm Summary

The novel methodology described in Chapter 6 is adapted for this study by modifying the

structural analysis algorithms used to evaluate the designs created in the optimization

routine. Static and natural frequency computations are still performed using dynamic

relaxation and eigenanalysis algorithms. Chapters 3 and 4 introduced these static and

natural frequency calculations, respectively. The lumped mass matrix used in the

eigenanlyses performed in this chapter is similar to the one utilized in Chapters 4 and 5.

The mass contributions in the longitudinal directions are significantly reduced to ensure

that the modes found are predominately in the vertical direction. In this chapter, the mass

values in the longitudinal direction are set to 0.1% of the values in the vertical direction.

The dynamic pedestrian excitation analysis is executed using modal calculations

introduced in Chapter 5.

The optimization and performance measurement algorithms described in Chapter 6

are still utilized in this chapter. The optimization uses a non-dominated sorting genetic

algorithm, NSGA-II (Deb et al. 2002) with the genetic parameters listed in Table 7.2 to

find a non-dominated solution set. For each case under consideration, the optimization

algorithm is run ten times. Then, the S-metric performance measure (Zitzler 1999) is used

to compare non-dominated solution sets to determine which run is producing the non-

dominated set that is closest to the true Pareto-optimal set.

Table 7.2. Genetic parameters for the optimization problems.


Parameter Value
Number of generations 45
Population size 100
Simulated binary crossover probability 0.75
Simulated binary mutation probability 0.2
Simulated binary crossover distribution index 10
Simulated binary mutation distribution index 20

120
)

7.4.2 Objective Function Calculations

In all cases, both objective functions are determined directly from the design variable

vector x, rather than a load combination.

7.4.3 Constraint Calculations

Constraints are calculated using the load combinations (LCs) defined in Section 7.2.3.

Table 7.3 shows which LC is used to determine the constraint values in each case.

Table 7.3. Load combinations (LCs) used in the structural analysis to determine the
constraint values.
Static constraints Dynamic constraints
Problem Case
g1(x) g2(x) g3(x) g4(x) g5(x) g6(x)

iii d9 LC2 LC2 LC1, LC3 - - -

iv d10, d11 LC2 LC2 LC1, LC3 LC4, LC5 LC4, LC5 LC4, LC5

Static stress (g1(x)) and slope (g2(x)) constraint values are calculated using LC2 in a

static analysis. By inspection, the LC2 stresses and slope are greater than the results for

the other static load combinations, LC1 and LC3.

The natural frequency constraint (g3(x)) in both problems and the additional dynamic

constraints (g4(x), g5(x), and g6(x)) required in problem iv if the natural frequency

constraint is not satisfied, are evaluated using Sétra’s dynamic analysis methodology for

a Class III footbridge (Sétra 2006). The natural frequency calculations are performed

using LC1 and LC3. Values for the acceleration (g4(x)), and dynamic stress (g5(x), and

g6(x)) constraints are computed using LC4 and LC5 in a pedestrian excitation analysis.

These dynamic calculations are repeated for every bridge mode with a vertical natural

121
)

frequency between 1.7 Hz and 2.1 Hz. For a design to be feasible all constraints must be

met for all the modes evaluated with LC4 and LC5.

A bridge whose fundamental frequency is less than 2.1 Hz (i.e., g3(x) is not met), but

whose other frequencies fall outside the range of 1.7 Hz to 2.1 Hz, is not at maximum

risk of experiencing resonance and per Sétra’s (2006) methodology, does not require a

pedestrian excitation analysis to evaluate its dynamic performance. However, in this

study at least one mode found using LC4 and LC5 is always checked when the

fundamental natural frequency is less than 2.1 Hz. Performing a dynamic analysis with at

least one mode is conservative, but ensures that if a bridge’s natural frequencies shift

slightly between the numerical predictions and the built structure, that pedestrian

vibration issues have still been addressed. For the scenario when the fundamental

frequency is less than 2.1 Hz, but when none of the natural frequencies are in the range of

1.7 Hz to 2.1 Hz, the mode with the natural frequency closest to the range is used in the

calculations.

7.5 Results and Discussion


This section presents and discusses the results for the cases evaluated in optimization

problems iii and iv. Table 7.4 includes the mean value (μs), standard deviation (σs), and

maximum value (ms) results for the S-metric computations.

Table 7.4. S-metric results for the cases evaluated in problems iii and iv.
Case μs (cm4) σs/μs (%) ms/μs – 1 (%)
d9 2.75x108 0.47 0.47
d10 1.68x109 2.41 2.42
d11 1.47x109 1.14 1.23

122
)

In Table 7.4, σs/μs ranges from 0.47% to 2.41% and ms/μs – 1 ranges from 0.47% to

2.42%. These results suggest that the spread of solutions between the ten runs performed

for each case is typically low. For each of the three cases in Table 7.4, the non-dominated

set for the run with the maximum S-metric value is plotted in Figure 7.2. Additionally,

the Figure 7.2 plot includes the unstayed, prestressed polyester-rope bridge cases

investigated in Chapter 6 to enable comparisons between cases subject to additional sets

of constraints.

5
x 10
5
d1 − static constraints
4.5 d5 − static + AASHTO’s natural frequency constraints
/
d9 − static + Setra’s natural frequency constraints
4 /
d10 − static + Setra’s complete dynamic constraints with ξ = 2%
/

3.5 d11 − static + Setra’s complete dynamic constraints with ξ = 6%

3
f2 (cm3)

2.5

1.5

0.5

0
−65 −60 −55 −50 −45 −40 −35 −30 −25 −20 −15
f1 (m)

Figure 7.2. Volume versus span results for unstayed, prestressed polyester-rope bridge
cases optimized in problems i - iv. Black markers are used for the cases evaluated in this
chapter (d9 - d11). Red markers are used for the cases introduced in Chapter 6 (d1 and
d5).

The Chapter 6 cases included in the Figure 7.2 plot are d1 and d5. Case d1 was

subjected to only static constraints (g1(x) and g2(x)). Case d5 was subjected to static

constraints (g1(x) and g2(x)) as well as a natural frequency constraint whose limit was

123
)

AASHTO’s (2009) 3 Hz criterion under prestress + DL. In the Figure 7.2 plot, the spans

are negative values to coincide with the objective function definition described in Section

7.3.2.

Figure 7.2 reiterates what was shown in Chapter 6. For case d5, the inclusion of the

AASHTO (2009) natural frequency criterion leads to solutions with much greater

volumes than are needed for systems subjected to just static constraints (case d1).

Designs for case d5 approach the upper area limit specified in Section 7.3.5 (i.e., 100

cm2) at spans less than 64 m. These high volume solutions are found because significant

prestress (strains are approaching the 0.0749 upper limit specified in Section 7.3.5) is

required to meet the natural frequency constraint. Figure 7.2 shows that case d9 nearly

aligns with case d5. Case d9 has a lower natural frequency limit (i.e., Sétra’s (2006) 2.1

Hz criterion rather than AASHTO’s (2009) 3 Hz limit), but must meet this limit under

LC3, a load combination that includes LL and consequently has greater mass than the

load combination used in case d5.

Figure 7.2 shows that cases d10 and d11 dominate cases d5 and d9. This indicates

that when dynamic constraints must be considered, meeting acceleration and dynamic

stress constraints in lieu of a natural frequency constraint, produces lower volume

designs. Results at a span of approximately 29.4 m are presented in Table 7.5 to provide

an example of the differences in volumes among the cases evaluated that have dynamic

constraints (d5, d9, d10, and d11). The cases whose only dynamic constraint is natural

frequency have approximately the same volume, 2.69x105 cm3 and 2.68x105 cm3 for

cases d5 and d9, respectively. Including the acceleration and dynamic stress constraints

124
)

and assuming damping ratios (ξ) of 2% (case d10) and 6% reduces the volume by

approximately 56%, and 67%, respectively.

Table 7.5. Volume results for solutions at a span of approximately 29.4 m for the cases
with dynamic constraints.
Case V (cm3)
d5 – static + AASHTO’s natural frequency constraints 2.69x105
d9 – Static + Sétra’s natural frequency constraints 2.68x105
d10 – Static + Sétra’s complete dynamic constraints with ξ = 2% 1.17x105
d11 – Static + Sétra’s complete dynamic constraints with ξ = 6% 8.75x104

Figure 7.2 also shows that both cases d10 and d11 are able to extend across the span

range considered with areas less than the 100 cm2 upper limit. In these cases, the natural

frequency constraint is never satisfied and instead the acceleration under LC4 is the

active dynamic constraint. Acceleration under LC5 is not active because that load

combination includes greater mass than LC4 which aids in reducing accelerations.

Dynamic stresses are not active. This was anticipated based on the results for the case

study bridge analyzed in chapter 5. While the natural frequency constraint is not satisfied

for the span range considered, the results (Figure 7.2) for cases d10 and d11 are

converging with the results for cases d5 and d9 near a span of 15 m. This convergence

suggests that for bridges with spans less than 15 m, satisfying the natural frequency

constraint may result in lower volume solutions than solutions found by satisfying the

acceleration and dynamic stress constraints.

To meet the acceleration constraint, superimposed dead load is required. The

superimposed dead load ranges found are [4.12, 9.67] kN/m and [2.15, 5.79] kN/m for

cases d10 and d11, respectively. In both cases, greater uniformly distributed

superimposed dead loads are required at shorter spans. Examining the dynamic live load

used in the analysis and returning to the idea presented in Chapter 4 and 5 that the

125
)

behavior of a polyester-rope suspended footbridge is approaching that of a taut string,

demonstrates the observed relationship between superimposed dead load and span.

Equation 7.1 indicates that the dynamic live load magnitude (wv) from Sétra’s dynamic

analysis methodology is inversely proportional to the square root of the span ( L ).

