Distributed Strain Measurement in Steel Bridge with FiberOptic Sensors: Validation through Diagnostic Load Test

Fabio Matta1; Filippo Bastianini2; Nestore Galati3; Paolo Casadei4; and Antonio Nanni, F.ASCE5

Abstract: Fiber optic sensing technologies are emerging as valid alternatives for the health monitoring of civil structures. Distributedsensors based on Brillouin scattering add the unique capability of measuring strain and temperature profiles along optical fibers. Mea-surement is performed by establishing the correlation between fiber strain and temperature, and the frequency shift of the Brillouinbackscattered light induced by a monochromatic light pulse. The technology holds potential for use on large structures and integratedtransportation infrastructure. Its effectiveness has been assessed through scaled laboratory experiments, whereas field validation is limitedto very few demonstration projects conducted to date. This paper presents a pilot application of Brillouin optical time domain reflecto-metry to measure strain profiles along the steel girders of a continuous slab-on-girder bridge subjected to diagnostic load testing. One ofthe exterior continuous girders required heat-straightening after falling during construction due to wind. The significance of applying adistributed measurement technique lies in the potential to assess the global girder response, which would be impractical and uneconomicalusing discrete measurement techniques. A 1.16 km long sensing circuit was installed onto the web of four girders. The circuit comprisesbare optical fiber sensors, and a novel adhesively bonded fiberglass tape with embedded sensing fibers for strain measurement and thermalcompensation. The strain profiles were first converted into deflection profiles and validated against discrete deflection measurementsperformed with a high-precision total station system. Structural assessment based on comparison of the strain profiles with the results ofthree-dimensional finite-element analysis of the bridge superstructure, and with specification mandated criteria, indicated that the responseof the girder under investigation was within the design limits, and did not pose serviceability concerns. Factors that may affect measure-ment accuracy are finally discussed on the basis of the experimental and numerical results.

DOI: 10.1061/�ASCE�0887-3828�2008�22:4�264�

CE Database subject headings: Assessments; Bridge tests; Fiber optics; Sensors; Strain measurement; Bridges, steel.

Introduction

Recent advances in fiber optic sensor �FOS� technologies havefostered the development of innovative solutions for the healthmonitoring of civil engineering structures �Uttamchandani 1994;Casas and Cruz 2003; Li et al. 2004�. Compelling advantages ofFOSs over conventional electrical sensors such as resistancestrain gauges are: the easy embeddability, due to their lightweight, high degree of miniaturization, and corrosion and chemi-cal resistance; the multiplexability that allows one to series-

1Research Assistant Professor, Dept. of Civil, Architectural, andEnvironmental Engineering, and Associate Director, NSF I/UCRU�Repair of Buildings and Bridges with Composites�, Univ. of Miami,Coral Gables, FL 33146 �corresponding author�. E-mail: fmatta@miami.edu

2Sesto Sensor—Metalmobile S.r.l., Zola Predosa, Bologna, Italy.3Structural Group, Inc., Elkridge, MD 21075.4Tec.Inn. S.r.l.—Innovative Technologies, Milan, Italy.5Lester & Gwen Fisher Endowed Scholar, Professor and Chair, Dept.

of Civil, Architectural, and Environmental Engineering, Univ. of Miami,Coral Gables, FL 33146.

Note. Discussion open until January 1, 2009. Separate discussionsmust be submitted for individual papers. To extend the closing date byone month, a written request must be filed with the ASCE ManagingEditor. The manuscript for this paper was submitted for review and pos-sible publication on October 25, 2007; approved on March 31, 2008. Thispaper is part of the Journal of Performance of Constructed Facilities,Vol. 22, No. 4, August 1, 2008. ©ASCE, ISSN 0887-3828/2008/4-264–

273/$25.00.

264 / JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES © AS

Downloaded 05 Jan 2009 to 131.151.86.242. Redistribution subject to

connect multiple FOSs; the immunity to electromagnetic fields,due to the sensors dielectric nature; and the suitability for remotemonitoring.

Brillouin optical time domain reflectometry �BOTDR� FOSsbased on Brillouin scattering �Brillouin 1922� add the unique abil-ity to measure the long-range �of the order of tens of kilometers�distributed strain and temperature along standard telecom-gradeoptical fibers, otherwise attainable in a quasidistributed fashiononly with several sensors applied at discrete locations. In addi-tion, access to only one end of the circuit is required, therebyenabling measurements even in case of damage or interruption ofthe sensing circuit at a random point. The technology holds sig-nificant potential for the health monitoring of large structures,including bridges, buildings, dams, nuclear reactors, pipelines,stadiums, tunnels �Komatsu et al. 2002�, and, in perspective, in-tegrated transportation systems �Fujihashi et al. 2003�.

