Structural Health Monitoring of a Cable-Stayed Bridge Using Regularly Conducted Diagnostic Load Tests

摘要

正文

Structural Health Monitoring of a Cable-Stayed Bridge Using Regularly Conducted Diagnostic Load Tests

Front.BuiltEnviron.5:41.

Cable-Stayed Bridge Using Regularly Conducted Diagnostic Load Tests HadiT.Al-Khateeb1,HarryW.ShentonIII2,MichaelJ.Chajes2*andChristosAloupis2 1JacobsEngineering,NewYork,NY,UnitedStates,2DepartmentofCivilandEnvironmentalEngineering,Univer sityof Delaware,Newark,DE,UnitedStates The management and maintenance of cable-stayed bridges rep resents a major investment of human and financial capital. One possible appr oach to reducing the cost while simultaneously improving the process is by utilizing structural health monitoring (SHM) systems to enable diagnostic load tests to be regularl y and efficiently conducted.

The Indian River Inlet Bridge (IRIB), a 533-m long cable stay ed bridge, was opened for traffic in 2012. From the very early stages of the design pr ocess, the Center for InnovativeBridge Engineering (CIBrE) attheUniversityof Delaware(UD) worked withthe DelawareDepartmentofTransportation(DelDOT)andtheird esign-buildteamofSkanska and AECOM to plan and install a comprehensive structural hea lth monitoring (SHM) system. The SHM system is a fiber-optic based design with more than 120 sensors of varying type distributed throughout the bridge. The system , which not only collects data continuously during normal operation, has also been utiliz ed during regularly scheduled controlled diagnostic load tests being used to monitor ongo ing bridge performance.

This paper presents results from a unique series of six diagn ostic load tests which have been performed over the first 6 years of the bridge’s serv ice life (just prior to the bridge’s opening, and then again at 6 months, 1, 2, 4, and 6 yea rs). The results of this extendedsetofdiagnosticloadtestshaveenabledthebridg e’sbaselineperformanceto berigorouslyestablished.Thisinturnhasprovidedtheopp ortunitytodevelopaprocess for conducting future biennial tests to and adding their res ults to an evolving database, thereby enhancing DelDOT’s ability to operate and maintain the bridge.

Keywords: diagnostic, load, test, structural, health, moni toring, cable-stayed, bridge INTRODUCTION In order to ensure the structural integrity of a bridge throu ghout its life, it is essential that the structuralcomponentsofthebridgeareroutinelyinspecteda ndevaluated.Inspectionsresultsand ensuing evaluations are used to classify the physical and fun ctional condition of the bridge. The datageneratedfromobservationalinspectionsarequalitati veandrelyontheinspector’sexperience, skill, and primarily focus on components of the bridge that can b e readily seen. Other evaluation methods can be used, along with visual techniques, to improve the load rating process such as non-destructive evaluation technologies of bridge load te sting. In a bridge load test, instruments such as strain gauges tilt meters, deflection devices, or oth er instruments are strategically located andattachedtothebridge.Aload,typicallyaheavilyloaded vehicle,isthenplacedordrivenacross thebridgeandthebridgeresponseismeasured.

Al-Khateeb et al. SHM of a Cable-Stayed Bridge Bridge load tests are often categorized into the categories of (1)prooftests,(2)diagnostictests,and(3)in-servicetes ts.Proof load tests are used in verifying the load carrying capacity of the bridge. A truck, weighing the load the bridge is intended to be able to carry safely, crosses the bridge. If the load cro sses thebridgewithoutdamageandwithinthedesignatedacceptab le stress range, it was deemed as proof the bridge can carry the load.Diagnosticloadtestsareusedtoquantifyabridgesre sponse to heavy loads, and the response is then used to either directl y evaluate the bridge or calibrate a numerical model which is in turn used to evaluate the bridge. In an in-service load tes t, instrumentation is used to measure the response of the bridge duetoambienttrafficoveraspecifiedamountoftime.Statistic al analysesarethenusedtocorrelatethecollecteddatatothe traffic loading. In all of these different types of load tests, the respon se of the bridge to the load is used to determine an acceptable loa d rating for the bridge, and that rating is ultimately compared to theratingcalculatedusingconventionalmethods.

The following publication provide both guidelines for using load tests to evaluate bridges as well as provide numerous applications ( Pinjarkar et al., 1990; Fu and Tang, 1992; Moses et al., 1994; Lichtenstein, 1995; Nowak and Saraf, 1996; Cha jes et al., 1997, 1999, 2000; Fu et al., 1997; NCHRP, 1998; The Institution of Civil Engineers, 1998; AASHTO, 2003; Chajes and Shenton, 2006; Schiff et al., 2006; Jeffrey et al., 2009; Hosteng and Phares, 2013; Olaszek et al., 2014; Peiris and Harik, 201 6;

Al-Khateeb et al., 2018 ). Most recently, Bayraktar et al. (2017) , has employed static and dynamic field testing on a cable- stayedbridge.

Historically, bridges have undergone “one-off” load tests for specific reasons (i.e., low rating, damage, load carrying capacity validation, numerical model validation, assessme nt of repair effectiveness, lack of construction drawings, etc.) an d have been instrumented with temporary sensors for a specific test. As such, there is little documented history of owners conducting a series of controlled load tests to quantify and monitor bridge health. This paper documents the initiation of a long-term monitoring program involving regularly conduct ed loadtestsusedincombinationwithotherlong-termmonitor ing, allperformedutilizingacomprehensiveSHMsystem.

SHM has been a topic of intense research for some time;

an early review of research in this area can be found in Doebling et al. (1996) . The literature review was updated through 2001, in Sohn et al. (2003) .Carden and Fanning (2004)also provided an updated review on vibration based SHM, picking up where Doebling et al. (1996) left off. More recent updates on SHM research include Das et al. (2016) , Mesquita et al. (2016) , andSeo et al. (2016) . Finally, Li and Ou (2016)provide a detailed review of SHM of cable-stayed bridges which is particularly relevant here. Their paper includes a lis t of significant cable-stayed bridges that have installed on t hem SHM systems; many are located in Asia. The paper outlines the many uses for SHM data in the operation and maintenance of a cable-stayed bridge including the use of diagnostic loadtests.

SHM systems have primarily been used for long-term in- service monitoring. In this paper, the focus is on demonstrati nghow SHM systems can make it possible to collect data during regularlyperformeddiagnosticloadtests.