Equation C.19 in Appendix C shows that for a structure whose behavior is approaching

that of a taut string the maximum acceleration is proportional to the maximum amplitude

of wv (defined as u) and inversely proportional to the system’s linear mass (m):

u
a max ∝ (7.10)
m

Consequently, for polyester-rope suspended bridges analyzed with Sétra’s dynamic

analysis methodology (Sétra 2006), the relationship between the maximum acceleration

(amax), span (L), and linear mass (m) is:

1
amax ∝ (7.11)
Lm

This expression shows that to have the same accelerations at two spans, a greater linear

mass is required at the shorter span. Superimposed dead load is used to provide this

greater linear mass.

Less superimposed dead load is required for case d11 than case d10 to meet the

acceleration constraint because the former has greater damping than the latter (ξ = 6%

versus ξ = 2%). These lower superimposed dead loads in case d11 lead to lower total

loads used in the static calculations for case d11 than case d10. Consequently, the case

d11 designs require less area than and dominates case d10.

Similarly, the primary difference between case d1 and cases d10 and d11 that leads to

case d1 having lower volume solutions is the lower total load used in the static

126
)

calculations. The static analysis for case d1 does not include any superimposed DL. In

cases d1, d10, and d11, the static stress and slope constraints are active. The prestress

strain ranges required for these cases to meet the static constraints are similar to the range

found for case d1. These ranges are presented in Table 7.6.

Table 7.6. Prestress stain ranges cases d1, d10, and d11.
Case ε1
d1 – Static constraints [0.0630, 0.0670]
d10 – Static + Sétra’s complete dynamic constraints with ξ = 2% [0.0628, 0.0684]
d11 – Static + Sétra’s complete dynamic constraints with ξ = 6% [0.0623, 0.0685]

7.6 Conclusions
This chapter utilized a set of multi-objective optimization problems to establish that when

an unstayed polyester-rope suspended footbridge must meet dynamic criteria related to

vertical pedestrian excitations, acceleration and dynamic stress constraints should be

considered in conjunction with natural frequency limits. Including acceleration and

dynamic stress constraints that can be satisfied in lieu of a natural frequency constraint

led to lower rope volume solutions for medium-span (15 m to 64 m) bridges. The

problems were evaluated using a modified version of the optimization methodology

introduced in Chapter 6. In this chapter, the methodology was extended to include modal

calculations for performing the pedestrian excitation analysis used to determine the

acceleration and dynamic stress constraints.

The acceleration constraint was found to be the controlling dynamic constraint. For

designs to meet the acceleration limit, superimposed dead load was required. Adding

load to the structure to decrease rope volume initially seems counterintuitive from the

perspective of a global minimization of materials. However, this strategy has two

127
)

benefits. First, lower rope volumes are accompanied by lower ultimate tension forces.

Therefore, connection, anchorage, and foundation requirements decrease as rope volume

decreases. Second, it may be possible to source the superimposed dead load from natural

materials available near the bridge. In these instances, minimizing the superimposed dead

load will not be critical from the perspective of transport. Meanwhile, the associated

reduction in rope volume can ease transport requirements.

128
)

Chapter 8

Conclusions and Future Work

8.1 Research Contributions


The goals of this dissertation were to:

i. Characterize the static and dynamic behavior of polyester-rope suspended

footbridges in the in-plane vertical direction.

ii. Establish the key design criteria and system parameters for polyester-rope

suspended footbridges.

This research was performed because there is limited design guidance available for

polyester-rope suspended footbridges. The goals were accomplished by combining

numerical and analytical computations, full-scale physical testing, and multi-objective

optimization techniques. This dissertation’s contributions to the research on polyester-

rope suspended footbridges are summarized in the remainder of this section.

8.1.1 Static and Dynamic Bridge Behavior

This dissertation investigated the impact of material stiffness on the in-plane vertical

static and dynamic behavior of polyester-rope suspended footbridges. Polyester rope in

129
)

general has low material stiffness. Lower-bound and upper-bound approximations of the

material stiffness for the 12-strand commercially available product considered in this

dissertation were 2.86 GPa and 5.96 GPa, respectively. By comparison, steel rope has a

modulus of elasticity of 196.5 GPa. Consequently, polyester-rope suspended footbridges

have greater flexibility than steel-rope structures. For the static analyses performed in this

dissertation, this high flexibility led to large deformations. In some circumstances it may

be desirable to use these deformations as a way to limit the live load on the bridge by

dissuading pedestrians from walking across the structure until those currently on the

bridge have finished crossing. In other scenarios, deformations may need to be limited to

comply with slope requirements. When trying to meet slope limits, applying additional

rope prestress rather than increasing the ropes’ cross-sectional area leads to a lower

volume design.

The large deformations that arise due to the flexibility of polyester-rope systems also

result in significant geometric nonlinear behavior; the systems undergo geometric

stiffening as they are loaded. This behavior leads to the bridge being able to support high

levels of live load (LL) beyond the service LL. For minimum volume, 64 m span

polyester-rope and steel-rope unstayed bridges (optimized with static stress and slope

constraints) the LL overload factors (ratio of applied LL-to-service LL) at their strength

limits were 7.1 and 3.05, respectively. Consequently, one advantage that the more

flexible polyester-rope system offers over the steel-rope system is greater overload

resistance. High levels of overload resistance are appealing for bridges built in rural areas

where maintenance may be inconsistent. Additionally, without established factors of

safety for polyester-rope bridges, overload resistance can provide an alternative measure

130
)

of safety that designers can consider when evaluating proposed polyester-rope suspended

footbridges.

Natural frequency, acceleration, and rope tension results found in analyses using the

lower and upper bound modulus of elasticity values were similar. These results suggest

that natural frequency and dynamic calculations can neglect polyester rope’s nonlinear

stress-strain behavior and utilize a single modulus of elasticity value.

This dissertation also showed that simple, single suspended rope and taut string

models that can be evaluated analytically have potential to approximate the static and

dynamic behavior, respectively, of more complex polyester-rope suspended bridge

models. These surrogate models were less computationally expensive to analyze than

complex models that require numerical calculations. Consequently, these surrogate

models may be beneficial to designers during the conceptual design of a footbridge.

8.1.2 Damping Characteristics

This dissertation presented the first set of damping ratios available for polyester-rope

suspended footbridges. Prior to this dissertation, damping characteristics for these

structures were not available in the literature. A set of free vibration tests was performed

on the Ait Bayoud case study bridge to determine the structure’s modal damping. The

damping ratio ranges found for the first three modes identified in the tests were

[2.6, 5.2] %, [3.0, 4.0] %, and [2.8, 6.1] %, respectively. These values can provide a

starting point for designers performing dynamic analyses on polyester-rope suspended

footbridges.

131
)

8.1.3 Key Design Criteria and System Parameters

This dissertation presented a novel multi-objective optimization methodology that

combined a non-dominated sorting genetic algorithm for searching the design space with

dynamic relaxation, eigenanalysis, and modal analysis algorithms for the static and

dynamic analyses. The methodology was applied to a set of suspended footbridge

problems in which volume was minimized and span was maximized while constraints

and system configurations varied. Including span as an objective function that was in

competition with the volume objective function allowed minimum volume designs to be

generated across the medium span range (15 m to 64 m). The plots of non-dominated

solution sets were provided as functions of span and can be used as visual design aids to

compare the bridge requirements for different systems subject to various constraints.

The optimization results indicated that while the steel-rope suspended footbridges did

not require prestress to meet static stress and slope constraints, the polyester-rope

suspended footbridges needed to be prestressed to meet these criteria. The resulting

polyester-rope bridges had higher volumes, but lower self-weights than the steel-rope

bridges. For example, at a span of 64 m, the polyester-rope bridge had a volume that was

3.8 times greater and a self-weight that was 1.5 times less than the steel-rope bridge

volume and self-weight, respectively. Low weight eases material transport and handling,

which is especially crucial in rural areas where materials may have to travel long

distances.

With the addition of vertical natural frequency constraints polyester-rope suspended

bridges required even greater levels of prestress. In these cases, the prestress strain

approached the allowable strain limit of 0.0749, indicating that prestress was the

132
)

dominant load for these structures. The inclusion of a vertical below-deck stay at midspan

reduced the material required to meet the constraints. For example, at a span of

approximately 29.4 m (the longest span achievable with the unstayed configuration given

the optimization problem specifications) the stay decreased the total required rope

volume by 51%. Inclined stay arrangements will lead to total system rope volumes that

fall between the volumes of the unstayed and vertically stayed configurations.

Considering dynamic acceleration and stress constraints that can be met instead of

natural frequency constraints resulted in lower volume designs. For example, the volume

of the 29.4 m unstayed configuration subject to only a natural frequency was reduced by

56% when the additional dynamic acceleration and stress constraints were included and

2% damping was assumed. This volume reduction increased to 67% if 6% damping was

assumed. To achieve these rope volume reductions required the addition of supplemental

mass to the structure. While the supplemental mass increased the overall weight of the

structure, the material contributing this supplemental mass could be sourced locally in

rural areas and not add to transport costs. Additionally, lower rope volumes are correlated

to lower ultimate tension forces and consequently, lower demands on connections,

anchorages, and foundations.

Based on the results from the optimization studies, static stress, static slope, and

accelerations are the key criteria for medium span polyester-rope suspended footbridges.

The important system parameters for these structures are prestress, damping, and

superimposed mass.

133
)

8.2 Areas for Future Research


There are many opportunities to make future research contributions to the limited

available design guidelines for polyester-rope suspended footbridges. This section

describes three potential areas for exploration that extend the work presented in this

dissertation. These areas for future research are:

i. Further characterize polyester rope’s properties as they relate to suspended

footbridges.

ii. Extend the presented optimization methodology to include additional/alternative

objective functions, constraints, configurations, and/or robust measures

iii. Perform large scale-physical tests on built structures to establish additional

damping information and verify the numerically and/or analytically predicted

structural behavior.