Advancing the validation of BOTDR techniques is critical toaddress the development of monitoring systems with improvedaccuracy and spatial resolution tailored for civil applications.However, very few demonstrative field applications have substan-tiated the promising outcomes of a number of large-scale labora-tory experiments �Naruse et al. 2000; Yasue et al. 2000;Bastianini et al. 2003; Murayama et al. 2003; Wu et al. 2006�.Thévenaz et al. �1988� used a sensing system based on stimulatedBrillouin scattering to monitor the temperature of concrete duringthe curing process of a 15�20�3 m3 casting in a dam in Luz-zone �Switzerland�, and to observe the temperature fluctuations ofdeep waters in relation to seasonal conditions in the Lake of

Geneva. Ohno et al. �2002� measured the vertical strain of two

CE / JULY/AUGUST 2008

ASCE license or copyright; see http://pubs.asce.org/copyright

cast-in-place concrete foundation piles with a diameter of 1.2 mand lengths of 11.0 and 7.6 m, respectively, which were tested toevaluate the contribution of frictional forces to the bearing capac-ity. FOSs were installed into grooves cut along the steel rebarsand filled with epoxy resin. Kihara et al. �2002� measured strainsin the Nyodo River �Japan� levee to assess effectiveness in de-tecting early stages of collapse due to water penetration. Shi et al.�2003� report on the six-month monitoring of a 750 m portion ofthe concrete box structure of the Gulou Tunnel in Nanjing �P.R. ofChina�, whose deformations were verified to lie within the safetylimits. Bastianini et al. �2005a� used a BOTDR system to measurethe deformations of externally bonded “smart” fiber reinforcedpolymer �FRP� sheets with embedded FOSs, which were used forthe seismic retrofit of masonry vaults and walls of the historicalElmi-Pandolfi building �1600� in Foligno, Italy, during an in situload test. In another field project on two small concrete bridgessubjected to load test, Bastianini et al. �2005c� used bonded smartFRP tapes to assess effectiveness with respect to stand-alone FOScables.

This paper presents a pilot application of BOTDR for the dis-tributed strain measurement in the high performance steel I gird-ers of a multispan highway bridge in Missouri, which wassubjected to a diagnostic load test. The research objectives werethe following.• Assess the structural response of an exterior girder that had

required heat straightening after falling during constructionfrom a height of about 15 m, due to the effect of high wind.The implementation of a distributed measurement techniquebecomes relevant to attempt studying the global girder re-sponse, when discrete measurement techniques present objec-tive practical and economical limitations; and

• Evaluate the performance of a commercially availableBOTDR system in conjunction with a novel fiberglass tapewith embedded FOSs for strain and temperature measurement,when implemented on a large bridge structure subjected tocontrolled loads, and with the FOSs being installed in the fieldand in nonartificial conditions compared to that of thelaboratory.

First, the measured strain profiles are rendered as deflectioncurves and validated against benchmark vertical displacementsmeasured at discrete locations using a state-of-the-art high-precision automated total station �ATS� system. Then, the vali-dated strain measurements are discussed and compared to thetheoretical strains from three-dimensional finite-element analysis,and from one-dimensional beam analysis that accounts for thegirder load distribution factor mandated by the specifications usedin design �MoDOT 2002�.

BOTDR Sensing Principles

Spontaneous Brillouin scattering arises from the interactionbetween optical and acoustic waves propagating in the samephysical medium. When the medium is illuminated with a mono-chromatic light source, a partial energy transfer occurs betweenthe colliding photons and phonons, the latter being generated byeither pressure or temperature fluctuations. The resulting changein density of the medium, and therefore in its refractive index andmechanical properties �elastic modulus and Poisson’s ratio�, de-termines a Doppler shift in the frequency of the backscattered

photons �Brillouin 1922�, which is given by

JOURNAL OF PERFORMANCE O

Downloaded 05 Jan 2009 to 131.151.86.242. Redistribution subject to

�B = 2nVa/� �1�

where n�refractive index; Va�velocity of acoustic wave; and��wavelength of incident light. Recently developed techniqueshave enabled to scan the Brillouin scattered light spectrum insingle-mode optical fibers with high resolution by using a coher-ent receiver. The Brillouin frequency shift was found to increaselinearly with strain �Horiguchi et al. 1989� and temperature�Kurashima et al. 1990�, i.e.

�B��� � �B�0��1 + Cs�� �2�

�B�t� � �B�t0��1 + Ct�t − t0�� �3�

where t0�reference temperature and Cs and Ct�proportional co-efficients of strain and temperature, respectively, which are char-acteristic of the optical fiber.

The basic concept is illustrated in Fig. 1. At any point alongthe fiber, the Brillouin spectrum detected is accurately approxi-mated by a Lorentzian function with resonance frequency of �B,and a Gauss function, from which either the strain or temperaturecan be derived by applying the specific �B��� and �B�t� correlationfunctions of the FOS. The measurement position along the FOS iscomputed from the light velocity in the fiber core and the timeelapsed between launching the pulsed light and detecting thebackscattered light. The spatial resolution is determined as

�z =c�

2n�4�

where c�light velocity in a vacuum and ��pulse width of inci-dent light, which typically translates into values up to 2 m, de-pending on the selected pulse width. When operating on a largestructure, comparable resolutions may be attained only by meansof a very high number of discrete sensors, which may not bepractical or cost effective.

