research article long-term structural health monitoring...
TRANSCRIPT
Research ArticleLong-Term Structural Health Monitoring System fora High-Speed Railway Bridge Structure
You-Liang Ding1 Gao-Xin Wang1 Peng Sun2 Lai-Yi Wu3 and Qing Yue3
1Key Laboratory of Concrete and Prestressed Concrete Structures of Ministry of Education Southeast UniversityNanjing 210096 China2Department of Civil and Environmental Engineering Rice University Houston TX 77005 USA3China Railway Major Bridge (Nanjing) Bridge and Tunnel Inspect amp Retrofit Co Ltd Nanjing 210032 China
Correspondence should be addressed to You-Liang Ding civilchinahotmailcom
Received 27 February 2015 Revised 30 July 2015 Accepted 24 August 2015
Academic Editor Ahmed El-Shafie
Copyright copy 2015 You-Liang Ding et alThis is an open access article distributed under the Creative CommonsAttribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
Nanjing Dashengguan Bridge which serves as the shared corridor crossing Yangtze River for both Beijing-Shanghai high-speedrailway and Shanghai-Wuhan-Chengdu railway is the first 6-track high-speed railway bridge with the longest span throughoutthe world In order to ensure safety and detect the performance deterioration during the long-time service of the bridge aStructural Health Monitoring (SHM) system has been implemented on this bridge by the application of modern techniques insensing testing computing and network communication The SHM system includes various sensors as well as correspondingdata acquisition and transmission equipment for automatic data collection Furthermore an evaluation system of structural safetyhas been developed for the real-time condition assessment of this bridge The mathematical correlation models describing theoverall structural behavior of the bridge can be obtained with the support of the health monitoring system which includes cross-correlation models for accelerations correlation models between temperature and static strains of steel truss arch and correlationmodels between temperature and longitudinal displacements of piers Some evaluation results using the mean value control chartbased onmathematical correlationmodels are presented in this paper to show the effectiveness of this SHM system in detecting thebridgersquos abnormal behaviors under the varying environmental conditions such as high-speed trains and environmental temperature
1 Introduction
The environmental actions (wind temperature rain earth-quake etc) and the bridge usage (traffic loading) can con-tinually modify the behavior and cause deterioration of long-span bridges in their long-term lifespan Over the past severaldecades a significant research effort has focused on the devel-opment of long-term Structural Health Monitoring (SHM)systems [1 2] Generally the objective of the SHM systemis to gather various reliable information that can be used todetect and follow the evolution of the bridgersquos condition statelocate and quantify damage and perform reliability-basedassessment so as to give the engineer or administrant a greatvariety of options with respect to maintenance interventionIn China governments as well as scientists and engineerspay much attention to the long-term health monitoring oflong-span highway bridges Typical SHM implementations
in highway bridges include Tsing Ma Bridge [3] Kap ShuiMun Bridge [3] Ting Kau Bridge [3] Stonecutters Bridge[3] Sutong Bridge [4] and Runyang Bridge [5] The modularconcept used in the design of the SHM systems for Tsing MaBridge Kap Shui Mun Bridge and Ting Kau Bridge has asignificant influence on the design of new SHM systems inHong Kong and China The six integrated modules are thesensory system the data acquisition and transmission systemthe data processing and control system the structural healthevaluation system the structural health data managementsystem and the inspection and maintenance system A SHMsystem for long-span bridges should be able to monitor theloading and structural parameters set by the bridge designerso that the bridge performance under current and futureloading conditions can be evaluated
With the recent development of design theories and con-struction techniques for high-speed railway in China many
Hindawi Publishing Corporatione Scientific World JournalVolume 2015 Article ID 250562 17 pageshttpdxdoiorg1011552015250562
2 The Scientific World Journal
(a) View of the Nanjing Dashengguan Bridge
192336
1stcross-section
1st cross-section 2nd cross-section
336
2ndcross-section
Beijing
1272
Shanghai
108 192 108
(b) Elevation drawing of the bridge (unit m)
Middle truss arch
Side truss arch
Orthotropic steel deck
Downstream15000
Side truss arch
Upstream15000
(c) Design drawing of the deck cross-section (unit mm)
Figure 1 Nanjing Dashengguan Bridge
high-speed railway bridges have been built during the lastfew years Maintaining the safety of these high-speed railwaybridges is crucial to the safety of passengers and trainsTherefore research efforts should be focused on the SHMsystems of high-speed railway bridges This paper reportsthe development and implementation of a SHM system ofNanjing Dashengguan Bridge Nanjing Dashengguan Bridgewhich serves as the shared corridor crossing Yangtze Riverfor both Beijing-Shanghai high-speed railway and Shanghai-Wuhan-Chengdu railway is the first 6-track high-speedrailway bridge with the longest span throughout the worldIt is the largest bridge with the heaviest design loading of allthe high-speed railway bridges by far because its main spanis 336m and its railway contains 6 tracks Also the designspeed of 300 kmh of the bridge is on the advanced level in theworld Due to these remarkable characteristics including longspan of themain girder heavy design loading and high speedof trains it is necessary to monitor the real-time health statusof the bridge during its service period With the state-of-the-art techniques of current monitoring systems of highwaybridges the health monitoring system for this bridge wasdesigned to be concise practical reliable and cost efficientThe main purpose of this system is to monitor the bridgersquosbehaviors under the varying environmental conditions and
detect the bridgersquos abnormal behaviors which may threatenthe structural safety This paper introduces the deploymentand functions of this system including a sensor system adata acquisition and transmission system a datamanagementsystem and a structural evaluation system Some evaluationresults of the bridge subjected to varying environmentalconditions such as high-speed trains and environmentaltemperature are also presented
2 Structural Health Monitoring System
Nanjing Dashengguan Bridge shown in Figure 1(a) is a steeltruss arched bridge with the span arrangement (108 + 192 +2 times 336 + 192 + 108)m which supports a six-line railwayincluding two regular rails two high-speed railways andtwo subway lines The elevation drawing of the bridge andthe deck cross-section are shown in Figures 1(b) and 1(c)respectively The three truss planes in the longitudinal direc-tion were used with a spacing of 15m which was selected toprovide the large stiffness for such heavy live loadsThe use ofa steel box as the deck system is another feature in the designof this steel truss arc which is adopted as the bottom chord ofthe entire truss segment to reduce uneven deflection on thedeck The bearing system of the bridge is that the 7 pier is
The Scientific World Journal 3
Table 1 Sensitivity analysis of bridge parameters for SHM design
Structure Key position Critical member Parameter ImportanceVariation undertemperature
actionSensitivity Decision
Pier Side spans 4 and 10 piers Longitudinaldisplacement Critical 229mm Sensitive Monitor
Main girder In the middle ofthe main span Steel deck Vertical
displacementLess-
important 35mm Less-sensitive No
Truss arch In the middle ofthe main span
Top chordmember diagonalweb member archrib chord memberand bottom chord
member
Static strain Critical 280 120583120576 Sensitive Monitor
Truss arch In the middle ofthe main span
Top chord memberand bottom chord
memberTemperature Critical 60∘C Sensitive Monitor
Main girder In the middle ofthe main span Steel deck Temperature Less-
important 60∘C Less-sensitive No
Whole bridge Frequency Important Nonsensitive Monitor one
Truss arch In the middle ofthe main span Top chord member Wind Less-
important Less-sensitive No
linked to the bridge deck so that the bridge deck is restrictedfrom moving in the longitudinal direction at the 7 pierAnd there have no longitudinal restraints between the bridgedeck and other piers Due to the remarkable characteristics ofNanjingDashengguanBridge including long span of themaingirder heavy design loading and high speed of trains a long-term SHM system was designed and installed on the NanjingDashengguan Bridge shortly after it was opened to railwaytraffic This online SHM system was designed to be a concisemonitoring system with a minimum of 28 sensors to monitorthe key parametersThe entire system consists of a sensor sys-tem a data acquisition system a data transmission system adatamanagement system and a structural evaluation systemThe primary purpose of the system is to monitor in-serviceperformance of the bridge structure under the varying envi-ronmental conditions such as high-speed trains and environ-mental temperature and to provide early warning of abnor-mal changes in in-service performance of the bridge thatmight be unsafe to bridge structureThe procedure of design-ing and implementing the SHM system includes three majorstages as follows (1) the stage of planning and design includesthe consideration of the physical quantities to be measuredthe selection of sensors appropriate for measuring such phys-ical parameters and the optimal placement of the sensors (2)The stage of installation and calibration includes a series ofstatic and dynamic experiments performed to check the oper-ation of the monitoring system after the completion of thebridge (3)The stage of main operation includes the test oper-ation of the monitoring system training of the maintenancecrew and the replacement ofmalfunctioning sensing devices
The design objective of the SHM system is to monitor thestructural health under daily service condition With regardto the bridge site of Nanjing Dashengguan Bridge there arevery lowprobability events of Typhoon and earthquakeThus
the train loading and temperature action are the most impor-tant factors that affect the daily performance of the bridgeDue to the design requirements the structural responses ofthe bridge under train loading are small and conservativeHowever the axial movements and thermal stresses of thebridge under temperature action are considerable due to thelong span of the bridge For instance the maximum thermalstrain and train-induced strain in the steel truss arch areabout 250 120583120576 and 50 120583120576 respectively For bridge conditionassessment it is of paramount importance to characterize nor-mal variability of structural parameters due to temperatureeffectsTherefore in order to control the overall budget of theSHM system without losing major information on structuraldaily behavior the structural parameters to be monitoredin the Nanjing Dashengguan Bridge were determined usingthe structural sensitivity analysis with the finite elementmodel under the designed temperature action as shown inTable 1 Other important parameters including accelerationsof steel deck and fatigue of steel deck were also monitoredConsidering that the displacement amplitudes of steel deckunder the train loading can be derived from the accelerationof steel deck the vertical displacements of steel deck were notchosen to monitor
21 Temperature and Static Strain Monitoring of Truss ArchAs shown in Figure 1(b) temperature and static strain mon-itoring of truss arch is performed at the 1st cross-section inthe middle of the first main span of the bridge Eight FBGtemperature and strain gauges are both installed on the 1stcross-section to simultaneously measure temperature fielddata and static strain field data in the steel truss arch as shownin Figure 2 It can be seen from Figures 2(c)sim2(f) that thesensors were installed at the top chordmember diagonal webmember arch rib chord member and bottom chord member
4 The Scientific World Journal
Downstream Upstream
Horizontal bracing of top chord member
Vertical bracing
Vertical web memberof side truss
Vertical web memberof middle truss
Top plate
Transverse stiffening girder
Horizontal bracing ofarch rib chord member
Vertical web memberof side truss
W1
W2
W3
W4
W5
W6
W7 W8
Y1
Y2
Y3Y4
Y5
Y6
Y7 Y8
15000 15000
68007
12000
2
2
(a) 1st cross-section of steel truss arch
Shanghai
Top chord member
Arch rib chord member
Vertical web member
Beijing
Diagonal web member
Bottom chord member
W1
W2
W3
W4
W5
W6
W7
W8
Y1Y2
Y3Y4
Y5
Y6
Y7
Y8
3
3
4
4
5
5
6
6
(b) 2-2 section of steel truss arch
Downstream Upstream
W1 W2Y1 Y2
1000
1000
850 850
150
150
1400
(c) 3-3 section of top chord member
Downstream Upstream
W3W4Y3Y4
150
150
610
610
1398
(d) 4-4 section of diagonal web member
Downstream Upstream
W5
W6Y5
Y6
150
150
1400
650 650
800
(e) 5-5 section of arch rib chord member
Downstream Upstream
W7 W8Y7 Y8
150
150
1266
1266
1400
1416
(f) 6-6 section of bottom chord member
Figure 2 Layout of temperature and static strain gauges on the steel truss arch (unit mm)
The Scientific World Journal 5
DYB-1 DYB-2
2500mm2500mm
Figure 3 Layout of dynamic strain gauges on the steel deck
Vertical accelerometer Z2Z4Transverse accelerometer Z1Z3
Figure 4 Layout of accelerometers on the steel deck
of the truss arch (where 119882119894denotes 119894th temperature sensor
119894 = 1 2 8 119884119894denotes 119894th strain sensor 119894 = 1 2 8)
The specification of FBG temperature sensors and straingauges is FBG4700 and FBG4100 supplied by China GeokonInstrumentsCo Ltd Sampling frequency of temperature andstatic strain data collection is set to 1Hz
22 Dynamic StrainMonitoring of Steel Deck Dynamic straingauges are installed at the 1st cross-section in the middleof the first main span of the bridge so as to monitor thefatigue performance of orthotropic steel deck as shown inFigure 3 It can be seen that the dynamic strain gauges DYB-1and DYB-2 monitor the long-term fatigue effects of deck-ribwelded detail induced by the traveling high-speed trainsThespecification of dynamic strain gauges is FBG4100 suppliedby China Geokon Instruments Co Ltd Sampling frequencyof dynamic strain data collections is set to 50Hz
23 Acceleration Monitoring of Steel Deck In order to moni-tor the transverse and vertical acceleration responses on thetwo main spans induced by the traveling high-speed trainsone transverse accelerometer and one vertical accelerometerwere installed at the 1st cross-section and 2nd cross-sectionin the middle of two main spans respectively as shown inFigure 4 (where 119885
119894denotes 119894th accelerometer 119894 = 1 2 4)
The specification of accelerometers is DH610 supplied byDonghua Testing Technology Co Ltd Sampling frequencyof acceleration data collections is set to 200Hz
24 Longitudinal Displacement Monitoring of Piers In thedesign of Nanjing Dashengguan Bridge the 7 pier is linkedto the bridge deck so that the bridge deck is restricted frommoving in the longitudinal direction at the 7 pier And therehave no longitudinal restraints between the bridge deck andother piersTherefore 6 displacement transducers are locatedon the piers and used for the measurement of longitudinaldisplacements at the 4 pier 5 pier 6 pier 8 pier 9pier and 10 pier as shown in Figure 5 (where 119871
119894denotes 119894th
displacement transducer 119894 = 1 2 6) The specificationof displacement transducers is CLMD2 supplied by ASMrsquosin Germany Finally it should be noted that the DAQs of theSHM system are located in the 6 and 8 piers
3 Evaluation of Structural Condition UsingLong-Term Monitoring Data
Using long-term monitoring data collected from the SHMsystem of Nanjing Dashengguan Bridge structural conditionassessment involves four aspects evaluation of vibration per-formance of themain girder evaluation of static performanceof steel truss arch evaluation of movement performance ofpier and evaluation of fatigue performance of steel deck
31 Evaluation of Vibration Performance of the Main GirderThe main function for the SHM system is to detect theperformance degradation of structures as early as possibleand to provide essential references for maintenance of actualbridges For Nanjing Dashengguan Bridge the vibration andstrain of the main girder together with the longitudinaldisplacement of piers are three important items to moni-tor Three kinds of performance evaluation are conductedaccordingly Those are evaluation of vibration performanceof the main girder evaluation of static performance of steeltruss arch and evaluation of movement performance of pierFirstly degradation or damage of local members may occurwith the increase of service period which will reduce thestructural stiffness of bridge Thus the dynamic propertyof bridge structure is affected significantly by the structuralstiffness reduction and will cause unusual variations ofthe structural vibration Such unusual behavior will affectrunning safety of the train directly Secondly damage of localmembers may induce redistribution of the internal forces inthe whole bridge under the temperature action The reasonis that Nanjing Dashengguan Bridge belongs to staticallyindeterminate structure Therefore static strain monitoringis necessary for typical members under temperature actions
6 The Scientific World Journal
Beijing Shanghai
4 pier 5 pier 6 pier 7 pier 8 pier 9 pier 10 pier
L1 L2 L3 L4 L5 L6
Figure 5 Layout of displacement transducers on the pier
Acce
lera
tion
(mm
s2)
1 80012001 4001 6001 10001
Data point
minus40
minus30
minus20
minus10
0
10
20
30
40
(a) Accelerometer 1198851
Acce
lera
tion
(mm
s2)
1 80012001 4001 6001 10001
Data point
minus40
minus30
minus20
minus10
0
10
20
30
40
(b) Accelerometer 1198853
Figure 6 Transverse acceleration time histories of the main girder induced by a high-speed train
Thirdly the longitudinal displacement behavior of the bridgepier is crucial for running safety of the train under thetemperature action Once the abnormal variation of the itemis detected inspection may be evolved timely for detection ofthe bridge piers In this section the evaluation of vibrationperformance of main girder is firstly discussed
For high-speed railway bridges vibration performanceof the main girder subjected to the high-speed trains isimportant for passenger comfort and traffic safety [6 7]For instance Zeng et al [8] point out that the essentialreason of train derailment is that the transverse vibrationof the train-railway-bridge system becomes unstable Henceit is necessary to present a strategy for early warning ofdeterioration of vibration performance of the main girderusing long-term health monitoring data
Figures 6 and 7 illustrate the typical transverse and ver-tical acceleration time histories of the main girder measuredfrom accelerometers Z
1simZ4 respectively It can be observed
that when high-speed train passed the bridge the transverseaccelerations in the middle of the first span are much smallerthan that of the second span and the vertical accelerationsin the middle of the first span are much larger than that ofthe second span These indicate that although the structurallayouts of two main spans of the main girder are the samethere exists significant difference between the transverseand vertical vibration characteristics of two main spans due
to the rail irregularity in transverse and vertical directionsThus there is a need to monitor the transverse and verticalaccelerations of two main spans in the long term so as torealize anomaly alarms for vibration performance of themaingirder
Hence in the present study the RMS values of theaccelerations are utilized as the monitoring parameters torepresent the transverse and vertical vibration characteristicof the main girder The cross-correlation between the RMSvalues of the accelerations in the middle of two main spans isfurther investigated when the high-speed trains pass throughthe bridge Figure 8 shows the scatter plots between the RMSvalues of the accelerations and shows the fitted results byusing quadratic polynomial as well
By using the fitting formula constructed in Figure 8scatter plot between theRMSvalues of the accelerations in themiddle of the second span and the corresponding fitted valuesis drawn in Figure 9 where the fitted values were obtainedby taking the RMS values of the accelerations in the middleof the first span into the fitting formula of the quadraticpolynomialThe correlation coefficient of the measured RMSvalues and fitted values on February 1st 2013 is 09391 and0918 for transverse and vertical accelerations respectively asshown in Figure 9 Computed using the same method all thedaily correlation coefficients in 2013 are more than 090 fortransverse and vertical accelerations as shown in Figure 10
The Scientific World Journal 7Ac
cele
ratio
n (m
ms2)
minus150
minus100
minus50
0
50
100
150
200
1 4001 6001 8001 100012001
Data point
(a) Accelerometer 1198852
Acce
lera
tion
(mm
s2)
1 4001 6001 8001 100012001
Data point
minus80
minus60
minus40
minus20
0
20
40
60
(b) Accelerometer 1198854
Figure 7 Vertical acceleration time histories of the main girder caused by a high-speed train
y = minus02642x2 + 37075x minus 56125
53 4 62 25 5535 45
RMS values of the 1st span (mms2)
0
2
4
6
8
10
RMS
valu
es o
f the
2nd
span
(mm
s2)
(a) Transverse accelerations
y = minus00131x2 + 07296x + 35995
50 15 20 2510
RMS values of the 1st span (mms2)
0
2
4
6
8
10
12
14
16RM
S va
lues
of t
he2
nd sp
an (m
ms2)
(b) Vertical accelerations
Figure 8 Cross-correlations between RMS values of accelerations measured in the middle of two main spans
Fitte
d RM
S va
lues
(mm
s2)
R = 09391
0
2
4
6
8
10
63 54 72 25 35 6555 7545
Measured RMS values (mms2)
(a) Transverse accelerations
R = 09518
4
6
8
10
12
14
16
Fitte
d RM
S va
lues
(mm
s2)
86 12 14 1610
Measured RMS values (mms2)
(b) Vertical accelerations
Figure 9 Fitting effect of cross-correlation between RMS values of accelerations by using quadratic polynomial
8 The Scientific World Journal
0 62 9331 186 217124 310279248155
Day
09
091
092
093
094
095
096
097C
orre
latio
n co
effici
ent
(a) Transverse accelerations
0 60 9030 180 210120 300270240150
Day
09
092
094
096
098
Cor
relat
ion
coeffi
cien
t
(b) Vertical accelerations
Figure 10 Daily computed correlation coefficients in 2013
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
s2)
minus2
minus1
0
1
2
20 40 60 80 100 120 140 1600Sample
Figure 11 Mean value control chart for accelerations in the middle of the second span
The analysis results indicate that good cross-correlationexists between the RMS values of the transverse and verticalaccelerations on the two main spans Multivariable controlchart is used here to monitor the measured changes in thetransverse and vertical accelerations caused by deteriorationof the vibration performance Firstly the condition index 119890
for early warning of abnormal vibration behavior is definedas the difference between the measured and predicted RMSvalues of the accelerations in the middle of the second span
119890 = RMS119898minus RMS
119901 (1)
where RMS119898is themeasured RMS values of the accelerations
in the middle of the second span RMS119901is the predicted
RMS values of the accelerations in the middle of the secondspan which is obtained by taking the RMS values of theaccelerations in the middle of the first span into the fittingformula of the quadratic polynomial in the healthy conditionThen a mean value control chart is employed to monitorthe time series of 119890 with regard to both transverse andvertical accelerations The time series of 119890 taken when thestructural vibration is in good condition will have somedistribution with mean 120583 and variance 120590
2 If the mean andstandard deviation are known a control chart is constructedby drawing a horizontal line CL at 120583 and twomore horizontallines representing the upper and lower control limits Theupper limit UCL is drawn at 120583 + 119896120590 and the lower limit LCL
at 120583 minus 119896120590 For online monitoring the controlling parameter119896 is chosen so that when the structural vibration is in goodcondition all observation samples fall between the controllimits When the new measurement is made the structuralabnormal vibration condition can be detected if an unusualnumber of samples fall beyond the control limits Figure 11shows the mean value control chart using the measuredand predicted RMS values of the accelerations shown inFigure 9 The upper and lower limits are determined usingthe controlling parameter 119896 = 2612 Hence a long-termmonitoring on the cross-correlation model of RMS values ofthe transverse and vertical accelerations can realize the earlywarning of deterioration of the vibration performance
32 Evaluation of Static Performance of Steel Truss Arch
321 Data of Temperature Field and Static Strains Tem-perature data from temperature sensor 119882
119894is denoted by
119879119894 temperature difference data acquired by 119879
119894minus 119879
119895is
denoted by119879119894119895(119879119894119895= 119879119894minus119879119895) and static strain data from strain
sensor119884119894is denoted by 119878
119894(119894 119895 = 1 2 8) As for illustration
the time-history curves of annual change and daily change intemperature 119879
1and temperature difference 119879
12are shown in
Figures 12 and 13 respectively And the time-history curvesof annual change and daily change in static strain data 119878
1
are shown in Figure 14 (negative values denote compressive
The Scientific World Journal 9
4 5 6 7 8 9 10 113Time (month)
minus5
5
15
25
35
45Te
mpe
ratu
re (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
5
10
15
20
Tem
pera
ture
(∘C)
(b) Daily curve (March 4th 2013)
Figure 12 Time-history curves of temperature data 1198791
4 5 6 7 8 9 10 113Time (month)
minus25
minus15
minus5
5
15
Tem
pera
ture
diff
eren
ce (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
minus10
minus5
0
5Te
mpe
ratu
re (∘
C)
(b) Daily curve (March 4th 2013)
Figure 13 Time-history curves of temperature difference data 11987912
4 5 6 7 8 9 10 113Time (month)
minus300
minus200
minus100
0
100
200
Stat
ic st
rain
(120583120576)
(a) Annual curve fromMarch 2013 to October 2013
Sharp peak
4 8 12 16 20 240Time (h)
minus120
minus65
minus10
45
100
Stat
ic st
rain
(120583120576)
(b) Daily curve (March 4th 2013)
Figure 14 Time-history curves of static strain data 1198781
10 The Scientific World Journal
4 8 12 16 20 240Time (h)
Stat
ic st
rain
S III
(120583120576)
minus120
minus65
minus10
45
100
(a) Daily extraction result 119878III of 1198781
4 5 6 7 8 9 10 113Time (month)
Stat
ic st
rain
S III
(120583120576)
minus200
minus130
minus60
10
80
150
(b) Annual extraction result 119878III of 1198781 fromMarch 2013 to October 2013
Figure 15 Extraction results 119878III of static strain data 1198781
static strain and positive values denote tension static strain)From Figure 14 it can be seen that time-history curve of 119878
1
contains many sharp peaks caused by high-speed trains witheach peak corresponding to the time when one train passesthrough the bridge And the static strains caused by high-speed trains are obviously lower than that of temperature fieldincluding temperature data and temperature difference data
As shown in Figure 14 static strain data in steel truss archmainly contains three parts static strain 119878I caused by tem-perature static strain 119878II caused by temperature differenceand static strain 119878III caused by train In order to construct thecorrelation between temperature field and static strain in steeltruss arch static strain 119878I and static strain 119878II (denoted by 119878III)must be extracted from the static strain data In the presentstudy wavelet packet decomposition method is used toextract 119878III considering the high-frequency characteristics ofstatic strain data caused by trains By using the wavelet packetdecomposition static strain data can be decomposed scale byscale into different frequency bands and each decompositioncoefficient corresponds to its frequency band [9 10] Eachdecomposition coefficient can be reconstructed into time-domain signals with constraints of its own frequency bandTaking the daily time-history curve of static strain data1198781shown in Figure 14(b) as an example it is decomposed
within 8 scales and the decomposition coefficient 11986208
119895
isspecially selected and then reconstructed as 119878III shown inFigure 15(a) It can be seen that sharp peaks caused by trainsare removed effectively and the static strains which are causedby temperature field are retained as well Furthermore theannual extraction result 119878III of 1198781 fromMarch 2013 toOctober2013 is shown in Figure 15(b)
322 Correlation Model between Temperature Field andStatic Strains In constructing the correlationmodel betweentemperature field and static strains data from March 2013to October 2013 are chosen as training data and data inNovember 2013 are chosen as test data for verification of
the correlation model The static strain 119878III is expressed asfollows
119878III =8
sum
119894=1
120572119894119879119894+
16
sum
119895=1
120573119895119863119895+ 119888 (2)
where 119879119894is 119894th temperature data from set 119879
1 1198792 1198793 1198794
1198795 1198796 1198797 1198798 119863119895is 119895th temperature difference data from set
11987912 11987934 11987956 11987978 11987913 11987915 11987917 11987935 11987937 11987957 11987924 11987926 11987928 11987946
11987948 11987968 120572119894is 119894th linear expansion coefficient of 119879
119894 120573119895is 119895th
linear expansion coefficient of 119879119894119895 and 119888 is the constant term
Considering that total 24 linear expansion coefficientsand one constant term in (2) are very complicated forcalculation principal component analysis is further usedto simplify (2) which aims at finding small amounts ofcomprehensive factors to represent main information of tem-perature and temperature difference data By using principalcomponent analysis the comprehensive factors can be foundby transforming temperature and temperature difference datainto factor component space [11 12] For temperature data theexplained variance of the first potential comprehensive factor1198751reaches 967 apparently larger than the other potential
comprehensive factors Thus 1198751is selected as the compre-
hensive factor Meanwhile for temperature difference datausing principal component analysis the explained varianceof the first three potential comprehensive factors 119877
1 1198772 and
1198773reaches 988 which are selected as the comprehensive
factorsTherefore the correlation models between 119878III and tem-
perature field can be expressed as follows
119878III = 1205821([119864]1times8
997888
1198798times1
) +9978881205741times3
([119865]3times16
997888
11986316times1
) + 119888 (3)
where997888
1198798times1
denotes one vector containing eight temperaturedata 997888119863
16times1denotes one vector containing sixteen temper-
ature difference data 119863119895(119895 = 1 2 16) [119864]
1times8and
The Scientific World Journal 11
b = minus3632
k = 0946
minus100 minus55 minus10 8035Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus100
minus55
minus10
35
80
Sim
ulat
iveS
III
(120583120576)
(a) 1198781
b = minus2013k = 0898
minus45minus75 minus15 45 7515Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus75
minus45
minus15
15
45
75
Sim
ulat
iveS
III
(120583120576)
(b) 1198782
b = minus1844k = 0904
minus60
minus30
0
30
60
90
Sim
ulat
iveS
III
(120583120576)
minus30minus60 30 60 900Monitoring SIII (120583120576)
Scattered pointsFitting curve
(c) 1198783
b = 2057k = 1126
minus15minus30 15 300Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus40
minus25
minus10
5
20
35Si
mul
ativ
eSII
I(120583120576)
(d) 1198784
b = minus4139k = 0935
minus100
minus65
minus30
5
40
75
Sim
ulat
iveS
III
(120583120576)
minus105 4515 75minus45minus75 minus15
Monitoring SIII (120583120576)
Scattered pointsFitting curve
(e) 1198785
b = minus4094k = 0918
minus145minus205 minus25minus85 9535Monitoring SIII (120583120576)
minus185
minus130
minus75
minus20
35
90
Sim
ulat
iveS
III
(120583120576)
Scattered pointsFitting curve
(f) 1198786
Figure 16 Continued
12 The Scientific World Journal
b = 2148k = 0881
minus50
minus28
minus6
16
38
60Si
mul
ativ
eSII
I(120583120576)
minus45 minus23 minus1 43 6521Monitoring SIII (120583120576)
Scattered pointsFitting curve
(g) 1198787
b = 0615k = 1118
minus30
minus15
0
15
30
Sim
ulat
iveS
III
(120583120576)
minus15minus25 minus5 15 255Monitoring SIII (120583120576)
Scattered pointsFitting curve
(h) 1198788
Figure 16 Correlation between simulative 119878III and monitoring 119878III
Table 2 Performance parameters 1205821
1205741
1205742
1205743
and 119862 in correlationmodels
Static straindata 120582
1
1205741
1205742
1205743
119862
119878III of 1198781 0454 minus4711 minus2597 minus4944 minus46806119878III of 1198782 1585 0013 4104 minus1247 minus70238119878III of 1198783 0301 3416 minus2098 2968 minus6884119878III of 1198784 0350 minus0734 0351 0948 minus23597119878III of 1198785 0179 minus4219 1827 3165 minus22049119878III of 1198786 0062 minus3245 8242 minus5398 minus22862119878III of 1198787 minus0367 2411 minus0598 5531 17075119878III of 1198788 0611 2188 minus1130 2018 minus32799
[119865]3times16
denote transformation matrix which can be cal-culated using principal component analysis [11 12] 120582
1is
performance parameter of first potential comprehensivefactor 119875
1 9978881205741times3
= [1205741 1205742 1205743] is one vector containing
three performance parameters corresponding to first threepotential comprehensive factors 119877
1 1198772 and 119877
3 The values of
performance parameters 1205821 9978881205741times3
and 119888 are estimated usingtraining data by multivariate linear regression method [13]and are shown in Table 2
In order to verify the effectiveness of the correlationmodels shown in (2) test data in November 2013 are usedto obtain the simulative 119878III The scattered points betweensimulative 119878III and monitoring 119878III are plotted as shown inFigure 16 which are further fitted by linear function
119878119904= 119896119878119898+ 119887 (4)
where 119878119904denotes simulative 119878III 119878119898 denotes monitoring
119878III and 119896 119887 are fitting parameters with the fitting resultsshown in Figure 16 as well The fitting results verify theeffectiveness of the correlationmodels Amultivariable mean
value control chart is also employed tomonitor the measuredabnormal changes in the static strains of the steel truss archThe condition index 119890 for early warning of deterioration ofthe static performance of steel truss arch is defined as thedifference between the measured and predicted static strains
119890 = 119878119898minus 119878119901 (5)
where 119878119898is the measured static strain 119878III 119878119901 is the predicted
static strain 119878III which is obtained by using the correlationmodels shown in (2) in the healthy condition Figure 17 showsthemean value control chart for static strains of the steel trussarch using measurement data in November 2013 The upperand lower limits are determined using the controlling param-eter 119896 = 2235 Hence a long-term monitoring on the corre-lation models between temperature field and static strains issuitable for early warning of deterioration of the static per-formance of steel truss arch
33 Evaluation of Movement Performance of Pier Longi-tudinal displacements of piers and temperature field datain steel truss arch collected from March 2013 to October2013 are used for evaluation of movement performanceof piers Longitudinal displacement data from sensors 119871
119894
(119894 = 1 2 6) are denoted by 119889119894(119905) (where 119905 means time)
respectively Meanwhile temperature field data from sensors119882119894(119894 = 1 2 8) are denoted by 119879
119894(119905) respectively and
their averaged value 119879(119905) is used to represent the averagetemperature of the steel truss arch For simplicity taking10 minutes as basic time interval the 10 min average dis-placements and temperatures are calculated and used as therepresentative values of this time interval By this way thedisplacements and temperatures in one day are reduced to144 representative samples Furthermore the temperature-displacement scatter diagrams are plotted in Figure 18 Anoverall increase in longitudinal displacements of piers isobserved with an increase in the average temperature of the
The Scientific World Journal 13
UCL
LCL
CL
Con
ditio
n in
dexe
(120583120576)
minus15
minus10
minus5
0
5
10
15
10 20 30 40 50 600Sample
Figure 17 Mean value control chart for static strains of the steel truss arch
Table 3 Summary of linear regression models
Pier Regression function4 pier 119889
1
= minus11524 + 696119879
5 pier 1198892
= minus9253 + 562119879
6 pier 1198893
= minus6628 + 382119879
8 pier 1198894
= minus4658 + 336119879
9 pier 1198895
= minus7605 + 530119879
10 pier 1198896
= minus9850 + 664119879
bridge And the correlation coefficients of the longitudinaldisplacements of piers and average temperature of the steeltruss arch are all larger than 092 as shown in Figure 18
A linear regression analysis between the 10 min averagedisplacement 119889
119894(119905) and the temperature 119879(119905) is further per-
formed to model the temperature-displacement correlationsusing the following equation [14]
119889119894(119905) = 120573
0+ 1205731119879 (119905) (6)
where the regression coefficients 1205731and 120573
0were obtained by
the least-squares method as
1205731=
119878119889119894119879
119878119879119879
1205730= 119889119894minus 1205731119879
(7)
where 119878119889119894119879
is the covariance between the displacement andthe temperature sequences 119878
119879119879is the variance of the mea-
sured temperature sequences and 119879 and 119889119894are the means
of the measured temperature and displacement sequencesrespectively Table 3 summarizes the expressions of linearregression models of the displacement versus temperature
The analysis results indicate that good correlation existsbetween the longitudinal displacements of piers and averagetemperature of steel truss arch A multivariable mean valuecontrol chart is also employed to monitor the measuredabnormal changes in the longitudinal displacements of piersThe condition index 119890 for early warning of abnormal behaviorof piers is defined as the difference between themeasured andpredicted longitudinal displacements
119890 = 119889119898minus 119889119901 (8)
where 119889119898is the measured displacement of the piers 119889
119901is
the predicted displacement of the piers which is obtainedby using the linear regression models shown in Table 3in the healthy condition Figure 19 shows the mean valuecontrol chart for longitudinal displacements of piers usingmeasurement data in March 2013 The upper and lowerlimits are determined using the controlling parameter 119896 =
2804 Hence a long-term monitoring on the temperature-displacement correlation model is suitable for real-timecondition assessment for movement performance of piers
34 Evaluation of Fatigue Performance of Steel Deck Fatiguecracking is universal for orthotropic steel deck of highwaybridges Once the fatigue crack appears extensive fatiguecracking may occur and repair is very difficult Henceinfinite-fatigue-life design method is applied for NanjingDashengguan Bridge instead of finite-fatigue-life designmethod which is applied extensively in highway bridges Theinfinite-fatigue-life design method is to control the fatiguestress amplitude in the range less than the value of fatiguecut-off limit refrained from fatigue cracking Consequentlythe purpose of equipping stain gauges in the SHM system ofDashengguan Bridge is to monitor the stress amplitude andthe validity of fatigue design can be achieved by comparingthe stress amplitude with fatigue stress limit
The structural dynamic response is asymmetrical becauseof the irregularity of railway and braking force from thetrain even though the structure of Nanjing DashengguanBridge is spatially symmetrical Hence the stresses are notsymmetrical measured by strain gauges from the upstreamand downstream directions However the fatigue amplitudesof two strain gauges induced by trains are both extremelysmall Thus the strain history data of the strain gauge DYB-2are merely selected in this paper to illustrate that the fatiguestress amplitude is much lower than the value of fatigue stresslimit The typical strain time-history curves of DYB-2 whensingle train with 8 carriages or 16 carriages passed throughthe bridge are shown in Figure 20 respectively It can be seenthat during the passing of high-speed trains both 8 carriagesand 16 carriages have led to strain variations reaching to 2ndash8 120583120576 And the number of strain peaks is 16 and 32 for trainswith 8 carriages and 16 carriages respectively because theaxle number of each carriage is 2
To perform fatigue evaluation simplified rainflow cyclecounting algorithm was firstly used to process strain history
14 The Scientific World Journal
R = 09748
10 20 30 40 500
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
(a) Correlation between 1198891of the 4 pier and average temperature 119879 of
steel truss arch
10 20 30 40 500
Temperature (∘C)
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
R = 09746
(b) Correlation between 1198892of the 5 pier and average temperature 119879 of
steel truss arch
10 200 40 5030
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
Disp
lace
men
t (m
m)
R = 09471
(c) Correlation between 1198893of the 6 pier and average temperature 119879 of
steel truss arch
R = 09276
10 20 30 40 500
Temperature (∘C)
minus60
minus30
0
30
60
90
120D
ispla
cem
ent (
mm
)
(d) Correlation between 1198894of the 8 pier and average temperature 119879 of
steel truss arch
R = 09536
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(e) Correlation between 1198895of the 9 pier and average temperature 119879 of
steel truss arch
R = 09537
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(f) Correlation between 1198896of the 10 pier and average temperature 119879 of
steel truss arch
Figure 18 Correlation between longitudinal displacement and average temperature
The Scientific World Journal 15
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
)
500 1000 1500 2000 2500 3000 3500 4000 45000Sample
minus15
minus10
minus5
0
5
10
15
Figure 19 Mean value control chart for longitudinal displacements of piers
Stra
in (120583
120576)
minus174
minus172
minus170
minus168
minus166
minus164
minus162
1 2714 53
66
79
92
40
118
209
183
222
105
157
196
144
131
235
248
170
Data point
(a) Single train with 8 carriages passed through the bridge
Stra
in (120583
120576)
minus180minus178minus176minus174minus172minus170minus168minus166
1
61
91
31
181
211
121
421
301
481
511
361
391
271
451
241
331
541
571
151
Data point
(b) Single train with 16 carriages passed through the bridge
Figure 20 Typical strain time-history curves of DYB-2 when single train passed through the bridge
data and the spectrum of stress amplitude was obtained[15] The spectra of stress amplitude calculated using thestrain history data with regard to 8 carriages or 16 carriagesare shown in Figure 21 respectively It can be seen that themaximum stress amplitude is smaller than 10MPa obtainedfrom strain history curves under single trainThen accordingtoMinerrsquos law and standard of Eurocode 3 fatigue parametersof 119878eq (equivalent stress amplitude) and 119873
119904(stress cycle
number) under single train were calculated as follows [16]
119878eq = (
1198991198941198785
119894
sum119899119894
)
15
119873s = sum119899119894
(9)
where 119878119894is 119894th stress amplitude 119899
119894is the number of
stress cycles corresponding to 119878119894 and 15 is slope value of
log 119878- log119873 curve in the scope of lower stress amplitudeaccording to the standard of Eurocode 3 [17]
Using (9) the equivalent stress amplitudes 119878eq with regardto 8 carriages or 16 carriages are 075MPa and 071MParespectively And the stress cycle numbers 119873
119904are 22 and 45
respectively Thus the value of 119878eq with regard to 8 carriagesis similar to that of 16 carriages However the value of 119873
119904
with regard to 8 carriages is about twice as much as that of 16carriages Furthermore as recommended by Eurocode 3 forrib-to-deck welded joint the fatigue limit Δ120590
119871is 29MPa [17]
and the value of 119878eq when the high-speed train passes throughbridge is far less than Δ120590
119871 which indicates that fatigue life of
rib-to-deck welded joint in Nanjing Dashengguan Bridge canbe considered infinite Thus the fatigue performance of steel
deck satisfies the requirement of infinite-fatigue-life designmethod
4 Conclusions
In the last few decades Structural Health Monitoring (SHM)system has been widely installed in many highway bridgesto continuously monitor the structural condition over anextended period of time However application of SHMsystem in high-speed railway bridges is far insufficient Thispaper presents an analysis of the SHM system installedin Nanjing Dashengguan Bridge in China The monitoringsystem is composed of a sensor system a data acquisitionand transmission system a data management system and astructural evaluation system The monitoring system is cur-rently in full operation accumulating the bridgersquos behaviorsunder the varying environmental conditions such as high-speed trains and environmental temperature
Based on long-termhealthmonitoring data general prin-ciples and analytical processes for structural evaluation areprovided and some evaluation results are reported It can beshown from the analysis in this paper that the mathematicalcorrelation models describing the overall structural behaviorof the bridge can be obtained with the support of the healthmonitoring system which includes cross-correlation modelsfor accelerations correlation models between temperatureand static strains of steel truss arch and correlation mod-els between temperature and longitudinal displacements ofpiers The mean value control chart is further employed tomonitor the time series of differences between the measuredstructural parameters and the predicted parameters usingthese correlation models The upper and lower control limits
