Multi scale smart sensing of vibration and impedance for structural health monitoring of cable stayed bridge

Thesis for the Degree of Doctor of PhilosophyMulti-scale Smart Sensing of Vibration and Impedance for Structural Health Monitoring of Cable-stayed Bridge by Duc-Duy Ho Department of Ocea

Thesis for the Degree of Doctor of Philosophy

Multi-scale Smart Sensing of Vibration and Impedance for Structural Health Monitoring of Cable-stayed Bridge

by Duc-Duy Ho Department of Ocean Engineering

The Graduate School Pukyong National University

February 2012

Multi-scale Smart Sensing of Vibration and Impedance for Structural Health Monitoring of Cable-stayed Bridge

Advisor: Prof Jeong-Tae Kim

by

Duc-Duy Ho

A thesis submitted in partial fulfillment of the requirements

for the degree ofDoctor of Philosophy

in Department of Ocean Engineering, the Graduate School,

Pukyong National University

February 2012

TABLE OF CONTENTS

TABLE OF CONTENTS i

LIST OF FIGURES v

LIST OF TABLES x

ABSTRACT xii

CHAPTER 1 INTRODUCTION 1

1.1 Problem Statement 1

1.2 Research Backgrounds 1

1.3 Objective Statement 12

1.4 Organization of the Dissertation 13

CHAPTER 2 THEORY OF MULTI-SCALE VIBRATION AND IMPEDANCE RESPONSES 15

2.1 Overview 15

2.2 Theories of Vibration Responses 15

2.2.1 Acceleration Responses 15

2.2.2 PZT’s Dynamic Strain Responses 16

2.3 Theory of Impedance Response 22

2.4 Vibration and Impedance Features 24

2.4.1 Vibration Features 24

2.4.2 Impedance Features 26

2.5 Feasibility of Multi-scale Vibration and Impedance Responses 27

2.5.1 Test Structure and Experimental Setup 27

2.5.2 Multi-scale Vibration and Impedance Responses 29

CHAPTER 3 MULTI-SCALE VIBRATION AND IMPEDANCE

RESPONSES OF PYLON-CABLE-DECK SYSTEM 31

3.1 Overview 31

3.2 Pylon-Cable-Deck System and Damage Types 31

3.3 SHM Schemes for Cable-stayed Bridge 33

3.4 Multi-scale Vibration and Impedance Responses 35

3.4.1 Vibration Responses of Pylon-Cable-Deck System 35

3.4.2 Impedance Response of Cable-Anchorage Subsystem 37

3.5 Experimental Analysis of Cable-Girder Model 38

CHAPTER 4 VIBRATION AND IMPEDANCE-BASED SHM METHODS 46

4.1 Overview 46

4.2 Vibration-based SHM Methods 46

4.2.1 Damage Alerting Methods 47

4.2.2 MSE-based Damage Index Method 48

4.2.3 MSE-based System Identification Method 50

4.2.4 Frequency-based Cable Force Model 51

4.3 Impedance-based SHM Methods 53

CHAPTER 5 MULTI-SCALE SMART SENSOR SYSTEM 56

5.1 Overview 56

5.2 Hardware Design 56

5.2.1 Schematic of Multi-scale Sensor Node 56

5.2.2 Imote2 Sensor Platform 57

5.2.3 Vibration Sensor Boards 59

5.2.4 Impedance Sensor Board 62

5.2.5 Solar Power Harvesting Unit 64

5.3 Embedded Software Design 65

5.3.1 Software for Solar Power Harvesting Unit 65

5.3.2 Vibration-based SHM Software 66

5.3.3 Impedance-based SHM Software 69

5.4 Sensing Performance of Multi-scale Smart Sensor System 71

CHAPTER 6 LABORATORY EVALUATION OF MULTI-SCALE SMART SENSOR SYSTEM 80

6.1 Overview 80

6.2 Experimental Setup and Damage Scenarios 80

6.3 Vibration and Impedance Responses 83

6.4 Cable Force Monitoring by Multi-scale Smart Sensor System 89

6.4.1 Vibration-based Cable Force Monitoring 89

6.4.2 Impedance-based Cable Force Monitoring 91

CHAPTER 7 FIELD APPLICATION OF MULTI-SCALE SMART SENSOR SYSTEM 93

7.1 Overview 93

7.2 Description of Test Bridge 93

7.3 Field Deployment of Multi-scale Smart Sensor System 98

7.4 Long-term SHM Performance on Test Bridge 105

7.4.1 Test History 105

7.4.2 Wireless Communication and Power Consumption 107

7.4.3 Vibration Monitoring under Typhoon Condition 109

7.4.4 Long-term SHM under Weather Change 112

7.5 Cable Force Monitoring for Test Bridge 116

7.5.1 Vibration-based Cable Force Monitoring 116

7.5.2 Impedance-based Cable Force Monitoring 120

7.5.3 Long-term Cable Force Monitoring 121

7.6 Experimental Modal Identification of Test Bridge 123

7.7 System Identification of Test Bridge 128

CHAPTER 8 SUMMARY AND CONCLUSION 133

APPENDIX A PZT’S MULTIPLE PIEZOELECTRIC RESPONSES

FOR DAMAGE MONITORING IN BEAMS 137

A.1 Target Structure and Experimental Setup 137

A.2 PZT Sensor for Vibration-based Damage Monitoring 138

A.3 PZT Sensor for Impedance-based Damage Monitoring 141

REFERENCES 143

CURRICULUM VITAE 156

ACKNOWLEDGEMENTS 165

LIST OF FIGURES

Figure 1.1 Wireless Sensor Platforms 7

Figure 1.2 SHM Systems of Cable-stayed Bridges 11

Figure 2.1 Piezoelectric Material: Direct and Inverse Effects 17

Figure 2.2 Relationship between Strain and Electrical Field for PZT, PMN, and Terfenol D (Giurgiutiu 2008) 18

Figure 2.3 Schematic of PZT-Beam Interaction in Direct Piezoelectric Effect 19

Figure 2.4 Calibration Test Setup for PZT’s Dynamic Strain in Test Beam A 20

Figure 2.5 Time History Responses of PZT Sensor and Strain Gage 21

Figure 2.6 PZT Sensor’s Dynamic Voltage vs Strain Gage’s Dynamic Strain 21

Figure 2.7 1-D Model Electro-Mechanical Interaction between Piezoelectric Patch and Host Structure (Liang et al 1996) 22