Bridge Description

Bridge No. A6358 carries U.S. Route 54 over the Osage Riverin Osage Beach, Mo. The superstructure, built in 2004, has fivecontinuous symmetric spans with no skew, as shown in Fig. 2�a�,with a total bridge length of 263.4 m. The cross section is

Fig. 1. Schematic of Brillouin frequency shift in FOS subjected tomechanical- or temperature-induced deformation

shown in Fig. 2�b� and comprises five identical welded plate I

F CONSTRUCTED FACILITIES © ASCE / JULY/AUGUST 2008 / 265

ASCE license or copyright; see http://pubs.asce.org/copyright

girders equally spaced at 2.6 m on-center, and a 216-mm-thickreinforced concrete �RC� deck with New Jersey continuousconcrete barriers. The out-to-out deck and clear roadway widthsare 12.4 and 11.6 m, respectively. Bolted splices at the contra-flexure areas connect the girder sections designed to resist themaximum positive and negative moments. The girders were fab-ricated using ASTM A709 Grade HPS 345W steel �yield strengthFy �345 MPa� except at the intermediate supports, where the hy-brid sections include top and bottom flanges made of ASTMA709 Grade HPS 485W steel �Fy �485 MPa�. Bridge design iscomposite in the positive moment regions, and non-composite inthe sections resisting maximum negative moments over the inte-rior supports. Fixed and sliding bearings were built on the bents atPiers 2–3 in Fig. 2�a�, and on the other supports, respectively.

Diagnostic Load Test

Load Passes and Procedure

The load test was conducted using six ten-wheel, three-axle dumptrucks prior to opening the bridge to traffic. The vehicles werefully loaded at a gross weight load between 192.8 and 268.1 kN,with approximate distribution of 3/8 in the front axle and 5/8 inthe rear axles, in the configuration illustrated in Fig. 3�a�. Thefour load passes detailed in Figs. 3�b and c� were designed toproduce maximum strains in the exterior girders using two sym-metric lanes of two and three trucks each on Span 1 �Girders 1and 5, Passes A and B�, and a single train of six trucks on Span 2�Girder 5, Pass C� and Span 1 �Girder 1, Pass D�. Focus on theexterior girders was aimed at assessing the response of Girder 1along Span 1, which was heat straightened after falling duringconstruction due to high wind, as documented in Fig. 4. The

Fig. 2. Bridge No. A6358: �a� elevation of steel girder; �b� cross sectLengths in meters, steel plate dimensions in millimeters.

wheel loading locations were marked on the deck to set a distance

266 / JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES © AS

Downloaded 05 Jan 2009 to 131.151.86.242. Redistribution subject to

of 2.74 m between the front axle of each truck and the rear axle ofthe preceding one. A 15 min interval was allowed at each loadpass before performing any measurements.

BOTDR Setup

Tight-buffered FOSs �9 /125 m single-mode optical fibers with900 m diameter and tight PA buffer coating� were used forstrain measurement. Loose-buffered FOSs �9 /125 m silicasingle-mode optical fibers with 900 m diameter, and either loosedry-coupled PVA or wet-coupled PE buffer coating�, which aresensitive to temperature-induced deformations only, were used forthermal compensation and installed parallel to the mechanicalstrain sensors. The sensing circuit included bare optical cables,and a tape made of woven E-glass fiber strands that carried twostrain- and two temperature-sensing FOSs, designated with theletters S and T, respectively, in Fig. 5. The tape was used due tothe ease of handling and to reduce the risk of damaging the opti-cal fibers during installation under nonideal field conditions.

The empirical correlation functions between mechanical strain� and Brillouin frequency shift �B were established by testingeach FOS in the laboratory under predetermined strain levels,which were applied using a specially made fixture. Piecewise-linear �B��� functions used for field measurement were then cali-brated by defining specific sets of correction coefficients, thusoptimizing accuracy and linearity of the strain analyzer readings.Suitable �B�t� correlation functions were defined in a similar fash-ion by testing T fibers in an environmental chamber at increasingtemperatures. The FOSs were adhesively bonded onto the web ofGirders 1, 2, 4, and 5 with a two-part epoxy resin, upon pretreat-ment of the steel surface by manual surface roughening followedby solvent wiping. The girders were instrumented at differentdepths with either bare optical cables or tape running continu-ously through the gaps at the transverse stiffeners, as detailed in

superstructure. ASTM A709 Grade HPS 345W used unless specified.

ion of

Figs. 6�a and b�. The sensors were installed along Span 1 and

CE / JULY/AUGUST 2008

ASCE license or copyright; see http://pubs.asce.org/copyright

Span 2 starting at about 6 m from Abutment 1, and up to the thirdbolted joint at a distance of 87.2 m from Abutment 1. A cartdesigned to move along the entire bridge by rolling over the bot-tom flanges of two adjacent girders was used for the installationwork �Matta et al. 2005�.

