The 2011 World Congress on

Advances in Structural Engineering and Mechanics (ASEM'11+)

Seoul, Korea, 18-22 September, 2011

Wireless Smart Sensor Technology for Monitoring Civil Infrastructure:

Technological Developments and Full-scale Applications

*B.F. Spencer Jr.1) and Soojin Cho2)

1), 2) Department of Civil and Environmental Engineering, University of Illinois at

Urbana-Champaign, Urbana, IL 61801, USA

1)bfs@illinois.edu,

2)soojin@illinois.edu

ABSTRACT

While much of the technology associated with wireless smart sensors (WSS) has been

available for over a decade, only a limited number of full-scale implementations have been realized, primarily due to the lack of critical hardware and software elements. Using MEMSIC’s Imote2, researchers at the University of Illinois at Urbana-Champaign (UIUC) have developed a flexible WSS framework for full-scale, autonomous SHM that integrates the necessary software and hardware elements, while addressing key implementation requirements. This paper discusses the recent advances in the development of this framework, as well as its successful deployment at full-scale on the 2nd Jindo Bridge in South Korea and the Government Bridge at the Rock Island Arsenal in Illinois. 1. INTRODUCTION

The United States has estimated its investment in built civil infrastructure to be $20 trillion.

For most industrialized nations, their annual expenditure on civil infrastructure amounts to between 8-15% of their GDP (US. Census Bureau 2004; Jensen 2003). This investment is only likely to increase; for example, the U.S. Department of Transportation (FHWA 2007) has reported that the capital investment to preserve highways and bridges increased 58.3% from $23.0 billion in 1997 to $36.4 billion in 2004. The growth in American spending on system rehabilitation is outpacing the growth in expenditures on system expansion, growing 27.9% from $21.5 billion in 1997 to $27.5 billion in 2004. Indeed, recent catastrophic failures have focused public attention on the declining state of the aging American infrastructure. These concerns apply not only to the nation’s civil engineering structures, such as bridges, highways, and buildings, but also to other structures such as the aging aircraft fleet commercial airlines currently in use.

Bridges account for a large part of the capital investment in the construction of road networks and represent a key element in terms of the safety and functionality of the entire highway system. Visual inspection is the current practice in monitoring the safety of bridges. However, high costs limit the use of visual inspection to infrequent

1) Nathan M. and Anne M. Newmark Endowed Chair in Civil Engineering

2) Postdoctoral Research Associate

ICOSSS Keynote Paper

occurrences. Moreover, a recent study in the United Sates by the Federal Highway Administration (FHWA) reported significant variability in the ratings assigned by a cohort of highly trained inspectors (Moore et al. 2001). As a result, failure of bridges in the United States is not as rare of an event as the public may believe. For example, between 1989 and 2000, a total of 134 bridges are reported to have partially or totally collapsed in the United States due to triggering events (e.g., earthquake or vehicle collision), design and construction error, and undetected structural deterioration (e.g., scour, fatigue) (Wardhana and Hadipriono 2003). Visual inspection alone does not appear to be adequate to protect civil infrastructure.

Structural health monitoring (SHM) of civil infrastructure in real time offers the potential to reduce inspection and repair costs, as well as the associated downtime, all while providing increased public safety. The principles of SHM and its application promise to aid not only in the inspection of existing infrastructure, but also lifetime monitoring of future construction projects. Moreover, after natural disasters, it is imperative that emergency facilities and evacuation routes, including bridges and highways, be assessed for safety in a timely manner. Continuous, automated structural health monitoring can facilitate these goals.

Many recently constructed bridges have in-depth, yet costly, monitoring systems. For example, the total cost of the monitoring system on the Bill Emerson Memorial Bridge in Cape Girardeau, Missouri is approximately $1.3 million for 86 accelerometers, making the average installed cost per sensor a little more than $15,000; this cost is not atypical of today’s wired SHM systems (Celebi 2006). Wireless sensors are an attractive alternative to such wired systems, offering the potential for low-cost and reliable structural health monitoring.

While wireless sensor has been commercially available for over a decade, only a limited number of full-scale implementations have been realized, primarily due to the lack of critical hardware and software elements. An example of one such critical issue is network scalability. A wireless sensor network implemented on the Golden Gate Bridge in 2008 took approximately 10 hours to collect 80 seconds of data (sampled at 1000 Hz) from 56 sensor nodes to a central location (Pakzad et al. 2008). To assist in dealing with the large amount of data that is generated by a monitoring system, on-board processing can be done locally on the sensor’s embedded microprocessor. This strategy is a radical departure from the conventional approach to monitoring structures.

This article discusses recent advances in development of low-cost, wireless means for continuous and reliable structural health monitoring, as well as its successful deployment at full scale on the 2nd Jindo Bridge in South Korea. The SHM system on the Jindo Bridge consititutes the first long-term, dense deployment of a wireless sensor network to monitor civil infrastructure and demonstrates the tremendous potential of this technology.

2. WIRELESS SMART SENSORS

Wireless smart sensors (WSS) differ from traditional wired sensors in significant ways.

Each sensor has an on-board microprocessor that can be used for digital signal processing, self-diagnosis, self-calibration, self-identification, and self-adaptation functions. Furthermore, all WSS platforms have thus far employed wireless communication technology. WSS technology has seen substantial progress through interdisciplinary research efforts to address issues in sensors, networks, and application-specific algorithms. This section discusses this progress in terms of both hardware and software.

2.1 Imote

Nearl

and data

with this

(MEMSI

SHM. T

can be se

416MHz

onboard

Sensing w

MEM

can meas

with digi

rate erro

inadequa

board can

Fi

2.2 Enab

This

specifica

Multi-m

versatile,

MEMSIC

applicatio

4-channe

anti-alias

board co

ranging ±

light).

High sen

performa

high-sens

low-leve

e2 and Basi

ly all comm

throughput

hardware.

IC 2010), is

The Imote2

elected base

z. It has 25

calculation

with the Im

MSIC provid

sure 3-axes

ital convers

ors, and ina

ate for SHM

n be found i

ig. 1. Imote2

bling Hardw

section des

ally designed

metric senso

, multi-met

C as the I

ons (Rice a

el Quickfil

sing filter, u

ontains the

±2g), the S

nsitivity se

ance, its r

sitivity sens

l vibration

ic Sensor Bo

mercially av

t rates; meet

The Imot

s a wireless

2 includes a

ed on applic

56K SRAM

required fo

mote2 is facil

des a Basic

of accelerat

sion. Due

adequate an

M applicatio

in Nagayam

2: top (left)

ware for SHM

scribes sens

d to meet th

or board

tric sensor

ISM400; s

and Spencer

lter analog

user-selectab

ST Micro

Sensirion ST

nsor board

resolution

sor board (d

(i.e., less

oard

vailable wir

ting the dem

te2 (see Fig

sensor plat

a high-perf

cation dema

M, 32MB FL

or SHM app

litated by se

Sensor Boa

tion, light (

to the lack

ntialiasing

ns. More

ma et al. (20

, right (mid

batte

M

sor boards d

he needs of

To addres

board, de

see Fig. 2)

r 2009). T

g-to-digital

ble samplin

electronics

TH11 (for t

d While t

may be i

designated S

than 1.0 m

reless senso

mands of SH

g. 1), develo

tform desig

formance X

ands and po

LASH, and

plications, a

ensor board

ard (ITS400

TSL2561),

of flexibili

filters in t

details rega

007).

ddle), and ba

ery board (r

developed a

SHM applic

s the short

signated SH

), was dev

The key com

converter

ng rates and

LIS344AL

temperature

the LIS344A

insufficient

SHM-H; see

mg) (Jo et

or nodes w

HM has prov

oped by Inte

gned for dat

X-scale proc

ower manag

32MB SDR

as well as s

d(s) stacked

0) to interfa

temperatur

ty in select

the ITS400

arding the l

asic sensor b

ight).

at Illinois fo

cations.

tcomings o

HM-A (sub

veloped for

mponent of

(ADC) (Q

customizab

LH accelero

e and humi

ALH accele

for low-l

e Fig. 3) ha

al. 2010a,b

were designe

ven to be ex

el and avail

ta intensive

cessor (PXA

gement, rang

RAM, which

storage of lo

on Imote2 v

ce with the

e, and relati

ting samplin

0, this boar

imitations o

board stack

or use with t

of the ITS4

bsequently

r structural

the ISM40

QF4A512)

ble digital f

ometer (for

dity), and t

erometer pr

level ambi

as been deve

b). The SH

ed for low

xtremely ch

lable from M

application

A27x), who

ging from 1

h enables th

onger measu

via two con

Imote2. Th

ive humidit

ng rate, the

rd was fou

of the ITS4

ked on Imote

the Imote2

400 sensor

commercia

l health m

0 board is t

that prov

filters. The

r 3-axis acc

the TAOS

rovides goo

ient vibrat

eloped for m

HM-H boar

sampling

hallenging

MEMSIC

ns such as

ose speed

13MHz to

he intense

urements.

nnectors.

he ITS400

ty (SHTx)

sampling

und to be

00 sensor

e2 and

that were

board, a

alized by

monitoring

the 16-bit

vides the

e ISM400

celeration

2561 (for

od overall

tion. A

measuring

rd uses a

low-nois

in the z-

y-axis).

