Long- Term Structural Performance
Monitoring of Bridges
Phase II: Development of Baseline Model and Methodology for
Health Monitoring and Damage Assessment
Final Report
Report CA07- 0245
December 2008
Division of Research
& Innovation
Long- Term Structural Performance Monitoring of Bridges
Phase II: Development of Baseline Model and Methodology for Health Monitoring and Damage Assessment
Final Report
Report No. CA07- 0245
December 2008
Prepared By:
Department of Civil and Environmental Engineering
University of California, Irvine
Irvine, CA 92697
Prepared For:
California Department of Transportation
Engineering Services Center
1801 30th Street
Sacramento, CA 95816
California Department of Transportation
Division of Research and Innovation, MS- 83
1227 O Street
Sacramento, CA 95814
DISCLAIMER STATEMENT
This document is disseminated in the interest of information exchange. The contents of this report
reflect the views of the authors who are responsible for the facts and accuracy of the data presented
herein. The contents do not necessarily reflect the official views or policies of the State of California
or the Federal Highway Administration. This publication does not constitute a standard,
specification or regulation. This report does not constitute an endorsement by the Department of
any product described herein.
STATE OF CALIFORNIA DEPARTMENT OF TRANSPORTATION
TECHNICAL REPORT DOCUMENTATION PAGE
TR0003 ( REV. 10/ 98)
1. REPORT NUMBER
CA07- 0245
2. GOVERNMENT ASSOCIATION NUMBER
3. RECIPIENT’S CATALOG NUMBER
4. TITLE AND SUBTITLE
Long- Term Structural Performance Monitoring of Bridges
Phase II: Development of Baseline Model and Methodology for Health
Monitoring and Damage Assessment
5. REPORT DATE
December, 2008
6. PERFORMING ORGANIZATION CODE
7. AUTHOR( S)
Maria Q. Feng, Yoshio Fukuda, Yangbo Chen, Serdar Soyoz, Sungchil Lee
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Department of Civil and Environmental Engineering
University of California, Irvine
Irvine, CA 92697
10. WORK UNIT NUMBER
11. CONTRACT OR GRANT NUMBER
DRI Research Task No. 0245
Contract No. 59A0311
12. SPONSORING AGENCY AND ADDRESS
California Department of Transportation
Engineering Services Center
1801 30th Street
Sacramento, CA 95816
California Department of Transportation
Division of Research and Innovation, MS- 83
1227 O Street
Sacramento, CA 95814
13. TYPE OF REPORT AND PERIOD COVERED
Final Report
14. SPONSORING AGENCY CODE
913
15. SUPPLEMENTAL NOTES
.
16. ABSTRACT
This project explores the use of the sensor technology for long- term bridge structural health monitoring. In Phase I of
the project, sensors were installed on two highway bridges, and vibration data analysis was reported in a Caltrans
technical report. In this Phase- II study, an additional highway bridge was instrumented, but the focus was on to develop
methodologies for analyzing the sensor data and diagnosing the on- going “ health” of the structure. Stiffness of structural
elements of a bridge is considered as an indicator of structural “ health”. As a structure deteriorates due to aging or suffers
from damage caused by extreme loads such as earthquakes, stiffness of the damaged structural elements would
decrease, and as a result, the global vibration characteristics of the structure would change. Therefore, by monitoring the
structural vibration, one can identify the change in structural vibration characteristics and then further identify the element
stiffness. A number of system identification methods were developed in this study for identifying the structural element
stiffness based on measurement of bridge vibrations caused by traffic and seismic excitations. A unique traffic excitation
model was proposed for more reliable stiffness identification based on traffic- induced vibrations. The effectiveness of
these methods in evaluating seismic damage on a bridge structure was demonstrated through seismic shaking table tests
of a multi- bent multi- column concrete bridge model. Long- term monitoring data from the instrumented bridges were
analyzed and developed into a structural stiffness database using a software platform developed in this study.
17. KEY WORDS
Structural Health Monitoring, Stiffness Identification,
Database, Shaking Table Test
18. DISTRIBUTION STATEMENT
No restrictions. This document is available to the public
through the National Technical Information Service,
Springfield, VA 22161
19. SECURITY CLASSIFICATION ( of this report)
Unclassified
20. NUMBER OF PAGES
252 Pages
21. PRICE
Reproduction of completed page authorized
LONG- TERM STRUCTURAL PERFORMANCE
MONITORING OF BRIDGES
Phase II: Development of Baseline Model and Methodology
for Health Monitoring and Damage Assessment
- Report to the California Department of Transportation –
By
Maria Q. Feng, Professor
Yoshio Fukuda, Post- doctoral Researcher
Yangbo Chen, Serdar Soyoz, and Sungchil Lee, Graduate Students
Department of Civil & Environmental Engineering
University of California, Irvine
October 31, 2006
ii
STATE OF CALIFORNIA ⋅ DEPARTMENT OF TRASPORTATION
TECHNICAL REPORT DOCUMENTAION PAGE
TR0003 ( REV. 9/ 99)
1. REPORT NUMBER
2. GOVERNMENT ASSOCIATION NUMBER 3. RECIPIENT’S CATALOG NUMBER
5. REPORT DATE
October, 2006
4. TITLE AND SUBTITLE
LONG- TERM STRUCTURAL PERFORMANCE
MONITORING OF BRIDGES
6. PERFORMING ORGANIZATION CODE
UC Irvine
7. AUTHOR
Maria Q. Feng, Yoshio Fukuda, Yangbo Chen, Serdar Soyoz,
and Sungchil Lee
8. PERFORMING ORGANIZATION REPORT
NO.
10. WORK UNIT NUMBER
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Civil and Environmental Engineering
E4120 Engineering Gateway
University of California, Irvine
Irvine, CA 92697- 2175
11. CONTACT OR GRANT NUMBER
13. TYPE OF REPORT AND PERIOD
COVERED
Final Report
12. SPONSORING AGENCY AND ADDRESS
California Department of Transportation ( Caltrans)
Sacramento, CA 14. SPONSORING AGENT CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
This project explores the use of the sensor technology for long- term bridge structural health
monitoring. In Phase I of the project, sensors were installed on two highway bridges, and vibration
data analysis was reported in a Catlrans technical report. In this Phase- II study, an additional
highway bridge was instrumented, but the focus was on to develope methodologies for analyzing the
sensor data and diagnosing the on- going “ health” of the structure. Stiffness of structural elements of
a bridge is considered as an indicator of structural “ health”. As a structure deteriorates due to aging
or suffers from damage caused by extreme loads such as earthquakes, stiffness of the damaged
structural elements would decrease, and as a result, the global vibration characteristics of the
structure would change. Therefore, by monitoring the structural vibration, one can identify the
change in structural vibration characteristics and then further identify the element stiffness. A
number of system identification methods were developed in this study for identifying the structural
element stiffness based on measurement of bridge vibrations caused by traffic and seismic
excitations. A unique traffic excitation model was proposed for more reliable stiffness identification
based on traffic- induced vibrations. The effectiveness of these methods in evaluating seismic
damage on a bridge structure was demonstrated through seismic shaking table tests of a multi- bent
multi- column concrete bridge model. Long- term monitoring data from the instrumented bridges
were analyzed and developed into a structural stiffness database using a software platform
developed in this study.
17. KEYWORDS
Structural Health Monitoring, Stiffness
Identification, Database, Shaking table
test
18. DISTRIBUTION STATEMENT
No restrictions.
19. SECURITY CLASSIFICATION ( of this report)
Unclassified
20. NUMBER OF PAGES 21. COST OF REPORT CHARGED
iii
DISCLAIMER: The contents of this report reflect the views of
the authors who are responsible for the facts and
the accuracy of the data presented herein. The
contents do not necessarily reflect the official
views or policies of the State of California or the
Federal Highway Administration. This report
does not constitute a standard, specification or
regulation.
The United States Government does not endorse
products or manufacturers. Trade and
manufacturers’ names appear in this report only
because they are considered essential to the
object of the document.
iv
SUMMARY
This project explores the use of the sensor technology for long- term bridge structural
health monitoring. In Phase I of the project, accelerometers and other types of
sensors were installed on two new highway bridges in Orange County, CA, and
vibration measurement data were analyzed, as reported to Caltrans by Feng and Kim
( 2001). In this Phase- II study, an additional highway bridge was instrumented with
sensors, but the focus is on the development of methodologies for analyzing the
vibration data gathered by the sensors and, based on the results, diagnosing the on-going
“ health” of the structure. In this study, the stiffness of structural elements of
the bridge structure is considered as an indication of structural “ health”. As a
structure deteriorates due to aging or suffers from damage caused by extreme loads
such as earthquakes, stiffness of the damaged structural elements would decrease, and
as a result, the global vibration characteristics of the structure would change.
Therefore, by monitoring the structural vibration, one can identify the change in
structural vibration characteristics and furthermore change in element stiffness. As
the change exceeds certain threshold, the structure is considered damaged.
Identification of structural stiffness enables assessment of not only extent but also
locations of structural damage. A stiffness established when the structure is new can
be used as a baseline for long- term structural health monitoring. The baseline based
on vibration measurement can also be used for verifying the design of the structure.
In this report, Chapter 1 provides background information for the project and Chapter
2 a literature review on the sensor- based monitoring technology. Chapter 3 describes
v
the installation of sensors, data logs, and communication devices on the three
highway bridges. Chapter 4 presents acceleration data recorded at the instrumented
bridges including traffic- and earthquake- induced vibration data, including 1712 sets
of traffic- induced vibration data collected at the Jamboree Road Overcrossing.
Chapter 5 presents measured strain data and comparison between the results obtained
from strain and acceleration data analyses.
Chapter 6 develops a number of methods and algorithms for identifying element
stiffness of a bridge structure based on vibration measurement. The chapter is divided
into two sections: one focuses on bridge super- structures and the other on bridge sub-structures
( columns). The super- structure stiffness is identified based on traffic
excitations, while sub- structure stiffness is identified by earthquake excitations for the
purpose of seismic damage assessment and thus nonlinear identification methods
become necessary. A Bayesian updating method and a neural network method were
developed for identifying super- structure stiffness based on traffic- induced vibration
measurement. In this regard, an innovative traffic excitation model was proposed in
this study that is more realistic and thus reliable than the conventional white noise
model because of the use of available traffic information from video monitoring. For
identifying bridge column stiffness, the neural network method and the extended
Kalman filter method were formulated based on earthquake vibration data. These
system identification methods can also be categorized as frequency- domain and time-domain
methods. Some of these methods can be performed online in real time and
deal with nonlinear structural response.
Chapter 7 presents experimental validation of the system identification methods
developed in Chapter 6. Two shaking table tests were performed on large- scale
concrete bridge models involving multiple bents and multiple columns, progressively
inducing seismic damage to the models. The stiffness reduction at the bridge columns
vi
identified based on the measured vibration data is consistent with the actual damage,
in terms of the damage extents and locations. These shaking table tests represent the
first effort in experimentally validating such damage identification methods using
realistic structural models suffering from realistic seismic damage.
In Chapter 8, a vibration test under controlled traffic excitation was performed on one
of the instrumented bridges, and the results were compared with those obtained from
a similar test performed when the bridge was new. Chapter 9 develops a database for
modal frequencies and furthermore for element stiffness values for each of the
instrumented bridges based on vibration measurement. The stiffness values were
identified using the neural network- based system identification method, and the
results are consistent with those identified by the other methods presented in Chapter
6. Variations in the identified frequencies ( as large as +/- 10%) and in stiffness values
( as large as 6%) for one of the bridges were observed over a four- year period, due to
change in environmental conditions. From the element stiffness database, it is
observed that the superstructure stiffness identified from vibration measurement
fluctuates around 95% of the design values, while the column stiffness is 85% of the
design value. Similar observations were made for the other two bridges. The
stiffness database serves as the baseline for long- term monitoring.
Chapter 10 describes a software platform developed in this project based on the
stiffness identification methods developed in this study for automated data processing,
analysis, stiffness identification, baseline updating, and database development.
Finally, Chapter 11 summarizes the conclusions made in this project and suggests
important topics for future study such as estimation of remaining capacity of bridges
based on stiffness evaluation.