Delaware’s Indian River Inlet Bridge (IRIB), a cable-stayed bridge located in southern Delaware, has a permanent array of instrumentsthatwereinstalledintheformofastructuralh ealth monitoring system. Immediately after the IRIB was opened to traffic, a series of diagnostic load tests were conducted to establish the “healthy condition” or baseline behavior of t he bridge. Additional diagnostic load tests or “physicals” have been conducted every 2 years to create an evolving “health record” for the bridge. To date, six load tests have been performed. This paper presents the methodology employed to build a comprehensive health record of the IRIB from biennial diagnostic load tests conducted utilizing the bri dge’s SHMsystem.

DESCRIPTION OF THE CABLE-STAYED BRIDGE AND THE STRUCTURAL HEALTH MONITORING SYSTEM ThefollowingsectionsprovidebothdetailsoftheIRIBbridgea nd oftheinstalledSHMsystem.Moreextensivedescriptionscanb e foundinShentonetal.(2017a,b) .

Cable-Stayed Bridge Description The Charles W. Cullen Bridge at the Indian River Inlet, also called the Indian River Inlet Bridge (IRIB), is a 1,749 ft (533m ) long cable-stayed bridge with a 948 ft (289m) main span and two 397 ft (121) m back spans. The bridge was designed using a combination of precast and cast-in-place reinforced concre te.

Thebridgeis105ft(32m)inwidthwithtwolanesoftrafficanda shoulderineachdirection.A11ft9¾in(3.6m)widepedestria n walkway is located on the east side of the bridge. This causes t he centerline of the roadway to shift toward the west edge girde r.

Thebridgeisfixedatthenorthpylonbutisfreetoexpandatthe southpylonandtheabutments.

The deck is comprised of two edge girders, transverse floor beams spaced at 11 ft 9 ¾ in (3.6m) on center, and a cast-in- placedeck.Thecontinuouscast-in-placeedgegirdersarerou ghly rectangular in shape with dimensions of 5 ft 7/8 in (1.8m deep) and 4 ft 11 in (1.5m wide). The 8 ½ in (21.6cm) thick cast-in- place deck has 1 5/8 in (4.13cm) of latex modified concrete as a wearing surface. The bridge has two twin pylons that reach a height of 248 ft (75.6m) above the ground. The pylons have a hollowboxcross-sectionthatisuniformbelowthedeckleve land abovedeckleveltaperstothetopofthepylon.Thereisatotalo f 152stays,38perpylon.Nineteenstaysemanatefromeachsideo f the pylons and are anchored to the edge girder on 24 ft (7.3m) centers. The stay cables consist of seven wire strands in bundl es of19–61.Thestrandsarewaxedandencapsulatedinhigh-dens ity polyethylene sheathing. The stays are enclosed in a helical hi gh- density polythene pipe with a raised helical strake to minimize thepotentialforwind-raininducedvibrations.

Construction of the bridge started in 2009 with the driving of the piles for the pylons. The bridge was opened to limited traffic in the winter of 2012 and was completed and opened Al-Khateeb et al. SHM of a Cable-Stayed Bridge FIGURE 1 | Layout of sensors in the SHM system.

to full traffic in May of 2012. Additional details of the bridge design and construction can be found in Delaware Department ofTransportation(2019) andNelson(2011) .

Structural Health Monitoring System TheIRIBbridgewasbuiltwithafiber-opticSHMsysteminstalle d throughout the full length of the bridge to monitor a variety of types of structural response. The system includes seven differe nt types of sensors, with a total of 144 individual sensors insta lled on the bridge ( Figure1). The different sensors are designed to measure the structural response of the bridge under various environmentalloadsandliveloadconditions.Theyinclude : •70strainsensors,locatedintheedgegirders,pylons,andde ck •44 accelerometers, mounted to the deck, pylons, and staycables •9tiltmetersmountedalongtheeastedgegirder •3 displacement sensors, one at each of the bridge expansion bearings(thetwoabutmentsandthesoutheastpylon) •2anemometersthatmeasurewindspeedanddirection,oneat decklevelandoneatthetopofonepylon •16chloridesensorsinthedeckin10locations All of the sensors are optical sensors, with the exception of the anemometers and 10 of the chloride sensors, which are conventionalanalogdevices( Shentonetal.,2017a ).

The SHM system operates 24/7. During operation two basic types of data are collected: “monitor” data and “event” data.

Monitor data is collected continuously at both low and high frequency. For low frequency data, a single average sensor reading, computed from data taken at 125Hz over a 10- min period, is recorded. For high frequency data, the data is continuously recorded at 25Hz. The monitor data is used to quantify the response due to ambient live loads and well as to monitor long-term, gradual variations in bridge behavio r.

These long-term variations might be due to daily or seasonal thermal variations or slow degradation due to environmenta leffects or sustained load. In particular, ongoing data from SHM system is being used to evaluate long-term effects including cable forces (using cable vibrations), bearing condition ( using bearing displacements), bridge ratings (using strain gauges ), bridge deflections (using inclinometers), and thermal respo nse (using multiple types of sensors). These applications have been outlinedin Chajesetal.(2018) .

USING A SERIES OF DIAGNOSTIC TESTS TO MONITOR BRIDGE HEALTH As mentioned, the SHM system is being used to monitor long- term response of the bridge due to ambient traffic and thermal changes/windeffects.However,noneofthisresponsedataisdu e tocontrolledloads.Whileitwouldtypicallybeveryexpensive to instrument and test a long-span bridge using controlled loads, even one time, by leveraging the existence of the permanent SHM system, it has become possible to conduct an ongoing series of controlled and calibrated diagnostic tests on the I RIB.

While the predominant stresses long-span bridges come from dead loads, calibrated live load tests can be effectively used to assess change in bridge response and associated change in condition. In order to accurately and effectively evaluate th e response of the bridge over time using the diagnostic load tests , a set of standard test procedures and the determination of the baseline response of the bridge that represents the “healthy” establishing a standard testing protocol, (2) establishing the baseline loading and baseline response, and (3) establishin g key responseparametersforfuturecomparison.

Establishing Test Protocol The determination of testing protocols takes place before any testing occurs. It involves, among other lesser details , determining the number of trucks to be used to load the bridge, Al-Khateeb et al. SHM of a Cable-Stayed Bridge FIGURE 2 | Key strain sensors on edge girder.

their weight, the configurations (or passes) to be used to load the bridge, the number of passes, the timing of the test, and the required traffic control. Also included is whether the loa d passes will be static (stationary trucks or applied loads), pseu do- static(trucksmovingataslowcrawl),ordynamic(trucksmo ving atfull-speed).