(z-axis),

0.3mg (x

This

reference

taking th

data, the

sensors c

Fig.

General

Quickfilt

(0~5V o

external

accommo

accommo

but it has

e accelerom

direction, a

The typic

that would

x- and y-axe

Fig. 2.

high-sensit

e sensor for

he cross-corr

noise levels

can improve

3. SHM-H

purpose

ter-based se

or -5~5V ra

sensors.

odated thro

odate a 3D

s been exten

Fig. 4. S

meter (Silico

and the ST

al RMS no

be enough

es).

ISM400 bo

tivity senso

r decentraliz

relation (or

s can be low

e the perform

board: pers

data acqu

ensor board

ange). Th

Additional

ugh the ter

ultra-sonic

nded subseq

HM-DAQ b

on Designs

Microelectr

oise levels o

h for measu

oard: perspe

or board h

zed data ag

r cross spect

wered consi

mance of th

spective (lef

uisition boa

, opens all

he SHM-DA

lly, digital

rminal block

anemomet

quently to in

board: persp

SD1221-00

ronics LIS3

of the SHM

uring the am

ective (left),

has been sh

gregation s

trum) of a l

iderably; as

he entire sen

ft), top (mid

ard The

the 4 chann

AQ has ter

sensors u

k. The SH

ter for wind

nterface wit

pective (left

02L) having

344ALH for

M-H board a

mbient vibra

, top (middl

hown to b

trategies em

low-noise s

a result, a s

nsor network

ddle), and b

SHM-DAQ

nels of the A

rminal bloc

using the I

HM-DAQ b

d environme

h generic an

ft), top (mid

g the noise d

r the other

at 20Hz ba

ation of civ

le), and bott

e exception

mployed in

ensor data w

small numb

k (Jo et al. 2

bottom (righ

Q board (s

ADC for ex

cks for dir

I2C or SP

board was in

ent monitor

nalog senso

ddle), and bo

density of 5

acceleratio

andwidth ar

vil infrastruc

tom (right).

nally effect

large WSSN

with low-co

er of high-s

2010b).

ht). (Jo et al.

see Fig. 4)

xternal anal

rect interfac

PI interface

nitially dev

ring (Jo et a

ors.

ottom (right

5.0μg/√Hz

on (x- and

re 0.05mg

cture, and

tive as a

Ns . By

ost sensor

sensitivity

. 2010a).

, another

log inputs

cing with

e can be

veloped to

al 2011a),

t).

2.3 Enabling Software for SHM SHM applications implemented on WSSNs require complex programming, ranging from

network functionality to algorithm implementation. Software development is made even more difficult by the fact that many smart sensor platforms employ special-purpose operating systems without support for common programming environments. The extensive expertise required to develop SHM applications has severely limited the use of smart sensing technology for monitoring of civil infrastructure. This section discusses an enabling software framework for SHM.

ISHMP Services Toolsuite A software framework for continuous and reliable monitoring of civil infrastructure using a dense network of smart sensors was developed under the Illinois Structural Health Monitoring Project (ISHMP), a collaborative effort between researchers in civil engineering and computer science at the University of Illinois at Urbana-Champaign. ISHMP addressed the complexity associated with creating WSSN applications by developing a software framework based on the design principles of Service Oriented Architecture (SOA) (Singh and Huhns 2005, Tsai 2005). This framework provides a suite of services implementing the key middleware infrastructure necessary to provide high-quality sensor data and to reliably communicate it within the sensor network, as well as number of commonly used numerical algorithms. This framework is intended to allow researchers and engineers to focus their attention on the advancement of SHM approaches without having to concern themselves with low level networking, communication and numerical sub-routines. This software is open-source and available for public use at http://shm.cs.uiuc.edu/software.html.

The components of the service-based framework provided by the ISHMP Services Toolsuite can be divided into four primary categories: (1) foundation services to provide commonly used wireless sensor functionalities required to support high-level applications, including reliable communication and sensing functionalities, (2) application services to provide the numerical algorithms necessary to implement SHM applications on the Imote2s, (3) continuous and autonomous monitoring services to support continuous and autonomous WSSN operation with efficient power management, and (4) tools and utilities for testing and debugging that are essential for any large-scale or long-term WSSN deployment. More detailed information about each category can be found in Rice and Spencer (2009), Rice et al. (2010) and Sim and Spencer (2009).

Autonomous operation and power management Autonomous operation is a key issue in managing a large-scale WSSN for SHM. One of the services in the ISHMP Service Toolsuite, AutoMonitor, was designed to facilitate autonomous operation of all of the necessary functions in a full-scale WSSN deployment.

To minimize power usage in the WSSN, AutoMonitor incorporates with power-efficient management services for sensor nodes, called SnoozeAlarm and ThresholdSentry. The Imote2 allows the processor to be put into a deep sleep mode with minimal power consumption by making only the processor clock awake. Effective utilization of the WSS’s sleep mode is fundamental to achieving power efficient WSSN. The SnoozeAlarm service places the sensor nodes in the deep-sleep state (for 10 seconds), waking periodically for a brief time (in current implementation, 0.6 sec for single-hop and 0.75 sec for multi-hop) to listen for beacon signals sent from the gateway node; should a signal be heard, the node becomes fully active (Rice and

Spencer 2009). To monitor activity in the network, a threshold triggering strategy is utilized. With parameters received from gateway node at the beginning and stored in flash memory, the ThresholdSentry application runs on the designated sentry nodes periodically execute sentry task, i.e., collecting data for given period of time and checking if the threshold value is exceeded. If this threshold is exceeded, the sentry nodes send notification to the gateway node, then the gateway node wakes up the entire network and initiates designated network-wide tasks, such as centralized sensing or decentralized data aggregation.

The carefully designed scheduler embodied in AutoMonitor enables power-efficient and continuous management of the WSSN, as well as combined operation of the other such as decentralized analysis, sensor diagnosis, and multi-hop communications (Rice and Spencer 2009).

Multi-hop communication Large-scale deployment of WSSN gives rise to the need for multi-hop communication to provide adequate wireless coverage. The limited radio range of general WSS using IEEE802.11 or IEEE802.15 wireless protocols (IEEE Std 1990), combined with the impact of various environmental effects on the radio transmission, make direct communication between all nodes impractical. On the other hand, an important requirement of any communication scheme is data transfer reliability. Multi-hop communication, together with appropriate packet-loss compensation, addresses these issues by allowing sensors to cooperate to reliably deliver data between nodes outside of direct communication range.

The Ad-hoc On-demand Distance Vector (AODV) protocol is widely used routing methods to discover optimal routes for multi-hop communication (Perkins et al. 2003). Fig. 5 illustrates the AODV method. The route request (RREQ) message initiated from a source node is rebroadcasted by neighbor nodes until it reaches the destination node. Then, a route reply (RREP) message originating at the destination or intermediate nodes knowing a path to the destination is sent back to the source node, establishing the route in the reverse order. The route is selected with minimum hop count among the received routes.

Fig. 5. Example of AODV route discovery (Nagayama et al. 2010).

The modified AODV protocol, termed General Purpose Multi-hop (GPMH), is developed to support diverse data flow patterns such as central data collection, dissemination as well as decentralized communication that are possible in SHM applications (Nagayama et al. 2010). The standard AODV protocol uses periodic probe messages to update routing information

frequently between mobile nodes, and it consumes significant power. The GPMH omits the periodic probe messages, because the sensors mobility is not an issue in SHM system. To reduce the delay caused by the routing protocol, the GPMH does not regenerate route request (RREQ) message when route discovery is unsuccessful; instead, the task is handled by the reliable data transfer service in ISHMP Services Toolsuite. The GPMH implementation of AODV employs an alternative to the standard hop-count routing metric used for evaluating different paths. The hop-count routing metric may lead to construction of long links, which may result in significant radio loss and power consumption. The new metric uses the link quality indicator (LQI), calculated using the received energy level and signal to noise ratio (SNR). Additional details may be found in Nagayama et al. (2010). 2.4 Energy Harvesting for Sustainable Operation

Availability of a sustainable power supply is one of the biggest consideration when a large-scale WSSN is deployed. Though the battery life of a WSS can be extended by efficient power management (see Section 2.2), data condensation, and in-network processing, the use of ordinary batteries still requires regular battery replacement. The Imote2 has a DA9030 power management integrated circuit (PMIC) from Dialog Semiconductor that facilitates sustainable energy harvesting. The PMIC interfaces directly with Li-ion battery pack and handles unregulated power from energy harvesters up to 10 V. The PMIC charger manipulates voltage and current from energy sources for faster and more stable charging at Li-ion battery.