Figure 2.8 Experiment Setup for Test Beam B 28

Figure 2.9 Time History Responses of Test Beam B 30

Figure 2.10 Frequency Responses of Test Beam B 30

Figure 2.11 Mode Shape Curvatures of Test Beam B 30

Figure 2.12 PZT1’s Impedance Signatures of Test Beam B 30

Figure 3.1 Schematic of Cable-stayed Bridge 32

Figure 3.2 Force Equilibrium of Pylon-Cable-Deck System 33

Figure 3.3 Schematic of Vibration-based SHM 34

Figure 3.4 Schematic of Impedance-based SHM 34

Figure 3.5 Multi-Scale Vibration and Impedance Sensing Scheme for Pylon-Cable-Deck System 35

Figure 3.6 Schematic of PZT-Cable Interaction 36

Figure 3.7 Schematic of Anchorage Force versus Structural Parameters in Cable-Anchorage Subsystem 37

Figure 3.8 Experiment Setup for Cable-Girder Model 39

Figure 3.9 Girder’s Vibration Responses in Cable-Girder Model 40

Figure 3.10 Cable’s Vibration Responses in Cable-Girder Model 40

Figure 3.11 Wind-induced Vibration Test on External Cable 41 Figure 3.12 Vibration Responses from Three Sensor Types at Wind

Speed of 0 m/s 42

Figure 3.13 Vibration Responses from Three Sensor Types at Wind Speed of 1 m/s 43

Figure 3.14 Vibration Responses from Three Sensor Types at Wind Speed of 2.2 m/s 43

Figure 3.15 Vibration Responses from Three Sensor Types at Wind Speed of 4.5 m/s 44

Figure 3.16 Anchorage’s Impedance Responses in Cable-Girder Model 45

Figure 3.17 Anchorage’s Impedance Responses versus Wind Speeds 45

Figure 5.1 Schematic of Multi-scale Sensor Node 57

Figure 5.2 Prototype of Multi-scale Sensor Node on Imote2 Platform 57

Figure 5.3 Schematic and Prototype of Imote2 Sensor Platform (Memsic Co 2010) 59

Figure 5.4 Schematic of SHM-H Sensor Board (Jo et al 2010) 61

Figure 5.5 Schematic of SHM-A (AS) Sensor Board 62

Figure 5.6 Schematic and Prototype of SSeL-I Sensor Board (Kim et al 2011) 63

Figure 5.7 Schematic of Solar Power Supply Design (Crossbow Technology 2007) 65

Figure 5.8 Flowchart of ChagerControl Component (Miller et al 2010) 66 Figure 5.9 Schematic of Vibration-based SHM Software 67

Figure 5.10 Flowchart of RemoteSensing Component (Rice and Spencer 2009) 68

Figure 5.11 Flowchart of AutoMonitor Component 68

Figure 5.12 Schematic of Impedance-based SHM Software 69

Figure 5.13 Flowchart of Impedance Component 70

Figure 5.14 Flowchart of ImpAutoMonitor Component 71

Figure 5.15 Experimental Setup for Test Beam C 72

Figure 5.16 Temperature Monitoring Results in Test Beam C 73

Figure 5.17 Power Monitoring Results in Test Beam C 73

Figure 5.18 Acceleration Responses of Test Beam C by Imote2/SHM-AS 75

Figure 5.19 PZT’s Dynamic Strain Responses of Test Beam C by Imote2/SHM-AS 75

Figure 5.20 CCs of PSDs vs Various Temperatures in Test Beam C 76

Figure 5.21 SSI Method’s Stabilization Charts for Test Beam C 77 Figure 5.22 First Natural Frequency Changes versus Various

Temperatures in Test Beam C 77 Figure 5.23 Second Natural Frequency Changes versus Various

Temperatures in Test Beam C 78 Figure 5.24 Real Impedance Signatures of Test Beam C by

Imote2/SSeL-I 79 Figure 5.25 Impedance Signatures vs Various Temperatures in Test

Beam C 79 Figure 6.1 Experiment Setup for Cable-Girder Model 81 Figure 6.2 External Cable of Cable-Girder Model 82 Figure 6.3 Acceleration Responses of Cable-Girder Model: Wired PCB

System vs Wireless Imote2/SHM-AS 84 Figure 6.4 PZT’s Dynamic Strain Responses of Cable-Girder Model:

Wired PCB System vs Wireless Imote2/SHM-AS 84 Figure 6.5 Impedance Responses of Cable-Girder Model: Wired

HIOKI System vs Wireless Imote2/SSeL-I 84 Figure 6.6 Interaction between Girder and Cable’s Vibrations 85 Figure 6.7 Wind Speed Effect on Vibration Responses of Cable-Girder

Model: 0 m/s 86 Figure 6.8 Wind Speed Effect on Vibration Responses of Cable-Girder

Model: 1 m/s 87 Figure 6.9 Wind Speed Effect on Vibration Responses of Cable-Girder

Model: 2.2 m/s 87 Figure 6.10 Wind Speed Effect on Vibration Responses of Cable-Girder

Model: 4.5 m/s 88 Figure 6.11 Wind Speed Effect on Impedance Responses of Cable-

Girder Model 88 Figure 6.12 Freq Responses vs Cable Forces of Cable-Girder Model:

Acceleration 89 Figure 6.13 Freq Responses vs Cable Forces of Cable-Girder Model:

PZT’s Dynamic Strain 89 Figure 6.14 Cable Force Estimation for Cable-Girder Model by

Imote2/SHM-AS 91 Figure 6.15 Real Impedances vs Cable Forces of Cable-Girder Model by

Imote2/SSeL-I 92

Figure 6.16 Impedance Signatures vs Cable Forces of Cable-Girder Model by Imote2/SSeL-I 92

Figure 7.1 On-Site View of Hwamyung Cable-stayed Bridge 94

Figure 7.2 Geometry of Hwamyung Cable-stayed Bridge 95

Figure 7.3 Enclosure Assemblies of Sensor Nodes 99

Figure 7.4 Installation of Cable Nodes 99

Figure 7.5 Field Deployment of Sensor Nodes on Test Bridge 100

Figure 7.6 Field Sensor Layout on Test Bridge 101

Figure 7.7 PZT’s Dynamic Strain Measurement System on Cables 102

Figure 7.8 PZT’s Impedance Measurement System on Cable Anchorages 102

Figure 7.9 Wi-Fi and Imote2 Radio Channel (Illinois Structural Health Monitoring Project 2011) 103