8.2.1 Polyester-Rope Material Characterization for Suspended

Footbridges

After identifying polyester rope as the best synthetic rope material for medium-span

suspended footbridges in rural areas, this dissertation focused on the relationship between

material stiffness and global bridge behavior. Future work could investigate other

features of polyester rope and the impact of these features on suspended footbridge

design. One potential area of research involves examining which fatigue failure modes

are critical for polyester-rope suspended footbridges. Examples of fatigue failure modes

are tensile fatigue, internal abrasion, hysteresis heating, and axial compression fatigue

(McKenna et al. 2004). Tensile fatigue begins with the formation of a transverse crack

134
)

which under cyclical loading causes a fiber within the rope to fail in tension. Internal

abrasion can arise when adjacent fibers slide against each other. Hysteresis heating

occurs during load cycling because of energy dissipation (the loading and unloading

curves do not align). Axial compression fatigue can arise due to the repeated bending of

fibers within the rope. These individual fibers can buckle in compression during cyclic

loading even though globally the rope is under tension (McKenna et al. 2004).

Polyester rope comes in a variety of grades and constructions (e.g., braided, twisted,

and parallel-lay) so a range of rope products could be evaluated. These studies can

contribute to the development of appropriate factors of safety specific to polyester rope in

suspended footbridges.

8.2.2 Extensions to the Presented Optimization Methodology

The flexibility of the novel optimization methodology presented in this dissertation

allows for the exploration of large design spaces. This methodology can be adapted to

suspended bridge problems with other objective functions, constraints and/or

configurations to further characterize the static and dynamic behavior of polyester-rope

suspended footbridges. Additionally, measures of system robustness can be integrated

into the optimization methodology.

8.2.2.1 Alternative Objective Function

This dissertation considered volume and span objective functions simultaneously to

produce sets of minimum volume designs as a function of span. Cost is an objective

function that can be used as an alternative to volume to compare suspended rope systems.

In addition to direct material costs, future studies could account for material transport and

135
)

handling, as well as the influence of material on required connection, anchorage, and

foundation costs.

8.2.2.2 Additional Constraints

This dissertation focused on vertical in-plane behavior that was assessed with two-

dimensional models. The solutions found in the optimization problems considered in this

dissertation not only provide insight into key parameters for polyester-rope suspended

footbridges, but can serve as preliminary designs that inform the development of more

detailed systems. Traditional iterative design approaches can be used to translate the

preliminary two-dimensional designs into final three-dimensional designs. Analyses on

the three-dimensional structures should account for the lateral and torsional static and

dynamic effects that were not considered in this dissertation. These analyses should

include prestress, dead, live, wind, snow, thermal, and/or seismic loads, where applicable

for a given site.

The optimization study can also be expanded to include constraints related to stress

and serviceability limits that accompany the additional listed loads. Plotting the

minimum volume results of polyester-rope suspended footbridge systems subject to these

different constraints, as was done in Chapters 6 and 7, can provide a convenient visual

tool for assessing the impact of each constraint or combination of constraints on bridge

requirements at different spans.

Future studies that include pedestrian excitation constraints, could compare different

dynamic analysis methodologies/dynamic load models. In this dissertation, the dynamic

load determined with the dynamic analysis methodology presented in the Service

d’Études Techniques des Routes et Autoroutes (Sétra) technical guide (2006) was used.

136
)

Van Nimmen et al. (2014) compared the differences in dynamic results for a set of eight

footbridges that were found using Sétra’s dynamic analysis methodology (Sétra 2006)

and the methodology presented in HiVoSS, a European guideline (RFCS 2008).

Similarly, the HiVoSS dynamic analysis methodology and load model could be used as

an alternative to Sétra’s methodology and load model for evaluating polyester-rope

suspended footbridges.

8.2.2.3 Additional Configurations

This dissertation focused on unstayed and stayed suspended bridge configurations. The

stayed bridge consisted of a single vertical element at midspan. Bridges with inclined

stays connecting at midspan will have a stiffness that falls between the low-stiffness

unstayed bridge and the high-stiffness vertically stayed bridge. Therefore, the

configurations examined in this dissertation bound the range of minimum volume designs

possible for stays that connect at midspan. However, additional configurations consisting

of multiple stays or rope trusses oriented to provide stiffness in vertical and/or lateral

directions could be considered.

8.2.2.4 Robust Optimization

Robust design is important because all structures are characterized by random

uncertainties on geometric and physical properties that may arise during manufacturing,

construction, etc. Park et al. (2006) and Beyer and Sendhoff (2007) provide detailed

reviews of robust design approaches. The author of this dissertation conducted an initial

study (Segal et al. 2014) in which robustness was considered in the optimization

methodology used to evaluate a polyester-rope suspended footbridge. The procedure

137
)

utilized to account for robustness in the optimization methodology was developed by the

third author of that initial study, Rajan D. Filomeno Coelho, and adapted from the

methodology presented by Filomeno Coelho et al. (2011) for the optimization of space

truss structures. In the study, robustness was defined as trying to reduce the sensitivity of

the final design to uncertainties on the model inputs. The author of this dissertation

applied the methodology to a bridge model consisting of a single suspended rope. Two

objective functions (volume which was minimized and natural frequency which was

maximized) and two static constraints (maximum stress and maximum slope were

limited) were considered simultaneously. Test cases in which the rope’s modulus of

elasticity and prestress were taken as random variables were examined. The non-

dominated sets of robust designs for all cases considered were found to be similar to the

non-dominated sets found when robustness was not considered. Consequently, robust

optimization did not make an impact in the investigated case study. However, the robust

optimization methodology can be adapted to include additional/alternative objective

functions, constraints, and configurations like the ones discussed in Sections 8.2.2.1 –

8.2.2.3 to assess when considering robustness does influence the required parameters for

suspended footbridges.

8.2.3 Large-Scale Physical Testing

This dissertation presented a set of large-scale free vibration tests performed on the Ait

Bayoud case study bridge to determine the structure’s damping ratios. Further dynamic

testing of the Ait Bayoud bridge as well as testing of new polyester-rope suspended

138
)

footbridge should to be conducted to provide additional information on the damping

characteristics of these structures that can be incorporate into numerical models.

Future work can also include additional static and/or dynamic testing to verify the

predicted numerical and/or analytical bridge response. For example, physical tests can be

utilized to confirm the pedestrian excitation response of polyester-rope suspended

footbridges that is expected based on the alternative dynamic analysis methodologies and

load models suggested in Section 8.2.2.2. Full-scale pedestrian excitation tests have been

performed on suspension (Dallard et al. 2001) and stress ribbon (Caetano and Cunha

2004; Magalhães et al. 2007; Caetano et al. 2011) footbridges.

Appendix D presents an initial set of pedestrian excitation tests performed on the Ait

Bayoud bridge to characterize the actual bridge accelerations under pedestrian loading

(Figure 8.1).

Figure 8.1. Polyester-rope suspended footbridge in Ait Bayoud, Morocco during a


pedestrian excitation test.

139
)

These tests were conducted with a single pedestrian and a group of four synchronized

pedestrians. Typical traffic on this specific bridge is closer to this load range than was

utilized in the Chapter 5 and 7 analyses. Differences between the experimental and

numerical loads preclude direct comparisons between observed and predicted

accelerations. Future pedestrian excitation tests should involve additional pedestrians to

enable these comparisons.

8.3 Summary
This chapter described this dissertation’s research contributions and potential areas for

future investigations related to medium-span polyester-rope suspended footbridges.

These contributions as well as additional research are crucial for establishing design

guidelines for these structures.

140
)

Appendix A

Comprehensive Analytical Static

Calculations for the Single Suspended

Rope Model Presented in Chapter 3

A.1 Introduction
This appendix presents the comprehensive analytical static calculations for the single

suspended rope model of the Ait Bayoud case study bridge presented in Chapter 3,

Section 3.3. These calculations are performed with Irvine’s flat cable theory (Irvine

1981). As noted in Chapter 3, Woodward (2012a), the designer of the Ait Bayoud

footbridge, performed similar analytical calculations to the ones presented in this

dissertation. This dissertation’s unique contribution is the comparison (Section 3.3) of the

analytical results (Section 3.3) to the results from numerical computations performed on a

detailed model of the case study bridge (Section 3.2).

141
)

A.2 Analytical Model


The model consists of a single suspended rope with pinned end supports (Figure A.1).

Figure A.1. Elevation of the single suspended rope model.

All ten 18 mm diameter (nominal) ropes used in the case study bridge are lumped

together in this single suspended rope. The rope has a 64 m span (L), 1.15 m sag (d)

under the combination of prestress and dead load, and 20.71 cm2 cross-sectional area (A).

The lower bound material stiffness case c1 introduced in Section 3.2.2.2 is investigated.

This case has an elastic modulus (E) of 2.86 GPa. The lumped strength limit (Tult) is 992

kN (group of ten ropes each with a limit of 99.2 kN).

The calculations include prestress, dead (DL), and service live (LL) loads. In Irvine’s

theory (Irvine 1981), prestress is accounted for indirectly. The sag (d) under the applied

DL is dependent on the amount of prestress that has been applied to the rope. The DL and

LL are uniformly distributed gravity loads with magnitudes of 0.56 kN/m (qDL) and 3.84

kN/m (qLL), respectively. Three load combinations (LCs) are considered in the analysis:

LC1: prestress

LC2: prestress + DL

LC3: prestress + DL + LL

142
)

A.3 Analytical Calculations


This section presents the calculations for LC1 – LC3. The calculations are performed to

determine the maximum rope tensions (T) and minimum factors of safety (FSmin) for each

load combination as well as the maximum LL deflection (δLL) under LC3.