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
2 The Scientific World Journal
(a) View of the Nanjing Dashengguan Bridge
192336
1stcross-section
1st cross-section 2nd cross-section
336
2ndcross-section
Beijing
1272
Shanghai
108 192 108
(b) Elevation drawing of the bridge (unit m)
Middle truss arch
Side truss arch
Orthotropic steel deck
Downstream15000
Side truss arch
Upstream15000
(c) Design drawing of the deck cross-section (unit mm)
Figure 1 Nanjing Dashengguan Bridge
high-speed railway bridges have been built during the lastfew years Maintaining the safety of these high-speed railwaybridges is crucial to the safety of passengers and trainsTherefore research efforts should be focused on the SHMsystems of high-speed railway bridges This paper reportsthe development and implementation of a SHM system ofNanjing Dashengguan Bridge Nanjing Dashengguan Bridgewhich serves as the shared corridor crossing Yangtze Riverfor both Beijing-Shanghai high-speed railway and Shanghai-Wuhan-Chengdu railway is the first 6-track high-speedrailway bridge with the longest span throughout the worldIt is the largest bridge with the heaviest design loading of allthe high-speed railway bridges by far because its main spanis 336m and its railway contains 6 tracks Also the designspeed of 300 kmh of the bridge is on the advanced level in theworld Due to these remarkable characteristics including longspan of themain girder heavy design loading and high speedof trains it is necessary to monitor the real-time health statusof the bridge during its service period With the state-of-the-art techniques of current monitoring systems of highwaybridges the health monitoring system for this bridge wasdesigned to be concise practical reliable and cost efficientThe main purpose of this system is to monitor the bridgersquosbehaviors under the varying environmental conditions and
detect the bridgersquos abnormal behaviors which may threatenthe structural safety This paper introduces the deploymentand functions of this system including a sensor system adata acquisition and transmission system a datamanagementsystem and a structural evaluation system Some evaluationresults of the bridge subjected to varying environmentalconditions such as high-speed trains and environmentaltemperature are also presented
2 Structural Health Monitoring System
Nanjing Dashengguan Bridge shown in Figure 1(a) is a steeltruss arched bridge with the span arrangement (108 + 192 +2 times 336 + 192 + 108)m which supports a six-line railwayincluding two regular rails two high-speed railways andtwo subway lines The elevation drawing of the bridge andthe deck cross-section are shown in Figures 1(b) and 1(c)respectively The three truss planes in the longitudinal direc-tion were used with a spacing of 15m which was selected toprovide the large stiffness for such heavy live loadsThe use ofa steel box as the deck system is another feature in the designof this steel truss arc which is adopted as the bottom chord ofthe entire truss segment to reduce uneven deflection on thedeck The bearing system of the bridge is that the 7 pier is
The Scientific World Journal 3
Table 1 Sensitivity analysis of bridge parameters for SHM design
Structure Key position Critical member Parameter ImportanceVariation undertemperature
actionSensitivity Decision
Pier Side spans 4 and 10 piers Longitudinaldisplacement Critical 229mm Sensitive Monitor
Main girder In the middle ofthe main span Steel deck Vertical
displacementLess-
important 35mm Less-sensitive No
Truss arch In the middle ofthe main span
Top chordmember diagonalweb member archrib chord memberand bottom chord
member
Static strain Critical 280 120583120576 Sensitive Monitor
Truss arch In the middle ofthe main span
Top chord memberand bottom chord
memberTemperature Critical 60∘C Sensitive Monitor
Main girder In the middle ofthe main span Steel deck Temperature Less-
important 60∘C Less-sensitive No
Whole bridge Frequency Important Nonsensitive Monitor one
Truss arch In the middle ofthe main span Top chord member Wind Less-
important Less-sensitive No
linked to the bridge deck so that the bridge deck is restrictedfrom moving in the longitudinal direction at the 7 pierAnd there have no longitudinal restraints between the bridgedeck and other piers Due to the remarkable characteristics ofNanjingDashengguanBridge including long span of themaingirder heavy design loading and high speed of trains a long-term SHM system was designed and installed on the NanjingDashengguan Bridge shortly after it was opened to railwaytraffic This online SHM system was designed to be a concisemonitoring system with a minimum of 28 sensors to monitorthe key parametersThe entire system consists of a sensor sys-tem a data acquisition system a data transmission system adatamanagement system and a structural evaluation systemThe primary purpose of the system is to monitor in-serviceperformance of the bridge structure under the varying envi-ronmental conditions such as high-speed trains and environ-mental temperature and to provide early warning of abnor-mal changes in in-service performance of the bridge thatmight be unsafe to bridge structureThe procedure of design-ing and implementing the SHM system includes three majorstages as follows (1) the stage of planning and design includesthe consideration of the physical quantities to be measuredthe selection of sensors appropriate for measuring such phys-ical parameters and the optimal placement of the sensors (2)The stage of installation and calibration includes a series ofstatic and dynamic experiments performed to check the oper-ation of the monitoring system after the completion of thebridge (3)The stage of main operation includes the test oper-ation of the monitoring system training of the maintenancecrew and the replacement ofmalfunctioning sensing devices
The design objective of the SHM system is to monitor thestructural health under daily service condition With regardto the bridge site of Nanjing Dashengguan Bridge there arevery lowprobability events of Typhoon and earthquakeThus
the train loading and temperature action are the most impor-tant factors that affect the daily performance of the bridgeDue to the design requirements the structural responses ofthe bridge under train loading are small and conservativeHowever the axial movements and thermal stresses of thebridge under temperature action are considerable due to thelong span of the bridge For instance the maximum thermalstrain and train-induced strain in the steel truss arch areabout 250 120583120576 and 50 120583120576 respectively For bridge conditionassessment it is of paramount importance to characterize nor-mal variability of structural parameters due to temperatureeffectsTherefore in order to control the overall budget of theSHM system without losing major information on structuraldaily behavior the structural parameters to be monitoredin the Nanjing Dashengguan Bridge were determined usingthe structural sensitivity analysis with the finite elementmodel under the designed temperature action as shown inTable 1 Other important parameters including accelerationsof steel deck and fatigue of steel deck were also monitoredConsidering that the displacement amplitudes of steel deckunder the train loading can be derived from the accelerationof steel deck the vertical displacements of steel deck were notchosen to monitor
21 Temperature and Static Strain Monitoring of Truss ArchAs shown in Figure 1(b) temperature and static strain mon-itoring of truss arch is performed at the 1st cross-section inthe middle of the first main span of the bridge Eight FBGtemperature and strain gauges are both installed on the 1stcross-section to simultaneously measure temperature fielddata and static strain field data in the steel truss arch as shownin Figure 2 It can be seen from Figures 2(c)sim2(f) that thesensors were installed at the top chordmember diagonal webmember arch rib chord member and bottom chord member
4 The Scientific World Journal
Downstream Upstream
Horizontal bracing of top chord member
Vertical bracing
Vertical web memberof side truss
Vertical web memberof middle truss
Top plate
Transverse stiffening girder
Horizontal bracing ofarch rib chord member
Vertical web memberof side truss
W1
W2
W3
W4
W5
W6
W7 W8
Y1
Y2
Y3Y4
Y5
Y6
Y7 Y8
15000 15000
68007
12000
2
2
(a) 1st cross-section of steel truss arch
Shanghai
Top chord member
Arch rib chord member
Vertical web member
Beijing
Diagonal web member
Bottom chord member
W1
W2
W3
W4
W5
W6
W7
W8
Y1Y2
Y3Y4
Y5
Y6
Y7
Y8
3
3
4
4
5
5
6
6
(b) 2-2 section of steel truss arch
Downstream Upstream
W1 W2Y1 Y2
1000
1000
850 850
150
150
1400
(c) 3-3 section of top chord member
Downstream Upstream
W3W4Y3Y4
150
150
610
610
1398
(d) 4-4 section of diagonal web member
Downstream Upstream
W5
W6Y5
Y6
150
150
1400
650 650
800
(e) 5-5 section of arch rib chord member
Downstream Upstream
W7 W8Y7 Y8
150
150
1266
1266
1400
1416
(f) 6-6 section of bottom chord member
Figure 2 Layout of temperature and static strain gauges on the steel truss arch (unit mm)
The Scientific World Journal 5
DYB-1 DYB-2
2500mm2500mm
Figure 3 Layout of dynamic strain gauges on the steel deck
Vertical accelerometer Z2Z4Transverse accelerometer Z1Z3
Figure 4 Layout of accelerometers on the steel deck
of the truss arch (where 119882119894denotes 119894th temperature sensor
119894 = 1 2 8 119884119894denotes 119894th strain sensor 119894 = 1 2 8)
The specification of FBG temperature sensors and straingauges is FBG4700 and FBG4100 supplied by China GeokonInstrumentsCo Ltd Sampling frequency of temperature andstatic strain data collection is set to 1Hz
22 Dynamic StrainMonitoring of Steel Deck Dynamic straingauges are installed at the 1st cross-section in the middleof the first main span of the bridge so as to monitor thefatigue performance of orthotropic steel deck as shown inFigure 3 It can be seen that the dynamic strain gauges DYB-1and DYB-2 monitor the long-term fatigue effects of deck-ribwelded detail induced by the traveling high-speed trainsThespecification of dynamic strain gauges is FBG4100 suppliedby China Geokon Instruments Co Ltd Sampling frequencyof dynamic strain data collections is set to 50Hz
23 Acceleration Monitoring of Steel Deck In order to moni-tor the transverse and vertical acceleration responses on thetwo main spans induced by the traveling high-speed trainsone transverse accelerometer and one vertical accelerometerwere installed at the 1st cross-section and 2nd cross-sectionin the middle of two main spans respectively as shown inFigure 4 (where 119885
119894denotes 119894th accelerometer 119894 = 1 2 4)
The specification of accelerometers is DH610 supplied byDonghua Testing Technology Co Ltd Sampling frequencyof acceleration data collections is set to 200Hz
24 Longitudinal Displacement Monitoring of Piers In thedesign of Nanjing Dashengguan Bridge the 7 pier is linkedto the bridge deck so that the bridge deck is restricted frommoving in the longitudinal direction at the 7 pier And therehave no longitudinal restraints between the bridge deck andother piersTherefore 6 displacement transducers are locatedon the piers and used for the measurement of longitudinaldisplacements at the 4 pier 5 pier 6 pier 8 pier 9pier and 10 pier as shown in Figure 5 (where 119871
119894denotes 119894th
displacement transducer 119894 = 1 2 6) The specificationof displacement transducers is CLMD2 supplied by ASMrsquosin Germany Finally it should be noted that the DAQs of theSHM system are located in the 6 and 8 piers
3 Evaluation of Structural Condition UsingLong-Term Monitoring Data
Using long-term monitoring data collected from the SHMsystem of Nanjing Dashengguan Bridge structural conditionassessment involves four aspects evaluation of vibration per-formance of themain girder evaluation of static performanceof steel truss arch evaluation of movement performance ofpier and evaluation of fatigue performance of steel deck
31 Evaluation of Vibration Performance of the Main GirderThe main function for the SHM system is to detect theperformance degradation of structures as early as possibleand to provide essential references for maintenance of actualbridges For Nanjing Dashengguan Bridge the vibration andstrain of the main girder together with the longitudinaldisplacement of piers are three important items to moni-tor Three kinds of performance evaluation are conductedaccordingly Those are evaluation of vibration performanceof the main girder evaluation of static performance of steeltruss arch and evaluation of movement performance of pierFirstly degradation or damage of local members may occurwith the increase of service period which will reduce thestructural stiffness of bridge Thus the dynamic propertyof bridge structure is affected significantly by the structuralstiffness reduction and will cause unusual variations ofthe structural vibration Such unusual behavior will affectrunning safety of the train directly Secondly damage of localmembers may induce redistribution of the internal forces inthe whole bridge under the temperature action The reasonis that Nanjing Dashengguan Bridge belongs to staticallyindeterminate structure Therefore static strain monitoringis necessary for typical members under temperature actions
6 The Scientific World Journal
Beijing Shanghai
4 pier 5 pier 6 pier 7 pier 8 pier 9 pier 10 pier
L1 L2 L3 L4 L5 L6
Figure 5 Layout of displacement transducers on the pier
Acce
lera
tion
(mm
s2)
1 80012001 4001 6001 10001
Data point
minus40
minus30
minus20
minus10
0
10
20
30
40
(a) Accelerometer 1198851
Acce
lera
tion
(mm
s2)
1 80012001 4001 6001 10001
Data point
minus40
minus30
minus20
minus10
0
10
20
30
40
(b) Accelerometer 1198853
Figure 6 Transverse acceleration time histories of the main girder induced by a high-speed train
Thirdly the longitudinal displacement behavior of the bridgepier is crucial for running safety of the train under thetemperature action Once the abnormal variation of the itemis detected inspection may be evolved timely for detection ofthe bridge piers In this section the evaluation of vibrationperformance of main girder is firstly discussed
For high-speed railway bridges vibration performanceof the main girder subjected to the high-speed trains isimportant for passenger comfort and traffic safety [6 7]For instance Zeng et al [8] point out that the essentialreason of train derailment is that the transverse vibrationof the train-railway-bridge system becomes unstable Henceit is necessary to present a strategy for early warning ofdeterioration of vibration performance of the main girderusing long-term health monitoring data
Figures 6 and 7 illustrate the typical transverse and ver-tical acceleration time histories of the main girder measuredfrom accelerometers Z
1simZ4 respectively It can be observed
that when high-speed train passed the bridge the transverseaccelerations in the middle of the first span are much smallerthan that of the second span and the vertical accelerationsin the middle of the first span are much larger than that ofthe second span These indicate that although the structurallayouts of two main spans of the main girder are the samethere exists significant difference between the transverseand vertical vibration characteristics of two main spans due
to the rail irregularity in transverse and vertical directionsThus there is a need to monitor the transverse and verticalaccelerations of two main spans in the long term so as torealize anomaly alarms for vibration performance of themaingirder
Hence in the present study the RMS values of theaccelerations are utilized as the monitoring parameters torepresent the transverse and vertical vibration characteristicof the main girder The cross-correlation between the RMSvalues of the accelerations in the middle of two main spans isfurther investigated when the high-speed trains pass throughthe bridge Figure 8 shows the scatter plots between the RMSvalues of the accelerations and shows the fitted results byusing quadratic polynomial as well
By using the fitting formula constructed in Figure 8scatter plot between theRMSvalues of the accelerations in themiddle of the second span and the corresponding fitted valuesis drawn in Figure 9 where the fitted values were obtainedby taking the RMS values of the accelerations in the middleof the first span into the fitting formula of the quadraticpolynomialThe correlation coefficient of the measured RMSvalues and fitted values on February 1st 2013 is 09391 and0918 for transverse and vertical accelerations respectively asshown in Figure 9 Computed using the same method all thedaily correlation coefficients in 2013 are more than 090 fortransverse and vertical accelerations as shown in Figure 10
The Scientific World Journal 7Ac
cele
ratio
n (m
ms2)
minus150
minus100
minus50
0
50
100
150
200
1 4001 6001 8001 100012001
Data point
(a) Accelerometer 1198852
Acce
lera
tion
(mm
s2)
1 4001 6001 8001 100012001
Data point
minus80
minus60
minus40
minus20
0
20
40
60
(b) Accelerometer 1198854
Figure 7 Vertical acceleration time histories of the main girder caused by a high-speed train
y = minus02642x2 + 37075x minus 56125
53 4 62 25 5535 45
RMS values of the 1st span (mms2)
0
2
4
6
8
10
RMS
valu
es o
f the
2nd
span
(mm
s2)
(a) Transverse accelerations
y = minus00131x2 + 07296x + 35995
50 15 20 2510
RMS values of the 1st span (mms2)
0
2
4
6
8
10
12
14
16RM
S va
lues
of t
he2
nd sp
an (m
ms2)
(b) Vertical accelerations
Figure 8 Cross-correlations between RMS values of accelerations measured in the middle of two main spans
Fitte
d RM
S va
lues
(mm
s2)
R = 09391
0
2
4
6
8
10
63 54 72 25 35 6555 7545
Measured RMS values (mms2)
(a) Transverse accelerations
R = 09518
4
6
8
10
12
14
16
Fitte
d RM
S va
lues
(mm
s2)
86 12 14 1610
Measured RMS values (mms2)
(b) Vertical accelerations
Figure 9 Fitting effect of cross-correlation between RMS values of accelerations by using quadratic polynomial
8 The Scientific World Journal
0 62 9331 186 217124 310279248155
Day
09
091
092
093
094
095
096
097C
orre
latio
n co
effici
ent
(a) Transverse accelerations
0 60 9030 180 210120 300270240150
Day
09
092
094
096
098
Cor
relat
ion
coeffi
cien
t
(b) Vertical accelerations
Figure 10 Daily computed correlation coefficients in 2013
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
s2)
minus2
minus1
0
1
2
20 40 60 80 100 120 140 1600Sample
Figure 11 Mean value control chart for accelerations in the middle of the second span
The analysis results indicate that good cross-correlationexists between the RMS values of the transverse and verticalaccelerations on the two main spans Multivariable controlchart is used here to monitor the measured changes in thetransverse and vertical accelerations caused by deteriorationof the vibration performance Firstly the condition index 119890
for early warning of abnormal vibration behavior is definedas the difference between the measured and predicted RMSvalues of the accelerations in the middle of the second span
119890 = RMS119898minus RMS
119901 (1)
where RMS119898is themeasured RMS values of the accelerations
in the middle of the second span RMS119901is the predicted
RMS values of the accelerations in the middle of the secondspan which is obtained by taking the RMS values of theaccelerations in the middle of the first span into the fittingformula of the quadratic polynomial in the healthy conditionThen a mean value control chart is employed to monitorthe time series of 119890 with regard to both transverse andvertical accelerations The time series of 119890 taken when thestructural vibration is in good condition will have somedistribution with mean 120583 and variance 120590
2 If the mean andstandard deviation are known a control chart is constructedby drawing a horizontal line CL at 120583 and twomore horizontallines representing the upper and lower control limits Theupper limit UCL is drawn at 120583 + 119896120590 and the lower limit LCL
at 120583 minus 119896120590 For online monitoring the controlling parameter119896 is chosen so that when the structural vibration is in goodcondition all observation samples fall between the controllimits When the new measurement is made the structuralabnormal vibration condition can be detected if an unusualnumber of samples fall beyond the control limits Figure 11shows the mean value control chart using the measuredand predicted RMS values of the accelerations shown inFigure 9 The upper and lower limits are determined usingthe controlling parameter 119896 = 2612 Hence a long-termmonitoring on the cross-correlation model of RMS values ofthe transverse and vertical accelerations can realize the earlywarning of deterioration of the vibration performance
32 Evaluation of Static Performance of Steel Truss Arch
321 Data of Temperature Field and Static Strains Tem-perature data from temperature sensor 119882
119894is denoted by
119879119894 temperature difference data acquired by 119879
119894minus 119879
119895is
denoted by119879119894119895(119879119894119895= 119879119894minus119879119895) and static strain data from strain
sensor119884119894is denoted by 119878
119894(119894 119895 = 1 2 8) As for illustration
the time-history curves of annual change and daily change intemperature 119879
1and temperature difference 119879
12are shown in
Figures 12 and 13 respectively And the time-history curvesof annual change and daily change in static strain data 119878
1
are shown in Figure 14 (negative values denote compressive
The Scientific World Journal 9
4 5 6 7 8 9 10 113Time (month)
minus5
5
15
25
35
45Te
mpe
ratu
re (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
5
10
15
20
Tem
pera
ture
(∘C)
(b) Daily curve (March 4th 2013)
Figure 12 Time-history curves of temperature data 1198791
4 5 6 7 8 9 10 113Time (month)
minus25
minus15
minus5
5
15
Tem
pera
ture
diff
eren
ce (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
minus10
minus5
0
5Te
mpe
ratu
re (∘
C)
(b) Daily curve (March 4th 2013)
Figure 13 Time-history curves of temperature difference data 11987912
4 5 6 7 8 9 10 113Time (month)
minus300
minus200
minus100
0
100
200
Stat
ic st
rain
(120583120576)
(a) Annual curve fromMarch 2013 to October 2013
Sharp peak
4 8 12 16 20 240Time (h)
minus120
minus65
minus10
45
100
Stat
ic st
rain
(120583120576)
(b) Daily curve (March 4th 2013)
Figure 14 Time-history curves of static strain data 1198781
10 The Scientific World Journal
4 8 12 16 20 240Time (h)
Stat
ic st
rain
S III
(120583120576)
minus120
minus65
minus10
45
100
(a) Daily extraction result 119878III of 1198781
4 5 6 7 8 9 10 113Time (month)
Stat
ic st
rain
S III
(120583120576)
minus200
minus130
minus60
10
80
150
(b) Annual extraction result 119878III of 1198781 fromMarch 2013 to October 2013
Figure 15 Extraction results 119878III of static strain data 1198781
static strain and positive values denote tension static strain)From Figure 14 it can be seen that time-history curve of 119878
1
contains many sharp peaks caused by high-speed trains witheach peak corresponding to the time when one train passesthrough the bridge And the static strains caused by high-speed trains are obviously lower than that of temperature fieldincluding temperature data and temperature difference data
As shown in Figure 14 static strain data in steel truss archmainly contains three parts static strain 119878I caused by tem-perature static strain 119878II caused by temperature differenceand static strain 119878III caused by train In order to construct thecorrelation between temperature field and static strain in steeltruss arch static strain 119878I and static strain 119878II (denoted by 119878III)must be extracted from the static strain data In the presentstudy wavelet packet decomposition method is used toextract 119878III considering the high-frequency characteristics ofstatic strain data caused by trains By using the wavelet packetdecomposition static strain data can be decomposed scale byscale into different frequency bands and each decompositioncoefficient corresponds to its frequency band [9 10] Eachdecomposition coefficient can be reconstructed into time-domain signals with constraints of its own frequency bandTaking the daily time-history curve of static strain data1198781shown in Figure 14(b) as an example it is decomposed
within 8 scales and the decomposition coefficient 11986208
119895
isspecially selected and then reconstructed as 119878III shown inFigure 15(a) It can be seen that sharp peaks caused by trainsare removed effectively and the static strains which are causedby temperature field are retained as well Furthermore theannual extraction result 119878III of 1198781 fromMarch 2013 toOctober2013 is shown in Figure 15(b)
322 Correlation Model between Temperature Field andStatic Strains In constructing the correlationmodel betweentemperature field and static strains data from March 2013to October 2013 are chosen as training data and data inNovember 2013 are chosen as test data for verification of
the correlation model The static strain 119878III is expressed asfollows
119878III =8
sum
119894=1
120572119894119879119894+
16
sum
119895=1
120573119895119863119895+ 119888 (2)
where 119879119894is 119894th temperature data from set 119879
1 1198792 1198793 1198794
1198795 1198796 1198797 1198798 119863119895is 119895th temperature difference data from set
11987912 11987934 11987956 11987978 11987913 11987915 11987917 11987935 11987937 11987957 11987924 11987926 11987928 11987946
11987948 11987968 120572119894is 119894th linear expansion coefficient of 119879
119894 120573119895is 119895th
linear expansion coefficient of 119879119894119895 and 119888 is the constant term
Considering that total 24 linear expansion coefficientsand one constant term in (2) are very complicated forcalculation principal component analysis is further usedto simplify (2) which aims at finding small amounts ofcomprehensive factors to represent main information of tem-perature and temperature difference data By using principalcomponent analysis the comprehensive factors can be foundby transforming temperature and temperature difference datainto factor component space [11 12] For temperature data theexplained variance of the first potential comprehensive factor1198751reaches 967 apparently larger than the other potential
comprehensive factors Thus 1198751is selected as the compre-
hensive factor Meanwhile for temperature difference datausing principal component analysis the explained varianceof the first three potential comprehensive factors 119877
1 1198772 and
1198773reaches 988 which are selected as the comprehensive
factorsTherefore the correlation models between 119878III and tem-
perature field can be expressed as follows
119878III = 1205821([119864]1times8
997888
1198798times1
) +9978881205741times3
([119865]3times16
997888
11986316times1
) + 119888 (3)
where997888
1198798times1
denotes one vector containing eight temperaturedata 997888119863
16times1denotes one vector containing sixteen temper-
ature difference data 119863119895(119895 = 1 2 16) [119864]
1times8and
The Scientific World Journal 11
b = minus3632
k = 0946
minus100 minus55 minus10 8035Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus100
minus55
minus10
35
80
Sim
ulat
iveS
III
(120583120576)
(a) 1198781
b = minus2013k = 0898
minus45minus75 minus15 45 7515Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus75
minus45
minus15
15
45
75
Sim
ulat
iveS
III
(120583120576)
(b) 1198782
b = minus1844k = 0904
minus60
minus30
0
30
60
90
Sim
ulat
iveS
III
(120583120576)
minus30minus60 30 60 900Monitoring SIII (120583120576)
Scattered pointsFitting curve
(c) 1198783
b = 2057k = 1126
minus15minus30 15 300Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus40
minus25
minus10
5
20
35Si
mul
ativ
eSII
I(120583120576)
(d) 1198784
b = minus4139k = 0935
minus100
minus65
minus30
5
40
75
Sim
ulat
iveS
III
(120583120576)
minus105 4515 75minus45minus75 minus15
Monitoring SIII (120583120576)
Scattered pointsFitting curve
(e) 1198785
b = minus4094k = 0918
minus145minus205 minus25minus85 9535Monitoring SIII (120583120576)
minus185
minus130
minus75
minus20
35
90
Sim
ulat
iveS
III
(120583120576)
Scattered pointsFitting curve
(f) 1198786
Figure 16 Continued
12 The Scientific World Journal
b = 2148k = 0881
minus50
minus28
minus6
16
38
60Si
mul
ativ
eSII
I(120583120576)
minus45 minus23 minus1 43 6521Monitoring SIII (120583120576)
Scattered pointsFitting curve
(g) 1198787
b = 0615k = 1118
minus30
minus15
0
15
30
Sim
ulat
iveS
III
(120583120576)
minus15minus25 minus5 15 255Monitoring SIII (120583120576)
Scattered pointsFitting curve
(h) 1198788
Figure 16 Correlation between simulative 119878III and monitoring 119878III
Table 2 Performance parameters 1205821
1205741
1205742
1205743
and 119862 in correlationmodels
Static straindata 120582
1
1205741
1205742
1205743
119862
119878III of 1198781 0454 minus4711 minus2597 minus4944 minus46806119878III of 1198782 1585 0013 4104 minus1247 minus70238119878III of 1198783 0301 3416 minus2098 2968 minus6884119878III of 1198784 0350 minus0734 0351 0948 minus23597119878III of 1198785 0179 minus4219 1827 3165 minus22049119878III of 1198786 0062 minus3245 8242 minus5398 minus22862119878III of 1198787 minus0367 2411 minus0598 5531 17075119878III of 1198788 0611 2188 minus1130 2018 minus32799
[119865]3times16
denote transformation matrix which can be cal-culated using principal component analysis [11 12] 120582
1is
performance parameter of first potential comprehensivefactor 119875
1 9978881205741times3
= [1205741 1205742 1205743] is one vector containing
three performance parameters corresponding to first threepotential comprehensive factors 119877
1 1198772 and 119877
3 The values of
performance parameters 1205821 9978881205741times3
and 119888 are estimated usingtraining data by multivariate linear regression method [13]and are shown in Table 2
In order to verify the effectiveness of the correlationmodels shown in (2) test data in November 2013 are usedto obtain the simulative 119878III The scattered points betweensimulative 119878III and monitoring 119878III are plotted as shown inFigure 16 which are further fitted by linear function
119878119904= 119896119878119898+ 119887 (4)
where 119878119904denotes simulative 119878III 119878119898 denotes monitoring
119878III and 119896 119887 are fitting parameters with the fitting resultsshown in Figure 16 as well The fitting results verify theeffectiveness of the correlationmodels Amultivariable mean
value control chart is also employed tomonitor the measuredabnormal changes in the static strains of the steel truss archThe condition index 119890 for early warning of deterioration ofthe static performance of steel truss arch is defined as thedifference between the measured and predicted static strains
119890 = 119878119898minus 119878119901 (5)
where 119878119898is the measured static strain 119878III 119878119901 is the predicted
static strain 119878III which is obtained by using the correlationmodels shown in (2) in the healthy condition Figure 17 showsthemean value control chart for static strains of the steel trussarch using measurement data in November 2013 The upperand lower limits are determined using the controlling param-eter 119896 = 2235 Hence a long-term monitoring on the corre-lation models between temperature field and static strains issuitable for early warning of deterioration of the static per-formance of steel truss arch
33 Evaluation of Movement Performance of Pier Longi-tudinal displacements of piers and temperature field datain steel truss arch collected from March 2013 to October2013 are used for evaluation of movement performanceof piers Longitudinal displacement data from sensors 119871
119894
(119894 = 1 2 6) are denoted by 119889119894(119905) (where 119905 means time)
respectively Meanwhile temperature field data from sensors119882119894(119894 = 1 2 8) are denoted by 119879
119894(119905) respectively and
their averaged value 119879(119905) is used to represent the averagetemperature of the steel truss arch For simplicity taking10 minutes as basic time interval the 10 min average dis-placements and temperatures are calculated and used as therepresentative values of this time interval By this way thedisplacements and temperatures in one day are reduced to144 representative samples Furthermore the temperature-displacement scatter diagrams are plotted in Figure 18 Anoverall increase in longitudinal displacements of piers isobserved with an increase in the average temperature of the
The Scientific World Journal 13
UCL
LCL
CL
Con
ditio
n in
dexe
(120583120576)
minus15
minus10
minus5
0
5
10
15
10 20 30 40 50 600Sample
Figure 17 Mean value control chart for static strains of the steel truss arch
Table 3 Summary of linear regression models
Pier Regression function4 pier 119889
1
= minus11524 + 696119879
5 pier 1198892
= minus9253 + 562119879
6 pier 1198893
= minus6628 + 382119879
8 pier 1198894
= minus4658 + 336119879
9 pier 1198895
= minus7605 + 530119879
10 pier 1198896
= minus9850 + 664119879
bridge And the correlation coefficients of the longitudinaldisplacements of piers and average temperature of the steeltruss arch are all larger than 092 as shown in Figure 18
A linear regression analysis between the 10 min averagedisplacement 119889
119894(119905) and the temperature 119879(119905) is further per-
formed to model the temperature-displacement correlationsusing the following equation [14]
119889119894(119905) = 120573
0+ 1205731119879 (119905) (6)
where the regression coefficients 1205731and 120573
0were obtained by
the least-squares method as
1205731=
119878119889119894119879
119878119879119879
1205730= 119889119894minus 1205731119879
(7)
where 119878119889119894119879
is the covariance between the displacement andthe temperature sequences 119878
119879119879is the variance of the mea-
sured temperature sequences and 119879 and 119889119894are the means
of the measured temperature and displacement sequencesrespectively Table 3 summarizes the expressions of linearregression models of the displacement versus temperature
The analysis results indicate that good correlation existsbetween the longitudinal displacements of piers and averagetemperature of steel truss arch A multivariable mean valuecontrol chart is also employed to monitor the measuredabnormal changes in the longitudinal displacements of piersThe condition index 119890 for early warning of abnormal behaviorof piers is defined as the difference between themeasured andpredicted longitudinal displacements
119890 = 119889119898minus 119889119901 (8)
where 119889119898is the measured displacement of the piers 119889
119901is
the predicted displacement of the piers which is obtainedby using the linear regression models shown in Table 3in the healthy condition Figure 19 shows the mean valuecontrol chart for longitudinal displacements of piers usingmeasurement data in March 2013 The upper and lowerlimits are determined using the controlling parameter 119896 =
2804 Hence a long-term monitoring on the temperature-displacement correlation model is suitable for real-timecondition assessment for movement performance of piers
34 Evaluation of Fatigue Performance of Steel Deck Fatiguecracking is universal for orthotropic steel deck of highwaybridges Once the fatigue crack appears extensive fatiguecracking may occur and repair is very difficult Henceinfinite-fatigue-life design method is applied for NanjingDashengguan Bridge instead of finite-fatigue-life designmethod which is applied extensively in highway bridges Theinfinite-fatigue-life design method is to control the fatiguestress amplitude in the range less than the value of fatiguecut-off limit refrained from fatigue cracking Consequentlythe purpose of equipping stain gauges in the SHM system ofDashengguan Bridge is to monitor the stress amplitude andthe validity of fatigue design can be achieved by comparingthe stress amplitude with fatigue stress limit
The structural dynamic response is asymmetrical becauseof the irregularity of railway and braking force from thetrain even though the structure of Nanjing DashengguanBridge is spatially symmetrical Hence the stresses are notsymmetrical measured by strain gauges from the upstreamand downstream directions However the fatigue amplitudesof two strain gauges induced by trains are both extremelysmall Thus the strain history data of the strain gauge DYB-2are merely selected in this paper to illustrate that the fatiguestress amplitude is much lower than the value of fatigue stresslimit The typical strain time-history curves of DYB-2 whensingle train with 8 carriages or 16 carriages passed throughthe bridge are shown in Figure 20 respectively It can be seenthat during the passing of high-speed trains both 8 carriagesand 16 carriages have led to strain variations reaching to 2ndash8 120583120576 And the number of strain peaks is 16 and 32 for trainswith 8 carriages and 16 carriages respectively because theaxle number of each carriage is 2
To perform fatigue evaluation simplified rainflow cyclecounting algorithm was firstly used to process strain history
14 The Scientific World Journal
R = 09748
10 20 30 40 500
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
(a) Correlation between 1198891of the 4 pier and average temperature 119879 of
steel truss arch
10 20 30 40 500
Temperature (∘C)
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
R = 09746
(b) Correlation between 1198892of the 5 pier and average temperature 119879 of
steel truss arch
10 200 40 5030
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
Disp
lace
men
t (m
m)
R = 09471
(c) Correlation between 1198893of the 6 pier and average temperature 119879 of
steel truss arch
R = 09276
10 20 30 40 500
Temperature (∘C)
minus60
minus30
0
30
60
90
120D
ispla
cem
ent (
mm
)
(d) Correlation between 1198894of the 8 pier and average temperature 119879 of
steel truss arch
R = 09536
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(e) Correlation between 1198895of the 9 pier and average temperature 119879 of
steel truss arch
R = 09537
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(f) Correlation between 1198896of the 10 pier and average temperature 119879 of
steel truss arch
Figure 18 Correlation between longitudinal displacement and average temperature
The Scientific World Journal 15
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
)
500 1000 1500 2000 2500 3000 3500 4000 45000Sample
minus15
minus10
minus5
0
5
10
15
Figure 19 Mean value control chart for longitudinal displacements of piers
Stra
in (120583
120576)
minus174
minus172
minus170
minus168
minus166
minus164
minus162
1 2714 53
66
79
92
40
118
209
183
222
105
157
196
144
131
235
248
170
Data point
(a) Single train with 8 carriages passed through the bridge
Stra
in (120583
120576)
minus180minus178minus176minus174minus172minus170minus168minus166
1
61
91
31
181
211
121
421
301
481
511
361
391
271
451
241
331
541
571
151
Data point
(b) Single train with 16 carriages passed through the bridge
Figure 20 Typical strain time-history curves of DYB-2 when single train passed through the bridge
data and the spectrum of stress amplitude was obtained[15] The spectra of stress amplitude calculated using thestrain history data with regard to 8 carriages or 16 carriagesare shown in Figure 21 respectively It can be seen that themaximum stress amplitude is smaller than 10MPa obtainedfrom strain history curves under single trainThen accordingtoMinerrsquos law and standard of Eurocode 3 fatigue parametersof 119878eq (equivalent stress amplitude) and 119873
119904(stress cycle
number) under single train were calculated as follows [16]
119878eq = (
1198991198941198785
119894
sum119899119894
)
15
119873s = sum119899119894
(9)
where 119878119894is 119894th stress amplitude 119899
119894is the number of
stress cycles corresponding to 119878119894 and 15 is slope value of
log 119878- log119873 curve in the scope of lower stress amplitudeaccording to the standard of Eurocode 3 [17]
Using (9) the equivalent stress amplitudes 119878eq with regardto 8 carriages or 16 carriages are 075MPa and 071MParespectively And the stress cycle numbers 119873
119904are 22 and 45
respectively Thus the value of 119878eq with regard to 8 carriagesis similar to that of 16 carriages However the value of 119873
119904
with regard to 8 carriages is about twice as much as that of 16carriages Furthermore as recommended by Eurocode 3 forrib-to-deck welded joint the fatigue limit Δ120590
119871is 29MPa [17]
and the value of 119878eq when the high-speed train passes throughbridge is far less than Δ120590
119871 which indicates that fatigue life of
rib-to-deck welded joint in Nanjing Dashengguan Bridge canbe considered infinite Thus the fatigue performance of steel
deck satisfies the requirement of infinite-fatigue-life designmethod
4 Conclusions
In the last few decades Structural Health Monitoring (SHM)system has been widely installed in many highway bridgesto continuously monitor the structural condition over anextended period of time However application of SHMsystem in high-speed railway bridges is far insufficient Thispaper presents an analysis of the SHM system installedin Nanjing Dashengguan Bridge in China The monitoringsystem is composed of a sensor system a data acquisitionand transmission system a data management system and astructural evaluation system The monitoring system is cur-rently in full operation accumulating the bridgersquos behaviorsunder the varying environmental conditions such as high-speed trains and environmental temperature
Based on long-termhealthmonitoring data general prin-ciples and analytical processes for structural evaluation areprovided and some evaluation results are reported It can beshown from the analysis in this paper that the mathematicalcorrelation models describing the overall structural behaviorof the bridge can be obtained with the support of the healthmonitoring system which includes cross-correlation modelsfor accelerations correlation models between temperatureand static strains of steel truss arch and correlation mod-els between temperature and longitudinal displacements ofpiers The mean value control chart is further employed tomonitor the time series of differences between the measuredstructural parameters and the predicted parameters usingthese correlation models The upper and lower control limits
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
The Scientific World Journal 3
Table 1 Sensitivity analysis of bridge parameters for SHM design
Structure Key position Critical member Parameter ImportanceVariation undertemperature
actionSensitivity Decision
Pier Side spans 4 and 10 piers Longitudinaldisplacement Critical 229mm Sensitive Monitor
Main girder In the middle ofthe main span Steel deck Vertical
displacementLess-
important 35mm Less-sensitive No
Truss arch In the middle ofthe main span
Top chordmember diagonalweb member archrib chord memberand bottom chord
member
Static strain Critical 280 120583120576 Sensitive Monitor
Truss arch In the middle ofthe main span
Top chord memberand bottom chord
memberTemperature Critical 60∘C Sensitive Monitor
Main girder In the middle ofthe main span Steel deck Temperature Less-
important 60∘C Less-sensitive No
Whole bridge Frequency Important Nonsensitive Monitor one
Truss arch In the middle ofthe main span Top chord member Wind Less-
important Less-sensitive No
linked to the bridge deck so that the bridge deck is restrictedfrom moving in the longitudinal direction at the 7 pierAnd there have no longitudinal restraints between the bridgedeck and other piers Due to the remarkable characteristics ofNanjingDashengguanBridge including long span of themaingirder heavy design loading and high speed of trains a long-term SHM system was designed and installed on the NanjingDashengguan Bridge shortly after it was opened to railwaytraffic This online SHM system was designed to be a concisemonitoring system with a minimum of 28 sensors to monitorthe key parametersThe entire system consists of a sensor sys-tem a data acquisition system a data transmission system adatamanagement system and a structural evaluation systemThe primary purpose of the system is to monitor in-serviceperformance of the bridge structure under the varying envi-ronmental conditions such as high-speed trains and environ-mental temperature and to provide early warning of abnor-mal changes in in-service performance of the bridge thatmight be unsafe to bridge structureThe procedure of design-ing and implementing the SHM system includes three majorstages as follows (1) the stage of planning and design includesthe consideration of the physical quantities to be measuredthe selection of sensors appropriate for measuring such phys-ical parameters and the optimal placement of the sensors (2)The stage of installation and calibration includes a series ofstatic and dynamic experiments performed to check the oper-ation of the monitoring system after the completion of thebridge (3)The stage of main operation includes the test oper-ation of the monitoring system training of the maintenancecrew and the replacement ofmalfunctioning sensing devices
The design objective of the SHM system is to monitor thestructural health under daily service condition With regardto the bridge site of Nanjing Dashengguan Bridge there arevery lowprobability events of Typhoon and earthquakeThus