Figure 7.10 Real-Time Monitoring Operation on Test Bridge 105

Figure 7.11 Performance Evaluation of Wireless Communication 108

Figure 7.12 Performance Evaluation of Power Consumption 109

Figure 7.13 Typhoon Meari’s Pass 110

Figure 7.14 Acceleration Responses and Power Spectral Densities of Top Pylon P1 111

Figure 7.15 Acceleration Responses and Power Spectral Densities of Deck D2 111

Figure 7.16 Acceleration Responses and Power Spectral Densities of Cable C4 111

Figure 7.17 Vibration Responses on Cable C3 112

Figure 7.18 CCs of PSDs versus Various Temperatures on Cable C3 113

Figure 7.19 Relative Changes in Natural Frequency versus Various Temperatures on Cable C3 114

Figure 7.20 Impedance Responses on Cable C3 115

Figure 7.21 RMSD and CC of Impedance Signatures versus Various Temperatures on Cable C3 115

Figure 7.22 Cable Force Estimation of Cables C1-C5 by Acceleration Responses 117

Figure 7.23 Cable Force Monitoring for Cable C3 by Acceleration and PZT’s Dynamic Strain Responses 119

Figure 7.24 PZT’s Dynamic Strain Responses versus Wind Speed: Cable

C3 120

Figure 7.25 Real Impedances versus Cable Forces: Cable C3 121

Figure 7.26 Impedance Signatures vs Cable Force Changes: Cable C3 121

Figure 7.27 Cable Force Estimation from Acceleration Responses for Cables C1-C5 versus Temperature 123

Figure 7.28 FE Model of Hwamyung Bridge 124

Figure 7.29 SSI Method: Stabilization Charts 126

Figure 7.30 Modal Parameters: Initial FE Model vs Experiment 127

Figure 7.31 Eigenvalue Sensitivities of Model-updating Parameters 129

Figure 7.32 Convergences of Natural Frequencies between FE Model and Experiment 132

Figure 7.33 Relative Changes in Model-updating Parameters of FE Model 132

Figure A.1 Experiment Setup for Test Beam B 137

Figure A.2 Power Spectral Densities of Test Beam B 139

Figure A.3 CC of PSDs of Test Beam B 139

Figure A.4 MSE-based Damage Localization for Test Beam B: Damage 1 140

Figure A.5 MSE-based Damage Localization for Test Beam B: Damage 2 140

Figure A.6 MSE-based Damage Localization for Test Beam B: Damage 3 140

Figure A.7 MSE-based Damage Localization for Test Beam B: Damage 4 141

Figure A.8 Changes in PZT1’s Real Impedance Signatures for Test Beam B 142

Figure A.9 RMSD of Impedance Signatures for Test Beam B 142

LIST OF TABLES

Table 1.1 Full-scale Implementation of Wireless Sensors 8

Table 2.1 Calibration for PZT’s Dynamic Strain in Test Beam A 21

Table 2.2 Vibration and Impedance Measurement Systems 29

Table 3.1 Natural Frequencies (Hz) Extracted by Three Sensor Types at Various Wind Speeds 42

Table 4.1 Vibration-based SHM Methods 46

Table 4.2 Impedance-based SHM Methods 53

Table 5.1 Comparison of Sensor Platforms 58

Table 5.2 Comparison of Vibration Measurement Systems 60

Table 5.3 Comparison of Impedance Measurement Systems 63

Table 5.4 Current Draw (mA) for Imote2/SHM-A(AS)/SHM-H/SSeL-I 65

Table 6.1 Specifications of External Cable 82

Table 6.2 Cable Force-loss Scenarios for Cable-Girder Model 83

Table 6.3 Natural Freqs (Hz) of External Cable at Various Wind Speeds 86

Table 6.4 Natural Frequencies (Hz) of External Cable Measured by Imote2/SHM-AS 90

Table 6.5 Cable Force Estimation for Cable-Girder Model by Imote2/SHM-AS 90

Table 7.1 Specifications of Cables in Hwamyung Cable-stayed Bridge 95 Table 7.2 Test History on Hwamyung Bridge (June 21 st - Sep 20 th , 2011) 106

Table 7.3 Wireless Communication Test (June 25 th -July 14 th , 2011) 108

Table 7.4 Structural Properties and Theoretical Natural Frequencies of Cables C1-C5 116

Table 7.5 Experimental Natural Frequencies (Hz) of Cables C1-C5 from Acceleration Responses at Temperature 27.9 o C 117

Table 7.6 Cable Force Estimation of Cables C1-C5 from Acceleration Responses at Temperature 27.9 o C 117

Table 7.7 Experimental Natural Frequencies (Hz) of Cable C3 Extracted from Acceleration and PZT’s Dynamic Strain Responses on August 17 th 118

Table 7.8 Cable Force Estimation (kN) from Acceleration and PZT’s

Dynamic Strain Responses for Cable C3 versus Temperature

on August 17 th 118

Table 7.9 Experimental Natural Frequencies (Hz) of Cables C1-C5 from Acceleration Responses versus Temperature 122

Table 7.10 Cable Force Estimation from Acceleration Responses for Cables C1-C5 versus Temperature 122

Table 7.11 FE Model’s Natural Frequencies (Hz) for Three Cases of Cable Forces 125

Table 7.12 Natural Frequencies: Initial FE Model vs Experiment 127

Table 7.13 Eigenvalue Sensitivities of Model-updating Parameters 129

Table 7.14 Natural Frequencies of FE Model during Model Update 131

Table 7.15 Model-updating Parameters: Initial Value versus Updated Value 131

Table A.1 Damage Scenarios on Test Beam B 137

Table A.2 Natural Frequencies (Hz) of Test Beam B by Two Sensor Types 138

Table A.3 MSE-based Damage Index for Test Beam B 140

Multi-scale Smart Sensing of Vibration and Impedance for Structural Health Monitoring of Cable-Stayed Bridge