In the calculations, numbered subscripts on variables correspond to the load

combination under investigation. For example, T1 and T2 are the rope tensions under LC1

and LC2, respectively. The calculations for LC2 are presented first because the results of

these calculations are utilized when evaluating LC1 and LC3.

A.3.1 Calculations for Load Combination 2

The free body diagram for the suspended rope under LC2 is shown in Figure A.2.

Figure A.2. Free body diagram of the suspended rope under LC2.

First, the horizontal reactions (H2) and vertical reactions (V2) are calculated using the

equations:

q DL L2
H2 = = 249.3 kN (A.1)
8d

qDL L
V2 = = 17.9 kN (A.2)
2

143
)

Next, the maximum rope tension (T2) is calculated using the horizontal and vertical

reactions:

T2 = H 22 + V22 = 249.9 kN (A.3)

The minimum rope factor of safety (FSmin2) is then calculated using the equation:

Tult
FSmin2 = = 3.97 (A.4)
T2

The rope strain (ε2) equivalent to the calculated tension is:

T2
ε2 = = 0.0422 (A.5)
EA

The stressed rope length (L2) is determined using the expression:

⎛ 8 ⎛ d ⎞ 2 32 ⎛ d ⎞ 4 ⎞
L2 ≅ L⎜1 + ⎜ ⎟ − ⎜ ⎟ ⎟ = 64.06 m (A.6)
⎜ 3⎝ L ⎠ 5 ⎝ L ⎠ ⎟⎠

A.3.2 Calculations for Load Combination 1

First, the initial unstressed length of the rope (Lo) is calculated using the LC2 stressed

rope length (L2) and strain (ε2) with the following equation:

L2
Lo = = 61.47 m (A.7)
1+ ε2

Next, the resulting prestress strain (εp) is determined using the initial unstressed (Lo)

and prestressed (Lp) rope lengths with the following expression:

L p − Lo
εp = = 0.0412 (A.8)
Lo

where Lp = L = 64 m.

144
)

The equivalent maximum rope tension (T1) and minimum factor of safety (FSmin1) for

LC1 are:

T1 = ε 1 EA = 244 kN (A.9)

Tult
FSmin1 = = 4.07 (A.10)
T1

A.3.3 Calculations for Load Combination 3

The free body diagram for the suspended rope under LC3 is shown in Figure A.3.

Figure A.3. Free body diagram of the suspended rope under LC3.

First, the length parameter (Le) is determined using the expression:

⎛ ⎛ d ⎞ ⎞⎟
2

Le ≅ L 1 + 8⎜ ⎟ = 64.17 m (A.11)
⎜ ⎝ L ⎠ ⎟⎠

Next, the system stiffness parameter (λ2) is calculated using the equation:

2
⎛ qDL L ⎞
⎜⎜ ⎟ L
⎝ H 2 ⎟⎠
λ =
2
= 0.49 (A.12)
⎛ H 2 Le ⎞
⎜ ⎟
⎝ EA ⎠

Then, the ratio of LL-to-DL (qr) is determined:

qLL
qr = = 6.86 (A.13)
qDL

145
)

Next, the dimensionless factor (hn) is found by solving the following cubic equation:

⎛ λ2 ⎞ ⎛ λ2 ⎞ λ2 ⎛ q ⎞
hn3 + ⎜⎜ 2 + ⎟⎟hn2 + ⎜⎜1 + ⎟⎟ hn − qr ⎜1 + r ⎟ = 0 (A.14)
⎝ 24 ⎠ ⎝ 12 ⎠ 12 ⎝ 2⎠

∴hn = 0.52

The LL horizontal reactions (HLL) are then calculated with the equation:

H LL = hn H2 = 129.6 kN (A.15)

Next, the LL vertical reactions (VLL) are found using the expression:

qLL L
VLL = = 122.9 kN (A.16)
2

Combining the LC2 and LL results, the LC3 horizontal reactions (H3), vertical

reactions (V3), and maximum rope tension (T3) are:

H3 = H2 + H LL = 378.9 kN (A.17)

V3 = V2 + VLL = 140.8 kN (A.18)

T3 = H 32 + V32 = 404.2 kN (A.19)

The minimum factor of safety (FSmin3) for LC3 is:

Tult
FS min3 = = 2.45 (A.20)
T3

The dimensionless factor (β) required to find the maximum (i.e. midspan) LL

deflection is calculated with the equation:

1 ⎛ hn ⎞
β= ⎜1 − ⎟ = 0.076 (A.21)
8(1 + hn ) ⎜⎝ qr ⎟⎠

146
)

The maximum LL deflection (δLL) is then determined with the expression:

β qLL L2
δ LL = = 4.79 m (A.22)
H2

A.4 Summary of Analytical Results

This section summarizes the results from the analytical calculations performed in

Appendix A. The maximum rope tensions and minimum factors of safety for LC1 – LC3

calculated in Section A.3 are summarized in Table A.1. The maximum LL deflection

(δLL) under LC3 calculated in Section A.3.3 is 4.79 m.

Table A.1. Maximum rope tension and minimum factor of safety results from the
analytical calculations.
LC T (kN) FS
1: prestress 244.0 4.07
2: prestress + DL 249.9 3.97
3: prestress + DL + LL 404.2 2.45

147
)

Appendix B

Accelerometer Alignment

B.1 Introduction
This appendix discusses the alignment of the accelerometers used in the full-scale

dynamic tests performed on the Ait Bayoud case study bridge that are described in

Chapter 4 and Appendix D. Specifically, this appendix first states why small angle

misalignments between the accelerometer and global axes may occur and then describes

why these misalignments should not significantly affect the acceleration results.

B.2 Discussion of Small Angle Misalignments


The accelerometers were not equipped with inclinometers that could record tilt

simultaneously with accelerations. Therefore, the acceleration results could not be

adjusted based on misalignment of the accelerometer’s axes with the global vertical,

lateral, and longitudinal directions. These misalignments include: (i) any errors in the

placement of the accelerometer on the bridge, (ii) bridge tilt when viewing the bridge

perpendicular to the span, and (iii) bridge slope in elevation. Additionally, no adjustments

were made for any further deviation between the accelerometer’s axes with the global

148
)

vertical, lateral, and longitudinal directions, due to bridge displacements that occurred

during testing.

The error introduced by not accounting for the bridge slope in elevation is described

to demonstrate the minor impact that small misalignments have on the results. The bridge

slopes in elevation at each of the six accelerometers introduced in Section 4.3.1 are listed

in Table B.1. Figure B.1 shows the relationship between accelerations in the

accelerometer’s axes (α and β) and the global axes (x and y).

Table B.1 Bridge slopes in elevation for the accelerometers.


Accelerometer θ (°) cos(θ) sin(θ)
a1 0.51 1 0.01
a2 0.28 1 0
a3a, a3b 2.32 1 0.04
a4 4.36 1 0.08
a5 4.90 1 0.09

Figure B.1. Elevation showing an accelerometer on a sloped bridge deck.

The equation to transform the accelerometer accelerations (aα and aβ) into the global

vertical acceleration (ay) is:

ay = aβ _ y + aα _ y = aβ cos(θ ) + aα sin(θ ) (B.1)

149
)

As indicated in Table B.1, for all the accelerometers, cos(θ) = 1 because the angle (θ) is

small. Therefore, equation B.1 simplifies to:

ay = aβ + aα sin(θ ) (B.2)

The percent error between what is assumed in this dissertation to be the vertical

acceleration (aβ) and the true global vertical acceleration (ay) is then:

a y − aβ aα sin(θ )
% error = x 100 = x 100 (B.3)
ay aβ + aα sin(θ )

Since aα should be much less than aβ for vertical excitations and sin(θ) is a maximum of

0.09 (from Table B.1) it is anticipated that aαsin(θ) << aβ and the percent error will be

small. The similarities between results at different accelerometers during free vibration

tests, such as those presented in Figure 4.8, indicate that these small bridge slopes are

indeed not greatly influencing the assumed and true vertical accelerations.

Similarly, the other small angle misalignments between the accelerometer’s axes and

the global vertical, lateral, and longitudinal directions should not significantly change the

results presented in this dissertation.

150
)

Appendix C

Equations for the Dynamic Analysis

Presented in Chapter 5

C.1 Introduction
This appendix presents supplementary equations referenced in the pedestrian excitation

dynamic analysis of the Ait Bayoud case study bridge described in Chapter 5. First, the

dynamic load calculations are described. Next, the equations used to calculate the modal

analysis steady-state displacements and accelerations are presented. Then, the

calculations used to determine the rope forces in the numerical calculations are described.

Finally, the taut string acceleration equation is presented.

C.2 Dynamic Load Calculations


This section presents the calculations for the dynamic load appearing in equation 5.1 in

Chapter 5. The calculation of the dynamic load follows from the methodology presented

in the Service d’Études Techniques des Routes et Autoroutes (Sétra) technical guide

151
)

(2006). For a Class III footbridge, the only dynamic load case that applies is Case 1

(Sétra 2006). The other dynamic load cases (Cases 2 and 3) account for denser crowds

and second harmonic effects that are not considered when assessing a Class III bridge.