vii
viii
List of Figures
Figure 1.1.1 Visual Inspection
Figure 3.1.1 Sensors on JRO
Figure 3.1.2 Web Pages of the Wireless Bridges
Figure 3.1.3 Directional Antennas
Figure 3.1.4 TS- Terminal Software
Figure 3.1.5 Server Software
Figure 3.1.6 Java Applet – Client Software
Figure 3.1.7 Battery Box
Figure 3.2.1 Configuration of Wireless Transmission
Figure 3.3.1 Location of the FROO
Figure 3.3.2 Plan view of the FROO
Figure 3.3.3 Elevation of the FROO
Figure 3.3.4 Bearing at Abutment 1
Figure 3.3.5 Typical Cross- section of the Box- girder
Figure 3.3.6 Schematic Layout of the Accelerometers
Figure 3.3.7 Picture of a Uni- axial Accelerometer
Figure 3.3.8 Pictures of a Bi- axial Accelerometer
Figure 3.3.9 Pictures of a Tri- axial Accelerometer
Figure 3.3.10 Pictures Accelerometers Mounted in the Box- girder
Figure 3.3.11 Picture of the Data- logger Container and the Junction Box
Figure 3.3.12 Data- logger and the Strain Meter Conditioner in the Container
Figure 3.3.13 Accelerometer Splicing in the Junction Box
Figure 3.3.14 LVDT Type Strain Meters Installation
Figure 3.3.15 Position of the Strain Meters on the Deck
Figure 3.3.16 Strain Meter Splicing in the Junction Box
ix
Figure 3.3.17 Conventional Strain Gauge and LVDT Strain Meter installed in
Column 3
Figure 3.3.18 Conventional Strain Gauges Installed in Footings
Figure 3.3.19 Installation of Signal Box for Conventional Strain Gauges
Figure 3.3.20 Soil Pressure Sensor
Figure 3.3.21 Backfilling at Abutment 1
Figure 3.3.22 Displacement Sensor
Figure 3.3.23 Thermocouple and Signal Conditioner
Figure 3.3.24 Thermocouple Wiring Map
Figure 3.3.25 Map of Channel Assignment of the Data- logger
Figure 4.1.1 Traffic- Induced Vibrations Saved at JRO
Figure 4.1.2 Traffic- Induced Vibrations Streamed at JRO
Figure. 4.2.1 Instrumental Intensity Map by Shake- Map
Figure 4.2.2 Earthquake Response Records at JRO
Figure 4.3.1 Traffic- Induced Vibrations Saved at WSO
Figure 4.4.1 Ground Motion of a Moderate Earthquake at WSO
Figure 4.5.1 Traffic- Induced Vibrations Saved at FROO
Figure 5.1.1 Strain Gauge Locations of WSO
Figure 5.2.1 Influence Line for R1 and R10
Figure 5.2.2 Strain Time History of R1
Figure 5.2.3 Strain Time History of R10
Figure. 5.2.4 Influence Line of R2
Figure. 5.2.5 Influence Line of R3
Figure. 5.2.6 Strain Time History of R2
Figure. 5.2.7 Strain Time History of R3
Figure. 5.2.8 Influence Line of R4
Figure. 5.2.9 Strain Time History of R4
Figure. 5.2.10 Strain Time History of R5
x
Figure. 5.2.11 Strain Time History of R6
Figure. 5.2.12 Influence Line of R7
Figure. 5.2.13 Influence Line of R8 and R9
Figure. 5.2.14 Strain Time History of R7
Figure. 5.2.15 Strain Time History of R8
Figure. 5.2.16 Strain Time History of R9
Figure. 5.3.1 Maximum Strain of R1 and R10
Figure. 5.3.2 Maximum Strain of R2 and R3
Figure. 5.3.3 Maximum and Minimum Strain of R4
Figure. 5.3.4 Maximum Strain of R5 and Minimum Strain of R6
Figure. 5.3.5 Maximum Strain of R6 and Minimum Strain of R5
Figure. 5.3.6 Minimum Strain of R7
Figure. 5.3.7 Maximum Strain of R8 and R9
Figure. 5.3.8 Finite Element Model of WSO
Figure. 5.3.9 HS 20- 44 Load
Figure. 5.3.10 Lanes for Moving Load Analysis
Figure. 5.3.11 Strain R1 and R10 due to Vehicle Locations
Figure. 5.3.12 Strain R2 and R3 due to Vehicle Locations
Figure. 5.3.13 Strain R4 due to Vehicle Locations
Figure. 5.3.14 Strain R5 due to Vehicle Locations
Figure. 5.3.15 Strain R6 due to Vehicle Locations
Figure. 5.3.16 Strain R7, R8, and R9 from Finite Element Analysis
Figure. 5.3.17 Long- term Monitored Data of R5
Figure. 5.3.18 Monitored Time Histories and Their Power Spectral Density
( 6/ 28/ 2006)
Figure. 5.3.19 Third Mode Shape ( Transverse Direction)
Figure. 5.3.20 The Fourth Mode Shape ( Transverse Direction)
Figure 6.2.1 System Identification Methodologies
xi
Figure 6.3.1 Beam- Moving Force Model
Figure 6.3.2 ENFs at Node i.
Figure 6.3.3 Geometry of a FE Model
Figure 6.3.4 cov[ 1 Q( ) Q( )]
i F t F s for Case 1: γ = 2 / s, SP = 1, v μ
= 20 m/ s, v σ = 5 m/ s
Figure 6.3.5 1 cov[ Q( ) Q( )]
i F t F s for Case 2: γ = 2 / s, SP = 1, v μ
= 30 m/ s, v σ = 5 m/ s
Figure 6.3.6 1 cov[ M( ) Q( )]
i F t F s for Case 1: γ = 2 / s, SP = 1, v μ
= 20 m/ s, v σ = 5 m/ s
Figure 6.3.7 1 cov[ M( ) Q( )]
i F t F s for Case 2: γ = 2 / s, SP = 1, v μ
= 30 m/ s, v σ = 5 m/ s
Figure 6.3.8 Distortion of the Response Spectrum
Figure 6.3.9 Captured Video Images of a Vehicle Traveling on the Bridge
Figure 6.3.10 Infer Parameters from Traffic Video Information
Figure 6.3.11 Sensor System and Finite Element Model of JRO
Figure 6.3.12 Elements from Excitation Covariance ΣF ( t)
Figure 6.3.13 Element of Predicted Response Covariance 1 ΣY( t− s)
Figure 6.3.14 Element of Experimental Response Covariance Y( )
ex Σ t− s
Figure 6.3.15 Results Using 0- 60 Second Data
Figure 6.3.16 Results Using Entire 0- 360 Second Data
Figure 6.3.17 Distribution of β at Selected Instants
Figure 6.4.1 Architecture of the Neural Network
Figure 7.1.1 Layout of the Specimen in Experiment 1
Figure 7.1.2 Reinforcement Details of the Specimen in Experiment 1
Figure 7.1.3 Schematic Plot of Experiment 1
Figure 7.1.4 Damage at Flared Portion of Columns
Figure 7.1.5 Design of the Bridge Specimen in Experiment 2
Figure 7.1.6 Design of the Post- tension Ducts in Experiment 2
Figure 7.1.7 Design of the Three Bents in Experiment 2
Figure 7.1.8 Illustration of Experiment 2
xii
Figure 7.1.9 Schematic Plot of the Sensor Layout in Experiment 2
Figure 7.1.10 Damage Observed at a Column of Bent 1
Figure. 7.1.11 Strain Measurements at Bent- 3 During Test 14
Figure 7.2.1 Measured Acceleration Responses at Ch- 5 to Ambient Excitations
Figure 7.2.2 Peak- Picking of Power Spectrum Density Functions
Figure 7.2.3 Free vibration II ( after 100% Sylmar) and Its Time- frequency Plot
Figure 7.2.4 Simulated and Measured Responses for 15% Sylmar Motion
Figure 7.2.5 Acceleration Responses at Ch- 4 to White Noise Disturbances at
Various Damage Stages
Figure 7.2.6 Simulated and Measured Ch- 4 Responses for WN- X- 2 Motion
Figure 7.2.7 Simulated and Measured Ch- 4 Responses for WN- X- 5 Motion
Figure. 7.3.1 Fourier Spectral Results Obtained from White Noise Input
Figure 7.3.2 Sensor Locations and System Identification Methodology
Figure 7.3.3 Time History Portion Used for the Identification
Figure 7.3.4 Identified Natural Frequency with Different Starting Points of the Ending
Portion
Figure 7.3.5 Identified Natural Frequencies Using End Portions
Figure. 7.3.6 Schematic View of Finite Element Model
Figure. 7.3.7 Comparison of Responses at DOF 10 for T- 13
Figure. 7.3.8 Stiffness Reduction During T- 13
Figure. 7.3.9 Comparison of Responses at DOF 10 for T- 13 ( Filtered)
Figure. 7.3.10 Stiffness Reduction During T- 14
Figure. 7.3.11 Stiffness Reduction During T- 15
Figure. 7.3.12 Stiffness Reduction During T- 19
Figure. 7.3.13 Comparison of Responses at DOF 10 for Test- 19
Figure 8.1.1 Test Vehicle Locations in Transverse Direction
Figure 8.1.2 Test Vehicle Locations in Longitudinal Direction
Figure 8.1.3 Axle Load of the Test Vehicle
xiii
Figure 8.1.4 Strain Time History of All Sensors
Figure 8.2.1 Exciting Force Location of Braking and Bumping Test
Figure 8.2.2 Locations of Fiber Optic Accelerometers
Figure 8.2.3 Acceleration Time History from Braking Test
Figure 8.2.4 PSD from Conventional Accelerometer
Figure 8.2.5 PSD from Fiber Optic Accelerometer
Figure 8.2.6 Mode Shape from Braking Test ( Vertical Direction)
Figure 8.2.7 Mode Shape from Braking Test ( Transverse Direction)
Figure 8.2.8 Time History of Acceleration of A10
Figure 8.2.9 Time History of Acceleration of A3
Figure 8.2.10 Time History of Acceleration of A5
Figure 8.2.11 PSD from Bumping Test
Figure 8.3.1 Time History of Strain at R3
Figure 8.3.2 Time History of Strain at R5
Figure 8.3.3 Time History of Strain at R8
Figure 8.3.4 Time History of Acceleration at A10
Figure 8.3.5 Time History of Acceleration at A5
Figure 8.3.6 Time History of Acceleration at A3
Figure 9.1.1 First Modal Identification Results
Figure 9.1.2 Second Modal Identification Results
Figure 9.1.3 Third Modal Identification Results
Figure 9.1.4 Fourth Modal Identification Results
Figure 9.1.5 Distribution of Data Recording Time
Figure 9.1.6 Typical Traffic- induced Accelerations in Middle of Spans 1 and 2
Figure 9.1.7 Mode Frequency Change at WSO
Figure 9.1.8 Ground Motion of the Yucaipa Earthquake
Figure 9.1.9 Frequency Response Functions of WSO under Earthquake
Figure 9.1.10 Finite Element Model for FROO
xiv
Figure 9.1.11 Mode Shapes of FROO
Figure 9.1.12 Typical Traffic- induced Accelerations
Figure 9.1.13 Mode Frequency Change at FROO
Figure. 9.1.14 Acceleration Time History of the FROO
Figure. 9.1.15 PSD from All Acceleration Channels
Figure. 9.1.16 PSD from Vertical Acceleration Channels
Figure. 9.1.17 PSD from Transverse Acceleration Channels
Figure 9.2.1 Architecture of the Neural Network
Figure 9.2.2 Column Stiffness Correction Coefficient
Figure 9.2.3 Superstructure Stiffness Correction Coefficient
Figure 10.1.1 Major Functions of the Software
Figure 10.1.2 Bridge Selection Window
Figure 10.1.3 Bridge Selection Window
Figure 10.1.4 Main Interface for the Software
Figure 10.2.1 Data Format Conversion Window
Figure 10.2.2 Time History Display Window
Figure 10.2.3 Data Combination Window
Figure 10.2.4 FDD Analysis Window
Figure 10.2.5 Neural Network Analysis Window
Figure 10.2.6 Animation Window
Figure 10.2.7 Location Map Window
Figure 10.2.8 Sensor Configuration Window
Figure 10.2.9 File Naming Window
xv
xvi
List of Tables
Table 3.3.1 Installation of Accelerometers
Table 3.3.2 Splicing and Wiring Map
Table 4.1.1 Summary of Triggered Data on the JRO
Table 4.1.2 Summary of Streamed Data on the JRO
Table 4.3.1 Summary of Triggered Data on the WSO
Table 4.4.1 Peak Ground Motion at the WSO
Table 4.5.1 Summary of Collected data on the FROO
Table 5.1.1 Strain Gauge Specification
Table 5.3.1 Measured Strain
Table 5.3.2 Strain from Measurement and Analysis
Table 6.3.1 Traffic Information Extracted from the Video Images
Table 7.1.1 Test Procedure
Table 7.2.1 Damping Ratios ζ
Table 7.2.2 Identified Correction Coefficients
Table 7.2.3 Comparison of the Modal Characteristics
Table 7.2.4 Identified Correction Coefficients
Table 7.3.1 Comparison of Identified Natural Frequency Using White Noise and End
Portions
Table 7.3.2 Finite Element Model Calibration
Table. 7.3.3 Comparison of First Modal Frequency
Table 8.1.1 Test Vehicle Locations in Longitudinal Direction
Table 8.1.2 Measured Strain from Static Test
Table 8.1.3 Computed Maximum Strain
Table 8.2.1 Mode Frequency from Braking Test
Table 8.2.2 Mode Frequency from Bumping Test
xvii
Table 8.3.1 Dynamic Load Cases
Table 8.3.2 Measured Dynamic Strain
Table 8.3.3 Comparison of Strain
Table 9.1.1 Modal Frequencies from Controlled Vibration Test
Table 9.1.2 Modal Frequencies Identified from Ambient Vibration Records
Table 9.1.3 Modal Frequencies Identified from the Yucaipa Earthquake
Table 9.1.4 Modal Frequencies of FROO
Table 9.1.5 Mode Frequencies Identified from Ambient Vibration Records
Table 9.2.1 Moments of Inertia of JRO
Table 9.2.2 Moments of Inertia of WSO
Table 9.2.3 Spring Stiffness of Abutment Boundary Conditions
Table 9.2.4 Training Patterns for the WSO
Table 9.2.5 Verification of Trained Neural Network
Table 9.2.6 Modal Frequencies Identified from Ambient Vibration Records
Table 9.2.7 Updated Moment of Inertia of Column and Superstructure
Table 9.2.8 Updated Spring Stiffness of Abutment Boundary Conditions
xviii
TABLE OF CONTENTS
TECHNICAL REPORT PAGE……………………………………………………..... ii
DISCLAMER…………………………………….…………………………………... iii
SUMMARY………………………………………………………………….……..... iv
LIST OF FIGURES………………………………………………………………… viii
LIST OF TABLES……………………………………….…………………………. xvi
TABLE OF CONTENTS…………………………………………..…………........ xviii
1. INTRODUCTION……………………………………………..……………........... 1
1.1 Vibration- Based Bridge Structural Health Monitoring: Concept and
Advantages……………………………………………………………….……….. 1
1.2 Objective and Scope………………………………………………………...… 5
1.3 Overview of Phase- I Work and Phase- II Report……………………………… 6
2. LITERATURE REVIEW………………………………………………….………. 8
3. MONITORING SYSTEM INSTALLATION AND UPGRADES………………. 15
3.1 Upgrades of Phase- I JRO Monitoring System………………………………. 15
3.1.1 Addition of Temporary Sensors……………………………………….. 15
3.1.2 Installation of Wireless Remote Data Acquisition System……………. 16
3.1.3 Development of Communication Software TS- Terminal……………... 20
3.1.4 Development of Server/ Client Solution for Real- Time Internet
Waveform Display and Data Acquisition…………………………………….…. 21
3.1.5 Power System Upgrade………………………………………..………. 22
xix
3.2 Upgrades of Phase- I WSO Monitoring System……………………………... 23
3.3 Instrumentation of the 3rd Bridge: FROO……………………………………. 24
3.3.1 Bridge Description…………………………………………………….. 24
3.3.2 Monitoring System Design and Installation…………………………… 29
4. VIBRATION DATA……………………………………………………………... 47
4.1 Ambient Vibration Data on JRO…………………………………………….. 48
4.1.1 Triggered Data…………...…………………………………………….. 48
4.1.2 Streamed Data……………………………..…………………………... 49
4.2 Moderate Earthquake on JRO……………………………………………..… 51
4.3 Ambient/ Traffic- Induced Vibration on WSO……………………………….. 53
4.4 Moderate Earthquake on WSO………………………………………………. 53
4.5 Ambient Vibration Data on FROO………………………………………….. 55
5. STRAIN DATA AND ANALYSIS………………………………………………. 57
5.1 Strain Sensors and Locations…………………………………………………. 57
5.2 Characteristics of Dynamic Strain Data………………………………………. 59
5.2.1 R1 and R10……………………………………………………………… 59
5.2.2 R2 and R3……………………………………………………………….. 60
5.2.3 R4, R5 and R6…………………………………………………………... 62
5.2.4 R7, R8 and R9…………………………………………………………... 64
5.3 Comparison of Measured and Computed Strain……………………………... 66
xx
5.3.1 Measured Maximum Strain…………………………………………….. 66
5.3.2 Finite Element Analysis under Design Live Load……………………... 72
5.3.3 Comparison of Strain Data……………………………………………... 80
5.4 Summary……………………………………………………………………... 84
6. DEVELOPMENT OF STRUCTURAL HEALTH MONITORING
METHODOLOGIES……………………………………………………………..….. 85
6.1 Definition of Structural Health and Damage………………………………… 85
6.2 System Identification Methodologies………………………………………... 86
6.3 Traffic Excitation Modeling and Super- Structure Condition Assessment…... 89
6.3.1 Output only System Identification……………………………………. 89
6.3.2 Physical Formulation of Traffic Loads on a Bridge…………………... 91
6.3.3 Traffic Excitation Covariance Model…………………………………. 95
6.3.4 Distortion on the Response Spectrum due to Spatially Correlated
Excitation…………………………………………………………………...…... 101
6.3.5 Video Based Traffic Monitoring and Processing……………………. 104
6.3.6 Structural Condition Assessment…………..………………………... 106
6.3.6.1 Bayesian Updating………………………………………..…. 107
6.3.6.2 Estimation of Response Covariance Matrix…………………. 111