Duc-Duy Ho

Department of Ocean Engineering

The Graduate School Pukyong National University

ABSTRACT

In this study, a multi-scale smart sensor system is developed forvibration and impedance-based structural health monitoring (SHM) ofpylon-cable-deck system in cable-stayed bridge The following approachesare implemented to achieve the objective Firstly, theories of multi-scaleresponses of acceleration, PZT’s dynamic strain and electro-mechanical(E/M) impedance are described Secondly, multi-scale vibration andimpedance responses of pylon-cable-deck system in cable-stayed bridge areexamined Thirdly, vibration and impedance-based SHM methods suitablefor the pylon-cable-deck system are examined Fourthly, multi-scale smartsensor system is designed for the vibration and impedance-based SHM Thefeasibility of the multi-scale smart sensor system is evaluated on a lab-scalecable-girder model Finally, the practicality of the multi-scale smart sensorsystem is evaluated on a real cable-stayed bridge, Hwamyung Bridge inKorea The system’s performance is experimentally analyzed under variouscable forces and weather conditions The experimental modal parameters areidentified by numerical modal analyses of the target bridge Also, itsstructural parameters are estimated from the modal strain energy-basedsystem identification using experimental modal parameters

CHAPTER 1 INTRODUCTION

1.1 Problem Statement

This study deals with the general problem of developing a multi-scalesmart sensor system for vibration and impedance-based structural healthmonitoring (SHM) of cable-stayed bridge The problem is necessary for atleast three reasons Firstly, the multi-scale sensing mechanism of vibrationand impedance should be analyzes to enhance the detection capacity onmulti-types of damage in cable-stayed bridge Secondly, the multi-scale smartsensor system should be designed to improve the accuracy of SHM incomplex structures Thirdly, the multi-scale smart sensor system should beevaluated to secure the practical SHM in on-site cable-stayed bridges

1.2 Research Backgrounds

Overview

Recent tragic collapses of bridges and buildings have awoken the public

on the need of SHM that can play an important role in the safety and servicelife of civil infrastructures For most developed countries, the budget formanagement of civil infrastructures has been annually expended up to theamount of 8-15% GDP Despite the efforts, the occurrence of damage inthose structures is tended to be inevitable since they are subjected to extremeloading and environmental conditions that did not considered in designprocess One of the promising ways to guarantee the structural safety andintegrity is to enact SHM in a regular periodic manner and to detect criticaldamage in its early stage Additionally, the time and cost associated withappropriate maintenance and repair should be also well managed to produceefficient outputs (Doebling et al 1998, Farrar 2001)

In Korea, the number of deteriorated infrastructures built in the 1970shas been increased rapidly (Yun et al 2009) After the tragic collapse ofSungsoo Bridge in 1994, “Special Act on the Safety Control of Installations”was established for those previously built in Korea According to the Act,structures with significant importance or large occupancy (e.g., bridgeslonger than 100 m, tunnels longer than 500 m, building with 16 or morestories, railway stations, and dams) are forced to get the regular monitoringand maintenance by the authorities Many defective bridges have been foundfrom the active monitoring of structures, and many of them have beenretrofitted or reconstructed in early 2000’s It is also noticed that the costassociated with the structural monitoring and maintenance has beenincreasing rapidly during the same period.

As the concern is limited to the SHM in civil infrastructures, there havebeen many research attempts on structural response analysis, development ofnew sensing mechanism, adaptation of SHM method suitable to the structure,and field evaluation and application Along with the research track, this studyfocuses on the following subjects: 1) multi-scale sensing, 2) vibration andimpedance-based SHM, 3) smart sensor system for SHM, and 4) SHM oncable-stayed bridge

Multi-scale Sensing

For SHM applications, physical quantities such as acceleration, strain,impedance, guided wave and temperature were commonly utilized asseparate ones Multi-scale sensing may be defined as the capacity ofintegration of multiple physical quantities into one system One feature of themulti-scale sensing is the use of multi-types of sensors distributed on wholestructure to measure structural responses

Studer and Peters (2004) demonstrated that the multi-scale sensing

yields better damage identification results than single-scale measurements.They proposed a damage identification approach for composite structuresusing multi-metric data including strain, integrated strain, and strain gradientsmeasured from optical fiber sensors Law et al (2005) presented a wavelet-based approach combining acceleration and strain responses to obtain betterdamage detection results than using any of the two responses separately Kim

et al (2010) proposed a multi-scale sensing system to measure vibration andimpedance responses of a prestressed concrete girder model It was foundthat this system was good for examining the global and local behaviors of thestructure

In summary, the multi-scale sensing provides an effective way for SHM.However, the multi-scale sensing to examine the behaviors of cable-stayedbridge, which has a variety of member types and many critical joints, hasbeen rarely studied Therefore, there is a need of developing the multi-scalesensing for SHM in cable-stayed bridge

Vibration and Impedance-based SHM

Since the early 1970s, many researchers have used vibration features ofstructural responses as an indication of global damage (Adams et al 1978,Stubbs and Osegueda 1990, Doebling et al 1998, Zou et al 2000, Farrar

2001, Sohn et al 2003) The basic concept of the vibration-based damagedetection is to examine changes in modal parameters (e.g., natural frequency,mode shape, modal flexibility, and modal strain energy) which representchanges in structural properties (e.g., mass, stiffness, damping, and boundarycondition)

For the vibration-based SHM, damage is usually estimated in forms ofdamage indices that identify changes in modal parameters (Lin 1990, Pandey

et al 1991, Kim et al 1992, Pandey and Biswas 1994, Koh et al 1995, Kim

and Stubbs 1995, Morassi and Rovere 1997, Lam et al 1998, Sampaio et al.

1999, Gawronski and Sawicki 2000, Kim and Stubbs 2002, Kim et al 2003,Kim et al 2004, Huth et al 2005, Koo 2008, Park et al 2011) or changes intime-frequency domain features (Hou et al 2000, Zou et al 2000, Sohn andFarrar 2001, Quek et al 2001, Law and Lu 2005) Recently, soft computingtechniques such as neural networks and genetic algorithms have been utilizedfor damage pattern recognition and system identification (Wu et al 1992,Mares and Surace 1996, Yun and Bahng 2000, Chou and Ghaboussi 2001,Hao and Xia 2002, Kim et al 2007b, Shin et al 2008)