The FOSs were series connected with a fusion splicer to forma 1,159 m long circuit. The optical attenuation was contained

Fig. 3. Diagnostic load test: �a� dump truck configuration; �b�longitudinal; and �c� transverse schematic of load test passes withtruck loads. Dimensions in meters and loads in kilo-Newtons.

within 6 dB in the first 1,026 m of the circuit, as shown in Fig. 7,

JOURNAL OF PERFORMANCE O

Downloaded 05 Jan 2009 to 131.151.86.242. Redistribution subject to

thus excluding the end portion with the tape along Girder 4 atLocation I, where a steep power decay from 5.8 to over37.6 dB /km was detected. A BOTDR AQ8603 optical fiberstrain/loss analyzer, manufactured by Yokogawa Electric Corpo-ration, was used to measure the strain along the FOSs from oneend of the sensing circuit. A minimum accuracy of 40 � isspecified by the manufacturer for measurements using FOS cir-cuits with steep geometric discontinuities. The spatial resolutionis 1 m at a pulse width �=10 ns �2 m at �=20 ns� within a 2 dB�6 dB� optical loss, and repeatability �0.04% �0.02%�. The strainanalyzer was connected to a laptop PC via standard ethernet in-terface, and a proprietary software was used to process the strainand temperature data in real-time �Bastianini et al. 2005b�. Thesetup, which is shown in Fig. 6�c�, accounted for a measurement

Fig. 4. Photograph of exterior girder fallen during erection due tohigh wind �Courtesy of Missouri Dept. of Transportation�

Fig. 5. Fiberglass tape with embedded FOSs for strain �S� andtemperature �T� measurement

F CONSTRUCTED FACILITIES © ASCE / JULY/AUGUST 2008 / 267

ASCE license or copyright; see http://pubs.asce.org/copyright

accuracy of 40 � on a length resolution of 2 m, with the strainanalyzer set for 20 ns laser pulses with wavelength of 1.55 m.The accuracy was expected to considerably improve away fromthe geometric discontinuities along the optical circuit �e.g., at thebolted joints, where the FOSs were bonded onto the adjoiningweb plates in the gap between the bottom flange and the websplice plates�, and the near-zero strain areas �such as at the con-traflexure regions�. Selected sections were instrumented withstrain gauges to measure mechanical- and temperature-induceddeformations, respectively.

Prior to load testing, a 24 h temperature monitoring test wasconducted to assess the selected FOS calibration by comparing

Fig. 6. BOTDR setup: �a� schematic of FOS locations; �b� photo-graph of sensors installed along Girder 5 �tape at Location I, and bareFO cables at Location III�; and �c� data acquisition and processingstation

the temperatures computed from the measured strains with that

268 / JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES © AS

Downloaded 05 Jan 2009 to 131.151.86.242. Redistribution subject to

directly measured via type-T thermocouples. The minimum andmaximum temperatures measured on Girder 5, which was themost exposed to sunlight, ranged between 10 and 44°C.

ATS Setup

To provide a reliable benchmark for the validation of the BOTDRresults, girder deflections were measured at discrete points using ahigh-precision ATS system. The total station combines the fea-tures of an electronic distance measurement unit with an elec-tronic theodolite �Wolf 2002� and is extensively used in practice.The instrument sends a laser ray to a set of reflecting prisms�targets� mounted on the structure and to fixed reference prisms,thus enabling the built-in computer to determine the movementsof the target points based on triangulation. The system is basedon the Leica TCA2003 total station, which allows us to automati-cally measure vertical displacements of preselected targets withan accuracy of 0.5 s on angular measurement, and 1 mm+1 ppm on distance measurements in average atmospheric con-ditions, with a working range up to 150 m. A total of 22 prisms,denoted as P1 to P22 in Fig. 8�a�, were mounted onto the bottomflange of the girders along Spans 1 and 2. Four reference targetswere used. Figs. 8�b and c� show photographs of a prism and ofthe test setup, respectively. The total station was secured on acustom-built leveling steel plate mounted on top of a concrete pileto minimize systematic errors. Four readings were taken for eachtarget, two direct and two inverse, to further reduce measurementerrors.

Validation of Distributed Strain Measurement

The strain profiles measured along the girders through the FOSscloser to the bottom flanges �Location I in Fig. 6�a�� were con-verted into vertical deflection profiles to allow validation via di-rect comparison with the benchmark ATS measurements. Thecurvature profiles were computed by dividing the strains, hereindenoted as �BOTDR�x� where x indicates the distance of the refer-ence section from Abutment 1 in Fig. 3�a�, by the vertical distancehS�x� between the measurement point and the neutral axis at thecorrespondent section. Due to the relatively high effective span-to-girder spacing ratio �in excess of 12�, steel-concrete composite

Fig. 7. Optical power attenuation of Brillouin backscattered lightalong sensing circuit �wavelength of incident pulse light=1.55 m�

girders with effective flange width replicate of the girder spacing

CE / JULY/AUGUST 2008

ASCE license or copyright; see http://pubs.asce.org/copyright

of 2.64 m were assumed to determine hS�x� as 1788 and 1658 mmalong Span 1 and Span 2, respectively, under service loads �Ama-dio and Fragiacomo 2002; Ahn et al. 2004�. In order to providecontinuous and yet accurate curvature functions, the profiles wereapproximated through least squares fitting of Nth-order polyno-mial functions as

y��x� =�BOTDR�x�

hS�x�� a0 + a1x + a2x2 + ¯ + aNxN �5�

where N was selected to ensure a coefficient of determination R2

greater than 0.95. The computation of the vertical deflection func-tion y�x� at a given load condition reduces to the solution of theboundary value problem

�y��x� = �i=0

N

aixi

y�0� = y�44.8 m� = 0 for Span 1

y�44.8 m� = y�101.2 m� = 0 for Span 2

�6�

where the ordinary differential equation is the Euler–Bernoullibeam equation for the selected composite girder and load testpass, and the boundary conditions impose zero vertical displace-ment at the supports.

