See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/253662420

Structural health monitoring system of a cable-stayed bridge using a dense

array of scalable smart sensor network

Article  in  Proceedings of SPIE - The International Society for Optical Engineering · March 2010
DOI: 10.1117/12.852272

CITATIONS READS

8 682

10 authors, including:

Soojin Cho Hongki Jo

Ulsan National Institute of Science and Technology The University of Arizona
53 PUBLICATIONS   1,538 CITATIONS    66 PUBLICATIONS   1,321 CITATIONS   

SEE PROFILE SEE PROFILE

Jennifer Bridge Hyung-Jo Jung

University of Florida Korea Advanced Institute of Science and Technology
64 PUBLICATIONS   1,657 CITATIONS    242 PUBLICATIONS   4,466 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Flexible Fillers in PT Bridge Applications View project

Sustainability of Building Skin (A Building Integrated Photovoltaic and Wind Turbine System) View project

All content following this page was uploaded by Billie F. Spencer on 18 September 2014.

The user has requested enhancement of the downloaded file.

 Structural health monitoring system of a cable-stayed bridge using a
dense array of scalable smart sensor network

Soojin Choa, Shin Ae Jangb, Hongki Job, Kirill Mechitovc, Jennifer A. Riced, Hyung-Jo Jung*a,
Chung-Bang Yuna, Billie F. Spencer Jr.b, Tomonori Nagayamae, and Juwon Seof
a
Dept. of Civil and Environmental Engrg., KAIST, Daejeon 305-701, Korea;
b
Dept. of Civil and Environmental Engrg., Univ. of Illinois, Urbana, IL 61801, USA;
c
Dept. of Computer Science., Univ. of Illinois, Urbana, IL 61801, USA;
d
Dept. of Civil and Environmental Engrg., Texas Tech Univ., Lubbock, TX 79409, USA;
e
Dept. of Civil Engrg., Univ. of Tokyo, Tokyo 113-8656, Japan;
a
Long Span Bridge Research Team, Hyundai Inst. Const. Tech., Yongin 449-716, Korea;

ABSTRACT

This paper presents a structural health monitoring (SHM) system using a dense array of scalable smart wireless sensor
network on a cable-stayed bridge (Jindo Bridge) in Korea. The hardware and software for the SHM system and its
components are developed for low-cost, efficient, and autonomous monitoring of the bridge. 70 sensors and two base
station computers have been deployed to monitor the bridge using an autonomous SHM application with consideration of
harsh outdoor surroundings. The performance of the system has been evaluated in terms of hardware durability, software
reliability, and power consumption. 3-D modal properties were extracted from the measured 3-axis vibration data using
output-only modal identification methods. Tension forces of 4 different lengths of stay-cables were derived from the
ambient vibration data on the cables. For the integrity assessment of the structure, multi-scale subspace system
identification method is now under development using a neural network technique based on the local mode shapes and
the cable tensions.
Keywords: Smart sensor, wireless sensor network, structural health monitoring, cable-stayed bridge, test-bed

1. INTRODUCTION
Bridges are valuable national assets which ensure economic prosperity and public safety. However, many of them in
modern countries have reached their design life so that they need to be retrofitted, or even reconstructed, to be remaining
in service. In the United States, more than 149,000 bridges – almost quarter of total number (603,000) of bridge - are
estimated to be structurally deficient or functionally obsolete [1]. Thus, the ability to assess the structural condition and
possibly to increase the service life has been pursued widely by many researchers and engineers. By the efforts, many
bridges get to have their own structural health monitoring (SHM) system to assess the structural conditions and defeats.
However, the cost of SHM on large structures is still high, especially due to cabling work for constructing tethered
centralized systems, which limits the number of sensors to be installed and wider implementation to the other structures.
For example, it has been reported that the Bill Emerson Memorial Bridge is instrumented with 84 wired accelerometers
with an average cost of over $15K per channel [2], and Tsing Ma Bridge with more than 600 sensors with an average
cost of over $27K per channel [3]. If expensive cost of data acquisition systems, such as corresponding signal
conditioner, analog filter, and amplifier for each type of sensor, and maintenance cost are additionally considered, then
the total cost of a SHM system becomes higher.
Wireless smart sensor technology is getting attention as an economic alternative of long-term, scalable SHM of civil
infrastructure. With its wireless communication and onboard computing capability, it provides flexible deployment of
sensors, easy installation without tethering work, decentralized signal conditioning, and efficient data management
strategy at a lower cost than traditional wired monitoring systems. There are many wireless smart sensor platforms

Sensors and Smart Structures Technologies for Civil, Mechanical, and Aerospace Systems 2010,
edited by Masayoshi Tomizuka, Chung-Bang Yun, Victor Giurgiutiu, Jerome P. Lynch, Proc. of SPIE
Vol. 7647, 764707 · © 2010 SPIE · CCC code: 0277-786X/10/$18 · doi: 10.1117/12.852272