Adams et al (1978) introduced a damage detection method usingchanges in natural frequencies of a few modes Also, Cawley and Adams(1979) developed a method using sensitivity analysis to locate damage inplates Gudmundson (1982) investigated the effect of small cracks on naturalfrequencies by using a modal perturbation method Stubbs (1985) developed

a non-destructive theory using the linear inverse technique to identify damagelocation and severity Stubbs and Osegueda (1990) further evaluated theapplicability of this method on beams, plates, and shells However, thoseprevious works have the difficulty when the number of modes is much lessthan the number of damage parameters To overcome the problem, Kim andStubbs (2003) proposed a damage index method based on a sensitivity ratioconcept by using experimental natural frequencies and their correspondingnumerical modal strain energies of a few modes Based on the study, Kim et

al (2003) proposed modal strain energy (MSE)-based damage index methodfor damage detection in beam structures Then, this method was successfullyverified for steel plate-girder bridge model and prestressed concrete girderbridge model (Kim et al 2007a, Kim et al 2010)

The vibration-based SHM methods are very useful to detect the globaldamages in a structure by using a few sensors However, those methods have

some weak points as follows: 1) the methods are not accurate for detectingincipient damage of small size, 2) the methods detect approximate damagezone but not pin-point damage location, and 3) the methods cannotdistinguish individual damage types from mixed damage scenarios occurred

in complex structures (Kim et al 2007a, Kim et al 2010)

Impedance-based SHM techniques have been developed for local SHM

by utilizing the electro-mechanical (E/M) coupling property of piezoelectricmaterials The basic concept of the impedance-based damage monitoringmethods is to monitor the changes in structural mechanical impedance caused

by the presence of damage The use of impedance signatures for damagedetection was first proposed by Liang et al (1994) Since then, manyresearchers have improved the method and applied the method to variousdamage detection problems (Sun et al 1995, Liang et al 1996, Kabeya 1998,Park et al 1999, Giurgiutiu and Zagrai 2000, Park et al 2000, Giurgiutiu andZagrai 2002, Bhalla and Soh 2003, Park et al 2003, Giurgiutiu and Zagrai

2005, Tseng and Wang 2005, Park et al 2006, Yang et al 2008) Bhalla andSoh (2003) proved that the real part of impedance is more sensitive tostructural damage than the imaginary part

To monitor the occurrence of damage, Sun et al (1995) proposed asimple statistical algorithm based on the root mean square deviation (RMSD)

of impedance signatures Raju et al (1998) suggested another method based

on the correlation coefficient (CC) of impedance signatures Zagrai andGiurgiutiu (2001) investigated several statistical methods including RMSD,mean absolute percentage deviation, covariance change, and correlationcoefficient deviation The impedance-based methods have been successfullyapplied to truss structures (Sun et al 1995), steel bridge sections (Ayres et al.1998), composite patches (Chaudhry et al 1996), concrete bridges (Soh et al.2000), steel structures (Park et al 2005), thin plates and aerospace structures

(Giurgiutiu and Zagrai 2005), prestress-loss in PSC girders (Kim et al 2010),and bolted connection in full-scale steel bridge (Min et al 2010).

The impedance-based SHM methods are useful to detect damage incritical local members Those methods are sensitive to small incipientdamage; however, there are several drawbacks needed to be overcome: 1)they have limited sensing ranges that confine their practical usages in largestructures, 2) the equipment used to measure impedance is bulky, expensiveand not suitable for deployment on typical environmental conditions, and 3)

an impedance measurement might be made with a damaged PZT patch whichmight give a false positive indication on structural damage (Kim et al 2010)

In summary, the vibration and impedance-based SHM strategies arecomplementary each other since one is efficient for global damagemonitoring while other is good at local damage detection The combination

of those two strategies can provided an efficient way to detect the multi-types

of damage in complex structures In cable-stayed bridge, for example, thevibration-based SHM methods are capable to recognize the global damageoccurrence in deck and pylon, and to detect the local cable force loss Theimpedance-based SHM methods are capable to recognize the local incipientdamage in the cable-anchorage subsystems Therefore, the suitable vibrationand impedance-based SHM methods need to be selected for damagemonitoring in cable-stayed bridge

Smart Sensor System for SHM

Recently, smart sensing technologies of vibration and impedance usingwireless sensors have been developed by many research groups The costassociated with wired SHM systems can be greatly reduced by adoptingwireless, smart sensing technologies Straser and Kiremidjian (1998) firstproposed a design of a low-cost wireless modular monitoring system for

SHM applications Since then, many researchers have developed wirelesssensors based on a variety of sensor platforms as shown in Fig 1.1 (Lynch et

al 2003, Lynch et al 2004, Spencer et al 2004, Kurata et al 2005, Sandoval et al 2006, Lynch and Loh 2006, Lynch et al 2006, Mascarenas et

Ruiz-al 2007, Zimmerman et Ruiz-al 2008, Nagayama et Ruiz-al 2009, Rice et Ruiz-al 2010,Kim et al 2011, Spencer and Cho 2011) Nagayama (2007) investigated theeffects of data loss and reliable communication of wireless sensors on SHMapplications Sim et al (2009) proposed an automated decentralized approachfor modal analysis using smart sensors

Furthermore, smart devices have been also designed to improve thesensing capacity of smart sensors Kim et al (2011) proposed smart PZT-interface as a complement device for impedance wireless sensor which hasthe measurable frequency range of 10 kHz - 100 kHz By using the interface,the impedance features were more sensitive to the change in structural systemand the frequency range was fixed Rice et al (2010) developed a multi-scalesensor board which can measure acceleration, temperature, and humidity.Since structural behaviors can be changed due to environmental conditionssuch as temperature and humidity, smart sensing devices should be improved

to cope with the restrictions during long-term operation Therefore, scale smart sensor nodes are needed to provide reliable SHM practices

multi-Lynch et al (2006)

Crossbow Mica2 (2004) SmartDust WeC (1999) EYES (2003)

Intel Imote (2004) Imote2 (2006)

Figure 1.1 Wireless Sensor Platforms

The previous works have been mainly focused on implementingwireless sensor nodes for SHM on lab-scale structures As listed in Table 1.1,there have been a few full-scale implementations of wireless sensors forbridge monitoring so far (Lynch et al 2003, Arms et al 2004, Lynch et al.