6.3.6.3 Validation on a Test- bed Bridge…………………………...... 112
6.3.7 Summary…………………………………………………………….. 118
6.4 Sub- Structure Condition Assessment………………………………………. 121
6.4.1 Frequency Domain Identification……………………………………. 121
xxi
6.4.1.1 Least Squares Estimation for Modal Parameters……………. 121
6.4.1.2 Neural Network Based Identification………………………... 123
6.4.2 Time Domain Identification…………………………………………. 125
6.4.2.1 Least Squares Estimation for Structural Parameters………… 125
6.4.2.2 Extended Kalman Filter Based Identification……………….. 126
7. EXPERIMENTAL VERIFICATION OF METHODOLOGIES……………….. 131
7.1 Large- Scale Shake Table Test Verification………………………………… 131
7.1.1 Two Column Bent Test ( Experiment 1)……………………………… 131
7.1.2 Full Bridge Test ( Experiment 2)……………………………………... 134
7.2 Damage Identification Based on Low- Level Excitation…………………… 140
7.2.1 Frequency Domain Identification……………………………………. 141
7.2.2 Time Domain Identification…………………………………………. 143
7.2.2.1 Experiment 1………………………………………………... 143
7.2.2.2 Experiment 2………………………………………………... 145
7.3 Damage Identification Based on Earthquake Excitations…………………. 149
7.3.1 Frequency Domain Identification…………………………………… 149
7.3.1.1. Frequency Domain Identification Using White Noise Input.. 149
7.3.1.2. Frequency Domain Identification Using Ending Portions of
Earthquake Motions………………………..………………. 150
7.3.2 Time Domain Identification……………………………………….... 154
8. FIELD TEST ON WEST STREET ON- RAMP………………………………… 161
xxii
8.1 Static Load Test……………………………………………………………… 161
8.1.1 Load Cases…………………………………………………………….. 161
8.1.2 Test Vehicle……………………………………………………………. 163
8.1.3 Static Test Results and Comparison with Analysis……………………. 163
8.2 Braking and Bumping Tests………………………………………………….. 165
8.2.1 Test Procedure…………………………………………………………. 166
8.2.2 Braking Test Results…………………………………………………... 166
8.2.3 Bumping Test Results…………………………………………………. 172
8.3 Dynamic Load Test…………………………………………………………... 174
8.3.1 Dynamic Load Cases…………………………………………………... 174
8.3.2 Dynamic Strain Results………………………………………………... 174
8.3.3 Mode Frequency……………………………………………………….. 177
9. DEVELOPMENT OF DATABASE……………………………………………. 179
9.1 Database for Modal Parameters…………………………………………….. 179
9.1.1 JRO…………………………………………………………………… 179
9.1.2 WSO………………………………………………………………….. 182
9.1.3 FROO………………………………………………………………… 189
9.1.3.1 Finite Element Modeling and Analysis………………………. 189
9.1.3.2 Ambient Vibration Data and Modal Frequency Identification. 192
9.2 Database for Structural Parameters…………………………………………. 197
9.2.1 JRO…………………………………………………………………… 199
9.2.2 WSO…………………………………………………………………... 201
xxiii
9.3 Summary and Design Recommendations…………………………………… 206
10. DEVELOPMENT OF SOFTWARE…………………………………………… 207
10.1 List of Software Modules…………………………………………………. 207
10.2 Description of Usage of Modules…………………………………………. 211
11. CONCLUSIONS AND RECOMMENDED FUTURE WORK…..................... 216
11.1 Conclusions………………………………………………………………. 217
11.2 Recommended Future Work……………………………………………… 219
REFERENCES……………………………...……………………………………… 220
1
Chapter 1
INTRODUCTION
This chapter first describes motivations of this research project on long- term
performance monitoring of Caltrans highway bridges, by introducing the concept of
vibration- based highway bridge structural health monitoring and its potential
advantages. Then, this chapter summarizes the overall scope of this two- phased
project. As this is the second report focusing on the Phase- II research, the work
accomplished in Phase I of this project will then be briefly reviewed.
1.1 Vibration- Based Bridge Structural Health
Monitoring: Concept and Advantages
Structural condition assessment of highway bridges has long relied on visual
inspection ( Fig. 1.1.1, courtesy of FHWA), which involves subjective judgment of
inspectors and detects only local and visible flaws. The frequency of visual
inspection and the qualification of the inspectors are regulated by a standard, the
National Bridge Inspection Standards ( NBIS 1996). And the Federal Highway
Administration ( FHWA) Recoding and Coding Guide ( FHWA, 1995) was also
provided to guide the procedure including the condition ratings and the
documentation in current practice. Even with these provisions, a recent investigation
initiated by FHWA to examine the reliability of visual inspections reveals significant
2
variability in the structural condition assignments by inspectors ( Phares et al., 2004).
Moreover, visual inspection cannot quantitatively evaluate the strength and/ or
deformation capacity reservation of a bridge. Local defects or flaws might or might
not have a significant effect on the bridge global performance.
Figure 1.1.1 Visual Inspection
Sensor- based structural health monitoring can revolutionize the traditional way we
inspect structures, in a more timely, objective, and quantitative fashion. By installing
appropriate sensors at critical locations on a bridge structure, transmitting the sensor
data through a communications network, and analyzing the data through a software
platform, the location and severity of bridge deterioration and damage can be
automatically, remotely, and rapidly assessed, without sending inspection crews to the
site. As the sensor, networking, and communication technologies advance, the
sensor- based structural health monitoring ( SHM) has become an intensively
investigated subject ( e. g., Aktan et al, 1997; Doebling et al, 1998, Feng and Kim,
1998, Feng and Bhang, 1999, Aktan et al, 2000; Park, et al, 2001, Peeters et al, 2001,
3
Catbas and Aktan, 2002; Chang and Liu, 2003; Chen and Feng, 2003, Kim, et al,
2003, Sohn, et al, 2003, Feng, et al, 2004).
In addition to the potential benefits to bridge inspection and maintenance, sensor-based
monitoring results can also be used to verify the current bridge design
approaches and suggest future improvement. The monitoring results can be used for
making more scientific decisions in terms of prioritization of bridges for structural
retrofit and strengthening. Furthermore, the sensor- based continuous monitoring will
potentially enable real- time and remote post- event damage assessment of highway
bridges and early warning, significantly improving emergency response operations.
As a branch of the wide- ranging subjects in SHM research, many researchers seek to
measure the structural vibration behavior ( dynamic response of a structure with or
without measuring the exerting excitations), and infer from the vibration data the
level of structural global and/ or local integrity. This is partially because vibration
sensors ( such as accelerometers) can be easily attached to the surface of an existing
structure, compared with other sensors ( such as strain sensors) that require
embedment during the construction ( for concrete structures). The concept of
vibration- based SHM comes from a fact that, when the structure is subjected to
damage or deterioration, the stiffness of some structural components or the support
conditions will change, and as a result, the global vibration characteristics of the
structure will change accordingly. Therefore, by monitoring the vibration and
detecting changes in the vibration characteristics, and further interpreting such
changes in terms of element stiffness changes, one can assess quantitatively the
structural health condition. Besides its global and quantitative natures, vibration
4
monitoring is a nondestructive condition assessment method that can be implemented
continuously on highway bridges without interrupting traffic. This has made it
particularly attractive.
However, two major obstacles remain against successful implementations of the
vibration- based SHM in real- life bridge structures. One is the lack of low- cost high-performance
vibration sensors and data acquisition systems, the other is the lack of
proper methodologies to interpret vibration data in terms of structural integrity.
5
1.2 Research Object and Scope
The overall objective of this project is to explore the use of the sensor technology for
long- term bridge structural performance monitoring, by ( 1) demonstrating the
installation of sensor and monitoring systems on three typical highway bridges and ( II)
developing methodologies and software for vibration data analysis and interpretation.
As reported by Feng and Kim ( 2001), the Phase- I effort focuses on the
instrumentation of two highway bridges and preliminary data measurement and
analysis. The Phase- II research included the instrumentation of an additional
highway bridge, upgrade of communication links for the monitoring systems, but the
major focus was on the development of methods for interpreting bridge vibration data
into the on- going structural health, defined as element stiffness of the bridge structure
in this study. The methods are mainly based on bridge responses to traffic loads.
Using traffic- induced vibration data has a few practical advantages over other bridge
structural condition assessment methods: ( I) It does not interrupt traffic; ( II) It
captures the in- situ dynamic behavior of the bridge undergoing its normal service; ( III)
It can be performed continuously, scheduled periodically or triggered automatically
and ( IV) It requires no special experimental arrangement or a heavy shaker/ hammer.
During Phase II of the research, the authors obtained unique opportunities to verify
the SHM methods developed in this study by performing seismic shake table tests on
large- scale realistic bridge models. These experiments demonstrate that the proposed
vibration- based methods can quantitatively assess the bridge structural conditions,
locate the damage zone and provide a mean to evaluate the bridge remaining capacity.
6
1.3 Overview of Phase- I Work and Phase- II
Report
The Phase- I work of this project has been summarized in a Caltrans report by Feng
and Kim ( 2001). In Phase I, sensor systems for long- term structural performance
monitoring were installed on two new highway bridges in Orange County, California:
the Jamboree Road Overcrossing ( JRO) and the West Street On- Ramp ( WSO). They
include accelerometers, strain gauges, pressure sensors, displacement sensors,
installed or embedded at strategic locations of both super- and substructures. Data
recorders and power supplies were also installed at the bridge sites. Preliminary
vibration measurement and data analysis were performed on these two instrumented
bridges. On the JRO bridge, ambient or traffic- induced vibration data were collected,
based on which natural frequencies and mode shapes were extracted using peak-picking,
randomdec and frequency domain decomposition methods, assuming the
excitation is a spatially uncorrelated white noise process. These results were
compared with those obtained by the preliminary finite element analysis. On the
WSO bridge, braking and bumping vibration tests were carried out using a water
truck. Natural frequencies were derived using similar methods as for the JRO bridge.
The JRO bridge and the WSO bridge instrumented in Phase I, are short or medium
span reinforced concrete box girder bridges, where the mechanical properties of the
abutments, including its support condition, mass and interaction with soil and
foundations, and its constrain stiffness to the superstructure, have significant
influence of the bridge dynamic behavior. To enrich the spectrum of the monitoring
bridges, a 3rd bridge, the Fairview Road On- Ramp Overcrossing ( FROO), with longer
span length and more number of spans, was instrumented with a denser sensor system
in Phase II.
7
In Phase II, the existing monitoring system on the JRO and the WSO underwent
major upgrades to accommodate wireless remote data acquisition. Such upgrades
ease the data collection, and are highly valuable for establishing a database to monitor
the long- term behaviors of these bridges. They also enable on- line real- time data
visualization and sharing on the Internet.
This report documents the Phase- II study. A literature review on structural
instrumentation and performance monitoring in provided in Chapter 2. The
instrumentation of the FROO and the system upgrades in the JRO and WSO are
documented in Chapter 3. Recorded data from ambient vibration and due to
earthquakes are shown in Chapter 4. As stated before measurements are taken not
only from accelerometers but also from strain gauges. In chapter 5 results obtain from
strain measurement and analysis are discussed. More importantly, the Phase- II
research focus on the development methods for analyzing and interpreting the
vibration data into structural health. Chapter 6 describes the vibration- based SHM
methods proposed and developed in this study, and Chapter 7 documents the unique
shaking table tests performed in this study to verify the SHM methods. Chapter 8
discusses the field tests conducted on WSOO using water trucks under controlled
environments. It has been well known that the environmental changes have
considerable effects on modal identification results. Chapter 9 shows the variation in
modal identification results throughout the last four years. Chapter 10 describes the
software developed in this study that implements the proposed and developed SHM
methods. Finally, Chapter 11 summarizes this project by providing concluding
remarks and suggesting future research topics.
8
Chapter 2
LITERATURE REVIEW
Structural condition assessment of highway bridges has long relied on visual
inspection, which involves subjective judgment of inspectors and detects only local
and visible flaws. The frequency of visual inspection and the qualification of the
inspectors were regulated by a standard, the National Bridge Inspection Standards
( NBIS 1996). The Federal Highway Administration ( FHWA) Recoding and Coding
Guide ( FHWA, 1995) was also provided to guide the procedure including the
condition ratings and the documentation in current practice. Even with these
provisions, a recent investigation initiated by FHWA to examine the reliability of
visual inspections reveals significant variability in the structural condition
assignments by inspectors ( Phares et al., 2004). Moreover, visual inspection cannot
quantitatively evaluate the strength and/ or deformation capacity reservation of a
bridge.
In order to investigate the global structural condition of bridges in an automated,
continuous, objective and quantitative manner, structural health monitoring ( SHM)
has been promoted by researchers ( e. g., Aktan et al, 1997; Doebling et al, 1998, Feng
and Kim, 1998, Feng and Bhang, 1999, Aktan et al, 2000; Park, 2001, Peeters et al,
2001, Catbas and Aktan, 2002; Chang and Liu, 2003; Chen and Feng, 2003, Kim, et
al, 2003, Sohn, et al, 2003, Feng, et al, 2004). Recently, SHM has been an intensively
investigated subject.