Among various available energy sources, solar energy is now the only sustainable power source sufficient to reliably operate Imote2s (Miller et al. 2010). Validation of solar energy harvesting has been successfully carried out in the Jindo Bridge deployment using solar panels (SCM-3.3W from SolarCenter (9V-370mA)) and lithium-polymer rechargeable battery (the Ainsys 3.7V-10,000mAh) (Jang et al. 2010). In addition, a prototype wind turbine (HR-W35V, Hankukrelay) can be used to power Imote2 in the windy area like Jindo Bridge (Park et al. 2010).

Fig. 6. Rechargeable battery (middle) and energy harvesters;

solar panel (left) and wind turbine (right).

The ChargerControl of ISHMP Services Toolsuite allows the energy harvesting system to work during the SnoozeAlarm mode (Miller et al. 2010). Enabling ChargerControl, the Imote2 determines whether it will continue in sleep mode or initiate charging mode by assessing

the battery voltage and charging. If the battery voltage is low and the power is sufficient, the Imote2 will start charge until the battery voltage achieves the target value of 4.2V. 3. RECENT ADVANCES FOR IMPROVED SHM USING WSS

Many recent hardware and software advances have been achieved toward improved SHM of civil structures, including new sensor boards, decentralized computing strategies, multi-hop communication protocols, streaming data transmission, and a more fault-tolerant ISHMP Services Toolsuite. These advances, detailed in this section, are expected to increase the robustness of WSSN, broaden the application fields, and support comprehensive SHM.

3.1 Recently Developed Sensor Boards This section describes sensor boards recently developed at Illinois for use with the Imote2 that expand the functionality of this WSS platform. Low-cost GPS sensors Global Positioning Systems (GPS) provide the possibility to measure a structural displacements. While survey-quality GPS technology is capable of measuring such displacements with sub-centimeter precision, the associated cost is too high to allow for routine deployment. Low-cost GPS chips commonly found in mobile phones and automobile navigation equipment are attractive in terms of size, cost, and power consumption; however, the displacement accuracy of these GPS chips is on the order of several meters, which is insufficient for SHM applications. Inspired by sensory information processing strategies of weakly electric fish, this section reports on the use of dense arrays of relatively low-precision sensors to achieve high-precision displacement estimates (Jo, et al. 2011b).

To assess the feasibility of using low-cost GPS sensors, four GlobalTop sensors (Gms-u1LP module, ~$20) are mounted together on a plank rotating horizontally in a 0.5m arc. The rotational displacement is measured at a 5Hz sampling rate when the rotor runs at 26 RPM. Despite the low fidelity, as shown in Fig. 7, the dynamic displacement measurements of ±0.5m range are encouraging. Further, averaging of the multiple GPS measurements further lowered noise levels and improved the frequency resolution. Although more static and dynamic tests have uncovered several issues that need to be resolved (e.g., significant low-frequency drift and directional difference in accuracy), the dense array of low-cost GPS sensors are promising.

Strain sensor board Strain provides another important measure of a structure’s health. A new strain sensor board for Imote2, designated SHM-S board (see Fig. 8), has been developed which includes a precisely controllable Wheatstone bridge circuit. In the SHM-S board, the Wheatstone bridge is carefully balanced prior to amplification so that the ADC is not saturated. To this end, digital potentiometers are used and autonomously controlled by software (RemoteCommand SHMSAutoBalance in ISHMP Services Toolsuite). The SHM-S board supports up to 2500 times signal amplification, 0.3 µε resolution below 20Hz, and temperature compensation. In addition, any analog signal from other types of sensors can be amplified with this sensor board, as long as the signal is in the range of the ADC. Fig. 9 shows typical strain data measured using SHM-S board on a pylon of the 2nd Jindo Bridge in Korea. Ambient strains at the 1~2 µε level are successfully captured. Note that the peak at 2.4Hz in power spectral density (PSD) corresponds to the 1st bending mode of the pylon.

F

F

Pressure

has been

5812 hig

ADC; it c

ig. 7. GPS m

Fig. 8. SHM

F

e sensor bo

n developed

h-precision

can be used

Fig. 10. S

0 5

-2

0

2

strain(u-strain)

measuremen

M-S board: to

Fig. 9. Strain

t

oard Addi

. The SHM

n analog pre

d with SHM

HM-P boar

-2

-1

0

1

2

dis

pla

cem

en

t (m

)

0

-40

-20

0

20

dB

10 15

time(s

nts: time hi

op (left), wi

n measurem

time history

itionally, a s

M-P board

ssure senso

M-DAQ boar

rd: top (left)

16:28:24 16

0.2 0.4 0.6

20 25

ec)

stories (top)

ith SHM-DA

ment on the

y (left) and

sensor boar

provides w

or. This bo

rd as shown

), bottom (m

:28:42 16:28:59

time (hh:mm:ss)

Time history

0.8 1 1

frequency (Hz)

Power Spectral Density

300

10-5

100

strain-P

SD

) and power

AQ (middle

steel pylon

PSD (right)

rd to measur

wind pressur

oard is a sim

n in Fig. 10.

middle), and

16:29:16

1.2 1.4 1.6

y

5 10 15

r spectral de

e), and with

of Jindo br

).

re wind pre

re measurem

mplified sen

d with SHM

16:29:34

1.8 2

GPS-1

GPS-2

GPS-3

GPS-4

Average

GPS-1

GPS-2

GPS-3

GPS-4

Average

20 25 30 35

freq(hz)

ensities (bot

h ISM400 (r

ridge;

essure (SHM

ments using

nsor board w

M-DAQ (righ

5 40 45 50

ttom).

right).

M-P board)

g AMSYS

without an

ht).

Digital-to-analog converter The SHM-D2A board (see Fig. 11) has a digital-to-analog converter (DAC) to support output of signals. The 4 channels of digital inputs delivered from the I2C interface of Imote2 are converted to external analog outputs ranging 0~2.5V using the TI-DAC8565. This DAC functionality provides an effective tool for wireless control.

Fig. 11. SHM-D2A board: perspective (left), top (middle), and bottom (right).

3.2 Decentralized In-network Processing Centralized data acquisition and processing schemes (See Fig. 12a) that are commonly used

in traditional wired sensor systems are not tractable in WSSNs due to the limitation in wireless data communication speeds; bringing all data to a centralized location will result in severe data congestion in the WSSN (Sim et al. 2009). Applications of two decentralized processing approaches will be discussed in this section; one uses independent data processing to estimate the tension in stay cables (see Figs. 12b) and the other uses coordinated computing to aggregate data (see 12c).

Stay-cables are the members resisting large portion of the loads on a cable-stayed structure (e.g., cable-stayed bridges). Tension forces of cables are direct indicators of a cable’s integrity as well as overall structural health of the cable-stayed structure. Several methods for estimating the cable tension are available including the direct measurement using the load cell, non-contact technique using the electromagnetic (EM) stress sensor (Wang et al. 2005), and vibration-based methods. Due to the convenience and cost effectiveness of sensor and sensor installation, the vibration-based methods have been recognized to be efficient in practice (Kim and Park 2007). Cho et al. (2010a) implemented a vibration-based method proposed by Zui et al. (1996) on an academic WSS platform and experimentally verified in laboratory testing using a string with both ends fixed. Because the vibration method derived by Zui et al. (1996) utilizes three lower natural frequencies of a cable, it may be impractical if vibration of the resisting structure is exciting the cable, called cable-deck interaction. The cable-deck interaction is dominant in the lower natural frequency region, which may distort the vibration signal of a cable, disabling reliable automated peak-picking. As a more practical alternative, a closed form relationship between natural frequencies and the tension force (Shimada 1994) is selected for automated cable tension estimation.

CableTensionEstimation (Sim et al. 2011) is a service that autonomously interrogates cable tensions based on the vibration-based method developed by Shimada (1994). Because this method only requires the natural frequencies of the cable, information sharing with sensor nodes on different cables is not necessary. The IndependentProcessingPSD service (Sim and Spencer 2009) which performs unsynchronized sensing and estimate the power spectrum, and the automated peak-picking algorithm (Cho et al. 2010a) are thus combined to estimate cable tension autonomously.

As opposed to the independent processing approach employed in

CableTensionEstimation, coordinated processing allows the sensor nodes to communicate

with each other and share information. The sensor network is divided into local sensor communities where data communication and processing are taking place to extract meaningful information from raw sensor data. Coordinated processing also allows estimation of spatial information that can be used to produce a global picture of a structural system (Sim et al. 2010a).