2006, Pakzad et al 2008, Hoult et al 2010, Spencer and Cho 2011).Additionally, solar energy harvesting system has been adopted to extend thelifetime of smart sensors, especially, for long-term autonomous monitoringapplications (Niyato et al 2007, Torah et al 2008, Casciati and Rossi 2007,Caetano et al 2008, Miller et al 2010)

Table 1.1 Full-scale Implementation of Wireless Sensors

platform

No of sensor

Sensor (channel)

Alamosa Canyon Bridge

(Lynch et al 2003)

Ben Franklin Bridge

(Arms et al 2004)

Microstrain SG-Link

histories Guemdang Bridge

(Lynch et al 2006)

mode shapes Golden Gate Bridge

(Weng et al 2008)

mode shapes Stork Bridge

(Feltrin et al 2010)

Ferriby Road Bridge

(Hoult et al 2010)

Inclin (3) Temp (7) Humid (6)

Long-term Crack growth,

inclination of deck

For field SHM applications, the use of batteries for powering to sensornodes has a disadvantage since it is inconvenient to regularly replace low-capacity batteries located within complex structural geometries On cable-stayed bridges, for example, sensor nodes installed on cables and top pylonsare difficult for replacing batteries Hence, solar power is one of the bestways to supply electricity for autonomous smart sensor nodes.

SHM on Cable-stayed Bridge

Recently, the interest on the safety assessment of cable-stayed bridgeshas been increasing The cable stayed bridge represents the potential socio-economic, and cultural symbolic of a nation For a cable-stayed bridge,critical damage may be occurred in main structural components such as deck,cable, and pylon by resulting in stiffness-loss, crack growth, concretedegradation, etc Critical damage in cable-anchorage subsystem may includecable force loss, anchorage damage and anchorage force loss Eventuallythese damages may result in local and global failure of the bridge system Toavoid the above-mentioned situations, therefore, the cable-stayed bridge must

be secured by a suitable SHM system that identifies damage occurrence andassesses the location and severity of damage on timely manner

Over the last decade, many research groups have worked on SHM oncable stayed bridges Koh et al (2005) reported the SHM results on twocable-stayed bridges in Korea as Seohae Bridge and Samcheonpo Bridge ForSeohae Bridge, more than 10 types of sensors with a total of 180 units wereinstalled The structural behaviors of the bridge were observed and analyzedduring 2 years after its completion The SHM results showed that verticaldeflection and stress in the stiffening girder satisfied the allowable designlimits Cable forces ranged within 95%-104% of the initial value attesting forthe stability of the bridge For Samcheonpo Bridge, a forced vibration test

was performed for system identification purpose just after its completion andbefore opening The excitation force was induced by a truck of 29.7 ton andacceleration responses were measured at 14 locations A total of seven naturalfrequencies and the corresponding mode shapes were extracted.

Many researchers have proposed model update methods for systemidentification by using vibration characteristics (Friswell and Mottershead

1995, Kim and Stubbs 1995, Stubbs and Kim 1996, Zhang et al 2000, Jaishiand Ren 2005, Yang and Chen 2009, Ho et al 2012) Among those methods,eigenvalue sensitivity-based algorithm has become one of the most popularand effective methods to produce the baseline model for structural healthassessment (Brownjohn et al 2001, Wu and Li 2004) However, theapplication of finite element (FE) model updating on cable-stayed bridgeshave rarely reported Zhang et al (2000) updated the FE model of Kap ShuiMun Bridge located in Hong Kong The model updating was carried outusing ambient vibration test results and sensitivity-based model updatingtechnique The natural frequency differences of the initial FE model up to28% have been reduced to less than 11% Also, Brownjohn and Xia (2000)updated the FE model of Safti Link cable-stayed bridge in Singapore Theupdated FE model still had the large natural frequency difference, about 9%.The application of Bridge Monitoring System (BRIMOS) on two cable-stayed bridges (Shandong Binzhou Yellow River Bridge and Harbin SonghuaRiver Bridge) in China was reported by Wenzel and Furtner (2006) In theirstudy, the advantages and disadvantages of permanent and period monitoringsystems in the sense of performance and costs were examined Zhang et al.(2007) reported the SHM of a long span cable-stayed bridge, Queshi Bridge,

in China Cable force, main structural deformation and vibration weremonitored with different methods over a few years Figure 1.2 shows twoexamples of wired SHM systems on cable-stayed bridges (Spencer and Cho

2011) The cost was evaluated for one sensing channel as $US 6,200 and $US10,750 for Sutong Bridge and Stonecutters Bridge, respectively It is obviousthat the cost associated with wired SHM system of cable-stayed bridges isvery high.

Main Span Length: 1,088 m

Sensor Channels: 600

Cost: $US 3.7 Million ($US 6,200 / Ch.)

Main Span Length: 1,018 m Sensor Channels: 1200 Cost: $US 12.9 Million ($US 10,750 / Ch.)

(a) Sutong Bridge (b) Stonecutters Bridge

Figure 1.2 SHM Systems of Cable-stayed Bridges

Weng et al (2008) presented a wireless SHM system installed on 240-mGi-Lu cable-stayed bridge in Taiwan Two output-only modal identificationmethods (i.e., stochastic subspace identification method and frequencydomain decomposition method) were used to extract dynamic characteristics

of the bridge from ambient vibration data A total of 10 modal frequenciesand their associated mode shapes were identified The interaction betweendeck and cables were also investigated Feltrin et al (2010) discussed severalbasic aspects of data processing and management for long-term monitoring ofStork Bridge using wireless sensor networks Spencer and Cho (2011)discussed recent advances in the deployment of wireless smart sensors, aswell as its successful deployment at full-scale the 2nd Jindo Bridge in Korea.The sensor network size was of 669 channels on 113 sensor nodes which iscurrently the world largest wireless sensor network for SHM Modalparameters, cable forces, and wind conditions have been monitored for long-term period

1.3 Objective Statement

Objective and Approaches

On the basis of the above mentioned research needs, the goal of thisstudy is to develop a multi-scale smart sensor system for vibration andimpedance-based SHM of pylon-cable-deck system in cable-stayed bridge.The following approaches are implemented to achieve the goal:

1) The theories of multi-scale responses of acceleration, PZT’s dynamicstrain and E/M impedance are described;

2) The multi-scale vibration and impedance responses of pylon-cable-decksystem in cable-stayed bridge are examined;

3) The vibration and impedance-based SHM methods are selected for thepylon-cable-deck system;

4) The multi-scale smart sensor system is designed for the vibration andimpedance-based SHM; and

5) The feasibility and the practicality of the multi-scale smart sensorsystem are evaluated on a cable-girder model and a real cable-stayedbridge (Hwamyung Bridge) in Korea

2) Secondly, although many SHM methods have been developed for civilinfrastructures, few successful attempts have been made to adapt theiralgorithms suitable for SHM in cable-stayed bridge

3) Thirdly, although a variety of wireless sensor system have beendeveloped, a multi-scale sensor system specified to wirelessly monitorboth global vibration responses (e.g., acceleration and PZT’s dynamicstrain) and local impedance response has not been developed yet.