9
As a branch of the wide- ranging efforts of SHM, many researchers seek to measure
the structural vibration behavior ( dynamic response of a structure with or without
measuring the exerting excitations), and infer from the vibration data the level of
structural integrity. Among many nondestructive evaluation methods, vibration
monitoring is one that can be implemented continuously on highway bridges without
interrupting traffic.
A thorough literature review on vibration- based SHM was first presented by Doebling
et al. ( 1996), summarizing hundreds of publications up to 1995. A four- level
hierarchy, namely, ( I) detecting the existence of damage, ( II) locating damaged
portions, ( III) evaluating the severity of damage and ( IV) predicting its future
consequences, proposed by Rytter ( 1993) and defined as the goals of SHM. Recently,
an updated review of the state was presented by Sohn et al. ( 2003), summarizing
publications from 1996 to 2001. This review interprets vibration- based SHM
following a statistical pattern recognition paradigm, consisting of a four- part process:
( I) operational evaluation, ( II) data acquisition, fusion, and cleansing, ( III) feature
extraction and information condensation, and ( IV) statistical model development for
feature discrimination. In this paradigm, features that are believed damage sensitive
are extracted from vibration data, and a pattern recognition procedure is employed to
classify the feature vectors to determine the existence, location and severity of
structural damage. While the important role of statistical methods in SHM was
recognized, the ultimate goal of SHM is still damage evaluation, as was defined by
the four- level hierarchy in the previous review and by Sikorsky ( 2005). In view of
difficulties associated with mathematical models ( often referring to finite element
models) of structural systems, especially the difficulty in quantifying the modeling
uncertainty and the bias due to modeling errors, the reviewers uphold methods that are
10
not based on such models as more attractive. However, difficulties of non- model-based
methods were also recognized, especially in quantifying the severity of damage
where a supervised learning mode is usually adopted. Training patterns have to be
generated by a mathematical model whose fidelity remains to be verified, because
data sets from a damaged structure are seldom obtained and if exist, not adequate to
cover all possible damage scenarios. A sufficient coverage on various scenarios by
the training patterns, nonetheless, is essential in the supervised learning procedure.
Research in vibration- based SHM has produced substantial literature, with many
conferences and journals held for information exchange and demonstration of
research results ( e. g. Ghanem and Shinozuka, 1995; Safak 1989; Safak 1991; Feng
and Kim, 1998, Feng and Bhang, 1999; Feng and Kim, 2001, Park, et al, 2001, Feng et
al, 2003, Kim, et al, 2003). These methods can be grouped into two depending on
whether the identification is carried out in frequency or time domain. If it is in
frequency domain, basically the changes in modal values; frequency, damping, shape,
are used as an indication of damage. However; if one wants to identify the changes
more in detail like changes in elemental stiffness, time domain identification methods
might be more appropriate. Time domain methods can be grouped into two depending
on whether they are purely data driven or they are incorporating finite element ( FE)
model. If it is aimed to determine the changes in the stiffness values, FE model must
always be used. Within time domain identification methods, the most common one is
the least squares estimation ( LSE). It is basically performing an optimization for the
parameters such as stiffness and damping so that the error between the measured and
the simulated responses is minimized. LSE is useful as a system identification
technique, when used in combination with a damage detection algorithm ( e. g., Stubbs
et al, 200). However, there are some drawbacks of LSE. Firstly, physical insight can
11
be easily lost and a local maximum can be chosen over a global one ( Udwadia, 1988).
Secondly, LSE is very time consuming and cannot be applied for “ on- line” SHM and
damage detection. To overcome this difficulty, the recursive least squares ( RLS)
technique is proposed so that any time varying property in a system caused by damage
can be tracked in real time. However in this case incorporation of FE is sacrificed, i. e.
it is purely data driven so change in the system parameters can be tracked but it is not
possible to link this to the change in structural stiffness and damping. Also, RLS is
susceptible to even low level of noise. As can be seen every method has some
drawbacks and is not effective for on- line identification of stiffness values under
realistic conditions.
Kalman filtering was a break- through in system engineering field when first proposed
four decades ago. It not only uses the data in a probabilistic sense but also gets
information from structural model ( Kalman, 1960). Results obtained by the Extended
Kalman Filter ( EKF) approach from simulated data and well defined models with
known damage scenarios were reported ( Yun and Shinozuka, 1980; Hoshiya and
Saito, 1984; Yang et al, 2005; Straser and Kiremidjian, 1996; Loh and Chung, 1993;
Loh and Tou, 1995, Ghanem and Ferro, 2006). However, applicability of the EKF
approach to civil engineering structures involving high uncertainties in structures and
loadings under realistic damaging events has not yet been studied.
Evident by these reviews and more recent papers ( e. g. Bolton et al., 2001; Hera, 2004;
Koh et al., 2003; Lam et al., 2004; Yang and Lin, 2005), despite significant efforts,
damage identification by SHM is still a highly challenging problem. When
implementing vibration- based SHM to real- life structures, the limitation of sensing
capacity ( e. g. spatial limitation due to insufficient number of sensors or prohibitive
positions of instrumentation, and temporal limitation due to insufficient sensor
12
frequency range and excitation bandwidth), and the operational and environmental
variations of the structures have significantly increased the difficulties.
Nonetheless, it is believed that part of the challenges in SHM can be attributed to a
scholars’ preference of an inductive, objective and entirely data- driven methodology.
A shift of epistemology from a purely inductive to a deductive- inductive hybrid
methodology might help to ease the problem and bring forward useful results. In the
deductive- inductive methodology, a priori knowledge, derived either from established
theories, engineering experiences, or even subjective postulations, is incorporated in a
probabilistic model of the structural system. In this model, the extent of knowledge
limitation is represented by the uncertainty of the model structure and parameters.
This model is subjected to correction or refinement based on sensor data, by first
deducing the expected vibration behaviors from the a priori model, and then
comparing them with the sensor observations and updating the model in a systematic
induction to reconcile the predicted and observed vibration. The advantage of this
approach is that gaps of necessary information not provided by sensor data are filled
in with the currently available best understanding of the system. Therefore, SHM is
no longer merely a means of nondestructive damage evaluation, but a procedure of
information collection to correct/ refine the probabilistic model of the structural
system so as to gradually diminish the system uncertainty.
The above methodology is essentially a Bayesian approach. This vision of SHM can
be traced back to Beck ( 1989), where a Bayesian framework was laid down for
structural system identification that selects the most probable model from a class of
models based on input/ output measurement. The major usage of this data- improved
model is for response prediction for future loads, which was shown asymptotically
correct as the sample size of measurement increases. Later in Beck and Katafygiotis
13
( 1998), this vision was formalized to not only update the model, but also assess the
uncertainties of the model itself and its predictions. This formulation addresses
explicitly the difficult problem in parameter identification: the inherent ill-conditioning
and non- uniqueness. If the a posteriori probability of the parameters has
mono- mode, the system is globally identifiable; or if it has multiple but distinct peaks,
the system is locally identifiable; when it has sustained support in a manifold within
the parameter space, the system is unidentifiable. In the latter two cases, prediction of
structural behaviors is still possible in this framework, using more than one candidate
model, but weighting their predictions according to their model a posteriori
probability. The last case was treated in Beck and Au ( 2002) using a Markov chain
Monte Carlo method. The Bayesian framework was extended in Beck and Yuen
( 2004) to address the modeling error issue arising when the ‘ true’ system is not within
the class of models being examined. Classes of models were compared based on the
Bayesian a posteriori probability, which was revealed to consist of two parts: one
appreciates the fitness of the model to the data, and the other appreciates the model
parsimony. The capacity of a data- updated model to predict in a probabilistic sense
the structural response to future loads was utilized to make a connection between
SHM results and structural reliability evaluation ( e. g., Park, et al, 1997, Papadimitriou
et al., 2001; Beck and Au, 2002). Solutions to the implemental difficulties in SHM
due to operational and environmental variations were suggested also in a Bayesian
framework. In Yuen et al. ( 2002) a time- domain Bayesian updating was proposed
when system inputs are not measured, and in Yuen and Beck ( 2003), the same
problem is addressed by a frequency- domain approach. In Vanik et al. ( 2000)
variation of modal parameters ( frequencies and mode shapes) was treated in a
Bayesian framework to set a probabilistic measure of the significance of modal
14
feature changes. Although damage identification is not the major concern of the
model updating procedure, it is also possible if damage can be defined quantitatively
in terms of parameter changes ( Yuen et al., 2004).
This approach is certainly model- dependant. However, it can be argued that models
are almost inevitable anyway in structural condition assessment ( e. g., in training
pattern generation) and in evaluation of current and expected future performance of a
structure. To minimize the disadvantage caused by modeling errors, one may need to
avoid a deterministic perspective of a model, but instead, use a probability measure to
represent modeling uncertainty.
15
Chapter 3
HARDWARE INSTALLATION
AND UPGRADES
This chapter reports the upgrades on the monitoring systems at the Jamboree Road
Overcrossing ( JRO) and the West Street On- Ramp ( WSO) that were installed during
the Phase- I study, and the instrumentation of the 3rd bridge, the Fairview Road On-
Ramp Overcrossing ( FROO).
3.1 Upgrades of Phase- I JRO Monitoring
System
The monitoring system at the JRO underwent the following major upgrades in Phase
II of this research.
3.1.1 Addition of Temporary Sensors
In the spring of 2002, four additional temporary accelerometers were installed on the
JRO bridge. The purpose of such additional instrumentation is two- folded: Firstly,
analysis of the vibration data obtained from the permanent sensors only shows that
the number of sensors is not sufficient for mode shape identification therefore
additional sensors are needed; and secondly, it is to obtain data comparable to the
initial data sets collected when temporary sensors were on the bridge at the very
16
beginning of the monitoring project. Channel 13 to 16 are the temporary
accelerometers added ( Fig. 3.1.1). Channel 16 was later found to be out of order.
Therefore, the JRO monitoring system currently has 14 accelerometers and one
displacement sensor ( Channel 12). Due to the limited funding, we could not install
sensors at all the desirable locations such as Abutment 1.
Figure 3.1.1 Sensors on JRO
3.1.2 Installation of Wireless Remote Data Acquisition System
To overcome a distance of 6 miles and remotely access the monitoring system and
verify its working conditions, a wireless data acquisition system was installed on the
JRO bridge during the first quarter in 2004. The system includes the following
acquired hardware and software developed in this project.
17
Hardware:
( 1) A pair of Cisco Aironet 350 Wireless Bridges, working in IEEE 802.11b Network
Standard, 2.4 to 2.497 GHz frequency range. One was installed at the bridge site,
mounted inside the existing data logger box, configured as the civil- eng- root end with
IP address 128.200.109.194. The other was installed in the facility room in
Engineering Tower at UC Irvine, configured as the civil- eng- nonroot end with IP
address 128.200.109.195. Figure 3.1.2 shows the web pages where the status of these
pair of devices are displayed and their working parameters can be configured by a
system administrator.
Figure 3.1.2 Web Pages of the Wireless Bridges
( 2) A pair of Cisco AIR- ANT3338 Aironet Antennas, with gain 21dBi, capable of
approximate range of 25 miles ( at 2Mbps) or 11.5 miles ( at 11Mbps). One was
mounted on top of a steel pole at the bridge site; the other was mounted on top of a
pole on the roof of Engineering Tower at UC Irvine ( Fig. 3.1.3). The steel pole at the
bridge site was designed and constructed by K. A. Wang & Associates. Inc.
18
Figure 3.1.3 Directional Antennas
( 3) A LAN converter provided by Tokyo Sokushin Co., Ltd, the sensor and data
logger maker, to connect the data logger to the Internet through it RS 232 series port.
The LAN converter is configured to listen on 128.200.109.205: 23, and connected to
the civil- eng- root wireless bridge. This LAN converter converts the data logger to a
TCP- IP device enabling the networking.
Software:
The first software used for this remote data acquisition system is TS- Terminal V2.4, a
wireless data acquisition software initially by Tokyo Sokushin Co. Ltd. ( Fig. 3.1.4).
With this software, virtually any computer running TS- Terminal and connected to the
Internet can access the data logger remotely. Data can be monitored almost in real
time on remote terminals. Data files on the flash memory card at the bridge site data
logger can be downloaded to the remote terminal and deleted from the flash memory
( a) Antenna mounted on a steel pole at JRO
( b) Antenna mounted on a pole at Engineering
Tower on UC Irvine Campus
19
card. The remote terminal can send commands to the data logger to trigger the
recording ( to the flash memory card only), calibrate the sensor and change the data
logger’s setting.
However, there several fatal problems were discovered in the is project with this
software: 1) A remote terminal running the TS- Terminal software cannot record real
time data stream on its own hard disk; 2) The stability of the software is not
satisfactory: especially, system frequently breaks down during downloading multiple
files from the flash memory card; and 3) most importantly, TS- Terminal was
developed as a remote terminal, not as a server, therefore, it supports only one online
user at one time and it does not support data visualization and distribution on internet.
Figure 3.1.4 TS- Terminal Software
20
3.1.3 Development of Communication Software TS- Terminal
Seeking a solution to the above problems of the TS- Terminal, a new software with a
remote data acquisition capability was developed by this team at UC Irvine based on
the platform of TS- Terminal. The newly developed software ( Fig. 3.1.5) has been
installed on a computer on UCI campus, and functions as a server that receives
streaming data from the data logger on the remote bridge site, saves them in the local
computer and buffers them for Internet publication. The new software has algorithm
to accommodate data transmission errors during wireless communication, thus
suffering much less interruptions during data transmission.
Figure 3.1.5 Server Software
21
3.1.4 Development of Server/ Client Solution for Real- time Internet
Waveform Display and Data Acquisition
Besides this server software, a Java applet was further developed in this project for
displaying real- time data on Internet. This Java applet is a client agent that displays
the waveforms of the data in the buffer of the server ( Fig. 3.1.6). This applet provides
a way for the public as well as Caltrans to view the real- time data on Internet. It is
available at http:// mfeng. calit2. uci. edu/ ( Special approval from Caltrans is needed for
downloading the data). This pair of server/ client software also provides a way to
verify the working status of the JRO monitoring system.
Figure 3.1.6 Java Applet – Client Software
3.1.5 Power System Upgrade
To provide sufficient power for the existing data logger, and also the devices added
for the wireless remote data acquisition system, two additional deep- cycle auto
rechargeable batteries were installed at JRO in the batteries box ( Fig. 3.1.7). A
transformer was used to provide DC 38V for the Cisco wireless bridge and the Cisco
22
AIR- ANT3338 Aironet Antenna by these two additional batteries. Three charging
controllers were integrated into the system to protect the batteries from over- charging
or discharging ( currently, these controllers are configured to auto- reset after several
hours if over- charging or discharging is detected).