The DecentralizedDataAggregation service (Sim and Spencer 2009) is a decentralized application for data acquisition and processing implementing the natural excitation technique (NExT) (James et al. 1993) and the random decrement (RD) technique (Cole 1968) (Fig. 12c).

DecentralizedDataAggregation outputs either correlation or RD function, depending on

the user-specified input. The network is divided into local communities in which correlation or RD functions are calculated at each node and gathered by the cluster-heads. The correlation or RD functions can be either collected at the base station or retained at the cluster-heads for further analysis such as in-network system identification and damage detection.

DecentralizedDataAggregation can be used as a foundation for other application

development that requires the decentralized network such as decentralized modal analysis (Sim et al. 2010a; Sim et al. 2010b), decentralized system identification (Sim and Spencer 2009), and decentralized damage detection (Jang et al. 2010b).

3.3 Multi-hop Communication

The choice of routing metric plays an important role in multi-hop communication. An efficient routing metric should address energy limitations, link quality variation, and diverse radio environments. Thus, improvements in the reliable multi-hop communication approach described above are developed by taking into account the topology and link failure to optimize the throughput, while minimizing the energy consumption. This work identified the principal factors affecting the performance of multi-hop routing, and developed a variant of the AODV protocol to provide multi-hop routing and data transfer.

Link estimation is a critical part of every sensor network routing protocol as the packet reception rate of candidate neighbors allows a protocol to choose the most energy efficient next

Fig. 12. Data acquisition and processing schemes: (a) centralized data collection, (b) independent processing, and (c) coordinated computing strategy (Sim et al. 2009).

routing h

routing m

carries a

metric is

Quality I

symbol p

of link o

the Chip

highly co

sensitivit

topology

applicatio

with the

sufficient

of RSSI

solution d

been dev

Receptio

to throug

data tran

while add

To ac

on both

built betw

3.4 Strea

Real-

acquisitio

Also, thi

structura

Imote2 is

hop. The p

metric (e.g.

a routing m

s developed

Indicator (L

periods and

ver which t

pcon CC242

orrelated w

ty. LQI is

y is sparse,

ons). As a

packet rece

t to confide

and the L

due to its si

veloped wit

on Rate (PR

ghput (see F

nsfer. The

ding close t

ccount for l

directions;

ween a sour

Fig. 1

aming Data

-time wirele

on systems

is approach

l control).

s developed

proposed me

, AODV) u

metric, whic

d based on

LQI). RSS

stored in th

the packet w

20 radio, a

with packet

expected to

, and poor

an aggregat

eption rate;

ently estima

LQI value f

implicity an

th two regr

RR). PRR

Fig. 13), and

proposed m

to zero over

ink asymme

on the rou

rce and a de

3. Through

ess data acq

to be mim

h is critical

Thus, high

d (Linderma

ethod of rou

used for ev

ch is calcul

the Receiv

SI is the es

he RSSI val

was transm

and measure

reception

o have a hig

r-quality lin

te metric, L

however, d

ate link qua

for the rece

nd low comp

ressor varia

is chosen a

d the strong

metric provi

rhead to the

etry, link qu

ute to, and f

estination.

put as a fun

quisition ex

icked and o

when real-

h-throughpu

an et al. 201

uting emplo

valuating di

lated on ea

ved Signal

stimate of t

lue register

mitted. LQI

es the rece

rate except

gher correla

nks are pr

LQI average

due to high

ality. The n

eived packe

putation cos

ables RSSI

as the depen

g requiremen

ides a suffic

protocol.

uality estim

from the de

nction of lin

xpands the

offering rea

-time state

ut, real-time

1).

oys an altern

ifferent path

ach interme

Strength In

the signal p

. The LQI

I is calculat

eived energ

t when ope

ation with li

revalent (as

d over man

h variance, i

new routing

et. First-or

st. A mult

and LQI

ndent variab

nt of SHM

ciently relia

mation and m

estination.

nk quality (p

application

al-time visu

knowledge

e, wireless

native to the

hs. Each r

diate node.

ndication (R

power and

I value char

ted for each

y level and

erating at t

ink reliabili

s may be

ny packets i

individual L

g metric is

rder regres

tiple linear r

and depend

ble due to i

application

able assessm

metric calcu

Also, non

packet recep

ns of WSS,

alization of

e is require

data acquis

e standard h

route reque

. A uniqu

RSSI) and

is calculate

racterizes th

h received p

d/or SNR.

the edge of

ity when the

the case f

is strongly c

LQI reading

a linear com

sion is an

regression m

dent variab

its direct rel

s on high th

ment of pat

ulation are p

n-identical r

ption rate).

allowing w

f structural

d (i.e., feed

sition servic

hop-count

est packet

ue routing

the Link

ed over 8

he quality

packet by

RSSI is

f receiver

e network

for SHM

correlated

gs are not

mbination

attractive

model has

le Packet

lationship

hroughput

th quality,

performed

routes are

wired data

response.

dback for

ce for the

Despite of the onboard computation and communication capabilities of WSS, real-time data acquisition using WSS is challenging due to operating system limitations, tight timing requirements, sharing of transmission bandwidth, and unreliable radio communication. In the case of Imote2, the event driven concurrency model of TinyOS, the embedded operating system, and hardware limitations make the real-time scheduling and control required for real-time data acquisition challenging. In TinyOS, tasks are completed in a first-in-first-out (FIFO) manner along with interrupts to facilitate interaction with hardware (Levis et al. 2005). The FIFO approach requires any processing of a data sample, including temperature correction, time stamping, and sending, to be completed before the next data sample is acquired; as a result, the maximum sampling rate is limited by the time required for each of these steps. Furthermore, the clear channel assessment (CCA) used with the Chipcon CC2420 radio can increase the communication time when multiple nodes transmit at the same time. Thus, the sending time, which is necessary for determining the sampling rate, can be hard to predict. Finally, a random offset in the time of the start of sensing due to the sensing approach and hardware variability requires the timestamps to be returned with the data sample, which increases the overhead for each sample (Nagayama et al. 2009).

A tightly coordinated scheduling approach is used to overcome the event driven nature of TinyOS, communication latency, along with the existing sensing framework to achieve real-time data acquisition. This approach requires limiting the time of each processing and sending step to achieve high data throughput. First, the data payload is limited to one packet to reduce the communication time. The maximum packet payload is 112 bytes, which corresponds to nine samples of four channels of 16-bit data and a 32-bit timestamp (TinyOS 2006). Thus, nine samples can be buffered prior to sending. Second, a staggered time-division- multiple-access (TDMA) scheme, which only allows each node to communicate with the gateway node during a specified timeslot, is used to limit the variability in communication time due to contention and back-off delays and reduce packet loss due to collisions. The number of nodes and sending/processing times are accounted for in the TDMA approach making this approach different from other MAC-layer protocols, which cannot account for these variables (van Hoesel and Havinga 2004; Gobriel et al. 2009).

The complete application combines time synchronization, reliably sent commands to initialize sensing and compute the TDMA send time, and the scheduled communication protocol to achieve high throughput near-real-time data acquisition that is viable for an extended sampling interval. Because a scheduled approach is used, there is a tradeoff between network size and maximum sampling rate. The communication and processing protocols allow for near-real-time sensing of 108 channels across 27 nodes at up to 25 Hz with minimal data loss as shown in Table 1. As the network size increases, the corresponding maximum sampling rate decreases and the data throughput remains relatively unchanged due to the increased number of sensor nodes.

Table 1. Real-Time Wireless Data Acquisition Performance.

Number of Nodes Sampling Rate (Hz) Max. Data Throughput (Kbps)

1 – 9 75 43.2 10 – 18 40 46 19 – 27 25 43.2

3.5 ISHMP Services Toolsuite From extensive laboratory and full-scale deployments, the importance of fault tolerant is

found to be critical in realizing a robust and stable WSSN. To this end, a number of improvements and new features have been implemented and added to the ISHMP Services Toolsuite, as described in this section.

Monitoring of sensor power The utility commands Vbat and ChargeStatus of ISHMP Services Toolsuite provide information about power system for the Imote2, such as battery voltage, and charging voltage and current. AutoUtilsCommand enables the autonomous tracking of battery voltage and charging status as scheduled within AutoMonitor.

Data storage in non-volatile memory Data is now stored into non-volatile flash memory on Imote2 prevent sensor nodes from staying awake after data collection waiting for subsequent transmission to the gateway node. In large-scale networks, the waiting time for each sensor node can be very long, causing significant energy consumption. Now the sensors turn into sleep mode after data collection, and data can then be retrieved by waking up the leaf node. This approach also improves the reliability the network.