Proc. of SPIE Vol. 7647 764707-1

Downloaded from SPIE Digital Library on 12 Apr 2010 to 143.248.120.215. Terms of Use: http://spiedl.org/terms
developed in academia and industry at a low cost less than $1000 USD per unit with more than one sensing channel [4,
5].
Several researchers have employed wireless smart sensors to monitor bridge structures [4, 5, 6, 7, 8], providing important
insight into the opportunities and challenges for wireless smart sensor network (WSSN) technology for long-term
monitoring. Critical issues identified include: (1) data synchronization and recovery of missing data, (2) on-board and
decentralized processing, (3) autonomous operation, (4) power management and energy harvesting, and (5)
environmental hardening. To address these research challenges, an international test bed employing a cable-stayed bridge,
the second Jindo Bridge, in South Korea was developed using a state-of-the-art wireless smart sensor technology through
a trilateral research collaboration among Korea (KAIST), the US (University of Illinois at Urbana-Champaign), and
Japan (University of Tokyo). The hardware and software of wireless smart sensors are developed to overcome the issues
while keeping low cost. A multi-metric sensor board compatible with Imote2 [9] is developed for vibration-based
monitoring of large structures, and it is modified to be interfaced with an anemometer for wind environment
investigation around the structures. Software for the wireless smart sensors is developed in service-oriented architecture
(SOA), which has been proposed as a way to use this design philosophy in building dynamic, heterogeneous distributed
applications [10, 11]. Operating software for base stations are also developed for autonomous and effective monitoring
of large structures. After serious consideration of environmental hardening, a total of 70 wireless smart sensors and two
base stations have been deployed to monitor the bridge. Measurement has been carried out during 4 months, and the
performance of the system has been evaluated in terms of hardware durability, software stability, power management,
and energy harvesting options. Measured acceleration data is analyzed in the frequency domain to be used for monitoring
of the bridge. For deck and pylons, output-only modal identification are carried out using two output-only modal
identification method, FDD and SSI, and for cables, tension forces are estimated by a vibration method. Based on the
results, discussions are made on the monitoring strategy utilizing the WSSN for comprehensive SHM of the cable-stayed
bridge.

2. BRIDGE SHM SYSTEM USING WIRELESS SMART SENSORS

Figure 1 shows the concept of a bridge SHM system constructed using WSSN. For effective communication, wireless
smart sensor nodes can be clustered according to their positions, and hierarchy is formed in the cluster. A cluster is
governed by a head node, which has more powerful functionality than the others (leaf nodes) in the cluster. The head
node gets commands from the gateway node, interacts with leaf nodes in the cluster, and sends data to the gateway node.
A gateway node, which is different from a wireless smart sensor node due to its lack of sensing capability, is connected
to the base station and communicates with head nodes directly. If there is no head node, then it communicates with all
leaf nodes directly. In this section, the hardware and software of bridge SHM system using WSSN will be described.
Cluster n
Head node
Cluster n+2
Cluster n‐1
Cluster n+1

Leaf nodes

Base Station

1. When head nodes exist, gateway node communicates with

head nodes, not leaf nodes.
2. When head nodes do not exist, gateway node communicates
with leaf nodes directly. Gateway node

Figure 1. Concept of bridge SHM system using wireless smart sensor networks

Proc. of SPIE Vol. 7647 764707-2

Downloaded from SPIE Digital Library on 12 Apr 2010 to 143.248.120.215. Terms of Use: http://spiedl.org/terms
 2.1 Hardware of wireless smart sensors
The key hardware components of a wireless smart sensor node are an Imote2 [9], a sensor board, and a battery board.
The Imote2 is one of high-performance wireless computing modules, having an Intel’s PXA271 XScale® processor
running at 13-416 MHz and an MMX DSP coprocessor. It contains sufficient memory capacity of 256kB SRAM, 32MB
FLASH, and 32MB SDRAM, which enables long-time measurement as well as on-board manipulation of the measured
data. It has an onboard antenna for 2.4GHz wireless communication, and an alternative 2.4GHz antenna can be equipped
additionally for field applications which require longer wireless communication.
An Imote2 can be interfaced with a sensor board, either an SHM-A (structural health monitoring-acceleration) board to
measure multi-metric data or an SHM-W (structural health monitoring-wind) sensor board to measure wind speed and
direction by interfacing with a 3-axis anemometer. The sensor boards have been designed to monitor civil infrastructure
through the Illinois SHM Project [12], an interdisciplinary collaborative effort by the researchers in civil engineering and
computer science at the University of Illinois at Urbana-Champaign. As shown in Figure 2(b), the SHM-A board has a 3-
axis accelerometer (ST Microelectronic’s LIS344ALH), and the analog acceleration signals from the accelerometer are
digitized by the embedded Quickfilter QF4A512, which has a 4-channel, 16-bit analog to digital converter (ADC) and
programmable signal conditioner with user-selectable sampling rates and programmable digital filters. The SHM-A
board also contains temperature, humidity, and light sensors. This board also provides extra external input connector to
measure data from various types of sensors such as anemometers and strain gages. Four sampling frequencies (10, 25,
50, 100 Hz) have been pre-programmed on the SHM-A board for bridge monitoring applications, and the sampling rate
can be modified by designing appropriate filters for the QF4A512. To monitor the wind environment, which is one of the
most critical factors which affect the bridge responses and condition in Korea, a SHM-W board has also been developed
by modifying the SHM-A board to have three external 0–5V input channels to be interfaced with a 3-D ultrasonic
anemometer as shown in Figure 2(c). The RM Young Model 81000 3-D ultra-sonic anemometer is selected due to its
high resolution (wind speed: 0.01m/s, wind direction: 0.1 degree), good accuracy (wind speed: ±1%, wind direction: ±2
degrees), and the long-term durability against harsh outdoor environment. The wind speed in horizontal and vertical
directions is recorded after the precise synchronization with the bridge acceleration data obtained from wireless smart
sensors adapting the SHM-A boards.

