Monitori Ing Case ST Udy: Gove Ernment B Bridge

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Monitoriing case study: Goveernment B

Bridge
C 1, Ryan Giles 2, Kiriill Mechitov3 , and Billie F. Spencer, Jr.4
Soojin Cho
1
University
ty of Seoul, Seooul, South Koreea, soojin@uoss.ac.kr
2
Stony
S Brook Un niversity, Stonyy Brook, NY, ryyan.giles@stonnybrook.edu
3
Dept. of Computer Science,
S Univerrsity of Illinois at Urbana-Chhampaign., Urbbana, IL, mechhitov@illinois.eedu
4
Dept. of Civil and Environmental Engineering, University of IIllinois at Urbana-Champaiggn., Urbana, IL L,
bfs@
@illinois.edu

ABSTRA ACT: This caase study sum mmarizes thee deploymentt and operatiion of a wireeless smart ssensor
network for
f the structu ural health mo
onitoring of the
t swing spaan of the Govvernment Briddge at Rock IIsland
Arsenal, a historic bridge over the Mississippi River. The w wireless monnitoring system m was installled to
supplemen g fiber optic sensor netwo
nt an existing ork to enablee continuous multimetric bbridge monitoring.
More infoormation abou yment is available in reseaarch literaturee (Cho et al., 2015; Giles et al.,
ut this deploy
2012; Giles et al., 2011
1; Giles and Spencer,
S 2015
5).

ucture and Measured


Test Stru M Datta

The Rock k Island Arssenal Govern nment Bridgee, built


over the Mississippi River in 18 896 between n Rock
Island, IL
L and Davenp port, IA, is ju
ust one of ovver two
hundred bridges
b owneed by the United States Army.
The swiing span of o the Rock k Island Arsenal
A
Governmeent Bridge (F Figure 1) has the ability to
o rotate
360° in either
e directio
on and can lo ock each end of the Figgure 1: Swing SSpan of the Goovernment Briddge at 
span on either
e abutment. The Army nspects thee Rock Island A
y regularly in Arsenal. 
and maintains their brridges to ensu ure their fun ctionality. o supplementt the regular inspections oof the
To
swing spaan of the Gov vernment Briidge, a structu ural health mmonitoring syystem compossed of both a fiber
optic senssor network and
a a wirelesss smart senso or network hhas been instaalled on the bbridge. This mmulti-
frameworrk system meeasures strain n, acceleratioon, and orienntation and uuses these m metrics to peerform
structural health monitoring of th he bridge. Th he monitorinng is designeed to automaatically recorrd the
measured changes in strain
s caused by swing ev vents and reccord the accellerations meaasured duringg train
events.
The goal of the SHM M system is to help enssure public ssafety while also using llimited repaiir and
maintenan nce funds as efficiently as possible. In the case of thhe Governmeent Bridge, thhe unique abillity of
this historric structure to
t rotate neceessitated a multimetric
m appproach to SHHM. Each meetric providess vital
informatio on on the reall status of thee bridge, but only
o togetherr can the metrrics give a fulll understandiing of
the real tim
me status of the
t bridge and d the long term
m monitoringg of its properrties. 

In order too help maintaain and preserrve the historiic Governmennt Bridge at tthe Rock Islannd Arsenal, thhe US
Army Eng gineering Ressearch and Deevelopment Center
C (ERDC C) contracted to install a S
SHM system oon the
swing spaan. A multimeetric SHM sysstem has been n developed tthat leveragess the capabilitties of many ssensor
types to provide
p the deesired functio
onality. The syystem was deesigned to proovide real timme informatioon that
can be viiewed on a web-based
w interface and also
a track inddicators of hhealth such ass strain levells and
natural freequencies oveer time. The multimetric
m SHM system innstalled on thhe Governmennt Bridge hass three
principal components: a fiber op ptic strain monitoring
m syystem, a diggital compasss, and a wiireless
acceleration monitorin ng system. Figure
F 2 shoows the locaation on the sensors on the bridge. Each
componen nt part has itss strengths and limitations while providding an essenttial understannding to the syystem
as a wholee.