Skipping of unresponsive nodes Measurement is carried out using sensor nodes awake. However, some nodes can be unresponsive due to low-power battery, bad radio communication or hardware malfunction before or after sensing, even if they notify that they were awake. In that case, the unresponsive nodes are excluded in the network to avoid the unnecessary waiting time and excessive use of power.

Exclusion of low-power sensor nodes Sensing is one of the main sources of power consumption. Even if a sensor node is awake, it can be powered off during sensing or after sensing if the battery power is not sufficient. Once turned off, it cannot be recharged. Now low-power nodes are excluded in the measurement and participate again in sensing after the battery is recharged.

Autonomous resuming of AutoMonitor During the long-term operation of the network, AutoMonitor, which is running on the gateway node, may experience unexpected resets. The capability of resuming AutoMonitor after unexpected resets is now added by creating a checkpoint which saves into flash memory the parameters defining the current state of AutoMonitor after the completion of each scheduled task. When the gateway reboots from an unexpected reset, the parameters in flash memory are reloaded and the AutoMonitor application restarts from the last saved checkpoint.

ThresholdSentry for multi-hop communication In multi-hop, when gateway node wants to wake up the sentry node, it broadcasts the wakeup message and any node that hears the message will stay awake for route establishment. Therefore, waking up the sentry nodes by gateway causes more energy consumption in the case of multi-hop. Now the sentry tasks are scheduled from sentry nodes which are autonomously checking threshold periodically based on the parameters received from gateway and stored in flash memory at the beginning, without the need to be woken up by gateway.

Email n

Toolsuite

are autom

This feat

4. FULL

The m

been vali

test-bed

describes

4.1 US-K

The J

between

National

the first

infrastruc

The

southwes

Bridge c

steel-box

by the 60

Initial de

The 70 se

side and

nodes, 3

nodes, 3

Each bas

XP profe

Each gat

board, an

antenna a

otification

e is email n

matically em

ture has bee

L-SCALE IM

majority of

idated throu

in Korea

s these depl

Korea-Japan

Jindo Bridg

the US (Un

University

long-term

cture.

Jindo Brid

stern tip of

constructed

x girder with

0 stay cables

eployment

ensor nodes

the other o

pylon node

pylon node

se station c

essional OS,

teway node

nd antenna.

and environ

of structu

notification.

mailed to a

n a big help

MPLEMEN

the hardwar

ugh full-sca

and on the

oyments.

n Testbed Pr

ge test-bed

niversity of

y), and Japa

and the w

dges are tw

Korean Pen

in 2006 (l

h 344m of

s connectin

efforts T

s in the netw

on the Haen

es, 8 cable n

es, 7 cable

consisted of

, an uninterr

interfaces

. Each lea

nmentally h

ural respon

. Any ano

predefined

p in debugg

NTATION

re and softw

ale deploym

e Governm

roject for W

is construc

f Illinois at U

an (Univers

world’s large

Fig. 14 T

win cable-st

ninsula near

eft side of

main span

g the two A

The 1st deplo

work were d

nam side.

nodes). Th

nodes). O

f an industr

ruptable ba

with the re

af node is c

hardened en

nse and ne

omalies in t

group of in

ing the WS

NS

ware develo

ment on a c

ment Bridge

WSSN

cted as an

Urbana-Cha

ity of Toky

est deploym

The Jindo T

tayed bridg

r the town o

Fig. 14).

and 70m of

A-shape stee

oyment on t

divided into

The Jindo s

he Haenam

One base sta

rial-grade P

ckup power

espective su

composed o

nclosure. 3

etwork ano

the response

ndividuals r

SN.

opments and

cable-stayed

e at Rock

internation

ampaign), S

yo). The c

ment of wi

Twin Bridge

ges, which

of Haenam.

The bridg

f two side s

el pylons on

the bridge t

o two sub-ne

sub-network

sub-networ

ations was

PC (AAEON

r supply (AP

ubnets and c

of an Imote

3 D-cell bat

omalies A

e of the stru

responsible

d advances m

d bridge (th

Island Ars

nal collabor

South Kore

collaborativ

reless sens

es

connect Ji

. The test-

ge has a thr

spans. The

n concrete p

test-bed was

etworks: on

k consists o

rk consists o

to collect d

N AEC-690

PC ES550),

consists of

e2, battery b

tteries were

A new featu

ucture or th

for the dep

mentioned a

he 2nd Jindo

senal. Thi

ative resear

a (KAIST a

ve project re

ors to mon

indo Island

-bed is the 2

ree-span str

e girder is s

iers.

s carried ou

ne on the Jin

of 33 nodes

of 37 nodes

data for eac

05) running

, and a gate

an Imote2,

board, sens

used to po

ure to the

he WSSN

ployment.

above has

o Bridge)

is section

rch effort

and Seoul

esulted in

nitor civil

d and the

2nd Jindo

reamlined

supported

ut in 2009.

ndo island

(22 deck

s (26 deck

ch subnet.

g Window

way node.

interface

sor board,

ower most

of the leaf nodes, 8 nodes of 70 were self-powered using solar panels and rechargeable batteries. SHM-A boards were used to measure acceleration, temperature, humidity, and light for most of nodes; the SHM-W sensor board (an early version of the SHM-DAQ) was used to measure the signal from the 3D ultra-sonic anemometer. The deployment details, evaluation, and data analysis can be found out in Jang et al. (2010) and Cho et al. (2010b).

2010-2011 deployments Advanced hardware and software were implemented on the test-bed bridge in the 2010 deployment. Energy harvesting strategies were employed for all sensor nodes based on satisfactory performance during 2009; additionally, a mini wind turbine was installed on one node to assess the potential for wind energy harvesting. The network size was also increased to a total of 669 channels on 113 sensor nodes; which is currently the world largest WSSN for SHM. For better understanding of wind environment, two more 3D ultra-sonic anemometers were installed with the newly developed SHM-DAQ boards. Ten SHM-H boards were implemented as cluster heads of DDA. The enclosure for the sensor nodes was upgraded, including better connections to the rechargeable battery and solar panels and a weather-resistant antenna (see Figs. 15-16).

Fig. 15. Enclosure assembly (left), sensor module mounting (middle) and installation using magnet (right).

Fig. 16. Solar panel on cable node (left), mini wind turbine (middle), and 3D ultra sonic anemometer (right).

To efficiently operate such a large sensor network, the WSSN is divided into four

sub-networks, considering the functionalities, network size, communication range, and communication protocol of each network. All four sub-networks share a common software configuration that includes autonomous operation by AutoMonitor, power harvesting by ChargerControl, sleep-cycling by SnoozeAlarm, monitoring of the network status by AutoUtilsCommand, and centralized data acquisition by RemoteSensing. In addition to these common features, the subnetworks have their own unique software applications: deck

4dBi antenna

magnet

bolted

acrylic jig

networks

bandwidt

by the C

on the Ji

shown in

F

To ef

(see Fig.

simplify

Moreove

0

-20

0

20

acc(m

g)

s employ D

th by conde

ableTensi

indo side fe

n Fig. 17.

Fig. 17. Sen

ffectively m

18). In the

scheduling

er, the fault t

200 400 60

time(s)

Decentral

ensing senso

ionEstima

feature mult

nsor topolog

manage all

2nd year de

. This car

tolerance of

Fig. 18. C

Fig. 19. E

at

00 800 1000

lizedData

or data, wh

ation (CT

ti-hop comm

gy with node

these featur

eployment,

refully desig

f the variou

Comprehen

Example ver

node 20 (le

0 20

-20

0

20

acc(m

g)

aAggregat

hile cable ne

TE) applicat

munication

e IDs (2010

res, well-or

AutoMoni

gned schedu

us software a

sive SHM f

rtical accele

eft), 89(midd

00 400 600

time(s)

ion (DDA

etworks can

tion. Furthe

. The senso

0 deploymen

rganized ma

itor was re

uler allows

applications

framework u

eration of Ji

dle), 25(righ

800 1000

A) to effici

n obtain ten

ermore, the

or topology

nt on the 2n

anagement

ewritten to i

for effectiv

s was impro

using WSSN

ndo-side de

ht).

0 200

-20

0

20

acc(m

g)

ently utiliz

sion forces

nodes on t

y of each n

nd Jindo Bri

software is

improve sta

ve network o

oved substan

N.

eck

0 400 600

time(s)

ze limited

of cables

the cables

etwork is

idge).

s required

ability and

operation.

ntially.

800 1000

In the 3rd year deployment in 2011, additional features were added in the WSSN to improve performance. To reduce energy consumption, ThresholdSentry was rewritten to operate autonomously, which was particularly important for operation of networks requiring multi-hop communication. Additionally, a long-term sleep mode was implemented to improve the long-term performance of the network (see Section 3.5). And email notification function enables the gateway nodes to notify the network manager when abnormalities are detected in the WSSN. An improved multi-hop protocol was deployed which includes a better optimized routing algorithm. Finally, the newly developed strain sensor boards (SHM-S) were installed and validated on the bridge (see Figs. 8 and 9).