Establishing Baseline Loading and Baseline Response To establish the baseline loading and baseline response, a se ries of diagnostic load tests should be conducted a few months apar t over the first year of service of the bridge. The first test shou ld be conducted as close to the completion of bridge constructio n as possible. The remaining tests should be conducted when tim e dependent effects, such the increase in strength and stiffness of the concrete, concrete creep and shrinkage and associate d pre-stressed losses, and any other ongoing changes that will affect bridge response, are believed to have stabilized. For so me bridges this might take 6 months to a year after the bridge construction is completed. When comparing the response from these preliminary tests, one can look at both the nature of the time-history response due to the slowly moving truck loading or the magnitude of sensor data. While one should not expect perfectcorrelation,ifthecomparisonofresultsfromthesei nitial tests are consistent and within an established variability of the data, the baseline test can be selected. The earliest test dur ing which the bridge response is deemed to have stabilized will be selected as the baseline. While initial baseline tests may u tilize a wide variety of load passes, based on evaluating the results f rom each pass, a baseline set of load passes should be determined.

This baseline set of load passes should be the minimum number of passes that will yield comprehensive response results, and i s called the “baseline loading.” The response resulting from t he baseline loading is called the “baseline response.” The test for whichthesebaselineresultscomeiscalledthe“baselinete st.” Establishing Key Parameters for Future Evaluation While a large number and wide distribution of sensors may be neededtoensurethatacomprehensiverecordofbridgerespons eis captured, to simplify comparisons of future response to the baselineresponse,itisusefultoidentifyasmallernumbero fkey sensorsforuseininitialcomparisons.Ifthecomparisonsindica te changesinresponsehasoccurred,theuseofamoreextensivese t ofsensorscanthenbeemployed.Thekeysensorswilltypicallyb e those located at regions associated with maximum load effects .

The key locations can be defined based on analytical results from the design process (such as locations that govern the loa d rating),aswellasfromtherecordedresponsedatafromthein itial load tests. At the key locations, both time histories and peak response resulting from various load passes canbe used to make the initial evaluation of response compared to the baseline. F or the IRIB bridge, the key sensors (see Figure2) will be the strain gauges located on the west and east edge girder at midpsan (S- W7/8 and S-E7/8), at the controlling location (S-W21/22 and S-E21/22), as well as the strain gauge located in pylon 6 west just above the deck level (S-W24S). The controlling locatio n is the longitudinal location along the edge girder that govern s the bridge load rating. This happens to be at the quarter point of the backspan, and is within a few meters of strain gauges S- W21/22 and S-E21/22 (see Figure2). The governing computed loadratingis1.17( Al-Khateeb,2016 ).

Other parameters that can be used to evaluate structural response are computed parameters such as the summation of girder strains across the bridge cross-section, load distri bution factors, or girder neutral axis location. Which particular response parameters to use will depend on the specific bridge beingevaluated.

DIAGNOSTIC LOAD TESTS CONDUCTED ON THE IRIB The following sections describe the series of six diagnostic l oad teststhatwereconductedoverthefirst6yearsofoperationof the IRIB and used to both establish the baseline response and track theconditionofthebridgeovertime.

Testing Protocol A final testing protocol for how the load tests should be conducted and what passes should be included was determined Al-Khateeb et al. SHM of a Cable-Stayed Bridge TABLE 1 | Baseline loading.

Pass Identifier Description Direction of travel ONE TRUCK 1, 7 1e Southbound shoulder Northbound (NB) 2, 8 1a Southbound slow-lane 3, 9 1b Southbound fast-lane 6, 12 1f Northbound shoulder 5, 11 1d Northbound slow-lane 4, 10 1c Northbound fast-lane FOUR TRUCKS 13, 14 4a Side by side, one in each travel lane NB SIX TRUCKS 15, 16 6a Side by side, one in each lane and shoulder NB after the initial three load tests were completed and evaluate d.

The final protocol included the minimum number of passes that should be conducted during each load test in order to assessbridge’scondition.Duringseveralofthetests,ext rapasses were conducted to examine specific phenomena, however will not be discussed here as they were for independent focused researchstudies.

The final test protocol is comprised of 16 slow crawl passes.

The first 12 passes involve single trucks traveling in one of th e fourtravellanesoroneofthetwoshoulders.Next,twofourt ruck passesinvolvetrucksinaside-by-sideformationtraveling inthe four travel lanes. Finally, two six truck passes involve truc ks in a side-by-sideformationacrossallsixlanes(travelandsho ulders).

Thepassnumber,passidentifier,truckformations,anddirect ion oftravelaregivenin Table1andpassconfigurationsareshownin Figure3.Duringtheloadtests,liveloadsareappliedusingupto sixtesttruckswithamaximumcombinedweightofroughly380 kips (1,690 kN). To minimize thermal effects during the testing period, all tests have been conducted at night, generally star ting no earlier than 10 pm. This also minimizes traffic disruption.

A complete load test report is submitted to DelDOT following eachtest.

Load Tests Conducted SixdiagnosticloadtestshavebeenconductedontheIRIBbrid ge since it was built. The initial test coincided with the openin g of the bridge to full traffic. The next two tests were performed after 6 months of service and after 1 year of service (these th ree were used to establish the baseline). Following the initial three tests, ongoing tests have been conducted to provide response data at 2-year intervals. Thus, far, the 2, 4, and 6-years tes ts have been completed. DelDOT’s plan is to continue conducting these biennial diagnostic tests as they will become increas ingly valuable as the bridge ages. One important question that shou ld be asked is whether the SHM system is robust enough to last years into the future when changes in bridge condition is muc h more likely to be seen. This is a very valid concern. To addres s this concern fiber optic sensors were selected as they are know n for theirexcellent durability. Furthermore, the strainse nsors are FIGURE 3 | Truck configurations for baseline load passes.

embeddedintheconcreteandthisshouldincreasethechance sof theirsurvival.Redundancyofsensorlocationshasbeenbuilt into the system, and DelDOT is allocating ongoing funds to active ly replacedsensorsthathavestoppedworkingordon’thavereliabl e measurements. There is also a plan to duplicate the key sensors byinstallingadditionalsurfacemountedsensors.However, until suchlong-termdemonstrationprojectsplayout,wecannotkno w thiscanwelearnhowtodesignandimplementSHMsystemsthat willhavelongservicelives.