Collection of SHM Casee Studies by A


ASCE SEI Meethods of Monnitoring Comm
mittee
Figure 2: SSensor deploym
ment map, botth trusses are iinstrumented  Figure 3:: Installed wireeless sensor in
n an 
identicallyy.  enclosurre with solar panel. 
The fiber optic (FO) sttrain monitoring system prreviously insttalled on the Government Bridge consiists of
a Micron Optics sm13 30 interrogato ptical Bragg ggrating strainn gauges multtiplexed through a
or and 34 op
Micron Optics
O sm041 multiplexer.. The sensorss were installled symmetrrically on a vvariety of meember
types as sh
hown in Figu
ure 2.
The wirelless accelerattion monitoring system usses a base staation node annd twenty-tw wo Imote2 wiireless
sensor noodes with a triaxial acceeleration senssor board deesigned speciifically for ccivil infrastruucture
monitorinng projects. The
T selection of the sensorr locations w was a result off several facttors. Primarilly, the
goal of thhe wireless sensor
s deploy
yment was to o be able to capture as m many of the first modes oof the
structure as possible. A finite elem
ment model off the bridge ((Giles 2011) indicated thaat many of thee first
modes off the structuree had symmeetric and antissymmetric shhapes. In ordeer to differenntiate these m modes,
sensors were
w required on o all four qu
uadrants of thee structure.
The senssor nodes arre housed in n plastic en nvironmental
enclosures and are powered with h a rechargeaable lithium
battery an
nd a solar pan
nel, as shown in Figure 3. The
T wireless
sensor sooftware contaains a contin
nuous monito oring routine
that run ns autonomo ously once started. Within
W this
autonomo ous monitorin ng frameworrk, synchron nous data is
taken wh hen predeterrmined acceeleration thresholds are
exceeded..
The advan ntages of usiing a wirelesss sensor are that they do
not requirre extensive wiring suscep ptible to elecctromagnetic
influencess and each sensor has onboard co omputational
Figure 4: SStrain record sshowing two sswing 
processingg allowing forf the deveelopment of hierarchical events aand a train for three FO senssors. 
computing g algorithms.  
Figure 4 shows
s 30 minnutes of data for three straain sensors thhat has been nnormalized suuch that each starts
at zero microstrain.
m At
A 20:40, the bridge is clo osed with the stairs facingg upstream. A At about 20:444 the
bridge unlocks to swin ng causing a significant
s change in the sttrain of the thhree sensors. A
At about 20:551 the
bridge clo oses again affter making a 180° turn such that the stairs are noow in the dow wnstream possition.
Though sttrains in L31 and R4 drop and the strain n in R2 rises,, they do not return to the same level ass they
started. This
T indicatess that the bridge has three different looading condiitions: open, closed with stairs
upstream,, and closed with w stairs do ownstream. When
W the briddge opens up again at 20:555 the value of the
change in n strain from the
t stairs dowwnstream posittion to the oppen position iss the same in all three gaugges as
it was goiing from the open o n to the stairs downstream position. Usiing only straiin data in real time
position
it is difficcult to conclussively determ
mine the position of the briddge.

Figure 5 shows the trriaxial acceleration for a typical node on the Government Briddge. The dataa was
sampled ata 50 Hz for ten
t minutes while
w the briddge was closeed in the dow
wnstream possition. The prrimary
feature off this record is that ordin
nary traffic caauses detectab
able accelerattions. This is in contrast tto the
strain record (as show wn in Figure 4) where even in the cloosed positionn, the level of strain caused by
ordinary traffic
t barely exceeds the noise
n floor of the FO strainn system. 

Collection of SHM Casee Studies by A


ASCE SEI Meethods of Monnitoring Comm
mittee
-2
10
1st svv
2nd s v

Singular Values
-4
10

-6
10

-8
10
0 1 2 3 4 5 6 7 8 9 10
Frequency (Hz)

Figure  6: Results of FFDD for accelerration record iin 
the closed position.. 
-2
10
1st svv
2nd svv

Singular Values
-4
10

-6
10

-8
10
Figure 5: Triaxial acceleration reco
ord for one of tthe  0 1 2 3 4 5 6 7 8 9 10
Frequency (Hz)
eless sensors (ffrom top to bo
wire ottom: verticall, 
Figure  7: Results of FFDD for accelerration record iin 
lateral, and longitudin
nal). 
the oopen position. 
SHM Meethodology an
nd Results
The accelleration records collected by
b the wirelesss SHM systeem are processsed using syystem identificcation
techniquees such as freequency dom main decompo osition (FDD D) to determinne the naturaal frequenciees and
mode shap pes of the stru
ucture. Figuree 6 and Figurre 7 show the result of FDD D (i.e., first aand second sinngular
values of power spectral density maatrix) when th he bridge is inn a closed andd open positioon, respectivelly. As
expected, the natural frrequencies off the bridge arre not the sam
me in the two positions beccause of the chhange
in boundaary condition ns. There aree more naturral frequenciees less than 3 Hz in thee open positiion as
comparedd to the closed position. Using the reesults of thee FDD analyysis, mode shhapes can also be
computedd to determinee what the dettermined mod des look like.