the train loading and temperature action are the most impor-tant factors that affect the daily performance of the bridgeDue to the design requirements the structural responses ofthe bridge under train loading are small and conservativeHowever the axial movements and thermal stresses of thebridge under temperature action are considerable due to thelong span of the bridge For instance the maximum thermalstrain and train-induced strain in the steel truss arch areabout 250 120583120576 and 50 120583120576 respectively For bridge conditionassessment it is of paramount importance to characterize nor-mal variability of structural parameters due to temperatureeffectsTherefore in order to control the overall budget of theSHM system without losing major information on structuraldaily behavior the structural parameters to be monitoredin the Nanjing Dashengguan Bridge were determined usingthe structural sensitivity analysis with the finite elementmodel under the designed temperature action as shown inTable 1 Other important parameters including accelerationsof steel deck and fatigue of steel deck were also monitoredConsidering that the displacement amplitudes of steel deckunder the train loading can be derived from the accelerationof steel deck the vertical displacements of steel deck were notchosen to monitor
21 Temperature and Static Strain Monitoring of Truss ArchAs shown in Figure 1(b) temperature and static strain mon-itoring of truss arch is performed at the 1st cross-section inthe middle of the first main span of the bridge Eight FBGtemperature and strain gauges are both installed on the 1stcross-section to simultaneously measure temperature fielddata and static strain field data in the steel truss arch as shownin Figure 2 It can be seen from Figures 2(c)sim2(f) that thesensors were installed at the top chordmember diagonal webmember arch rib chord member and bottom chord member
4 The Scientific World Journal
Downstream Upstream
Horizontal bracing of top chord member
Vertical bracing
Vertical web memberof side truss
Vertical web memberof middle truss
Top plate
Transverse stiffening girder
Horizontal bracing ofarch rib chord member
Vertical web memberof side truss
W1
W2
W3
W4
W5
W6
W7 W8
Y1
Y2
Y3Y4
Y5
Y6
Y7 Y8
15000 15000
68007
12000
2
2
(a) 1st cross-section of steel truss arch
Shanghai
Top chord member
Arch rib chord member
Vertical web member
Beijing
Diagonal web member
Bottom chord member
W1
W2
W3
W4
W5
W6
W7
W8
Y1Y2
Y3Y4
Y5
Y6
Y7
Y8
3
3
4
4
5
5
6
6
(b) 2-2 section of steel truss arch
Downstream Upstream
W1 W2Y1 Y2
1000
1000
850 850
150
150
1400
(c) 3-3 section of top chord member
Downstream Upstream
W3W4Y3Y4
150
150
610
610
1398
(d) 4-4 section of diagonal web member
Downstream Upstream
W5
W6Y5
Y6
150
150
1400
650 650
800
(e) 5-5 section of arch rib chord member
Downstream Upstream
W7 W8Y7 Y8
150
150
1266
1266
1400
1416
(f) 6-6 section of bottom chord member
Figure 2 Layout of temperature and static strain gauges on the steel truss arch (unit mm)
The Scientific World Journal 5
DYB-1 DYB-2
2500mm2500mm
Figure 3 Layout of dynamic strain gauges on the steel deck
Vertical accelerometer Z2Z4Transverse accelerometer Z1Z3
Figure 4 Layout of accelerometers on the steel deck
of the truss arch (where 119882119894denotes 119894th temperature sensor
119894 = 1 2 8 119884119894denotes 119894th strain sensor 119894 = 1 2 8)
The specification of FBG temperature sensors and straingauges is FBG4700 and FBG4100 supplied by China GeokonInstrumentsCo Ltd Sampling frequency of temperature andstatic strain data collection is set to 1Hz
22 Dynamic StrainMonitoring of Steel Deck Dynamic straingauges are installed at the 1st cross-section in the middleof the first main span of the bridge so as to monitor thefatigue performance of orthotropic steel deck as shown inFigure 3 It can be seen that the dynamic strain gauges DYB-1and DYB-2 monitor the long-term fatigue effects of deck-ribwelded detail induced by the traveling high-speed trainsThespecification of dynamic strain gauges is FBG4100 suppliedby China Geokon Instruments Co Ltd Sampling frequencyof dynamic strain data collections is set to 50Hz
23 Acceleration Monitoring of Steel Deck In order to moni-tor the transverse and vertical acceleration responses on thetwo main spans induced by the traveling high-speed trainsone transverse accelerometer and one vertical accelerometerwere installed at the 1st cross-section and 2nd cross-sectionin the middle of two main spans respectively as shown inFigure 4 (where 119885
119894denotes 119894th accelerometer 119894 = 1 2 4)
The specification of accelerometers is DH610 supplied byDonghua Testing Technology Co Ltd Sampling frequencyof acceleration data collections is set to 200Hz
24 Longitudinal Displacement Monitoring of Piers In thedesign of Nanjing Dashengguan Bridge the 7 pier is linkedto the bridge deck so that the bridge deck is restricted frommoving in the longitudinal direction at the 7 pier And therehave no longitudinal restraints between the bridge deck andother piersTherefore 6 displacement transducers are locatedon the piers and used for the measurement of longitudinaldisplacements at the 4 pier 5 pier 6 pier 8 pier 9pier and 10 pier as shown in Figure 5 (where 119871
119894denotes 119894th
displacement transducer 119894 = 1 2 6) The specificationof displacement transducers is CLMD2 supplied by ASMrsquosin Germany Finally it should be noted that the DAQs of theSHM system are located in the 6 and 8 piers
3 Evaluation of Structural Condition UsingLong-Term Monitoring Data
Using long-term monitoring data collected from the SHMsystem of Nanjing Dashengguan Bridge structural conditionassessment involves four aspects evaluation of vibration per-formance of themain girder evaluation of static performanceof steel truss arch evaluation of movement performance ofpier and evaluation of fatigue performance of steel deck
31 Evaluation of Vibration Performance of the Main GirderThe main function for the SHM system is to detect theperformance degradation of structures as early as possibleand to provide essential references for maintenance of actualbridges For Nanjing Dashengguan Bridge the vibration andstrain of the main girder together with the longitudinaldisplacement of piers are three important items to moni-tor Three kinds of performance evaluation are conductedaccordingly Those are evaluation of vibration performanceof the main girder evaluation of static performance of steeltruss arch and evaluation of movement performance of pierFirstly degradation or damage of local members may occurwith the increase of service period which will reduce thestructural stiffness of bridge Thus the dynamic propertyof bridge structure is affected significantly by the structuralstiffness reduction and will cause unusual variations ofthe structural vibration Such unusual behavior will affectrunning safety of the train directly Secondly damage of localmembers may induce redistribution of the internal forces inthe whole bridge under the temperature action The reasonis that Nanjing Dashengguan Bridge belongs to staticallyindeterminate structure Therefore static strain monitoringis necessary for typical members under temperature actions
6 The Scientific World Journal
Beijing Shanghai
4 pier 5 pier 6 pier 7 pier 8 pier 9 pier 10 pier
L1 L2 L3 L4 L5 L6
Figure 5 Layout of displacement transducers on the pier
Acce
lera
tion
(mm
s2)
1 80012001 4001 6001 10001
Data point
minus40
minus30
minus20
minus10
0
10
20
30
40
(a) Accelerometer 1198851
Acce
lera
tion
(mm
s2)
1 80012001 4001 6001 10001
Data point
minus40
minus30
minus20
minus10
0
10
20
30
40
(b) Accelerometer 1198853
Figure 6 Transverse acceleration time histories of the main girder induced by a high-speed train
Thirdly the longitudinal displacement behavior of the bridgepier is crucial for running safety of the train under thetemperature action Once the abnormal variation of the itemis detected inspection may be evolved timely for detection ofthe bridge piers In this section the evaluation of vibrationperformance of main girder is firstly discussed
For high-speed railway bridges vibration performanceof the main girder subjected to the high-speed trains isimportant for passenger comfort and traffic safety [6 7]For instance Zeng et al [8] point out that the essentialreason of train derailment is that the transverse vibrationof the train-railway-bridge system becomes unstable Henceit is necessary to present a strategy for early warning ofdeterioration of vibration performance of the main girderusing long-term health monitoring data
Figures 6 and 7 illustrate the typical transverse and ver-tical acceleration time histories of the main girder measuredfrom accelerometers Z
1simZ4 respectively It can be observed
that when high-speed train passed the bridge the transverseaccelerations in the middle of the first span are much smallerthan that of the second span and the vertical accelerationsin the middle of the first span are much larger than that ofthe second span These indicate that although the structurallayouts of two main spans of the main girder are the samethere exists significant difference between the transverseand vertical vibration characteristics of two main spans due
to the rail irregularity in transverse and vertical directionsThus there is a need to monitor the transverse and verticalaccelerations of two main spans in the long term so as torealize anomaly alarms for vibration performance of themaingirder
Hence in the present study the RMS values of theaccelerations are utilized as the monitoring parameters torepresent the transverse and vertical vibration characteristicof the main girder The cross-correlation between the RMSvalues of the accelerations in the middle of two main spans isfurther investigated when the high-speed trains pass throughthe bridge Figure 8 shows the scatter plots between the RMSvalues of the accelerations and shows the fitted results byusing quadratic polynomial as well
By using the fitting formula constructed in Figure 8scatter plot between theRMSvalues of the accelerations in themiddle of the second span and the corresponding fitted valuesis drawn in Figure 9 where the fitted values were obtainedby taking the RMS values of the accelerations in the middleof the first span into the fitting formula of the quadraticpolynomialThe correlation coefficient of the measured RMSvalues and fitted values on February 1st 2013 is 09391 and0918 for transverse and vertical accelerations respectively asshown in Figure 9 Computed using the same method all thedaily correlation coefficients in 2013 are more than 090 fortransverse and vertical accelerations as shown in Figure 10
The Scientific World Journal 7Ac
cele
ratio
n (m
ms2)
minus150
minus100
minus50
0
50
100
150
200
1 4001 6001 8001 100012001
Data point
(a) Accelerometer 1198852
Acce
lera
tion
(mm
s2)
1 4001 6001 8001 100012001
Data point
minus80
minus60
minus40
minus20
0
20
40
60
(b) Accelerometer 1198854
Figure 7 Vertical acceleration time histories of the main girder caused by a high-speed train
y = minus02642x2 + 37075x minus 56125
53 4 62 25 5535 45
RMS values of the 1st span (mms2)
0
2
4
6
8
10
RMS
valu
es o
f the
2nd
span
(mm
s2)
(a) Transverse accelerations
y = minus00131x2 + 07296x + 35995
50 15 20 2510
RMS values of the 1st span (mms2)
0
2
4
6
8
10
12
14
16RM
S va
lues
of t
he2
nd sp
an (m
ms2)
(b) Vertical accelerations
Figure 8 Cross-correlations between RMS values of accelerations measured in the middle of two main spans
Fitte
d RM
S va
lues
(mm
s2)
R = 09391
0
2
4
6
8
10
63 54 72 25 35 6555 7545
Measured RMS values (mms2)
(a) Transverse accelerations
R = 09518
4
6
8
10
12
14
16
Fitte
d RM
S va
lues
(mm
s2)
86 12 14 1610
Measured RMS values (mms2)
(b) Vertical accelerations
Figure 9 Fitting effect of cross-correlation between RMS values of accelerations by using quadratic polynomial
8 The Scientific World Journal
0 62 9331 186 217124 310279248155
Day
09
091
092
093
094
095
096
097C
orre
latio
n co
effici
ent
(a) Transverse accelerations
0 60 9030 180 210120 300270240150
Day
09
092
094
096
098
Cor
relat
ion
coeffi
cien
t
(b) Vertical accelerations
Figure 10 Daily computed correlation coefficients in 2013
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
s2)
minus2
minus1
0
1
2
20 40 60 80 100 120 140 1600Sample
Figure 11 Mean value control chart for accelerations in the middle of the second span
The analysis results indicate that good cross-correlationexists between the RMS values of the transverse and verticalaccelerations on the two main spans Multivariable controlchart is used here to monitor the measured changes in thetransverse and vertical accelerations caused by deteriorationof the vibration performance Firstly the condition index 119890
for early warning of abnormal vibration behavior is definedas the difference between the measured and predicted RMSvalues of the accelerations in the middle of the second span
119890 = RMS119898minus RMS
119901 (1)
where RMS119898is themeasured RMS values of the accelerations
in the middle of the second span RMS119901is the predicted
RMS values of the accelerations in the middle of the secondspan which is obtained by taking the RMS values of theaccelerations in the middle of the first span into the fittingformula of the quadratic polynomial in the healthy conditionThen a mean value control chart is employed to monitorthe time series of 119890 with regard to both transverse andvertical accelerations The time series of 119890 taken when thestructural vibration is in good condition will have somedistribution with mean 120583 and variance 120590
2 If the mean andstandard deviation are known a control chart is constructedby drawing a horizontal line CL at 120583 and twomore horizontallines representing the upper and lower control limits Theupper limit UCL is drawn at 120583 + 119896120590 and the lower limit LCL
at 120583 minus 119896120590 For online monitoring the controlling parameter119896 is chosen so that when the structural vibration is in goodcondition all observation samples fall between the controllimits When the new measurement is made the structuralabnormal vibration condition can be detected if an unusualnumber of samples fall beyond the control limits Figure 11shows the mean value control chart using the measuredand predicted RMS values of the accelerations shown inFigure 9 The upper and lower limits are determined usingthe controlling parameter 119896 = 2612 Hence a long-termmonitoring on the cross-correlation model of RMS values ofthe transverse and vertical accelerations can realize the earlywarning of deterioration of the vibration performance
32 Evaluation of Static Performance of Steel Truss Arch
321 Data of Temperature Field and Static Strains Tem-perature data from temperature sensor 119882
119894is denoted by
119879119894 temperature difference data acquired by 119879
119894minus 119879
119895is
denoted by119879119894119895(119879119894119895= 119879119894minus119879119895) and static strain data from strain
sensor119884119894is denoted by 119878
119894(119894 119895 = 1 2 8) As for illustration
the time-history curves of annual change and daily change intemperature 119879
1and temperature difference 119879
12are shown in
Figures 12 and 13 respectively And the time-history curvesof annual change and daily change in static strain data 119878
1
are shown in Figure 14 (negative values denote compressive
The Scientific World Journal 9
4 5 6 7 8 9 10 113Time (month)
minus5
5
15
25
35
45Te
mpe
ratu
re (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
5
10
15
20
Tem
pera
ture
(∘C)
(b) Daily curve (March 4th 2013)
Figure 12 Time-history curves of temperature data 1198791
4 5 6 7 8 9 10 113Time (month)
minus25
minus15
minus5
5
15
Tem
pera
ture
diff
eren
ce (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
minus10
minus5
0
5Te
mpe
ratu
re (∘
C)
(b) Daily curve (March 4th 2013)
Figure 13 Time-history curves of temperature difference data 11987912
4 5 6 7 8 9 10 113Time (month)
minus300
minus200
minus100
0
100
200
Stat
ic st
rain
(120583120576)
(a) Annual curve fromMarch 2013 to October 2013
Sharp peak
4 8 12 16 20 240Time (h)
minus120
minus65
minus10
45
100
Stat
ic st
rain
(120583120576)
(b) Daily curve (March 4th 2013)
Figure 14 Time-history curves of static strain data 1198781
10 The Scientific World Journal
4 8 12 16 20 240Time (h)
Stat
ic st
rain
S III
(120583120576)
minus120
minus65
minus10
45
100
(a) Daily extraction result 119878III of 1198781
4 5 6 7 8 9 10 113Time (month)
Stat
ic st
rain
S III
(120583120576)
minus200
minus130
minus60
10
80
150
(b) Annual extraction result 119878III of 1198781 fromMarch 2013 to October 2013
Figure 15 Extraction results 119878III of static strain data 1198781
static strain and positive values denote tension static strain)From Figure 14 it can be seen that time-history curve of 119878
1
contains many sharp peaks caused by high-speed trains witheach peak corresponding to the time when one train passesthrough the bridge And the static strains caused by high-speed trains are obviously lower than that of temperature fieldincluding temperature data and temperature difference data
As shown in Figure 14 static strain data in steel truss archmainly contains three parts static strain 119878I caused by tem-perature static strain 119878II caused by temperature differenceand static strain 119878III caused by train In order to construct thecorrelation between temperature field and static strain in steeltruss arch static strain 119878I and static strain 119878II (denoted by 119878III)must be extracted from the static strain data In the presentstudy wavelet packet decomposition method is used toextract 119878III considering the high-frequency characteristics ofstatic strain data caused by trains By using the wavelet packetdecomposition static strain data can be decomposed scale byscale into different frequency bands and each decompositioncoefficient corresponds to its frequency band [9 10] Eachdecomposition coefficient can be reconstructed into time-domain signals with constraints of its own frequency bandTaking the daily time-history curve of static strain data1198781shown in Figure 14(b) as an example it is decomposed
within 8 scales and the decomposition coefficient 11986208
119895
isspecially selected and then reconstructed as 119878III shown inFigure 15(a) It can be seen that sharp peaks caused by trainsare removed effectively and the static strains which are causedby temperature field are retained as well Furthermore theannual extraction result 119878III of 1198781 fromMarch 2013 toOctober2013 is shown in Figure 15(b)
322 Correlation Model between Temperature Field andStatic Strains In constructing the correlationmodel betweentemperature field and static strains data from March 2013to October 2013 are chosen as training data and data inNovember 2013 are chosen as test data for verification of
the correlation model The static strain 119878III is expressed asfollows
119878III =8
sum
119894=1
120572119894119879119894+
16
sum
119895=1
120573119895119863119895+ 119888 (2)
where 119879119894is 119894th temperature data from set 119879
1 1198792 1198793 1198794
1198795 1198796 1198797 1198798 119863119895is 119895th temperature difference data from set
11987912 11987934 11987956 11987978 11987913 11987915 11987917 11987935 11987937 11987957 11987924 11987926 11987928 11987946
11987948 11987968 120572119894is 119894th linear expansion coefficient of 119879
119894 120573119895is 119895th
linear expansion coefficient of 119879119894119895 and 119888 is the constant term
Considering that total 24 linear expansion coefficientsand one constant term in (2) are very complicated forcalculation principal component analysis is further usedto simplify (2) which aims at finding small amounts ofcomprehensive factors to represent main information of tem-perature and temperature difference data By using principalcomponent analysis the comprehensive factors can be foundby transforming temperature and temperature difference datainto factor component space [11 12] For temperature data theexplained variance of the first potential comprehensive factor1198751reaches 967 apparently larger than the other potential
comprehensive factors Thus 1198751is selected as the compre-
hensive factor Meanwhile for temperature difference datausing principal component analysis the explained varianceof the first three potential comprehensive factors 119877
1 1198772 and
1198773reaches 988 which are selected as the comprehensive
factorsTherefore the correlation models between 119878III and tem-
perature field can be expressed as follows
119878III = 1205821([119864]1times8
997888
1198798times1
) +9978881205741times3
([119865]3times16
997888
11986316times1
) + 119888 (3)
where997888
1198798times1
denotes one vector containing eight temperaturedata 997888119863
16times1denotes one vector containing sixteen temper-
ature difference data 119863119895(119895 = 1 2 16) [119864]
1times8and
The Scientific World Journal 11
b = minus3632
k = 0946
minus100 minus55 minus10 8035Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus100
minus55
minus10
35
80
Sim
ulat
iveS
III
(120583120576)
(a) 1198781
b = minus2013k = 0898
minus45minus75 minus15 45 7515Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus75
minus45
minus15
15
45
75
Sim
ulat
iveS
III
(120583120576)
(b) 1198782
b = minus1844k = 0904
minus60
minus30
0
30
60
90
Sim
ulat
iveS
III
(120583120576)
minus30minus60 30 60 900Monitoring SIII (120583120576)
Scattered pointsFitting curve
(c) 1198783
b = 2057k = 1126
minus15minus30 15 300Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus40
minus25
minus10
5
20
35Si
mul
ativ
eSII
I(120583120576)
(d) 1198784
b = minus4139k = 0935
minus100
minus65
minus30
5
40
75
Sim
ulat
iveS
III
(120583120576)
minus105 4515 75minus45minus75 minus15
Monitoring SIII (120583120576)
Scattered pointsFitting curve
(e) 1198785
b = minus4094k = 0918
minus145minus205 minus25minus85 9535Monitoring SIII (120583120576)
minus185
minus130
minus75
minus20
35
90
Sim
ulat
iveS
III
(120583120576)
Scattered pointsFitting curve
(f) 1198786
Figure 16 Continued
12 The Scientific World Journal
b = 2148k = 0881
minus50
minus28
minus6
16
38
60Si
mul
ativ
eSII
I(120583120576)
minus45 minus23 minus1 43 6521Monitoring SIII (120583120576)
Scattered pointsFitting curve
(g) 1198787
b = 0615k = 1118
minus30
minus15
0
15
30
Sim
ulat
iveS
III
(120583120576)
minus15minus25 minus5 15 255Monitoring SIII (120583120576)
Scattered pointsFitting curve
(h) 1198788
Figure 16 Correlation between simulative 119878III and monitoring 119878III
Table 2 Performance parameters 1205821
1205741
1205742
1205743
and 119862 in correlationmodels
Static straindata 120582
1
1205741
1205742
1205743
119862
119878III of 1198781 0454 minus4711 minus2597 minus4944 minus46806119878III of 1198782 1585 0013 4104 minus1247 minus70238119878III of 1198783 0301 3416 minus2098 2968 minus6884119878III of 1198784 0350 minus0734 0351 0948 minus23597119878III of 1198785 0179 minus4219 1827 3165 minus22049119878III of 1198786 0062 minus3245 8242 minus5398 minus22862119878III of 1198787 minus0367 2411 minus0598 5531 17075119878III of 1198788 0611 2188 minus1130 2018 minus32799
[119865]3times16
denote transformation matrix which can be cal-culated using principal component analysis [11 12] 120582
1is
performance parameter of first potential comprehensivefactor 119875
1 9978881205741times3
= [1205741 1205742 1205743] is one vector containing
three performance parameters corresponding to first threepotential comprehensive factors 119877
1 1198772 and 119877
3 The values of
performance parameters 1205821 9978881205741times3
and 119888 are estimated usingtraining data by multivariate linear regression method [13]and are shown in Table 2
In order to verify the effectiveness of the correlationmodels shown in (2) test data in November 2013 are usedto obtain the simulative 119878III The scattered points betweensimulative 119878III and monitoring 119878III are plotted as shown inFigure 16 which are further fitted by linear function
119878119904= 119896119878119898+ 119887 (4)
where 119878119904denotes simulative 119878III 119878119898 denotes monitoring
119878III and 119896 119887 are fitting parameters with the fitting resultsshown in Figure 16 as well The fitting results verify theeffectiveness of the correlationmodels Amultivariable mean
value control chart is also employed tomonitor the measuredabnormal changes in the static strains of the steel truss archThe condition index 119890 for early warning of deterioration ofthe static performance of steel truss arch is defined as thedifference between the measured and predicted static strains
119890 = 119878119898minus 119878119901 (5)
where 119878119898is the measured static strain 119878III 119878119901 is the predicted
static strain 119878III which is obtained by using the correlationmodels shown in (2) in the healthy condition Figure 17 showsthemean value control chart for static strains of the steel trussarch using measurement data in November 2013 The upperand lower limits are determined using the controlling param-eter 119896 = 2235 Hence a long-term monitoring on the corre-lation models between temperature field and static strains issuitable for early warning of deterioration of the static per-formance of steel truss arch
33 Evaluation of Movement Performance of Pier Longi-tudinal displacements of piers and temperature field datain steel truss arch collected from March 2013 to October2013 are used for evaluation of movement performanceof piers Longitudinal displacement data from sensors 119871
119894
(119894 = 1 2 6) are denoted by 119889119894(119905) (where 119905 means time)
respectively Meanwhile temperature field data from sensors119882119894(119894 = 1 2 8) are denoted by 119879
119894(119905) respectively and
their averaged value 119879(119905) is used to represent the averagetemperature of the steel truss arch For simplicity taking10 minutes as basic time interval the 10 min average dis-placements and temperatures are calculated and used as therepresentative values of this time interval By this way thedisplacements and temperatures in one day are reduced to144 representative samples Furthermore the temperature-displacement scatter diagrams are plotted in Figure 18 Anoverall increase in longitudinal displacements of piers isobserved with an increase in the average temperature of the
The Scientific World Journal 13
UCL
LCL
CL
Con
ditio
n in
dexe
(120583120576)
minus15
minus10
minus5
0
5
10
15
10 20 30 40 50 600Sample
Figure 17 Mean value control chart for static strains of the steel truss arch
Table 3 Summary of linear regression models
Pier Regression function4 pier 119889
1
= minus11524 + 696119879
5 pier 1198892
= minus9253 + 562119879
6 pier 1198893
= minus6628 + 382119879
8 pier 1198894
= minus4658 + 336119879
9 pier 1198895
= minus7605 + 530119879
10 pier 1198896
= minus9850 + 664119879
bridge And the correlation coefficients of the longitudinaldisplacements of piers and average temperature of the steeltruss arch are all larger than 092 as shown in Figure 18
A linear regression analysis between the 10 min averagedisplacement 119889
119894(119905) and the temperature 119879(119905) is further per-
formed to model the temperature-displacement correlationsusing the following equation [14]
119889119894(119905) = 120573
0+ 1205731119879 (119905) (6)
where the regression coefficients 1205731and 120573
0were obtained by
the least-squares method as
1205731=
119878119889119894119879
119878119879119879
1205730= 119889119894minus 1205731119879
(7)
where 119878119889119894119879
is the covariance between the displacement andthe temperature sequences 119878
119879119879is the variance of the mea-
sured temperature sequences and 119879 and 119889119894are the means
of the measured temperature and displacement sequencesrespectively Table 3 summarizes the expressions of linearregression models of the displacement versus temperature
The analysis results indicate that good correlation existsbetween the longitudinal displacements of piers and averagetemperature of steel truss arch A multivariable mean valuecontrol chart is also employed to monitor the measuredabnormal changes in the longitudinal displacements of piersThe condition index 119890 for early warning of abnormal behaviorof piers is defined as the difference between themeasured andpredicted longitudinal displacements
119890 = 119889119898minus 119889119901 (8)
where 119889119898is the measured displacement of the piers 119889
119901is
the predicted displacement of the piers which is obtainedby using the linear regression models shown in Table 3in the healthy condition Figure 19 shows the mean valuecontrol chart for longitudinal displacements of piers usingmeasurement data in March 2013 The upper and lowerlimits are determined using the controlling parameter 119896 =
2804 Hence a long-term monitoring on the temperature-displacement correlation model is suitable for real-timecondition assessment for movement performance of piers
34 Evaluation of Fatigue Performance of Steel Deck Fatiguecracking is universal for orthotropic steel deck of highwaybridges Once the fatigue crack appears extensive fatiguecracking may occur and repair is very difficult Henceinfinite-fatigue-life design method is applied for NanjingDashengguan Bridge instead of finite-fatigue-life designmethod which is applied extensively in highway bridges Theinfinite-fatigue-life design method is to control the fatiguestress amplitude in the range less than the value of fatiguecut-off limit refrained from fatigue cracking Consequentlythe purpose of equipping stain gauges in the SHM system ofDashengguan Bridge is to monitor the stress amplitude andthe validity of fatigue design can be achieved by comparingthe stress amplitude with fatigue stress limit
The structural dynamic response is asymmetrical becauseof the irregularity of railway and braking force from thetrain even though the structure of Nanjing DashengguanBridge is spatially symmetrical Hence the stresses are notsymmetrical measured by strain gauges from the upstreamand downstream directions However the fatigue amplitudesof two strain gauges induced by trains are both extremelysmall Thus the strain history data of the strain gauge DYB-2are merely selected in this paper to illustrate that the fatiguestress amplitude is much lower than the value of fatigue stresslimit The typical strain time-history curves of DYB-2 whensingle train with 8 carriages or 16 carriages passed throughthe bridge are shown in Figure 20 respectively It can be seenthat during the passing of high-speed trains both 8 carriagesand 16 carriages have led to strain variations reaching to 2ndash8 120583120576 And the number of strain peaks is 16 and 32 for trainswith 8 carriages and 16 carriages respectively because theaxle number of each carriage is 2
To perform fatigue evaluation simplified rainflow cyclecounting algorithm was firstly used to process strain history
14 The Scientific World Journal
R = 09748
10 20 30 40 500
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
(a) Correlation between 1198891of the 4 pier and average temperature 119879 of
steel truss arch
10 20 30 40 500
Temperature (∘C)
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
R = 09746
(b) Correlation between 1198892of the 5 pier and average temperature 119879 of
steel truss arch
10 200 40 5030
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
Disp
lace
men
t (m
m)
R = 09471
(c) Correlation between 1198893of the 6 pier and average temperature 119879 of
steel truss arch
R = 09276
10 20 30 40 500
Temperature (∘C)
minus60
minus30
0
30
60
90
120D
ispla
cem
ent (
mm
)
(d) Correlation between 1198894of the 8 pier and average temperature 119879 of
steel truss arch
R = 09536
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(e) Correlation between 1198895of the 9 pier and average temperature 119879 of
steel truss arch
R = 09537
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(f) Correlation between 1198896of the 10 pier and average temperature 119879 of
steel truss arch
Figure 18 Correlation between longitudinal displacement and average temperature
The Scientific World Journal 15
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
)
500 1000 1500 2000 2500 3000 3500 4000 45000Sample
minus15
minus10
minus5
0
5
10
15
Figure 19 Mean value control chart for longitudinal displacements of piers
Stra
in (120583
120576)
minus174
minus172
minus170
minus168
minus166
minus164
minus162
1 2714 53
66
79
92
40
118
209
183
222
105
157
196
144
131
235
248
170
Data point
(a) Single train with 8 carriages passed through the bridge
Stra
in (120583
120576)
minus180minus178minus176minus174minus172minus170minus168minus166
1
61
91
31
181
211
121
421
301
481
511
361
391
271
451
241
331
541
571
151
Data point
(b) Single train with 16 carriages passed through the bridge
Figure 20 Typical strain time-history curves of DYB-2 when single train passed through the bridge
data and the spectrum of stress amplitude was obtained[15] The spectra of stress amplitude calculated using thestrain history data with regard to 8 carriages or 16 carriagesare shown in Figure 21 respectively It can be seen that themaximum stress amplitude is smaller than 10MPa obtainedfrom strain history curves under single trainThen accordingtoMinerrsquos law and standard of Eurocode 3 fatigue parametersof 119878eq (equivalent stress amplitude) and 119873
119904(stress cycle
number) under single train were calculated as follows [16]
119878eq = (
1198991198941198785
119894
sum119899119894
)
15
119873s = sum119899119894
(9)
where 119878119894is 119894th stress amplitude 119899
119894is the number of
stress cycles corresponding to 119878119894 and 15 is slope value of
log 119878- log119873 curve in the scope of lower stress amplitudeaccording to the standard of Eurocode 3 [17]
Using (9) the equivalent stress amplitudes 119878eq with regardto 8 carriages or 16 carriages are 075MPa and 071MParespectively And the stress cycle numbers 119873
119904are 22 and 45
respectively Thus the value of 119878eq with regard to 8 carriagesis similar to that of 16 carriages However the value of 119873
119904
with regard to 8 carriages is about twice as much as that of 16carriages Furthermore as recommended by Eurocode 3 forrib-to-deck welded joint the fatigue limit Δ120590
119871is 29MPa [17]
and the value of 119878eq when the high-speed train passes throughbridge is far less than Δ120590
119871 which indicates that fatigue life of
rib-to-deck welded joint in Nanjing Dashengguan Bridge canbe considered infinite Thus the fatigue performance of steel
deck satisfies the requirement of infinite-fatigue-life designmethod
4 Conclusions
In the last few decades Structural Health Monitoring (SHM)system has been widely installed in many highway bridgesto continuously monitor the structural condition over anextended period of time However application of SHMsystem in high-speed railway bridges is far insufficient Thispaper presents an analysis of the SHM system installedin Nanjing Dashengguan Bridge in China The monitoringsystem is composed of a sensor system a data acquisitionand transmission system a data management system and astructural evaluation system The monitoring system is cur-rently in full operation accumulating the bridgersquos behaviorsunder the varying environmental conditions such as high-speed trains and environmental temperature
Based on long-termhealthmonitoring data general prin-ciples and analytical processes for structural evaluation areprovided and some evaluation results are reported It can beshown from the analysis in this paper that the mathematicalcorrelation models describing the overall structural behaviorof the bridge can be obtained with the support of the healthmonitoring system which includes cross-correlation modelsfor accelerations correlation models between temperatureand static strains of steel truss arch and correlation mod-els between temperature and longitudinal displacements ofpiers The mean value control chart is further employed tomonitor the time series of differences between the measuredstructural parameters and the predicted parameters usingthese correlation models The upper and lower control limits
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
4 The Scientific World Journal
Downstream Upstream
Horizontal bracing of top chord member
Vertical bracing
Vertical web memberof side truss
Vertical web memberof middle truss
Top plate
Transverse stiffening girder
Horizontal bracing ofarch rib chord member
Vertical web memberof side truss
W1
W2
W3
W4
W5
W6
W7 W8
Y1
Y2
Y3Y4
Y5
Y6
Y7 Y8
15000 15000
68007
12000
2
2
(a) 1st cross-section of steel truss arch
Shanghai
Top chord member
Arch rib chord member
Vertical web member
Beijing
Diagonal web member
Bottom chord member
W1
W2
W3
W4
W5
W6
W7
W8
Y1Y2
Y3Y4
Y5
Y6
Y7
Y8
3
3
4
4
5
5
6
6
(b) 2-2 section of steel truss arch
Downstream Upstream
W1 W2Y1 Y2
1000
1000
850 850
150
150
1400
(c) 3-3 section of top chord member
Downstream Upstream
W3W4Y3Y4
150
150
610
610
1398
(d) 4-4 section of diagonal web member
Downstream Upstream
W5
W6Y5
Y6
150
150
1400
650 650
800
(e) 5-5 section of arch rib chord member
Downstream Upstream
W7 W8Y7 Y8
150
150
1266
1266
1400
1416
(f) 6-6 section of bottom chord member
Figure 2 Layout of temperature and static strain gauges on the steel truss arch (unit mm)
The Scientific World Journal 5
DYB-1 DYB-2
2500mm2500mm
Figure 3 Layout of dynamic strain gauges on the steel deck
Vertical accelerometer Z2Z4Transverse accelerometer Z1Z3
Figure 4 Layout of accelerometers on the steel deck
of the truss arch (where 119882119894denotes 119894th temperature sensor
119894 = 1 2 8 119884119894denotes 119894th strain sensor 119894 = 1 2 8)
The specification of FBG temperature sensors and straingauges is FBG4700 and FBG4100 supplied by China GeokonInstrumentsCo Ltd Sampling frequency of temperature andstatic strain data collection is set to 1Hz
22 Dynamic StrainMonitoring of Steel Deck Dynamic straingauges are installed at the 1st cross-section in the middleof the first main span of the bridge so as to monitor thefatigue performance of orthotropic steel deck as shown inFigure 3 It can be seen that the dynamic strain gauges DYB-1and DYB-2 monitor the long-term fatigue effects of deck-ribwelded detail induced by the traveling high-speed trainsThespecification of dynamic strain gauges is FBG4100 suppliedby China Geokon Instruments Co Ltd Sampling frequencyof dynamic strain data collections is set to 50Hz
23 Acceleration Monitoring of Steel Deck In order to moni-tor the transverse and vertical acceleration responses on thetwo main spans induced by the traveling high-speed trainsone transverse accelerometer and one vertical accelerometerwere installed at the 1st cross-section and 2nd cross-sectionin the middle of two main spans respectively as shown inFigure 4 (where 119885
119894denotes 119894th accelerometer 119894 = 1 2 4)
The specification of accelerometers is DH610 supplied byDonghua Testing Technology Co Ltd Sampling frequencyof acceleration data collections is set to 200Hz
24 Longitudinal Displacement Monitoring of Piers In thedesign of Nanjing Dashengguan Bridge the 7 pier is linkedto the bridge deck so that the bridge deck is restricted frommoving in the longitudinal direction at the 7 pier And therehave no longitudinal restraints between the bridge deck andother piersTherefore 6 displacement transducers are locatedon the piers and used for the measurement of longitudinaldisplacements at the 4 pier 5 pier 6 pier 8 pier 9pier and 10 pier as shown in Figure 5 (where 119871
119894denotes 119894th
displacement transducer 119894 = 1 2 6) The specificationof displacement transducers is CLMD2 supplied by ASMrsquosin Germany Finally it should be noted that the DAQs of theSHM system are located in the 6 and 8 piers
3 Evaluation of Structural Condition UsingLong-Term Monitoring Data
Using long-term monitoring data collected from the SHMsystem of Nanjing Dashengguan Bridge structural conditionassessment involves four aspects evaluation of vibration per-formance of themain girder evaluation of static performanceof steel truss arch evaluation of movement performance ofpier and evaluation of fatigue performance of steel deck
31 Evaluation of Vibration Performance of the Main GirderThe main function for the SHM system is to detect theperformance degradation of structures as early as possibleand to provide essential references for maintenance of actualbridges For Nanjing Dashengguan Bridge the vibration andstrain of the main girder together with the longitudinaldisplacement of piers are three important items to moni-tor Three kinds of performance evaluation are conductedaccordingly Those are evaluation of vibration performanceof the main girder evaluation of static performance of steeltruss arch and evaluation of movement performance of pierFirstly degradation or damage of local members may occurwith the increase of service period which will reduce thestructural stiffness of bridge Thus the dynamic propertyof bridge structure is affected significantly by the structuralstiffness reduction and will cause unusual variations ofthe structural vibration Such unusual behavior will affectrunning safety of the train directly Secondly damage of localmembers may induce redistribution of the internal forces inthe whole bridge under the temperature action The reasonis that Nanjing Dashengguan Bridge belongs to staticallyindeterminate structure Therefore static strain monitoringis necessary for typical members under temperature actions
6 The Scientific World Journal
Beijing Shanghai
4 pier 5 pier 6 pier 7 pier 8 pier 9 pier 10 pier
L1 L2 L3 L4 L5 L6
Figure 5 Layout of displacement transducers on the pier
Acce
lera
tion
(mm
s2)
1 80012001 4001 6001 10001
Data point
minus40
minus30
minus20
minus10
0
10
20
30
40
(a) Accelerometer 1198851
Acce
lera
tion
(mm
s2)
1 80012001 4001 6001 10001
Data point
minus40
minus30
minus20
minus10
0
10
20
30
40
(b) Accelerometer 1198853
Figure 6 Transverse acceleration time histories of the main girder induced by a high-speed train
Thirdly the longitudinal displacement behavior of the bridgepier is crucial for running safety of the train under thetemperature action Once the abnormal variation of the itemis detected inspection may be evolved timely for detection ofthe bridge piers In this section the evaluation of vibrationperformance of main girder is firstly discussed
For high-speed railway bridges vibration performanceof the main girder subjected to the high-speed trains isimportant for passenger comfort and traffic safety [6 7]For instance Zeng et al [8] point out that the essentialreason of train derailment is that the transverse vibrationof the train-railway-bridge system becomes unstable Henceit is necessary to present a strategy for early warning ofdeterioration of vibration performance of the main girderusing long-term health monitoring data
Figures 6 and 7 illustrate the typical transverse and ver-tical acceleration time histories of the main girder measuredfrom accelerometers Z
1simZ4 respectively It can be observed
that when high-speed train passed the bridge the transverseaccelerations in the middle of the first span are much smallerthan that of the second span and the vertical accelerationsin the middle of the first span are much larger than that ofthe second span These indicate that although the structurallayouts of two main spans of the main girder are the samethere exists significant difference between the transverseand vertical vibration characteristics of two main spans due
to the rail irregularity in transverse and vertical directionsThus there is a need to monitor the transverse and verticalaccelerations of two main spans in the long term so as torealize anomaly alarms for vibration performance of themaingirder
Hence in the present study the RMS values of theaccelerations are utilized as the monitoring parameters torepresent the transverse and vertical vibration characteristicof the main girder The cross-correlation between the RMSvalues of the accelerations in the middle of two main spans isfurther investigated when the high-speed trains pass throughthe bridge Figure 8 shows the scatter plots between the RMSvalues of the accelerations and shows the fitted results byusing quadratic polynomial as well
By using the fitting formula constructed in Figure 8scatter plot between theRMSvalues of the accelerations in themiddle of the second span and the corresponding fitted valuesis drawn in Figure 9 where the fitted values were obtainedby taking the RMS values of the accelerations in the middleof the first span into the fitting formula of the quadraticpolynomialThe correlation coefficient of the measured RMSvalues and fitted values on February 1st 2013 is 09391 and0918 for transverse and vertical accelerations respectively asshown in Figure 9 Computed using the same method all thedaily correlation coefficients in 2013 are more than 090 fortransverse and vertical accelerations as shown in Figure 10
The Scientific World Journal 7Ac
cele
ratio
n (m
ms2)
minus150
minus100
minus50
0
50
100
150
200
1 4001 6001 8001 100012001
Data point
(a) Accelerometer 1198852
Acce
lera
tion
(mm
s2)
1 4001 6001 8001 100012001
Data point
minus80
minus60
minus40
minus20
0
20
40
60
(b) Accelerometer 1198854
Figure 7 Vertical acceleration time histories of the main girder caused by a high-speed train
y = minus02642x2 + 37075x minus 56125
53 4 62 25 5535 45
RMS values of the 1st span (mms2)
0
2
4
6
8
10
RMS
valu
es o
f the
2nd
span
(mm
s2)
(a) Transverse accelerations
y = minus00131x2 + 07296x + 35995
50 15 20 2510
RMS values of the 1st span (mms2)
0
2
4
6
8
10
12
14
16RM
S va
lues
of t
he2
nd sp
an (m
ms2)
(b) Vertical accelerations
Figure 8 Cross-correlations between RMS values of accelerations measured in the middle of two main spans
Fitte
d RM
S va
lues
(mm
s2)
R = 09391
0
2
4
6
8
10
63 54 72 25 35 6555 7545
Measured RMS values (mms2)
(a) Transverse accelerations
R = 09518
4
6
8
10
12
14
16
Fitte
d RM
S va
lues
(mm
s2)
86 12 14 1610
Measured RMS values (mms2)
(b) Vertical accelerations
Figure 9 Fitting effect of cross-correlation between RMS values of accelerations by using quadratic polynomial
8 The Scientific World Journal
0 62 9331 186 217124 310279248155
Day
09
091
092
093
094
095
096
097C
orre
latio
n co
effici
ent
(a) Transverse accelerations
0 60 9030 180 210120 300270240150
Day
09
092
094
096
098
Cor
relat
ion
coeffi
cien
t
(b) Vertical accelerations
Figure 10 Daily computed correlation coefficients in 2013
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
s2)
minus2
minus1
0
1
2
20 40 60 80 100 120 140 1600Sample
Figure 11 Mean value control chart for accelerations in the middle of the second span