Figure 3.1.7 Battery Box
3.2 Upgrades of Phase– I WSO Monitoring
System
During Phase- II of the research project, data retrieval of the WSO system has been
proven very difficult. The major difficulty comes from that fact that the data logger
was installed inside the box girder due to the unavailability of an easy- to- access
space. To access the data logger or to retrieve the data recorded in the memory card,
one needs to climb into the enclosed box- girder through a man hole. Entering such an
23
enclosure environment requires special training. To access the man hole, one needs a
ladder which requires a pick- up truck for its transference. For safety, accessing the
man hole is not recommended without proper guidance.
To cope with these problems, a wireless LAN router and a serial to LAN converter
were installed inside the box- girder of the WSO. Figure 3.2.1 shows a system
configuration of this wireless transmission setup. With this wireless transmission
setup, recorded vibration data can be retrieved from the outside box- girder of the
WSO.
Recorded vibration data is retrieved from the data logger through a serial
communication line. The serial to LAN converter, which is connected to the data
logger, converts this serial data to TCP/ IP format in order to connect to the wireless
LAN router. This converted data is transmitted to the commercially available
wireless LAN router, which is placed close to the man- hole in the box girder, by
wired connection. The wireless LAN router establishes a local area network by using
Data Logger
Serial connection LAN
Serial to LAN
converter
Wireless LAN
router
Figure 3.2.1 Configuration of Wireless Transmission
Notebook
computer
Outside box girder
Inside box girder
24
private IP address and broadcasts the vibration data to the outside box girder. A
notebook computer, which has a wireless NIC, can receive the broadcasted vibration
data from the wireless LAN router without entering the enclosed box- girder by
connecting established local area network. Although limitation of transmission
distance of the wireless LAN router is 50 [ m] according to its specification, it is
possible to extend this distance by installing a wireless access point to provide more
convenience.
This wireless transmission setup is working stably, and many vibration data has been
collected on the WSO wirelessly.
3.3 Instrumentation of the 3rd Bridge: FROO
During Phase II of the research project, a third bridge is instrumented with a sensor
system consisting of accelerometers, LVDT type strain meters and conventional strain
gauges, displacement meters, pressure sensors and thermocouples.
3.3.1 Bridge Description
The Fairview Road On- Ramp Overcrossing ( FROO), located in Costa Mesa, Orange
County, California, is the on- ramp of Fairview Road onto the north bound of I- 405
freeway, overcrossing the Harbor Boulevard off- ramp. Figure 3.3.1 is a map from
Google Local showing its location.
25
Figure 3.3.1 Location of the FROO
The FROO is a four- span continuous cast- in- place pre- stressed post- tension box-girder
bridge ( Fig. 3.3.2). The total length of the bridge is 224.0 m ( 734.9 ft.), in
which the lengths of spans are 52.5, 59.5, 59.5 and 52.5 m ( 172.2, 195.2, 195.2 and
172.2 ft), from span 1 to span 4 respectively ( Fig. 3.3.3). The bridge is supported on
three monolithic single columns and sliding bearings on both abutments. The sliding
bearings ( Fig. 3.3.4) allow creep, shrinkage, and thermal expansion or contraction.
The typical cross section of the box- girder is shown in Fig. 3.3.5.
Compared with the other two instrumented and monitored bridges ( the JRO and the
WSO), the FROO has more and longer spans. It enriches the spectrum of the
monitored bridges. Instrumentation of this bridge offers opportunities to study and
understand behaviors of longer span RC bridges where the abutments are expected to
affect relatively less on the overall bridge dynamic behaviors. It will be of great
interest to monitor and evaluate the long- term structural performance of such bridges
under not only seismic but also service loads, and to compare their performance with
that of the bridges with less and shorter spans.
Fairview Road On-
Ramp Overcorossing
26
During Phase II of this research project, the FROO was under going construction. It
was completed and opened to the traffic in 2004. Thus it provided excellent
opportunity for embedding strain sensors in concrete and pressure sensors in the
abutments during the construction. Accelerometers were mounted inside the bridge
box- girder for better protection. Based on the experience of data analysis of the other
two instrumented bridges, the FROO was instrumented with a denser sensor system
with more accelerometers and strain gauges in comparison with the JRO and WSO.
27
Figure 3.3.2 Plan view of the FROO
Figure 3.3.3 Elevation of the FROO
28
Figure 3.3.4 Bearing at Abutment 1
Figure 3.3.5 Typical Cross- section of the Box- girder
29
3.3.2 Monitoring System Design and Installation
Accelerometers
A total of 21 channels of acceleration sensors were installed both on the bridge super-and
substructures. As shown in Table 3.3.1 and Fig. 3.3.6, one tri- axial accelerometer
( A0), five bi- axial ( A2, A3, A5, A9 and A12), and eight uni- axial ( A1, A4, A6, A7, A8,
A10, A11 and A13) were installed. Pictures of a uni- axial, a bi- axial and a tri- axial
accelerometers are shown in Figures 3.3.7 to 3.3.9. Except for A0, which was installed
against the end wall at Abutment 1 to measure the ground motion in the three directions,
accelerometers ( A1 to A13) were mounted on the floor surface inside the box- girder, by
brackets bolted into the concrete ( as shown in Fig. 3.3.10), to measure the superstructure
vibration at different positions. A1 to A13 were aligned along the longitudinal center
line of the box- girder to mitigate the effect of torsional modes. Again due to the limited
funding, it was not possible to install additional sensors to measure the torsional modes.
We recommend to add more sensors to measure the torsional modes when funding
becomes available in the future. The positions of A1 to A13 are in respectively in the
middle and quarter points of the spans.
The positive directions follow the sign conventions as noted in Table 3.3.1, which
documents the orientations of the accelerometers. The sub- column ‘ Marked’ in column
‘ Direction’ documents the assigned directions by the sensor manufacturer that were
marked on the enclosure box of each sensor. The sub- column ‘ Planned’ denotes the
installation plan. However, due to the actual difficulties of installation, for example the
obstacle during concrete drilling, or the miss- match of the bracket orientation, the actual
‘ Installed’ orientation can be different from the plan. For example, the row for A0
should be interpreted as: we intended to install the accelerometer marked as (+ X) along
30
the positive longitudinal direction ( planned + X), but we ended up with installing it along
the negative vertical direction ( installed – Z); similarly, we ended up with installing the
accelerometer marked with (+ Y) and (+ Z) along the negative transverse (- Y) and positive
longitudinal (+ X) directions, respectively.
The cables of the accelerometers ( and those of the embedded strain meters) run inside the
box- girder and through the cap beams on top of the bents ( through pre- installed PVC
pipes). After the installation, the ceiling slabs of the box- girder were cast and there is no
access to the sensors on span 2 to 4. One accelerometer, A6, on span 2 was found to be
shorted somewhere inside the box- girder and thus not functional.
For the convenience of future system maintenance, this report documents the detailed
wiring and splicing maps used in the accelerometer installation. Figure 3.3.11 shows the
container box on a concrete pad and the junction box mounted on the wall of Abutment
1. The container box houses the data logger, the strain meter conditioner and the
uninterrupted power supply ( UPS) unit, as in Fig. 3.3.12. The 48- channel 22- bit A/ D
data logger provides the A/ D conversion and controls the triggering, timing, sampling,
recording and data streaming for all the sensors in this system. It also supplies DC ± 15V
power for the accelerometers. The 11- channel strain meter conditioner, on the other
hand, is for the strain gauges and the pressure sensors only ( this will be discussed in
detail later). The strain and pressure signals, conditioned by this device, are further
connected to the data logger for A/ D.
The cables of the sensors ( the accelerometers, strain gauges, pressure sensors,
displacement sensor and GPS antenna), going through the conduits, are spliced in the
junction box following Table 3.3.2. The spliced cables are then wired to either the data
logger or the strain meter conditioner, depending on the sensor types. The DC ± 15V
31
power is spliced inside the junction box to provide power for all the accelerometers.
Figure 3.3.13 documents the detail splicing of the accelerometer cables in the junction
box.
Strain Meters ( LVDT type)
Seven LVDT type strain meters were embedded in the bridge superstructure ( Fig.
3.3.14a), and three were embedded in Column 3 ( Fig. 3.3.14b). All these strain meters
were built on dummy rebars and attached to the steel cage before concrete casting. After
the concrete cured, the strain meters are assumed to develop deformations consistent with
the concrete surrounding them, thus measuring the strain of the concrete at that position.
Figure 3.3.15 shows the installation positions of the strain meters in the superstructure
( denoted as SD1 to SD7) . The purpose of installing these strain meters is to monitor the
evolution and the lose of pre- stress in the superstructure. Therefore, they were installed
along a pre- stressing tendon and aligned horizontally. The other three strain meters were
installed in Column 3 ( denoted as SC1 to SC3) at the same elevation, measuring vertical
strains at the three equally dividend points of the periphery of a circular cross section.
However, one of these three sensors ( either SC1 or SC2) was damaged during the
construction of the bridge. Nonetheless, the remaining sensors can still serve the major
purpose of this instrumentation: to obtain information of the static gravity load on
Column 3.
A strain meter conditioner supplies 5V DC power to the strain meters, and at the same
time, conditions the strain signals ( Channels 1 to 9 of the conditioner) before sampled by
a data log ( whose channel connection is documented in Table 3.3.2). A detailed cable
splicing map is documented in Fig. 3.3.16.
32
Strain Gauge ( Resister type)
In addition to the LVDT type strain meters, conventional resister type strain gauges were
also embedded in the substructure ( Figures 3.3.17 and 3.3.18). They are used to measure
strain distribution in the reinforced concrete footing of the columns ( Fig. 3.3.18) and as a
comparison to the LVDT strain meters in Column 3 ( Fig. 3.3.17). Conventional strain
gauges are not expected to last as long as the LVDT type strain meters, therefore not
wired to the data logger. A portable strain reader and a temperature compensator can be
used to acquire data from these strain gauges. Boxes housing the signal conditioner and
the data log were installed at the column surface above the ground ( Fig. 3.3.19).
Soil Pressure Sensors
Two soil pressure sensors ( P1 and P2, Fig. 3.3.20) were installed between the soil and the
end walls of Abutment 1 and Abutment 4, respectively. Sensor installation was
performed before the backfill of the soil ( Fig. 3.3.21). Pressure sensors are of similar
sensing mechanism as the LVDT type strain meters, and thus conditioned by the strain
meter conditioner ( Channel 10 and 11) and wired to Channel 42 and 43 of the data log
( Table 3.3.2).
Displacement Sensor
A displacement sensor ( D1, Fig. 3.3.22) was installed at Abutment 1 to measure the
relative displacement between the abutment and the superstructure along the longitudinal
direction. This sensor requires 5V DC power which is supplied by a
converter/ transformer installed in the data log housing box. The sensor data are acquired
to Channel 44 of the data log ( Table 3.3.2).
33
Thermocouples
Three thermocouples were installed in the superstructure in span 1. One of them
measures the outside temperature and the other two the inside temperature of the box
girder, with the first one installed near the ceiling and the second one near the floor of the
box girder. These thermocouples were connected to a signal conditioner that is located
inside the box girder ( Figure 3.3.23). The conditioner takes in ± 15V DC power from the
data log and supplies to the thermocouples, and at the same time, reads the outputs of the
thermocouples. Table 3.3.2 and Fig. 3.3.24 show the details of the splicing.
Figure 3.3.25 summarizes the current channel assignment of the data logger ( SAMTAC-
700). There are some spare channels for further expansion of the instrumentation system.
34
Table 3.3.1 Installation of Accelerometers
No. Model No. Serial No. Location Bracket Marked DPliarencnteiodn Installed
A0 SV355T 020723 Abutment 1 Ground No + X,+ Y,+ Z + X,+ Y,+ Z - Z, - Y, + X
A1 SV155T 020729 Beginning of span1 No + X + Y - Y
A2 SV255T 020724 Middle of span1 Type2 + X, + Y + Y, + Z + Y, + Z
A3 SV255T 020725 End of span1 No + X, + Y + X, + Y - X, + Y
A4 SV155T 020730 1/ 4 point of span2 Type1 + X + Z + Z
A5 SV255T 020726 1/ 2 point of span2 Type2 + X, + Y + Y, + Z + Z, + Y
A6 SV155T 020731 3/ 4 point of span2 Type1 + X + Z + Z
A7 SV155T 020732 End of span2 No + X + Y + Y
A8 SV155T 020733 1/ 4 point of span3 Type1 + X + Z + Z
A9 SV255T 020727 1/ 2 point of span3 Type2 + X, + Y + Y, + Z + Y, + Z
A10 SV155T 020734 3/ 4 point of span3 Type1 + X + Z + Z
A11 SV155T 020735 End of span3 No + X + Y + Y
A12 SV255T 020728 Middle of span4 Type2 + X, + Y + Y, + Z + Y, + Z
A13 SV155T 020736 End of span4 No + X + Y + X
Notes: ( a) + X: longitudinal, ( from abutment 1 to aboutment5), + Y: transverse, from North to South, + Z: vertical, from bottom to top. ( b) Bracket Type1
is for uni- directional accelerometer, and Type 2 is for bi- directional accelerometer.