Baseline Diagnostic Load Tests As described, the first three load tests were all performed within the first year of service of the bridge and were used to establish the baseline response. Details of these three test s are summarizednext.

Loadtest1—April30,2012 In the first load test, conducted right before the bridge was f ully openedtotraffic,fourtruckswereused.Theaveragetruckweigh t was 63.5 kips (282 kN). A total of 17 passes were made in this first load test, 15 slow crawl passes and two dynamic passes. The first four passes were single truck passes and were all conducted using the same truck in each of the four travel lanes. Next, six passes were made with two trucks in specified formations.

Finally,fivepassesweremadeinwhichallfourtruckswereplac ed in different formations. The dynamic, or high-speed, tests were conductedwithallfourtruckstravelingat ∼55mph(88.5km/h) withapproximatelya100ft.(30.5m)intervalbetweeneachtru ck.

Loadtest2—November28,2012 Afteranalyzingtheresultsofthefirstloadtest,andreviewi ngthe test procedures, a decision was made to add two more trucks in could be loaded simultaneously, thereby creating the maximum possibleloadingacrossthewidthofthebridge.Theaveragetr uck weightforthistestwas62.4kips(282kN).

Al-Khateeb et al. SHM of a Cable-Stayed Bridge Atotalof25passesweremadeinthesecondloadtest,23slow crawl passes and two dynamic passes. The first six passes were single truck passes in each of the four lanes and two shoulders .

Next, eight two-truck passes were made in different formations and alignments. This was followed by two, three-truck passes , and then five four-truck passes. Finally, six trucks were used to make two passes in a side-by-side formation. The dynamic, or high-speed, tests were conducted with all four trucks travel ing at∼55 mph (88.5km/h) with approximately a 100 ft. (30.5m) intervalbetweeneachtruck.

Loadtest3—May9,2013 In the third load test, conducted 1 year after the bridge was fully opened to traffic, six trucks were used, with an average truckweightof60.2kips(268kN).Duringthistest,ninedist inct pass configurations were used, and each pass configuration was repeated (repeatability of data is an important step in data validation). A total of 18 passes were made in the third load test, 16 slow crawl passes and two dynamic passes. The first 12 passes were single truck passes in each of the four lanes and two shoulders.Next,twofourtruckpassesweremadewiththetruc ks in a side-by-side formation and two six-truck passes were mad e withthetrucksinaside-by-sideformation.Finally,thedy namic, or high-speed tests, were conducted with a single truck travel ing atapproximately55mph(88.5km/h)insouthboundslowlane.

Diagnostic Load Tests Used to Monitor the Bridge’s Health As described earlier, the next three load tests were performe d at 2-yearintervals.Detailsofthesethreetestsaresummariz ednext.

Loadtest4—May7,2014 The fourth load test, conducted 2 years after the bridge was fully opened to traffic, was conducted following the standard test protocol with an average truck weight of 63.4 kips (282 kN). This test included additional passes with specific resear ch objectives. In addition to the standard 16 slow crawl passes described in the protocol, there were also two dynamic, or hig h- speed, tests conducted with a single truck traveling at ∼55 mph (88.5km/h) in southbound slow lane. In addition, a couple of passes were conducted multiple times (six times) to assist in t he quantification of measurement variability. Finally, at one point during the testing, the bridge was closed to traffic for 5min an d ambientmeasurementsweretakentofurtherassistinquanti fying low-levelsensor“noise.” Loadtest5—May18,2016 The fourth load test, conducted 4 years after the bridge was fully opened to traffic, was conducted following the standard test protocol with an average truck weight of 63.1 kips (281 kN). This test included additional passes with specific resear ch objectives. In addition to the standard 16 slow crawl passes describedintheprotocol,thistestincluded10additionalpa sses.

A total of 26 passes were conducted, 20 slow crawl passes and 6 dynamic passes. The first 18 passes were identical to the 16 included in protocol. To better assess the effect of the high-speed passes, dynamic passes were made of a singletruck in the southbound shoulder, in the southbound fast- lane, in the northbound fast-lane, and in the northbound slo w lane. Theses dynamic, or high-speed tests, were conducted with trucks traveling at approximately 55 mph (88.5km/h).

Finally,tosimulatelongtrucksandtheireffect,additional passes were made using a two-truck train in the southbound slow- lane (twice), a three-truck train in the southbound slow-la ne, and a four-truck train in the southbound slow-lane (all of these passes were conducted with the truck trains moving at a crawlspeed).

Loadtest6—June6,2018 Finally, the sixth load test, conducted 6 years after the bri dge was fully opened to traffic, was identical to third load test (the standard 16 slow crawl passes described in the protocol plus two dynamic passes) with an average truck weight of 62.1kips(276kN).

RESULTS Thefollowingsectionscontain(1)areviewthebaselinerespo nse and associated response parameters that were established based on the first three load tests, (2) an evaluation as to how the response during the ensuing three load tests (years 2, 4, and 6) compares to the baseline response, and (3) a qualitative assessment as to how the response of the bridge has varied over time. Ongoing work is being conducted to determine at what level do individual changes i n response, or a changing trend in response represent changes in bridge behavior. This work is aimed at establishing how severe a change in condition must be before the response parameters are “significantly” affected. Having data from this series of six tests has been very useful for the ongoing sensitivityevaluation.

Baseline Response The baseline response was found from the three tests conducte d withinthefirstyearofserviceofthebridge.Fromthesecond test (6 months) on, it was found that the bridge response stabilize d.

As such, the second load test was deemed to be the baseline tes t, andresultsfromthattesthavebeendefinedasthebaselinere sults.

Thefollowingwillserveasasummaryofthoseresults.

Post-processing and Interpreting Data Before looking at individual load test results, it is importan t to notethatthesameprocedurewasusedtopost-processtestresults from each test. For each sensor, the time-history record was fi rst “zeroed” by taking the average of the first 25 data points and subtracting that value from the entire time history. In this way any initial offset in the record was eliminated. Next a moving average was computed using a window of 1.6s (25 data points for data recorded at 15.6Hz). This smoothing was performed to eliminatetheinherentlow-levelnoiseinthesensordata.Fina lly, the maximum and minimum (i.e., peak) values of the record weredetermined.