External Antenna COUIJGCOL

(Antenova 2.4GHz)

SHM-A Board (rev. 4.0)

Imote2

Battery Board
(IBB2400CA)
Oh VVJb
w/ 3 AAA COUIJGCOL COULJGCOL
Batteries
V BO9L V BO9L

(a) Wireless smart sensor (b) Components of SHM-A Board

(c) SHM-W Board and Anemometer (d) Solar panel (SPE-350-6) and lithium-polymer rechargeable
(RM Young Model 81000) battery for energy harvesting
Figure 2. Hardware components of wireless smart sensor

Proc. of SPIE Vol. 7647 764707-3

Downloaded from SPIE Digital Library on 12 Apr 2010 to 143.248.120.215. Terms of Use: http://spiedl.org/terms
The IBB2400CA battery board powers the Imote2 using a series of three 1.5V AAA batteries. It contains a power
management integrated circuit (PMIC) which enables powering the node using energy harvesting devices. In this study,
solar panels with rechargeable batteries have been adapted on several wireless smart sensor nodes to investigate the
feasibility of sustainable energy harvesting. The Solarworld SPE-350-6 solar panel (9V-350mA) was employed to the
battery board with the Powerizer lithium-polymer rechargeable battery, of which voltage can be fully charged up to 4.2V
with 10,000-mAh capacity.
For the field deployment, some adjustments have been made to the wireless smart sensor nodes. An Antenova gigaNova
Titanis 2.4 GHz external dipole antenna has been additionally equipped to secure longer wireless communication. The
battery boards with AAA batteries have been modified to employ a series of three 1.5V D-cell batteries with large
capacities (21000mAh) for longer operation without changing batteries. All the components of a wireless smart sensor is
stacked as shown in Figure 2(a) and contained in a hardened plastic enclosure to be protected from harsh outdoor
environment.
2.2 Software for wireless smart sensors
An open-source middleware services toolsuite, which interacts between the target SHM applications and operating
software of wireless smart sensors, was developed by ISHMP to help the civil engineers easily code into the wireless
smart sensor nodes [12]. The toolsuite, called ISHMP Services Toolsuite, contains basic middleware to provide high-
quality sensor data and to transfer the data reliably to the base station via wireless communication as well as a library of
numerical algorithms.
The toolsuite components are categorized into foundation services, tools and utilities, application services, and
continuous and autonomous monitoring services. The foundation services provide the fundamental functionalities to
measure synchronized sensor data with high confidence [13, 14]. The application services are the numerical algorithms
to implement SHM applications on the Imote2, including modal identification and damage detection algorithms. The
tools and utilities support network maintenance and debugging. This category has essential services for full-scale
monitoring as well as sensor maintenance. Several key services will be explained here: RemoteSensing is the application
for the remote data measurement; DecentralizedDataAggretation is the application for remote data measurement and
subsequent decentralized on-board computation of the correlation functions of measured data in localized clusters;
imote2comm is a terminal to interact with a gateway node; RemoteVbat is an application for checking the battery level of
remote sensors; and TestRadio is the tool for assessing radio communication quality.
2.3 Base station
The base station controls the network by (1) sending messages to the leaf nodes, (2) storing the transmitted data from the
WSSN, (3) processing received data, and (4) transferring the data to the remote server via internet. In this study, a base
station is composed of an industrial personal computer (PC), an uninterrupted power supply (UPS) backup, a gateway
node, and ADSL internet modem contained in an environmentally hardened enclosure for outdoor application as shown
in Figure 6. The gateway node consists of an Imote2 stacked on an IIB2400 interface board [9] connected to the PC via a
USB/UART port without a sensor board.
2.4 Operating software of wireless smart sensor network
To operate WSSN with large number of wireless smart sensor nodes, 3 problems are seriously considered: (1) energy
saving software architecture, (2) data inundation, and (3) continuous and autonomous operation. To save batteries of
large number of nodes, SnoozeAlarm, which allows the network to sleep default and wakes up periodically for a short
time to listen to broadcasted commands, is embedded into the wireless smart sensor nodes. To prevent data inundation
from large number of sensor nodes, ThresholdSentry, which wakes up sentry nodes at a preset interval to measure a short
period of data, is applied to the network along with selection of several sentry nodes. If the measured data of sentry
nodes exceeds a preset threshold which may indicate a severe event to the bridge, the sentry nodes send alarm to the
gateway node, and the base station subsequently wakes up the entire network for data measurement. AutoMonitor, an
autonomous SHM network management application which combines RemoteSensing, SnoozeAlarm, and
ThresholdSentry, is installed on the PC of the base station to enable the automatic, continuous monitoring with reduced
power consumption. For remote operation of WSSN via internet, VNC (Virtual Network Computing) server and FTP
(File Transfer Protocol) server are additionally installed on the PC to control the PC remotely and to download the
measured data remotely, respectively.

Proc. of SPIE Vol. 7647 764707-4

Downloaded from SPIE Digital Library on 12 Apr 2010 to 143.248.120.215. Terms of Use: http://spiedl.org/terms
 3. DEPLOYMENT OF BRIDGE SHM SYSTEM ON SECOND JINDO BRIDGE
3.1 Bridge Description
The Jindo Bridges are twin cable-stayed bridges connecting Haenam (a city on the mainland of Korea) with the Jindo
Island as shown in Figure 3. Each of these bridges consists of a 344-m main span and two 70-m lateral spans. The 2nd
Jindo Bridge has a streamlined steel box girder supported by 60 parallel wire strand (PWS) cables. The 2nd Jindo Bridge
is selected as the test bed under the permission of unfettered access by the authority because of (i) the existing SHM
system to provide the comparable data with measured one by WSSN and (ii) the complete design and construction
documents for full interpretation of the bridge.