Figure 3.3.6 Schematic Layout of the Accelerometers
L1/ 2 L1/ 2 L2/ 4 L2/ 4 L2/ 4 L2/ 4 L3/ 4 L3/ 4 L3/ 4 L3/ 4 L4/ 2 L4/ 2
A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 A11 A12 A13
Abutment1 Bent2 Bent3 Bent4 Abutment5
A0
35
Figure 3.3.7 Picture of a Uni- axial Accelerometer
Figure 3.3.8 Pictures of a Bi- axial Accelerometer
36
Figure 3.3.9 Pictures of a Tri- axial Accelerometer
Figure 3.3.10 Pictures of Accelerometers Mounted in the Box- girder
( a) A uni- axial accelerometer
mounted in box- girder without
bracket
( b) A bi- axial accelerometer mounted in box-girder
with a bracket to adjust the sensor
orientations
37
Figure 3.3.11 Picture of the Data- logger Container and the Junction Box
Figure 3.3.12 Data- logger and the Strain Meter Conditioner in the Container
Junction Box
Data logger
Container
38
Table 3.3.2 Splicing and Wiring Map
Sensor
No. & Dir
as Installed
Sensor
No. & Dir
as Marked
Other
Cables
Splicing
Cable Tag
Ch. No. in
Strain Meter
Conditioner
Ch. No. in
Data Logger
(-) A0Z A0X Thick Cable - 19
(-) A0Y A0Y Thick Cable - 20
A0X A0Z Thick Cable - 21
(-) A1Y A1X A1 blue - 9
A2Y A2X A2 blue - 10
A2Z A2Y 8 red - 1
(-) A3X A3X A3 blue - 17
A3Y A3Y 4 red - 11
A4Z A4X 11 red - 2
A5Z A5X A5 blue - 3
A5Y A5Y 7 red - 12
A6Z A6X A6 blue - --
A7Y A7X A7 blue - 13
A8Z A8X A8 blue - 5
A9Y A9X A9 blue - 14
A9Z A9Y 12 red - 6
A10Z A10X A10 blue - 7
A11Y A11X A11 blue - 15
A12Y A12X A12 blue - 16
A12Z A12Y 2 red - 8
A13X A13X A13 blue - 18
± 15V 13 red - --
ANT GPS 9 red - ANT
SD1 SD1 SD1 blue 1 33
SD2 SD2 SD2 blue 2 34
SD3 SD3 SD3 blue 3 35
SD4 SD4 SD4 blue 4 36
SD5 SD5 SD5 blue 5 37
SD6 SD6 SD6 blue 6 38
SD7 SD7 SD7 blue 7 39
SC1/ 2 SC1 SC1/ 2 blue 8 40
SC3 SC3 SC3 blue 9 41
P1 1 red 10 42
P2 3 red 11 43
T DC IN T1 T1 blue - --
T1/ T2/ T3
OUT T2 T2 blue - 22, 23, 24
D1 D1 6 red - 44
Notes: ( a) Symbol (-) marks the sensor with an orientation that is opposite to the assigned positive direction.
( b) Channel 4 of the datalogger is currently not used, because A6 is found to malfunction. ( c) Currently,
temperature sensors T1 and T2 are not connected to the datalogger. ( d) The antenna has not received
GPS signal up to date. ( e) Due to the fading of the marks on SC1, SC2, T1 and T2, SC1 is not
distinguishable from SC2; to distinguish T1 and T2 it will rely on future data reading and reasonable engineer
judgment. ( f) Channels 22 to 32, and channels 45 to 48 are currently unused.
39
Figure 3.3.13 Accelerometer Splicing in the Junction Box
Uni-axial
Red ( thick)-------+ 15V ( Source)
Black ( thick)----- 0V ( Source)
Blue ( thick)------ - 15V ( Source)
Brown------------- Signal output
Black-------------- Signal Com
Yellow----------- - Cal +
Green------------ Cal –
Shield
Bi-axial
Red ( thick)------- + 15V ( Source)
Black ( thick)----- 0V ( Source)
Blue ( thick)------ - 15V ( Source)
Brown------------- Signal output ( X)
Red---------------- Signal output ( Y)
Black/ Rose----- Signal Com( X. Y)
Yellow------------- X- axis CAL +
Green-------------- X- axis CAL -
Blue---------------- Y- axis CAL +
Purple------------- Y- axis CAL -
Tri-axial
Shield
Shield
Red ( thick)------ + 15V
Black ( thick)----- GND
Blue ( thick)----- - 15V
Brown------------ Signal Output ( X)
Red--------------- Signal Output ( Y)
Orange---------- Signal Output ( Z)
Black/ Rose----- X, Y, Z COM
Yellow------------ X- axis CAL +
Green---------- -- X- axis CAL -
Blue------------- - Y- axis CAL +
Purple------------ Y- axis CAL -
Gray ------------- Z- axis CAL +
White ------------ Z- axis CAL -
Red
Black
White
Brown
Black
DC ± 15V
Signal Cable
Red
Black
White
Brown
Brown
Black
DC ± 15V
X Signal Cable
Y Signal Cable
Red
Black
White
Brown
Brown
Brown
Black
DC ± 15V
X Signal Cable
Y Signal Cable
Z Signal Cable
Sensor Cables Splicing Cables
40
( a) LVDT Type Strain Meter Installed in the Deck
( b) LVDT Type Strain Meter Installed in the Column
Figure 3.3.14 LVDT Type Strain Meters Installation
41
Figure 3.3.15 Position of the Strain Meters on the Deck
Figure 3.3.16 Strain Meter Splicing in the Junction Box
Figure 3.3.17 Conventional Strain Gauge and LVDT Strain Meter Installed in Column 3
Strain Meter
( BF/ ES)
Red-------+ 5V ( Source)
Black----- 0V ( Source)
White----- 0V ( Output Com)
Green----- Signal
Yellow
Shield
Red
Black
White
Green
Strain Cable
Sensor Cables Splicing Cables
Abut 1 Bent 2 Bent 3 Bent 4 Abut 5
SD1 SD3
SD2
SD4 SD5 SD6 SD7
16in
61in
8in
52in 43in 52in 1041in 51in 1040in 52in
42
Figure3.3.18 Conventional Strain Gauges Installed in Footings
North
East
1
3 2
4
North
East
5
7 6
8
North
East
9
11 10
12
Bent 2 Bent 3 Bent 4
3, 4
1, 2
7, 8
5, 6
9, 10
11, 12
2, 6, 10
1, 5, 9
4, 8, 12
3, 7, 11
43
Figure 3.3.19 Installation of Signal Box for Conventional Strain Gauges
Figure 3.3.20 Soil Pressure Sensor
44
Figure 3.3.21 Backfilling at Abutment 1
Figure 3.3.22 Displacement Sensor
45
Figure 3.3.23 Thermocouple and Signal Conditioner
Figure 3.3.24 Thermocouple Wiring Map
( a) Thermocouple Sensor head
( b) Signal Conditioner: Front ( c) Signal Conditioner: Back
SENSOR- IN
OUT DC15V IN
1 2 3
White
Black
SQT- 51
To
Shield
CH- 22
CH- 23
CH- 24
+ 15V
0V
Red
Green
White
Black
46
Figure 3.3.25 Map of Channel Assignment of the Data- logger
IC Card
Dig ital Reco rder
SAMTAC- 700
( 48CH)
IC C ard 96MB
AC- 120 V
Rela tive
Disp laceme nt M e ter
DP- 100
GPS
DC So u rce
CH1 ～ 21
Arrester
Acce le rome ter
SV- 155T×8
Soil P res sure
Senso r
BE- 2KRS12×2
Reber
Stra in M eter
ES- 500T×9
Signa l
C ond itio ner
UPS
SC10R- 3 0
30min .
Back - up
Time
Acce le rome ter
SV- 355T×1
Acce le rome ter
SV- 255T×5
Thermome ter
SQT- 51×2
Ethe rnet
SQT- 51
（ 3CH）
AL- 10
（ 9CH）
AL- 10
( 2CH)
+ 1 5V
CH33 ～ 40
CH22 ～ 24
+ 5V
CH41
CH44
CH42- 4 3
Down
T rans.
3
10
8
8
2
1
3
1
47
Chapter 4
VIBRATION DATA
This chapter documents vibration data collected on the three instrumented highway
bridges during Phase II of the research project.
Each of the data loggers on the three monitoring systems can be set to continuously
monitor 3 channels of accelerometers and if the signals of these three channels meet
the selected triggering criteria, the data logger will be automatically triggered to
record vibration signals of all sensors. They can also be manually triggered in the
control panel to record a 1- minute vibration data file. If a pair of triggering jumper
wires is used, the data logger can record continuously as long as the jumper is
engaged. In this report, the data files recorded in these 3 modes are cataloged as the
‘ triggered’ data, which are recorded in the compact flash memory cards on the data
logs and retrieved and analyzed off- line. On the JRO, however, after the system
upgrades, in addition to these 3 modes we are also able to continuously receive 9
channels of the on- line data streamed through the wireless system, and save them on
the server computer. Data collected in this mode are the ‘ streamed’ data.
Besides the working modes of the data loggers, vibration data are also cataloged by
the different types of excitation sources. The bridge vibrations due to ambient effects
( e. g. wind) or traffic loads constitute the majority of the collected data. In this case,
the excitation on the bridge structure is not measured, but the bridge response to such
excitation is recorded. Usually in such ambient/ traffic- induced vibration, the
superstructure response exhibits much larger amplitude than the substructure
48
response, and the vibration is mainly in the vertical direction. Another excitation
source is ground motion. During Phase II of the research project, two moderate
earthquakes were recorded by the monitoring systems. These ground motion- induced
vibrations are both in the transverse and vertical directions. The ground motion
sensors pick up considerable vibration at the footing of the substructure, which can be
considered as the time- history of the ground motion acceleration that excited the
bridge.
4.1 Ambient/ Traffic- Induced Vibration on JRO
Since the JRO was instrumented with the monitoring system, total of 1712 data sets
have been collected on this bridge.
4.1.1 Triggered Data
After analyzing all the collected data it was observed that the maximum transverse
acceleration in the middle of the span is between 2- 20 gal; whereas the maximum
vertical acceleration ranges between 10- 80 gal. Table 4.1.1 documents the triggered
vibration data that have been collected.
Table 4.1.1 Summary of Triggered Data on the JRO
Date Time
( a4) max Date Time
4 max ( a )
05/ 03/ 2002 09.29 12 09/ 10/ 2004 18.11 21
01/ 24/ 2003 08.39 42 05/ 24/ 2005 19.07 34
… … … … … …
49
Typical traffic- induced time histories for the vertical and transverse accelerations at
middle of Span 2 are shown in Fig 4.1.1.
0 10 20 30 40 50 60
- 30
- 20
- 10
0
10
20
30
Vertical Acceleration ( gal)
0 10 20 30 40 50 60
- 10
- 5
0
5
10
time ( sec)
Transverse Acceleration ( gal)
Fig 4.1.1 Saved Traffic- Induced Vibrations at JRO
4.1.2. Streamed Data
Starting August 2006, 5 min long data have been automatically collected every hour.
The increase of the data length enabled more precise modal identification results.
Table 4.1.2 documents the streamed vibration data that have been collected.
50
Table 4.1.2 Summary of Streamed Data on the JRO
Date Time
( a4) max Date Time
4 max ( a )
08/ 30/ 2006 11.00 16 09/ 17/ 2006 11.00 23
09/ 02/ 2006 11.00 22 09/ 30/ 2006 11.00 36
… … … … … …
Typical traffic- induced time history for the vertical and transverse accelerations at the
middle of Span 2 are shown in Fig 4.1.2.
0 50 100 150 200 250 300
- 40
- 20
0
20
40
Vertical Acceleration ( gal)
0 50 100 150 200 250 300
- 10
- 5
0
5
10
time ( sec)
Transverse Acceleration ( gal)
Fig 4.1.2 Scheduled Traffic- Induced Vibrations at JRO
51
4.2 Moderate Earthquake on JRO
On June 16, 2005, a moderate earthquake occurred at 1: 53 pm ( PDT) in Yucaipa, CA.
The local magnitude is between 4 to 5 MI, and the distance from the epicenter to the
JRO is about 105 km ( 65 miles). The monitoring system at the JRO was triggered by
this ground motion and recorded this event. The record shows a peak ground
acceleration in North- South of 11.6 gal, in East- West of 13.0 gal and vertical of 3.55
gal. These values are consistent with the Shake- Map instrumental intensity maps ( Fig.
4.2.1).
Fig. 4.2.1 Instrumental Intensity Map by Shake- map
52
The earthquake records of the selected channels are plotted in Fig. 4.2.2. One can
observe that the earthquake excited bridge vibration in the transverse direction more
than the traffic does. Transverse vibration ( Ch- 3) in the middle of span 2 has an
amplitude of 25 gal, comparable to that of the vertical direction ( Ch- 4) in the same
event, but much larger than the transverse vibration induced by traffic. The bridge
vibration near the ground, such as Ch- 10, is much stronger than that under traffic
excitation. Peak ground accelerations for this event are given in Table 4.4.1. Also
note that the vertical vibration remains in the same level for both traffic excited and
earthquake excited vibrations. One can see the impulse- like pattern in the vertical
vibration record during the event, indicating vehicles passing the bridge during the
earthquake event.
0 10 20 30 40 50 60
- 10
0
10
20
0 10 20 30 40 50 60
- 5
0
5
Acc. ( gal) Acc. ( gal)
( )
Time ( sec)
Time ( sec)
Input in the transverse direction
Input in the vertical direction
( a) Input Motion
0 10 20 30 40 50 60
- 50
0
50
Time ( sec)
Acc. ( gal)
0 10 20 30 40 50 60
- 50
0
50
20 Acc. ( gal)
Time ( sec)
Response in the transverse direction
Response in the vertical direction
( b) Earthquake Response
Figure 4.2.2 Typical Earthquake Records on JRO
53
Table 4.4.1 Peak Ground Motion at the JRO
Direction Longitudinal Transverse Vertical
Peak Ground Acceleration ( gal) 11.6 13.0 3.6
4.3 Ambient/ Traffic- Induced Vibration on
WSO
Ever since the WSO bridge was instrumented with a monitoring system, total of 92
data sets have been collected on this bridge. Some examples can be seen in Table
4.3.1.
Table 4.3.1 Summary of Collected Triggered Data on the WSO
Date Time
9 max ( a ) File Name Date Time
10 max ( a ) File Name
5/ 17/ 05 1: 21: 29 0.0256 1D62155D 5/ 17/ 05 1: 21: 29 0.1028 1D62155D
9/ 26/ 05 11: 31: 16 0.0293 1E74B7D0 9/ 26/ 05 11: 31: 16 0.2032 1E74B7D0
… … … … … … … …
Typical traffic induced time history for the vertical and transverse accelerations
recorded at the middle of Span 2 are shown in Fig 4.3.1.
4.4 Moderate Earthquake on WSO
There was a moderate earthquake on 16 July, 2005. The ground motions of the
earthquake are shown in Fig. 4.4.1. The peak acceleration of each direction is shown
54
in Table 4.4.1. Unlike the vibration induced by traffic, the transverse direction is most
dominant component.
( a) Input ground motion
( b) Earthquake response
Figure 4.4.1 Typical Earthquake Records on WSO
55
Table 4.4.1 Peak Ground Motion at the WSO
Direction Longitudinal Transverse Vertical
Peak Ground Acceleration ( gal) 5.6 13.6 5.0
4.5 Ambient Vibration Data on FROO
Figure 4.5.1 shows the typical acceleration time history of the Fairview On Ramp at
the middle of span 3. In Table 4.5.1 the examples of monitored peak acceleration
values are shown.