The analysis results indicate that good cross-correlationexists between the RMS values of the transverse and verticalaccelerations on the two main spans Multivariable controlchart is used here to monitor the measured changes in thetransverse and vertical accelerations caused by deteriorationof the vibration performance Firstly the condition index 119890
for early warning of abnormal vibration behavior is definedas the difference between the measured and predicted RMSvalues of the accelerations in the middle of the second span
119890 = RMS119898minus RMS
119901 (1)
where RMS119898is themeasured RMS values of the accelerations
in the middle of the second span RMS119901is the predicted
RMS values of the accelerations in the middle of the secondspan which is obtained by taking the RMS values of theaccelerations in the middle of the first span into the fittingformula of the quadratic polynomial in the healthy conditionThen a mean value control chart is employed to monitorthe time series of 119890 with regard to both transverse andvertical accelerations The time series of 119890 taken when thestructural vibration is in good condition will have somedistribution with mean 120583 and variance 120590
2 If the mean andstandard deviation are known a control chart is constructedby drawing a horizontal line CL at 120583 and twomore horizontallines representing the upper and lower control limits Theupper limit UCL is drawn at 120583 + 119896120590 and the lower limit LCL
at 120583 minus 119896120590 For online monitoring the controlling parameter119896 is chosen so that when the structural vibration is in goodcondition all observation samples fall between the controllimits When the new measurement is made the structuralabnormal vibration condition can be detected if an unusualnumber of samples fall beyond the control limits Figure 11shows the mean value control chart using the measuredand predicted RMS values of the accelerations shown inFigure 9 The upper and lower limits are determined usingthe controlling parameter 119896 = 2612 Hence a long-termmonitoring on the cross-correlation model of RMS values ofthe transverse and vertical accelerations can realize the earlywarning of deterioration of the vibration performance
32 Evaluation of Static Performance of Steel Truss Arch
321 Data of Temperature Field and Static Strains Tem-perature data from temperature sensor 119882
119894is denoted by
119879119894 temperature difference data acquired by 119879
119894minus 119879
119895is
denoted by119879119894119895(119879119894119895= 119879119894minus119879119895) and static strain data from strain
sensor119884119894is denoted by 119878
119894(119894 119895 = 1 2 8) As for illustration
the time-history curves of annual change and daily change intemperature 119879
1and temperature difference 119879
12are shown in
Figures 12 and 13 respectively And the time-history curvesof annual change and daily change in static strain data 119878
1
are shown in Figure 14 (negative values denote compressive
The Scientific World Journal 9
4 5 6 7 8 9 10 113Time (month)
minus5
5
15
25
35
45Te
mpe
ratu
re (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
5
10
15
20
Tem
pera
ture
(∘C)
(b) Daily curve (March 4th 2013)
Figure 12 Time-history curves of temperature data 1198791
4 5 6 7 8 9 10 113Time (month)
minus25
minus15
minus5
5
15
Tem
pera
ture
diff
eren
ce (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
minus10
minus5
0
5Te
mpe
ratu
re (∘
C)
(b) Daily curve (March 4th 2013)
Figure 13 Time-history curves of temperature difference data 11987912
4 5 6 7 8 9 10 113Time (month)
minus300
minus200
minus100
0
100
200
Stat
ic st
rain
(120583120576)
(a) Annual curve fromMarch 2013 to October 2013
Sharp peak
4 8 12 16 20 240Time (h)
minus120
minus65
minus10
45
100
Stat
ic st
rain
(120583120576)
(b) Daily curve (March 4th 2013)
Figure 14 Time-history curves of static strain data 1198781
10 The Scientific World Journal
4 8 12 16 20 240Time (h)
Stat
ic st
rain
S III
(120583120576)
minus120
minus65
minus10
45
100
(a) Daily extraction result 119878III of 1198781
4 5 6 7 8 9 10 113Time (month)
Stat
ic st
rain
S III
(120583120576)
minus200
minus130
minus60
10
80
150
(b) Annual extraction result 119878III of 1198781 fromMarch 2013 to October 2013
Figure 15 Extraction results 119878III of static strain data 1198781
static strain and positive values denote tension static strain)From Figure 14 it can be seen that time-history curve of 119878
1
contains many sharp peaks caused by high-speed trains witheach peak corresponding to the time when one train passesthrough the bridge And the static strains caused by high-speed trains are obviously lower than that of temperature fieldincluding temperature data and temperature difference data
As shown in Figure 14 static strain data in steel truss archmainly contains three parts static strain 119878I caused by tem-perature static strain 119878II caused by temperature differenceand static strain 119878III caused by train In order to construct thecorrelation between temperature field and static strain in steeltruss arch static strain 119878I and static strain 119878II (denoted by 119878III)must be extracted from the static strain data In the presentstudy wavelet packet decomposition method is used toextract 119878III considering the high-frequency characteristics ofstatic strain data caused by trains By using the wavelet packetdecomposition static strain data can be decomposed scale byscale into different frequency bands and each decompositioncoefficient corresponds to its frequency band [9 10] Eachdecomposition coefficient can be reconstructed into time-domain signals with constraints of its own frequency bandTaking the daily time-history curve of static strain data1198781shown in Figure 14(b) as an example it is decomposed
within 8 scales and the decomposition coefficient 11986208
119895
isspecially selected and then reconstructed as 119878III shown inFigure 15(a) It can be seen that sharp peaks caused by trainsare removed effectively and the static strains which are causedby temperature field are retained as well Furthermore theannual extraction result 119878III of 1198781 fromMarch 2013 toOctober2013 is shown in Figure 15(b)
322 Correlation Model between Temperature Field andStatic Strains In constructing the correlationmodel betweentemperature field and static strains data from March 2013to October 2013 are chosen as training data and data inNovember 2013 are chosen as test data for verification of
the correlation model The static strain 119878III is expressed asfollows
119878III =8
sum
119894=1
120572119894119879119894+
16
sum
119895=1
120573119895119863119895+ 119888 (2)
where 119879119894is 119894th temperature data from set 119879
1 1198792 1198793 1198794
1198795 1198796 1198797 1198798 119863119895is 119895th temperature difference data from set
11987912 11987934 11987956 11987978 11987913 11987915 11987917 11987935 11987937 11987957 11987924 11987926 11987928 11987946
11987948 11987968 120572119894is 119894th linear expansion coefficient of 119879
119894 120573119895is 119895th
linear expansion coefficient of 119879119894119895 and 119888 is the constant term
Considering that total 24 linear expansion coefficientsand one constant term in (2) are very complicated forcalculation principal component analysis is further usedto simplify (2) which aims at finding small amounts ofcomprehensive factors to represent main information of tem-perature and temperature difference data By using principalcomponent analysis the comprehensive factors can be foundby transforming temperature and temperature difference datainto factor component space [11 12] For temperature data theexplained variance of the first potential comprehensive factor1198751reaches 967 apparently larger than the other potential
comprehensive factors Thus 1198751is selected as the compre-
hensive factor Meanwhile for temperature difference datausing principal component analysis the explained varianceof the first three potential comprehensive factors 119877
1 1198772 and
1198773reaches 988 which are selected as the comprehensive
factorsTherefore the correlation models between 119878III and tem-
perature field can be expressed as follows
119878III = 1205821([119864]1times8
997888
1198798times1
) +9978881205741times3
([119865]3times16
997888
11986316times1
) + 119888 (3)
where997888
1198798times1
denotes one vector containing eight temperaturedata 997888119863
16times1denotes one vector containing sixteen temper-
ature difference data 119863119895(119895 = 1 2 16) [119864]
1times8and
The Scientific World Journal 11
b = minus3632
k = 0946
minus100 minus55 minus10 8035Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus100
minus55
minus10
35
80
Sim
ulat
iveS
III
(120583120576)
(a) 1198781
b = minus2013k = 0898
minus45minus75 minus15 45 7515Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus75
minus45
minus15
15
45
75
Sim
ulat
iveS
III
(120583120576)
(b) 1198782
b = minus1844k = 0904
minus60
minus30
0
30
60
90
Sim
ulat
iveS
III
(120583120576)
minus30minus60 30 60 900Monitoring SIII (120583120576)
Scattered pointsFitting curve
(c) 1198783
b = 2057k = 1126
minus15minus30 15 300Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus40
minus25
minus10
5
20
35Si
mul
ativ
eSII
I(120583120576)
(d) 1198784
b = minus4139k = 0935
minus100
minus65
minus30
5
40
75
Sim
ulat
iveS
III
(120583120576)
minus105 4515 75minus45minus75 minus15
Monitoring SIII (120583120576)
Scattered pointsFitting curve
(e) 1198785
b = minus4094k = 0918
minus145minus205 minus25minus85 9535Monitoring SIII (120583120576)
minus185
minus130
minus75
minus20
35
90
Sim
ulat
iveS
III
(120583120576)
Scattered pointsFitting curve
(f) 1198786
Figure 16 Continued
12 The Scientific World Journal
b = 2148k = 0881
minus50
minus28
minus6
16
38
60Si
mul
ativ
eSII
I(120583120576)
minus45 minus23 minus1 43 6521Monitoring SIII (120583120576)
Scattered pointsFitting curve
(g) 1198787
b = 0615k = 1118
minus30
minus15
0
15
30
Sim
ulat
iveS
III
(120583120576)
minus15minus25 minus5 15 255Monitoring SIII (120583120576)
Scattered pointsFitting curve
(h) 1198788
Figure 16 Correlation between simulative 119878III and monitoring 119878III
Table 2 Performance parameters 1205821
1205741
1205742
1205743
and 119862 in correlationmodels
Static straindata 120582
1
1205741
1205742
1205743
119862
119878III of 1198781 0454 minus4711 minus2597 minus4944 minus46806119878III of 1198782 1585 0013 4104 minus1247 minus70238119878III of 1198783 0301 3416 minus2098 2968 minus6884119878III of 1198784 0350 minus0734 0351 0948 minus23597119878III of 1198785 0179 minus4219 1827 3165 minus22049119878III of 1198786 0062 minus3245 8242 minus5398 minus22862119878III of 1198787 minus0367 2411 minus0598 5531 17075119878III of 1198788 0611 2188 minus1130 2018 minus32799
[119865]3times16
denote transformation matrix which can be cal-culated using principal component analysis [11 12] 120582
1is
performance parameter of first potential comprehensivefactor 119875
1 9978881205741times3
= [1205741 1205742 1205743] is one vector containing
three performance parameters corresponding to first threepotential comprehensive factors 119877
1 1198772 and 119877
3 The values of
performance parameters 1205821 9978881205741times3
and 119888 are estimated usingtraining data by multivariate linear regression method [13]and are shown in Table 2
In order to verify the effectiveness of the correlationmodels shown in (2) test data in November 2013 are usedto obtain the simulative 119878III The scattered points betweensimulative 119878III and monitoring 119878III are plotted as shown inFigure 16 which are further fitted by linear function
119878119904= 119896119878119898+ 119887 (4)
where 119878119904denotes simulative 119878III 119878119898 denotes monitoring
119878III and 119896 119887 are fitting parameters with the fitting resultsshown in Figure 16 as well The fitting results verify theeffectiveness of the correlationmodels Amultivariable mean
value control chart is also employed tomonitor the measuredabnormal changes in the static strains of the steel truss archThe condition index 119890 for early warning of deterioration ofthe static performance of steel truss arch is defined as thedifference between the measured and predicted static strains
119890 = 119878119898minus 119878119901 (5)
where 119878119898is the measured static strain 119878III 119878119901 is the predicted
static strain 119878III which is obtained by using the correlationmodels shown in (2) in the healthy condition Figure 17 showsthemean value control chart for static strains of the steel trussarch using measurement data in November 2013 The upperand lower limits are determined using the controlling param-eter 119896 = 2235 Hence a long-term monitoring on the corre-lation models between temperature field and static strains issuitable for early warning of deterioration of the static per-formance of steel truss arch
33 Evaluation of Movement Performance of Pier Longi-tudinal displacements of piers and temperature field datain steel truss arch collected from March 2013 to October2013 are used for evaluation of movement performanceof piers Longitudinal displacement data from sensors 119871
119894
(119894 = 1 2 6) are denoted by 119889119894(119905) (where 119905 means time)
respectively Meanwhile temperature field data from sensors119882119894(119894 = 1 2 8) are denoted by 119879
119894(119905) respectively and
their averaged value 119879(119905) is used to represent the averagetemperature of the steel truss arch For simplicity taking10 minutes as basic time interval the 10 min average dis-placements and temperatures are calculated and used as therepresentative values of this time interval By this way thedisplacements and temperatures in one day are reduced to144 representative samples Furthermore the temperature-displacement scatter diagrams are plotted in Figure 18 Anoverall increase in longitudinal displacements of piers isobserved with an increase in the average temperature of the
The Scientific World Journal 13
UCL
LCL
CL
Con
ditio
n in
dexe
(120583120576)
minus15
minus10
minus5
0
5
10
15
10 20 30 40 50 600Sample
Figure 17 Mean value control chart for static strains of the steel truss arch
Table 3 Summary of linear regression models
Pier Regression function4 pier 119889
1
= minus11524 + 696119879
5 pier 1198892
= minus9253 + 562119879
6 pier 1198893
= minus6628 + 382119879
8 pier 1198894
= minus4658 + 336119879
9 pier 1198895
= minus7605 + 530119879
10 pier 1198896
= minus9850 + 664119879
bridge And the correlation coefficients of the longitudinaldisplacements of piers and average temperature of the steeltruss arch are all larger than 092 as shown in Figure 18
A linear regression analysis between the 10 min averagedisplacement 119889
119894(119905) and the temperature 119879(119905) is further per-
formed to model the temperature-displacement correlationsusing the following equation [14]
119889119894(119905) = 120573
0+ 1205731119879 (119905) (6)
where the regression coefficients 1205731and 120573
0were obtained by
the least-squares method as
1205731=
119878119889119894119879
119878119879119879
1205730= 119889119894minus 1205731119879
(7)
where 119878119889119894119879
is the covariance between the displacement andthe temperature sequences 119878
119879119879is the variance of the mea-
sured temperature sequences and 119879 and 119889119894are the means
of the measured temperature and displacement sequencesrespectively Table 3 summarizes the expressions of linearregression models of the displacement versus temperature
The analysis results indicate that good correlation existsbetween the longitudinal displacements of piers and averagetemperature of steel truss arch A multivariable mean valuecontrol chart is also employed to monitor the measuredabnormal changes in the longitudinal displacements of piersThe condition index 119890 for early warning of abnormal behaviorof piers is defined as the difference between themeasured andpredicted longitudinal displacements
119890 = 119889119898minus 119889119901 (8)
where 119889119898is the measured displacement of the piers 119889
119901is
the predicted displacement of the piers which is obtainedby using the linear regression models shown in Table 3in the healthy condition Figure 19 shows the mean valuecontrol chart for longitudinal displacements of piers usingmeasurement data in March 2013 The upper and lowerlimits are determined using the controlling parameter 119896 =
2804 Hence a long-term monitoring on the temperature-displacement correlation model is suitable for real-timecondition assessment for movement performance of piers
34 Evaluation of Fatigue Performance of Steel Deck Fatiguecracking is universal for orthotropic steel deck of highwaybridges Once the fatigue crack appears extensive fatiguecracking may occur and repair is very difficult Henceinfinite-fatigue-life design method is applied for NanjingDashengguan Bridge instead of finite-fatigue-life designmethod which is applied extensively in highway bridges Theinfinite-fatigue-life design method is to control the fatiguestress amplitude in the range less than the value of fatiguecut-off limit refrained from fatigue cracking Consequentlythe purpose of equipping stain gauges in the SHM system ofDashengguan Bridge is to monitor the stress amplitude andthe validity of fatigue design can be achieved by comparingthe stress amplitude with fatigue stress limit
The structural dynamic response is asymmetrical becauseof the irregularity of railway and braking force from thetrain even though the structure of Nanjing DashengguanBridge is spatially symmetrical Hence the stresses are notsymmetrical measured by strain gauges from the upstreamand downstream directions However the fatigue amplitudesof two strain gauges induced by trains are both extremelysmall Thus the strain history data of the strain gauge DYB-2are merely selected in this paper to illustrate that the fatiguestress amplitude is much lower than the value of fatigue stresslimit The typical strain time-history curves of DYB-2 whensingle train with 8 carriages or 16 carriages passed throughthe bridge are shown in Figure 20 respectively It can be seenthat during the passing of high-speed trains both 8 carriagesand 16 carriages have led to strain variations reaching to 2ndash8 120583120576 And the number of strain peaks is 16 and 32 for trainswith 8 carriages and 16 carriages respectively because theaxle number of each carriage is 2
To perform fatigue evaluation simplified rainflow cyclecounting algorithm was firstly used to process strain history
14 The Scientific World Journal
R = 09748
10 20 30 40 500
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
(a) Correlation between 1198891of the 4 pier and average temperature 119879 of
steel truss arch
10 20 30 40 500
Temperature (∘C)
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
R = 09746
(b) Correlation between 1198892of the 5 pier and average temperature 119879 of
steel truss arch
10 200 40 5030
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
Disp
lace
men
t (m
m)
R = 09471
(c) Correlation between 1198893of the 6 pier and average temperature 119879 of
steel truss arch
R = 09276
10 20 30 40 500
Temperature (∘C)
minus60
minus30
0
30
60
90
120D
ispla
cem
ent (
mm
)
(d) Correlation between 1198894of the 8 pier and average temperature 119879 of
steel truss arch
R = 09536
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(e) Correlation between 1198895of the 9 pier and average temperature 119879 of
steel truss arch
R = 09537
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(f) Correlation between 1198896of the 10 pier and average temperature 119879 of
steel truss arch
Figure 18 Correlation between longitudinal displacement and average temperature
The Scientific World Journal 15
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
)
500 1000 1500 2000 2500 3000 3500 4000 45000Sample
minus15
minus10
minus5
0
5
10
15
Figure 19 Mean value control chart for longitudinal displacements of piers
Stra
in (120583
120576)
minus174
minus172
minus170
minus168
minus166
minus164
minus162
1 2714 53
66
79
92
40
118
209
183
222
105
157
196
144
131
235
248
170
Data point
(a) Single train with 8 carriages passed through the bridge
Stra
in (120583
120576)
minus180minus178minus176minus174minus172minus170minus168minus166
1
61
91
31
181
211
121
421
301
481
511
361
391
271
451
241
331
541
571
151
Data point
(b) Single train with 16 carriages passed through the bridge
Figure 20 Typical strain time-history curves of DYB-2 when single train passed through the bridge
data and the spectrum of stress amplitude was obtained[15] The spectra of stress amplitude calculated using thestrain history data with regard to 8 carriages or 16 carriagesare shown in Figure 21 respectively It can be seen that themaximum stress amplitude is smaller than 10MPa obtainedfrom strain history curves under single trainThen accordingtoMinerrsquos law and standard of Eurocode 3 fatigue parametersof 119878eq (equivalent stress amplitude) and 119873
119904(stress cycle
number) under single train were calculated as follows [16]
119878eq = (
1198991198941198785
119894
sum119899119894
)
15
119873s = sum119899119894
(9)
where 119878119894is 119894th stress amplitude 119899
119894is the number of
stress cycles corresponding to 119878119894 and 15 is slope value of
log 119878- log119873 curve in the scope of lower stress amplitudeaccording to the standard of Eurocode 3 [17]
Using (9) the equivalent stress amplitudes 119878eq with regardto 8 carriages or 16 carriages are 075MPa and 071MParespectively And the stress cycle numbers 119873
119904are 22 and 45
respectively Thus the value of 119878eq with regard to 8 carriagesis similar to that of 16 carriages However the value of 119873
119904
with regard to 8 carriages is about twice as much as that of 16carriages Furthermore as recommended by Eurocode 3 forrib-to-deck welded joint the fatigue limit Δ120590
119871is 29MPa [17]
and the value of 119878eq when the high-speed train passes throughbridge is far less than Δ120590
119871 which indicates that fatigue life of
rib-to-deck welded joint in Nanjing Dashengguan Bridge canbe considered infinite Thus the fatigue performance of steel
deck satisfies the requirement of infinite-fatigue-life designmethod
4 Conclusions
In the last few decades Structural Health Monitoring (SHM)system has been widely installed in many highway bridgesto continuously monitor the structural condition over anextended period of time However application of SHMsystem in high-speed railway bridges is far insufficient Thispaper presents an analysis of the SHM system installedin Nanjing Dashengguan Bridge in China The monitoringsystem is composed of a sensor system a data acquisitionand transmission system a data management system and astructural evaluation system The monitoring system is cur-rently in full operation accumulating the bridgersquos behaviorsunder the varying environmental conditions such as high-speed trains and environmental temperature
Based on long-termhealthmonitoring data general prin-ciples and analytical processes for structural evaluation areprovided and some evaluation results are reported It can beshown from the analysis in this paper that the mathematicalcorrelation models describing the overall structural behaviorof the bridge can be obtained with the support of the healthmonitoring system which includes cross-correlation modelsfor accelerations correlation models between temperatureand static strains of steel truss arch and correlation mod-els between temperature and longitudinal displacements ofpiers The mean value control chart is further employed tomonitor the time series of differences between the measuredstructural parameters and the predicted parameters usingthese correlation models The upper and lower control limits
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
The Scientific World Journal 5
DYB-1 DYB-2
2500mm2500mm
Figure 3 Layout of dynamic strain gauges on the steel deck
Vertical accelerometer Z2Z4Transverse accelerometer Z1Z3
Figure 4 Layout of accelerometers on the steel deck
of the truss arch (where 119882119894denotes 119894th temperature sensor
119894 = 1 2 8 119884119894denotes 119894th strain sensor 119894 = 1 2 8)
The specification of FBG temperature sensors and straingauges is FBG4700 and FBG4100 supplied by China GeokonInstrumentsCo Ltd Sampling frequency of temperature andstatic strain data collection is set to 1Hz
22 Dynamic StrainMonitoring of Steel Deck Dynamic straingauges are installed at the 1st cross-section in the middleof the first main span of the bridge so as to monitor thefatigue performance of orthotropic steel deck as shown inFigure 3 It can be seen that the dynamic strain gauges DYB-1and DYB-2 monitor the long-term fatigue effects of deck-ribwelded detail induced by the traveling high-speed trainsThespecification of dynamic strain gauges is FBG4100 suppliedby China Geokon Instruments Co Ltd Sampling frequencyof dynamic strain data collections is set to 50Hz
23 Acceleration Monitoring of Steel Deck In order to moni-tor the transverse and vertical acceleration responses on thetwo main spans induced by the traveling high-speed trainsone transverse accelerometer and one vertical accelerometerwere installed at the 1st cross-section and 2nd cross-sectionin the middle of two main spans respectively as shown inFigure 4 (where 119885
119894denotes 119894th accelerometer 119894 = 1 2 4)
The specification of accelerometers is DH610 supplied byDonghua Testing Technology Co Ltd Sampling frequencyof acceleration data collections is set to 200Hz
24 Longitudinal Displacement Monitoring of Piers In thedesign of Nanjing Dashengguan Bridge the 7 pier is linkedto the bridge deck so that the bridge deck is restricted frommoving in the longitudinal direction at the 7 pier And therehave no longitudinal restraints between the bridge deck andother piersTherefore 6 displacement transducers are locatedon the piers and used for the measurement of longitudinaldisplacements at the 4 pier 5 pier 6 pier 8 pier 9pier and 10 pier as shown in Figure 5 (where 119871
119894denotes 119894th
displacement transducer 119894 = 1 2 6) The specificationof displacement transducers is CLMD2 supplied by ASMrsquosin Germany Finally it should be noted that the DAQs of theSHM system are located in the 6 and 8 piers
3 Evaluation of Structural Condition UsingLong-Term Monitoring Data
Using long-term monitoring data collected from the SHMsystem of Nanjing Dashengguan Bridge structural conditionassessment involves four aspects evaluation of vibration per-formance of themain girder evaluation of static performanceof steel truss arch evaluation of movement performance ofpier and evaluation of fatigue performance of steel deck
31 Evaluation of Vibration Performance of the Main GirderThe main function for the SHM system is to detect theperformance degradation of structures as early as possibleand to provide essential references for maintenance of actualbridges For Nanjing Dashengguan Bridge the vibration andstrain of the main girder together with the longitudinaldisplacement of piers are three important items to moni-tor Three kinds of performance evaluation are conductedaccordingly Those are evaluation of vibration performanceof the main girder evaluation of static performance of steeltruss arch and evaluation of movement performance of pierFirstly degradation or damage of local members may occurwith the increase of service period which will reduce thestructural stiffness of bridge Thus the dynamic propertyof bridge structure is affected significantly by the structuralstiffness reduction and will cause unusual variations ofthe structural vibration Such unusual behavior will affectrunning safety of the train directly Secondly damage of localmembers may induce redistribution of the internal forces inthe whole bridge under the temperature action The reasonis that Nanjing Dashengguan Bridge belongs to staticallyindeterminate structure Therefore static strain monitoringis necessary for typical members under temperature actions
6 The Scientific World Journal
Beijing Shanghai
4 pier 5 pier 6 pier 7 pier 8 pier 9 pier 10 pier
L1 L2 L3 L4 L5 L6
Figure 5 Layout of displacement transducers on the pier
Acce
lera
tion
(mm
s2)
1 80012001 4001 6001 10001
Data point
minus40
minus30
minus20
minus10
0
10
20
30
40
(a) Accelerometer 1198851
Acce
lera
tion
(mm
s2)
1 80012001 4001 6001 10001
Data point
minus40
minus30
minus20
minus10
0
10
20
30
40
(b) Accelerometer 1198853
Figure 6 Transverse acceleration time histories of the main girder induced by a high-speed train
Thirdly the longitudinal displacement behavior of the bridgepier is crucial for running safety of the train under thetemperature action Once the abnormal variation of the itemis detected inspection may be evolved timely for detection ofthe bridge piers In this section the evaluation of vibrationperformance of main girder is firstly discussed
For high-speed railway bridges vibration performanceof the main girder subjected to the high-speed trains isimportant for passenger comfort and traffic safety [6 7]For instance Zeng et al [8] point out that the essentialreason of train derailment is that the transverse vibrationof the train-railway-bridge system becomes unstable Henceit is necessary to present a strategy for early warning ofdeterioration of vibration performance of the main girderusing long-term health monitoring data
Figures 6 and 7 illustrate the typical transverse and ver-tical acceleration time histories of the main girder measuredfrom accelerometers Z
1simZ4 respectively It can be observed
that when high-speed train passed the bridge the transverseaccelerations in the middle of the first span are much smallerthan that of the second span and the vertical accelerationsin the middle of the first span are much larger than that ofthe second span These indicate that although the structurallayouts of two main spans of the main girder are the samethere exists significant difference between the transverseand vertical vibration characteristics of two main spans due
to the rail irregularity in transverse and vertical directionsThus there is a need to monitor the transverse and verticalaccelerations of two main spans in the long term so as torealize anomaly alarms for vibration performance of themaingirder
Hence in the present study the RMS values of theaccelerations are utilized as the monitoring parameters torepresent the transverse and vertical vibration characteristicof the main girder The cross-correlation between the RMSvalues of the accelerations in the middle of two main spans isfurther investigated when the high-speed trains pass throughthe bridge Figure 8 shows the scatter plots between the RMSvalues of the accelerations and shows the fitted results byusing quadratic polynomial as well
By using the fitting formula constructed in Figure 8scatter plot between theRMSvalues of the accelerations in themiddle of the second span and the corresponding fitted valuesis drawn in Figure 9 where the fitted values were obtainedby taking the RMS values of the accelerations in the middleof the first span into the fitting formula of the quadraticpolynomialThe correlation coefficient of the measured RMSvalues and fitted values on February 1st 2013 is 09391 and0918 for transverse and vertical accelerations respectively asshown in Figure 9 Computed using the same method all thedaily correlation coefficients in 2013 are more than 090 fortransverse and vertical accelerations as shown in Figure 10
The Scientific World Journal 7Ac
cele
ratio
n (m
ms2)
minus150
minus100
minus50
0
50
100
150
200
1 4001 6001 8001 100012001
Data point
(a) Accelerometer 1198852
Acce
lera
tion
(mm
s2)
1 4001 6001 8001 100012001
Data point
minus80
minus60
minus40
minus20
0
20
40
60
(b) Accelerometer 1198854
Figure 7 Vertical acceleration time histories of the main girder caused by a high-speed train
y = minus02642x2 + 37075x minus 56125
53 4 62 25 5535 45
RMS values of the 1st span (mms2)
0
2
4
6
8
10
RMS
valu
es o
f the
2nd
span
(mm
s2)
(a) Transverse accelerations
y = minus00131x2 + 07296x + 35995
50 15 20 2510
RMS values of the 1st span (mms2)
0
2
4
6
8
10
12
14
16RM
S va
lues
of t
he2
nd sp
an (m
ms2)
(b) Vertical accelerations
Figure 8 Cross-correlations between RMS values of accelerations measured in the middle of two main spans
Fitte
d RM
S va
lues
(mm
s2)
R = 09391
0
2
4
6
8
10
63 54 72 25 35 6555 7545
Measured RMS values (mms2)
(a) Transverse accelerations
R = 09518
4
6
8
10
12
14
16
Fitte
d RM
S va
lues
(mm
s2)
86 12 14 1610
Measured RMS values (mms2)
(b) Vertical accelerations
Figure 9 Fitting effect of cross-correlation between RMS values of accelerations by using quadratic polynomial
8 The Scientific World Journal
0 62 9331 186 217124 310279248155
Day
09
091
092
093
094
095
096
097C
orre
latio
n co
effici
ent
(a) Transverse accelerations
0 60 9030 180 210120 300270240150
Day
09
092
094
096
098
Cor
relat
ion
coeffi
cien
t
(b) Vertical accelerations
Figure 10 Daily computed correlation coefficients in 2013
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
s2)
minus2
minus1
0
1
2
20 40 60 80 100 120 140 1600Sample
Figure 11 Mean value control chart for accelerations in the middle of the second span
The analysis results indicate that good cross-correlationexists between the RMS values of the transverse and verticalaccelerations on the two main spans Multivariable controlchart is used here to monitor the measured changes in thetransverse and vertical accelerations caused by deteriorationof the vibration performance Firstly the condition index 119890
for early warning of abnormal vibration behavior is definedas the difference between the measured and predicted RMSvalues of the accelerations in the middle of the second span
119890 = RMS119898minus RMS
119901 (1)
where RMS119898is themeasured RMS values of the accelerations
in the middle of the second span RMS119901is the predicted
RMS values of the accelerations in the middle of the secondspan which is obtained by taking the RMS values of theaccelerations in the middle of the first span into the fittingformula of the quadratic polynomial in the healthy conditionThen a mean value control chart is employed to monitorthe time series of 119890 with regard to both transverse andvertical accelerations The time series of 119890 taken when thestructural vibration is in good condition will have somedistribution with mean 120583 and variance 120590
2 If the mean andstandard deviation are known a control chart is constructedby drawing a horizontal line CL at 120583 and twomore horizontallines representing the upper and lower control limits Theupper limit UCL is drawn at 120583 + 119896120590 and the lower limit LCL
at 120583 minus 119896120590 For online monitoring the controlling parameter119896 is chosen so that when the structural vibration is in goodcondition all observation samples fall between the controllimits When the new measurement is made the structuralabnormal vibration condition can be detected if an unusualnumber of samples fall beyond the control limits Figure 11shows the mean value control chart using the measuredand predicted RMS values of the accelerations shown inFigure 9 The upper and lower limits are determined usingthe controlling parameter 119896 = 2612 Hence a long-termmonitoring on the cross-correlation model of RMS values ofthe transverse and vertical accelerations can realize the earlywarning of deterioration of the vibration performance
32 Evaluation of Static Performance of Steel Truss Arch
321 Data of Temperature Field and Static Strains Tem-perature data from temperature sensor 119882
119894is denoted by
119879119894 temperature difference data acquired by 119879
119894minus 119879
119895is
denoted by119879119894119895(119879119894119895= 119879119894minus119879119895) and static strain data from strain
sensor119884119894is denoted by 119878
119894(119894 119895 = 1 2 8) As for illustration
the time-history curves of annual change and daily change intemperature 119879
1and temperature difference 119879
12are shown in
Figures 12 and 13 respectively And the time-history curvesof annual change and daily change in static strain data 119878
1
are shown in Figure 14 (negative values denote compressive
The Scientific World Journal 9
4 5 6 7 8 9 10 113Time (month)
minus5
5
15
25
35
45Te
mpe
ratu
re (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
5
10
15
20
Tem
pera
ture
(∘C)
(b) Daily curve (March 4th 2013)
Figure 12 Time-history curves of temperature data 1198791
4 5 6 7 8 9 10 113Time (month)
minus25
minus15
minus5
5
15
Tem
pera
ture
diff
eren
ce (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
minus10
minus5
0
5Te
mpe
ratu
re (∘
C)
(b) Daily curve (March 4th 2013)
Figure 13 Time-history curves of temperature difference data 11987912
4 5 6 7 8 9 10 113Time (month)
minus300
minus200
minus100
0
100
200
Stat
ic st
rain
(120583120576)
(a) Annual curve fromMarch 2013 to October 2013
Sharp peak
4 8 12 16 20 240Time (h)
minus120
minus65
minus10
45
100
Stat
ic st
rain
(120583120576)
(b) Daily curve (March 4th 2013)
Figure 14 Time-history curves of static strain data 1198781
10 The Scientific World Journal
4 8 12 16 20 240Time (h)
Stat
ic st
rain
S III
(120583120576)
minus120
minus65
minus10
45
100
(a) Daily extraction result 119878III of 1198781
4 5 6 7 8 9 10 113Time (month)
Stat
ic st
rain
S III
(120583120576)
minus200
minus130
minus60
10
80
150
(b) Annual extraction result 119878III of 1198781 fromMarch 2013 to October 2013
Figure 15 Extraction results 119878III of static strain data 1198781
static strain and positive values denote tension static strain)From Figure 14 it can be seen that time-history curve of 119878
1
contains many sharp peaks caused by high-speed trains witheach peak corresponding to the time when one train passesthrough the bridge And the static strains caused by high-speed trains are obviously lower than that of temperature fieldincluding temperature data and temperature difference data
As shown in Figure 14 static strain data in steel truss archmainly contains three parts static strain 119878I caused by tem-perature static strain 119878II caused by temperature differenceand static strain 119878III caused by train In order to construct thecorrelation between temperature field and static strain in steeltruss arch static strain 119878I and static strain 119878II (denoted by 119878III)must be extracted from the static strain data In the presentstudy wavelet packet decomposition method is used toextract 119878III considering the high-frequency characteristics ofstatic strain data caused by trains By using the wavelet packetdecomposition static strain data can be decomposed scale byscale into different frequency bands and each decompositioncoefficient corresponds to its frequency band [9 10] Eachdecomposition coefficient can be reconstructed into time-domain signals with constraints of its own frequency bandTaking the daily time-history curve of static strain data1198781shown in Figure 14(b) as an example it is decomposed
within 8 scales and the decomposition coefficient 11986208
119895
isspecially selected and then reconstructed as 119878III shown inFigure 15(a) It can be seen that sharp peaks caused by trainsare removed effectively and the static strains which are causedby temperature field are retained as well Furthermore theannual extraction result 119878III of 1198781 fromMarch 2013 toOctober2013 is shown in Figure 15(b)
322 Correlation Model between Temperature Field andStatic Strains In constructing the correlationmodel betweentemperature field and static strains data from March 2013to October 2013 are chosen as training data and data inNovember 2013 are chosen as test data for verification of
the correlation model The static strain 119878III is expressed asfollows
119878III =8
sum
119894=1
120572119894119879119894+
16
sum
119895=1
120573119895119863119895+ 119888 (2)
where 119879119894is 119894th temperature data from set 119879
1 1198792 1198793 1198794
1198795 1198796 1198797 1198798 119863119895is 119895th temperature difference data from set
11987912 11987934 11987956 11987978 11987913 11987915 11987917 11987935 11987937 11987957 11987924 11987926 11987928 11987946
11987948 11987968 120572119894is 119894th linear expansion coefficient of 119879
119894 120573119895is 119895th
linear expansion coefficient of 119879119894119895 and 119888 is the constant term
Considering that total 24 linear expansion coefficientsand one constant term in (2) are very complicated forcalculation principal component analysis is further usedto simplify (2) which aims at finding small amounts ofcomprehensive factors to represent main information of tem-perature and temperature difference data By using principalcomponent analysis the comprehensive factors can be foundby transforming temperature and temperature difference datainto factor component space [11 12] For temperature data theexplained variance of the first potential comprehensive factor1198751reaches 967 apparently larger than the other potential
comprehensive factors Thus 1198751is selected as the compre-
hensive factor Meanwhile for temperature difference datausing principal component analysis the explained varianceof the first three potential comprehensive factors 119877
1 1198772 and
1198773reaches 988 which are selected as the comprehensive
factorsTherefore the correlation models between 119878III and tem-
perature field can be expressed as follows
119878III = 1205821([119864]1times8
997888
1198798times1
) +9978881205741times3
([119865]3times16
997888
11986316times1
) + 119888 (3)
where997888
1198798times1
denotes one vector containing eight temperaturedata 997888119863
16times1denotes one vector containing sixteen temper-
ature difference data 119863119895(119895 = 1 2 16) [119864]
1times8and
The Scientific World Journal 11
b = minus3632
k = 0946
minus100 minus55 minus10 8035Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus100
minus55
minus10
35
80
Sim
ulat
iveS
III
(120583120576)
(a) 1198781
b = minus2013k = 0898
minus45minus75 minus15 45 7515Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus75
minus45
minus15
15
45
75
Sim
ulat
iveS
III
(120583120576)
(b) 1198782
b = minus1844k = 0904
minus60
minus30
0
30
60
90
Sim
ulat
iveS
III
(120583120576)
minus30minus60 30 60 900Monitoring SIII (120583120576)
Scattered pointsFitting curve
(c) 1198783
b = 2057k = 1126
minus15minus30 15 300Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus40
minus25
minus10
5
20
35Si
mul
ativ
eSII
I(120583120576)
(d) 1198784
b = minus4139k = 0935
minus100
minus65
minus30
5
40
75
Sim
ulat
iveS
III
(120583120576)
minus105 4515 75minus45minus75 minus15
Monitoring SIII (120583120576)
Scattered pointsFitting curve
(e) 1198785
b = minus4094k = 0918
minus145minus205 minus25minus85 9535Monitoring SIII (120583120576)
minus185
minus130
minus75
minus20
35
90
Sim
ulat
iveS
III
(120583120576)
Scattered pointsFitting curve
(f) 1198786
Figure 16 Continued
12 The Scientific World Journal
b = 2148k = 0881
minus50
minus28
minus6
16
38
60Si
mul
ativ
eSII
I(120583120576)
minus45 minus23 minus1 43 6521Monitoring SIII (120583120576)
Scattered pointsFitting curve
(g) 1198787
b = 0615k = 1118
minus30
minus15
0
15
30
Sim
ulat
iveS
III
(120583120576)
minus15minus25 minus5 15 255Monitoring SIII (120583120576)
Scattered pointsFitting curve
(h) 1198788
Figure 16 Correlation between simulative 119878III and monitoring 119878III
Table 2 Performance parameters 1205821
1205741
1205742
1205743
and 119862 in correlationmodels
Static straindata 120582
1
1205741
1205742
1205743
119862
119878III of 1198781 0454 minus4711 minus2597 minus4944 minus46806119878III of 1198782 1585 0013 4104 minus1247 minus70238119878III of 1198783 0301 3416 minus2098 2968 minus6884119878III of 1198784 0350 minus0734 0351 0948 minus23597119878III of 1198785 0179 minus4219 1827 3165 minus22049119878III of 1198786 0062 minus3245 8242 minus5398 minus22862119878III of 1198787 minus0367 2411 minus0598 5531 17075119878III of 1198788 0611 2188 minus1130 2018 minus32799
[119865]3times16
denote transformation matrix which can be cal-culated using principal component analysis [11 12] 120582
1is
performance parameter of first potential comprehensivefactor 119875
1 9978881205741times3
= [1205741 1205742 1205743] is one vector containing
three performance parameters corresponding to first threepotential comprehensive factors 119877
1 1198772 and 119877
3 The values of
performance parameters 1205821 9978881205741times3
and 119888 are estimated usingtraining data by multivariate linear regression method [13]and are shown in Table 2
In order to verify the effectiveness of the correlationmodels shown in (2) test data in November 2013 are usedto obtain the simulative 119878III The scattered points betweensimulative 119878III and monitoring 119878III are plotted as shown inFigure 16 which are further fitted by linear function
119878119904= 119896119878119898+ 119887 (4)
where 119878119904denotes simulative 119878III 119878119898 denotes monitoring
119878III and 119896 119887 are fitting parameters with the fitting resultsshown in Figure 16 as well The fitting results verify theeffectiveness of the correlationmodels Amultivariable mean
value control chart is also employed tomonitor the measuredabnormal changes in the static strains of the steel truss archThe condition index 119890 for early warning of deterioration ofthe static performance of steel truss arch is defined as thedifference between the measured and predicted static strains
119890 = 119878119898minus 119878119901 (5)
where 119878119898is the measured static strain 119878III 119878119901 is the predicted
static strain 119878III which is obtained by using the correlationmodels shown in (2) in the healthy condition Figure 17 showsthemean value control chart for static strains of the steel trussarch using measurement data in November 2013 The upperand lower limits are determined using the controlling param-eter 119896 = 2235 Hence a long-term monitoring on the corre-lation models between temperature field and static strains issuitable for early warning of deterioration of the static per-formance of steel truss arch