Figure 3. Jindo Bridges (left: 2nd Jindo Bridge, right: 1st Jindo Bridge)
3.2 Sensor deployment strategy
The developed hardware and software framework described in Section 2 has been deployed on the 2nd Jindo Bridge to
realize a large-scale and autonomous SHM system using WSSN. The network topology was carefully determined to
ensure the reliable communication on the bridge. Since the communication range of Imote2 with an external antenna is
about 200m shorter than the total length of the bridge, the network was divided into two sub-networks with different
radio channels: one on the Jindo side and the other on the Haenam side as in Figure 4. A total of 70 leaf nodes were
deployed on the bridge. They contain SHM-A sensor boards to measure 3-axis acceleration mainly, except one node
containing a SHM-W board interfaced with an anemometer. The Jindo sub-network consists of 33 nodes with 22 nodes
on the deck, 3 nodes on the pylon, and 8 nodes on the cables. The Haenam sub-network consists of 37 nodes with 26
nodes on the deck, 3 nodes on the pylon, and 7 nodes on the cables.
Figures 4-5 show the locations of 70 leaf nodes and photos of various types of nodes. They were enclosed in water-tight
plastic enclosures for protection from moisture and dust of harsh outdoor environment. The deck/pylon nodes were
mounted using one-directional magnets attached on the bottom of enclosures, and the cable nodes were mounted on
aluminum plates with round interface to fit the round cables. An anemometer was installed on a 5m-tall steel bar at the
center of deck to prevent any interruption of the bridge on the wind measurement, while the leaf node with SHM-W
board incorporating the anemometer was installed underneath of the deck to secure the line-of-sight to the base station.
3.3 Details of established wireless smart sensor network
Each sub-network is controlled by a corresponding base station located at the tops of the concrete piers supporting the
steel pylons of the 1st Jindo Bridge, which is parallel to the 2nd Jindo Bridge. The locations of base stations are selected
to secure consistent line-of-sight communication with the leaf nodes.
The parameters of WSSN for SHM of the 2nd Jindo Bridge have been set as Table 1 after a series of laboratory and
preliminary in-field tests. For ThresholdSentry application, two types of sentry nodes are selected: vibration-sentry nodes
for acceleration measurement and a wind-sentry node for wind measurement. The combination of two types of sentry
nodes provides efficient measurement strategy during the events causing large responses of the bridge by wind and
traffics. For both types of sentry nodes, two types of threshold values were preset: for normal condition and for extreme
condition, such as typhoon and earthquake. One time of measurement in a day is set to be carried out for exceeding
threshold values of normal condition, while unfettered measurement is set for exceeding threshold values of extreme
condition as shown in Table 1.

Proc. of SPIE Vol. 7647 764707-5

Downloaded from SPIE Digital Library on 12 Apr 2010 to 143.248.120.215. Terms of Use: http://spiedl.org/terms
 P- JT1 Jindo side (RF- CH25) Haenam side (RF- CH20) P- HT1
Solar
P- JT1(5)
P- HT1(48)

C- J4 C- J6 C- J7 C- J8 C- H7 C- H6 C- H5 C- H4 C- H3
C- J3 C- H2
C- J1 C- J2 D- H13 C- H1
C- J5 D- H10
P- J1 D- J9 D- J11 P- H1
D- J4 D- H4 D- H1
D- J1
D- J5 D- J6 D- J7 D- J8 D- J10 D- H7
D- J2 D- J3 D- H12 D- H9 D- H6 D- H5 D- H3 D- H2
D- H11 D- H8
2GU20L2
Sensors on Pylons & Cables
P- JT1 P- HT1
(5) (56) (99)
(118) (48)
Solar P- JW1 P- HW1
C- HW3 Solar
West

C- JE1 C- JE2 C- JE3 C- JE4 C- JE6 C- JE7 C- JE8 C- HE7 C- HE6 C- HE5 C- HE4 C- HE3 C- HE2 C- HE1 P- JE1(65) P- JW1(56)
P- JE1 C- JE5 P- HE1
(151) (38) (102) (64) (4) (36) (153) (16) (49) (153) (67) (137) (80) (33) P- HE1(52) P- HW1(99)
(144) (32) x (52)

z y
(135) (18) z
Sensors on Deck
x y y
z
(149) D- HW9
z
y (153)D- HW12 x x
z
y
z y
x
(3) D- HW13 (34)D- HW10 z y
x
Solar z y
x x
East

(134) (1) (23) (35) (131) (132) D- JW10 (66) D- HW11 (41) (51) (113) (85) (2) (68)
(37) (17) (86) (6) (89)
D- JW1 D- JW2 D- JW3 D- JW4 D- JW5 D- JW6 D- JW7 D- JW8 D- JW9 D- JW11 D- HW8 D- HW7 D- HW6 D- HW5 D- HW4 D- HW3 D- HW2 D- HW1

WindSent ry(75)

D- JE1 D- JE2 D- JE3 D- JE5 D- JE6 D- JE7 D- JE8 D- JE9 D- HE8 D- HE7 D- HE6 D- HE5 D- HE3 D- HE2 D- HE1
D- JE4 D- JE11 D- HE4
(101) (40) (135) (67) (170) (19) D- HE11 (65) (152) (120) (55) (81) (133)
(25) (21) (104) D- JE10 (18) (131) (54)
Sentry x (40) x x
Sentry x D- HE13 D- HE10 Sentry HighSensitivity Sentry
z z
y (85) y
bAIOU
z
y
z y
z
z y (145) D- HE12 (150) D- HE9 y z y
x z y x
(70) (20)
x x

x x
Note:
Sensor directions of Deck nodes z y
z y
z 1. ( ) values mean node ID
Sensor directions of Deck nodes y Wind Sentry node on Deck 2. z- axis direction of deck sensors is downward
x