The dynamic load (pv) corresponding to Case 1 is calculated using the equation:

ξ
pv = d (280N ) cos(2πft )(10.8) ψ N/m2 (C.1)
n

where d is the crowd density = 0.5 pedestrians/m2 for a Class III bridge, f is the natural

frequency under consideration, t is time, ξ is the damping ratio for the mode under

consideration, n is the number of pedestrians on the bridge, and ψ is a factor related to the

risk of resonance. For a Class III footbridge, ψ = 1. The number of unsynchronized

pedestrians on the bridge (n) is defined as:

n = dbL = 32.6 pedestrians (C.2)

where b is the bridge width (1.02 m) and L is the bridge span (64 m). Equation C.1 then

simplifies to:

pv = 264.7 cos(2πft) ξ N/m2 (C.3)

The equivalent uniformly distributed load (wv) is found by multiply equation C.3 by the

bridge width (b). For the case study bridge, the resulting expression is:

wv = 269.9 cos(2πft) ξ N/m (C.4)

Equation C.4 appears as equation 5.1 in Chapter 5.

152
)

C.3 Steady-State Displacement and Acceleration

Equations
This section presents the equations used for calculating the steady-state displacements

and accelerations in the modal analysis described in Section 5.3.2. Equations C.5 – C.8,

C.11, and C.13 follow from Sétra’s technical guide (2006).

The maximum (across time) steady-state displacement ( χ j ) and acceleration ( χ


&&j ) for

a single mode j are:

χj =
[φ ] [F ]
j
T
0
(C.5)
2ξ jω mj 2
j

χ&& j =
[φ ] [F ] = χ ω
j
T
0 2
j j (C.6)
2ξ j m j

where ξ j is the modal damping ratio, ωj is the angular natural frequency = 2πfj (where fj

is the natural frequency), [φj ] is the eigenvector, mj is the generalized modal mass, and

[F0] relates to the applied load function [F(t)]. The generalized modal mass is defined as:

mj = φj [ ] [M][φ ]
T
j (C.7)

where [M] is the lumped mass matrix. The load function is equivalent to the harmonic

function, wv(t) defined in equation C.4 and can be expressed as:

[F(t )] = [F0 ]cos(ω jt ) (C.8)

where the components of [F0] acting in the vertical direction at the nodes along the

below-deck rope ([F0v] ) are:

[F0v ] = lsu (C.9)

153
)

where ls is the tributary length contributing load to each node and u is the maximum

amplitude of wv(t). The maximum amplitude occurs when cos(2πft) = 1 and is defined as:

u = 269.9 ξ N/m (C.10)

The maximum displacement in mode j (qj) is transformed into a set of global total

displacements ([δ]) using the following equation (Sétra 2006):

[δ] = [δs ] ± [δd ] = [δs ] ± χ j [φ j ] (C.11)

where [δs] is the vector of static displacements found under just the application of the

static loads (refer to Chapter 5 for the static load combinations and Chapter 3 for a

description of the static calculation procedure) and [δd] is the vector of dynamic

displacements. Equation C.11 indicates that the static and dynamic displacements are

superimposed in two ways: (i) by adding the two displacements together and (ii) by

subtracting the dynamic displacement from the static displacements. Both combinations

are performed to account for the mode shapes either occurring as shown in Figures 5.2a,

5.3a, and 5.4a in Chapter 5 or being mirrored (in plane) about the deformed bridge

geometry and consequently having dynamic loads that are applied in the opposite

direction of those shown in Figures 5.1b, 5.2b, and 5.3b in Chapter 5. A simpler case that

helps illustrates the concept that a mode shape can have two possible orientations is the

first mode of a horizontal taut rope. The mode shape is a half sin curve that can be

indicated as occurring above or below the horizontal rope. The maximum combination of

the static and dynamic displacements will occur when the mode shape is oriented below

the horizontal rope. For more complex cases it may not be clear which of the two

possible orientations for a mode shape will lead to the greatest effect until the results for

both are calculated.

154
)

The following expression combines contributions from equations C.5 and C.11 to

show how the dynamic displacement vector [δd] is inversely proportional to the modal

2
frequency squared ( f j ):

1 1
[δd ] ∝ χ j ∝ ∝ (C.12)
ω 2
j f j2

The maximum acceleration in mode j ( χ


&&j ) is transformed into a set of global

accelerations ([a]) using the following equation:

[a] = χ&&j [φj ] (C.13)

This global acceleration vector represents the acceleration variation along the length of

the bridge. Since the variation is directly proportional to the eigenvector, the variation

will take on a shape similar to that of the mode shape. The maximum accelerations along

the length of the bridge are presented in Section 5.3.3.

The following expression combines contributions from equations C.6, C.9, C.10, and

C.13 to show how the global acceleration vector ([a]) is inversely proportional to the

square root of the damping ratio ( ξ ):

[a] ∝ χ&& j ∝ [F0 ] ∝ ξ



1
(C.14)
ξ ξ ξ

C.4 Rope Force Equations


This section presents the equations used for calculating the numerical rope force results

under the combination of static and dynamic loads. The global total displacements found

in equation C.11 are used to determine the final length (lf) of the rope elements or in the

155
)

case of the back stays and suspenders a rope cluster (refer to Chapter 3 for additional

details on rope clusters). This final rope length is then utilized to calculate the rope

elongations (e) using the equation:

e = l f − l0 (C.15)

where l0 is the initial element or cluster length. The elongations are then converted into

forces (T) with the expression:

eAE
T= (C.16)
l0

where A is the cross-sectional area and E is the modulus of elasticity of an element or

cluster.

C.5 Taut String Acceleration Equation


This section presents the equations used for calculating the maximum rope accelerations

of a taut string. While the Sétra technical guide (2006) does not provide an analytical

expression for a taut string, the equation presented in the guide for a simply supported

beam (equation C.19) can be shown to also apply for a taut string.

The general expression for the global acceleration vector [a] that results from

substituting equation C.6 into equation C.13 indicates that acceleration is only dependent

on a mode’s eigenvector ( [φj ] ), damping ratio (ξ j), and generalized mass (mj) as well as

the harmonic load amplitude ([F0]). This equation is:

[a ] =
[φ ] [F ] [φ ]
j
T
0
(C.17)
j
2ξ j m j

156
)

The eigenvectors for both a taut string and a simply supported beam are the same because

their mode shapes are both defined by the same equation (McConnell and Varoto 2008):

⎛ jπx ⎞
v j = sin⎜ ⎟ , for j = 1, 2, . . . (C.18)
⎝ L ⎠

where L is the rope span, x is a location along the span, and j is the mode number. The

remaining variables in equation C.17 are independent of whether the structure is a taut

string or a simply supported beam. Therefore, the following analytical expression for the

maximum acceleration of a simply supported beam (Sétra 2006) can also be used to

determine the maximum acceleration of a taut string:

2u
amax = (C.19)
ξ jπ m

where m is the linear mass for the load combination under consideration and u is the

maximum harmonic load amplitude defined in equation C.10. Substituting equation C.10

into equation C.19 and simplifying provides the expression:

539.8 N/m
amax = (C.20)
ξ jπ m

Equation C.20 appears as equation 5.6 in Chapter 5.

157
)

Appendix D

Initial Pedestrian Excitation Tests

Performed on the Ait Bayoud Bridge

D.1 Introduction
In Chapter 8, large-scale physical testing of polyester-rope suspended footbridges was

discussed as a way to verify numerical computations such as those performed in a

pedestrian excitation analysis. This appendix presents an initial set of pedestrian

excitation tests performed with a single pedestrian and a group of four synchronized

walkers on the Ait Bayoud case study bridge (introduced in Chapter 3). The objective of

the testing was to determine the bridge accelerations under typical traffic levels for this

specific bridge. Since these tests and the calculations performed on the bridge in Chapter

5 were conducted using different dynamic loads, direct comparisons between

experimental and numerical results are not appropriate. In future studies, the procedures

presented in this appendix can be implemented with additional pedestrians. These future

tests can be used to evaluate the extent to which numerical computations that utilize

158
)

dynamic analysis methodologies such as those presented in the Service d’Études

Techniques des Routes et Autoroutes (Sétra) technical guide (2006) and the HiVoSS

guideline (RFCS 2008) can predict the dynamic behavior of polyester-rope suspended

footbridges.

The remainder of the appendix is organized as follows. First, details for the

accelerometers used to record the bridge accelerations during testing are described.

Then, the single pedestrian and group pedestrian walking test procedures and results are

discussed. While this dissertation primarily focuses on in-plane vertical bridge behavior,

lateral in addition to vertical acceleration results are presented in this appendix. As

discussed in Chapter 8, characterizing the lateral dynamic response of polyester-rope

suspended bridges is a potential area of future research.

D.2 Accelerometer Details


The same accelerometers and accelerometer setup described in Chapter 4 for the ambient

and free vibration tests were used for this set of walking tests. The plan presented in

Chapter 4 that shows the accelerometer locations on the bridge is included again in

Figure D.1.

Figure D.1. Plan showing the accelerometer locations on the bridge. Note that the
drawing is not to scale.

159
)

Initial accelerometer deviations from zero are corrected by shifting the accelerations

based on average acceleration data found over 10 to 30 second intervals recorded before

testing. As discussed in Chapter 4, misalignments between the accelerometer and the

global axes may exist. Appendix B describes why the impact of these small angle

misalignments should not be significant.

D.3 Single Pedestrian Walking Tests


This section describes tests performed to characterize the response of the case study

bridge under a single pedestrian.