Table 4.5.1 Summary of collected data on the FROO
Date Time
9 max ( a ) * File Name Date Time
9 max ( a ) ** File Name
3/ 20/ 2006 16: 05: 35 0.3019 20E90163 3/ 20/ 2006 16: 05: 35 0.1826 20E90163
3/ 20/ 2006 16: 17: 34 0.1755 20E90462 3/ 20/ 2006 16: 17: 34 0.1336 20E90462
… … … … … … … …
*: Vertical direction, ** : Transverse direction.
56
Figure 4.5.1 Time History of FROO ( Middle of Span 3)
57
Chapter 5
STRAIN DATA AND ANALYSIS
In this chapter, dynamic strain data from the West St. On- Ramp ( WSO) under traffic
loads were analyzed and compared with the those based on moving- load analysis.
From the results, it was found that in the WSO the transverse mode was excited by
heavy moving vehicle and it caused higher strain in the column than that predicted in
the design.
5.1 Strain Sensors and Locations
The strain sensors were permanently embedded in concrete members of the West
Street On- Ramp ( WSO) during construction. The strain gauges were completely
welded to dummy reinforcing bars. The specification of the strain gauges is shown in
Table 5.1.1. The sensors were installed to measure the dynamic strains induced by
bending moments. The locations of strain gauges are shown in Fig 5.1.1
58
Table 5.1.1 Strain Gauge Specification
Parameter Specification
1. Model ES- 500T
2. Strain range ± 1000μ Srain
3. Average resolution 0.01μ Srain
4. Average sensitivity 0.55 mV/ μStrain
5. Temperature coefficient 0.7×10- 5/° C (- 20 to + 60° C) 85μV/ kg
6. Gauge length 500mm
7. Frequency response DC - 50Hz
8. Cable 4 Conductor, shielded ( Extensible length 300m)
Fig. 5.1.1 Strain gauge locations of WSO
59
5.2 Characteristics of Dynamic Strain Data
The dynamic strain time history of each sensor assembles the moment influence line
at the sensor location. In this section, the characteristics of the dynamic strain time
history were discussed in comparison with the static moment influence line.
5.2.1 R1 and R10
The strain gauges R1 and R10 are embedded in the girder near Abutment 1 of the
bridge. R1 is located in the outer girder while R10 inner girder. As shown in
Figure. 5.2.1, the influence line of the moment at R1 and R10 shows sharp increase
and gradual decrease. The same trend was observed from the recorded data for R1 as
shown in Figure 5.2.2. From the influence line and the recorded data, one can observe
that when a vehicle enters the bridge the strain at R1 increases abruptly and as the
vehicle passes through, the strain decreases gradually. However, as depicted in Figure
5.2.3, R10 does not show the same pattern as R1. It is considered that R10 is not
reliable.
Figure. 5.2.1 Influence Line for R1 and R10
60
Figure. 5.2.2 Strain Time History of R1
Figure. 5.2.3 Strain Time History of R10
5.2.2 R2 and R3
Sensors R2 and R3 are located in the outside girder above column 2 outside of the
diaphragm. R2 is embedded in the upper part of the girder while R3 in the lower part
of the girder. Influence lines for R2 and R3 are respectively shown in Figures 5.2.4
and 5.2.5. Because these two strain gauges are located in the upper and lower parts of
the same cross section, they show opposite strain signs at the same time. From the
influence lines, it can be inferred that the strain value is higher when vehicle traverses
on span 1 than other spans, and each influence line has two peaks.
61
Figure. 5.2.4 Influence Line of R2
Figure. 5.2.5 Influence Line of R3
Similar trend is observed from the time histories of R2 and R3. The recorded time
history of R2 and R3 are shown in Figures 5.2.6 and 5.2.7.
Figure. 5.2.6 Strain Time History of R2
62
Figure. 5.2.7 Strain Time History of R3
5.2.3 R4, R5, and R6
The strain sensors R4, R5, and R6 are located in the column 2 under the ground level.
The location of each sensor can be seen in Figure. 5.1.3. The influence line of R4 in
Figure. 5.2.8 shows one and half cycle. The same trend can be seen from strain time
history of R4 in Figure. 5.2.9. It should be noted that the strain sensors located in the
column shows both signs with almost the same strain values for both tension and
compression. It means that the column experiences both tension and compression
when vehicle traverses the bridge.
Figure. 5.2.8 Influence Line of R4
63
R5 and R6 are located on the opposite side of the column 2 in the transverse direction.
Thus, when R5 is in tension then R6 is in compression and vice versa. Therefore it
can be inferred that R5 and R6 show approximately the same strain value with
opposite signs. The recorded strain time histories of R5 and R6 are shown in Figures
5.2.10 and 5.2.11.
Figure. 5.2.9 Strain Time History of R4
Figure. 5.2.10 Strain Time History of R5
Figure. 5.2.11 Strain Time History of R6
64
5.2.4 R7, R8, and R9
The sensors R7, R8, and R9 are embedded in the middle of span 2. R7 is located at
the upper part of the outside girder. R8 and R9 are located at the lower part of the
girder. R8 is in the outside girder and R9 in the inside girder. The moment at the
middle of span 2 is negative when a moving vehicle is located on the span 1 and 3 but
it is positive on the span 2.
Figures 5.2.12 and 5.2.13 show the influence lines of R7 and R8 ( R9) respectively.
Since R7 is located at the upper part of the girder, the strain data should show the
exact opposite sign to R8 and R9. As depicted in Figure 5.2.14, however, the
monitored data at R7 did not show the expected trend. The sensor at R7 is believed to
be out of order. Figures 5.2.15 and 5.2.16 show the time histories of the strains at R8
and R9. The shape of the time histories is the same as the compressed shape of
influence line of R8.
Figure. 5.2.12 Influence Line of R7
Figure. 5.2.13 Influence Line of R8 and R9
65
Figure. 5.2.14 Strain Time History of R7
Figure. 5.2.15 Strain Time History of R8
Figure. 5.2.16 Strain Time History of R9
66
5.3 Comparison of Measured and Computed
Strain
The maximum and minimum strains measured by each strain sensor were compared
with that computed from the moving load analysis. The dynamic effect of the design
live load was represented by employing an impact factor. The centrifugal force due to
the curvature of the bridge was also considered in the analysis. The computed
maximum strains of in the girder showed higher values than those from the
measurement. However, the measured strains of the column were higher than the
computed ones. From analysis of the measured strains and accelerations, the high
strains at the columns were attributed to the transverse vibration excited by moving
vehicles.
5.3.1 Measured Maximum Strain
( 1) R1 and R10
The maximum strain of each data sets for R1 and R10 is shown in Figure. 5.3.1. The
maximum strain values of R1 and R10 are 2.282μ and 2.246μ respectively. Though
the maximum strain of these two strain sensors is nearly the same, the average strain
of R10 is much smaller than R1. The average strain of R1 and R10 is 1.047μ and
0.614μ. The maximum values and the average values of all the strain sensors are
shown in Table 5.3.1.
67
0.0
0.5
1.0
1.5
2.0
2.5
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89
No. of Data
Strain( μ)
Max R1
Max R10
Figure. 5.3.1 Maximum Strain of R1 and R10
Table 5.3.1 Measured Strain
Sensor Maximum ( μ) Max. Average ( μ) Minimum ( μ)
Min. Average
( μ)
R1 2.282 1.047 - 0.641 - 0.298
R2 6.686 2.726 - 3.222 - 0.635
R3 3.312 0.831 - 7.751 - 3.482
R4 5.364 1.072 - 4.138 - 1.451
R5 11.132 2.463 - 14.828 - 4.738
R6 12.745 2.379 - 7.326 - 2.049
R7 0.440 0.256 - 0.399 - 0.249
R8 18.513 6.494 - 5.204 - 1.640
R9 21.471 10.317 - 7.035 - 2.2132
R10 2.246 0.614 - 1.489 - 0.505
68
- 10.0
- 8.0
- 6.0
- 4.0
- 2.0
0.0
2.0
4.0
6.0
8.0
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89
No. of Data
Strain( μ)
Max R2
Min R3
Figure. 5.3.2 Maximum Strain of R2 and R3
( 2) R2 and R3
The maximum strain values of R2 and R3 are shown in Figure. 5.3.2. Because of the
sensor locations of R2 and R3, the maximum strain at R2 corresponds to the
minimum one at R3. The largest strain of R2 is 6.686μ while the smallest value of R3
is – 7.751μ. The average strain value of R2 and R3 is 2.726μ and – 3.482μ.
( 3) R4, R5, and R6
The maximum and minimum strain of R4, R5, and R6 are shown in Figures 5.3.3
through 5.3.5. From Figure 5.3.3 the maximum tensile and compressive strain of R4
are approximately the same. It means that the column experiences the same negative
and positive moment in the longitudinal direction. Figures 5.3.4 and 5.3.5 for R5 and
69
R6 indicate that column 2 is subject to both negative and positive moments in the
transverse direction as well.
The maximum tensile and compressive strains of R4 are 5.364μ and – 4.138μ
respectively. The maximum and minimum strain of R5 are 11.132μ, – 14.828μ, and
those of R6 are 12.745μ and – 7.326μ respectively.
- 6.0
- 4.0
- 2.0
0.0
2.0
4.0
6.0
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89
No. of Data
Strain( μ)
Max R4
Min R4
Figure. 5.3.3 Maximum and Minimum Strain of R4
70
- 10.0
- 5.0
0.0
5.0
10.0
15.0
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91
No. of Data
Strain( μ)
Max R5
Min R6
Figure. 5.3.4 Maximum Strain of R5 and Minimum strain of R6
- 20.0
- 15.0
- 10.0
- 5.0
0.0
5.0
10.0
15.0
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89
No. of Data
Strain( μ)
Max R6
Min R5
Figure. 5.3.5 Maximum Strain of R6 and Minimum Strain of R5
71
( 4) R7, R8, and R9
Figure 5.3.6 shows the minimum strain at R7 while Figure 5.3.7 the maximum strains
at R8 and R9. Considering the locations of R7 and R8 or R9, the absolute strain value
of R7 should be similar to that of R8 or R9, but all the strain values of R7 are between
0 and - 0.4μ. The strain sensor at R7 is thus considered to be out of order.
The maximum values of strains at R8 and R9 are respectively 18.513μ and 21.471μ.
It is found that the strain at R9 is larger than that at R8. The average strain of R8 is
6.494μ and that of R9 is 10.317μ. The strain difference between R8 and R9 is
attributed to the their locations; R9 is located in the inside girder while R8 the outside
girder.
- 0.5
- 0.4
- 0.3
- 0.2
- 0.1
0.0
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89
No. of Data
Strain( μ)
Min R7
Figure. 5.3.6 Minimum Strain of R7
72
0.0
5.0
10.0
15.0
20.0
25.0
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89
No. of Data
Strain( μ)
Max R9
Max R8
Figure. 5.3.7 Maximum Strain of R8 and R9
5.3.2 Finite Element Analysis under Design Live Load
( 1) Finite Element Model
For the comparison of the monitored strain data with analytical one, finite element
( FE) analysis was carried out. Three- dimensional beam elements were used for the
deck and column components of the bridge. The superstructure has 12% inclination in
the transverse direction and it was represented using the angular rotation of the
element local axis. The superstructure was modeled with totally 200 beam elements
and a column with 16 beam elements. Figure. 5.3.8 shows the FE model of the bridge.
73
Figure. 5.3.8 Finite Element Model of WSO
Figure. 5.3.9 HS 20- 44 Load
The most difficult aspect is to model the bridge boundary conditions - the column
footing and abutments realistically and accurately. Considering that the use of the
model is for analyzing the bridge response to operational traffic loads, the abutment
bearings of the bridge were modeled as linear horizontal, vertical, and rotational
springs, while the footing piles as fixed. The bearing stiffness values at both the
abutments were assigned according to FHWA ( 1996) as 6.58×104 kip/ ft for the
longitudinal springs, and 1.29×105 kip/ ft and 1.48×105 kip/ ft for the transverse and
vertical springs respectively. The rotational spring stiffness are 6.29×107 kip- ft/ rad
74
and 3.5×107 kip- ft/ rad for longitudinal and transverse direction axis. It is noted that
these values were used for the preliminary finite element analysis. They later were
identified and updated by the vibration measurement as shown in Chapter 9.
( 2) Moving vehicle load
The design live load HS 20- 44 load was used for moving vehicle load analysis. Figure
5.3.9 shows the axial load and spacing of the HS 20- 44 load. The total axial load of
HS20- 44 is 72kips and the width of the truck is 10 feet. The WSO has two traffic
lanes of 24 feet but the possible traffic passage lanes were defined based on the width
of HS20- 44. A total of 12 lanes were defined as shown in Figure5.3.10. The lanes
from R3 to R8 are located on the inside of the horizontal curvature of the bridge and
the lanes from L3 to L8 on the outside. The number after ‘ R’ and ‘ L’ represents the
eccentricity of the lane from the center of the bridge.
Figure. 5.3.10 Lanes for moving load analysis
( 3) Centrifugal Force
Centrifugal force exerted by moving vehicles due to the curvature of the bridge was
considered in the finite element model according to the design specification.
75
Centrifugal force was taken as the product of the axle weights of the design truck and
the factor C computed as:
gR
C V
2
3
= 4
where:
V= vehicle speed ( ft/ sec)
g = gravitational acceleration: 32.2 ( ft/ sec)
R= radius of curvature of traffic lane ( ft)
Centrifugal forces were applied horizontally at a distance 6.0 feet above the roadway
surface. It was found from the finite element analysis that girder strain under
centrifugal forces was less than 1.5μ but the column strain was more than 6μ. So
the centrifugal forces affected more on the column strain than the girder strain.
( 4) Strain sensitivity to lanes
Figure. 5.3.11 to 5.3.16 show the strain at each sensor location from the FE analysis
and Table 5.3.2 summarizes the FE analysis results together with the monitored data.
From the figures it can be seen that each strain varies according to the vehicle
location. Sensitivity coefficient S is defined in Eq. ( 5- 1) in order to compare the
sensitivities of each strain sensor to the vehicle location.
(%) 100
min
max min ×
−
=
μ
μ μ
S ( 5- 1)
where S : Sensitivity
max μ : Maximum strain value of FE results
min μ : Maximum strain value of FE results
76
The sensitivity coefficients for all the strain sensors are plotted in Figures 5.3.11
through 5.3. 16. The sensitivity of R1 and R10 is 2.82% and that of R2 and R3 is
52.82% while the sensitivity of R7, R8, and R9 is 40.68%. The sensitivity values of
R1 and R10 are very low compared with those of other sensors. These two sensors are
installed near the abutment ( entrance) of the bridge, and thus the moment does not
change much due to the different locations of the vehicle. The large sensitivity values
for the sensors installed in the girder above the column 2 and at the middle of span 2
imply that the moments at those locations are quite dependent on the location of
vehicle load in the transverse direction.