The calcuulated modes and natural frequencies


f are
a intrinsic pproperties of tthe system foor the given bbridge
position. When
W compaaring them it is essential to know for w which positioon the given aacceleration rrecord
was loggeed. Therefore,, like the chan nge in strain due
d to the swiing events, iff statistically ssignificant chhanges
in the natu
ural frequencies and modees are detected d for a given pposition, it iss an indicationn that somethhing in
the structu
ure has chang ged. In such ann instance, daamage detectiion algorithm ms or model uppdating proceedures
could be used
u to locatee possible sou
urces of the chhange. 
Based on the observattions of multiimetric data, an algorithm m has been deeveloped thatt can, in real time,
determinee the starting and stopping g times of th
he event, diffferentiate a swwing event ffrom a train eevent,
record thee starting and ending orien
ntations of thee bridge and tthe direction of the swing,, record the chhange
in strain caused by th he swing eveent as a healtth monitoringg metric, dettermine the ddirection a trrain is
traveling, and record the
t maximum m strain differrence caused by the train. When acceleeration recordds are
produced,, the detected b used to dettermine the bbridge positioon during the record. This event
d events can be
detection algorithm efffectively sortts through thee streaming sttrain and commpass data annd records thee vital
informatioon.
Figure 8 compares the full strain record with the data thatt has been obbtained by thhe event deteection
algorithm
m. In total, thee algorithm detected
d five swing
s As can be seen, the
eventss and three traain events. A
detected event
e data doees not track th ween swing annd train eventts. Neither does the
he strain data exactly betw
detected event
e data shoow all the dynnamic variation that takes place duringg a train eventt. Neverthelesss, the
detected event
e data keeeps the impo ortant inform
mation – the cchange in strain caused by the openinng and
closing off the bridge. By just keeping the pertin nent informattion that can be used in m model updatinng, the
event deteection algoritthm also reduuces the amou unt of data thhat needs to bbe stored. Thee strain system
m can

Collection of SHM Casee Studies by A


ASCE SEI Meethods of Monnitoring Comm
mittee
output datta at a rate off up to 125 Hz.
H At that maaximum rate, 12 GB of daata are producced every dayy. The
event dettection algorithm reducess the output to just
14kB.
Figure 8 also illustraates the need d for a mulltimetric
approach. The first deetected train event just affter 9:00
looks diffferent than th
hose at 10:000 and 12:00. This is
because this
t event rep
presents a trrain that enteered the
bridge onn the side opposite
o the R2 sensor location,
l
stopped, and
a then reveersed off the track to returrn to the
rail yard
ds. The asy ymmetric loaad caused a strain
decrease, like a swing g event wou uld. In this case,
c the
compass data, which h remained constant, becomes
b
importantt in identifyiing the even nt as a trainn event;
otherwisee the event cou
uld be mistak
ken for a swin
ng event.
Fiigure 8: Compaarison of unprrocessed strain
n data 
A database of the changes in straain caused by y swing annd the results of using the eevent detectionn 
events an nd the naturall frequencies and mode sh hapes of allgorithm for a four hour time span where three 
the bridgee in its variou us positions is being created
d. In the trrains and five sswing events ttook place. 
future, th he databasee will be used to deetermine
statisticallly if it is likeely that any damage
d has occurred
o overr time. The ddata could then also be ussed in
damage detection
d algo
orithms and model
m updatiing in a moddel that accoounts for the differences iin the
bridge’s positions
p to deetermine wheere corrosion or
o other damaage has occurrred.
Lessons Learned
L
This case study shows a successful application of o a SHM systtem composeed of various types of senssors to
provide th
he desired funnctionality in the highly efffective mannner. The fiberr optic strain monitoring syystem
monitoredd the strain buut could not figure
f out thee cause of strrain change clearly. The adddition of wiireless
acceleration monitorin ng system an nd a digital compass on the existing strain moniitoring system m has
enriched the
t understan nding of the in
nput loading,, consequencee of the loadiing, and strucctural conditiion by
working as
a a whole. TheT advantagees of the wireeless sensor nnetwork weree presented byy easy addition on
the operattional bridge with minimall control of trraffic, evasionn of electrom
magnetic influeences on the w
wires,
and onbo oard computtational proccessing allow wing for thee developmeent of hierarrchical compputing
algorithm
ms. Furthermore, this case study shows another monnumental deployment of laarge-scale wiireless
sensor nettwork for the SHM of a full-scale bridgge.

References
[1] Cho S,
S Giles R, an nd Spencer B (2015) Systeem Identificattion of Historric Swing Truuss Bridge Ussing a
Wireless Sensor Network Employin ng Orientatio
on Correctionn. Structural C
Control and H Health Monittoring
22(2): 2555-272.
[2] Giles R, Kim R, Sweeney
S S, Spencer B, Bergman
B L, SShield C, andd Olson S (22012). Multimmetric
Monitorin ng of a Historric Swing Bridge. Proceeddings of the 200th ASCE Anaalysis & Commputation Speecialty
Conferencce, 151–162.
[3] Giles R et al. (2011). Structuraal Health Inddices for Steeel Truss Briddges. Civil Enngineering T
Topics,
Volume 4: Proceedingss of the 29th IMAC,
I A Connference on Sttructural Dynnamics, 391–3398.
[4] Giles R and Speencer B (2015). Develop pment of a Long-term, M Multimetric Structural HHealth
Monitorin ng System forr a Historic Steel Truss Sw
wing Bridge. N Newmark Strructural Enginneering Laborratory
(NSEL) Report Serries, No. 39, Universitty of Illinoois at Urbaana-Champaiign, Urbana, IL.
http://hdl.handle.net/21142/78088

Collection of SHM Casee Studies by A


ASCE SEI Meethods of Monnitoring Comm
mittee

You might also like