D.3.1 Procedure for Single Pedestrian Walking Tests

A set of eleven single pedestrian walking tests was performed. In each test, the pedestrian

walked approximately in-sync with a metronome set to the desired frequency. One test

was performed for each frequency that is a multiple of 0.25 Hz in the range, 0.5 Hz to

3Hz. This range extends to the American Association of State and Transportation

Officials’ (AASHTO) upper limit of 3 Hz for potential vibration problems in the vertical

direction (AASHTO 2009) and includes frequencies close to the numerically determined

vertical natural frequencies: 0.52 Hz, 1.03 Hz, and 1.54 Hz. These numerical frequencies

are the values determined using the lower-bound modulus of elasticity for polyester rope

(El = 2.86 GPa) in Chapter 4. As discussed in Chapter 4, the set of ambient vibration and

free vibration tests that were performed on the case study bridge uncovered values that

differ from the numerical predictions. The results of the ambient vibration and free

vibration tests were not post-processed until after returning from Morocco, so the

160
)

walking tests were not performed at the natural frequencies found experimentally

(approximately 0.68 Hz, 1.32 Hz, and 2.12 Hz) and therefore, did not capture the bridge

behavior during resonance in a vertical mode. Additionally, since the lateral natural

frequencies were not determined numerically or through testing prior to performing these

walking tests, there was no attempt to walk at frequencies that would knowingly capture

the bridge behavior during resonance in a lateral mode.

In the tests, the pedestrian typically began at one end of the bridge, walked to the

other end of the bridge, turned around, and walked back to midspan where he stopped

abruptly. For the 0.5 Hz and 0.75 Hz tests, the pedestrian stopped after only walking

across the bridge in one direction. Figure D.2 shows the pedestrian mid-stride while

walking across the bridge.

Figure D.2. Single pedestrian walking across the bridge.

D.3.2 Results and Discussion for Single Pedestrian Walking Tests

Table D.1 presents the maximum (absolute values) for lateral (H) and vertical (V)

accelerations among all of the accelerometers for each test. In all cases the maximum

accelerations exceed the minimum comfort limits specified in Sétra’s technical guide

(2006). These limits are 0.08 g and 0.25 g in the lateral and vertical directions,

respectively. Figure D.3 presents the acceleration time series for the accelerometers with

the maximum accelerations in tests in which the walking frequency is closest to the

161
)

experimentally found vertical natural frequencies. These tests are s2 (walking frequency

of 0.75 Hz is closest to 0.68 Hz), s4 (walking frequency of 1.25 Hz is closest to 1.32 Hz),

s7, and s8 (walking frequencies of 2.00 Hz and 2.25 Hz bracket 2.12 Hz). During this set

of tests the lateral and vertical comfort limits (represented with dashed lines in the Figure

D.3 plots) are typically exceeded at a particular accelerometer for only short durations of

time.

Table D.1. Results for single pedestrian walking tests.


Maximum |acceleration| (g), accelerometer
Test Walking frequency (Hz)
H V
s1 0.50 0.39, a2 1.02, a2
s2 0.75 0.39, a2 0.81, a2
s3 1.00 0.34, a2 0.87, a4
s4 1.25 0.53, a2 0.96, a4
s5 1.50 0.27, a4 1.12, a4
s6 1.75 0.34, a5 0.92, a5
s7 2.00 0.42, a5 1.22, a2
s8 2.25 0.64, a2 1.57, a5
s9 2.50 0.65, a2 1.29, a4
s10 2.75 1.41, a2 2.28, a5
s11 3.00 0.86, a2 2.55, a1

162
)

0.6 1.5
0.4 1
H (g) 0.2 0.5

V (g)
0 0
−0.2 −0.5
−0.4 −1
−0.6 −1.5
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time (s) Time (s)

a. Test s2, accelerometer a2. e. Test s2, accelerometer a2.

0.6 1.5
0.4 1
0.2 0.5
H (g)

V (g)
0 0
−0.2 −0.5
−0.4 −1
−0.6 −1.5
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time (s) Time (s)

b. Test s4, accelerometer a2. f. Test s4, accelerometer a4.

0.6 1.5
0.4 1
0.2 0.5
H (g)

V (g)

0 0
−0.2 −0.5
−0.4 −1
−0.6 −1.5
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time (s) Time (s)

c. Test s7, accelerometer a5. g. Test s7, accelerometer a2.

0.6 1.5
0.4 1
0.2 0.5
H (g)

V (g)

0 0
−0.2 −0.5
−0.4 −1
−0.6 −1.5
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time (s) Time (s)

d. Test s8, accelerometer a2. h. Test s8, accelerometer a5.


Figure D.3. Acceleration time series for the lateral (a – d) and vertical (e – h)
accelerations for a set of the single pedestrian walking tests.

163
)

D.4 Group Walking Tests


This section describes tests performed to characterize the response of the case study

bridge under a group of four pedestrians.

D.4.1 Procedure for Group Walking Tests

A set of eight group walking tests was performed. In each test, four pedestrians walked

approximately in-sync with a metronome set to the desired frequency. One test was

performed for each frequency that is a multiple of 0.25 Hz in the range, 0.75 Hz to 2.5

Hz. These tests include the approximate vertical numerical (second and higher modes),

but not the experimentally found vertical natural frequencies. The range here differs from

the single pedestrian test. The lower end of the range was increased above that of the

numerically predicted vertical fundamental frequency because walking at 0.5 Hz was

determined to be an unrealistic walking pace; walking at 0.5 Hz was very slow. Testing

stopped after 2.5 Hz, because higher frequencies required walking at paces that were too

fast to allow the pedestrians to synchronize.

In all of the tests, the group began at one end of the bridge, walked to the other end of

the bridge, turned around, and walked back to midspan where the pedestrians stopped

abruptly. Figure D.4 shows the group of pedestrians during two different tests.

D.4.2 Results and Discussion for Group Walking Tests

Table D.2 presents the maximum (absolute values) for lateral (H) and vertical (V)

accelerations among all of the accelerometers for each test. As was the case in the single

pedestrian tests, the maximum accelerations for the crowd tests exceed Sétra’s (2006)

164
)

0.08 g and 0.25 g limits in the lateral and vertical directions, respectively. Figure D.5

presents a set of acceleration time series. Again, the acceleration time series shown are

for accelerometers with the maximum accelerations in tests in which the walking

frequency is closest to the experimentally found vertical natural frequencies.

These tests are g1, g3, g6, and g7. The plots in Figure D.5 show that during this set of

tests, accelerations frequently are close to or exceed the lateral and vertical comfort limits

(shown in dashed lines). Had the tests been performed at walking frequencies closer to

the actual natural frequencies of the bridge, it is expected that the resulting resonance

would lead to even higher accelerations.

a. Group walking in-sync with the b. Group walking near the bridge
metronome. centerline.
Figure D.4. Group of four pedestrians in two different tests walking across the bridge.

Table D.2. Results for group walking tests.


Maximum |acceleration| (g), accelerometer
Test Walking frequency (Hz)
H V
g1 0.75 0.57, a5 1.68, a1
g2 1.00 0.66, a2 3.41, a2
g3 1.25 0.46, a5 1.94, a1
g4 1.50 0.53, a5 2.41, a3b
g5 1.75 0.72, a2 2.52, a5
g6 2.00 0.68, a2 2.69, a2
g7 2.25 0.96, a5 2.27, a3a
g8 2.50 0.90, a5 1.85, a5

165
)

0.9 2.4
0.6 1.6
H (g) 0.3 0.8

V (g)
0 0
−0.3 −0.8
−0.6 −1.6
−0.9 −2.4
0 30 60 90 120 150 180 0 30 60 90 120 150 180
Time (s) Time (s)

a. Test g1, accelerometer a5. e. Test g1, accelerometer a1.

0.9 2.4
0.6 1.6
0.3 0.8
H (g)

V (g)
0 0
−0.3 −0.8
−0.6 −1.6
−0.9 −2.4
0 30 60 90 120 150 180 0 30 60 90 120 150 180
Time (s) Time (s)

b. Test g3, accelerometer a5. f. Test g3, accelerometer a1.

0.9 2.4
0.6 1.6
0.3 0.8
H (g)

V (g)

0 0
−0.3 −0.8
−0.6 −1.6
−0.9 −2.4
0 30 60 90 120 150 180 0 30 60 90 120 150 180
Time (s) Time (s)

c. Test g6, accelerometer a2. g. Test g6, accelerometer a2.

0.9 2.4
0.6 1.6
0.3 0.8
H (g)

V (g)

0 0
−0.3 −0.8
−0.6 −1.6
−0.9 −2.4
0 30 60 90 120 150 180 0 30 60 90 120 150 180
Time (s) Time (s)

d. Test g7, accelerometer a5. h. Test g7, accelerometer a3a.


Figure D.5. Acceleration time series for the lateral (a – d) and vertical (e – h)
accelerations for a set of the group walking tests.

166
)

D.5 Conclusions
Full-scale physical tests were performed to evaluate the case study bridge’s accelerations

under anticipated traffic levels on the bridge. The bridge had maximum accelerations

that did not meet Sétra’s minimum pedestrian comfort limits. As the number of people on

the bridge increased, the duration of time at which the limits were exceeded at a

particular bridge location also increased. Synchronization of the pedestrians with the

bridge’s natural frequencies was not done in this study; pedestrians walked at frequencies

near, but not exactly at the experimentally found vertical bridge natural frequencies.

Walking in-sync with the bridge’s natural frequencies is expected to lead to greater

accelerations and durations of time in which comfort limits are not met.

167
)

Figure Credits
Unless otherwise noted in this disseration, the figures were created by the author. This

section provides the complete credits for the figures included in this dissertation that were

generated by others.

Figure 2.3a.

Photograph courtesy of Bridges to Prosperity.

Link to photograph:

<http://commons.wikimedia.org/wiki/File%3ACollpatomaico_footbridge.jpg>.

Licensing: [CC BY-SA 3.0 <http://creativecommons.org/licenses/by-sa/3.0>], via

Wikimedia Commons.

The author of this dissertation cropped the photograph.

Figure 2.3b.

Photograph courtesy of Susan Williams.

Link to photograph: <https://www.flickr.com/photos/susan_w/3650404159/>.

Licensing: [CC BY-NC 2.0 <https://creativecommons.org/licenses/by-

nc/2.0/deed.en>], via Flikr.

The author of this dissertation made no modifications to the photograph.