The column, as mentioned earlier, shows both tensile and compressive strains when a
vehicle traverses the bridge. The sensitivity values for the column are quite different
for tension and compression. For example, the sensitivity value of R4 is 84.51% for
tensile strain and 29.51% for compressive strain, and those for R5 ( R6) are 226.64%
and 4.61%. Though the absolute strain values are not large compared with those of R8
and R9, the sensitivity values of the column are much larger than those of the girder.
77
0.765
0.770
0.775
0.780
0.785
0.790
0.795
0.800
0.805
L8 L7 L6 L5 L4 L3 R3 R4 R5 R6 R7 R8
Lane
Strain( μ)
R1
R10
Figure. 5.3.11 Strain R1 and R10 due to Vehicle Locations
- 30.0
- 25.0
- 20.0
- 15.0
- 10.0
- 5.0
0.0
5.0
10.0
15.0
L8 L7 L6 L5 L4 L3 R3 R4 R5 R6 R7 R8
Lane
Strain( μ)
R2
R3
Figure. 5.3.12 Strain R2 and R3 due to Vehicle Locations
78
- 20.0
- 15.0
- 10.0
- 5.0
0.0
5.0
10.0
15.0
20.0
L8 L7 L6 L5 L4 L3 R3 R4 R5 R6 R7 R8
Lane
Strain( μ)
R4_ Max
R4_ Min
Figure. 5.3.13 Strain R4 due to Vehicle Locations
- 12.0
- 10.0
- 8.0
- 6.0
- 4.0
- 2.0
0.0
2.0
4.0
6.0
8.0
L8 L7 L6 L5 L4 L3 R3 R4 R5 R6 R7 R8
Lane
Strain( μ)
R5_ max
R5_ min
Figure. 5.3.14 Strain R5 due to Vehicle Locations
79
- 8.0
- 6.0
- 4.0
- 2.0
0.0
2.0
4.0
6.0
8.0
10.0
12.0
L8 L7 L6 L5 L4 L3 R3 R4 R5 R6 R7 R8
Lane
Strain( μ)
R6_ max
R6_ min
Figure. 5.3.15 Strain R6 due to Vehicle Locations
- 20.0
- 15.0
- 10.0
- 5.0
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
L8 L7 L6 L5 L4 L3 R3 R4 R5 R6 R7 R8
Lane
Strain( μ)
R7
R8
R9
Figure. 5.3.16 Strain R7, R8, and R9 from Finite Element Analysis
80
5.3.3 Comparison of Strain Data
Table 5.3.2 shows the maximum and minimum strains of each strain sensor extracted
from the recorded 92 data sets, in comparison with those from the finite element
analysis. Sensors R1 and R10 are located at the girder near the entrance of the bridge.
The recorded maximum strains at R1 and R10 are nearly twice higher than those from
the analysis. This is due to the impact at the expansion joint of the bridge
superstructure at the entrance of the bridge. This impact was not considered in the
finite element analysis.
Table 5.3.2 Strain from measurement and analysis
Sensor Monitored ( μ) ( 1) Computed ( μ) ( 2)
Difference (%)
⎟ ⎟⎠
⎞
⎜ ⎜⎝
⎛
×
−
= 100
( 2)
( 2) ( 1)
R1 2.282 0.801 ( 185)
R2 6.686 10.595 37
R3 - 7.751 - 24.307 68
Max 5.364 15.280 65
R4
Min - 4.138 - 17.862 77
Max 11.132 6.391 ( 74)
R5
Min - 14.828 - 9.905 ( 50)
Max 12.745 9.905 ( 29)
R6
Min - 7.326 - 6.391 ( 15)
R7* 0.440 - 13.178 -
R8 18.513 19.360 4
R9 21.471 30.771 30
R10 2.246 0.801 ( 180)
* : out of order
81
From Table 5.3.2 the difference between the computed and measured strains at R2
and R3 ( on the girder on the top of column 2) are higher than those at R8 and R9 ( on
the girder in the middle of span 2). The measured maximum strains at the box girder
above column 2 are much higher than the computed ones, while the difference is
much smaller in the middle of span 2. This implies that the load capacity of the box
girder above the column is higher than that of the middle of span 2. On the other
hand, the strain difference inside column 2 depends on the direction. At R4 ( in the
longitudinal direction), the strain difference is 65% but at 5 and R6 ( in the transverse
direction) they are respectively – 74% and 22%. The negative strain difference of R5
means that the recorded maximum strain exceeds the computed maximum strain.
0
2
4
6
8
10
12
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91
No. of Data
Strain ( x106)
Measured
Computed
Fig. 5.3.17 Long- term monitored data of R5
Figure 5.3.17 shows the maximum strain envelop of R5 from the monitored data set.
The solid line is the strain obtained from the finite element analysis under the HS- 20
design load. A typical time history of R5, whose maximum value is near the
computed strain, is shown in Figs. 5.3.18 ( a) and ( b), in comparison with acceleration
A5 recorded at the same time.
82
( a) A5 ( deck- above column : trans. dir.)
( b) R5 ( bottom of column)
( c) PSD of A5 ( d) PSD of R5
Fig. 5.3.18 Monitored time histories and their power spectral density ( 6/ 28/ 2006)
83
Fig. 5.3.19 Power spectral density of acceleration ( transverse direction)
Fig. 5.3.20 The fourth mode shape ( transverse direction)
Figures 5.3.18 ( c) and ( d) plot the power spectral density ( PSD) of A5 and R5. The
dominant frequencies identified from the acceleration ( A5) and strain ( R5) time
histories are identical and close to the fourth mode frequency of the bridge in Fig.
5.3.19. Figure 5.3.20 shows the fourth mode shape of the bridge in the transverse
direction from the analysis and measurement. It is clear that the moving heavy vehicle
excited the fourth ode of the bridge, resulting in higher strain in column 2 in the
transverse direction, as shown in Fig. 5.3.18 ( b), than that expected from analysis.
84
5.3 Summary
In the strain analysis, the measured strain from the WSO was compared with
computed one. Because input vehicle loads could not be measured, the results of
strain analysis can be interpreted only in a qualitative way. From the comparison of
the measured and computed strain, it is found that generally the bridge superstructure
was more conservatively designed than the column under moving vehicle load. It is
also noted that the column of the WSO was more affected by heavy moving vehicles
than the superstructure. From the frequency analysis of the acceleration and strain
measured column, it was found that some vehicles excited the transverse mode of the
bridge resulting in higher strain in the column than that expected from analysis. The
design based on the dynamic factor underestimate the strain in the column.
85
Chapter 6
DEVELOPMENT OF STRUCTURAL
HEALTH MONITORING
METHODOLOGIES
This chapter presents methodologies developed for identifying the structural “ health”
conditions of highway bridges.
6.1 Definition of Structural Health and Damage
Structural elemental stiffness is proposed to be an indicator of the structural “ health”.
As a structure deteriorates due to aging or suffers from damage by extreme events
such as earthquakes, the structural stiffness will degrade, and as a result, the global
dynamic characteristics of the structure will change. Therefore, by measuring the
structural vibration, it is possible to identify the change in structural dynamic
characteristics, and furthermore change in structural stiffness. When the reduction in
structural stiffness exceeds a certain threshold, the structure is defined as damaged.
The use of structural stiffness enables assessment of not only extent but also locations
of damage.
86
6.2 System Identification Methodologies
In this project, a number of system identification methods were developed for
identifying structural elemental stiffness based on structural vibration responses to
traffic or earthquake excitations. For assessing the bridge superstructure health, it is
proposed to use traffic- induced vibration as the moving vehicle induces high-amplitude
vertical vibrations. For this purpose, a unique traffic excitation model was
developed that incorporates partial traffic information based on video monitoring, and
as a result it is more realistic than the conventional assumption of white noise.
Bayesian updating and neural network system identification methods were developed
for identification of bridge structures based on traffic excitations.
For assessing seismic damage that usually occurs in bridge columns, it is proposed to
use seismic- induced vibrations. Because the damaged structure is a nonlinear system
while most of the available system identification methods are for linear systems, the
project developed a special system identification method based on the extended
Kalman filtering that can deal with nonlinear systems.
The following provides a literature review of related system identification methods.
System identification methods for structures based on vibration measurement can be
grouped into two depending on whether the identification is carried out in frequency
or time domain, as shown in Figure 6.2.1. If it is in frequency domain, basically the
changes in modal values; frequency, damping, shape, are used as an indication of
damage. However; if one wants to identify the changes more in detail like changes in
elemental stiffness, time domain identification methods might be more appropriate.
Time domain methods can be grouped into two depending on whether they are purely
data driven or they are incorporating FE model. If it is aimed to determine the
changes in the stiffness values, FE model must always be used. Within time domain
identification methods, the most common one is the least squares estimation ( LSE). It
is basically performing an optimization for the parameters such as stiffness and
87
damping so that the error between the measured and the simulated responses is
minimized. LSE is useful as a system identification technique, when used in
combination with a damage detection algorithm ( Stubbs et al, 2000). However, there
are some drawbacks of LSE. Firstly, physical insight can be easily lost and a local
maximum can be chosen over a global one. Secondly, LSE is very time consuming
and cannot be applied for “ on- line” structural health monitoring and damage
detection. To overcome this difficulty, the recursive least squares ( RLS) technique is
proposed so that any time varying property in a system caused by damage can be
tracked in real time. However in this case incorporation of FE is sacrificed, i. e. it is
purely data driven so change in the system parameters can be tracked but it is not
possible to link this to the change in structural stiffness and damping. Also, RLS is
susceptible to even low level of noise. As can be seen every method has some
drawbacks and is not effective for on- line identification of stiffness values under
realistic conditions.
Kalman filtering was a break- through in system engineering field when first proposed
four decades ago. It not only uses the data in a probabilistic sense but also gets
information from structural model ( Kalman, 1960). Results obtained by the Extended
Kalman Filter ( EKF) approach from simulated data and well defined models with
known damage scenarios were reported ( Yun and Shinozuka, 1980; Hoshiya and
Saito, 1984; Yang et al, 2005; Straser and Kiremidjian, 1996; Loh and Chung, 1993;
Loh and Tou, 1995, Ghanem and Ferro, 2006). However, applicability of the EKF
approach to civil engineering structures involving high uncertainties in structures and
loadings under realistic damaging events has not yet been studied. This research effort
can be seen within this report.
88
Figure 6.2.1 System Identification Methodologies
System Identification Methodologies
Frequency Domain Identification Time Domain Identification
Least Squares
Estimation for
Modal Parameters
Neural Network
Based
Identification
Extended
Kalman Filter
Based
Identification
Least Squares
Estimation for
Structural
Parameters
89
6.3 Traffic Excitation Modeling and Super-
Structure Condition Assessment
Since it is impossible to measure the input traffic excitation on a bridge, a stochastic
model of traffic excitation on bridges is developed in this project, by assuming that
vehicles traversing a bridge ( modeled as an elastic beam) arrive in accordance with a
Poisson process, and that the contact force of a vehicle on the bridge deck can be
converted to equivalent dynamic loads at the nodes of the beam elements. The
parameters in this model, such as the Poisson arrival rate and the stochastic
distribution of vehicle speeds, are obtained by image processing of the traffic video.
The model reveals that traffic excitation on bridges is spatially correlated. Partial
traffic information expressed by the stochastic model is incorporated in a Bayesian
framework to evaluate the structural properties and update their uncertainty for
condition assessment of the bridge superstructure. The vehicle weights are also
estimated simultaneously in this procedure. This method is validated in the testbed.
6.3.1 Output- Only System Identification
The desirableness of measuring vibration responses of an instrumented highway
bridge to traffic excitations for a long- term SHM purpose has been addressed by
many authors. To list a few of its practical advantages over other bridge structural
condition assessment methods: ( I) It does not interrupt traffics; ( II) It captures the in-situ
dynamic behavior of the bridge undergoing its normal service; ( III) It can be
performed continuously, scheduled periodically or triggered automatically and ( IV) It
requires no special experimental arrangement or a heavy shaker/ hammer. During
such measurements, however, the excitation loads are neither controllable nor ( easily)
measurable. Thus, to extract the structural properties of the bridge from the vibration
data, system identification is performed based only on the measured time histories of
the bridge responses ( system output) without measuring the traffic excitations ( system
90
input). As a result, to facilitate such output- only identification of structural properties,
models or assumptions representing the stochastic characteristics of the input must be
established a priori, otherwise there can be various combinations between bridge
structural properties and excitation loads that might have resulted in the same
measured vibration responses.
In recent years, several output- only identification techniques have been developed.
These include the natural excitation technique ( Caicedo et al., 2004; James et al.,
1996; Shen et al., 2003), the frequency domain decomposition ( Brincker et al., 2001;
Feng et al. 2004), the subspace decomposition ( Peeters et al., 2001), the random
decrement technique ( Asmussen and Brincker, 1996; Feng and Kim, 1998) and
various types of ARMA model fitting techniques ( Garibaldi et al., 1998; Huang, 2001;
Jensen et al., 1992). A common assumption in these output- only techniques is the
spatially uncorrelated white noise input model ( referred to hereafter as the
conventional excitation model). In mathematical terms, the conventional model has
an input covariance matrix that conforms to cov[ F( t), F( t+ Δt)]= δ ( Δt)⋅ Σ , where Σ is a
matrix constant and the Dirac’s delta function δ( Δt) is non- zero only when Δt = 0 .
Note that F( t) is the input vector at time t, a multivariate random process with its i- th
component Fi( t) being the random input at the i- th spatial location ( or degree- of-freedom,
DOF). Despite its mathematical attractiveness, the conventional excitation
model can be inadequate to account for the operational variations of the excitation on
a bridge, and moreover, it incorrectly excludes the correlation between excitation
processes at different spatial points when Δt ≠ 0, which indeed, is an intrinsic
characteristic of the traffic excitation.
In this section, a stochastic model of traffic excitation on bridges is developed based
on the physics of moving loads traversing a beam, taking into account various sources
of randomness, to accommodate the operational variation of the traffic on a bridge.
91
6.3.2 Physical Formulation of Traffic Loads on a Bridge
When a vehicle traverses a short- to medium- span highway bridge, which is usually
rather rigid with, for example, concrete box- girders, the bridge- vehicle system can be
sufficiently decoupled to a beam- moving force model ( Cebon, 1999; Pan and Li, 2002;
Pesterev et al., 2003; Pesterev et al., 2004; Schenk and Bergman, 2003; Yang et al.,
2000), i. e., the bridge ( modeled as an elastic beam) is subjected to a time- variant tire
force P( t) mo