The experimental and approximated curvature functions for

Fig. 8. ATS setup: �a� schematic of location of targets; �b� reflectingprism; and �c� photograph of setup with total station. Dimensions inmeters.

Girder 1 at Passes A, B, and D are plotted in Fig. 9. The derived

JOURNAL OF PERFORMANCE O

Downloaded 05 Jan 2009 to 131.151.86.242. Redistribution subject to

vertical deflection profiles are shown in Fig. 10�a� together withthe discrete ATS measurements. It can be seen that the deflectionscomputed from the measured distributed strains are good agree-ment with those from the ATS system. Improved results are ob-tained as the coefficient of determination increases in Fig. 9,where the regression sum of squares approaches the total sum ofsquares from Passes A to B to D �R2 from 0.95 to 0.99�, meaningthat agreement between BOTDR and ATS improves as the loadlevels are increased and result in smoother strain profiles. This isalso noted in Fig. 10�b� in the case of Girder 5 at Pass C, wherethe curvature was approximated with a function having R2 of 0.99and translated into a deflection profile that matches the ATS mea-surements along Span 2.

On the basis of the comparative results presented, the strainmeasurements were validated and could be used for structuralassessment, which is addressed in the next section, and to discussthe performance of the BOTDR setup. In addition, it is clear fromFig. 10 that the maximum deflections of Girder 1 and Girder 5under a six-truck lane load in Pass D and Pass C, respectively,remained well below the optional limit of 1/800 times the spanlength set forth in the AASHTO LRFD Bridge Design Specifica-tions �2004�.

Structural Assessment

The diagnostic load test was designed primarily to assess thestructural response of Girder 1 along Span 1 under service loads.In particular, the effective implementation of a distributed strainmeasurement technique was intended to describe the deformationin a continuous fashion along the entire structural member, andpossibly overcome the limitations of discrete measurement solu-tions that provide data only at selected sections. Two criteria wereused for assessment. First, the maximum tensile strain must notexceed that associated with the design load distribution factor

Fig. 9. Least squares polynomial fitting of BOTDR curvature profilesalong Girder 1, Span 1 at Location I at Passes A, B, and D

�LDF�, which yields conservative estimates of the ratio between

F CONSTRUCTED FACILITIES © ASCE / JULY/AUGUST 2008 / 269

ASCE license or copyright; see http://pubs.asce.org/copyright

the maximum girder bending moment at a given load condition,and the maximum moment assuming the entire truck load appliedto a single composite girder. A value of S /3.36=0.788 was used,where S�2.64 m girder spacing, according to the semiempiricalspecifications used to design the bridge �MoDOT 2002�. Thesecond criterion required the strain profiles to be in qualitativeagreement with and not to exceed those from three-dimensionalfinite-element analysis of a bridge model, which was constructedto yield rational upper-bound strains by neglecting the stiffeningeffects of secondary structural members, such as cross frames andcontinuous reinforced concrete barriers �Mabsout et al. 1997;Eamon and Nowak 2002; Conner and Huo 2006�.

The finite-element model was developed using a commerciallyavailable software �Strand7 2005� and is illustrated in Fig. 11.Two-node shear-flexible Timoshenko beam elements �BEAM2�and eight-node quadrilateral shell elements �QUAD8� were used

Fig. 10. Deflection profiles derived from BOTDR measurements anddiscrete ATS measurements: �a� Girder 1, Span 1 at Passes A, B andD; �b� Girder 5, Span 2 at Pass C

to model the steel girders and the RC deck, respectively. The deck

270 / JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES © AS

Downloaded 05 Jan 2009 to 131.151.86.242. Redistribution subject to

and girder elements are eccentrically connected by means of rigidlinks in the sections where shear studs were designed, accountingfor an average deck haunch of 50 mm, thereby imposing the com-patibility of composite behavior. This numerical approach waschosen to produce accurate results while minimizing the compu-tational effort �Chung and Sotelino 2006�. Pinned links connectthe girder elements and the eccentric deck elements to model thenon-composite sections within the bolted joints over the interme-diate supports, where shear connectors were not present. The sup-port boundary conditions were approximated by imposing eithersimple support �Abutments 1 and 2, and Piers 1 and 4� or hinge�Piers 2 and 3� constraints at the corresponding beam centroidnodes. Lateral girder displacements were not constrained. Themesh of the deck shell elements is characterized by a maximumin-plane aspect ratio of 2:1, and was designed to accurately applythe transverse and longitudinal wheel loads replicating the testconditions. For convenience, 0.37 m2 square tire contact areas areused. The elastic moduli used for steel, Es, and concrete, Ec, are195 and 25 GPa, respectively. The latter was determined as Ec

=4,733�fc� �MPa�, where fc��design concrete cylinder compres-sive strength of 28 MPa �ACI 2005�. The stiffening effects ofsecondary structural members such as the cross frames and theconcrete barriers were neglected.