Analysis of measured data Each sensor node equipped with SHM-A or SHM-H sensor board provides 3-axes accelerations at a user-selectable sampling rate of 25Hz. Fig. 11 shows the sample acceleration time histories measured at the center span of Jindo-side deck (z-axis) during the typhoon Kompasu. The acceleration levels were about 20mg on average at that time and quite stationary. Fig. 20 shows the power spectral densities (PSD) for the z-axis acceleration of Haenam-side deck.

The acceleration responses were collected from the four subnets and used in the Natural Excitation Technique (NExT) in conjunction with Eigensystem Realization Algorithm (ERA) to identify natural modes as shown in Table 2 and Fig. 21. The modal properties from two deck networks are combined to provide the global information. To construct the global mode shapes, a least-square method is applied to link the modes together at the four overlapped sensor nodes at the center of the deck (see Fig. 17). The estimated modal properties are consistent with those from the wired sensor system (Cho et al. 2010b).

Fig. 20. Example PSD for vertical accelerations of Jindo-side deck.

Fig. 22 shows a typical charging and battery status monitoring for two days (Sep. 11~12th, 2010) using AutoUtilsCommand. The first day was sunny and the second day was rainy. In this deployment, the default checking cycle is 1 hour for the charging current and battery voltage of all sensor nodes. As shown in the Fig. 22 (bottom), the charging starts around 6~7AM, then the average charging current stays around 140mA during daytime and stops charging around 6 PM. The battery voltages (Fig. 14, top) during the day are not accurate, being

influenced by PMIC charging circuit; the actual battery voltages can be measured after charging is complete (at night). The actual voltage increase of the 10,000mAh Li-ion battery with one day charging was about only 0.02~0.04V around 4.1V level at that time; this modest increase is because the battery was almost fully charged at the time. As the charging level goes up, the charging speed becomes slower for safe charging. The second day was rainy, and the average charging current was around 70mA. The sentry nodes (150, 116, and 85) are equipped with two solar panels to compensate for increased power demands of ThresholdSentry; these nodes showed charging currents around 130mA. However, the charging currents of the sentry nodes during sunny day are still about 140mA, not double of the other sensor nodes, which is attributed to the fact that the batteries are almost fully charged and reduce charging speed.

Table 2. Identified natural frequencies and comparisons.

Mode NExT/ERA (Haenam)

NExT/ERA (Jindo)

Wired SHM (Cho et al. 2010b)

Deck Vertical 1 0.4462 0.4462 0.4395

DV 2 0.6454 0.6471 0.6592

DV 3 1.0331 1.0326 1.0498

DV 4 1.3559 1.3421 1.3672

DV 5 1.5549 1.5490 1.5869

DV 6 1.6528 1.6346 1.6602

Deck Torsion 1 1.7977 1.8022 -

DV 7 1.8710 1.8704 1.8555

DV 8 2.2594 2.2609 2.3193

DV 9 2.8121 2.8133 2.8076

Fig. 21. Identified mode shapes using NExT/ERA.

DV1 DV2 DV3

DV4 DT1 DV9

Fig. 22. Example of charging status monitoring using AutoUtilsCommand (Sep. 11~12th, 2010).

The Jindo Bridge experienced Typhoon Kompasu, having the 960 hPa of central pressure

and 40m/s of max wind speed after the 2010 deployment (Aug.31th ~ Sep.2nd, 2010). The typhoon passed the bridge quite closely, as shown in Fig. 23a. Based on the Korea Meteorological Administration (KMA) records, the wind speed was 14~20 m/s, and the wind was coming from the SSE (see the white arrow in Fig. 23b, in the Jindo area around 21:00 on Sep.1st, 2010. The wind data was measured on the same time using the 3D ultra-sonic anemometer on the bridge interfaced with the SHM-DAQ board as shown in Fig. 23c. It shows the 15~25 m/s of wind speed and 170~200 degree (meaning SE direction) of wind direction. Considering that the location of KMA (Mt. Chum-chal) and the Jindo Bridge (which is in a strait), the wind measurements are consistent.

Fig. 23. Typhoon Kompasu (a) path on Sep. 1st, 2010, (b) measured wind direction, and (c) measured data at 21:00 using 3D ultra-sonic anemometer with SHM-DAQ board.

0:00 2 4 6 8 10 12 14 16 18 20 22 0:00 2 4 6 8 10 12 14 16 18 20 22 0:004

4.05

4.1

4.15

4.2

4.25

Time (HH:MM)

Battery

Voltage (V

)

East-side nodes

0:00 2 4 6 8 10 12 14 16 18 20 22 0:00 2 4 6 8 10 12 14 16 18 20 22 0:00

0

100

200

Time (HH:MM)

Charg

ing c

urrent (m

A)

East-side nodes

151

96

116

144

64

70135

150

132

39

2

129

85

52

Aug. 29. 21:00Birth of Typhoon

Sep. 2, 3:00

Sep. 1, 21:00 (Closest to Jindo)

Sep. 2, 9:00

Sep. 2, 15:00

Sep. 2, 21:00

Sep. 3, 9:00Korea

Japan

China

Jindo Bridges

Jindo

Haenam

(Inland)

0 200 400 600 800 10000

10

20

30

time(s)

speed(m

/s)

wind speed

0 200 400 600 800 1000

140

160

180

200

220

time(s)

degree

wind direction (horizontal)

(b) (c)

4.2 Government Bridge at Rock Island Arsenal Currently, much of the national bridge stock of the United States has reached or is reaching

the end of its design life. SHM of the old and deteriorated structures is important in ensuring that they are still functional and safe for their intended uses. The continued monitoring and use of these structures represents a sustainable approach to meeting the transportation needs of today. An implementation of an integrated fiber optic and wireless SHM system on the 115 year-old steel truss Government Bridge at the Rock Island Arsenal illustrates this important use of SHM technology.

Introduction to the Government Bridge Located over the Mississippi River between Rock Island, IL and Davenport, IA the Government Bridge is one of over two hundred steel bridges owned and operated by the US Army and the Army Corps of Engineers. Authorized by an act of Congress in 1894 and built in 1896, the Government Bridge was designed by Ralph Modjeski as a replacement for a bridge built on the same piers that had become functionally obsolete in the years since its construction in 1872 (Modjeski 1897). The Government Bridge is an eight-span, double decker, steel truss bridge, as shown in Fig. 24a, where the upper deck carries rail traffic for the Iowa Interstate Railroad (IAIS) and the lower deck carries vehicular traffic. The second span of the bridge, shown in Fig. 24b, is a draw span that can swing 360° in either direction to allow boats to pass and has been in near continual operation since its construction. The Government Bridge is still a relevant and vital link in the national transportation network. In fact, the importance of the bridge has only increased in the last few years as bio-fuel plants built along the IAIS have meant that a large portion of the nation’s ethanol and bio-diesel cross the Mississippi over this bridge (Frailey 2011).

Development of integrated SHM system The Army Corps of Engineers has a strict maintenance and inspection schedule to keep the bridge functional and safe. To supplement the traditional visual inspections, the Corps installed a fiber optic SHM system with 34 FBG strain sensors on the draw span of the Government Bridge (Giles et al. 2011). To supplement the fiber optic strain system, a wireless SHM system, mainly armed with tri-axial MEMS accelerometers, was installed in July, 2011.

Initial observations of the strain data showed that the swinging bridge span had different strain levels in the measured members based on which direction the bridge closed in. Being a symmetric structure, the only distinguishing feature of the draw span is the stairs used to access the operator’s house. When these stairs upstream (original position), the strain levels differ from those when the stairs are in the downstream position (opposite position). Therefore, to confirm the orientation of the bridge, a wired digital compass is supplemented to the SHM system. These three systems, fiber optic, wireless, and wired, together make up the integrated SHM system for the draw span and generate rich data on the structure.

The integration of the measurements from all three systems into one system is expected to provide a more robust and complete view of the bridge condition. The strain data from the fiber optic system is processed to provide information on when train and swing events occur on the bridge and record the change in strain these events cause. The change in strain due to the bridge swinging is constant. The acceleration data from the wireless system is processed to determine the modal properties of the structure. Both the modal properties of the bridge and change in strain due to swing events can be fed into a model updating algorithm to determine

likely locations of changes to the health of the structure. The acceleration data can also be used in damage detection algorithms to determine damage indices for the members in the system (Giles et al. 2011).