When interpreting the results, it is important to note that strains, and associated stresses, with a positive value indic ates Al-Khateeb et al. SHM of a Cable-Stayed Bridge TABLE 2 | Baseline response peak strain.

Sensor Single truck Four trucks Six trucks Max.

strain (µε)Min.

strain (µε)Max.

strain (µε)Min.

strain (µε)Max.

strain (µε)Min.

strain (µε) S-W7 – −14 (1e) – −26 (4a) – −41 (6a) S-E7 – −13 (1f) – −21 (4a) – −31 (6a) S-W8 36 (1e) – 91 (4a) – 138 (6a) – S-E8 32 (1f) – 78 (4a) – 119 (6a) – S-W21 – −18 (1e) – −40 (4a) – −59 (6a) S-E21 – −16 (1f) – −31 (4a) – −48 (6a) S-W22 33 (1e) – 91 (4a) – 151 (6a) – S-E22 30 (1f) – 80 (4a) – 131 (6a) – S-W24S 9 (1e) −11 (1e) 25 (4a) −26 (4a) 36 (6a) −42 (6a) tension. That means that maximum positive strains indicate the largest live-load tensile strain recorded and maximum negative strains indicate the largest live-load compressio n strain recorded. Oneshouldfurthernotethathavingalive-loadte nsile strain/stress during the test does not necessarily mean tha t the element is in a state of net tension, as there can be a large initial compression component due to pre-stressing or post- tensioning that keeps the element in net compression. Where liveloadstressisreported,itisobtainedbymultiplyingstr ainby Young’s modulus of 29,000 ksi (200 GPa) for steel and 5,164 ks i (35.6 GPa) for concrete. For the concrete, the Young’s modulu s is based on an average compressive strength of 8,240 psi (56.8 MPa) determined from the tests of cylinders made during the concretepours.

Baseline Peak Values Table2shows the baseline peak strains for the key sensors for single, four, and six truck passes. For girder sensors, od d numbers indicate top sensors, where we will focus on peak negative values and even numbers indicate bottom sensors where we will focus on peak positive values. In the table, the peak positive values for the top sensors, and peak negative values for the bottom sensors are not shown as they are not significant compared to the other peaks. Gauge S-W24S is a pylon sensor and both positive (tension) and negative (compression) peaks are of similar magnitude and are both shown. The pass identifier for each peak value is given inparentheses.

One can see that the west girder experiences larger strains than the east girder. This is because traffic is skewed toward that side of the bridge due to a wide pedestrian sidewalk on the east side of the bridge. This causes the centroid of the traffic lanes to be closer to the west girder. For the single truc k passes, pass identifier (1e) produces the largest strains becau se that pass consists of a truck in the shoulder closest to the westgirder.

In terms of the magnitude of peak strains, the very largest occurredduringsixside-by-sidetruckpasses.Thelargestt ension strainrecordedduringanyoftheloadpasseswas151 µεatgauge S-W22. This gauge is located at the bottom of the western edgegirderbetweenpylon6Wandpier7(veryclosetothecontrolling location for load rating). The strain of 151 µεcorresponds to a live-load tensile stress in steel of 4.38 ksi (30.2 MPa) and a live-load tensile stress in concrete of 780 psi (5.38 MPa). The largestcompressionstrainrecordedduringanyoftheloadpas ses was−59µεat gauge S-W21. This gauge is located at the same location as gauge S-W22 (the controlling location), but is i n the top face of the western edge girder. This strain corresponds t o a live-loadcompressionstressof1.71ksi(11.8MPa)insteelan da live-load compression stress in concrete of 305 psi (2.10 MPa).

The maximum tension and compression strains in the pylon recorded during any of the load passes were 36 µεand−42µε, respectively.Thestrainof36 µεcorrespondstoalive-loadtensile stressinsteelof1.04ksi(7.17MPa)andalive-loadtensilest ressin concreteof184psi(1.27MPa).Thestrainof −42µεcorresponds to a live-load compression stress of 1.21 ksi (8.34 MPa) in stee l and a live-load compression stress in concrete of 215 psi (1.48 MPa).ThesepylonstrainswererecordedbystraingaugeS-W24S (locatedinpylon6westjustabovethedecklevel).

Baseline Time Histories Figure4 shows the baseline time histories for the key strain gauges (S-W7 & SW-8, S-W21 & S-W22, and S-W24S) due to the six side-by-side truck pass. These time histories will be used later for comparison to load tests 4, 5, and 6. Please note that the plots of the time history are strain vs. time and not strain vs. distance. As such, the length of the plots varie s depending on the exact velocity of the truck(s). Also note that the data recording always starts prior to any trucks coming onto the bridge, and so the first portion of the plot essentially measuresambientresponseandcanbeshortenedasappropriate forplottingtheresults.

Baseline Distribution Factors and Summation of Edge Girder Strains Transverse load distribution is an important characteristi c of a bridge and can be a useful quantity to track over time. The distribution factors for the IRIB were computed at midspan and at the controlling location based on one, four, and six loade d lanes.Table3showsthedistributionfactorsthatwerecomputed atmidspanandthecontrollinglocation.

To further improve the quantitative comparison, we can also look at the sum of the peak edge girder strains (top and bottom) at midspan (S-W7 +S-E7, S-W8 +S-E8) and at the controlling location (S-W21 +S-E21, S-W22 +S-E22). By using the sum of the two strains, some of the variability due to differences in transverse truck location can be eliminated . In a way, the sum of the peak strains is closely correlated to the total moment across the section at the two locations. Table4 shows the summation of the girder strains at midspan and the controllinglocation.

Itshouldbenotedthatwearenottrackingdistributionfacto rs for the floor system as those members are not instrumented. It is believed that the very basic distribution between the two e dge girders can be an indicator of change in behavior, perhaps due tochangesincableforceswhicharebeingmonitoredusingcab le vibrationsaspartoftheSHMlong-termmonitoringeffort.

Al-Khateeb et al. SHM of a Cable-Stayed Bridge FIGURE 4 | Baseline strain time histories for six truck pass: (A)midspan; (B) controlling location; (C)pylon.