4) Finally, although some researchers have made efforts on SHM in on-sitecable-stayed bridges using wireless sensor systems, more practicalefforts are still needed to better understand complex behaviors of thebridge, to improve the SHM methods, and to upgrade the smart sensorsystem

1.4 Organization of the Dissertation

The remainder of the work is divided into eight chapters, as follows:

In Chapter 2, theories of multi-scale vibration and impedance responsesfor SHM are examined Then, vibration and impedance features to beextracted for SHM purposes are outlined The feasibility of multi-scalevibration and impedance responses is evaluated on a lab-scale beam

In Chapter 3, multi-scale vibration and impedance responses of cable-deck system in cable-stayed bridge are examined Vibration andimpedance-based SHM schemes for cable-stayed bridge are outlined Then,multi-scale acceleration, PZT’s dynamic strain, and impedance responses ofpylon-cable-deck system are described Experimental analysis is performed

pylon-on a cable-girder model to examine the multi-scale resppylon-onses

In Chapter 4, vibration and impedance-based SHM methods selected forthe pylon-cable-deck system in cable-stayed bridge are described Fivevibration-based SHM methods utilizing several vibration features arepresented Two impedance-based SHM methods utilizing impedance featuresare described

In Chapter 5, the design of multi-scale smart sensor system for vibration

and impedance monitoring is described The hardware and embeddedsoftware are designed for the multi-scale sensor system The solar-powerharvesting unit is also implemented to the sensor system Then, the multi-scale smart sensor system is evaluated on a lab-scale beam to examine thelong-term monitoring capacities under various weather conditions.

In Chapter 6, the feasibility of the multi-scale smart sensor system isexperimentally evaluated on a cable-girder model Vibration and impedanceresponses measured from the target model are examined to verify thefeasibility of the multi-scale sensor system for SHM The cable force is alsoestimated from the monitored vibration and impedance signatures of themulti-scale sensor system

In Chapter 7, the practicality of the multi-scale smart sensor system isexperimentally evaluated on a real cable-stayed bridge The long-termmonitoring performance of the multi-scale smart sensor system isexperimentally analyzed under various cable forces and weather conditions.The experimental modal parameters of the target bridge are identified bynumerical modal analyses The structural parameters are estimated by themodal strain energy-based system identification using experimental modalparameters Finally, Chapter 8 states the summary and conclusion of thisstudy with some suggestions for future works

CHAPTER 2 THEORY OF MULTI-SCALE VIBRATION AND

IMPEDANCE RESPONSES

2.1 Overview

In this chapter, theories of multi-scale vibration and impedanceresponses are examined The following approaches are implemented Firstly,multi-scale vibration responses which include acceleration and PZT’sdynamic strain responses are theoretically described Secondly, electro-mechanical (E/M) impedance response is also theoretically described Thirdly,vibration and impedance features which are extracted from the vibration andimpedance responses are outlined, respectively Finally, the feasibility ofmulti-scale vibration and impedance responses is evaluated on a lab-scalebeam

2.2 Theories of Vibration Responses

2.2.1 Acceleration Responses

When a structure is vibrated, its motion can be described in terms ofdisplacement, velocity, and acceleration with respect to time Accelerationresponse depends on structural dynamic characteristics such as mass,stiffness, and damping which can be defined as (Clough and Penzien 1993)

characteristics that represent the structural behaviors Accelerometer isutilized to measure acceleration responses of a structure The mostwidespread type of accelerometers is the piezoelectric accelerometer For thisaccelerometer, the piezoelectric elements act as a spring and connect the base

of the accelerometer to a seismic mass When an input is introduced to thebase of the accelerometer, a force is created on piezoelectric material which

is proportional to the applied acceleration and the size of the seismic mass(Wilson 2005) The piezoelectric accelerometer provides high-fidelityacceleration due to its high-sensitivity

2.2.2 PZT’s Dynamic Strain Responses

Direct and Inverse Effects of Piezoelectric Materials

Over the last decade, piezoelectric materials have been widely adoptedfor SHM applications (Park et al 2003) The advantages of piezoelectricmaterial are cheap, lightweight, robust, and multi-form ranging from thinpatches to complex shapes (Giurgiutiu 2008) Piezoelectric materials arecommonly used as both sensor (direct effect) and actuator (inverse effect) forSHM applications A key characteristic of these materials is the utilization ofthe direct effect to sense structural deformation in addition to the inversepiezoelectric effect to actuate the structure

Figure 2.1 shows the concept of direct and inverse effects ofpiezoelectric materials As the direct effect shown in Fig 2.1(a), the electricaldisplacement (directly related to electrical current) is induced since amechanical stress field (or mechanical strain field) is applied to apiezoelectric sensor As the inverse effect shown in Fig 2.1(b), when anelectric field (directly related to electrical voltage) is applied to two surfaces

of a piezoelectric sensor, a strain field is induced According to these effects,

a piezoelectric sensor can be used as a sensor or an actuator

The constitutive equations in strain-charge relation for a piezoelectricmaterial can be expressed by tensor form as (Sirohi and Chopra 2000):

m

d im j ij

m

E km j

c jk

where D i is the electrical displacement; 

ij

e is the dielectric permittivity at

zero mechanical stress ( = 0); ij E j is the electric field vector; d

s is the elastic compliance of the piezoelectric material

at zero electric field (E j = 0)

V

Strain Voltage

V

Strain Voltage

(a) Direct Effect: Sensor (b) Inverse Effect: Actuator

Figure 2.1 Piezoelectric Material: Direct and Inverse Effects

Generally, there are many kinds of piezoelectric material which have theabove effects However, piezoelectric ceramic is widely adopted due to itslinear relationship between strain and electrical field Figure 2.2 shows therelationship between strain and electrical field for three materials includingpiezoelectric ceramic (PZT), electrostrictive ceramic (PMN), andmagnetostrictive compound (Terfenol D) (Giurgiutiu 2008) The relationshipbetween strain and electrical field of PZT is relatively linear as compared toTerfenol D and PMN Furthermore, electrical field of PZT is more sensitive

to strain, which is suitable to use a PZT patch as a sensor

Figure 2.2 Relationship between Strain and Electrical Field for PZT, PMN,

and Terfenol D (Giurgiutiu 2008)