168
)

Figure 2.3c.

Photgraph courtesy of Lev Yakupov.

Link to photograph:

<https://www.flickr.com/photos/28217469@N00/3125116579/in/photolist-5La3hk-

5N35SV-6LsuRY-6PPaKG-6Q2vcP-6RXHJA-7bzAcW-7juN5H-7nhTvp-7niDxK-

7nmykw-dKSeAX-dKT24r-abSNa2-btxCBP-abUesE-byPBJ5-bjKnXp-aFsYcV-

9eP7g7-bqrsSo-nBpLPb-8aCa6R-aHK47r-qtnc1Y-qr271X-7RnKuh-97FHY9-

p8y9i7-8aC2iR-a5Czhg-apqcxA-8Z8ZZx-eyGcRN-aRASkX-aRASLF-eN2fxP-

7YCt7k-8XQtSf-aQ8GJc-dVxkx1-7YmbUC-aTwtEa-7G3imF-aVEsVX-9Fd6Kz-

aP1qxR-7Hh9Fe-dVwEYW-btwFne-qpUxo2>.

Licensing: [CC BY-SA 2.0 < https://creativecommons.org/licenses/by-

sa/2.0/deed.en>], via Flikr.

The author of this dissertation cropped and sharpened the photograph.

Figure 2.3d.

Photgraph courtesy of John Pavelka.

Link to photograph: <https://www.flickr.com/photos/28705377@N04/7476664204/>.

Licensing: [CC BY 2.0 <https://creativecommons.org/licenses/by/2.0/deed.en>], via

Flikr.

The author of this dissertation made no modifications to the photograph.

169
)

Figure 2.4a.

Photograph courtesy of Edmund Candler.

Link to photograph: <http://commons.wikimedia.org/wiki/File%3AOld_Chain-

Bridge_at_Chaksam.jpg>.

Licensing: [Public domain], via Wikimedia Commons.

The author of this dissertation made no modifications to the photograph.

Figure 2.4b.

Photograph courtesy of Rutahsa Adventures <www.rutahsa.com>.

Link to photograph: <http://commons.wikimedia.org/wiki/File:IRB-6-

BringingDeckMat-KC603-8.jpg>.

Licensing: Uploaded with permission by User: Leonard G. at en.wikipedia [CC BY-

SA 1.0 <http://creativecommons.org/licenses/by-sa/1.0>], via Wikimedia Commons.

The author of this dissertation made no modifications to the photograph.

Figure 2.5b.

Photograph courtesy of William Plunkett.

The photograph was obtained directly from the creator.

The author of this dissertation made no modifications to the photograph.

Figure 3.2.

Drawing courtesy of Ryan Woodward.

The drawing was obtained directly from the creator.

The author of this dissertation made modifications to the drawing that were approved

by the creator.

170
)

Figure 3.3.

Drawing courtesy of Ryan Woodward.

The drawing was obtained directly from the creator.

The author of this dissertation made modifications to the drawing that were approved

by the creator.

171
)

Works Cited
AASHTO. (2002). Standard specifications for highway bridges, 17th edition, American

Association of State Highway and Transportation Officials (AASHTO), Washington,

D.C.

AASHTO. (2009). LRFD guide specifications for the design of pedestrian bridges, 2nd

edition, AASHTO, Washington, D.C.

ABS. (2011). Guidance notes on the application of fiber rope for offshore mooring,

American Bureau of Shipping (ABS), Houston, TX.

Bang, A. L. (2009). "Cable-supported pedestrian bridge design for rural construction.”

M.S. Thesis, University of Colorado, Boulder, CO.

Barnes, M. R. (1999). "Form finding and analysis of tension structures by dynamic

relaxation." International Journal of Space Structures, 14(2), 89-104.

Bel Hadj Ali, N., Rhode-Barbarigos, L., and Smith, I. F. C. (2011). "Analysis of clustered

tensegrity structures using a modified dynamic relaxation algorithm." International

Journal of Solids and Structures, 48(5), 637-647.

Beyer, H.-G., and Sendhoff, B. (2007). "Robust optimization - A comprehensive survey."

Computer Methods in Applied Mechanics and Engineering, 196(33-34), 3190-3218.

Billington, D. P., and Nazmy, A. (1990). "History and aesthetics of cable-stayed bridges."

Journal of Structural Engineering, 117(10), 3103-3134.

172
)

Bleicher, A., Schlaich, M., Fujino, Y., and Schauer, T. (2011). "Model-based design and

experimental validation of active vibration control for a stress ribbon bridge using

pneumatic muscle actuators." Engineering Structures, 33(8), 2237-2247.

Bloss, N., Rehm, K., and Bowser, M. (2014). "Footbridges save lives in a walking world

- Bridgin the Gap Africa." Footbridge 2014 - Past, Present & Future, Hemming

Information Services, London, United Kingdom.

Bridges to Prosperity. (2011). Bridges to Prosperity bridge manual.

Buonopane, S. G. (2006). "The Roeblings and the stayed suspension bridge: Its

development and propogation in 19th century United States." The Second

International Congresss on Construction History, Construction History Society,

Cambridge, United Kingdom, 441-460.

Buonopane, S. G. (2012). "A historical perspective on suspension bridges." Festschrift

Billington, E. M. Hines, S. G. Buonopane, and M. E. M. Garlock, International

Network for Structural Art, 38-76.

Burgoyne, C. J. (1993). "Parafil ropes for prestressing applications." Fiber-reinforced-

plastic (FRP) reinforcement for concrete structures: Properties and applications, A.

Nanni, ed., Elsevier Science B.V., Amsterdam, The Netherlands.

Burgoyne, C. J. (2001). "Rational use of advanced composites in concrete." Proceedings

of the Institution of Civil Engineers: Structures and Buildings, 146(3), 253-262.

173
)

Burgoyne, C. J., and Head, P. R. (1993). "Aberfeldy Bridge -- An advanced textile

reinforced footbridge." Proceedings of the TechTextil Symposium, Frankfurt,

Germany, paper 418, 1-9.

Caetano, E. and Cunha, Á. (2004). "Experimental and numerical assessment of the

dynamic behaviour of a stress-ribbon footbridge." Structural Concrete, 5(1), 28-38.

Caetano, E., Cunha, Á., and Moutinho, C. (2011). "Vandal loads and induced vibrations

on a footbridge." Journal of Bridge Engineering, 16(3), 375-382.

Chen, Z., Cao, H., Zhu, H., Hu, J., and Li, S. (2014). "A simplified structural mechanics

model for cable-truss footbridges and its implications for preliminary design."

Engineering Structures, 68, 121-133.

Chopra, A. K. (2007). Dynamics of structures: Theory and applications to earthquake

engineering, 3rd edition, Pearson Prentice Hall, Upper Saddle River, New Jersey.

Dallard, P., Fitzpatrick, T., Flint, A., Low, A., Smith, R. R., Willford, M., and Roche, M.

(2001). "London Millennium Bridge: Pedestrian-induced lateral vibration." Journal of

Bridge Engineering, 6(6), 412-417.

Day, A. S. (1965). "An introduction to dynamic relaxation." The Engineer, 219, 218-221.

Deb, K. (2001). Multi-objective optimization using evolutionary algorithms. John Wiley

& Sons, Ltd., Chichester, United Kingdom.

174
)

Deb, K., Pratap, A., Agarwal, S., and Meyarivan, T. (2002). "A fast and elitist multi-

objective genetic algorithm: NSGA-II.” IEEE Transactions on Evolutionary

Computation, 6(2), 182-197.

Ding, H., Zhang, R., Zhang, Y., Xia, W., and Huang, J. (2011). "A bridge in China

externally prestressed by carbon fiber reinforced polymer tendons." Structural

Engineering International, 21(3), 346-348.

fib. (2005). fib bulletin 32: Guidelines for the design of footbridges, International

Federation for Structural Concrete (fib), Laussane, Switzerland.

Filomeno Coelho, R., Lebon, J., and Bouillard, P. (2011). "Hierarchical stochastic

metamodels based on moving least squares and polynomial chaos expansion:

Application to the multiobjective reliability-based optimization of space truss

structures." Structural and Multidisciplinary Optimization, 43(5), 707-729.

Flory, J. F., Banfield, S. J., and Berryman, C. (2007). "Polyester mooring lines on

platforms and MODUs in deep water.” Proceedings of the Annual Offshore

Technology Conference, Offshore Technology Conference, Richardson, TX, OTC

18768.

Flory, J. F., Banfield, S. P., and Petruska, D. J. (2004). "Defining, measuring, and

calculating the properties of fiber rope deepwater mooring lines." Proceedings of the

Annual Offshore Technology Conference, Offshore Technology Conference,

Richardson, TX, OTC 16151.

175
)

Fujino, Y., Pacheco, B. M., Nakamura, S.-I, and Warnitchai, P. (1993). "Synchronization

of human walking observed during lateral vibration of a congested pedestrian bridge."

Earthquake Engineering and Structural Dynamics, 22(9), 741-758.

Fujino, Y. and Siringoringo, D. (2014). "Revisiting the pedestrian-induced lateral

vibration of footbridges and crowd synchronization problem." Footbridge 2014 -

Past, Present & Future, Hemming Information Services, London, United Kingdom.

Gentile, C., and Gallino, N. (2008). "Ambient vibration testing and structural evaluation

of an historic suspension footbridge." Advances in Engineering Software, 39(4), 356-

366.

Gerner, M. (2007). Chakzampa Thangtong Gyalpo: Architect, philosopher and iron chain

bridge builder." The Centre for Bhutan Studies, Thimphu, Bhutan, translated from the

German by G. Verhufen.