Figure 4. Locations of wireless smart sensor nodes deployed on the 2nd Jindo Bridge

(a) Node under deck (b) Node on pylon top (c) Node on cable (d) Anemometer at (e) Reference sensors at
(w/ solar panel) (w/ solar panel) center of deck center of the deck
Figure 5. Various types of wireless smart sensor nodes installed on the 2nd Jindo Bridge

Table 1. Network parameters

g96M
Parameter Value
No. of sampled data 5,000
Sampling rate 10 Hz
Anti-aliasing filter 4 Hz
No. of Sensing events 1 per day
EUCIO2flI
SnoozeAlarm sleep time 750 ms
SnoozeAlarm listen time 15 s
JoqGw
Threshold sensing interval 10 min
Threshold sensing time 10 sec
Threshold for vibration-sentry 10 mg (normal condition)
50 mg (extreme condition)
Threshold for wind-sentry 3 m/sec (normal condition)
8 m/sec (extreme condition)
Figure 6. Base station of Jindo-
side subnetwork

Proc. of SPIE Vol. 7647 764707-6

Downloaded from SPIE Digital Library on 12 Apr 2010 to 143.248.120.215. Terms of Use: http://spiedl.org/terms
 4. EVALUATION OF THE SHM SYSTEM
4.1 Battery power consumption
The battery voltage levels of leaf nodes have been monitored during 2 month using RemoteVbat command, provided by
ISHMP services toolsuite, and they are shown in Figure 7. The average voltage with three brand-new batteries on each
node was about 4.6V. From 27 August to 8 September, a series of measurements were carried out to optimize network
performance, which results in the rapid drop of battery power as shown in Figure 7(a). After 8 September, the
AutoMonitor application has controlled the network and the power consumption has become approximately linear. The
minimum onboard voltage required for sensing is about 3.6V so that it can be concluded that a series of three D-cell
batteries are able to last about 2 months of monitoring.
4.2 Energy harvesting using solar panel
The voltage levels of 8 nodes powered by solar panels and rechargeable batteries (5 on cables, 2 on pylon tops, and 1
under the deck) are also investigated to check the feasibility of solar-based energy harvesting. Figure 7(b) shows the
voltage levels of the rechargeable batteries of 6 nodes (for cables and deck) powered by solar panels during 1.5 months
of monitoring. It is shown that the voltage levels have stayed around 4.15V, except one node under the deck, whose solar
panel faces intentionally downward without being exposed to direct sunlight.
4.3 Measured acceleration and wind data
Figure 8 shows examples of the ambient acceleration data measured on the deck, pylon, and cable in 3 directions. The
amplitudes of the acceleration due to the passing traffic on the deck are found to be large enough for mode extraction,
especially for vertical modes (z-axis). Similar to the deck, the in-plane (perpendicular to the cable in the vertical plane)
vibration of the cable is much larger than the other components. The amplitudes are also found to be sufficiently large
for mode extraction, which will be used for estimation of the cable tension forces.

(a) Average on-board voltage (b) Battery status of rechargeable batteries with solar panels
Figure 7. Evaluation of battery and sustainable energy harvesting

Raw acceleration data (X axis) Raw acceleration data (X axis) Raw acceleration data (X axis)
5 5 20
acc.(mg)
acc.(mg)
acc.(mg)

0 0 0
-5 -5 -20
0 100 200 300 400 500 0 100 200 300 400 500 0 100 200 300 400 500
Raw acceleration data (Y axis) Raw acceleration data (Y axis) Raw acceleration data (Y axis)
5 5 20
acc.(mg)
acc.(mg)
acc.(mg)

0 0 0
-5 -5 -20
0 100 200 300 400 500 0 100 200 300 400 500 0 100 200 300 400 500
Raw acceleration data (Z axis) Raw acceleration data (Z axis) Raw acceleration data (Z axis)
20 5 20
acc.(mg)
acc.(mg)
acc.(mg)

0 0 0
-20 -5 -20
0 100 200 300 400 500 0 100 200 300 400 500 0 100 200 300 400 500
time(sec) time(sec) time(sec)

(a) On the deck of main span (b) At the top of a pylon (c) On a cable
Figure 8. Examples of measured acceleration (Jindo sub-network)

Proc. of SPIE Vol. 7647 764707-7

Downloaded from SPIE Digital Library on 12 Apr 2010 to 143.248.120.215. Terms of Use: http://spiedl.org/terms
The power spectral densities (PSD) of the vibration data have been investigated. Figure 9 shows PSDs of the deck
accelerations obtained from a wireless node and from the existing wired monitoring system. Though the data from wired
monitoring system was measured in 2007 while the wireless sensor data was measured in 2009, the peak frequencies of
two PSDs implying modal frequencies look very similar at around 0.44, 0.66, 1.05, 1.37Hz and so on.
The wind speed and direction has been successfully measured using the 3D ultra-sonic anemometer at the mid span. The
data are synchronized with vibration data measured by SHM-A sensor board. Figure 10 shows an example of measured
wind speed and direction. In the example data, the wind speed was 4-6 m/sec and the direction is between -10 to 20
degree to the longitudinal direction of the bridge.