Sensor Variability In comparing results of “duplicate” passes, it is important to quantifythevariabilityinthesensorreadingsthatcancom efrom (1) variations in truck location for duplicate passes, (2) bri dge vibrations even in relatively low winds, and (3) general sen sor noise related to sensor resolution. To establish this, (1) d ata can be collected for several minutes while no traffic is on the brid ge and the wind is calm, and (2) several replicates of specific truc k passes can be conducted. To accomplish this for the IRIB, traffic was stopped for 5min and no cars or trucks were permitted to cross the bridge while data was recorded at 125Hz (the test wa s conducted at night in calm wind conditions). In addition, si xTABLE 3 | Baseline response live load distribution factors.

Lanes loaded DF at midspan DF at controlling location

1 0.85 0.63

4 2.1 1.9

6 3.2 3.2

TABLE 4 | Baseline response summation of girder strains.

Sensor Summation of strains ( µε) S_W7+S_E7 −72 S_W8+S_E8 257 S_W21+S_E21 −107 S_W22+S_E22 282 replicatesofbothasingletruckpass(Pass1e)andsixside-by-s ide truck passes (Pass 6a) were conducted. By analyzing the resul ts from these two series of tests, the threshold for a meaningfu l differencebetweenmeasuredstrainvaluesfromdifferenttests but from similar passes was found to be ±4µε. This value can be usedwhenevaluatingtheresultsfromsuccessivediagnostic load tests.Additionaldetailsregardingitsdeterminationcan befound inAloupisetal.(2019) .

Bridge Response Compared to Baseline In the following sections, the recorded response of the IRIB bridge during tests conducted at 2 years (load test 4), 4 years (load test 5), and 6 years (load test 6) after the bridge was openedtotrafficareevaluatedbycomparingthemtothebaseline response.Whileitwouldnotbeexpectedthatsignificantchang es in condition and associated response would be noticed this early in the life of the structure, these test results represen t the beginning of the “medical file” for the bridge (as if the bridg e were a person undergoing biennial physicals). In fact, the bridg e remains in excellent “health” as evidenced by the data about to be shown, and also as evidenced by the biennial inspection reports on the bridge. As noted in section Sensor Variability , variability between tests and due to sensor accuracy should l ead toastrainvariabilityof ±4µε.Averyvalidquestioniswhatlevel of change in strain is needed to signal a change in condition?

This is an active area of ongoing research by the research tea m both for strain data due to diagnostic tests as well as all sensor data due to long-term ambient monitoring. Clearly the natur e and location of the change in condition will how much the measured strain will change. By repeating the test every 2 year s, both one-time changes and trending changes can be captured.

It is anticipated that changes that follow a trend will be the b est signalsofbridgeconditionchange.Finally,itisimportant tonote thatforallresultspresented,theresponsehasbeennormaliz edto theloadingmagnitudeofthebaselinetest.

Comparison of Time Histories Figure5 shows a comparison of the time history response of the key strain gauges at midspan (S-W7/8), the controlling Al-Khateeb et al. SHM of a Cable-Stayed Bridge FIGURE 5 | Baseline strain time histories compared to time histories f rom load tests 4, 5, and 6 for six truck pass: (A)west girder midspan; (B)west girder controlling location; (C)pylon.

location (S-W21/22), and in the pylon (S-24S). One can see that qualitatively, the response from test to test is very consistent. The peaks show some variation, but no trend of increasing nor decreasing magnitude is clear. In the next section, the specific values of the peaks will beinvestigated.

Comparison of Peak Values Table5presents a comparison of the peak strains recorded by keysensorsduringloadtests4,5,and6tothebaselinestrai ns.Of the30differencesfromthebaselinethatwerecomputed,24were <10%.Takingtheabsolutevalueofall30differences,theaverag e differenceis6.5%.Furthermore,whilethereisnoapparenttrend instraindata,ifanyslighttrendexists,itisforthepeakst rainsto begettingsmallerovertime.Comparedtothebaseline,onlyo neTABLE 5 | Comparison of peak baseline strain for key sensors to peak st rains for load tests 4, 5, and 6 for six truck passes.

Sensor Baseline TestLoad Test 4 Load Test 5 Load Test 6 Strain (µε)Strain (µε)Difference (%)Strain (µε)Difference (%)Strain (µε)Difference S_W7 −41−39 3.8 −33 19.2 −36 13.2 S_E7 −31−30 3.5 −28 9.0 −29 7.2 S_W8 138 149 −8.1 135 2.4 143 −3.7 S_E8 119 118 1.4 111 6.8 97 19.0 S_W21 −59−60−2.6 −53 9.6 −58 1.4 S_E21 −48−45 6.1 −42 12.4 −43 11.0 S_W22 151 153 −1.6 139 8.3 151 0.1 S_E22 131 125 4.9 116 11.7 119 9.8 S_W24S 36 36 −0.3 33 6.1 33 7.7 S_W24S −42−43−2.2 −41 0.4 −41 1.7 FIGURE 6 | Peak strains at midspan from load tests 4, 5, and 6 compared to the baseline value: (A)top of west edge girder; (B)bottom of west edge girder.

ofthetenpeakstrainvaluesduringloadtest6waslargerthan the baseline value. In Figures6 ,7, one can graphically see how the peak strains have varied over time in comparison to the baseline value(thehorizontallinesintheplots).Thisvisualrepresen tation of the peak values shows that the ongoing response is quite similartothebaselineresponse,andwithnodiscernablepatt ern ofchange.

Al-Khateeb et al. SHM of a Cable-Stayed Bridge FIGURE 7 | Peak strains at the controlling location from load tests 4, 5 , and 6 compared to the baseline value: (A)top of west edge girder; (B)bottom of west edge girder.

Comparison of Distribution Factors and Summation of Edge Girder Strains As mentioned earlier, both transverse load distribution fa ctors and summation of top and bottom edge girder strains (east and west)canbeusefulparameterstotrackovertime.

Table6provides a comparison of single, four, and six lane loadeddistributionfactorsfromloadtests4,5,and6ascom pared to the baseline values. Of the 18 differences that were computed (absolute values), 10 were below five percent and all were belo w 10%. Using the absolute value of all 18 differences, the average difference is 4.5%. Table7shows the summation of top and bottom edge girder strain gauges at midspan for load tests 4, 5 , and6ascomparedtothebaselinevalue.Ofthe12differencestha t were computed (absolute values), nine were below ten percent and all were below 15 percent. Using the absolute value of all 1 2 differences, the average difference is 6.4%. If any minor trend is noted, it is that the summation of strains is getting smaller over time.Figure8 shows graphically how the summation of strains havevariedovertime.