33 Evaluation of Movement Performance of Pier Longi-tudinal displacements of piers and temperature field datain steel truss arch collected from March 2013 to October2013 are used for evaluation of movement performanceof piers Longitudinal displacement data from sensors 119871
119894
(119894 = 1 2 6) are denoted by 119889119894(119905) (where 119905 means time)
respectively Meanwhile temperature field data from sensors119882119894(119894 = 1 2 8) are denoted by 119879
119894(119905) respectively and
their averaged value 119879(119905) is used to represent the averagetemperature of the steel truss arch For simplicity taking10 minutes as basic time interval the 10 min average dis-placements and temperatures are calculated and used as therepresentative values of this time interval By this way thedisplacements and temperatures in one day are reduced to144 representative samples Furthermore the temperature-displacement scatter diagrams are plotted in Figure 18 Anoverall increase in longitudinal displacements of piers isobserved with an increase in the average temperature of the
The Scientific World Journal 13
UCL
LCL
CL
Con
ditio
n in
dexe
(120583120576)
minus15
minus10
minus5
0
5
10
15
10 20 30 40 50 600Sample
Figure 17 Mean value control chart for static strains of the steel truss arch
Table 3 Summary of linear regression models
Pier Regression function4 pier 119889
1
= minus11524 + 696119879
5 pier 1198892
= minus9253 + 562119879
6 pier 1198893
= minus6628 + 382119879
8 pier 1198894
= minus4658 + 336119879
9 pier 1198895
= minus7605 + 530119879
10 pier 1198896
= minus9850 + 664119879
bridge And the correlation coefficients of the longitudinaldisplacements of piers and average temperature of the steeltruss arch are all larger than 092 as shown in Figure 18
A linear regression analysis between the 10 min averagedisplacement 119889
119894(119905) and the temperature 119879(119905) is further per-
formed to model the temperature-displacement correlationsusing the following equation [14]
119889119894(119905) = 120573
0+ 1205731119879 (119905) (6)
where the regression coefficients 1205731and 120573
0were obtained by
the least-squares method as
1205731=
119878119889119894119879
119878119879119879
1205730= 119889119894minus 1205731119879
(7)
where 119878119889119894119879
is the covariance between the displacement andthe temperature sequences 119878
119879119879is the variance of the mea-
sured temperature sequences and 119879 and 119889119894are the means
of the measured temperature and displacement sequencesrespectively Table 3 summarizes the expressions of linearregression models of the displacement versus temperature
The analysis results indicate that good correlation existsbetween the longitudinal displacements of piers and averagetemperature of steel truss arch A multivariable mean valuecontrol chart is also employed to monitor the measuredabnormal changes in the longitudinal displacements of piersThe condition index 119890 for early warning of abnormal behaviorof piers is defined as the difference between themeasured andpredicted longitudinal displacements
119890 = 119889119898minus 119889119901 (8)
where 119889119898is the measured displacement of the piers 119889
119901is
the predicted displacement of the piers which is obtainedby using the linear regression models shown in Table 3in the healthy condition Figure 19 shows the mean valuecontrol chart for longitudinal displacements of piers usingmeasurement data in March 2013 The upper and lowerlimits are determined using the controlling parameter 119896 =
2804 Hence a long-term monitoring on the temperature-displacement correlation model is suitable for real-timecondition assessment for movement performance of piers
34 Evaluation of Fatigue Performance of Steel Deck Fatiguecracking is universal for orthotropic steel deck of highwaybridges Once the fatigue crack appears extensive fatiguecracking may occur and repair is very difficult Henceinfinite-fatigue-life design method is applied for NanjingDashengguan Bridge instead of finite-fatigue-life designmethod which is applied extensively in highway bridges Theinfinite-fatigue-life design method is to control the fatiguestress amplitude in the range less than the value of fatiguecut-off limit refrained from fatigue cracking Consequentlythe purpose of equipping stain gauges in the SHM system ofDashengguan Bridge is to monitor the stress amplitude andthe validity of fatigue design can be achieved by comparingthe stress amplitude with fatigue stress limit
The structural dynamic response is asymmetrical becauseof the irregularity of railway and braking force from thetrain even though the structure of Nanjing DashengguanBridge is spatially symmetrical Hence the stresses are notsymmetrical measured by strain gauges from the upstreamand downstream directions However the fatigue amplitudesof two strain gauges induced by trains are both extremelysmall Thus the strain history data of the strain gauge DYB-2are merely selected in this paper to illustrate that the fatiguestress amplitude is much lower than the value of fatigue stresslimit The typical strain time-history curves of DYB-2 whensingle train with 8 carriages or 16 carriages passed throughthe bridge are shown in Figure 20 respectively It can be seenthat during the passing of high-speed trains both 8 carriagesand 16 carriages have led to strain variations reaching to 2ndash8 120583120576 And the number of strain peaks is 16 and 32 for trainswith 8 carriages and 16 carriages respectively because theaxle number of each carriage is 2
To perform fatigue evaluation simplified rainflow cyclecounting algorithm was firstly used to process strain history
14 The Scientific World Journal
R = 09748
10 20 30 40 500
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
(a) Correlation between 1198891of the 4 pier and average temperature 119879 of
steel truss arch
10 20 30 40 500
Temperature (∘C)
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
R = 09746
(b) Correlation between 1198892of the 5 pier and average temperature 119879 of
steel truss arch
10 200 40 5030
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
Disp
lace
men
t (m
m)
R = 09471
(c) Correlation between 1198893of the 6 pier and average temperature 119879 of
steel truss arch
R = 09276
10 20 30 40 500
Temperature (∘C)
minus60
minus30
0
30
60
90
120D
ispla
cem
ent (
mm
)
(d) Correlation between 1198894of the 8 pier and average temperature 119879 of
steel truss arch
R = 09536
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(e) Correlation between 1198895of the 9 pier and average temperature 119879 of
steel truss arch
R = 09537
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(f) Correlation between 1198896of the 10 pier and average temperature 119879 of
steel truss arch
Figure 18 Correlation between longitudinal displacement and average temperature
The Scientific World Journal 15
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
)
500 1000 1500 2000 2500 3000 3500 4000 45000Sample
minus15
minus10
minus5
0
5
10
15
Figure 19 Mean value control chart for longitudinal displacements of piers
Stra
in (120583
120576)
minus174
minus172
minus170
minus168
minus166
minus164
minus162
1 2714 53
66
79
92
40
118
209
183
222
105
157
196
144
131
235
248
170
Data point
(a) Single train with 8 carriages passed through the bridge
Stra
in (120583
120576)
minus180minus178minus176minus174minus172minus170minus168minus166
1
61
91
31
181
211
121
421
301
481
511
361
391
271
451
241
331
541
571
151
Data point
(b) Single train with 16 carriages passed through the bridge
Figure 20 Typical strain time-history curves of DYB-2 when single train passed through the bridge
data and the spectrum of stress amplitude was obtained[15] The spectra of stress amplitude calculated using thestrain history data with regard to 8 carriages or 16 carriagesare shown in Figure 21 respectively It can be seen that themaximum stress amplitude is smaller than 10MPa obtainedfrom strain history curves under single trainThen accordingtoMinerrsquos law and standard of Eurocode 3 fatigue parametersof 119878eq (equivalent stress amplitude) and 119873
119904(stress cycle
number) under single train were calculated as follows [16]
119878eq = (
1198991198941198785
119894
sum119899119894
)
15
119873s = sum119899119894
(9)
where 119878119894is 119894th stress amplitude 119899
119894is the number of
stress cycles corresponding to 119878119894 and 15 is slope value of
log 119878- log119873 curve in the scope of lower stress amplitudeaccording to the standard of Eurocode 3 [17]
Using (9) the equivalent stress amplitudes 119878eq with regardto 8 carriages or 16 carriages are 075MPa and 071MParespectively And the stress cycle numbers 119873
119904are 22 and 45
respectively Thus the value of 119878eq with regard to 8 carriagesis similar to that of 16 carriages However the value of 119873
119904
with regard to 8 carriages is about twice as much as that of 16carriages Furthermore as recommended by Eurocode 3 forrib-to-deck welded joint the fatigue limit Δ120590
119871is 29MPa [17]
and the value of 119878eq when the high-speed train passes throughbridge is far less than Δ120590
119871 which indicates that fatigue life of
rib-to-deck welded joint in Nanjing Dashengguan Bridge canbe considered infinite Thus the fatigue performance of steel
deck satisfies the requirement of infinite-fatigue-life designmethod
4 Conclusions
In the last few decades Structural Health Monitoring (SHM)system has been widely installed in many highway bridgesto continuously monitor the structural condition over anextended period of time However application of SHMsystem in high-speed railway bridges is far insufficient Thispaper presents an analysis of the SHM system installedin Nanjing Dashengguan Bridge in China The monitoringsystem is composed of a sensor system a data acquisitionand transmission system a data management system and astructural evaluation system The monitoring system is cur-rently in full operation accumulating the bridgersquos behaviorsunder the varying environmental conditions such as high-speed trains and environmental temperature
Based on long-termhealthmonitoring data general prin-ciples and analytical processes for structural evaluation areprovided and some evaluation results are reported It can beshown from the analysis in this paper that the mathematicalcorrelation models describing the overall structural behaviorof the bridge can be obtained with the support of the healthmonitoring system which includes cross-correlation modelsfor accelerations correlation models between temperatureand static strains of steel truss arch and correlation mod-els between temperature and longitudinal displacements ofpiers The mean value control chart is further employed tomonitor the time series of differences between the measuredstructural parameters and the predicted parameters usingthese correlation models The upper and lower control limits
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
6 The Scientific World Journal
Beijing Shanghai
4 pier 5 pier 6 pier 7 pier 8 pier 9 pier 10 pier
L1 L2 L3 L4 L5 L6
Figure 5 Layout of displacement transducers on the pier
Acce
lera
tion
(mm
s2)
1 80012001 4001 6001 10001
Data point
minus40
minus30
minus20
minus10
0
10
20
30
40
(a) Accelerometer 1198851
Acce
lera
tion
(mm
s2)
1 80012001 4001 6001 10001
Data point
minus40
minus30
minus20
minus10
0
10
20
30
40
(b) Accelerometer 1198853
Figure 6 Transverse acceleration time histories of the main girder induced by a high-speed train
Thirdly the longitudinal displacement behavior of the bridgepier is crucial for running safety of the train under thetemperature action Once the abnormal variation of the itemis detected inspection may be evolved timely for detection ofthe bridge piers In this section the evaluation of vibrationperformance of main girder is firstly discussed
For high-speed railway bridges vibration performanceof the main girder subjected to the high-speed trains isimportant for passenger comfort and traffic safety [6 7]For instance Zeng et al [8] point out that the essentialreason of train derailment is that the transverse vibrationof the train-railway-bridge system becomes unstable Henceit is necessary to present a strategy for early warning ofdeterioration of vibration performance of the main girderusing long-term health monitoring data
Figures 6 and 7 illustrate the typical transverse and ver-tical acceleration time histories of the main girder measuredfrom accelerometers Z
1simZ4 respectively It can be observed
that when high-speed train passed the bridge the transverseaccelerations in the middle of the first span are much smallerthan that of the second span and the vertical accelerationsin the middle of the first span are much larger than that ofthe second span These indicate that although the structurallayouts of two main spans of the main girder are the samethere exists significant difference between the transverseand vertical vibration characteristics of two main spans due
to the rail irregularity in transverse and vertical directionsThus there is a need to monitor the transverse and verticalaccelerations of two main spans in the long term so as torealize anomaly alarms for vibration performance of themaingirder
Hence in the present study the RMS values of theaccelerations are utilized as the monitoring parameters torepresent the transverse and vertical vibration characteristicof the main girder The cross-correlation between the RMSvalues of the accelerations in the middle of two main spans isfurther investigated when the high-speed trains pass throughthe bridge Figure 8 shows the scatter plots between the RMSvalues of the accelerations and shows the fitted results byusing quadratic polynomial as well
By using the fitting formula constructed in Figure 8scatter plot between theRMSvalues of the accelerations in themiddle of the second span and the corresponding fitted valuesis drawn in Figure 9 where the fitted values were obtainedby taking the RMS values of the accelerations in the middleof the first span into the fitting formula of the quadraticpolynomialThe correlation coefficient of the measured RMSvalues and fitted values on February 1st 2013 is 09391 and0918 for transverse and vertical accelerations respectively asshown in Figure 9 Computed using the same method all thedaily correlation coefficients in 2013 are more than 090 fortransverse and vertical accelerations as shown in Figure 10
The Scientific World Journal 7Ac
cele
ratio
n (m
ms2)
minus150
minus100
minus50
0
50
100
150
200
1 4001 6001 8001 100012001
Data point
(a) Accelerometer 1198852
Acce
lera
tion
(mm
s2)
1 4001 6001 8001 100012001
Data point
minus80
minus60
minus40
minus20
0
20
40
60
(b) Accelerometer 1198854
Figure 7 Vertical acceleration time histories of the main girder caused by a high-speed train
y = minus02642x2 + 37075x minus 56125
53 4 62 25 5535 45
RMS values of the 1st span (mms2)
0
2
4
6
8
10
RMS
valu
es o
f the
2nd
span
(mm
s2)
(a) Transverse accelerations
y = minus00131x2 + 07296x + 35995
50 15 20 2510
RMS values of the 1st span (mms2)
0
2
4
6
8
10
12
14
16RM
S va
lues
of t
he2
nd sp
an (m
ms2)
(b) Vertical accelerations
Figure 8 Cross-correlations between RMS values of accelerations measured in the middle of two main spans
Fitte
d RM
S va
lues
(mm
s2)
R = 09391
0
2
4
6
8
10
63 54 72 25 35 6555 7545
Measured RMS values (mms2)
(a) Transverse accelerations
R = 09518
4
6
8
10
12
14
16
Fitte
d RM
S va
lues
(mm
s2)
86 12 14 1610
Measured RMS values (mms2)
(b) Vertical accelerations
Figure 9 Fitting effect of cross-correlation between RMS values of accelerations by using quadratic polynomial
8 The Scientific World Journal
0 62 9331 186 217124 310279248155
Day
09
091
092
093
094
095
096
097C
orre
latio
n co
effici
ent
(a) Transverse accelerations
0 60 9030 180 210120 300270240150
Day
09
092
094
096
098
Cor
relat
ion
coeffi
cien
t
(b) Vertical accelerations
Figure 10 Daily computed correlation coefficients in 2013
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
s2)
minus2
minus1
0
1
2
20 40 60 80 100 120 140 1600Sample
Figure 11 Mean value control chart for accelerations in the middle of the second span
The analysis results indicate that good cross-correlationexists between the RMS values of the transverse and verticalaccelerations on the two main spans Multivariable controlchart is used here to monitor the measured changes in thetransverse and vertical accelerations caused by deteriorationof the vibration performance Firstly the condition index 119890
for early warning of abnormal vibration behavior is definedas the difference between the measured and predicted RMSvalues of the accelerations in the middle of the second span
119890 = RMS119898minus RMS
119901 (1)
where RMS119898is themeasured RMS values of the accelerations
in the middle of the second span RMS119901is the predicted
RMS values of the accelerations in the middle of the secondspan which is obtained by taking the RMS values of theaccelerations in the middle of the first span into the fittingformula of the quadratic polynomial in the healthy conditionThen a mean value control chart is employed to monitorthe time series of 119890 with regard to both transverse andvertical accelerations The time series of 119890 taken when thestructural vibration is in good condition will have somedistribution with mean 120583 and variance 120590
2 If the mean andstandard deviation are known a control chart is constructedby drawing a horizontal line CL at 120583 and twomore horizontallines representing the upper and lower control limits Theupper limit UCL is drawn at 120583 + 119896120590 and the lower limit LCL
at 120583 minus 119896120590 For online monitoring the controlling parameter119896 is chosen so that when the structural vibration is in goodcondition all observation samples fall between the controllimits When the new measurement is made the structuralabnormal vibration condition can be detected if an unusualnumber of samples fall beyond the control limits Figure 11shows the mean value control chart using the measuredand predicted RMS values of the accelerations shown inFigure 9 The upper and lower limits are determined usingthe controlling parameter 119896 = 2612 Hence a long-termmonitoring on the cross-correlation model of RMS values ofthe transverse and vertical accelerations can realize the earlywarning of deterioration of the vibration performance
32 Evaluation of Static Performance of Steel Truss Arch
321 Data of Temperature Field and Static Strains Tem-perature data from temperature sensor 119882
119894is denoted by
119879119894 temperature difference data acquired by 119879
119894minus 119879
119895is
denoted by119879119894119895(119879119894119895= 119879119894minus119879119895) and static strain data from strain
sensor119884119894is denoted by 119878
119894(119894 119895 = 1 2 8) As for illustration
the time-history curves of annual change and daily change intemperature 119879
1and temperature difference 119879
12are shown in
Figures 12 and 13 respectively And the time-history curvesof annual change and daily change in static strain data 119878
1
are shown in Figure 14 (negative values denote compressive
The Scientific World Journal 9
4 5 6 7 8 9 10 113Time (month)
minus5
5
15
25
35
45Te
mpe
ratu
re (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
5
10
15
20
Tem
pera
ture
(∘C)
(b) Daily curve (March 4th 2013)
Figure 12 Time-history curves of temperature data 1198791
4 5 6 7 8 9 10 113Time (month)
minus25
minus15
minus5
5
15
Tem
pera
ture
diff
eren
ce (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
minus10
minus5
0
5Te
mpe
ratu
re (∘
C)
(b) Daily curve (March 4th 2013)
Figure 13 Time-history curves of temperature difference data 11987912
4 5 6 7 8 9 10 113Time (month)
minus300
minus200
minus100
0
100
200
Stat
ic st
rain
(120583120576)
(a) Annual curve fromMarch 2013 to October 2013
Sharp peak
4 8 12 16 20 240Time (h)
minus120
minus65
minus10
45
100
Stat
ic st
rain
(120583120576)
(b) Daily curve (March 4th 2013)
Figure 14 Time-history curves of static strain data 1198781
10 The Scientific World Journal
4 8 12 16 20 240Time (h)
Stat
ic st
rain
S III
(120583120576)
minus120
minus65
minus10
45
100
(a) Daily extraction result 119878III of 1198781
4 5 6 7 8 9 10 113Time (month)
Stat
ic st
rain
S III
(120583120576)
minus200
minus130
minus60
10
80
150
(b) Annual extraction result 119878III of 1198781 fromMarch 2013 to October 2013
Figure 15 Extraction results 119878III of static strain data 1198781
static strain and positive values denote tension static strain)From Figure 14 it can be seen that time-history curve of 119878
1
contains many sharp peaks caused by high-speed trains witheach peak corresponding to the time when one train passesthrough the bridge And the static strains caused by high-speed trains are obviously lower than that of temperature fieldincluding temperature data and temperature difference data
As shown in Figure 14 static strain data in steel truss archmainly contains three parts static strain 119878I caused by tem-perature static strain 119878II caused by temperature differenceand static strain 119878III caused by train In order to construct thecorrelation between temperature field and static strain in steeltruss arch static strain 119878I and static strain 119878II (denoted by 119878III)must be extracted from the static strain data In the presentstudy wavelet packet decomposition method is used toextract 119878III considering the high-frequency characteristics ofstatic strain data caused by trains By using the wavelet packetdecomposition static strain data can be decomposed scale byscale into different frequency bands and each decompositioncoefficient corresponds to its frequency band [9 10] Eachdecomposition coefficient can be reconstructed into time-domain signals with constraints of its own frequency bandTaking the daily time-history curve of static strain data1198781shown in Figure 14(b) as an example it is decomposed
within 8 scales and the decomposition coefficient 11986208
119895
isspecially selected and then reconstructed as 119878III shown inFigure 15(a) It can be seen that sharp peaks caused by trainsare removed effectively and the static strains which are causedby temperature field are retained as well Furthermore theannual extraction result 119878III of 1198781 fromMarch 2013 toOctober2013 is shown in Figure 15(b)
322 Correlation Model between Temperature Field andStatic Strains In constructing the correlationmodel betweentemperature field and static strains data from March 2013to October 2013 are chosen as training data and data inNovember 2013 are chosen as test data for verification of
the correlation model The static strain 119878III is expressed asfollows
119878III =8
sum
119894=1
120572119894119879119894+
16
sum
119895=1
120573119895119863119895+ 119888 (2)
where 119879119894is 119894th temperature data from set 119879
1 1198792 1198793 1198794
1198795 1198796 1198797 1198798 119863119895is 119895th temperature difference data from set
11987912 11987934 11987956 11987978 11987913 11987915 11987917 11987935 11987937 11987957 11987924 11987926 11987928 11987946
11987948 11987968 120572119894is 119894th linear expansion coefficient of 119879
119894 120573119895is 119895th
linear expansion coefficient of 119879119894119895 and 119888 is the constant term
Considering that total 24 linear expansion coefficientsand one constant term in (2) are very complicated forcalculation principal component analysis is further usedto simplify (2) which aims at finding small amounts ofcomprehensive factors to represent main information of tem-perature and temperature difference data By using principalcomponent analysis the comprehensive factors can be foundby transforming temperature and temperature difference datainto factor component space [11 12] For temperature data theexplained variance of the first potential comprehensive factor1198751reaches 967 apparently larger than the other potential
comprehensive factors Thus 1198751is selected as the compre-
hensive factor Meanwhile for temperature difference datausing principal component analysis the explained varianceof the first three potential comprehensive factors 119877
1 1198772 and
1198773reaches 988 which are selected as the comprehensive
factorsTherefore the correlation models between 119878III and tem-
perature field can be expressed as follows
119878III = 1205821([119864]1times8
997888
1198798times1
) +9978881205741times3
([119865]3times16
997888
11986316times1
) + 119888 (3)
where997888
1198798times1
denotes one vector containing eight temperaturedata 997888119863
16times1denotes one vector containing sixteen temper-
ature difference data 119863119895(119895 = 1 2 16) [119864]
1times8and
The Scientific World Journal 11
b = minus3632
k = 0946
minus100 minus55 minus10 8035Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus100
minus55
minus10
35
80
Sim
ulat
iveS
III
(120583120576)
(a) 1198781
b = minus2013k = 0898
minus45minus75 minus15 45 7515Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus75
minus45
minus15
15
45
75
Sim
ulat
iveS
III
(120583120576)
(b) 1198782
b = minus1844k = 0904
minus60
minus30
0
30
60
90
Sim
ulat
iveS
III
(120583120576)
minus30minus60 30 60 900Monitoring SIII (120583120576)
Scattered pointsFitting curve
(c) 1198783
b = 2057k = 1126
minus15minus30 15 300Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus40
minus25
minus10
5
20
35Si
mul
ativ
eSII
I(120583120576)
(d) 1198784
b = minus4139k = 0935
minus100
minus65
minus30
5
40
75
Sim
ulat
iveS
III
(120583120576)
minus105 4515 75minus45minus75 minus15
Monitoring SIII (120583120576)
Scattered pointsFitting curve
(e) 1198785
b = minus4094k = 0918
minus145minus205 minus25minus85 9535Monitoring SIII (120583120576)
minus185
minus130
minus75
minus20
35
90
Sim
ulat
iveS
III
(120583120576)
Scattered pointsFitting curve
(f) 1198786
Figure 16 Continued
12 The Scientific World Journal
b = 2148k = 0881
minus50
minus28
minus6
16
38
60Si
mul
ativ
eSII
I(120583120576)
minus45 minus23 minus1 43 6521Monitoring SIII (120583120576)
Scattered pointsFitting curve
(g) 1198787
b = 0615k = 1118
minus30
minus15
0
15
30
Sim
ulat
iveS
III
(120583120576)
minus15minus25 minus5 15 255Monitoring SIII (120583120576)
Scattered pointsFitting curve
(h) 1198788
Figure 16 Correlation between simulative 119878III and monitoring 119878III
Table 2 Performance parameters 1205821
1205741
1205742
1205743
and 119862 in correlationmodels
Static straindata 120582
1
1205741
1205742
1205743
119862
119878III of 1198781 0454 minus4711 minus2597 minus4944 minus46806119878III of 1198782 1585 0013 4104 minus1247 minus70238119878III of 1198783 0301 3416 minus2098 2968 minus6884119878III of 1198784 0350 minus0734 0351 0948 minus23597119878III of 1198785 0179 minus4219 1827 3165 minus22049119878III of 1198786 0062 minus3245 8242 minus5398 minus22862119878III of 1198787 minus0367 2411 minus0598 5531 17075119878III of 1198788 0611 2188 minus1130 2018 minus32799
[119865]3times16
denote transformation matrix which can be cal-culated using principal component analysis [11 12] 120582
1is
performance parameter of first potential comprehensivefactor 119875
1 9978881205741times3
= [1205741 1205742 1205743] is one vector containing
three performance parameters corresponding to first threepotential comprehensive factors 119877
1 1198772 and 119877
3 The values of
performance parameters 1205821 9978881205741times3
and 119888 are estimated usingtraining data by multivariate linear regression method [13]and are shown in Table 2
In order to verify the effectiveness of the correlationmodels shown in (2) test data in November 2013 are usedto obtain the simulative 119878III The scattered points betweensimulative 119878III and monitoring 119878III are plotted as shown inFigure 16 which are further fitted by linear function
119878119904= 119896119878119898+ 119887 (4)
where 119878119904denotes simulative 119878III 119878119898 denotes monitoring
119878III and 119896 119887 are fitting parameters with the fitting resultsshown in Figure 16 as well The fitting results verify theeffectiveness of the correlationmodels Amultivariable mean
value control chart is also employed tomonitor the measuredabnormal changes in the static strains of the steel truss archThe condition index 119890 for early warning of deterioration ofthe static performance of steel truss arch is defined as thedifference between the measured and predicted static strains
119890 = 119878119898minus 119878119901 (5)
where 119878119898is the measured static strain 119878III 119878119901 is the predicted
static strain 119878III which is obtained by using the correlationmodels shown in (2) in the healthy condition Figure 17 showsthemean value control chart for static strains of the steel trussarch using measurement data in November 2013 The upperand lower limits are determined using the controlling param-eter 119896 = 2235 Hence a long-term monitoring on the corre-lation models between temperature field and static strains issuitable for early warning of deterioration of the static per-formance of steel truss arch
33 Evaluation of Movement Performance of Pier Longi-tudinal displacements of piers and temperature field datain steel truss arch collected from March 2013 to October2013 are used for evaluation of movement performanceof piers Longitudinal displacement data from sensors 119871
119894
(119894 = 1 2 6) are denoted by 119889119894(119905) (where 119905 means time)
respectively Meanwhile temperature field data from sensors119882119894(119894 = 1 2 8) are denoted by 119879
119894(119905) respectively and
their averaged value 119879(119905) is used to represent the averagetemperature of the steel truss arch For simplicity taking10 minutes as basic time interval the 10 min average dis-placements and temperatures are calculated and used as therepresentative values of this time interval By this way thedisplacements and temperatures in one day are reduced to144 representative samples Furthermore the temperature-displacement scatter diagrams are plotted in Figure 18 Anoverall increase in longitudinal displacements of piers isobserved with an increase in the average temperature of the
The Scientific World Journal 13
UCL
LCL
CL
Con
ditio
n in
dexe
(120583120576)
minus15
minus10
minus5
0
5
10
15
10 20 30 40 50 600Sample
Figure 17 Mean value control chart for static strains of the steel truss arch
Table 3 Summary of linear regression models
Pier Regression function4 pier 119889
1
= minus11524 + 696119879
5 pier 1198892
= minus9253 + 562119879
6 pier 1198893
= minus6628 + 382119879
8 pier 1198894
= minus4658 + 336119879
9 pier 1198895
= minus7605 + 530119879
10 pier 1198896
= minus9850 + 664119879
bridge And the correlation coefficients of the longitudinaldisplacements of piers and average temperature of the steeltruss arch are all larger than 092 as shown in Figure 18
A linear regression analysis between the 10 min averagedisplacement 119889
119894(119905) and the temperature 119879(119905) is further per-
formed to model the temperature-displacement correlationsusing the following equation [14]
119889119894(119905) = 120573
0+ 1205731119879 (119905) (6)
where the regression coefficients 1205731and 120573
0were obtained by
the least-squares method as
1205731=
119878119889119894119879
119878119879119879
1205730= 119889119894minus 1205731119879
(7)
where 119878119889119894119879
is the covariance between the displacement andthe temperature sequences 119878
119879119879is the variance of the mea-
sured temperature sequences and 119879 and 119889119894are the means
of the measured temperature and displacement sequencesrespectively Table 3 summarizes the expressions of linearregression models of the displacement versus temperature
The analysis results indicate that good correlation existsbetween the longitudinal displacements of piers and averagetemperature of steel truss arch A multivariable mean valuecontrol chart is also employed to monitor the measuredabnormal changes in the longitudinal displacements of piersThe condition index 119890 for early warning of abnormal behaviorof piers is defined as the difference between themeasured andpredicted longitudinal displacements
119890 = 119889119898minus 119889119901 (8)
where 119889119898is the measured displacement of the piers 119889
119901is
the predicted displacement of the piers which is obtainedby using the linear regression models shown in Table 3in the healthy condition Figure 19 shows the mean valuecontrol chart for longitudinal displacements of piers usingmeasurement data in March 2013 The upper and lowerlimits are determined using the controlling parameter 119896 =
2804 Hence a long-term monitoring on the temperature-displacement correlation model is suitable for real-timecondition assessment for movement performance of piers
34 Evaluation of Fatigue Performance of Steel Deck Fatiguecracking is universal for orthotropic steel deck of highwaybridges Once the fatigue crack appears extensive fatiguecracking may occur and repair is very difficult Henceinfinite-fatigue-life design method is applied for NanjingDashengguan Bridge instead of finite-fatigue-life designmethod which is applied extensively in highway bridges Theinfinite-fatigue-life design method is to control the fatiguestress amplitude in the range less than the value of fatiguecut-off limit refrained from fatigue cracking Consequentlythe purpose of equipping stain gauges in the SHM system ofDashengguan Bridge is to monitor the stress amplitude andthe validity of fatigue design can be achieved by comparingthe stress amplitude with fatigue stress limit
The structural dynamic response is asymmetrical becauseof the irregularity of railway and braking force from thetrain even though the structure of Nanjing DashengguanBridge is spatially symmetrical Hence the stresses are notsymmetrical measured by strain gauges from the upstreamand downstream directions However the fatigue amplitudesof two strain gauges induced by trains are both extremelysmall Thus the strain history data of the strain gauge DYB-2are merely selected in this paper to illustrate that the fatiguestress amplitude is much lower than the value of fatigue stresslimit The typical strain time-history curves of DYB-2 whensingle train with 8 carriages or 16 carriages passed throughthe bridge are shown in Figure 20 respectively It can be seenthat during the passing of high-speed trains both 8 carriagesand 16 carriages have led to strain variations reaching to 2ndash8 120583120576 And the number of strain peaks is 16 and 32 for trainswith 8 carriages and 16 carriages respectively because theaxle number of each carriage is 2
To perform fatigue evaluation simplified rainflow cyclecounting algorithm was firstly used to process strain history
14 The Scientific World Journal
R = 09748
10 20 30 40 500
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
(a) Correlation between 1198891of the 4 pier and average temperature 119879 of
steel truss arch
10 20 30 40 500
Temperature (∘C)
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
R = 09746
(b) Correlation between 1198892of the 5 pier and average temperature 119879 of
steel truss arch
10 200 40 5030
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
Disp
lace
men
t (m
m)
R = 09471
(c) Correlation between 1198893of the 6 pier and average temperature 119879 of
steel truss arch
R = 09276
10 20 30 40 500
Temperature (∘C)
minus60
minus30
0
30
60
90
120D
ispla
cem
ent (
mm
)
(d) Correlation between 1198894of the 8 pier and average temperature 119879 of
steel truss arch
R = 09536
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(e) Correlation between 1198895of the 9 pier and average temperature 119879 of
steel truss arch
R = 09537
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(f) Correlation between 1198896of the 10 pier and average temperature 119879 of
steel truss arch
Figure 18 Correlation between longitudinal displacement and average temperature
The Scientific World Journal 15
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
)
500 1000 1500 2000 2500 3000 3500 4000 45000Sample
minus15
minus10
minus5
0
5
10
15
Figure 19 Mean value control chart for longitudinal displacements of piers
Stra
in (120583
120576)
minus174
minus172
minus170
minus168
minus166
minus164
minus162
1 2714 53
66
79
92
40
118
209
183
222
105
157
196
144
131
235
248
170
Data point
(a) Single train with 8 carriages passed through the bridge
Stra
in (120583
120576)
minus180minus178minus176minus174minus172minus170minus168minus166
1
61
91
31
181
211
121
421
301
481
511
361
391
271
451
241
331
541
571
151
Data point
(b) Single train with 16 carriages passed through the bridge
Figure 20 Typical strain time-history curves of DYB-2 when single train passed through the bridge
data and the spectrum of stress amplitude was obtained[15] The spectra of stress amplitude calculated using thestrain history data with regard to 8 carriages or 16 carriagesare shown in Figure 21 respectively It can be seen that themaximum stress amplitude is smaller than 10MPa obtainedfrom strain history curves under single trainThen accordingtoMinerrsquos law and standard of Eurocode 3 fatigue parametersof 119878eq (equivalent stress amplitude) and 119873
119904(stress cycle
number) under single train were calculated as follows [16]
119878eq = (
1198991198941198785
119894
sum119899119894
)
15
119873s = sum119899119894
(9)
where 119878119894is 119894th stress amplitude 119899
119894is the number of
stress cycles corresponding to 119878119894 and 15 is slope value of
log 119878- log119873 curve in the scope of lower stress amplitudeaccording to the standard of Eurocode 3 [17]
Using (9) the equivalent stress amplitudes 119878eq with regardto 8 carriages or 16 carriages are 075MPa and 071MParespectively And the stress cycle numbers 119873
119904are 22 and 45
respectively Thus the value of 119878eq with regard to 8 carriagesis similar to that of 16 carriages However the value of 119873
119904
with regard to 8 carriages is about twice as much as that of 16carriages Furthermore as recommended by Eurocode 3 forrib-to-deck welded joint the fatigue limit Δ120590
119871is 29MPa [17]
and the value of 119878eq when the high-speed train passes throughbridge is far less than Δ120590
119871 which indicates that fatigue life of
rib-to-deck welded joint in Nanjing Dashengguan Bridge canbe considered infinite Thus the fatigue performance of steel
deck satisfies the requirement of infinite-fatigue-life designmethod
4 Conclusions
In the last few decades Structural Health Monitoring (SHM)system has been widely installed in many highway bridgesto continuously monitor the structural condition over anextended period of time However application of SHMsystem in high-speed railway bridges is far insufficient Thispaper presents an analysis of the SHM system installedin Nanjing Dashengguan Bridge in China The monitoringsystem is composed of a sensor system a data acquisitionand transmission system a data management system and astructural evaluation system The monitoring system is cur-rently in full operation accumulating the bridgersquos behaviorsunder the varying environmental conditions such as high-speed trains and environmental temperature
Based on long-termhealthmonitoring data general prin-ciples and analytical processes for structural evaluation areprovided and some evaluation results are reported It can beshown from the analysis in this paper that the mathematicalcorrelation models describing the overall structural behaviorof the bridge can be obtained with the support of the healthmonitoring system which includes cross-correlation modelsfor accelerations correlation models between temperatureand static strains of steel truss arch and correlation mod-els between temperature and longitudinal displacements ofpiers The mean value control chart is further employed tomonitor the time series of differences between the measuredstructural parameters and the predicted parameters usingthese correlation models The upper and lower control limits
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
The Scientific World Journal 7Ac
cele
ratio
n (m
ms2)
minus150
minus100
minus50
0
50
100
150
200
1 4001 6001 8001 100012001
Data point
(a) Accelerometer 1198852
Acce
lera
tion
(mm
s2)
1 4001 6001 8001 100012001
Data point
minus80
minus60
minus40
minus20
0
20
40
60
(b) Accelerometer 1198854
Figure 7 Vertical acceleration time histories of the main girder caused by a high-speed train
y = minus02642x2 + 37075x minus 56125
53 4 62 25 5535 45
RMS values of the 1st span (mms2)
0
2
4
6
8
10
RMS
valu
es o
f the
2nd
span
(mm
s2)
(a) Transverse accelerations
y = minus00131x2 + 07296x + 35995
50 15 20 2510
RMS values of the 1st span (mms2)
0
2
4
6
8
10
12
14
16RM
S va
lues
of t
he2
nd sp
an (m
ms2)
(b) Vertical accelerations
Figure 8 Cross-correlations between RMS values of accelerations measured in the middle of two main spans
Fitte
d RM
S va
lues
(mm
s2)
R = 09391
0
2
4
6
8
10
63 54 72 25 35 6555 7545
Measured RMS values (mms2)
(a) Transverse accelerations
R = 09518
4
6
8
10
12
14
16
Fitte
d RM
S va
lues
(mm
s2)
86 12 14 1610
Measured RMS values (mms2)
(b) Vertical accelerations
Figure 9 Fitting effect of cross-correlation between RMS values of accelerations by using quadratic polynomial
8 The Scientific World Journal
0 62 9331 186 217124 310279248155
Day
09
091
092
093
094
095
096
097C
orre
latio
n co
effici
ent
(a) Transverse accelerations
0 60 9030 180 210120 300270240150
Day
09
092
094
096
098
Cor
relat
ion
coeffi
cien
t
(b) Vertical accelerations
Figure 10 Daily computed correlation coefficients in 2013
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
s2)
minus2
minus1
0
1
2
20 40 60 80 100 120 140 1600Sample
Figure 11 Mean value control chart for accelerations in the middle of the second span
The analysis results indicate that good cross-correlationexists between the RMS values of the transverse and verticalaccelerations on the two main spans Multivariable controlchart is used here to monitor the measured changes in thetransverse and vertical accelerations caused by deteriorationof the vibration performance Firstly the condition index 119890
for early warning of abnormal vibration behavior is definedas the difference between the measured and predicted RMSvalues of the accelerations in the middle of the second span
119890 = RMS119898minus RMS
119901 (1)
where RMS119898is themeasured RMS values of the accelerations
in the middle of the second span RMS119901is the predicted
RMS values of the accelerations in the middle of the secondspan which is obtained by taking the RMS values of theaccelerations in the middle of the first span into the fittingformula of the quadratic polynomial in the healthy conditionThen a mean value control chart is employed to monitorthe time series of 119890 with regard to both transverse andvertical accelerations The time series of 119890 taken when thestructural vibration is in good condition will have somedistribution with mean 120583 and variance 120590
2 If the mean andstandard deviation are known a control chart is constructedby drawing a horizontal line CL at 120583 and twomore horizontallines representing the upper and lower control limits Theupper limit UCL is drawn at 120583 + 119896120590 and the lower limit LCL
at 120583 minus 119896120590 For online monitoring the controlling parameter119896 is chosen so that when the structural vibration is in goodcondition all observation samples fall between the controllimits When the new measurement is made the structuralabnormal vibration condition can be detected if an unusualnumber of samples fall beyond the control limits Figure 11shows the mean value control chart using the measuredand predicted RMS values of the accelerations shown inFigure 9 The upper and lower limits are determined usingthe controlling parameter 119896 = 2612 Hence a long-termmonitoring on the cross-correlation model of RMS values ofthe transverse and vertical accelerations can realize the earlywarning of deterioration of the vibration performance
32 Evaluation of Static Performance of Steel Truss Arch
321 Data of Temperature Field and Static Strains Tem-perature data from temperature sensor 119882
119894is denoted by
119879119894 temperature difference data acquired by 119879
119894minus 119879
119895is
denoted by119879119894119895(119879119894119895= 119879119894minus119879119895) and static strain data from strain
sensor119884119894is denoted by 119878
119894(119894 119895 = 1 2 8) As for illustration
the time-history curves of annual change and daily change intemperature 119879
1and temperature difference 119879
12are shown in
Figures 12 and 13 respectively And the time-history curvesof annual change and daily change in static strain data 119878
1
are shown in Figure 14 (negative values denote compressive
The Scientific World Journal 9
4 5 6 7 8 9 10 113Time (month)
minus5
5
15
25
35
45Te
mpe
ratu
re (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
5
10
15
20
Tem
pera
ture
(∘C)
(b) Daily curve (March 4th 2013)
Figure 12 Time-history curves of temperature data 1198791
4 5 6 7 8 9 10 113Time (month)
minus25
minus15
minus5
5
15
Tem
pera
ture
diff
eren
ce (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
minus10
minus5
0
5Te
mpe
ratu
re (∘
C)
(b) Daily curve (March 4th 2013)
Figure 13 Time-history curves of temperature difference data 11987912
4 5 6 7 8 9 10 113Time (month)
minus300
minus200
minus100
0
100
200
Stat
ic st
rain
(120583120576)
(a) Annual curve fromMarch 2013 to October 2013
Sharp peak
4 8 12 16 20 240Time (h)
minus120
minus65
minus10
45
100
Stat
ic st
rain
(120583120576)
(b) Daily curve (March 4th 2013)
Figure 14 Time-history curves of static strain data 1198781
10 The Scientific World Journal
4 8 12 16 20 240Time (h)
Stat
ic st
rain
S III
(120583120576)
minus120
minus65
minus10
45
100
(a) Daily extraction result 119878III of 1198781
4 5 6 7 8 9 10 113Time (month)
Stat
ic st
rain
S III
(120583120576)
minus200
minus130
minus60
10
80
150
(b) Annual extraction result 119878III of 1198781 fromMarch 2013 to October 2013
Figure 15 Extraction results 119878III of static strain data 1198781
static strain and positive values denote tension static strain)From Figure 14 it can be seen that time-history curve of 119878