15 Wind Speed

Wired system (in 2007) 8

WSSN (in 2009)

speed(m/s)
6

10
4
Amplitude

0 50 100 150 200 250 300 350 400 450 500

time(sec)
Wind Direction (vertical)
5 30

direction(degree)
20
10

0 -10
0 0.5 1 1.5 2 2.5 3 3.5 4 -20
0 50 100 150 200 250 300 350 400 450 500
Frequency(Hz) time(sec)

Figure 9. PSDs from the wired monitoring system Figure 10. Example of measured wind data
measured in 2007 and from deployed WSSN

5. DATA ANALYSES
5.1 Output-only modal identification
Modal properties such as natural frequencies, mode shapes, and modal damping ratios, play key roles for SHM of
bridges. To analyze the ambient (or operational) acceleration data excited by ambient sources, such as wind and traffic,
two output-only modal identification methods are employed in this study: frequency domain decomposition (FDD) [15]
and stochastic subspace identification (SSI) [16] methods.
Modal analyses are carried out on two sets of data obtained from Haenam- and Jindo- side WSSNs using both FDD and
SSI. Because the WSSNs are not synchronized to each other during the measurement, the data from each WSSN were
analyzed independently, and then combined subsequently. Table 2 gives the details of the identified modal properties
from independent WSSN, including the natural frequencies from two WSSNs. The results from different modal
identification methods are found to be consistent to each other. Several modes (DL1, DV2, DT1, and PB1) are found
undetected by FDD. Though longer acceleration records may result in modal properties in better fidelity extracted by
both FDD and SSI, longer data collection will drastically increase the time for transmitting data packets for a large-scale
WSSN. The identified natural frequencies by both methods are compared with those obtained from the existing wired
SHM system and from the FE analysis. The identified natural frequencies show excellent agreements with the
frequencies obtained from the wired monitoring system in 2007. The results are also found to be very close to those from
the FE analysis up to the 3rd vertical modes, while those for the higher modes are generally larger than the FE results.
However, the differences are found to be within 16%.
The modal properties from each WSSN are combined to provide the global information for SHM. Four overlapped
reference nodes at the mid span shown in Figure 5(e) are employed to construct the global mode shapes using the least-
square method. Examples of the combined mode shapes are compared with those from the FE analysis in Figure 11.
They show excellent agreements to each other with MAC values of 0.943-0.986, which reinforces the performance of the
current WSSNs. The accuracy of the combined modal information may be improved by a decentralized approach to
estimate global modal information, which is developed by [17]. Also, software is currently under development for
synchronization of two separated base stations and expected to be implemented on the 2nd Jindo Bridge

Proc. of SPIE Vol. 7647 764707-8

Downloaded from SPIE Digital Library on 12 Apr 2010 to 143.248.120.215. Terms of Use: http://spiedl.org/terms
 Table 2 Natural frequencies extracted by SSI and FDD (Haenam-side on 9/8/2009, Jindo-side on 9/11/2009)
No. Modes Mode Shape SSI (Hz) FDD (Hz) Wired FE analysis
Haenam- Jindo- Haenam- Jindo- monitoring (Hz)
side side side side system (Hz)
1 DL1 1st longitudinal 0.2998 0.2985 - - - 0.3137
2 DV1 1st vertical 0.4347 0.4380 0.4492 0.4492 0.4395 0.4422
3 DV2 2nd vertical 0.6619 0.6439 - 0.6445 0.6592 0.6471
4 DV3 3rd vertical 1.0371 1.0364 1.0352 1.0352 1.0498 1.0010
5 DV4 4th vertical 1.3481 1.3555 1.3379 1.3379 1.3672 1.2472
6 DV5 5th vertical 1.5755 1.5759 1.5723 1.5723 1.5869 1.3490
7 DV6 6th vertical 1.6618 1.6660 1.6602 1.6699 1.6602 1.4596
8 DT1 1st torsional 1.8278 1.8410 - - - 1.7888
9 DV7 7th vertical 1.8844 1.8860 1.8848 1.8848 1.8555 1.5858
10 DV8 8th vertical 2.2712 2.2731 2.2656 2.2754 2.3193 2.1154
11 PB1 1st bending 2.4107 2.3890 2.4121 - 2.3682 2.1392
12 DV9 9th vertical 2.8127 2.8266 2.8027 2.8320 2.8076 2.5612

(a) DV1 (MAC: 0.957)

(b) DV2 (MAC: 0.986)

(c) DV3 (MAC: 0.975)

(d) DV6 (MAC: 0.949)

(e) DV9 (MAC: 0.943)

Figure 11. Identified combined mode shapes (left) and those from FE model analysis (right)
5.2 Estimation of cable tension forces
The 2nd Jindo Bridge has a total of 60 stay cables of parallel wire strand (PWS). The bridge is symmetric along the
longitudinal as well as the lateral directions. Each pylon holds 30 cables; 15 cables on each of east and west sides. The
cables are categorized into 4 groups with different cross sections (i.e., φ 7 ×139 , φ 7 ×109 , φ 7 × 73 , and φ 7 ×151 ) as
shown in Figure 12. The above designations indicate the number of steel wires of a 7mm in diameter. High-damping
rubber dampers are installed on cable anchors to reduce the wind-induced vibration of the cables.