The experimental and theoretical strain profiles along Girder 1,Span 1 at Location I �Fig. 6�a�� are shown in Figs. 12�a–c� forPasses A, B, and D, respectively. The strain profiles measuredwith the FOSs in the FRP tape and located at 22 mm from thebottom flange surface are fairly regular and closely reproducethose from the FEA, and no anomalies were thus observed. Goodagreement is also noted with the measurements of strain gaugesmounted on the upper face of the bottom flange at 20.0 m fromAbutment 1, where the maximum strain at Pass D was expected.

The measured distributed strain peaked at Pass D at 413 �,which is 20% smaller than the value of 518 � associated withthe design LDF for the exterior girder �MoDOT 2002�, and overfour times smaller than the theoretical yield strain of the ASTMA709 Grade HPS 345W steel. It was concluded that Girder 1 metboth the assessment criteria defined and did not present any ser-viceability concern.

Performance of BOTDR System

Valid BOTDR measurements from the FOSs embedded in theFRP tapes were performed in the positive moment regions alongthe span directly subjected to truck loads, as shown in Fig. 12 forGirder 1. In particular, more regular profiles were obtained atincreasing strain levels, such as at Pass D in Fig. 12�c�, when a

Fig. 11. Schematic of finite-element model of bridge superstructure

simple second-order least square fitting polynomial function al-

CE / JULY/AUGUST 2008

ASCE license or copyright; see http://pubs.asce.org/copyright

lowed to attain a coefficient of determination of 0.99. This sug-gests that measurements may be affected by some initial fibermisalignment, whose effects are mitigated as the optical fibers arestretched, and may be overcome by pre-straining the FOSs usingspecial fixtures �Komatsu et al. 2002�. Such effect could not beexploited along the spans adjacent to that directly loaded, wherethe bending moment is steadily negative, and where a reducedmeasurement accuracy at the bottom flanges may result from rela-tively small compressive strains and localized distortion of theFOSs, as in the case of Girder 5, Span 2 at Pass A in Fig. 13�a�.

Another plausible source of reduced accuracy is the combina-

Fig. 12. Experimental and theoretical strain profiles at Location Ialong Girder 1, Span 1: �a� Pass A; �b� Pass B; and �c� Pass D

tion of relatively small strains with their transition from tensile to

JOURNAL OF PERFORMANCE O

Downloaded 05 Jan 2009 to 131.151.86.242. Redistribution subject to

compressive in the contraflexure regions. This can be clearly seenin Fig. 13�a� for Girder 5, Span 1 at Location I at Pass A, wherethe strain discontinuity peaks at the theoretical contraflexure sec-tion. Localized strain nonlinearities may also cause the BOTDRmeasurement accuracy to approach the 40 � limit suggestedby the manufacturer of the strain analyzer, such as in the case oflocal strain relief at the cross-frame connection sections, or incase of strain drop at the bolted joints, where the FOSs ran di-rectly across the butt joint between adjoined girder portions. Thelatter effect may be recognized through the steep strain disconti-

Fig. 13. Experimental and FEA strain profiles along Girder 5: �a� atLocation I at Pass A; �b� at Location I at Pass C; and �c� at LocationIII at Pass C

nuity in Girder 5, Span 1 at Location I at Pass C in Fig. 13�b�, and

F CONSTRUCTED FACILITIES © ASCE / JULY/AUGUST 2008 / 271

ASCE license or copyright; see http://pubs.asce.org/copyright

was likely enhanced by the local transition from composite tonon-composite behavior anticipated in the FEA profile, as well asby distortion of the FOSs under compression. Conversely, asmooth strain profile was obtained along Span 2, which was di-rectly loaded with a six-truck lane.

The tensile strain profile measured with the bare FOS cables atLocation III �close to the top flange, as shown illustrated Fig.6�a�� along Girder 5, Span 1 at Pass C is shown in Fig. 13�c�. Thereadings between the bolted joint and Pier 1 follow the pattern ofthose from the FEA, thus corroborating the assumption of non-composite behavior for relatively high service loads. This is alsoindicated by the tendency of the maximum negative strains mea-sured at Location I �close to the bottom flange� over the interme-diate support at Pier 1 to approach or exceed the FEA peaks inboth the exterior girders, as shown in Figs. 12 and 13 for Girders1 and 5, respectively.

The readings from Girder 4 at Location I were confirmed to benot relevant, due to the low accuracy determined by the opticalpower loss that was detected along the terminal portion of theFOS circuit, between 1,026 and 1,159 m. The results from thebare FOS cables along Girder 2 at Location II �at a vertical dis-tance of 711 mm from the upper face of the bottom flange werealso not meaningful, since approximately zero strain was mea-sured. This may be attributed to: low strain levels from the bench-mark strain gauge readings at 23.0 m from Abutment 1, which areone order of magnitude smaller than those from the strain gaugesmounted on the bottom flange of Girder 1 at 20.0 m from Abut-ment 1, as shown in Fig. 14�a�; and presence of several verticalcircuit portions at about 7.3 m intervals, usually partially un-bonded and with relatively sharp bends at the cross-frame sec-tions, where the FOSs were drawn through the gaps between thegusset plates and bottom flanges, as depicted in Fig. 14�b�, whichmay have resulted in undesired perturbations in the signalsdetected.