PZT’s Dynamic Voltage versus PZT’s Dynamic Strain

As previously mentioned, piezoelectric material can be used as a sensorfor dynamic strain measurement Strain is measured in terms of the chargegenerated by PZT patch as a result of direct piezoelectric effect When apiezoelectric material is mechanically strained on a beam, an electrical field

is produced as shown in Figure 2.3 The constitutive relations of the PZTstrain for 1-D PZT-beam interaction:

1 31 3 33

3 31

d is the piezoelectric coupling constant; 1 is the stress in direction 1; 1

is the strain in direction 1; Y is the complex Young’s modulus of the zero- E

electric field

By using Eq (2.5), the dynamic strain is calculated from the dynamicvoltage However, in order to improve measurement accuracy, the output

voltage of PZT sensor should be passed through some signal conditioningcircuits (Sirohi and Chopra 2000) It should be noted that the derivation of Eq.(2.5) was based on the assumption that only 1-D strain contributed to thecharge generation and that there is no loss of strain in the bond layer Inreality, however, a transverse component of strain exists and there are somelosses in finite thickness bond layer (Bhalla et al 2009, Shanker et al 2011).

As a result, the strain value as calculated by Eq (2.5) is not the actualstrain measured by the strain gage Some correction factors should berequired to account for transverse strain and shear lag losses in the bond layer.Moreover, the effect of temperature on sensor characteristics should beconsidered for accuracy measurement of dynamic strain using PZT sensor.The properties of PZT sensor are very sensitive to temperature Therefore, thedynamic strain should be measured at known temperature conditions, then, itshould be modified by using appropriate values of the constants forcalibration

1

3 2

Host Structure

Figure 2.3 Schematic of PZT-Beam Interaction in Direct Piezoelectric Effect

Calibration Experiment for PZT’s Dynamic Strain

Experiments on a lab-scale beam were carried out to calibrate strainfrom output voltage of PZT sensor As shown in Figure 2.4, the test beam (i.e.,Test Beam A) is a 6006010 mm aluminum cantilever beam For dynamicvoltage measurement from PZT sensor, a PZT sensor, FT-20T-3.6A1

produced by APC International, Ltd, was installed at the fixed end location.The data acquisition system consists of DAQ card, terminal block and Laptopwith MATLAB software For dynamic strain measurement from electricalstrain gage (ESG), a TML FLA-5-11-1L strain gage was also placed at thefixed end location The data acquisition system consists of TML SB120Bbridge box, Kyowa EDX-100A universal recorder and Laptop with DCS-100A software The impact force was applied by a hammer at a location 180

mm distanced from the free end

600

Excitation

PZT

Unit: mm ESG

180

PZT

ESG

Figure 2.4 Calibration Test Setup for PZT’s Dynamic Strain in Test Beam A

Dynamic responses (i.e., PZT’s voltage and ESG’s strain) weremeasured at the same time with sampling frequency of 1 kHz Figure 2.5shows the time history responses measured by PZT sensor and ESG The

maximum voltage (V max) and the corresponding maximum strain ( max) weredetermined from the time history responses Note that eight ensembles ofPZT’s voltage and ESG’s strain were recorded as listed in Table 2.1 Then,scale factors between PZT’s voltage and ESG’s strain was calculated Asgiven in Table 2.1, the scale factors were almost consistent with a mean value

of 17.5 As illustrated in Fig 2.6, the PZT’s voltage and ESG’s strain have alinear relationship

Table 2.1 Calibration for PZT’s Dynamic Strain in Test Beam A

2.3 Theory of Impedance Response

The electro-mechanical (E/M) impedance response is based on thecoupling of mechanical and electrical characteristics (Liang et al 1996) Inthis method, a piezoelectric patch is usually surface-bonded to a hoststructure The electrical effect of piezoelectric is partly controlled bymechanical effect of host structure As shown in Figure 2.7, the interactionbetween the piezoelectric patch and the host structure is conceptuallyexplained as an idealized 1-D electro-mechanical relation The host structure

is described as the effects of mass, stiffness, damping, and boundarycondition The PZT patch is modeled as a short circuit powered by aharmonic voltage or current

Figure 2.7 1-D Model Electro-Mechanical Interaction between Piezoelectric

Patch and Host Structure (Liang et al 1996)

When a PZT patch is surface-bonded to a structure, the electricaladmittance (the inverse of electro-mechanical impedance Z  ) of the patch,

s a

a E

p

p p

kl

kl Y

d Z Z

Z Y

d e t

l w j

31 11

2 31

k   11 is the wave number that depends on mass density  and

Young’s modulus Y11E of the piezoelectric material; and w p, l p, and t p

are the width, length, and thickness of the piezoelectric transducer,

respectively The parameters  and  are the structural damping loss

factor and the dielectric loss factor of piezoelectric material, respectively

In Eq (2.6), the first term of the equation is the capacitive admittance ofthe free piezoelectric patch The second term includes the mechanicalimpedance of both the piezoelectric patch and the host structure Themechanical impedance of the host structure Z s  is the ratio of PZT force,

PZT

e

F x

f Z

any changes in dynamic characteristics of the structure could be represented

in the change in E/M impedance

2.4 Vibration and Impedance Features

2.4.1 Vibration Features

In this section, several vibration features that can be extracted from bothacceleration response and PZT’s dynamic strain response are brieflydescribed The vibration features include: 1) power spectral density, 2)natural frequency and mode shape, and 3) modal strain energy

Power Spectral Density

Power spectral density (PSD) represents a positive real function of afrequency variable associated with a stationary stochastic process It shows atwhich frequencies variations are strong and at which frequencies variationsare weak Power spectral density of the vibration signals may contain lessnoise than an ordinary fast Fourier transform (FFT) result since it iscomputed from the average of FFT results For a vibration signal x (t), thePSD, S xx ( f), is calculated from Welch’s procedure as follows (Bendat andPiersol 1993):

S

1

2),(

1)

Natural Frequency and Mode Shape

Natural frequency and mode shape are vibration characteristics of astructural system The change in these modal parameters can be utilized forestimation of the change in structural properties Natural frequency generally