Proc. of SPIE Vol. 7647 764707-9

Downloaded from SPIE Digital Library on 12 Apr 2010 to 143.248.120.215. Terms of Use: http://spiedl.org/terms
Among the cables, 10 east-side cables with larger tension forces are selected to estimate the tension forces as shown in
Figure 12. Table 4 shows the general properties of the cables. The effective lengths of the cables are obtained from the
work by [18]. It is worthy to note that the leaf nodes monitoring the cables are installed at about 3m height from the
onset of the cables connected to the deck, which is available for regular battery change while the rubber dampers does
not affect the response much.

bsLs 1161 bsLs 1161

1!LJqC

Figure 12. Arrangement of stay cables and wireless sensors on cables (sensor numbers in parentheses)
Table 3. Properties of the cables monitored
Cables HC4, JC4 HC6, JC6 HC9, JC9 HC13, JC13 HC15, JC13
Cable type φ 7 × 151 φ 7 × 151 φ 7 × 73 φ 7 × 109 φ 7 × 139
Elasticity (tonf/mm2) 20.0 20.0 20.0 20.0 20.0
Area (mm2) 5811.0 5811.0 2809.0 4195.0 5349.0
Length (m) 97.10 65.00 83.17 141.76 174.15
Effective length (m) 95.38 63.33 79.01 136.87 169.69
Unit mass (ton/m) 0.00486 0.0486 0.00236 0.00354 0.00448
Design cable sag (mm) 256.0 96.0 221.0 537.0 809.0
Design tension force (tonf) 237.0 271.0 90.0 160.0 202.0
Allowable tension force (tonf) 470.0 470.0 227.0 339.0 433.0

Table 4. Comparison of estimated tension forces with those from previous regular inspections in 2007 and 2008
Cables (East-side) Estimated tension forces (tonf) Initial design Maintenance
WSSNs Previous inspections values (tonf) thresholds (tonf)
in 2009 in 2007 in 2008
Haenam- HC4 274.0 (2.04)* 262.7 268.4 246.2 329
side HC6 294.7 (-3.19)* 304.6 304.1 271.8 329
HC9 89.3 (0.90)* 86.9 88.5 87.6 158
HC13 170.2 (3.00)* 164.0 165.1 163.6 237
HC15 224.9 (2.18)* 219.9 220.0 204.8 303
Jindo- JC4 254.0 (1.30)* 245.1 250.7 245.9 329
side JC6 274.5 (-1.09)* 282.0 277.5 271.5 329
JC9 88.5 (2.15)* 85.5 86.6 88.2 158
JC13 154.3 (2.33)* 148.3 150.7 164.1 237
JC15 216.8 (0.14)* 214.1 216.5 201.3 303
*
The differences from regular inspection in 2008 are shown in the parentheses.
Figure 8(c) shows an example of the 3-axis acceleration data measured on a cable of Jindo-side span. The frequency-
domain analyses were carried out to obtain Fourier amplitude spectra (FAS) for the cable motions along with the FAS
for deck motions, and it was found out that many peaks are apparent in the FAS for the vertical cable vibration, while
some of them are from the deck motion owing to the interaction between the deck and cables particularly in the vertical
direction. Since the cables of the 2nd Jindo Bridge are slender with small sag, the modal properties of the cable are very
similar in the vertical and lateral directions. Taking advantage of this characteristics, the natural frequencies of cables,
not frequencies caused by the cable-deck interaction, are extracted. For example, five natural frequencies are identified

Proc. of SPIE Vol. 7647 764707-10

Downloaded from SPIE Digital Library on 12 Apr 2010 to 143.248.120.215. Terms of Use: http://spiedl.org/terms
for two cables: 0.645, 1.294, 1.948, 2.598, and 3.247Hz for Cable JC15 with Node C-JE8, and 0.772, 1.514, 2.275,
3.027, and 3.789Hz for Cable JC13 with Node C-JE7. The natural frequencies are found to be almost proportional to the
order of modes ( n ), which is also a dynamic characteristic of a slender cable with little bending and sag effect [19, 20].
Based on the identified dominant frequencies, the tension forces of 10 cables are estimated as Table 5. The estimated
tension forces are compared with those obtained from two previous regular inspections in 2007 and 2008, those from the
initial design, and those from the maintenance thresholds which are 60% of allowable tension forces of the cables [21].
The current estimations are found to be very close to the tension forces from two previous inspections with less than 4%
difference. The tension forces of 8 cables have increased slightly with time, while those of 2 cables (HC6 and JC6)
supporting the side spans have slightly decreased. The estimated cable tension forces are generally larger than the initial
design values (10% at maximum) except JC13. All cable tension values are well within the maintenance thresholds,
indicating that the cables are in safe operation.

6. CONCLUDING REMARKS
This paper reports on the collaborative research on a state-of-the-art wireless smart sensor technology among Korea
(KAIST), the US (University of Illinois at Urbana-Champaign), and Japan (University of Tokyo). A structural health
monitoring system using 70 wireless smart sensors and 2 base stations has been successfully deployed on a cable-stayed
bridge, the 2nd Jindo Bridge, in South Korea to verify the performance of the system along with its components. The
result of the deployment can be summarized as below:
1) Hardware and software of wireless smart sensor node are developed for low-cost and efficient monitoring. SHM-A
and SHM-W sensor boards were developed to be interfaced with Imote2 to measure 3-axis acceleration and wind. Basic
software is also developed in service oriented architecture.
2) Operating software for base stations are also developed for autonomous and effective monitoring of large structures in
consideration of energy saving software architecture (SnoozeAlarm), data inundation (ThresholdSentry), and continuous
and autonomous operation (AutoMonitor).
3) After the deployment, it is shown that the wireless smart sensor network can be operated for 2 month without battery
change with the help of autonomous and efficient monitoring strategy, and that the solar panels incorporating with
rechargeable batteries provides sustainable energy during the monitoring period, which shows the feasibility of solar-
based energy harvesting for wireless smart sensor network.
4) Modal properties of the bridge were successfully obtained from the ambient acceleration measurements through
WSSNs using both FDD and SSI. The natural frequencies currently identified using the WSSNs are found to be in
excellent agreements with those previously obtained by the existing wired sensors. The extracted mode shapes show
excellent agreements with those from the FE analysis.
5) By utilizing the symmetric property of slender cables with small sag, the natural frequencies of cables are extracted by
frequency-domain analysis from the measured 3-axis acceleration data with complementary use of the lateral vibration
data of the cables. Tension forces of 10 cables estimated from the natural frequencies of the cable using a vibration
method are found to be very close to those from 2 previous regular inspections within 4% difference.
6) A substructural damage identification method for a cable-stayed bridge is now under development with full utilization
of decentralized computing capability through WSSN. Substructural modal information of the deck/pylon and cable
tension forces may be effectively combined to assess the comprehensive structural integrity of the bridge structure.