Conclusions

A pilot application of a BOTDR-based optical fiber sensor systemto measure distributed strains along the steel I girders of a high-way bridge has been presented herein. The structure was sub-jected to a diagnostic load test, where a primary goal was toassess an exterior girder that had been repaired after falling dur-ing erection of the superstructure. The relevance of distributedstrain measurement lies in its unique potential for investigatingthe structural response along the girder length in a quasicontinu-ous fashion, thereby overcoming typical practical and economicallimitations of discrete sensing techniques. A 1.16 km optical cir-cuit was installed onto the girders that included sensing fibers forstrain measurement and thermal compensation. A high-precisiontotal station system for deflection measurement was used to pro-vide a solid benchmark.

The experimental strain profiles were converted into verticaldeflection profiles and were in good agreement with the totalstation measurements, thereby validating the BOTDR systemimplemented despite the non-ideal logistic and operational fieldconditions encountered during installation of the sensing circuit.The BOTDR measurements were then enlisted for structural as-sessment purposes. The tensile strain profiles were similar tothose from three-dimensional finite-element analysis of the bridgesuperstructure, where the finite-element model was designed to

yield rational maximum strain profiles. The maximum tensile

272 / JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES © AS

Downloaded 05 Jan 2009 to 131.151.86.242. Redistribution subject to

strain measured also remained well below the original designlimit, thus prompting no serviceability concerns.

The global girder response was effectively described by theBOTDR measurements. The accuracy largely outperformed the40 � limit suggested for the strain/loss analyzer, which con-templates worst-case scenarios that involve either step-like dis-continuities �such as at the bolted joint sections� or very smallstrain levels that exceed the measurement sensitivity �such as atthe contraflexure zones�, especially under negative strains whenlocal distortion of the optical fibers may occur, and when severalbends are imposed on the sensors to reach the desired locationsduring installation. The measurement quality appeared to improveat increasing strain levels, which may suggest the use of pre-straining fixtures when feasible. The characterization of the opti-cal attenuation in the Brillouin backscattered light spectrumenabled the systematic identification of a portion of the sensingcircuit where the measurement accuracy decayed.

Fig. 14. Strain gauge measurements on Girder 2 at Location II: �a�comparison with Girder 1 at Passes A, B and D; �b� photograph ofFOS arrangement at cross-frame sections

The project demonstrated the practical potential of BOTDR

CE / JULY/AUGUST 2008

ASCE license or copyright; see http://pubs.asce.org/copyright

distributed strain measurement for the structural health monitor-ing and assessment of large-scale structures. Further research isneeded to improve and refine the technology, develop dedicatedand practical solutions for the structural health monitoring of con-structed facilities, and advance their field validation.

Acknowledgments

The financial support from the Missouri Dept. of Transportationand the Univ. of Missouri-Rolla Univ. Transportation Center onAdvanced Materials and NDT Technologies is gratefully ac-knowledged. Special thanks are due to Leica Geosystems, Inc.,Isaberg Rapid, Pirelli Cavi e Sistemi S.p.A., SEAL S.p.A.,Yokogawa Corp. of America, Rolla Technical Institute-MachineTechnology and Precision Manufacturing, and Strongwell Corp.,industry member of the NSF Industry/University Cooperative Re-search Center “Repair of Buildings and Bridges with Compos-ites” �RB2C�, for providing precious technical and logisticassistance, materials, and equipment.

References

AASHTO. �2004�. Load and resistance factor design (LRFD) bridge de-sign specifications, 3rd Ed., Washington, D.C.

Ahn, I. S., Chiewanichakorn, M., Chen, S. S., and Aref, A. J. �2004�.“Effective flange width provisions for composite steel bridges.” Eng.Struct., 26�12�, 1843–1851.

Amadio, C., and Fragiacomo, M. �2002�. “Effective width evaluationfor steel-concrete composite beams.” J. Constr. Steel Res., 58�3�,373–388.

American Concrete Institute �ACI�. �2005�. “Building code requirementsfor structural concrete.” ACI318-05, Farmington Hills, Mich.

Bastianini, F., Cargnelutti, M., Di Tommaso, A., and Toffanin, M. �2003�.“Distributed Brillouin fiber optic strain monitoring applications in ad-vanced composite materials.” Proc. SPIE, 5057, 478–485.

Bastianini, F., Corradi, M., Borri, A., and Di Tommaso, A. �2005a�. “Ret-rofit and monitoring of an historical building using ‘smart’ CFRP withembedded fibre optic Brillouin sensors.” Constr. Build. Mater., 19�7�,525–535.

Bastianini, F., Matta, F., Galati, N., and Nanni, A. �2005b�. “A Brillouinsmart FRP material and a strain data post processing software forstructural health monitoring through laboratory testing and field ap-plication on a highway bridge.” Proc. SPIE, 5765, 600–611.

Bastianini, F., Rizzo, A., Deza, U., Galati, N., and Nanni, A. �2005c�.“Discontinuous Brillouin strain monitoring of small concrete bridges:Comparison between near-to-surface and ‘smart’ FRP fiber installa-tion techniques.” Proc. SPIE, 5765, 612–623.

Brillouin, L. �1922�. “Diffusion de la lumière et des rayons X par uncorps transparent homogéne—Influence de l’agitation thermique.”Ann. Phys. (Paris), 17, 88–122.

Casas, J. R., and Cruz, P. J. S. �2003�. “Fiber optic sensors for bridgemonitoring.” J. Bridge Eng., 8�6�, 362–373.