1
contains many sharp peaks caused by high-speed trains witheach peak corresponding to the time when one train passesthrough the bridge And the static strains caused by high-speed trains are obviously lower than that of temperature fieldincluding temperature data and temperature difference data
As shown in Figure 14 static strain data in steel truss archmainly contains three parts static strain 119878I caused by tem-perature static strain 119878II caused by temperature differenceand static strain 119878III caused by train In order to construct thecorrelation between temperature field and static strain in steeltruss arch static strain 119878I and static strain 119878II (denoted by 119878III)must be extracted from the static strain data In the presentstudy wavelet packet decomposition method is used toextract 119878III considering the high-frequency characteristics ofstatic strain data caused by trains By using the wavelet packetdecomposition static strain data can be decomposed scale byscale into different frequency bands and each decompositioncoefficient corresponds to its frequency band [9 10] Eachdecomposition coefficient can be reconstructed into time-domain signals with constraints of its own frequency bandTaking the daily time-history curve of static strain data1198781shown in Figure 14(b) as an example it is decomposed
within 8 scales and the decomposition coefficient 11986208
119895
isspecially selected and then reconstructed as 119878III shown inFigure 15(a) It can be seen that sharp peaks caused by trainsare removed effectively and the static strains which are causedby temperature field are retained as well Furthermore theannual extraction result 119878III of 1198781 fromMarch 2013 toOctober2013 is shown in Figure 15(b)
322 Correlation Model between Temperature Field andStatic Strains In constructing the correlationmodel betweentemperature field and static strains data from March 2013to October 2013 are chosen as training data and data inNovember 2013 are chosen as test data for verification of
the correlation model The static strain 119878III is expressed asfollows
119878III =8
sum
119894=1
120572119894119879119894+
16
sum
119895=1
120573119895119863119895+ 119888 (2)
where 119879119894is 119894th temperature data from set 119879
1 1198792 1198793 1198794
1198795 1198796 1198797 1198798 119863119895is 119895th temperature difference data from set
11987912 11987934 11987956 11987978 11987913 11987915 11987917 11987935 11987937 11987957 11987924 11987926 11987928 11987946
11987948 11987968 120572119894is 119894th linear expansion coefficient of 119879
119894 120573119895is 119895th
linear expansion coefficient of 119879119894119895 and 119888 is the constant term
Considering that total 24 linear expansion coefficientsand one constant term in (2) are very complicated forcalculation principal component analysis is further usedto simplify (2) which aims at finding small amounts ofcomprehensive factors to represent main information of tem-perature and temperature difference data By using principalcomponent analysis the comprehensive factors can be foundby transforming temperature and temperature difference datainto factor component space [11 12] For temperature data theexplained variance of the first potential comprehensive factor1198751reaches 967 apparently larger than the other potential
comprehensive factors Thus 1198751is selected as the compre-
hensive factor Meanwhile for temperature difference datausing principal component analysis the explained varianceof the first three potential comprehensive factors 119877
1 1198772 and
1198773reaches 988 which are selected as the comprehensive
factorsTherefore the correlation models between 119878III and tem-
perature field can be expressed as follows
119878III = 1205821([119864]1times8
997888
1198798times1
) +9978881205741times3
([119865]3times16
997888
11986316times1
) + 119888 (3)
where997888
1198798times1
denotes one vector containing eight temperaturedata 997888119863
16times1denotes one vector containing sixteen temper-
ature difference data 119863119895(119895 = 1 2 16) [119864]
1times8and
The Scientific World Journal 11
b = minus3632
k = 0946
minus100 minus55 minus10 8035Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus100
minus55
minus10
35
80
Sim
ulat
iveS
III
(120583120576)
(a) 1198781
b = minus2013k = 0898
minus45minus75 minus15 45 7515Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus75
minus45
minus15
15
45
75
Sim
ulat
iveS
III
(120583120576)
(b) 1198782
b = minus1844k = 0904
minus60
minus30
0
30
60
90
Sim
ulat
iveS
III
(120583120576)
minus30minus60 30 60 900Monitoring SIII (120583120576)
Scattered pointsFitting curve
(c) 1198783
b = 2057k = 1126
minus15minus30 15 300Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus40
minus25
minus10
5
20
35Si
mul
ativ
eSII
I(120583120576)
(d) 1198784
b = minus4139k = 0935
minus100
minus65
minus30
5
40
75
Sim
ulat
iveS
III
(120583120576)
minus105 4515 75minus45minus75 minus15
Monitoring SIII (120583120576)
Scattered pointsFitting curve
(e) 1198785
b = minus4094k = 0918
minus145minus205 minus25minus85 9535Monitoring SIII (120583120576)
minus185
minus130
minus75
minus20
35
90
Sim
ulat
iveS
III
(120583120576)
Scattered pointsFitting curve
(f) 1198786
Figure 16 Continued
12 The Scientific World Journal
b = 2148k = 0881
minus50
minus28
minus6
16
38
60Si
mul
ativ
eSII
I(120583120576)
minus45 minus23 minus1 43 6521Monitoring SIII (120583120576)
Scattered pointsFitting curve
(g) 1198787
b = 0615k = 1118
minus30
minus15
0
15
30
Sim
ulat
iveS
III
(120583120576)
minus15minus25 minus5 15 255Monitoring SIII (120583120576)
Scattered pointsFitting curve
(h) 1198788
Figure 16 Correlation between simulative 119878III and monitoring 119878III
Table 2 Performance parameters 1205821
1205741
1205742
1205743
and 119862 in correlationmodels
Static straindata 120582
1
1205741
1205742
1205743
119862
119878III of 1198781 0454 minus4711 minus2597 minus4944 minus46806119878III of 1198782 1585 0013 4104 minus1247 minus70238119878III of 1198783 0301 3416 minus2098 2968 minus6884119878III of 1198784 0350 minus0734 0351 0948 minus23597119878III of 1198785 0179 minus4219 1827 3165 minus22049119878III of 1198786 0062 minus3245 8242 minus5398 minus22862119878III of 1198787 minus0367 2411 minus0598 5531 17075119878III of 1198788 0611 2188 minus1130 2018 minus32799
[119865]3times16
denote transformation matrix which can be cal-culated using principal component analysis [11 12] 120582
1is
performance parameter of first potential comprehensivefactor 119875
1 9978881205741times3
= [1205741 1205742 1205743] is one vector containing
three performance parameters corresponding to first threepotential comprehensive factors 119877
1 1198772 and 119877
3 The values of
performance parameters 1205821 9978881205741times3
and 119888 are estimated usingtraining data by multivariate linear regression method [13]and are shown in Table 2
In order to verify the effectiveness of the correlationmodels shown in (2) test data in November 2013 are usedto obtain the simulative 119878III The scattered points betweensimulative 119878III and monitoring 119878III are plotted as shown inFigure 16 which are further fitted by linear function
119878119904= 119896119878119898+ 119887 (4)
where 119878119904denotes simulative 119878III 119878119898 denotes monitoring
119878III and 119896 119887 are fitting parameters with the fitting resultsshown in Figure 16 as well The fitting results verify theeffectiveness of the correlationmodels Amultivariable mean
value control chart is also employed tomonitor the measuredabnormal changes in the static strains of the steel truss archThe condition index 119890 for early warning of deterioration ofthe static performance of steel truss arch is defined as thedifference between the measured and predicted static strains
119890 = 119878119898minus 119878119901 (5)
where 119878119898is the measured static strain 119878III 119878119901 is the predicted
static strain 119878III which is obtained by using the correlationmodels shown in (2) in the healthy condition Figure 17 showsthemean value control chart for static strains of the steel trussarch using measurement data in November 2013 The upperand lower limits are determined using the controlling param-eter 119896 = 2235 Hence a long-term monitoring on the corre-lation models between temperature field and static strains issuitable for early warning of deterioration of the static per-formance of steel truss arch
33 Evaluation of Movement Performance of Pier Longi-tudinal displacements of piers and temperature field datain steel truss arch collected from March 2013 to October2013 are used for evaluation of movement performanceof piers Longitudinal displacement data from sensors 119871
119894
(119894 = 1 2 6) are denoted by 119889119894(119905) (where 119905 means time)
respectively Meanwhile temperature field data from sensors119882119894(119894 = 1 2 8) are denoted by 119879
119894(119905) respectively and
their averaged value 119879(119905) is used to represent the averagetemperature of the steel truss arch For simplicity taking10 minutes as basic time interval the 10 min average dis-placements and temperatures are calculated and used as therepresentative values of this time interval By this way thedisplacements and temperatures in one day are reduced to144 representative samples Furthermore the temperature-displacement scatter diagrams are plotted in Figure 18 Anoverall increase in longitudinal displacements of piers isobserved with an increase in the average temperature of the
The Scientific World Journal 13
UCL
LCL
CL
Con
ditio
n in
dexe
(120583120576)
minus15
minus10
minus5
0
5
10
15
10 20 30 40 50 600Sample
Figure 17 Mean value control chart for static strains of the steel truss arch
Table 3 Summary of linear regression models
Pier Regression function4 pier 119889
1
= minus11524 + 696119879
5 pier 1198892
= minus9253 + 562119879
6 pier 1198893
= minus6628 + 382119879
8 pier 1198894
= minus4658 + 336119879
9 pier 1198895
= minus7605 + 530119879
10 pier 1198896
= minus9850 + 664119879
bridge And the correlation coefficients of the longitudinaldisplacements of piers and average temperature of the steeltruss arch are all larger than 092 as shown in Figure 18
A linear regression analysis between the 10 min averagedisplacement 119889
119894(119905) and the temperature 119879(119905) is further per-
formed to model the temperature-displacement correlationsusing the following equation [14]
119889119894(119905) = 120573
0+ 1205731119879 (119905) (6)
where the regression coefficients 1205731and 120573
0were obtained by
the least-squares method as
1205731=
119878119889119894119879
119878119879119879
1205730= 119889119894minus 1205731119879
(7)
where 119878119889119894119879
is the covariance between the displacement andthe temperature sequences 119878
119879119879is the variance of the mea-
sured temperature sequences and 119879 and 119889119894are the means
of the measured temperature and displacement sequencesrespectively Table 3 summarizes the expressions of linearregression models of the displacement versus temperature
The analysis results indicate that good correlation existsbetween the longitudinal displacements of piers and averagetemperature of steel truss arch A multivariable mean valuecontrol chart is also employed to monitor the measuredabnormal changes in the longitudinal displacements of piersThe condition index 119890 for early warning of abnormal behaviorof piers is defined as the difference between themeasured andpredicted longitudinal displacements
119890 = 119889119898minus 119889119901 (8)
where 119889119898is the measured displacement of the piers 119889
119901is
the predicted displacement of the piers which is obtainedby using the linear regression models shown in Table 3in the healthy condition Figure 19 shows the mean valuecontrol chart for longitudinal displacements of piers usingmeasurement data in March 2013 The upper and lowerlimits are determined using the controlling parameter 119896 =
2804 Hence a long-term monitoring on the temperature-displacement correlation model is suitable for real-timecondition assessment for movement performance of piers
34 Evaluation of Fatigue Performance of Steel Deck Fatiguecracking is universal for orthotropic steel deck of highwaybridges Once the fatigue crack appears extensive fatiguecracking may occur and repair is very difficult Henceinfinite-fatigue-life design method is applied for NanjingDashengguan Bridge instead of finite-fatigue-life designmethod which is applied extensively in highway bridges Theinfinite-fatigue-life design method is to control the fatiguestress amplitude in the range less than the value of fatiguecut-off limit refrained from fatigue cracking Consequentlythe purpose of equipping stain gauges in the SHM system ofDashengguan Bridge is to monitor the stress amplitude andthe validity of fatigue design can be achieved by comparingthe stress amplitude with fatigue stress limit
The structural dynamic response is asymmetrical becauseof the irregularity of railway and braking force from thetrain even though the structure of Nanjing DashengguanBridge is spatially symmetrical Hence the stresses are notsymmetrical measured by strain gauges from the upstreamand downstream directions However the fatigue amplitudesof two strain gauges induced by trains are both extremelysmall Thus the strain history data of the strain gauge DYB-2are merely selected in this paper to illustrate that the fatiguestress amplitude is much lower than the value of fatigue stresslimit The typical strain time-history curves of DYB-2 whensingle train with 8 carriages or 16 carriages passed throughthe bridge are shown in Figure 20 respectively It can be seenthat during the passing of high-speed trains both 8 carriagesand 16 carriages have led to strain variations reaching to 2ndash8 120583120576 And the number of strain peaks is 16 and 32 for trainswith 8 carriages and 16 carriages respectively because theaxle number of each carriage is 2
To perform fatigue evaluation simplified rainflow cyclecounting algorithm was firstly used to process strain history
14 The Scientific World Journal
R = 09748
10 20 30 40 500
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
(a) Correlation between 1198891of the 4 pier and average temperature 119879 of
steel truss arch
10 20 30 40 500
Temperature (∘C)
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
R = 09746
(b) Correlation between 1198892of the 5 pier and average temperature 119879 of
steel truss arch
10 200 40 5030
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
Disp
lace
men
t (m
m)
R = 09471
(c) Correlation between 1198893of the 6 pier and average temperature 119879 of
steel truss arch
R = 09276
10 20 30 40 500
Temperature (∘C)
minus60
minus30
0
30
60
90
120D
ispla
cem
ent (
mm
)
(d) Correlation between 1198894of the 8 pier and average temperature 119879 of
steel truss arch
R = 09536
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(e) Correlation between 1198895of the 9 pier and average temperature 119879 of
steel truss arch
R = 09537
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(f) Correlation between 1198896of the 10 pier and average temperature 119879 of
steel truss arch
Figure 18 Correlation between longitudinal displacement and average temperature
The Scientific World Journal 15
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
)
500 1000 1500 2000 2500 3000 3500 4000 45000Sample
minus15
minus10
minus5
0
5
10
15
Figure 19 Mean value control chart for longitudinal displacements of piers
Stra
in (120583
120576)
minus174
minus172
minus170
minus168
minus166
minus164
minus162
1 2714 53
66
79
92
40
118
209
183
222
105
157
196
144
131
235
248
170
Data point
(a) Single train with 8 carriages passed through the bridge
Stra
in (120583
120576)
minus180minus178minus176minus174minus172minus170minus168minus166
1
61
91
31
181
211
121
421
301
481
511
361
391
271
451
241
331
541
571
151
Data point
(b) Single train with 16 carriages passed through the bridge
Figure 20 Typical strain time-history curves of DYB-2 when single train passed through the bridge
data and the spectrum of stress amplitude was obtained[15] The spectra of stress amplitude calculated using thestrain history data with regard to 8 carriages or 16 carriagesare shown in Figure 21 respectively It can be seen that themaximum stress amplitude is smaller than 10MPa obtainedfrom strain history curves under single trainThen accordingtoMinerrsquos law and standard of Eurocode 3 fatigue parametersof 119878eq (equivalent stress amplitude) and 119873
119904(stress cycle
number) under single train were calculated as follows [16]
119878eq = (
1198991198941198785
119894
sum119899119894
)
15
119873s = sum119899119894
(9)
where 119878119894is 119894th stress amplitude 119899
119894is the number of
stress cycles corresponding to 119878119894 and 15 is slope value of
log 119878- log119873 curve in the scope of lower stress amplitudeaccording to the standard of Eurocode 3 [17]
Using (9) the equivalent stress amplitudes 119878eq with regardto 8 carriages or 16 carriages are 075MPa and 071MParespectively And the stress cycle numbers 119873
119904are 22 and 45
respectively Thus the value of 119878eq with regard to 8 carriagesis similar to that of 16 carriages However the value of 119873
119904
with regard to 8 carriages is about twice as much as that of 16carriages Furthermore as recommended by Eurocode 3 forrib-to-deck welded joint the fatigue limit Δ120590
119871is 29MPa [17]
and the value of 119878eq when the high-speed train passes throughbridge is far less than Δ120590
119871 which indicates that fatigue life of
rib-to-deck welded joint in Nanjing Dashengguan Bridge canbe considered infinite Thus the fatigue performance of steel
deck satisfies the requirement of infinite-fatigue-life designmethod
4 Conclusions
In the last few decades Structural Health Monitoring (SHM)system has been widely installed in many highway bridgesto continuously monitor the structural condition over anextended period of time However application of SHMsystem in high-speed railway bridges is far insufficient Thispaper presents an analysis of the SHM system installedin Nanjing Dashengguan Bridge in China The monitoringsystem is composed of a sensor system a data acquisitionand transmission system a data management system and astructural evaluation system The monitoring system is cur-rently in full operation accumulating the bridgersquos behaviorsunder the varying environmental conditions such as high-speed trains and environmental temperature
Based on long-termhealthmonitoring data general prin-ciples and analytical processes for structural evaluation areprovided and some evaluation results are reported It can beshown from the analysis in this paper that the mathematicalcorrelation models describing the overall structural behaviorof the bridge can be obtained with the support of the healthmonitoring system which includes cross-correlation modelsfor accelerations correlation models between temperatureand static strains of steel truss arch and correlation mod-els between temperature and longitudinal displacements ofpiers The mean value control chart is further employed tomonitor the time series of differences between the measuredstructural parameters and the predicted parameters usingthese correlation models The upper and lower control limits
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
8 The Scientific World Journal
0 62 9331 186 217124 310279248155
Day
09
091
092
093
094
095
096
097C
orre
latio
n co
effici
ent
(a) Transverse accelerations
0 60 9030 180 210120 300270240150
Day
09
092
094
096
098
Cor
relat
ion
coeffi
cien
t
(b) Vertical accelerations
Figure 10 Daily computed correlation coefficients in 2013
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
s2)
minus2
minus1
0
1
2
20 40 60 80 100 120 140 1600Sample
Figure 11 Mean value control chart for accelerations in the middle of the second span
The analysis results indicate that good cross-correlationexists between the RMS values of the transverse and verticalaccelerations on the two main spans Multivariable controlchart is used here to monitor the measured changes in thetransverse and vertical accelerations caused by deteriorationof the vibration performance Firstly the condition index 119890
for early warning of abnormal vibration behavior is definedas the difference between the measured and predicted RMSvalues of the accelerations in the middle of the second span
119890 = RMS119898minus RMS
119901 (1)
where RMS119898is themeasured RMS values of the accelerations
in the middle of the second span RMS119901is the predicted
RMS values of the accelerations in the middle of the secondspan which is obtained by taking the RMS values of theaccelerations in the middle of the first span into the fittingformula of the quadratic polynomial in the healthy conditionThen a mean value control chart is employed to monitorthe time series of 119890 with regard to both transverse andvertical accelerations The time series of 119890 taken when thestructural vibration is in good condition will have somedistribution with mean 120583 and variance 120590
2 If the mean andstandard deviation are known a control chart is constructedby drawing a horizontal line CL at 120583 and twomore horizontallines representing the upper and lower control limits Theupper limit UCL is drawn at 120583 + 119896120590 and the lower limit LCL
at 120583 minus 119896120590 For online monitoring the controlling parameter119896 is chosen so that when the structural vibration is in goodcondition all observation samples fall between the controllimits When the new measurement is made the structuralabnormal vibration condition can be detected if an unusualnumber of samples fall beyond the control limits Figure 11shows the mean value control chart using the measuredand predicted RMS values of the accelerations shown inFigure 9 The upper and lower limits are determined usingthe controlling parameter 119896 = 2612 Hence a long-termmonitoring on the cross-correlation model of RMS values ofthe transverse and vertical accelerations can realize the earlywarning of deterioration of the vibration performance
32 Evaluation of Static Performance of Steel Truss Arch
321 Data of Temperature Field and Static Strains Tem-perature data from temperature sensor 119882
119894is denoted by
119879119894 temperature difference data acquired by 119879
119894minus 119879
119895is
denoted by119879119894119895(119879119894119895= 119879119894minus119879119895) and static strain data from strain
sensor119884119894is denoted by 119878
119894(119894 119895 = 1 2 8) As for illustration
the time-history curves of annual change and daily change intemperature 119879
1and temperature difference 119879
12are shown in
Figures 12 and 13 respectively And the time-history curvesof annual change and daily change in static strain data 119878
1
are shown in Figure 14 (negative values denote compressive
The Scientific World Journal 9
4 5 6 7 8 9 10 113Time (month)
minus5
5
15
25
35
45Te
mpe
ratu
re (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
5
10
15
20
Tem
pera
ture
(∘C)
(b) Daily curve (March 4th 2013)
Figure 12 Time-history curves of temperature data 1198791
4 5 6 7 8 9 10 113Time (month)
minus25
minus15
minus5
5
15
Tem
pera
ture
diff
eren
ce (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
minus10
minus5
0
5Te
mpe
ratu
re (∘
C)
(b) Daily curve (March 4th 2013)
Figure 13 Time-history curves of temperature difference data 11987912
4 5 6 7 8 9 10 113Time (month)
minus300
minus200
minus100
0
100
200
Stat
ic st
rain
(120583120576)
(a) Annual curve fromMarch 2013 to October 2013
Sharp peak
4 8 12 16 20 240Time (h)
minus120
minus65
minus10
45
100
Stat
ic st
rain
(120583120576)
(b) Daily curve (March 4th 2013)
Figure 14 Time-history curves of static strain data 1198781
10 The Scientific World Journal
4 8 12 16 20 240Time (h)
Stat
ic st
rain
S III
(120583120576)
minus120
minus65
minus10
45
100
(a) Daily extraction result 119878III of 1198781
4 5 6 7 8 9 10 113Time (month)
Stat
ic st
rain
S III
(120583120576)
minus200
minus130
minus60
10
80
150
(b) Annual extraction result 119878III of 1198781 fromMarch 2013 to October 2013
Figure 15 Extraction results 119878III of static strain data 1198781
static strain and positive values denote tension static strain)From Figure 14 it can be seen that time-history curve of 119878
1
contains many sharp peaks caused by high-speed trains witheach peak corresponding to the time when one train passesthrough the bridge And the static strains caused by high-speed trains are obviously lower than that of temperature fieldincluding temperature data and temperature difference data
As shown in Figure 14 static strain data in steel truss archmainly contains three parts static strain 119878I caused by tem-perature static strain 119878II caused by temperature differenceand static strain 119878III caused by train In order to construct thecorrelation between temperature field and static strain in steeltruss arch static strain 119878I and static strain 119878II (denoted by 119878III)must be extracted from the static strain data In the presentstudy wavelet packet decomposition method is used toextract 119878III considering the high-frequency characteristics ofstatic strain data caused by trains By using the wavelet packetdecomposition static strain data can be decomposed scale byscale into different frequency bands and each decompositioncoefficient corresponds to its frequency band [9 10] Eachdecomposition coefficient can be reconstructed into time-domain signals with constraints of its own frequency bandTaking the daily time-history curve of static strain data1198781shown in Figure 14(b) as an example it is decomposed
within 8 scales and the decomposition coefficient 11986208
119895
isspecially selected and then reconstructed as 119878III shown inFigure 15(a) It can be seen that sharp peaks caused by trainsare removed effectively and the static strains which are causedby temperature field are retained as well Furthermore theannual extraction result 119878III of 1198781 fromMarch 2013 toOctober2013 is shown in Figure 15(b)
322 Correlation Model between Temperature Field andStatic Strains In constructing the correlationmodel betweentemperature field and static strains data from March 2013to October 2013 are chosen as training data and data inNovember 2013 are chosen as test data for verification of
the correlation model The static strain 119878III is expressed asfollows
119878III =8
sum
119894=1
120572119894119879119894+
16
sum
119895=1
120573119895119863119895+ 119888 (2)
where 119879119894is 119894th temperature data from set 119879
1 1198792 1198793 1198794
1198795 1198796 1198797 1198798 119863119895is 119895th temperature difference data from set
11987912 11987934 11987956 11987978 11987913 11987915 11987917 11987935 11987937 11987957 11987924 11987926 11987928 11987946
11987948 11987968 120572119894is 119894th linear expansion coefficient of 119879
119894 120573119895is 119895th
linear expansion coefficient of 119879119894119895 and 119888 is the constant term
Considering that total 24 linear expansion coefficientsand one constant term in (2) are very complicated forcalculation principal component analysis is further usedto simplify (2) which aims at finding small amounts ofcomprehensive factors to represent main information of tem-perature and temperature difference data By using principalcomponent analysis the comprehensive factors can be foundby transforming temperature and temperature difference datainto factor component space [11 12] For temperature data theexplained variance of the first potential comprehensive factor1198751reaches 967 apparently larger than the other potential
comprehensive factors Thus 1198751is selected as the compre-
hensive factor Meanwhile for temperature difference datausing principal component analysis the explained varianceof the first three potential comprehensive factors 119877
1 1198772 and
1198773reaches 988 which are selected as the comprehensive
factorsTherefore the correlation models between 119878III and tem-
perature field can be expressed as follows
119878III = 1205821([119864]1times8
997888
1198798times1
) +9978881205741times3
([119865]3times16
997888
11986316times1
) + 119888 (3)
where997888
1198798times1
denotes one vector containing eight temperaturedata 997888119863
16times1denotes one vector containing sixteen temper-
ature difference data 119863119895(119895 = 1 2 16) [119864]
1times8and
The Scientific World Journal 11
b = minus3632
k = 0946
minus100 minus55 minus10 8035Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus100
minus55
minus10
35
80
Sim
ulat
iveS
III
(120583120576)
(a) 1198781
b = minus2013k = 0898
minus45minus75 minus15 45 7515Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus75
minus45
minus15
15
45
75
Sim
ulat
iveS
III
(120583120576)
(b) 1198782
b = minus1844k = 0904
minus60
minus30
0
30
60
90
Sim
ulat
iveS
III
(120583120576)
minus30minus60 30 60 900Monitoring SIII (120583120576)
Scattered pointsFitting curve
(c) 1198783
b = 2057k = 1126
minus15minus30 15 300Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus40
minus25
minus10
5
20
35Si
mul
ativ
eSII
I(120583120576)
(d) 1198784
b = minus4139k = 0935
minus100
minus65
minus30
5
40
75
Sim
ulat
iveS
III
(120583120576)
minus105 4515 75minus45minus75 minus15
Monitoring SIII (120583120576)
Scattered pointsFitting curve
(e) 1198785
b = minus4094k = 0918
minus145minus205 minus25minus85 9535Monitoring SIII (120583120576)
minus185
minus130
minus75
minus20
35
90
Sim
ulat
iveS
III
(120583120576)
Scattered pointsFitting curve
(f) 1198786
Figure 16 Continued
12 The Scientific World Journal
b = 2148k = 0881
minus50
minus28
minus6
16
38
60Si
mul
ativ
eSII
I(120583120576)
minus45 minus23 minus1 43 6521Monitoring SIII (120583120576)
Scattered pointsFitting curve
(g) 1198787
b = 0615k = 1118
minus30
minus15
0
15
30
Sim
ulat
iveS
III
(120583120576)
minus15minus25 minus5 15 255Monitoring SIII (120583120576)
Scattered pointsFitting curve
(h) 1198788
Figure 16 Correlation between simulative 119878III and monitoring 119878III
Table 2 Performance parameters 1205821
1205741
1205742
1205743
and 119862 in correlationmodels
Static straindata 120582
1
1205741
1205742
1205743
119862
119878III of 1198781 0454 minus4711 minus2597 minus4944 minus46806119878III of 1198782 1585 0013 4104 minus1247 minus70238119878III of 1198783 0301 3416 minus2098 2968 minus6884119878III of 1198784 0350 minus0734 0351 0948 minus23597119878III of 1198785 0179 minus4219 1827 3165 minus22049119878III of 1198786 0062 minus3245 8242 minus5398 minus22862119878III of 1198787 minus0367 2411 minus0598 5531 17075119878III of 1198788 0611 2188 minus1130 2018 minus32799
[119865]3times16
denote transformation matrix which can be cal-culated using principal component analysis [11 12] 120582
1is
performance parameter of first potential comprehensivefactor 119875
1 9978881205741times3
= [1205741 1205742 1205743] is one vector containing
three performance parameters corresponding to first threepotential comprehensive factors 119877
1 1198772 and 119877
3 The values of
performance parameters 1205821 9978881205741times3
and 119888 are estimated usingtraining data by multivariate linear regression method [13]and are shown in Table 2
In order to verify the effectiveness of the correlationmodels shown in (2) test data in November 2013 are usedto obtain the simulative 119878III The scattered points betweensimulative 119878III and monitoring 119878III are plotted as shown inFigure 16 which are further fitted by linear function
119878119904= 119896119878119898+ 119887 (4)
where 119878119904denotes simulative 119878III 119878119898 denotes monitoring
119878III and 119896 119887 are fitting parameters with the fitting resultsshown in Figure 16 as well The fitting results verify theeffectiveness of the correlationmodels Amultivariable mean
value control chart is also employed tomonitor the measuredabnormal changes in the static strains of the steel truss archThe condition index 119890 for early warning of deterioration ofthe static performance of steel truss arch is defined as thedifference between the measured and predicted static strains
119890 = 119878119898minus 119878119901 (5)
where 119878119898is the measured static strain 119878III 119878119901 is the predicted
static strain 119878III which is obtained by using the correlationmodels shown in (2) in the healthy condition Figure 17 showsthemean value control chart for static strains of the steel trussarch using measurement data in November 2013 The upperand lower limits are determined using the controlling param-eter 119896 = 2235 Hence a long-term monitoring on the corre-lation models between temperature field and static strains issuitable for early warning of deterioration of the static per-formance of steel truss arch
33 Evaluation of Movement Performance of Pier Longi-tudinal displacements of piers and temperature field datain steel truss arch collected from March 2013 to October2013 are used for evaluation of movement performanceof piers Longitudinal displacement data from sensors 119871
119894
(119894 = 1 2 6) are denoted by 119889119894(119905) (where 119905 means time)
respectively Meanwhile temperature field data from sensors119882119894(119894 = 1 2 8) are denoted by 119879
119894(119905) respectively and
their averaged value 119879(119905) is used to represent the averagetemperature of the steel truss arch For simplicity taking10 minutes as basic time interval the 10 min average dis-placements and temperatures are calculated and used as therepresentative values of this time interval By this way thedisplacements and temperatures in one day are reduced to144 representative samples Furthermore the temperature-displacement scatter diagrams are plotted in Figure 18 Anoverall increase in longitudinal displacements of piers isobserved with an increase in the average temperature of the
The Scientific World Journal 13
UCL
LCL
CL
Con
ditio
n in
dexe
(120583120576)
minus15
minus10
minus5
0
5
10
15
10 20 30 40 50 600Sample
Figure 17 Mean value control chart for static strains of the steel truss arch
Table 3 Summary of linear regression models
Pier Regression function4 pier 119889
1
= minus11524 + 696119879
5 pier 1198892
= minus9253 + 562119879
6 pier 1198893
= minus6628 + 382119879
8 pier 1198894
= minus4658 + 336119879
9 pier 1198895
= minus7605 + 530119879
10 pier 1198896
= minus9850 + 664119879
bridge And the correlation coefficients of the longitudinaldisplacements of piers and average temperature of the steeltruss arch are all larger than 092 as shown in Figure 18
A linear regression analysis between the 10 min averagedisplacement 119889
119894(119905) and the temperature 119879(119905) is further per-
formed to model the temperature-displacement correlationsusing the following equation [14]
119889119894(119905) = 120573
0+ 1205731119879 (119905) (6)
where the regression coefficients 1205731and 120573
0were obtained by
the least-squares method as
1205731=
119878119889119894119879
119878119879119879
1205730= 119889119894minus 1205731119879
(7)
where 119878119889119894119879
is the covariance between the displacement andthe temperature sequences 119878
119879119879is the variance of the mea-
sured temperature sequences and 119879 and 119889119894are the means
of the measured temperature and displacement sequencesrespectively Table 3 summarizes the expressions of linearregression models of the displacement versus temperature
The analysis results indicate that good correlation existsbetween the longitudinal displacements of piers and averagetemperature of steel truss arch A multivariable mean valuecontrol chart is also employed to monitor the measuredabnormal changes in the longitudinal displacements of piersThe condition index 119890 for early warning of abnormal behaviorof piers is defined as the difference between themeasured andpredicted longitudinal displacements
119890 = 119889119898minus 119889119901 (8)
where 119889119898is the measured displacement of the piers 119889
119901is
the predicted displacement of the piers which is obtainedby using the linear regression models shown in Table 3in the healthy condition Figure 19 shows the mean valuecontrol chart for longitudinal displacements of piers usingmeasurement data in March 2013 The upper and lowerlimits are determined using the controlling parameter 119896 =
2804 Hence a long-term monitoring on the temperature-displacement correlation model is suitable for real-timecondition assessment for movement performance of piers
34 Evaluation of Fatigue Performance of Steel Deck Fatiguecracking is universal for orthotropic steel deck of highwaybridges Once the fatigue crack appears extensive fatiguecracking may occur and repair is very difficult Henceinfinite-fatigue-life design method is applied for NanjingDashengguan Bridge instead of finite-fatigue-life designmethod which is applied extensively in highway bridges Theinfinite-fatigue-life design method is to control the fatiguestress amplitude in the range less than the value of fatiguecut-off limit refrained from fatigue cracking Consequentlythe purpose of equipping stain gauges in the SHM system ofDashengguan Bridge is to monitor the stress amplitude andthe validity of fatigue design can be achieved by comparingthe stress amplitude with fatigue stress limit
The structural dynamic response is asymmetrical becauseof the irregularity of railway and braking force from thetrain even though the structure of Nanjing DashengguanBridge is spatially symmetrical Hence the stresses are notsymmetrical measured by strain gauges from the upstreamand downstream directions However the fatigue amplitudesof two strain gauges induced by trains are both extremelysmall Thus the strain history data of the strain gauge DYB-2are merely selected in this paper to illustrate that the fatiguestress amplitude is much lower than the value of fatigue stresslimit The typical strain time-history curves of DYB-2 whensingle train with 8 carriages or 16 carriages passed throughthe bridge are shown in Figure 20 respectively It can be seenthat during the passing of high-speed trains both 8 carriagesand 16 carriages have led to strain variations reaching to 2ndash8 120583120576 And the number of strain peaks is 16 and 32 for trainswith 8 carriages and 16 carriages respectively because theaxle number of each carriage is 2
To perform fatigue evaluation simplified rainflow cyclecounting algorithm was firstly used to process strain history
14 The Scientific World Journal
R = 09748
10 20 30 40 500
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
(a) Correlation between 1198891of the 4 pier and average temperature 119879 of
steel truss arch
10 20 30 40 500
Temperature (∘C)
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
R = 09746
(b) Correlation between 1198892of the 5 pier and average temperature 119879 of
steel truss arch
10 200 40 5030
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
Disp
lace
men
t (m
m)
R = 09471
(c) Correlation between 1198893of the 6 pier and average temperature 119879 of
steel truss arch
R = 09276
10 20 30 40 500
Temperature (∘C)
minus60
minus30
0
30
60
90
120D
ispla
cem
ent (
mm
)
(d) Correlation between 1198894of the 8 pier and average temperature 119879 of
steel truss arch
R = 09536
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(e) Correlation between 1198895of the 9 pier and average temperature 119879 of
steel truss arch
R = 09537
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(f) Correlation between 1198896of the 10 pier and average temperature 119879 of
steel truss arch
Figure 18 Correlation between longitudinal displacement and average temperature
The Scientific World Journal 15
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
)
500 1000 1500 2000 2500 3000 3500 4000 45000Sample
minus15
minus10
minus5
0
5
10
15
Figure 19 Mean value control chart for longitudinal displacements of piers
Stra
in (120583
120576)
minus174
minus172
minus170
minus168
minus166
minus164
minus162
1 2714 53
66
79
92
40
118
209
183
222
105
157
196
144
131
235
248
170
Data point
(a) Single train with 8 carriages passed through the bridge
Stra
in (120583
120576)
minus180minus178minus176minus174minus172minus170minus168minus166
1
61
91
31
181
211
121
421
301
481
511
361
391
271
451
241
331
541
571
151
Data point
(b) Single train with 16 carriages passed through the bridge
Figure 20 Typical strain time-history curves of DYB-2 when single train passed through the bridge
data and the spectrum of stress amplitude was obtained[15] The spectra of stress amplitude calculated using thestrain history data with regard to 8 carriages or 16 carriagesare shown in Figure 21 respectively It can be seen that themaximum stress amplitude is smaller than 10MPa obtainedfrom strain history curves under single trainThen accordingtoMinerrsquos law and standard of Eurocode 3 fatigue parametersof 119878eq (equivalent stress amplitude) and 119873
119904(stress cycle
number) under single train were calculated as follows [16]
119878eq = (
1198991198941198785
119894
sum119899119894
)
15
119873s = sum119899119894
(9)
where 119878119894is 119894th stress amplitude 119899
119894is the number of
stress cycles corresponding to 119878119894 and 15 is slope value of
log 119878- log119873 curve in the scope of lower stress amplitudeaccording to the standard of Eurocode 3 [17]
Using (9) the equivalent stress amplitudes 119878eq with regardto 8 carriages or 16 carriages are 075MPa and 071MParespectively And the stress cycle numbers 119873
119904are 22 and 45
respectively Thus the value of 119878eq with regard to 8 carriagesis similar to that of 16 carriages However the value of 119873
119904
with regard to 8 carriages is about twice as much as that of 16carriages Furthermore as recommended by Eurocode 3 forrib-to-deck welded joint the fatigue limit Δ120590
119871is 29MPa [17]
and the value of 119878eq when the high-speed train passes throughbridge is far less than Δ120590
119871 which indicates that fatigue life of
rib-to-deck welded joint in Nanjing Dashengguan Bridge canbe considered infinite Thus the fatigue performance of steel
deck satisfies the requirement of infinite-fatigue-life designmethod
4 Conclusions
In the last few decades Structural Health Monitoring (SHM)system has been widely installed in many highway bridgesto continuously monitor the structural condition over anextended period of time However application of SHMsystem in high-speed railway bridges is far insufficient Thispaper presents an analysis of the SHM system installedin Nanjing Dashengguan Bridge in China The monitoringsystem is composed of a sensor system a data acquisitionand transmission system a data management system and astructural evaluation system The monitoring system is cur-rently in full operation accumulating the bridgersquos behaviorsunder the varying environmental conditions such as high-speed trains and environmental temperature
Based on long-termhealthmonitoring data general prin-ciples and analytical processes for structural evaluation areprovided and some evaluation results are reported It can beshown from the analysis in this paper that the mathematicalcorrelation models describing the overall structural behaviorof the bridge can be obtained with the support of the healthmonitoring system which includes cross-correlation modelsfor accelerations correlation models between temperatureand static strains of steel truss arch and correlation mod-els between temperature and longitudinal displacements ofpiers The mean value control chart is further employed tomonitor the time series of differences between the measuredstructural parameters and the predicted parameters usingthese correlation models The upper and lower control limits
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
The Scientific World Journal 9
4 5 6 7 8 9 10 113Time (month)
minus5
5
15
25
35
45Te
mpe
ratu
re (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
5
10
15
20
Tem
pera
ture
(∘C)
(b) Daily curve (March 4th 2013)
Figure 12 Time-history curves of temperature data 1198791
4 5 6 7 8 9 10 113Time (month)
minus25
minus15
minus5
5
15
Tem
pera
ture
diff
eren
ce (∘
C)
(a) Annual curve fromMarch 2013 to October 2013
4 8 12 16 20 240Time (h)
minus10
minus5
0
5Te
mpe
ratu
re (∘
C)
(b) Daily curve (March 4th 2013)
Figure 13 Time-history curves of temperature difference data 11987912
4 5 6 7 8 9 10 113Time (month)
minus300
minus200
minus100
0
100
200
Stat
ic st
rain
(120583120576)
(a) Annual curve fromMarch 2013 to October 2013
Sharp peak
4 8 12 16 20 240Time (h)
minus120
minus65
minus10
45
100
Stat
ic st
rain
(120583120576)
(b) Daily curve (March 4th 2013)
Figure 14 Time-history curves of static strain data 1198781
10 The Scientific World Journal
4 8 12 16 20 240Time (h)
Stat
ic st
rain
S III
(120583120576)
minus120
minus65
minus10
45
100
(a) Daily extraction result 119878III of 1198781
4 5 6 7 8 9 10 113Time (month)
Stat
ic st
rain
S III
(120583120576)
minus200
minus130
minus60
10
80
150
(b) Annual extraction result 119878III of 1198781 fromMarch 2013 to October 2013
Figure 15 Extraction results 119878III of static strain data 1198781
static strain and positive values denote tension static strain)From Figure 14 it can be seen that time-history curve of 119878
1
contains many sharp peaks caused by high-speed trains witheach peak corresponding to the time when one train passesthrough the bridge And the static strains caused by high-speed trains are obviously lower than that of temperature fieldincluding temperature data and temperature difference data
As shown in Figure 14 static strain data in steel truss archmainly contains three parts static strain 119878I caused by tem-perature static strain 119878II caused by temperature differenceand static strain 119878III caused by train In order to construct thecorrelation between temperature field and static strain in steeltruss arch static strain 119878I and static strain 119878II (denoted by 119878III)must be extracted from the static strain data In the presentstudy wavelet packet decomposition method is used toextract 119878III considering the high-frequency characteristics ofstatic strain data caused by trains By using the wavelet packetdecomposition static strain data can be decomposed scale byscale into different frequency bands and each decompositioncoefficient corresponds to its frequency band [9 10] Eachdecomposition coefficient can be reconstructed into time-domain signals with constraints of its own frequency bandTaking the daily time-history curve of static strain data1198781shown in Figure 14(b) as an example it is decomposed