Summary of Comparisons The comparisons of time histories, peak values, and distribut ion factors all indicate that the bridge condition has remained unchanged during the first 6 years of service. While this is wh at would be expected, the database of response will be extremely valuable in the years to come. This data does contain variabi lity,TABLE 6 | Baseline distribution factors at midspan and controlling l ocation compared to distribution factors from load tests 4, 5, and 6.

Lanes loadedBaseline testLoad test 4 Load test 5 Load test 6 DF DF Difference (%)DF Difference (%)DF Difference DISTRIBUTION FACTORS AT MIDSPAN (BOTTOM SENSORS)

1 0.84 0.83 1.06 0.82 2.74 0.79 6.72

4 2.1 2.2 −6.09 2.1 1.97 2.2 −4.48

6 3.2 3.5 −8.10 3.1 2.39 3.3 −3.70

DISTRIBUTION FACTORS AT CONTROL LOCATION (BOTTOM SENSORS)

1 0.70 0.67 4.00 0.65 6.53 0.64 8.39

4 1.9 2.0 −5.20 1.9 0.45 2.1 −8.33

6 3.2 3.3 −1.60 2.9 8.31 3.2 0.10

TABLE 7 | Comparison of baseline summation of peak top and bottom gird er strains at midspan and the controlling location to summatio n of peak top and bottom girder strains from load tests 4, 5, and 6 for six truck passes.

Sensors Baseline TestLoad Test 4 Load Test 5 Load Test 6 Strain (µε)Strain (µε)Difference (%)Strain (µε)Difference (%)Strain (µε)Difference S_W7+ S_E7−72−70 3.60 −61 14.8 −64 10.6 S_W8+ S_E8257 267 −3.89 246 4.33 240 6.70 S_W21+ S_E21−107−106 0.935 −95 10.8 −101 5.61 S_W22+ S_E22282 278 1.42 254 9.76 269 4.49 butabsenttrendsinthedata,suggestthatfuturevariabili tywithin the ranges seen here should be of no concern. On the other hand, trends in the data, or variability beyond what has been documented,wouldbecauseforfurtherinvestigation.

CONCLUSIONS AND FUTURE WORK This paper has described how SHM systems can be utilized to facilitate ongoing bridge health monitoring using regular ly performed diagnostic load tests. The process involves first establishing a baseline response and then comparing future response to that baseline. In essence, the series of tests is analogoustoaseriesof“physicalexams”andtogethertheycrea te a“healthrecord”forthebridge.Thebaselineresponserepres ents the “healthy” condition of the structure, and each successi ve test adds valuable information to record with which the change in conditionandassociatedhealthofthestructurecanbeasse ssed.

In the case of the IRIB, three diagnostic load tests were conducted to establish the baseline response of the bridge, a nd threeadditionaldiagnosticloadtests(physicalexams)hav ebeen conducted at 2-year intervals to create a health record for t he Al-Khateeb et al. SHM of a Cable-Stayed Bridge FIGURE 8 | Summation of west and east peak strains: (A)top of edge girders at midspan and controlling location; (B)bottom of edge girders at midspan and controlling location.

bridge. The results indicate that the bridge is performing as expected. While some variability in response is observed, no defined pattern or trend in the response over time is evident.

Future tests will be added to the health record (which also includedvisualinspectionresult),therebyenablingtheow nerto developamorequantitativemeasureofthebridge’sconditi on.

Should some event occur or condition arise in the future that raises concern about the health of the bridge, a load test can be quicklyandeasilyconducted,andtheresultsusedinconjun ction with visual inspection and theoretical analyses to fully asse ss the condition of the bridge. This becomes just one more tool in th e engineer’stoolboxforevaluatingthebridgeinsuchaninst ance.

The work presented shows how a bridge SHM system can provide value to a bridge owner. In addition to (ideally) providingautomaticearlycluestopotentialproblems,aperiod iccontrolled load test can provide confirmation that condition s havenotchanged.

Future work will focus on developing a more detailed characterization of test-related variability of the respon se parameters and on determining when changes in response indicates actual structural change and is not simply due to expected test-related variability. The preliminary result of that effortarepresentedin Aloupisetal.(2019) .

AUTHOR CONTRIBUTIONS HA-K, HS, and MC contributed to the conception and design of the study. HS was in charge of developing the SHM system, and oversaw all of the load tests. HA-K and CA performed the data analysis. HA-K, CA, and MC wrote portions of the first draft of the manuscript. All authors contributed to manuscript revisi on, andreadandapprovedthesubmittedversion.

FUNDING The project was supported by funds from both the Delaware Department of Transportation and the Federal Highway Administration under grants BRDG422145–09001448, BRDG422158–TASK 30A −1717, BRDG422161–TASK 30B−1717,andBRDG422162–TASK30C −1717.

ACKNOWLEDGMENTS The authors would also like to acknowledge a number of individuals, agencies, and firms for their support and role in developing and implementing the structural health monitorin g systemfortheIndianRiverInletBridgeandfortheirassistan cein conducting the controlled load test. These include the Delaw are Department of Transportation for the financial support to develop and implement the structural monitoring system, and for help during the load tests (Doug Robb, Craig Stevens, Marx Possible, David Gray, Alastair Probert, Jason Arndt, Cr aig Kursinski, Raymond Eskaros, and the crew from the southern district); the Federal Highway Administration for the finan cial support to develop and implement the structural monitoring system; Cleveland Electric Labs/Chandler Monitoring Systems (Jim Zammataro, Keith Chandler, Jennifer Chandler, and Abee Zeleke); and University of Delaware students for their help during the load tests and other activities (Pablo Marquez, N akul Ramana,JackCardinal,andPatrickCarson).

REFERENCES AASHTO(2003). ManualforConditionEvaluationandLoadandResistanceFact or Rating (LRFR) of Highway Bridges . Washington, DC: American Association of StateHighwayandTransportationOfficials.

Al-Khateeb,H.T.(2016). BridgeEvaluationUtilizingStructuralHealthMonitoring Data.Ph.D.Dissertation,UniversityofDelaware,Newark,DE.

Al-Khateeb, H. T., Shenton, III., H. W., Chajes, M. J., and Wencz el, G. (2018). Computing continuous load rating factors for bridges u sing structural health monitoring data. J. Civil Struct. Health Monitor 8, 721–735.

in diagnostic load test data,” in Proceedings of the Bridge Engineering Institute Conference (Honolulu,HI).