REFERENCES

1. http://www.fhwa.dot.gov/Bridge/deficient.cfm
2. Celebi, M. [Seismic instrumentation of buildings (with emphasis on federal buildings)], Special GSA/USGS Project,
an administrative report, United States Geological Survey, Menlo Park, CA (2002).
3. Farrar, C. R., “Historical overview of structural health monitoring”, Lecture Notes on Structural Health Monitoring
using Statistical Pattern Recognition, Los Alamos Dynamics, Los Alamos, NM (2001).

Proc. of SPIE Vol. 7647 764707-11

Downloaded from SPIE Digital Library on 12 Apr 2010 to 143.248.120.215. Terms of Use: http://spiedl.org/terms
 4. Lynch, J. P., Wang, Y., Loh, K. J., Yi, J. H., and Yun, C. B., “Performance Monitoring of the Geumdang Bridge
using a Dense Network of High-resolution Wireless Sensors”, Smart Mat. Struct. 15, 1561-1575 (2006).
5. Jang, S.A., Rice, J.A., Li, J., Jo, H., Spencer, B.F. Jr., and Wang, Z., “Structural Monitoring of a Historic Truss
Bridge using a Wireless Sensor Network”, Proc. Asia-Pacific Workshop on Struct. Health Monitoring (2009).
6. Nagayama, T. and Spencer Jr., B.F., [Structural Health Monitoring using Smart Sensors, NSEL Report Series No.
1], University of Illinois at Urbana-Champaign, Urbana, (http://hdl.handle.net/2142/3521) (2007).
7. Kim S., Pakzad, S., Culler, D., Demmel, D., Fenves, G., Glaser, S., and Turon, M., “Health Monitoring of Civil
Infrastructures using Wireless Sensor Networks”, Proc. 6th Int. Conf. Information Processing in Sensor Networks,
254-263 (2007).
8. Pakzad, S., [Statistical Approach to Structural Monitoring Using Scalable Wireless Sensor Networks - Ph.D.
Dissertation], University of California – Berkeley, Berkeley, CA (2008).
9. Crossbow, Inc., “Imote2 Hardware Reference Manual.” Revision A, PN: 7430-0409-01,
http://www.xbow.com/Support/Support_pdf_files/Imote2_Hardware_Reference_Manual.pdf (2007).
10. Singh, M.P. and Huhns, M.N., [Service-Oriented Computing: Semantics, Processes, Agents], John Wiley and Sons,
New Jersey (2005).
11. Tsai, W.T., “Service-Oriented System Engineering: A New Paradigm,” Proc. IEEE Int. Workshop on Service-
Oriented Systems Engineering, 3-8 (2005).
12. http://shm.cs.uiuc.edu
13. Mechitov, K., Kim, W. Y., Agha, G., and Nagayama, T., "High-Frequency Distributed Sensing for Structure
Monitoring," Proc. 1st Int. Workshop on Networked Sensing Systems, 101-104 (2004).
14. Rice, J.A., Mechitov, K.A., Spencer Jr., B.F. and Agha, G., “A Service-Oriented Architecture for Structural Health
Monitoring using Smart Sensors,” Proc. 14th World Conference on Earthquake Engineering, Beijing, China (2008).
15. Brincker, R. Zhang, L. and Andersen, P., “Modal identification of output-only systems using frequency domain
decomposition” Smart Mater. Struct. 10, 441-445 (2001).
16. Overschee, P. V. and De Moor, B., “Subspace algorithms for the stochastic identification problem”, Automatics
29(3), 649-660 (1993).
17. Sim, S. H., Spencer Jr., B. F., Zhang, M. and Xie, H., “Automated decentralized modal analysis using smart
sensors”, Struct. Control Health Monit.., DOI: 10.1002/stc.348, May (2009).
18. Park, Y. S., Choi, S. M., Yang, W. Y., Hong, H. J. and Kim, W. H., “A study on tension for cables of a cable-stayed
bridge damper is attached (in Korean)”, J. Korean Soc. Steel Constr. 20(5), 609-616 (2008).
19. Irvine, M., [Cable structures], Dover Publications, Inc., New York (1981).
20. Cho, S., Lynch, J. P., Lee, J-J. and Yun, C-B., “Development of an automated wireless tension force estimation
system for cable-stayed bridge”, J. Intell. Mat. Syst. Str., 21(3), 361-376 (2010).
21. ATMACS, [Consulting for the measurement systems for integrated entrusted management of long-span bridges -
The 1st and 2nd Jindo Bridges (in Korean)], Sungnam, Gyunggi-do, Korea (2008).

Proc. of SPIE Vol. 7647 764707-12

View publication stats Downloaded from SPIE Digital Library on 12 Apr 2010 to 143.248.120.215. Terms of Use: http://spiedl.org/terms