within 8 scales and the decomposition coefficient 11986208
119895
isspecially selected and then reconstructed as 119878III shown inFigure 15(a) It can be seen that sharp peaks caused by trainsare removed effectively and the static strains which are causedby temperature field are retained as well Furthermore theannual extraction result 119878III of 1198781 fromMarch 2013 toOctober2013 is shown in Figure 15(b)
322 Correlation Model between Temperature Field andStatic Strains In constructing the correlationmodel betweentemperature field and static strains data from March 2013to October 2013 are chosen as training data and data inNovember 2013 are chosen as test data for verification of
the correlation model The static strain 119878III is expressed asfollows
119878III =8
sum
119894=1
120572119894119879119894+
16
sum
119895=1
120573119895119863119895+ 119888 (2)
where 119879119894is 119894th temperature data from set 119879
1 1198792 1198793 1198794
1198795 1198796 1198797 1198798 119863119895is 119895th temperature difference data from set
11987912 11987934 11987956 11987978 11987913 11987915 11987917 11987935 11987937 11987957 11987924 11987926 11987928 11987946
11987948 11987968 120572119894is 119894th linear expansion coefficient of 119879
119894 120573119895is 119895th
linear expansion coefficient of 119879119894119895 and 119888 is the constant term
Considering that total 24 linear expansion coefficientsand one constant term in (2) are very complicated forcalculation principal component analysis is further usedto simplify (2) which aims at finding small amounts ofcomprehensive factors to represent main information of tem-perature and temperature difference data By using principalcomponent analysis the comprehensive factors can be foundby transforming temperature and temperature difference datainto factor component space [11 12] For temperature data theexplained variance of the first potential comprehensive factor1198751reaches 967 apparently larger than the other potential
comprehensive factors Thus 1198751is selected as the compre-
hensive factor Meanwhile for temperature difference datausing principal component analysis the explained varianceof the first three potential comprehensive factors 119877
1 1198772 and
1198773reaches 988 which are selected as the comprehensive
factorsTherefore the correlation models between 119878III and tem-
perature field can be expressed as follows
119878III = 1205821([119864]1times8
997888
1198798times1
) +9978881205741times3
([119865]3times16
997888
11986316times1
) + 119888 (3)
where997888
1198798times1
denotes one vector containing eight temperaturedata 997888119863
16times1denotes one vector containing sixteen temper-
ature difference data 119863119895(119895 = 1 2 16) [119864]
1times8and
The Scientific World Journal 11
b = minus3632
k = 0946
minus100 minus55 minus10 8035Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus100
minus55
minus10
35
80
Sim
ulat
iveS
III
(120583120576)
(a) 1198781
b = minus2013k = 0898
minus45minus75 minus15 45 7515Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus75
minus45
minus15
15
45
75
Sim
ulat
iveS
III
(120583120576)
(b) 1198782
b = minus1844k = 0904
minus60
minus30
0
30
60
90
Sim
ulat
iveS
III
(120583120576)
minus30minus60 30 60 900Monitoring SIII (120583120576)
Scattered pointsFitting curve
(c) 1198783
b = 2057k = 1126
minus15minus30 15 300Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus40
minus25
minus10
5
20
35Si
mul
ativ
eSII
I(120583120576)
(d) 1198784
b = minus4139k = 0935
minus100
minus65
minus30
5
40
75
Sim
ulat
iveS
III
(120583120576)
minus105 4515 75minus45minus75 minus15
Monitoring SIII (120583120576)
Scattered pointsFitting curve
(e) 1198785
b = minus4094k = 0918
minus145minus205 minus25minus85 9535Monitoring SIII (120583120576)
minus185
minus130
minus75
minus20
35
90
Sim
ulat
iveS
III
(120583120576)
Scattered pointsFitting curve
(f) 1198786
Figure 16 Continued
12 The Scientific World Journal
b = 2148k = 0881
minus50
minus28
minus6
16
38
60Si
mul
ativ
eSII
I(120583120576)
minus45 minus23 minus1 43 6521Monitoring SIII (120583120576)
Scattered pointsFitting curve
(g) 1198787
b = 0615k = 1118
minus30
minus15
0
15
30
Sim
ulat
iveS
III
(120583120576)
minus15minus25 minus5 15 255Monitoring SIII (120583120576)
Scattered pointsFitting curve
(h) 1198788
Figure 16 Correlation between simulative 119878III and monitoring 119878III
Table 2 Performance parameters 1205821
1205741
1205742
1205743
and 119862 in correlationmodels
Static straindata 120582
1
1205741
1205742
1205743
119862
119878III of 1198781 0454 minus4711 minus2597 minus4944 minus46806119878III of 1198782 1585 0013 4104 minus1247 minus70238119878III of 1198783 0301 3416 minus2098 2968 minus6884119878III of 1198784 0350 minus0734 0351 0948 minus23597119878III of 1198785 0179 minus4219 1827 3165 minus22049119878III of 1198786 0062 minus3245 8242 minus5398 minus22862119878III of 1198787 minus0367 2411 minus0598 5531 17075119878III of 1198788 0611 2188 minus1130 2018 minus32799
[119865]3times16
denote transformation matrix which can be cal-culated using principal component analysis [11 12] 120582
1is
performance parameter of first potential comprehensivefactor 119875
1 9978881205741times3
= [1205741 1205742 1205743] is one vector containing
three performance parameters corresponding to first threepotential comprehensive factors 119877
1 1198772 and 119877
3 The values of
performance parameters 1205821 9978881205741times3
and 119888 are estimated usingtraining data by multivariate linear regression method [13]and are shown in Table 2
In order to verify the effectiveness of the correlationmodels shown in (2) test data in November 2013 are usedto obtain the simulative 119878III The scattered points betweensimulative 119878III and monitoring 119878III are plotted as shown inFigure 16 which are further fitted by linear function
119878119904= 119896119878119898+ 119887 (4)
where 119878119904denotes simulative 119878III 119878119898 denotes monitoring
119878III and 119896 119887 are fitting parameters with the fitting resultsshown in Figure 16 as well The fitting results verify theeffectiveness of the correlationmodels Amultivariable mean
value control chart is also employed tomonitor the measuredabnormal changes in the static strains of the steel truss archThe condition index 119890 for early warning of deterioration ofthe static performance of steel truss arch is defined as thedifference between the measured and predicted static strains
119890 = 119878119898minus 119878119901 (5)
where 119878119898is the measured static strain 119878III 119878119901 is the predicted
static strain 119878III which is obtained by using the correlationmodels shown in (2) in the healthy condition Figure 17 showsthemean value control chart for static strains of the steel trussarch using measurement data in November 2013 The upperand lower limits are determined using the controlling param-eter 119896 = 2235 Hence a long-term monitoring on the corre-lation models between temperature field and static strains issuitable for early warning of deterioration of the static per-formance of steel truss arch
33 Evaluation of Movement Performance of Pier Longi-tudinal displacements of piers and temperature field datain steel truss arch collected from March 2013 to October2013 are used for evaluation of movement performanceof piers Longitudinal displacement data from sensors 119871
119894
(119894 = 1 2 6) are denoted by 119889119894(119905) (where 119905 means time)
respectively Meanwhile temperature field data from sensors119882119894(119894 = 1 2 8) are denoted by 119879
119894(119905) respectively and
their averaged value 119879(119905) is used to represent the averagetemperature of the steel truss arch For simplicity taking10 minutes as basic time interval the 10 min average dis-placements and temperatures are calculated and used as therepresentative values of this time interval By this way thedisplacements and temperatures in one day are reduced to144 representative samples Furthermore the temperature-displacement scatter diagrams are plotted in Figure 18 Anoverall increase in longitudinal displacements of piers isobserved with an increase in the average temperature of the
The Scientific World Journal 13
UCL
LCL
CL
Con
ditio
n in
dexe
(120583120576)
minus15
minus10
minus5
0
5
10
15
10 20 30 40 50 600Sample
Figure 17 Mean value control chart for static strains of the steel truss arch
Table 3 Summary of linear regression models
Pier Regression function4 pier 119889
1
= minus11524 + 696119879
5 pier 1198892
= minus9253 + 562119879
6 pier 1198893
= minus6628 + 382119879
8 pier 1198894
= minus4658 + 336119879
9 pier 1198895
= minus7605 + 530119879
10 pier 1198896
= minus9850 + 664119879
bridge And the correlation coefficients of the longitudinaldisplacements of piers and average temperature of the steeltruss arch are all larger than 092 as shown in Figure 18
A linear regression analysis between the 10 min averagedisplacement 119889
119894(119905) and the temperature 119879(119905) is further per-
formed to model the temperature-displacement correlationsusing the following equation [14]
119889119894(119905) = 120573
0+ 1205731119879 (119905) (6)
where the regression coefficients 1205731and 120573
0were obtained by
the least-squares method as
1205731=
119878119889119894119879
119878119879119879
1205730= 119889119894minus 1205731119879
(7)
where 119878119889119894119879
is the covariance between the displacement andthe temperature sequences 119878
119879119879is the variance of the mea-
sured temperature sequences and 119879 and 119889119894are the means
of the measured temperature and displacement sequencesrespectively Table 3 summarizes the expressions of linearregression models of the displacement versus temperature
The analysis results indicate that good correlation existsbetween the longitudinal displacements of piers and averagetemperature of steel truss arch A multivariable mean valuecontrol chart is also employed to monitor the measuredabnormal changes in the longitudinal displacements of piersThe condition index 119890 for early warning of abnormal behaviorof piers is defined as the difference between themeasured andpredicted longitudinal displacements
119890 = 119889119898minus 119889119901 (8)
where 119889119898is the measured displacement of the piers 119889
119901is
the predicted displacement of the piers which is obtainedby using the linear regression models shown in Table 3in the healthy condition Figure 19 shows the mean valuecontrol chart for longitudinal displacements of piers usingmeasurement data in March 2013 The upper and lowerlimits are determined using the controlling parameter 119896 =
2804 Hence a long-term monitoring on the temperature-displacement correlation model is suitable for real-timecondition assessment for movement performance of piers
34 Evaluation of Fatigue Performance of Steel Deck Fatiguecracking is universal for orthotropic steel deck of highwaybridges Once the fatigue crack appears extensive fatiguecracking may occur and repair is very difficult Henceinfinite-fatigue-life design method is applied for NanjingDashengguan Bridge instead of finite-fatigue-life designmethod which is applied extensively in highway bridges Theinfinite-fatigue-life design method is to control the fatiguestress amplitude in the range less than the value of fatiguecut-off limit refrained from fatigue cracking Consequentlythe purpose of equipping stain gauges in the SHM system ofDashengguan Bridge is to monitor the stress amplitude andthe validity of fatigue design can be achieved by comparingthe stress amplitude with fatigue stress limit
The structural dynamic response is asymmetrical becauseof the irregularity of railway and braking force from thetrain even though the structure of Nanjing DashengguanBridge is spatially symmetrical Hence the stresses are notsymmetrical measured by strain gauges from the upstreamand downstream directions However the fatigue amplitudesof two strain gauges induced by trains are both extremelysmall Thus the strain history data of the strain gauge DYB-2are merely selected in this paper to illustrate that the fatiguestress amplitude is much lower than the value of fatigue stresslimit The typical strain time-history curves of DYB-2 whensingle train with 8 carriages or 16 carriages passed throughthe bridge are shown in Figure 20 respectively It can be seenthat during the passing of high-speed trains both 8 carriagesand 16 carriages have led to strain variations reaching to 2ndash8 120583120576 And the number of strain peaks is 16 and 32 for trainswith 8 carriages and 16 carriages respectively because theaxle number of each carriage is 2
To perform fatigue evaluation simplified rainflow cyclecounting algorithm was firstly used to process strain history
14 The Scientific World Journal
R = 09748
10 20 30 40 500
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
(a) Correlation between 1198891of the 4 pier and average temperature 119879 of
steel truss arch
10 20 30 40 500
Temperature (∘C)
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
R = 09746
(b) Correlation between 1198892of the 5 pier and average temperature 119879 of
steel truss arch
10 200 40 5030
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
Disp
lace
men
t (m
m)
R = 09471
(c) Correlation between 1198893of the 6 pier and average temperature 119879 of
steel truss arch
R = 09276
10 20 30 40 500
Temperature (∘C)
minus60
minus30
0
30
60
90
120D
ispla
cem
ent (
mm
)
(d) Correlation between 1198894of the 8 pier and average temperature 119879 of
steel truss arch
R = 09536
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(e) Correlation between 1198895of the 9 pier and average temperature 119879 of
steel truss arch
R = 09537
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(f) Correlation between 1198896of the 10 pier and average temperature 119879 of
steel truss arch
Figure 18 Correlation between longitudinal displacement and average temperature
The Scientific World Journal 15
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
)
500 1000 1500 2000 2500 3000 3500 4000 45000Sample
minus15
minus10
minus5
0
5
10
15
Figure 19 Mean value control chart for longitudinal displacements of piers
Stra
in (120583
120576)
minus174
minus172
minus170
minus168
minus166
minus164
minus162
1 2714 53
66
79
92
40
118
209
183
222
105
157
196
144
131
235
248
170
Data point
(a) Single train with 8 carriages passed through the bridge
Stra
in (120583
120576)
minus180minus178minus176minus174minus172minus170minus168minus166
1
61
91
31
181
211
121
421
301
481
511
361
391
271
451
241
331
541
571
151
Data point
(b) Single train with 16 carriages passed through the bridge
Figure 20 Typical strain time-history curves of DYB-2 when single train passed through the bridge
data and the spectrum of stress amplitude was obtained[15] The spectra of stress amplitude calculated using thestrain history data with regard to 8 carriages or 16 carriagesare shown in Figure 21 respectively It can be seen that themaximum stress amplitude is smaller than 10MPa obtainedfrom strain history curves under single trainThen accordingtoMinerrsquos law and standard of Eurocode 3 fatigue parametersof 119878eq (equivalent stress amplitude) and 119873
119904(stress cycle
number) under single train were calculated as follows [16]
119878eq = (
1198991198941198785
119894
sum119899119894
)
15
119873s = sum119899119894
(9)
where 119878119894is 119894th stress amplitude 119899
119894is the number of
stress cycles corresponding to 119878119894 and 15 is slope value of
log 119878- log119873 curve in the scope of lower stress amplitudeaccording to the standard of Eurocode 3 [17]
Using (9) the equivalent stress amplitudes 119878eq with regardto 8 carriages or 16 carriages are 075MPa and 071MParespectively And the stress cycle numbers 119873
119904are 22 and 45
respectively Thus the value of 119878eq with regard to 8 carriagesis similar to that of 16 carriages However the value of 119873
119904
with regard to 8 carriages is about twice as much as that of 16carriages Furthermore as recommended by Eurocode 3 forrib-to-deck welded joint the fatigue limit Δ120590
119871is 29MPa [17]
and the value of 119878eq when the high-speed train passes throughbridge is far less than Δ120590
119871 which indicates that fatigue life of
rib-to-deck welded joint in Nanjing Dashengguan Bridge canbe considered infinite Thus the fatigue performance of steel
deck satisfies the requirement of infinite-fatigue-life designmethod
4 Conclusions
In the last few decades Structural Health Monitoring (SHM)system has been widely installed in many highway bridgesto continuously monitor the structural condition over anextended period of time However application of SHMsystem in high-speed railway bridges is far insufficient Thispaper presents an analysis of the SHM system installedin Nanjing Dashengguan Bridge in China The monitoringsystem is composed of a sensor system a data acquisitionand transmission system a data management system and astructural evaluation system The monitoring system is cur-rently in full operation accumulating the bridgersquos behaviorsunder the varying environmental conditions such as high-speed trains and environmental temperature
Based on long-termhealthmonitoring data general prin-ciples and analytical processes for structural evaluation areprovided and some evaluation results are reported It can beshown from the analysis in this paper that the mathematicalcorrelation models describing the overall structural behaviorof the bridge can be obtained with the support of the healthmonitoring system which includes cross-correlation modelsfor accelerations correlation models between temperatureand static strains of steel truss arch and correlation mod-els between temperature and longitudinal displacements ofpiers The mean value control chart is further employed tomonitor the time series of differences between the measuredstructural parameters and the predicted parameters usingthese correlation models The upper and lower control limits
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
10 The Scientific World Journal
4 8 12 16 20 240Time (h)
Stat
ic st
rain
S III
(120583120576)
minus120
minus65
minus10
45
100
(a) Daily extraction result 119878III of 1198781
4 5 6 7 8 9 10 113Time (month)
Stat
ic st
rain
S III
(120583120576)
minus200
minus130
minus60
10
80
150
(b) Annual extraction result 119878III of 1198781 fromMarch 2013 to October 2013
Figure 15 Extraction results 119878III of static strain data 1198781
static strain and positive values denote tension static strain)From Figure 14 it can be seen that time-history curve of 119878
1
contains many sharp peaks caused by high-speed trains witheach peak corresponding to the time when one train passesthrough the bridge And the static strains caused by high-speed trains are obviously lower than that of temperature fieldincluding temperature data and temperature difference data
As shown in Figure 14 static strain data in steel truss archmainly contains three parts static strain 119878I caused by tem-perature static strain 119878II caused by temperature differenceand static strain 119878III caused by train In order to construct thecorrelation between temperature field and static strain in steeltruss arch static strain 119878I and static strain 119878II (denoted by 119878III)must be extracted from the static strain data In the presentstudy wavelet packet decomposition method is used toextract 119878III considering the high-frequency characteristics ofstatic strain data caused by trains By using the wavelet packetdecomposition static strain data can be decomposed scale byscale into different frequency bands and each decompositioncoefficient corresponds to its frequency band [9 10] Eachdecomposition coefficient can be reconstructed into time-domain signals with constraints of its own frequency bandTaking the daily time-history curve of static strain data1198781shown in Figure 14(b) as an example it is decomposed
within 8 scales and the decomposition coefficient 11986208
119895
isspecially selected and then reconstructed as 119878III shown inFigure 15(a) It can be seen that sharp peaks caused by trainsare removed effectively and the static strains which are causedby temperature field are retained as well Furthermore theannual extraction result 119878III of 1198781 fromMarch 2013 toOctober2013 is shown in Figure 15(b)
322 Correlation Model between Temperature Field andStatic Strains In constructing the correlationmodel betweentemperature field and static strains data from March 2013to October 2013 are chosen as training data and data inNovember 2013 are chosen as test data for verification of
the correlation model The static strain 119878III is expressed asfollows
119878III =8
sum
119894=1
120572119894119879119894+
16
sum
119895=1
120573119895119863119895+ 119888 (2)
where 119879119894is 119894th temperature data from set 119879
1 1198792 1198793 1198794
1198795 1198796 1198797 1198798 119863119895is 119895th temperature difference data from set
11987912 11987934 11987956 11987978 11987913 11987915 11987917 11987935 11987937 11987957 11987924 11987926 11987928 11987946
11987948 11987968 120572119894is 119894th linear expansion coefficient of 119879
119894 120573119895is 119895th
linear expansion coefficient of 119879119894119895 and 119888 is the constant term
Considering that total 24 linear expansion coefficientsand one constant term in (2) are very complicated forcalculation principal component analysis is further usedto simplify (2) which aims at finding small amounts ofcomprehensive factors to represent main information of tem-perature and temperature difference data By using principalcomponent analysis the comprehensive factors can be foundby transforming temperature and temperature difference datainto factor component space [11 12] For temperature data theexplained variance of the first potential comprehensive factor1198751reaches 967 apparently larger than the other potential
comprehensive factors Thus 1198751is selected as the compre-
hensive factor Meanwhile for temperature difference datausing principal component analysis the explained varianceof the first three potential comprehensive factors 119877
1 1198772 and
1198773reaches 988 which are selected as the comprehensive
factorsTherefore the correlation models between 119878III and tem-
perature field can be expressed as follows
119878III = 1205821([119864]1times8
997888
1198798times1
) +9978881205741times3
([119865]3times16
997888
11986316times1
) + 119888 (3)
where997888
1198798times1
denotes one vector containing eight temperaturedata 997888119863
16times1denotes one vector containing sixteen temper-
ature difference data 119863119895(119895 = 1 2 16) [119864]
1times8and
The Scientific World Journal 11
b = minus3632
k = 0946
minus100 minus55 minus10 8035Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus100
minus55
minus10
35
80
Sim
ulat
iveS
III
(120583120576)
(a) 1198781
b = minus2013k = 0898
minus45minus75 minus15 45 7515Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus75
minus45
minus15
15
45
75
Sim
ulat
iveS
III
(120583120576)
(b) 1198782
b = minus1844k = 0904
minus60
minus30
0
30
60
90
Sim
ulat
iveS
III
(120583120576)
minus30minus60 30 60 900Monitoring SIII (120583120576)
Scattered pointsFitting curve
(c) 1198783
b = 2057k = 1126
minus15minus30 15 300Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus40
minus25
minus10
5
20
35Si
mul
ativ
eSII
I(120583120576)
(d) 1198784
b = minus4139k = 0935
minus100
minus65
minus30
5
40
75
Sim
ulat
iveS
III
(120583120576)
minus105 4515 75minus45minus75 minus15
Monitoring SIII (120583120576)
Scattered pointsFitting curve
(e) 1198785
b = minus4094k = 0918
minus145minus205 minus25minus85 9535Monitoring SIII (120583120576)
minus185
minus130
minus75
minus20
35
90
Sim
ulat
iveS
III
(120583120576)
Scattered pointsFitting curve
(f) 1198786
Figure 16 Continued
12 The Scientific World Journal
b = 2148k = 0881
minus50
minus28
minus6
16
38
60Si
mul
ativ
eSII
I(120583120576)
minus45 minus23 minus1 43 6521Monitoring SIII (120583120576)
Scattered pointsFitting curve
(g) 1198787
b = 0615k = 1118
minus30
minus15
0
15
30
Sim
ulat
iveS
III
(120583120576)
minus15minus25 minus5 15 255Monitoring SIII (120583120576)
Scattered pointsFitting curve
(h) 1198788
Figure 16 Correlation between simulative 119878III and monitoring 119878III
Table 2 Performance parameters 1205821
1205741
1205742
1205743
and 119862 in correlationmodels
Static straindata 120582
1
1205741
1205742
1205743
119862
119878III of 1198781 0454 minus4711 minus2597 minus4944 minus46806119878III of 1198782 1585 0013 4104 minus1247 minus70238119878III of 1198783 0301 3416 minus2098 2968 minus6884119878III of 1198784 0350 minus0734 0351 0948 minus23597119878III of 1198785 0179 minus4219 1827 3165 minus22049119878III of 1198786 0062 minus3245 8242 minus5398 minus22862119878III of 1198787 minus0367 2411 minus0598 5531 17075119878III of 1198788 0611 2188 minus1130 2018 minus32799
[119865]3times16
denote transformation matrix which can be cal-culated using principal component analysis [11 12] 120582
1is
performance parameter of first potential comprehensivefactor 119875
1 9978881205741times3
= [1205741 1205742 1205743] is one vector containing
three performance parameters corresponding to first threepotential comprehensive factors 119877
1 1198772 and 119877
3 The values of
performance parameters 1205821 9978881205741times3
and 119888 are estimated usingtraining data by multivariate linear regression method [13]and are shown in Table 2
In order to verify the effectiveness of the correlationmodels shown in (2) test data in November 2013 are usedto obtain the simulative 119878III The scattered points betweensimulative 119878III and monitoring 119878III are plotted as shown inFigure 16 which are further fitted by linear function
119878119904= 119896119878119898+ 119887 (4)
where 119878119904denotes simulative 119878III 119878119898 denotes monitoring
119878III and 119896 119887 are fitting parameters with the fitting resultsshown in Figure 16 as well The fitting results verify theeffectiveness of the correlationmodels Amultivariable mean
value control chart is also employed tomonitor the measuredabnormal changes in the static strains of the steel truss archThe condition index 119890 for early warning of deterioration ofthe static performance of steel truss arch is defined as thedifference between the measured and predicted static strains
119890 = 119878119898minus 119878119901 (5)
where 119878119898is the measured static strain 119878III 119878119901 is the predicted
static strain 119878III which is obtained by using the correlationmodels shown in (2) in the healthy condition Figure 17 showsthemean value control chart for static strains of the steel trussarch using measurement data in November 2013 The upperand lower limits are determined using the controlling param-eter 119896 = 2235 Hence a long-term monitoring on the corre-lation models between temperature field and static strains issuitable for early warning of deterioration of the static per-formance of steel truss arch
33 Evaluation of Movement Performance of Pier Longi-tudinal displacements of piers and temperature field datain steel truss arch collected from March 2013 to October2013 are used for evaluation of movement performanceof piers Longitudinal displacement data from sensors 119871
119894
(119894 = 1 2 6) are denoted by 119889119894(119905) (where 119905 means time)
respectively Meanwhile temperature field data from sensors119882119894(119894 = 1 2 8) are denoted by 119879
119894(119905) respectively and
their averaged value 119879(119905) is used to represent the averagetemperature of the steel truss arch For simplicity taking10 minutes as basic time interval the 10 min average dis-placements and temperatures are calculated and used as therepresentative values of this time interval By this way thedisplacements and temperatures in one day are reduced to144 representative samples Furthermore the temperature-displacement scatter diagrams are plotted in Figure 18 Anoverall increase in longitudinal displacements of piers isobserved with an increase in the average temperature of the
The Scientific World Journal 13
UCL
LCL
CL
Con
ditio
n in
dexe
(120583120576)
minus15
minus10
minus5
0
5
10
15
10 20 30 40 50 600Sample
Figure 17 Mean value control chart for static strains of the steel truss arch
Table 3 Summary of linear regression models
Pier Regression function4 pier 119889
1
= minus11524 + 696119879
5 pier 1198892
= minus9253 + 562119879
6 pier 1198893
= minus6628 + 382119879
8 pier 1198894
= minus4658 + 336119879
9 pier 1198895
= minus7605 + 530119879
10 pier 1198896
= minus9850 + 664119879
bridge And the correlation coefficients of the longitudinaldisplacements of piers and average temperature of the steeltruss arch are all larger than 092 as shown in Figure 18
A linear regression analysis between the 10 min averagedisplacement 119889
119894(119905) and the temperature 119879(119905) is further per-
formed to model the temperature-displacement correlationsusing the following equation [14]
119889119894(119905) = 120573
0+ 1205731119879 (119905) (6)
where the regression coefficients 1205731and 120573
0were obtained by
the least-squares method as
1205731=
119878119889119894119879
119878119879119879
1205730= 119889119894minus 1205731119879
(7)
where 119878119889119894119879
is the covariance between the displacement andthe temperature sequences 119878
119879119879is the variance of the mea-
sured temperature sequences and 119879 and 119889119894are the means
of the measured temperature and displacement sequencesrespectively Table 3 summarizes the expressions of linearregression models of the displacement versus temperature
The analysis results indicate that good correlation existsbetween the longitudinal displacements of piers and averagetemperature of steel truss arch A multivariable mean valuecontrol chart is also employed to monitor the measuredabnormal changes in the longitudinal displacements of piersThe condition index 119890 for early warning of abnormal behaviorof piers is defined as the difference between themeasured andpredicted longitudinal displacements
119890 = 119889119898minus 119889119901 (8)
where 119889119898is the measured displacement of the piers 119889
119901is
the predicted displacement of the piers which is obtainedby using the linear regression models shown in Table 3in the healthy condition Figure 19 shows the mean valuecontrol chart for longitudinal displacements of piers usingmeasurement data in March 2013 The upper and lowerlimits are determined using the controlling parameter 119896 =
2804 Hence a long-term monitoring on the temperature-displacement correlation model is suitable for real-timecondition assessment for movement performance of piers
34 Evaluation of Fatigue Performance of Steel Deck Fatiguecracking is universal for orthotropic steel deck of highwaybridges Once the fatigue crack appears extensive fatiguecracking may occur and repair is very difficult Henceinfinite-fatigue-life design method is applied for NanjingDashengguan Bridge instead of finite-fatigue-life designmethod which is applied extensively in highway bridges Theinfinite-fatigue-life design method is to control the fatiguestress amplitude in the range less than the value of fatiguecut-off limit refrained from fatigue cracking Consequentlythe purpose of equipping stain gauges in the SHM system ofDashengguan Bridge is to monitor the stress amplitude andthe validity of fatigue design can be achieved by comparingthe stress amplitude with fatigue stress limit
The structural dynamic response is asymmetrical becauseof the irregularity of railway and braking force from thetrain even though the structure of Nanjing DashengguanBridge is spatially symmetrical Hence the stresses are notsymmetrical measured by strain gauges from the upstreamand downstream directions However the fatigue amplitudesof two strain gauges induced by trains are both extremelysmall Thus the strain history data of the strain gauge DYB-2are merely selected in this paper to illustrate that the fatiguestress amplitude is much lower than the value of fatigue stresslimit The typical strain time-history curves of DYB-2 whensingle train with 8 carriages or 16 carriages passed throughthe bridge are shown in Figure 20 respectively It can be seenthat during the passing of high-speed trains both 8 carriagesand 16 carriages have led to strain variations reaching to 2ndash8 120583120576 And the number of strain peaks is 16 and 32 for trainswith 8 carriages and 16 carriages respectively because theaxle number of each carriage is 2
To perform fatigue evaluation simplified rainflow cyclecounting algorithm was firstly used to process strain history
14 The Scientific World Journal
R = 09748
10 20 30 40 500
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
(a) Correlation between 1198891of the 4 pier and average temperature 119879 of
steel truss arch
10 20 30 40 500
Temperature (∘C)
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
R = 09746
(b) Correlation between 1198892of the 5 pier and average temperature 119879 of
steel truss arch
10 200 40 5030
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
Disp
lace
men
t (m
m)
R = 09471
(c) Correlation between 1198893of the 6 pier and average temperature 119879 of
steel truss arch
R = 09276
10 20 30 40 500
Temperature (∘C)
minus60
minus30
0
30
60
90
120D
ispla
cem
ent (
mm
)
(d) Correlation between 1198894of the 8 pier and average temperature 119879 of
steel truss arch
R = 09536
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(e) Correlation between 1198895of the 9 pier and average temperature 119879 of
steel truss arch
R = 09537
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(f) Correlation between 1198896of the 10 pier and average temperature 119879 of
steel truss arch
Figure 18 Correlation between longitudinal displacement and average temperature
The Scientific World Journal 15
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
)
500 1000 1500 2000 2500 3000 3500 4000 45000Sample
minus15
minus10
minus5
0
5
10
15
Figure 19 Mean value control chart for longitudinal displacements of piers
Stra
in (120583
120576)
minus174
minus172
minus170
minus168
minus166
minus164
minus162
1 2714 53
66
79
92
40
118
209
183
222
105
157
196
144
131
235
248
170
Data point
(a) Single train with 8 carriages passed through the bridge
Stra
in (120583
120576)
minus180minus178minus176minus174minus172minus170minus168minus166
1
61
91
31
181
211
121
421
301
481
511
361
391
271
451
241
331
541
571
151
Data point
(b) Single train with 16 carriages passed through the bridge
Figure 20 Typical strain time-history curves of DYB-2 when single train passed through the bridge
data and the spectrum of stress amplitude was obtained[15] The spectra of stress amplitude calculated using thestrain history data with regard to 8 carriages or 16 carriagesare shown in Figure 21 respectively It can be seen that themaximum stress amplitude is smaller than 10MPa obtainedfrom strain history curves under single trainThen accordingtoMinerrsquos law and standard of Eurocode 3 fatigue parametersof 119878eq (equivalent stress amplitude) and 119873
119904(stress cycle
number) under single train were calculated as follows [16]
119878eq = (
1198991198941198785
119894
sum119899119894
)
15
119873s = sum119899119894
(9)
where 119878119894is 119894th stress amplitude 119899
119894is the number of
stress cycles corresponding to 119878119894 and 15 is slope value of
log 119878- log119873 curve in the scope of lower stress amplitudeaccording to the standard of Eurocode 3 [17]
Using (9) the equivalent stress amplitudes 119878eq with regardto 8 carriages or 16 carriages are 075MPa and 071MParespectively And the stress cycle numbers 119873
119904are 22 and 45
respectively Thus the value of 119878eq with regard to 8 carriagesis similar to that of 16 carriages However the value of 119873
119904
with regard to 8 carriages is about twice as much as that of 16carriages Furthermore as recommended by Eurocode 3 forrib-to-deck welded joint the fatigue limit Δ120590
119871is 29MPa [17]
and the value of 119878eq when the high-speed train passes throughbridge is far less than Δ120590
119871 which indicates that fatigue life of
rib-to-deck welded joint in Nanjing Dashengguan Bridge canbe considered infinite Thus the fatigue performance of steel
deck satisfies the requirement of infinite-fatigue-life designmethod
4 Conclusions
In the last few decades Structural Health Monitoring (SHM)system has been widely installed in many highway bridgesto continuously monitor the structural condition over anextended period of time However application of SHMsystem in high-speed railway bridges is far insufficient Thispaper presents an analysis of the SHM system installedin Nanjing Dashengguan Bridge in China The monitoringsystem is composed of a sensor system a data acquisitionand transmission system a data management system and astructural evaluation system The monitoring system is cur-rently in full operation accumulating the bridgersquos behaviorsunder the varying environmental conditions such as high-speed trains and environmental temperature
Based on long-termhealthmonitoring data general prin-ciples and analytical processes for structural evaluation areprovided and some evaluation results are reported It can beshown from the analysis in this paper that the mathematicalcorrelation models describing the overall structural behaviorof the bridge can be obtained with the support of the healthmonitoring system which includes cross-correlation modelsfor accelerations correlation models between temperatureand static strains of steel truss arch and correlation mod-els between temperature and longitudinal displacements ofpiers The mean value control chart is further employed tomonitor the time series of differences between the measuredstructural parameters and the predicted parameters usingthese correlation models The upper and lower control limits
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
The Scientific World Journal 11
b = minus3632
k = 0946
minus100 minus55 minus10 8035Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus100
minus55
minus10
35
80
Sim
ulat
iveS
III
(120583120576)
(a) 1198781
b = minus2013k = 0898
minus45minus75 minus15 45 7515Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus75
minus45
minus15
15
45
75
Sim
ulat
iveS
III
(120583120576)
(b) 1198782
b = minus1844k = 0904
minus60
minus30
0
30
60
90
Sim
ulat
iveS
III
(120583120576)
minus30minus60 30 60 900Monitoring SIII (120583120576)
Scattered pointsFitting curve
(c) 1198783
b = 2057k = 1126
minus15minus30 15 300Monitoring SIII (120583120576)
Scattered pointsFitting curve
minus40
minus25
minus10
5
20
35Si
mul
ativ
eSII
I(120583120576)
(d) 1198784
b = minus4139k = 0935
minus100
minus65
minus30
5
40
75
Sim
ulat
iveS
III
(120583120576)
minus105 4515 75minus45minus75 minus15
Monitoring SIII (120583120576)
Scattered pointsFitting curve
(e) 1198785
b = minus4094k = 0918
minus145minus205 minus25minus85 9535Monitoring SIII (120583120576)
minus185
minus130
minus75
minus20
35
90
Sim
ulat
iveS
III
(120583120576)
Scattered pointsFitting curve
(f) 1198786
Figure 16 Continued
12 The Scientific World Journal
b = 2148k = 0881
minus50
minus28
minus6
16
38
60Si
mul
ativ
eSII
I(120583120576)
minus45 minus23 minus1 43 6521Monitoring SIII (120583120576)
Scattered pointsFitting curve
(g) 1198787
b = 0615k = 1118
minus30
minus15
0
15
30
Sim
ulat
iveS
III
(120583120576)
minus15minus25 minus5 15 255Monitoring SIII (120583120576)
Scattered pointsFitting curve
(h) 1198788
Figure 16 Correlation between simulative 119878III and monitoring 119878III
Table 2 Performance parameters 1205821
1205741
1205742
1205743
and 119862 in correlationmodels
Static straindata 120582
1
1205741
1205742
1205743
119862
119878III of 1198781 0454 minus4711 minus2597 minus4944 minus46806119878III of 1198782 1585 0013 4104 minus1247 minus70238119878III of 1198783 0301 3416 minus2098 2968 minus6884119878III of 1198784 0350 minus0734 0351 0948 minus23597119878III of 1198785 0179 minus4219 1827 3165 minus22049119878III of 1198786 0062 minus3245 8242 minus5398 minus22862119878III of 1198787 minus0367 2411 minus0598 5531 17075119878III of 1198788 0611 2188 minus1130 2018 minus32799
[119865]3times16
denote transformation matrix which can be cal-culated using principal component analysis [11 12] 120582
1is
performance parameter of first potential comprehensivefactor 119875
1 9978881205741times3
= [1205741 1205742 1205743] is one vector containing
three performance parameters corresponding to first threepotential comprehensive factors 119877
1 1198772 and 119877
3 The values of
performance parameters 1205821 9978881205741times3
and 119888 are estimated usingtraining data by multivariate linear regression method [13]and are shown in Table 2
In order to verify the effectiveness of the correlationmodels shown in (2) test data in November 2013 are usedto obtain the simulative 119878III The scattered points betweensimulative 119878III and monitoring 119878III are plotted as shown inFigure 16 which are further fitted by linear function
119878119904= 119896119878119898+ 119887 (4)
where 119878119904denotes simulative 119878III 119878119898 denotes monitoring
119878III and 119896 119887 are fitting parameters with the fitting resultsshown in Figure 16 as well The fitting results verify theeffectiveness of the correlationmodels Amultivariable mean
value control chart is also employed tomonitor the measuredabnormal changes in the static strains of the steel truss archThe condition index 119890 for early warning of deterioration ofthe static performance of steel truss arch is defined as thedifference between the measured and predicted static strains
119890 = 119878119898minus 119878119901 (5)
where 119878119898is the measured static strain 119878III 119878119901 is the predicted
static strain 119878III which is obtained by using the correlationmodels shown in (2) in the healthy condition Figure 17 showsthemean value control chart for static strains of the steel trussarch using measurement data in November 2013 The upperand lower limits are determined using the controlling param-eter 119896 = 2235 Hence a long-term monitoring on the corre-lation models between temperature field and static strains issuitable for early warning of deterioration of the static per-formance of steel truss arch
33 Evaluation of Movement Performance of Pier Longi-tudinal displacements of piers and temperature field datain steel truss arch collected from March 2013 to October2013 are used for evaluation of movement performanceof piers Longitudinal displacement data from sensors 119871
119894
(119894 = 1 2 6) are denoted by 119889119894(119905) (where 119905 means time)
respectively Meanwhile temperature field data from sensors119882119894(119894 = 1 2 8) are denoted by 119879
119894(119905) respectively and
their averaged value 119879(119905) is used to represent the averagetemperature of the steel truss arch For simplicity taking10 minutes as basic time interval the 10 min average dis-placements and temperatures are calculated and used as therepresentative values of this time interval By this way thedisplacements and temperatures in one day are reduced to144 representative samples Furthermore the temperature-displacement scatter diagrams are plotted in Figure 18 Anoverall increase in longitudinal displacements of piers isobserved with an increase in the average temperature of the
The Scientific World Journal 13
UCL
LCL
CL
Con
ditio
n in
dexe
(120583120576)
minus15
minus10
minus5
0
5
10
15
10 20 30 40 50 600Sample
Figure 17 Mean value control chart for static strains of the steel truss arch
Table 3 Summary of linear regression models
Pier Regression function4 pier 119889
1
= minus11524 + 696119879
5 pier 1198892
= minus9253 + 562119879
6 pier 1198893
= minus6628 + 382119879
8 pier 1198894
= minus4658 + 336119879
9 pier 1198895
= minus7605 + 530119879
10 pier 1198896
= minus9850 + 664119879
bridge And the correlation coefficients of the longitudinaldisplacements of piers and average temperature of the steeltruss arch are all larger than 092 as shown in Figure 18
A linear regression analysis between the 10 min averagedisplacement 119889
119894(119905) and the temperature 119879(119905) is further per-
formed to model the temperature-displacement correlationsusing the following equation [14]
119889119894(119905) = 120573
0+ 1205731119879 (119905) (6)
where the regression coefficients 1205731and 120573
0were obtained by
the least-squares method as
1205731=
119878119889119894119879
119878119879119879
1205730= 119889119894minus 1205731119879
(7)
where 119878119889119894119879
is the covariance between the displacement andthe temperature sequences 119878
119879119879is the variance of the mea-
sured temperature sequences and 119879 and 119889119894are the means
of the measured temperature and displacement sequencesrespectively Table 3 summarizes the expressions of linearregression models of the displacement versus temperature
The analysis results indicate that good correlation existsbetween the longitudinal displacements of piers and averagetemperature of steel truss arch A multivariable mean valuecontrol chart is also employed to monitor the measuredabnormal changes in the longitudinal displacements of piersThe condition index 119890 for early warning of abnormal behaviorof piers is defined as the difference between themeasured andpredicted longitudinal displacements
119890 = 119889119898minus 119889119901 (8)
where 119889119898is the measured displacement of the piers 119889
119901is