Gibbs, M., Trost, D., Morgenthal, G., and Kareem, A. (2014). "Monitoring dynamics of

suspension footbridges using novel sensing technology." Footbridge 2014 - Past,

Present & Future, Hemming Information Services, London, United Kingdom.

Gimsing, N. J., and Georgakis, C. T. (2012). Cable supported bridges: concept and

design, 3rd edition, John Wiley & Sons, Ltd., Chichester, United Kingdom.

Greiner, D., Emperador, J. M., Winter, G., and Galván, B. (2007). "Improving

computational mechanics optimum design using helper objectives: An application in

frame bar structures." Evolutionary Multi-Criterion Optimization: 4th International

Conference EMO 2007, Matsushima, Japan, March 5-8, 2007. Proceedings, S.

176
)

Obayashi, K. Deb, C. Poloni, T. Hiroyasu, and T. Murata, eds., Springer, Berlin,

Germany, 575-589.

Gupte, S., Betti, R., and Zoli, T. (2010). "Synthetic fiber ropes to replace steel wire in

pedestrian suspension bridges." Bridge Maintenance, Safety, Management and Life-

Cycle Optimization – Proceedings of the 5th International Conference on

Maintenance, Safety and Management, Taylor and Francis Group, London, United

Kingdom, 3228-3234.

Hamming, R. W. (1989). Digital filters, 3rd edition, Prentice Hall, Englewood Cliffs, NJ.

Heinemeyer, C., Butz, C., Keil, A., Schlaich, M., Goldack, A., Trometer, S., Lukić, M.,

Chabrolin, B., Lemaire, A, Martin, P.-O., Cunha, Á., and Caetano, E. (2009). JRC

scientific and technical reports: Design of lightweight footbridges for human induced

vibrations, European Commission, Office for Official Publications of the European

Communities, Luxembourg.

Hennessey, C. M., Pearson, N. J., and Plaut, R. H. (2005). "Experimental snap loading of

synthetic ropes." Shock and Vibration, 12(3), 163-175.

HMG, TBSSP/Helvetas. (2003). Short span trail bridge standard: suspended, SKAT.

Huang, M.-H., Thambiratnam, D. P., and Perera, N. J. (2005a). "Load deformation

characteristics of shallow suspension footbridge with reverse profiled pre-tensioned

cables." Structural Engineering and Mechanics, 21(4), 375-392.

177
)

Huang, M.-H., Thambiratnam, D. P., and Perera, N. J. (2005b). "Resonanant vibration of

shallow suspension footbridges." Proceedings of the Institution of Civil Engineers:

Bridge Engineering, 158(BE4), 201-209.

Huang, M.-H., Thambiratnam, D. P., and Perera, N. J. (2005c). "Vibration characteristics

of shallow suspension bridge with pre-tensionsed cables." Engineering Structures,

27(8), 1220-1233.

Irvine, H. M. (1975). "Statics of Suspended Cables." Journal of the Engineering

Mechanics Divsion, 101(EM3), 187-205.

Irvine, M. (1981). Cable structures, Dover Publications, Inc., Mineola, NY.

Lebo, J., and Schelling, D. (2001). “Design and Appraisal of Rural Transport

Infrastructure: Ensuring Basic Access for Rural Communities." World Bank technical

paper no. 496, The World Bank, Washington D.C.

Magalhães, F., Cunha, Á., and Caetano, E. (2007). "Dynamic testing of the New Coimbra

Footbridge before implementation of control devices." Conference Proceedings of the

Society for Experimental Mechanics Series, Springer, New York, NY.

Magalhães, F., Cunha, Á., Caetano, E., and Brincker, R. (2010). "Damping estimation

using free decays and ambient vibration tests." Mechanical Systems and Signal

Processing, 24(5), 1274-1290.

MATLAB. (2012). Version 7.14.0.739 (R2012a).

178
)

McConnell, K. G., and Varoto, P. S. (2008). Vibration testing: Theory and practice, 2nd

edition, John Wiley & Sons, Inc., Hoboken, NJ.

McKenna, H. A., Hearle, J. W. S., and O'Hear, N. (2004). Handbook of fibre rope

technology, Woodhead Publishing Ltd., Cambridge, United Kingdom.

Meier, U. (2012). "Carbon fiber reinforced polymer cables: Why? Why Not? What If?"

Arabian Journal for Science and Engineering, 37(2), 399-411.

Meng, X., Dodson, A. H., and Roberts, G. W. (2007). "Detecting bridge dynamics with

GPS and triaxial accelerometers." Engineering Structures, 29(11), 3178-3184.

Moored, K. W., and Bart-Smith, H. (2009). "Investigation of clustered actuation in

tensegrity structures." International Journal of Solids and Structures, 46(17), 3272-

3281.

Moschas, F., and Stiros, S. (2011). "Measurement of the dynamic displacements and of

the modal frequencies of a short-span pedestrian bridge using GPS and an

accelerometer." Engineering Structures, 33(1), 10-17.

Ochsendorf, J. (2004). "Sustainable structural design: Lessons from history." Structural

Engineering International, 14(3), 192-194.

Park, G.-J., Lee, T.-H., Lee, K. H., and Hwang, K.-H. (2006). "Robust design: An

overview." AIAA Journal, 44(1), 181-191.

179
)

Press, W. H., Teukolsky, S. A., Vetterling, W. T., and Flannery, B. P. (1992). Numerical

recipes in C: The art of scientific computing, 2nd edition, Cambridge University

Press, Cambridge, United Kingdom.

Ramberg, S. E., and Griffin, O. M. (1977). "Free vibrations of taut and slack marine

cables." Journal of the Structural Division, Proceedings of the American Society of

Civil Engineers, 103(ST11), 2079-2092.

Raphael, B., and Smith, I. F. C. (2003). Fundamentals of computer-aided engineering,

John Wiley & Sons Ltd., Chichester, United Kingdom.

RFCS (2008). Human induced vibration of steel structures (Hivoss) - Design of

footbridges: Guidelines, Research Fund for Coal and Steel (RFCS).

Roberts, P., Shyam, K. C., and Cordula, R. (2006). “Rural access index: A key

development indicator.” Transport paper (TP-10). The International Bank for

Reconstruction and Development/The World Bank, Washington, D.C.

Roberts, T. M. (2005). "Lateral pedestrian excitation of footbridges." Journal of Bridge

Engineering, 10(1), 107-112.

Sarner, C-M., Wachter, M., and Plunkett, W. (2010). “A stiffnened rope catenary bridge:

Exploration of concept and constructability.” Unpublished report, Princeton

University, Princeton, NJ.

Segal, E. M., Rhode-Barbarigos, L., Filomeno Coelho, R. D., and Adriaenssens, S.

(2014). "An Automated Robust Design Methodology for Suspended Structures."

180
)

Proceedings of the IASS-SLTE 2014 Symposiuim - Shells, Membranes, and Spatial

Structures: Footprints, The International Association for Shell and Spatial Structures

(IASS), Brasília, Brazil.

SENSR. (2014). “Practical guide to accelerometers.” SENSR,

<http://www.sensr.com/pdf/practical-guide-to-accelerometers.pdf> (Nov. 30, 2014).

Sétra. (2006). Technical guide - Footbridges: Assessment of vibrational behaviour of

footbridges under pedestrian loading, Service d’Études techniques des routes et

autoroutes (Sétra), Paris, France.

Smith III, J. O. (2007). Introduction to digital filters with audio applications, W3K

Publishing.

Strasky, J. (2005). Stress ribbon and cable-supported pedestrian bridges, Thomas

Telford Publishing, London, United Kingdom.

Van Nimmen, K., Lombaert, G., De Roeck, G., and Van den Broeck, P. (2014).

"Vibration serviceability of footbridges: Evaluation of the current codes of practice."

Engineering Structures, 59, 448-461.

Welch, P. D. (1967). "The use of fast fourier transform for the estimation of power

spectra: A method based on time averaging over short, modified periodograms."

IEEE Transactions on Audio and Electroacoustics, AU-15(2), 70-73.

181
)

Winslow, P., Pellegrino, S., and Sharma, S. B. (2010). "Multi-objective optimization of

free-form grid structures." Structural and Multidisciplinary Optimization, 40(1-6),

257-269.

Woodward, R. (2012a). “EWB Morocco Footbridge, Ait Bayoud site 1." Calculations,

April 27, 2012.

Woodward, R. (2012b). "Pedestrian bridge at site 1, Ait Bayoud, commune of Bizdad,

Morocco." Structural drawings, S3 - S12.

World Bank. (2007). “Rural access index (RAI).” The World Bank Group,

<http://www.worldbank.org/transport/transportresults/headline/rural-access.html>

(Mar. 14, 2015).

Zavala, G. R., Nebro, A. J., Luna, F., and Coello Coello, C. A. (2014). "A survey of

multi-objective metaheuristics applied to structural optimization ." Structural and

Multidisciplinary Optimization, 49(4), 537-558.

Zimmerman, A. T., Swartz, R. A., Lynch, J. P. (2008). “Automated identification of

modal properties in a steel bridge instrumented with a dense wireless sensor

network.” Bridge Maintenance, Safety, Management, Health Monitoring and

Informatics – Proceedings of the 4th International Conference on Bridge

Maintenance, Safety and Management, Taylor & Francis Group, London, United

Kingdom, 1608-1615.

182
)

Zitzler, E. (1999). “Evolutionary algorithms for multiobjective optimization: Methods

and applications." Doctor of Technical Sciences dissertation, Swiss Federal Institute

of Technology (ETH) Zürich, Zürich, Switzerland.

Zitzler, E., Thiele, L., Laumanns, M., Fonseca, C. M., and Grunert da Fonseca, V. (2002).

“Performance assessment of multiobjective optimizers: An analysis and review."

TIK- Report No. 139, ETH Zürich, Zürich, Switzerland.

183

You might also like