Chung, W., and Sotelino, E. D. �2006�. “Three-dimensional finite-elementmodeling of composite girder bridges.” Eng. Struct., 28�1�, 63–71.

Conner, S., and Huo, X. S. �2006�. “Influence of parapets and aspect ratioon live-load distribution.” J. Bridge Eng., 11�2�, 188–196.

Eamon, C. D., and Nowak, A. S. �2002�. “Effects of edge-stiffeningelements and diaphragms on bridge resistance and load distribution.”

JOURNAL OF PERFORMANCE O

Downloaded 05 Jan 2009 to 131.151.86.242. Redistribution subject to

J. Bridge Eng., 7�5�, 258–266.Fujihashi, K., Uehara, H., Okutsu, M., and Komatsu, K. �2003�. “De-

velopment of a road disaster monitoring system using optical fibersensing technology.” NTT Tech. Rev., 1�9�, 48–52.

Horiguchi, T., Kurashima, T., and Tateda, M. �1989�. “Tensile strain de-pendence of Brillouin frequency shift in silica optical fibers.” IEEEPhotonics Technol. Lett., 1�5�, 107–108.

Kihara, M., Hiramatsu, K., Shima, M., and Ikeda, S. �2002�. “Distributedoptical fiber strain sensor for detecting river embankment collapse.”IEICE Trans. Electron., E85-C�4�, 952–960.

Komatsu, K., Fujihashi, K., and Okutsu, M. �2002�. “Application ofoptical sensing technology to the civil engineering field with opticalfiber strain measurement device �BOTDR�.” Proc. SPIE, 4920,352–361.

Kurashima, T., Horiguchi, T., and Tateda, M. �1990�. “Thermal effects ofBrillouin gain spectra in single-mode fibers.” IEEE Photonics Tech-nol. Lett., 2�10�, 718–720.

Li, H.-N., Li, D.-S., and Song, G.-B. �2004�. “Recent applications of fiberoptic sensors to health monitoring in civil engineering.” Eng. Struct.,26�11�, 1647–1657.

Mabsout, M. E., Tarhini, K. M., Frederick, G. M., and Kobrosly, M.�1997�. “Influence of sidewalks and railings on wheel load distribu-tion in steel girder bridges.” J. Bridge Eng., 2�3�, 88–96.

Matta, F., Vath, M., Galati, N., and Nanni, A. �2005�. “Modular frameconcept cart for inspection of slab-on-girder bridge superstructure.”Proc., 5th Int. Conf. on Bridge Management, G. A. R. Parke and P.Disney, eds., Thomas Telford, London, 187–194.

Missouri Dept. of Transportation �MoDOT�. �2002�. Bridge designmanual, Jefferson City, Mo.

Murayama, H., Kageyama, K., Naruse, H., Shimada, A., and Uzawa, K.�2003�. “Application of fiber-optic distributed sensors to health moni-toring for full-scale composite structures.” J. Intell. Mater. Syst.Struct., 14�1�, 3–13.

Naruse, H., Uchiyama, Y., Kurashima, T., and Unno, S. �2000�. “Riverlevee change detection using distributed fiber optic strain sensor.”IEICE Trans. Electron., E83-C�3�, 462–467.

Ohno, H., Naruse, H., Kurashima, T., Nobiki, A., Uchiyama, Y., andKusakabe, Y. �2002�. “Application of Brillouin scattering-based dis-tributed optical fiber strain sensor to actual concrete piles.” IEICETrans. Electron., E85-C�4�, 945–951.

Shi, B., Xu, H., Chen, B., Zhang, D., Ding, Y., Cui, H., and Gao, J.�2003�. “Feasibility study on the application of fiber-optic distributedsensors for strain measurement in the Taiwan Strait Tunnel project.”Mar. Geotech., 21�3–4�, 333–343.

Strand7 Pty Ltd. �2005�. “Using Strand7—Introduction to the Strand7finite-element analysis system.” Strand7 Pty Ltd., Sydney, Australia,292.

Thévenaz, L., Niklès, M., Fellay, A., Facchini, M., and Robert, P. �1998�.“Truly distributed strain and temperature sensing using embedded op-tical fibers.” Proc. SPIE, 3330, 301–314.

Uttamchandani, D. �1994�. “Fibre-optic sensors and smart structures:Developments and prospects.” Electron. Commun. Eng. J., 6�5�,237–246.

Wolf, P. R. �2002�. “Surveying and mapping: History, current status, andfuture projections.” J. Surv. Eng., 128�3�, 79–107.

Wu, Z., Xu, B., Hayashi, K., and Machida, A. �2006�. “Distributed opticfiber sensing for a full-scale PC girder strengthened with prestressedPBO sheets.” Eng. Struct., 28�7�, 1049–1059.

Yasue, N., Naruse, H., Masuda, J., Kino, H., Nakamura, T., and Yamaura,T. �2000�. “Concrete pipe strain measurement using optical fiber sen-sor.” IEICE Trans. Electron., E83-C�3�, 468–474.

F CONSTRUCTED FACILITIES © ASCE / JULY/AUGUST 2008 / 273

ASCE license or copyright; see http://pubs.asce.org/copyright