the predicted displacement of the piers which is obtainedby using the linear regression models shown in Table 3in the healthy condition Figure 19 shows the mean valuecontrol chart for longitudinal displacements of piers usingmeasurement data in March 2013 The upper and lowerlimits are determined using the controlling parameter 119896 =
2804 Hence a long-term monitoring on the temperature-displacement correlation model is suitable for real-timecondition assessment for movement performance of piers
34 Evaluation of Fatigue Performance of Steel Deck Fatiguecracking is universal for orthotropic steel deck of highwaybridges Once the fatigue crack appears extensive fatiguecracking may occur and repair is very difficult Henceinfinite-fatigue-life design method is applied for NanjingDashengguan Bridge instead of finite-fatigue-life designmethod which is applied extensively in highway bridges Theinfinite-fatigue-life design method is to control the fatiguestress amplitude in the range less than the value of fatiguecut-off limit refrained from fatigue cracking Consequentlythe purpose of equipping stain gauges in the SHM system ofDashengguan Bridge is to monitor the stress amplitude andthe validity of fatigue design can be achieved by comparingthe stress amplitude with fatigue stress limit
The structural dynamic response is asymmetrical becauseof the irregularity of railway and braking force from thetrain even though the structure of Nanjing DashengguanBridge is spatially symmetrical Hence the stresses are notsymmetrical measured by strain gauges from the upstreamand downstream directions However the fatigue amplitudesof two strain gauges induced by trains are both extremelysmall Thus the strain history data of the strain gauge DYB-2are merely selected in this paper to illustrate that the fatiguestress amplitude is much lower than the value of fatigue stresslimit The typical strain time-history curves of DYB-2 whensingle train with 8 carriages or 16 carriages passed throughthe bridge are shown in Figure 20 respectively It can be seenthat during the passing of high-speed trains both 8 carriagesand 16 carriages have led to strain variations reaching to 2ndash8 120583120576 And the number of strain peaks is 16 and 32 for trainswith 8 carriages and 16 carriages respectively because theaxle number of each carriage is 2
To perform fatigue evaluation simplified rainflow cyclecounting algorithm was firstly used to process strain history
14 The Scientific World Journal
R = 09748
10 20 30 40 500
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
(a) Correlation between 1198891of the 4 pier and average temperature 119879 of
steel truss arch
10 20 30 40 500
Temperature (∘C)
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
R = 09746
(b) Correlation between 1198892of the 5 pier and average temperature 119879 of
steel truss arch
10 200 40 5030
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
Disp
lace
men
t (m
m)
R = 09471
(c) Correlation between 1198893of the 6 pier and average temperature 119879 of
steel truss arch
R = 09276
10 20 30 40 500
Temperature (∘C)
minus60
minus30
0
30
60
90
120D
ispla
cem
ent (
mm
)
(d) Correlation between 1198894of the 8 pier and average temperature 119879 of
steel truss arch
R = 09536
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(e) Correlation between 1198895of the 9 pier and average temperature 119879 of
steel truss arch
R = 09537
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(f) Correlation between 1198896of the 10 pier and average temperature 119879 of
steel truss arch
Figure 18 Correlation between longitudinal displacement and average temperature
The Scientific World Journal 15
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
)
500 1000 1500 2000 2500 3000 3500 4000 45000Sample
minus15
minus10
minus5
0
5
10
15
Figure 19 Mean value control chart for longitudinal displacements of piers
Stra
in (120583
120576)
minus174
minus172
minus170
minus168
minus166
minus164
minus162
1 2714 53
66
79
92
40
118
209
183
222
105
157
196
144
131
235
248
170
Data point
(a) Single train with 8 carriages passed through the bridge
Stra
in (120583
120576)
minus180minus178minus176minus174minus172minus170minus168minus166
1
61
91
31
181
211
121
421
301
481
511
361
391
271
451
241
331
541
571
151
Data point
(b) Single train with 16 carriages passed through the bridge
Figure 20 Typical strain time-history curves of DYB-2 when single train passed through the bridge
data and the spectrum of stress amplitude was obtained[15] The spectra of stress amplitude calculated using thestrain history data with regard to 8 carriages or 16 carriagesare shown in Figure 21 respectively It can be seen that themaximum stress amplitude is smaller than 10MPa obtainedfrom strain history curves under single trainThen accordingtoMinerrsquos law and standard of Eurocode 3 fatigue parametersof 119878eq (equivalent stress amplitude) and 119873
119904(stress cycle
number) under single train were calculated as follows [16]
119878eq = (
1198991198941198785
119894
sum119899119894
)
15
119873s = sum119899119894
(9)
where 119878119894is 119894th stress amplitude 119899
119894is the number of
stress cycles corresponding to 119878119894 and 15 is slope value of
log 119878- log119873 curve in the scope of lower stress amplitudeaccording to the standard of Eurocode 3 [17]
Using (9) the equivalent stress amplitudes 119878eq with regardto 8 carriages or 16 carriages are 075MPa and 071MParespectively And the stress cycle numbers 119873
119904are 22 and 45
respectively Thus the value of 119878eq with regard to 8 carriagesis similar to that of 16 carriages However the value of 119873
119904
with regard to 8 carriages is about twice as much as that of 16carriages Furthermore as recommended by Eurocode 3 forrib-to-deck welded joint the fatigue limit Δ120590
119871is 29MPa [17]
and the value of 119878eq when the high-speed train passes throughbridge is far less than Δ120590
119871 which indicates that fatigue life of
rib-to-deck welded joint in Nanjing Dashengguan Bridge canbe considered infinite Thus the fatigue performance of steel
deck satisfies the requirement of infinite-fatigue-life designmethod
4 Conclusions
In the last few decades Structural Health Monitoring (SHM)system has been widely installed in many highway bridgesto continuously monitor the structural condition over anextended period of time However application of SHMsystem in high-speed railway bridges is far insufficient Thispaper presents an analysis of the SHM system installedin Nanjing Dashengguan Bridge in China The monitoringsystem is composed of a sensor system a data acquisitionand transmission system a data management system and astructural evaluation system The monitoring system is cur-rently in full operation accumulating the bridgersquos behaviorsunder the varying environmental conditions such as high-speed trains and environmental temperature
Based on long-termhealthmonitoring data general prin-ciples and analytical processes for structural evaluation areprovided and some evaluation results are reported It can beshown from the analysis in this paper that the mathematicalcorrelation models describing the overall structural behaviorof the bridge can be obtained with the support of the healthmonitoring system which includes cross-correlation modelsfor accelerations correlation models between temperatureand static strains of steel truss arch and correlation mod-els between temperature and longitudinal displacements ofpiers The mean value control chart is further employed tomonitor the time series of differences between the measuredstructural parameters and the predicted parameters usingthese correlation models The upper and lower control limits
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
12 The Scientific World Journal
b = 2148k = 0881
minus50
minus28
minus6
16
38
60Si
mul
ativ
eSII
I(120583120576)
minus45 minus23 minus1 43 6521Monitoring SIII (120583120576)
Scattered pointsFitting curve
(g) 1198787
b = 0615k = 1118
minus30
minus15
0
15
30
Sim
ulat
iveS
III
(120583120576)
minus15minus25 minus5 15 255Monitoring SIII (120583120576)
Scattered pointsFitting curve
(h) 1198788
Figure 16 Correlation between simulative 119878III and monitoring 119878III
Table 2 Performance parameters 1205821
1205741
1205742
1205743
and 119862 in correlationmodels
Static straindata 120582
1
1205741
1205742
1205743
119862
119878III of 1198781 0454 minus4711 minus2597 minus4944 minus46806119878III of 1198782 1585 0013 4104 minus1247 minus70238119878III of 1198783 0301 3416 minus2098 2968 minus6884119878III of 1198784 0350 minus0734 0351 0948 minus23597119878III of 1198785 0179 minus4219 1827 3165 minus22049119878III of 1198786 0062 minus3245 8242 minus5398 minus22862119878III of 1198787 minus0367 2411 minus0598 5531 17075119878III of 1198788 0611 2188 minus1130 2018 minus32799
[119865]3times16
denote transformation matrix which can be cal-culated using principal component analysis [11 12] 120582
1is
performance parameter of first potential comprehensivefactor 119875
1 9978881205741times3
= [1205741 1205742 1205743] is one vector containing
three performance parameters corresponding to first threepotential comprehensive factors 119877
1 1198772 and 119877
3 The values of
performance parameters 1205821 9978881205741times3
and 119888 are estimated usingtraining data by multivariate linear regression method [13]and are shown in Table 2
In order to verify the effectiveness of the correlationmodels shown in (2) test data in November 2013 are usedto obtain the simulative 119878III The scattered points betweensimulative 119878III and monitoring 119878III are plotted as shown inFigure 16 which are further fitted by linear function
119878119904= 119896119878119898+ 119887 (4)
where 119878119904denotes simulative 119878III 119878119898 denotes monitoring
119878III and 119896 119887 are fitting parameters with the fitting resultsshown in Figure 16 as well The fitting results verify theeffectiveness of the correlationmodels Amultivariable mean
value control chart is also employed tomonitor the measuredabnormal changes in the static strains of the steel truss archThe condition index 119890 for early warning of deterioration ofthe static performance of steel truss arch is defined as thedifference between the measured and predicted static strains
119890 = 119878119898minus 119878119901 (5)
where 119878119898is the measured static strain 119878III 119878119901 is the predicted
static strain 119878III which is obtained by using the correlationmodels shown in (2) in the healthy condition Figure 17 showsthemean value control chart for static strains of the steel trussarch using measurement data in November 2013 The upperand lower limits are determined using the controlling param-eter 119896 = 2235 Hence a long-term monitoring on the corre-lation models between temperature field and static strains issuitable for early warning of deterioration of the static per-formance of steel truss arch
33 Evaluation of Movement Performance of Pier Longi-tudinal displacements of piers and temperature field datain steel truss arch collected from March 2013 to October2013 are used for evaluation of movement performanceof piers Longitudinal displacement data from sensors 119871
119894
(119894 = 1 2 6) are denoted by 119889119894(119905) (where 119905 means time)
respectively Meanwhile temperature field data from sensors119882119894(119894 = 1 2 8) are denoted by 119879
119894(119905) respectively and
their averaged value 119879(119905) is used to represent the averagetemperature of the steel truss arch For simplicity taking10 minutes as basic time interval the 10 min average dis-placements and temperatures are calculated and used as therepresentative values of this time interval By this way thedisplacements and temperatures in one day are reduced to144 representative samples Furthermore the temperature-displacement scatter diagrams are plotted in Figure 18 Anoverall increase in longitudinal displacements of piers isobserved with an increase in the average temperature of the
The Scientific World Journal 13
UCL
LCL
CL
Con
ditio
n in
dexe
(120583120576)
minus15
minus10
minus5
0
5
10
15
10 20 30 40 50 600Sample
Figure 17 Mean value control chart for static strains of the steel truss arch
Table 3 Summary of linear regression models
Pier Regression function4 pier 119889
1
= minus11524 + 696119879
5 pier 1198892
= minus9253 + 562119879
6 pier 1198893
= minus6628 + 382119879
8 pier 1198894
= minus4658 + 336119879
9 pier 1198895
= minus7605 + 530119879
10 pier 1198896
= minus9850 + 664119879
bridge And the correlation coefficients of the longitudinaldisplacements of piers and average temperature of the steeltruss arch are all larger than 092 as shown in Figure 18
A linear regression analysis between the 10 min averagedisplacement 119889
119894(119905) and the temperature 119879(119905) is further per-
formed to model the temperature-displacement correlationsusing the following equation [14]
119889119894(119905) = 120573
0+ 1205731119879 (119905) (6)
where the regression coefficients 1205731and 120573
0were obtained by
the least-squares method as
1205731=
119878119889119894119879
119878119879119879
1205730= 119889119894minus 1205731119879
(7)
where 119878119889119894119879
is the covariance between the displacement andthe temperature sequences 119878
119879119879is the variance of the mea-
sured temperature sequences and 119879 and 119889119894are the means
of the measured temperature and displacement sequencesrespectively Table 3 summarizes the expressions of linearregression models of the displacement versus temperature
The analysis results indicate that good correlation existsbetween the longitudinal displacements of piers and averagetemperature of steel truss arch A multivariable mean valuecontrol chart is also employed to monitor the measuredabnormal changes in the longitudinal displacements of piersThe condition index 119890 for early warning of abnormal behaviorof piers is defined as the difference between themeasured andpredicted longitudinal displacements
119890 = 119889119898minus 119889119901 (8)
where 119889119898is the measured displacement of the piers 119889
119901is
the predicted displacement of the piers which is obtainedby using the linear regression models shown in Table 3in the healthy condition Figure 19 shows the mean valuecontrol chart for longitudinal displacements of piers usingmeasurement data in March 2013 The upper and lowerlimits are determined using the controlling parameter 119896 =
2804 Hence a long-term monitoring on the temperature-displacement correlation model is suitable for real-timecondition assessment for movement performance of piers
34 Evaluation of Fatigue Performance of Steel Deck Fatiguecracking is universal for orthotropic steel deck of highwaybridges Once the fatigue crack appears extensive fatiguecracking may occur and repair is very difficult Henceinfinite-fatigue-life design method is applied for NanjingDashengguan Bridge instead of finite-fatigue-life designmethod which is applied extensively in highway bridges Theinfinite-fatigue-life design method is to control the fatiguestress amplitude in the range less than the value of fatiguecut-off limit refrained from fatigue cracking Consequentlythe purpose of equipping stain gauges in the SHM system ofDashengguan Bridge is to monitor the stress amplitude andthe validity of fatigue design can be achieved by comparingthe stress amplitude with fatigue stress limit
The structural dynamic response is asymmetrical becauseof the irregularity of railway and braking force from thetrain even though the structure of Nanjing DashengguanBridge is spatially symmetrical Hence the stresses are notsymmetrical measured by strain gauges from the upstreamand downstream directions However the fatigue amplitudesof two strain gauges induced by trains are both extremelysmall Thus the strain history data of the strain gauge DYB-2are merely selected in this paper to illustrate that the fatiguestress amplitude is much lower than the value of fatigue stresslimit The typical strain time-history curves of DYB-2 whensingle train with 8 carriages or 16 carriages passed throughthe bridge are shown in Figure 20 respectively It can be seenthat during the passing of high-speed trains both 8 carriagesand 16 carriages have led to strain variations reaching to 2ndash8 120583120576 And the number of strain peaks is 16 and 32 for trainswith 8 carriages and 16 carriages respectively because theaxle number of each carriage is 2
To perform fatigue evaluation simplified rainflow cyclecounting algorithm was firstly used to process strain history
14 The Scientific World Journal
R = 09748
10 20 30 40 500
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
(a) Correlation between 1198891of the 4 pier and average temperature 119879 of
steel truss arch
10 20 30 40 500
Temperature (∘C)
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
R = 09746
(b) Correlation between 1198892of the 5 pier and average temperature 119879 of
steel truss arch
10 200 40 5030
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
Disp
lace
men
t (m
m)
R = 09471
(c) Correlation between 1198893of the 6 pier and average temperature 119879 of
steel truss arch
R = 09276
10 20 30 40 500
Temperature (∘C)
minus60
minus30
0
30
60
90
120D
ispla
cem
ent (
mm
)
(d) Correlation between 1198894of the 8 pier and average temperature 119879 of
steel truss arch
R = 09536
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(e) Correlation between 1198895of the 9 pier and average temperature 119879 of
steel truss arch
R = 09537
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(f) Correlation between 1198896of the 10 pier and average temperature 119879 of
steel truss arch
Figure 18 Correlation between longitudinal displacement and average temperature
The Scientific World Journal 15
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
)
500 1000 1500 2000 2500 3000 3500 4000 45000Sample
minus15
minus10
minus5
0
5
10
15
Figure 19 Mean value control chart for longitudinal displacements of piers
Stra
in (120583
120576)
minus174
minus172
minus170
minus168
minus166
minus164
minus162
1 2714 53
66
79
92
40
118
209
183
222
105
157
196
144
131
235
248
170
Data point
(a) Single train with 8 carriages passed through the bridge
Stra
in (120583
120576)
minus180minus178minus176minus174minus172minus170minus168minus166
1
61
91
31
181
211
121
421
301
481
511
361
391
271
451
241
331
541
571
151
Data point
(b) Single train with 16 carriages passed through the bridge
Figure 20 Typical strain time-history curves of DYB-2 when single train passed through the bridge
data and the spectrum of stress amplitude was obtained[15] The spectra of stress amplitude calculated using thestrain history data with regard to 8 carriages or 16 carriagesare shown in Figure 21 respectively It can be seen that themaximum stress amplitude is smaller than 10MPa obtainedfrom strain history curves under single trainThen accordingtoMinerrsquos law and standard of Eurocode 3 fatigue parametersof 119878eq (equivalent stress amplitude) and 119873
119904(stress cycle
number) under single train were calculated as follows [16]
119878eq = (
1198991198941198785
119894
sum119899119894
)
15
119873s = sum119899119894
(9)
where 119878119894is 119894th stress amplitude 119899
119894is the number of
stress cycles corresponding to 119878119894 and 15 is slope value of
log 119878- log119873 curve in the scope of lower stress amplitudeaccording to the standard of Eurocode 3 [17]
Using (9) the equivalent stress amplitudes 119878eq with regardto 8 carriages or 16 carriages are 075MPa and 071MParespectively And the stress cycle numbers 119873
119904are 22 and 45
respectively Thus the value of 119878eq with regard to 8 carriagesis similar to that of 16 carriages However the value of 119873
119904
with regard to 8 carriages is about twice as much as that of 16carriages Furthermore as recommended by Eurocode 3 forrib-to-deck welded joint the fatigue limit Δ120590
119871is 29MPa [17]
and the value of 119878eq when the high-speed train passes throughbridge is far less than Δ120590
119871 which indicates that fatigue life of
rib-to-deck welded joint in Nanjing Dashengguan Bridge canbe considered infinite Thus the fatigue performance of steel
deck satisfies the requirement of infinite-fatigue-life designmethod
4 Conclusions
In the last few decades Structural Health Monitoring (SHM)system has been widely installed in many highway bridgesto continuously monitor the structural condition over anextended period of time However application of SHMsystem in high-speed railway bridges is far insufficient Thispaper presents an analysis of the SHM system installedin Nanjing Dashengguan Bridge in China The monitoringsystem is composed of a sensor system a data acquisitionand transmission system a data management system and astructural evaluation system The monitoring system is cur-rently in full operation accumulating the bridgersquos behaviorsunder the varying environmental conditions such as high-speed trains and environmental temperature
Based on long-termhealthmonitoring data general prin-ciples and analytical processes for structural evaluation areprovided and some evaluation results are reported It can beshown from the analysis in this paper that the mathematicalcorrelation models describing the overall structural behaviorof the bridge can be obtained with the support of the healthmonitoring system which includes cross-correlation modelsfor accelerations correlation models between temperatureand static strains of steel truss arch and correlation mod-els between temperature and longitudinal displacements ofpiers The mean value control chart is further employed tomonitor the time series of differences between the measuredstructural parameters and the predicted parameters usingthese correlation models The upper and lower control limits
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
The Scientific World Journal 13
UCL
LCL
CL
Con
ditio
n in
dexe
(120583120576)
minus15
minus10
minus5
0
5
10
15
10 20 30 40 50 600Sample
Figure 17 Mean value control chart for static strains of the steel truss arch
Table 3 Summary of linear regression models
Pier Regression function4 pier 119889
1
= minus11524 + 696119879
5 pier 1198892
= minus9253 + 562119879
6 pier 1198893
= minus6628 + 382119879
8 pier 1198894
= minus4658 + 336119879
9 pier 1198895
= minus7605 + 530119879
10 pier 1198896
= minus9850 + 664119879
bridge And the correlation coefficients of the longitudinaldisplacements of piers and average temperature of the steeltruss arch are all larger than 092 as shown in Figure 18
A linear regression analysis between the 10 min averagedisplacement 119889
119894(119905) and the temperature 119879(119905) is further per-
formed to model the temperature-displacement correlationsusing the following equation [14]
119889119894(119905) = 120573
0+ 1205731119879 (119905) (6)
where the regression coefficients 1205731and 120573
0were obtained by
the least-squares method as
1205731=
119878119889119894119879
119878119879119879
1205730= 119889119894minus 1205731119879
(7)
where 119878119889119894119879
is the covariance between the displacement andthe temperature sequences 119878
119879119879is the variance of the mea-
sured temperature sequences and 119879 and 119889119894are the means
of the measured temperature and displacement sequencesrespectively Table 3 summarizes the expressions of linearregression models of the displacement versus temperature
The analysis results indicate that good correlation existsbetween the longitudinal displacements of piers and averagetemperature of steel truss arch A multivariable mean valuecontrol chart is also employed to monitor the measuredabnormal changes in the longitudinal displacements of piersThe condition index 119890 for early warning of abnormal behaviorof piers is defined as the difference between themeasured andpredicted longitudinal displacements
119890 = 119889119898minus 119889119901 (8)
where 119889119898is the measured displacement of the piers 119889
119901is
the predicted displacement of the piers which is obtainedby using the linear regression models shown in Table 3in the healthy condition Figure 19 shows the mean valuecontrol chart for longitudinal displacements of piers usingmeasurement data in March 2013 The upper and lowerlimits are determined using the controlling parameter 119896 =
2804 Hence a long-term monitoring on the temperature-displacement correlation model is suitable for real-timecondition assessment for movement performance of piers
34 Evaluation of Fatigue Performance of Steel Deck Fatiguecracking is universal for orthotropic steel deck of highwaybridges Once the fatigue crack appears extensive fatiguecracking may occur and repair is very difficult Henceinfinite-fatigue-life design method is applied for NanjingDashengguan Bridge instead of finite-fatigue-life designmethod which is applied extensively in highway bridges Theinfinite-fatigue-life design method is to control the fatiguestress amplitude in the range less than the value of fatiguecut-off limit refrained from fatigue cracking Consequentlythe purpose of equipping stain gauges in the SHM system ofDashengguan Bridge is to monitor the stress amplitude andthe validity of fatigue design can be achieved by comparingthe stress amplitude with fatigue stress limit
The structural dynamic response is asymmetrical becauseof the irregularity of railway and braking force from thetrain even though the structure of Nanjing DashengguanBridge is spatially symmetrical Hence the stresses are notsymmetrical measured by strain gauges from the upstreamand downstream directions However the fatigue amplitudesof two strain gauges induced by trains are both extremelysmall Thus the strain history data of the strain gauge DYB-2are merely selected in this paper to illustrate that the fatiguestress amplitude is much lower than the value of fatigue stresslimit The typical strain time-history curves of DYB-2 whensingle train with 8 carriages or 16 carriages passed throughthe bridge are shown in Figure 20 respectively It can be seenthat during the passing of high-speed trains both 8 carriagesand 16 carriages have led to strain variations reaching to 2ndash8 120583120576 And the number of strain peaks is 16 and 32 for trainswith 8 carriages and 16 carriages respectively because theaxle number of each carriage is 2
To perform fatigue evaluation simplified rainflow cyclecounting algorithm was firstly used to process strain history
14 The Scientific World Journal
R = 09748
10 20 30 40 500
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
(a) Correlation between 1198891of the 4 pier and average temperature 119879 of
steel truss arch
10 20 30 40 500
Temperature (∘C)
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
R = 09746
(b) Correlation between 1198892of the 5 pier and average temperature 119879 of
steel truss arch
10 200 40 5030
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
Disp
lace
men
t (m
m)
R = 09471
(c) Correlation between 1198893of the 6 pier and average temperature 119879 of
steel truss arch
R = 09276
10 20 30 40 500
Temperature (∘C)
minus60
minus30
0
30
60
90
120D
ispla
cem
ent (
mm
)
(d) Correlation between 1198894of the 8 pier and average temperature 119879 of
steel truss arch
R = 09536
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(e) Correlation between 1198895of the 9 pier and average temperature 119879 of
steel truss arch
R = 09537
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(f) Correlation between 1198896of the 10 pier and average temperature 119879 of
steel truss arch
Figure 18 Correlation between longitudinal displacement and average temperature
The Scientific World Journal 15
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
)
500 1000 1500 2000 2500 3000 3500 4000 45000Sample
minus15
minus10
minus5
0
5
10
15
Figure 19 Mean value control chart for longitudinal displacements of piers
Stra
in (120583
120576)
minus174
minus172
minus170
minus168
minus166
minus164
minus162
1 2714 53
66
79
92
40
118
209
183
222
105
157
196
144
131
235
248
170
Data point
(a) Single train with 8 carriages passed through the bridge
Stra
in (120583
120576)
minus180minus178minus176minus174minus172minus170minus168minus166
1
61
91
31
181
211
121
421
301
481
511
361
391
271
451
241
331
541
571
151
Data point
(b) Single train with 16 carriages passed through the bridge
Figure 20 Typical strain time-history curves of DYB-2 when single train passed through the bridge
data and the spectrum of stress amplitude was obtained[15] The spectra of stress amplitude calculated using thestrain history data with regard to 8 carriages or 16 carriagesare shown in Figure 21 respectively It can be seen that themaximum stress amplitude is smaller than 10MPa obtainedfrom strain history curves under single trainThen accordingtoMinerrsquos law and standard of Eurocode 3 fatigue parametersof 119878eq (equivalent stress amplitude) and 119873
119904(stress cycle
number) under single train were calculated as follows [16]
119878eq = (
1198991198941198785
119894
sum119899119894
)
15
119873s = sum119899119894
(9)
where 119878119894is 119894th stress amplitude 119899
119894is the number of
stress cycles corresponding to 119878119894 and 15 is slope value of
log 119878- log119873 curve in the scope of lower stress amplitudeaccording to the standard of Eurocode 3 [17]
Using (9) the equivalent stress amplitudes 119878eq with regardto 8 carriages or 16 carriages are 075MPa and 071MParespectively And the stress cycle numbers 119873
119904are 22 and 45
respectively Thus the value of 119878eq with regard to 8 carriagesis similar to that of 16 carriages However the value of 119873
119904
with regard to 8 carriages is about twice as much as that of 16carriages Furthermore as recommended by Eurocode 3 forrib-to-deck welded joint the fatigue limit Δ120590
119871is 29MPa [17]
and the value of 119878eq when the high-speed train passes throughbridge is far less than Δ120590
119871 which indicates that fatigue life of
rib-to-deck welded joint in Nanjing Dashengguan Bridge canbe considered infinite Thus the fatigue performance of steel
deck satisfies the requirement of infinite-fatigue-life designmethod
4 Conclusions
In the last few decades Structural Health Monitoring (SHM)system has been widely installed in many highway bridgesto continuously monitor the structural condition over anextended period of time However application of SHMsystem in high-speed railway bridges is far insufficient Thispaper presents an analysis of the SHM system installedin Nanjing Dashengguan Bridge in China The monitoringsystem is composed of a sensor system a data acquisitionand transmission system a data management system and astructural evaluation system The monitoring system is cur-rently in full operation accumulating the bridgersquos behaviorsunder the varying environmental conditions such as high-speed trains and environmental temperature
Based on long-termhealthmonitoring data general prin-ciples and analytical processes for structural evaluation areprovided and some evaluation results are reported It can beshown from the analysis in this paper that the mathematicalcorrelation models describing the overall structural behaviorof the bridge can be obtained with the support of the healthmonitoring system which includes cross-correlation modelsfor accelerations correlation models between temperatureand static strains of steel truss arch and correlation mod-els between temperature and longitudinal displacements ofpiers The mean value control chart is further employed tomonitor the time series of differences between the measuredstructural parameters and the predicted parameters usingthese correlation models The upper and lower control limits
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
14 The Scientific World Journal
R = 09748
10 20 30 40 500
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
(a) Correlation between 1198891of the 4 pier and average temperature 119879 of
steel truss arch
10 20 30 40 500
Temperature (∘C)
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
R = 09746
(b) Correlation between 1198892of the 5 pier and average temperature 119879 of
steel truss arch
10 200 40 5030
Temperature (∘C)
minus150
minus100
minus50
0
50
100
150
Disp
lace
men
t (m
m)
R = 09471
(c) Correlation between 1198893of the 6 pier and average temperature 119879 of
steel truss arch
R = 09276
10 20 30 40 500
Temperature (∘C)
minus60
minus30
0
30
60
90
120D
ispla
cem
ent (
mm
)
(d) Correlation between 1198894of the 8 pier and average temperature 119879 of
steel truss arch
R = 09536
minus100
minus50
0
50
100
150
200
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(e) Correlation between 1198895of the 9 pier and average temperature 119879 of
steel truss arch
R = 09537
minus100
minus50
0
50
100
150
200
250
Disp
lace
men
t (m
m)
10 20 30 40 500
Temperature (∘C)
(f) Correlation between 1198896of the 10 pier and average temperature 119879 of
steel truss arch
Figure 18 Correlation between longitudinal displacement and average temperature
The Scientific World Journal 15
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
)
500 1000 1500 2000 2500 3000 3500 4000 45000Sample
minus15
minus10
minus5
0
5
10
15
Figure 19 Mean value control chart for longitudinal displacements of piers
Stra
in (120583
120576)
minus174
minus172
minus170
minus168
minus166
minus164
minus162
1 2714 53
66
79
92
40
118
209
183
222
105
157
196
144
131
235
248
170
Data point
(a) Single train with 8 carriages passed through the bridge
Stra
in (120583
120576)
minus180minus178minus176minus174minus172minus170minus168minus166
1
61
91
31
181
211
121
421
301
481
511
361
391
271
451
241
331
541
571
151
Data point
(b) Single train with 16 carriages passed through the bridge
Figure 20 Typical strain time-history curves of DYB-2 when single train passed through the bridge
data and the spectrum of stress amplitude was obtained[15] The spectra of stress amplitude calculated using thestrain history data with regard to 8 carriages or 16 carriagesare shown in Figure 21 respectively It can be seen that themaximum stress amplitude is smaller than 10MPa obtainedfrom strain history curves under single trainThen accordingtoMinerrsquos law and standard of Eurocode 3 fatigue parametersof 119878eq (equivalent stress amplitude) and 119873
119904(stress cycle
number) under single train were calculated as follows [16]
119878eq = (
1198991198941198785
119894
sum119899119894
)
15
119873s = sum119899119894
(9)
where 119878119894is 119894th stress amplitude 119899
119894is the number of
stress cycles corresponding to 119878119894 and 15 is slope value of
log 119878- log119873 curve in the scope of lower stress amplitudeaccording to the standard of Eurocode 3 [17]
Using (9) the equivalent stress amplitudes 119878eq with regardto 8 carriages or 16 carriages are 075MPa and 071MParespectively And the stress cycle numbers 119873
119904are 22 and 45
respectively Thus the value of 119878eq with regard to 8 carriagesis similar to that of 16 carriages However the value of 119873
119904
with regard to 8 carriages is about twice as much as that of 16carriages Furthermore as recommended by Eurocode 3 forrib-to-deck welded joint the fatigue limit Δ120590
119871is 29MPa [17]
and the value of 119878eq when the high-speed train passes throughbridge is far less than Δ120590
119871 which indicates that fatigue life of
rib-to-deck welded joint in Nanjing Dashengguan Bridge canbe considered infinite Thus the fatigue performance of steel
deck satisfies the requirement of infinite-fatigue-life designmethod
4 Conclusions
In the last few decades Structural Health Monitoring (SHM)system has been widely installed in many highway bridgesto continuously monitor the structural condition over anextended period of time However application of SHMsystem in high-speed railway bridges is far insufficient Thispaper presents an analysis of the SHM system installedin Nanjing Dashengguan Bridge in China The monitoringsystem is composed of a sensor system a data acquisitionand transmission system a data management system and astructural evaluation system The monitoring system is cur-rently in full operation accumulating the bridgersquos behaviorsunder the varying environmental conditions such as high-speed trains and environmental temperature
Based on long-termhealthmonitoring data general prin-ciples and analytical processes for structural evaluation areprovided and some evaluation results are reported It can beshown from the analysis in this paper that the mathematicalcorrelation models describing the overall structural behaviorof the bridge can be obtained with the support of the healthmonitoring system which includes cross-correlation modelsfor accelerations correlation models between temperatureand static strains of steel truss arch and correlation mod-els between temperature and longitudinal displacements ofpiers The mean value control chart is further employed tomonitor the time series of differences between the measuredstructural parameters and the predicted parameters usingthese correlation models The upper and lower control limits
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
The Scientific World Journal 15
UCL
LCL
CL
Con
ditio
n in
dexe
(mm
)
500 1000 1500 2000 2500 3000 3500 4000 45000Sample
minus15
minus10
minus5
0
5
10
15
Figure 19 Mean value control chart for longitudinal displacements of piers
Stra
in (120583
120576)
minus174
minus172
minus170
minus168
minus166
minus164
minus162
1 2714 53
66
79
92
40
118
209
183
222
105
157
196
144
131
235
248
170
Data point
(a) Single train with 8 carriages passed through the bridge
Stra
in (120583
120576)
minus180minus178minus176minus174minus172minus170minus168minus166
1
61
91
31
181
211
121
421
301
481
511
361
391
271
451
241
331
541
571
151
Data point
(b) Single train with 16 carriages passed through the bridge
Figure 20 Typical strain time-history curves of DYB-2 when single train passed through the bridge
data and the spectrum of stress amplitude was obtained[15] The spectra of stress amplitude calculated using thestrain history data with regard to 8 carriages or 16 carriagesare shown in Figure 21 respectively It can be seen that themaximum stress amplitude is smaller than 10MPa obtainedfrom strain history curves under single trainThen accordingtoMinerrsquos law and standard of Eurocode 3 fatigue parametersof 119878eq (equivalent stress amplitude) and 119873
119904(stress cycle
number) under single train were calculated as follows [16]
119878eq = (
1198991198941198785
119894
sum119899119894
)
15
119873s = sum119899119894
(9)
where 119878119894is 119894th stress amplitude 119899
119894is the number of
stress cycles corresponding to 119878119894 and 15 is slope value of
log 119878- log119873 curve in the scope of lower stress amplitudeaccording to the standard of Eurocode 3 [17]
Using (9) the equivalent stress amplitudes 119878eq with regardto 8 carriages or 16 carriages are 075MPa and 071MParespectively And the stress cycle numbers 119873
119904are 22 and 45
respectively Thus the value of 119878eq with regard to 8 carriagesis similar to that of 16 carriages However the value of 119873
119904
with regard to 8 carriages is about twice as much as that of 16carriages Furthermore as recommended by Eurocode 3 forrib-to-deck welded joint the fatigue limit Δ120590
119871is 29MPa [17]
and the value of 119878eq when the high-speed train passes throughbridge is far less than Δ120590
119871 which indicates that fatigue life of
rib-to-deck welded joint in Nanjing Dashengguan Bridge canbe considered infinite Thus the fatigue performance of steel
deck satisfies the requirement of infinite-fatigue-life designmethod
4 Conclusions
In the last few decades Structural Health Monitoring (SHM)system has been widely installed in many highway bridgesto continuously monitor the structural condition over anextended period of time However application of SHMsystem in high-speed railway bridges is far insufficient Thispaper presents an analysis of the SHM system installedin Nanjing Dashengguan Bridge in China The monitoringsystem is composed of a sensor system a data acquisitionand transmission system a data management system and astructural evaluation system The monitoring system is cur-rently in full operation accumulating the bridgersquos behaviorsunder the varying environmental conditions such as high-speed trains and environmental temperature
Based on long-termhealthmonitoring data general prin-ciples and analytical processes for structural evaluation areprovided and some evaluation results are reported It can beshown from the analysis in this paper that the mathematicalcorrelation models describing the overall structural behaviorof the bridge can be obtained with the support of the healthmonitoring system which includes cross-correlation modelsfor accelerations correlation models between temperatureand static strains of steel truss arch and correlation mod-els between temperature and longitudinal displacements ofpiers The mean value control chart is further employed tomonitor the time series of differences between the measuredstructural parameters and the predicted parameters usingthese correlation models The upper and lower control limits
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
16 The Scientific World Journal
02 03 04 05 06 07 08 09 1001
Stress amplitude (MPa)
Num
ber o
f stre
ss cy
cles
100
80
70
60
50
40
30
20
10
90
0
(a) 8 carriages
0
5
10
15
20
25
30
35
40
Num
ber o
f stre
ss cy
cles
020 080010 050 060 070030 090040
Stress amplitude (MPa)
(b) 16 carriages
Figure 21 Spectra of stress amplitude calculated using the strain history data
of the control chart are then determined and abnormalbehaviors can be detected if the time series of differencesfall beyond the control limits during the monitoring periodwhich can result in useful suggestions to bridge managementauthorities by reasonably evaluating the monitored data
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The authors gratefully acknowledge the NationalBasic Research Program of China (973 Program) (no2015CB060000) the National Science and TechnologySupport ProgramofChina (no 2014BAG07B01) theKey Pro-gram of National Natural Science Foundation (no 51438002)and the Program of ldquoSix Major Talent Summitrdquo Foundation(no 1105000268)
References
[1] E Watanabe H Furuta T Yamaguchi and M Kano ldquoOnlongevity and monitoring technologies of bridges a surveystudy by the Japanese Society of Steel Constructionrdquo Structureand Infrastructure Engineering vol 10 no 4 pp 471ndash491 2014
[2] S L LiH Li Y Liu C LanWZhou and JOu ldquoSMCstructuralhealth monitoring benchmark problem using monitored datafrom an actual cable-stayed bridgerdquo Structural Control andHealth Monitoring vol 21 no 2 pp 156ndash172 2014
[3] K-Y Wong ldquoInstrumentation and health monitoring of cable-supported bridgesrdquo Structural Control and Health Monitoringvol 11 no 2 pp 91ndash124 2004
[4] A Q Li Y L Ding H Wang and T Guo ldquoAnalysis and assess-ment of bridge health monitoring mass datamdashprogress inresearchdevelopment of lsquoStructural Health Monitoringrsquordquo Sci-ence China Technological Sciences vol 55 no 8 pp 2212ndash22242012
[5] D Yang D Youliang and L Aiqun ldquoStructural conditionassessment of long-span suspension bridges using long-termmonitoring datardquo Earthquake Engineering and EngineeringVibration vol 9 no 1 pp 123ndash131 2010
[6] C Y Xia H Xia and G De Roeck ldquoDynamic response of atrain-bridge system under collision loads and running safetyevaluation of high-speed trainsrdquo Computers and Structures vol140 pp 23ndash38 2014
[7] S H Ju ldquoImprovement of bridge structures to increase thesafety of moving trains during earthquakesrdquo Engineering Struc-tures vol 56 pp 501ndash508 2013
[8] Q Y Zeng J Xiang P Lou and Z Zhou ldquoMechanical mech-anism of derailment and theory of derailment preventionrdquo Jour-nal of Railway Science and Engineering vol 1 no 1 pp 19ndash312004
[9] T Figlus S Liscak A Wilk and B Łazarz ldquoCondition moni-toring of engine timing system by using wavelet packet decom-position of a acoustic signalrdquo Journal of Mechanical Science andTechnology vol 28 no 5 pp 1663ndash1671 2014
[10] R Wald T M Khoshgoftaar and J C Sloan ldquoFeature selectionfor optimization of wavelet packet decomposition in reliabilityanalysis of systemsrdquo International Journal on Artificial Intelli-gence Tools vol 22 no 5 Article ID 1360011 2013
[11] H F Zhou Y Q Ni and J M Ko ldquoConstructing input toneural networks for modeling temperature-caused modal vari-ability mean temperatures effective temperatures and princi-pal components of temperaturesrdquo Engineering Structures vol32 no 6 pp 1747ndash1759 2010
[12] A Singh S Datta and S S Mahapatra ldquoPrincipal componentanalysis and fuzzy embedded Taguchi approach for multi-res-ponse optimization in machining of GFRP polyester compos-ites a case studyrdquo International Journal of Industrial and SystemsEngineering vol 14 no 2 pp 175ndash206 2013
[13] A Amiri A Saghaei M Mohseni and Y Zerehsaz ldquoDiagnosisaids in multivariate multiple linear regression profiles monitor-ingrdquo Communications in StatisticsmdashTheory and Methods vol43 no 14 pp 3057ndash3079 2014
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
The Scientific World Journal 17
[14] Y Q Ni X G Hua K Y Wong and J M Ko ldquoAssessment ofbridge expansion joints using long-term displacement and tem-perature measurementrdquo Journal of Performance of ConstructedFacilities vol 21 no 2 pp 143ndash151 2007
[15] S Ya K Yamada and T Ishikawa ldquoFatigue evaluation of rib-to-deck welded joints of orthotropic steel bridge deckrdquo Journal ofBridge Engineering vol 16 no 4 pp 492ndash499 2011
[16] Y S Song and Y L Ding ldquoFatigue monitoring and analysis oforthotropic steel deck considering traffic volume and ambienttemperaturerdquo Science China Technological Sciences vol 56 no7 pp 1758ndash1766 2013
[17] European Committee for Standardization ldquoEurocode 3 designof steel structures part 1ndash9 fatiguerdquo BS EN 1993-1-92005European Committee for Standardization 2005
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of