Bayraktar, A., Turker, T., Tadla, J., Kursun, A., and Erdis, A. (2 017). Static and dynamicfieldloadtestingofthelongspannissibicable-stayedbrid ge.SoilDyn.

Carden, E. P., and Fanning, P. (2004). Vibration based conditi on monitoring: a Chajes, M. J., Mertz, D. R., and Commander, B. (1997). Experimental load rating of a posted girder–and–slab bridge. J. Bridge Eng. 2, 1–10.

Al-Khateeb et al. SHM of a Cable-Stayed Bridge Chajes, M. J., and Shenton, H. W. (2006). Using diagnostic load tests for accurate load rating of typical bridges. J. Bridge Struct. 2, 13–23.

Chajes, M. J., Shenton, H. W., Al-Khateeb, H. T., Wenczel, G., Nat alicchio, C., Chen, J., et al. (2018). “ Structural Health Monitoring of Delaware’s Indian River Inlet Bridge ,”inProceedings of the International Bridge Conference (Washington,DC).

Chajes, M. J., Shenton, H. W. III, and O’Shea, D. (2000). Bridge condition assessment and load rating using nondestructive evaluation metho ds.J.

Chajes, M. J., Shenton, III., H. W., and O’Shea, D. (1999). “Us e of field testing in delaware’sbridgemanagementprogram,”in Proceedingofthe8thInternational BridgeManagementConference,TRB,NationalResearchCounc il(Denver,CO).

Das, S., Saha, P., and Patro, S. K., (2016). Vibration-based da mage detection techniques used for health monitoring of structures: a review. J. Civil Struct.

Delaware Department of Transportation (2019). Indian River Inlet Bridge .

river_bridge/(accessedJanuary1,2019).

Doebling, S. W., Farrar, C. R., Prime, M. B., and Shevitz, D. W. (199 6).Damage IdentificationandHealthMonitoringofStructuralandMecha nicalSystemsfrom Changes in Their Vibration Characteristics: A Literature R eview. Los Alamos NationalLaboratoryReport.LA-13070-MS,LosAlamos,NM.

Fu, G., Pezze, F., and Alampalli, S. (1997). Diagnostic load testin g for bridge load Fu, G., and Tang, J. (1992). Proof Load Formula for Highway Bridge Rating .

TransportResearchRecord,1371,Washington,DC.

Hosteng, T., and Phares, B. (2013). “Demonstration of load rating c apabilities through physical load testing: ida county bridge case study,” in Part of InTrans Project12-444,BridgeEngineeringCenter .(Ames,IA:IowaStateUniversity).

Jeffrey, A., Breña, S. F., and Civjay, S. (2009). Evaluation of Bridge Performance and Rating through Nondestructive Load Testing. Amherst, MA: University ofMassachusetts.

Li, H., and Ou, J. P. (2016). The state of the art in structural healt h monitoring of cable-stayed bridges. J. Civil Struct. Health Monitor. 6, 43–67.

Lichtenstein, A. G. (1995). Bridge Rating Through Nondestructive Load Testing .

FinalReport,NCHRPProject12-28A,Washington,DC.

Mesquita,E.,Antunes,P.,Coelho,F.,Andre,P.,Andre,A.,and Varum,H.,(2016).

Globaloverviewonadvancesinstructuralhealthmonitoringplatforms .J.Civil Moses, F., Lebet, J. P., and Bez, R. (1994). Applications of field testing to bridge evaluation. J. Struct .Eng. ASCE 120, 1745–1762.

NCHRP (1998). Manual for Bridge Rating Through Load Testing. NCHRP Project l2-28A,TransportationResearchBoard,Washington,DC.

Nelson, E. T. (2011) Indian River Inlet Bridge – Surviving the Storms. Chicago, IL: AspiretheConcreteBridgeMagazine,Precast/PrestressedInstitu te.Nowak, A. S., and Saraf, V. K. (1996). Load Testing of Bridges, Research Report UMCEE96-10. AnnArbor,MI:UniversityofMichigan.

Olaszek, P., Lagoda, M., and Casas, J. R, (2014). Diagnostic lo ad testing and assessment of existing bridges: examples of application. Struct.

Infrastruct. Eng. Mainten. Manag. Life-Cycle Design Perform. 10, 834–842.

Peiris, A., and Harik, I. (2016). “Load testing of bridges for loa d rating,” in 7th InternationalConferenceonSustainableBuiltEnvironment (Kandy).

Pinjarkar, S. G., Guedelhoefer, O. C., Smith, B. J., and Kritzler, R. W. (1990).

Nondestructive Load Testing for Evaluation and Rating . NCHRP Project 12-28 FinalReport.

Schiff,S.D.,Piccirilli,J.J.,Iser,C.M.,andAnderson,K.J.( 2006)LoadTestingfor AssessmentandRatingofHighwayBridges .ResearchProjectNo.655,Clemson University,Clemson,SC.

Seo, J., Hu, J. W., and Lee, J. (2016). Summary review of structural health monitoring applications for highway bridges. J. Perform 0000824 Shenton, H. W. III, Al-Khateeb, H. T., Chajes, M. J., and Wenczel, G.

(2017a) Indian river inlet bridge (Part A): description of the bridg e and the 170111 Shenton, H. W. III, Al-Khateeb, H. T., Chajes, M. J., Wenczel, G., Arndt, J., and Stevens, C. (2017b) Indian river inlet bridge (Part B): lessons lea rned from the design, installation and operation of the structural health monitori ng system.

Sohn,H.,Farrar,C.R.,Hemez,F.M.,Shunk,D.D.,Stimemates,D.W., andNadler, B. R. (2003). A Review of Structural Health Monitoring Literature: 1996–2001.

LosAlamosNationalLaboratoryReport,LA-13976-MS,LosAlamos,NM.

The Institution of Civil Engineers (1998). Guidelines for the Supplementary Load Testing of Bridges, The Institution of Civil Engineers , National Steering Committee for the Load Testing of Bridges. London: Thomas TelfordPublications.

Conflict of Interest Statement: HA-K was employed by the company Jacobs Engineering The remaining authors declare that the research was conducted in t he absence of any commercial or financial relationships that could be construed as a potential conflictofinterest.

articledistributedunderthetermsoftheCreativeCommons AttributionLicense(CC BY). The use, distribution or reproduction in other forums is permitted, provided publication in this journal is cited, in accordance with acc epted academic practice.

No use, distribution or reproduction is permitted which doe s not comply with these terms.