(a) Full spans

(b) Draw span

Fig. 24. Government Bridge. Installation of wireless SHM system In July of 2011, a WSSN consisting of 22 leaf nodes was deployed on the Government Bridge at the Rock Island Arsenal. The leaf nodes are composed of Imote2, battery board, sensor board, antenna, and environmentally hardened enclosure, similar to those deployed of the Jindo Bridge. Based on a finite element model of the bridge, sensor nodes were distributed optimally in a pattern designed to capture the most modal data possible. The sensor locations are placed to allow recharging their batteries with the solar panels and to keep them out of reach of curious pedestrians along the lower chord. Analysis of measured data The three systems installed work together to monitor activities on the bridge. The compass gives the heading of the bridge and indicates when the bridge is turning. Fig. 26a shows the compass heading plotted with acceleration data taken from one of the sensor nodes. During the opening and closing of the bridge, the locking of the roller jacks on the end of the bridge cause a large impulse in the acceleration record. Approximately 30 seconds after the pulse indicating the bridge has closed, normal traffic is allowed to cross the bridge. Without the compass interpreting the data is challenging due to the lack of knowledge about the orientation and state of the bridge (i.e., closed and open). The SHM system to starts recording data (a lag of approximately two minutes in taking the date is due to network synchronization), and the compass indicates the bridge has left its initial heading and begun to turn counter clockwise; the wireless system recorded the vibration of the bridge caused by its rotation (higher frequency vibration caused by the mechanical system not a rotational acceleration). The bridge stops for a period of time half-way through its turn while the boats are passing; subsequently, the bridge begins to rotate and the vibration increases.

(a) FOS stain gauges (circle), digital compass (square), and WSS (star)

(b) WSS on top chord (c) WSS on railway deck

Fig. 25 Installed sensors on Government Bridge

(a) acceleration and heading when bridge swings

(b) acceleration, heading, and strain when a train passes with a pause

Fig. 26. Example measured acceleration (by WSS), heading (by digital compass), and strain (by FOS strain gauge).

Gateway

Operator’s

House

Digital

Compass Base Station

PC

20:18 20:20 20:22 20:24 20:26 20:28 20:30

-300

-200

-100

0

100

200

300

July 17, 2011

mg &

° H

eadin

g

Acceleration (24V)

Heading

17:28 17:30 17:32 17:34 17:36 17:38-1000

-800

-600

-400

-200

0

200

400

600

800

1000

July 17, 2011

mg, °

Headin

g &

µε

Acceleration (24V)

Heading

Strain (R2 Raw FBG)

The strain record also helps interpret what is occurring in the accelerations recorded by the wireless SHM system. Fig. 26b shows the vertical acceleration of one sensor node plotted with the compass heading and the strain from the member closest to the wireless node. The compass heading remains constant throughout the record, indicating the acceleration was not caused by a swing event. In this instance, the train entered the bridge at about 17:28, as indicated by the increase in the strain, and the wireless accelerometer began recording data

about three minutes later. Just before 17:34, the train stopped moving on the bridge as the strain levels are still elevated but are not experiencing any dynamic amplification. While the train was stopped, the acceleration levels are low. After the brief stop, the train again began to move and the acceleration levels again picked up. Without the strain record to supplement the acceleration measurements, determining if two trains had passed in short succession or one train with a pause in the middle of the passage would be quite difficult. Whereas normal traffic does not represent a significant mass in proportion to structure itself, fully loaded trains and their engines cause a strain change similar to when the bridge opens and closes and should be used to qualify any modal data taken during train crossings.

Frequency domain decomposition (FDD) (Brincker et al. 2001) is used to identify the modal properties of the bridge. With the help of digital compass and strain gauges, the acceleration data excited by automobile traffic while in the closed position is selected. Fig. 27 shows the result of FDD obtained using 60000 data points taken at a 100Hz sampling rate. Fig. 27a is the lower three singular values (SV); the first SV shows several dominant peaks. Figs. 27b-27f show the lower 5 modes of the bridge. The shape of the bridge is long and narrow; thus, the lateral mode shapes are dominant. The first vertical mode is 4.273Hz, which is about 2.5 times of the first natural frequency. The experimental mode shapes appear reasonable, considering the bridges shape and geometry. Using the modal information, many vibration-based system identification techniques, such as FE model updating, modal strain energy method, and damage locating vector method, can be employed in assessment of bridge integrity.

(a) singular values from FDD (b) 1st lateral mode: 1.636Hz

(c) 2nd lateral mode: 1.904Hz (d) 3rd lateral mode: 3.076Hz

(e) 1st vertical mode: 4.273Hz (f) 4th lateral mode: 5.591Hz

Fig. 27. Modal analysis result of Government Bridge.

0 5 10 15 20 2510

-8

10-6

10-4

Frequency (Hz)

Sin

gula

r V

alu

es

Result for Singluar Value Decomposition

1st sv

2nd sv

3rd sv

0

20

40

60

80

100

-50

5

0

10

( p g g p )

0

20

40

60

80

100

-50

5

0

10

0

20

40

60

80

100

-50

5

0

10

0 20 40 60 80 100

0

10

20

Elevation View

0

20

40

60

80

100

-50

5

0

10

5 CONCLUSIONS This paper discusses recent advances and field validation of a state-of-the-art framework

developed at the University of Illinois at Urbana-Champaign for wireless smart sensor networks (WSSN). Various hardware and software issues taken into account based on the lessons learned from recent efforts to develop, stabilize, and operate full-scale wireless smart sensor networks for SHM of civil infrastructure. High-fidelity, multi-scale responses can now be captured using variety of sensor boards recently developed. Decentralized computing is realized for decentralized data aggregation and autonomous cable tension estimation by fully utilizing computing capabilities and wireless communication. Routing in multi-hop communication is advanced to minimize the data loss and energy consumption. Data streaming is available for mimicking wired sensors and near-real-time wireless control. The ISHMP Services Toolsuite is updated with more services and fault-tolerant features. Using new (the 2nd Jindo Bridge) and historic bridges (the Government Bridge), full-scale WSSNs are realized for the purpose of SHM. The evaluation of the WSSN as well as the data analysis show high practicality of wireless SHM systems for civil infrastructure.

The success of the full-scale deployments discussed in this paper demonstrates the tremendous potential of WSSN for SHM. Indeed, wireless smart sensor technology has reached a level of maturity that makes is an important tool for management of civil infrastructure. ACKNOWLEDGEMENTS

This study is supported by National Science Foundation Grant CMS 06-00433 (Dr. S.C. Liu, program manager), the Korea Research Foundation (NRF-2008-220-D00117), Smart Infrastructure Technology Center (SISTeC) at KAIST, and the US Army Corps of Engineers (MEC W9132T-ILL-006). This support is gratefully acknowledged. The authors also would like to express their appreciation the following individuals for their efforts in performing the research reported herein: Gul Agha, Yozo Fujino, Shinae Jang, Hongki Jo, Ryan Giles, Hyung-Jo Jung, Jongwoong Park, Ho-Kyung Kim, Robin Eunju Kim, Jian Li, Kirill Mechitov, Parya Moinzadeh, Tomonori Nagayama, Jennifer Rice, Sung-Han Sim, and Chung-Bang Yun. REFERENCES

Brincker, R., Zhang, L., and Anderson, P. (2001). “Modal identification of output-only systems

using frequency domain decomposition”, Smart Materials and Structures 10(3): 441–445. Çelebi, M. (2006). “Real-Time Seismic Monitoring of the New Cape Girardeau Bridge and

Preliminary Analyses of Recorded Data: An Overview”, Earthquake Spectra 22:. 609-630. Cho, S.., Lynch, J.P., Lee, J-J. and Yun, C-B. (2010a). “Development of an automated wireless

tension force estimation system for cable-stayed bridge”, Journal of Intelligent Material

Systems and Structures 21(3): 361-376. Cho, S., Jang, S., Jo, H., Park, J., Jung, H-J., Yun, C-B., Spencer Jr., B.F. and Seo, J. (2010b).

“Structural Health Monitoring of a Cable-stayed Bridge Using Smart Sensor Technology: Data Analysis”, Smart Stuctures and Systems 6(5-6): 461-480.

Cole, H.A. (1968). “On-the-analysis of random vibrations”, Paper No. 68-288. American

Institute of Aeronautics and Astronautics.

Federal Highway Administration U.S. Department of Transportation and Federal Transit Administration (2007). Status of the National Highways, Bridges, and Transit: Conditions

and Performance, Report to congress, 2007. Frailey, F.W. (2011). “The Iowa Interstate Story”, Trains: 30-37. Giles, R.K., Kim, R., Spencer Jr., B.F., Bergman, L.A., Shield, C.K. and Sweeney, S.C. (2011).

“Structural Health Indices for Steel Truss Bridges”, Proceedings of the International Modal

Analysis Conference (IMAC XXIX), Jacksonville, FL Gobriel, S., Mosse, D., and Cleric, R. (2009). “TDMA-ASAP: Sensor Network TDMA

Scheduling with Adaptive Slot-Stealing and Parallelism”, Proceedings of the 29th

International Conference on Distributed Computing Systems (ICDCS’09). IEEE, Montreal, QC, 458-465.

IEEE Std 802-1990 (1990). IEEE Standards for Local and Metropolitan Networks: Overview

and Architecture, New York. James, G.H., Carne, T.G. and Lauffer, J.P. (1993). The natural excitation technique (NExT) for

modal parameter extraction from operating wind turbines. Sandia Report, SAND92-1666, Sandia National Laboratories, Albuquerque, NM.

Jang, S., Jo, H., Cho, S., Mechitov, K.A., Rice, J.A., Sim, S. H., Jung, H-J., Yun, C-B., Spencer Jr., B.F. and Agha, G. (2010a). “Structural Health Monitoring of a Cable-stayed Bridge Using Smart Sensor Technology: Deployment and Evaluation”, Smart Structures and Systems 6(5-6): 439-459

Jang, S., Sim, S.-H., Jo, H. and Spencer Jr., B.F. (2010b). “Decentralized bridge health monitoring using wireless smart sensors”, Proceedings of the SPIE Smart Structures/NDE

Conference, Vol.7647, 76473I. Jensen, S. (2003). Summary Outlook to 2005 for the European Construction Market. Available

at http://www.cifs.dk/scripts/artikel.asp?id=775&lng=2. Jo, H, Rice, J.A., Spencer Jr., B.F., and Nagayama, T. (2010a). “Development of a

High-sensitivity Accelerometer Board for Structural Health Monitoring”, Proceedings of the

SPIE Smart Structures/NDE Conference, Vol.7647, 76473I Jo, H., Sim, S.-H., Nagayama, T. and Spencer Jr., B.F. (2010b). “Decentralized Stochastic

Modal Identification using High-sensitivity Wireless Smart Sensors”, Proceedings of the 5th

World Conference on Structural Control and Monitoring. Jo, H., Sim, S-H., Mechitov, K. A., Kim, R., Li, J., Moinzadeh, P., Spencer Jr., B. F., Park, J.W.,

Cho, S., Jung, H-J., Yun, C-B., Rice, J. A. and Nagayama, T. (2011a). “Hybrid Wireless Smart Sensor Network for Full-scale Structural Health Monitoring of a Cable-stayed Bridge”, Proceedings of the SPIE Smart Structures/NDE Conference.

Jo, H., Sim, S.-H., Tatkowski, A., Spencer Jr., B.F., and Nelson, M.E. (2011b). “Feasibility of Displacement Monitoring using Low-cost GPS Receivers”, Proceedings of the World

Congress on Advances in Structural Engineering and Mechanics (ASEM’11+). Kim, B.H. and Park, T. (2007). “Estimation of cable tension force using the frequency-based

system identification method”, Journal of Sound and Vibration 304: 660-676. Levis, P, Madden, S., Polastre, J., Szewczyk, R., Woo, A., Gay, D., Hill, J., Welsh, M., Brewer,

E. and Culler, D. (2005). “TinyOS: An Operating System for Sensor Networks”, Ambient

Intelligence: 115-147. Springer, Berlin, Heidelberg.

Linderman, L.E., Mechitov, K.A., and Spencer Jr., B.F. (2011). Real-Time Wireless Data

Acquisition for Structural Health Monitoring and Control. NSEL Report No. 029, University of Illinois at Urbana-Champaign, Illinois. http://hdl.handle.net/2142/25420.

MEMSIC, Inc. (2010). ISM400, Imote2 Structural Health Monitoring Board, Andover, MA. Miller, T.I., Spencer Jr., B.F., Li, J. and Jo, H. (2010). Solar Energy Harvesting and Software

Enhancements for Autonomous Wireless Smart Sensor Networks. NSEL Report No. 022, University of Illinois at Urbana-Champaign. Available at http://hdl.handle.net/2142/16300.

Modjeski, R. (1897). “Erection of the Draw Span of the New Rock Island Bridge”, Journal of

the Western Society of Engineers 2(2): 135-168. Moore, M., Phares, B., Graybeal, B., Rolander, D. and Washer, G. (2001). Reliability of Visual

Inspection for Highway Bridges, Technical Report #FHWA-RD-01-020, Federal Highway Administration, Washington DC.

Nagayama, T., Sim, S-H., Miyamori, Y. and Spencer Jr., B. F. (2007). “Issues in structural health monitoring employing smart sensors”, Smart Structures and Systems 3(3): 299-320.

Nagayama, T., Spencer Jr., B.F., Mechitov, K.A, and Agha, G.A. (2009). “Middleware services for structural health monitoring using smart sensors”, Smart Structures and Systems 5(2), 119–137.

Nagayama, T., Moinzadeh, P., Mechitov, K.A., Ushita, M., Makihata, N., Leiri, M., Agha, G., Spencer Jr., B.F., Fujino, Y. and Seo, J. (2010). “Reliable Multi-hop Communication for Structural Health Monitoring”, Smart Stuctuctures and Systems 6(5-6): 481-504.

Park, J., Jung, H-J., Jo, H., Jang, S., and Spencer Jr., B.F. (2010). “Feasibility study of wind generator for smart wireless sensor node in cable-stayed bridge”, Proceedings of the SPIE

Smart Structures/NDE Conference 7647, 764747. doi:10.1117/12.853600. Pakzad, S.N., Fenves, G.L., Kim, S. and Culler, D.E. (2008). “Design and Implementation of

Scalable Wireless Sensor Network for Structural Monitoring”, Journal of Infrastructure

System 14(1): 89-101. Perkins, C., Belding-Royer, E., Das, S. (2003). Ad hoc On-Demand Distance Vector (AODV)

Routing. IETF. RFC 3561. Rice, J.A. and Spencer Jr., B.F. (2009). Flexible Smart Sensor Framework for Autonomous

Full-scale Structural Health Monitoring. NSEL Report Series, No. 18, University of Illinois at Urbana-Champaign. Available at http://hdl.handle.net/2142/13635.

Shimada T., (2004). “Estimating Method of Cable Tension from Natural Frequency of High Mode”, Proceeding of JSCE 501/1-29: 163-171 (in Japanese).

Sim, S.H. and Spencer Jr., B.F. (2009). Decentralized Strategies for Monitoring Structures

using Wireless Smart Sensor Networks. NSEL Report Series, 019, University of Illinois at Urbana-Champaign, Available at http:// www.ideals.illinois.edu/handle/2142/14280.

Sim, S.-H., Spencer Jr., B.F., Zhang, M., and Xie, H. (2010a). “Automated decentralized modal analysis using smart sensors”, J. Structural Control and Health Monitoring, (doi: 10.1002/stc.348).

Sim, S.-H., Carbonell-Marquez, J.F., Spencer, Jr. B.F., and Jo, H. (2010b). “Decentralized Random Decrement Technique for efficient data aggregation and system identification in wireless smart sensor networks”, Probabilistic Engineering Mechanics 26: 81-91.

Sim, S.-H., Li, J., Cho, S., Jo, H., Park, J., Spencer Jr., B.F. and Jo, H. (2011). “Automated Cable Tension Monitoring using Smart Sensors”, Proceedings of the World Congress on

Advances in Structural Engineering and Mechanics (ASEM’11+).

Singh, M.P. and Huhns, M.N. (2005). Service-Oriented Computing: Semantics, Processes,

Agents. John Wiley and Sons, New Jersey. TinyOS, http://www.tinyos.net, (2006). Tsai, W.T. (2005). “Service-Oriented System Engineering: A New Paradigm”, Proceedings of

the IEEE International Workshop on Service-Oriented Systems Engineering, 3-8. U.S. Census Bureau (2004). Value of Construction Put in Place. van Hoesel, L.F.W. and Havinga, P.J.M. (2004). “A TDMA-based MAC protocol for WSNs”,

Proceedings of the 2nd International Conference on Embedded Networked Sensor Systems

(SenSys '04). ACM, New York, NY, USA. Wang, M.L., Wang, G. and Zhao, Y. (2005). “Application of EM stress sensors in large steel

cables”, Sensing Issues in Civil Structural Health Monitoring, Chapter III, 145-154, (doi: 10.1007/1-4020-3661-2_15).

Wardhana, K. and Hadipriono, F.C. (2003). “Analysis of Recent Bridge Failures in the United States”, Journal of Performance of Constructed Facilities 17:3 144-150.

Zui, H., Shinke, T. and Namita, Y.H. (1996) “Practical formulas for estimation of cable tension by vibration method”, Journal of Structural Engineering, ASCE, 122(6): 651–656.