ACI Structural Journal/November-December 2005 823

ACI Structural Journal, V. 102, No. 6, November-December 2005.MS No. 04-151 received September 7, 2004, and reviewed under Institute publication

policies. Copyright © 2005, American Concrete Institute. All rights reserved, including themaking of copies unless permission is obtained from the copyright proprietors. Pertinentdiscussion including author’s closure, if any, will be published in the September-October 2006 ACI Structural Journal if the discussion is received by May 1, 2006.

ACI STRUCTURAL JOURNAL TECHNICAL PAPER

This paper describes a laboratory study of the effects of trafficloads applied during and after strengthening on the performanceof a reinforced concrete (RC) bridge strengthened with externallybonded fiber-reinforced polymer (FRP) reinforcement. Eightbeams were tested to failure. Seven of the beams were strengthenedwith a precured, unidirectional, carbon-fiber laminate strip repre-sentative of the material used on the actual bridge. Test variablesincluded the intensity and frequency of load cycles applied duringthe epoxy-curing period, the thickness of the epoxy layer, and thethickness of the FRP strip. Failure of all strengthened specimensinitiated with FRP debonding in the maximum moment region. Forall of the traffic load regimens applied during and after installation,no reduction in the effectiveness of the strengthening was observed.Use of ACI 440.2R-02 design recommendations resulted inunconservative predictions of the FRP strain at failure and thestrength increase attributable to the FRP.

Keywords: anchorage; bridge; debonding; epoxy; fiber-reinforced concrete;flexural strength.

INTRODUCTIONAs a demonstration project, the Alabama Department of

Transportation (ALDOT), along with Auburn Universityresearchers, strengthened the War Memorial Bridge inMacon County, Alabama, with externally bonded fiber-reinforced polymer (FRP) reinforcement in November 2001.Swenson and Barnes1 describe the reinforced concretebridge, which was constructed in 1945, and the design of theFRP strengthening system. The bridge is shown in Fig. 1.

Prior to installation of the FRP, the manufacturer recom-mended that the bridge be closed for several hours duringand after strengthening to allow the epoxy to cure. Thebridge was not subjected to a high volume of traffic;however, potential detour routes were lengthy. Thus,ALDOT personnel decided that closing the bridge for anylength of time would not be feasible. Instead, efforts weremade to limit the effects of traffic loads during strengtheningoperations. This was achieved by diverting traffic away fromthe lane directly above the FRP installation each day. Inaddition, trucks were not allowed on the bridge on the daysthe interior girders were strengthened.

From ALDOT’s point-of-view, strengthening of bridgeswith externally bonded FRP is an attractive method if it canbe reliably used to strengthen bridges without an interruptionof service. A literature review yielded little information onthe effects that traffic loads during the epoxy-curing periodhave on the effectiveness of externally bonded FRP rein-forcement. Because the FRP anchorage most likely controlsthe ultimate strength of the strengthened War MemorialBridge, it was critical that the anchorage capacity of the FRPbe examined under actual installation conditions. For thisreason, a laboratory study was initiated to investigate theactual ultimate strength of the FRP-strengthened girders and

whether the capacity is significantly affected by variousregimens of traffic loading that might be experienced duringstrengthening operations. Complete details of the study arereported by Reed.2

RESEARCH SIGNIFICANCEMuch research has been conducted on the topic of FRP

strengthening of structural concrete members. The attrac-tiveness of FRP strengthening is greatly enhanced if bridgescan remain open to traffic during installation. Very fewinvestigations, however, have been performed on the effec-tiveness of externally bonded FRP when the strengthenedstructure is subjected to significant transient loads duringepoxy curing. This paper describes an investigation of thistopic. In addition, the results indicate that ACI 440.2R-023

design recommendations for computing the limiting FRPstrain for flexural strengthening with externally bonded FRPcan be unconservative for the materials used in this study.

BACKGROUNDPrevious research concerning effects of transient loads during strengthening

MacDonald4 describes a study of single-lap shear steelspecimens that were subjected to cyclic movements whilethe epoxy was curing. Each specimen was subjected tocycles, intended to be representative of highway trafficconditions, with a frequency of 1 Hz while the epoxy wascuring and for several days afterwards. At least 500,000cycles with a strain range of 50 microstrain were applied by

Title no. 102-S83

Fiber-Reinforced Polymer Strengthening of Concrete Bridges that Remain Open to Trafficby Michael W. Reed, Robert W. Barnes, Anton K. Schindler, and Hwan-Woo Lee

Fig. 1—War memorial bridge, AL 81 at Uphapee Creek,Macon County, Ala.

ACI Structural Journal/November-December 2005824

full cure. Two different types of epoxy were used. For Type Aepoxy, a reduction in strength between 7 and 31% with anaverage value of 16% resulted. There was no reduction instrength with Type B epoxy. Type A was described as “verymuch” stiffer than Type B.

Barnes and Mays5 tested steel and carbon-fiber-reinforcedpolymer (CFRP) lap-joint specimens. Cyclic loading duringcuring caused a progressive reduction in strength withincreasing strain levels. An 8% strength reduction resultedfor steel joints subjected to 50 microstrain cycles. This valuelies between the average reduction values for the two epoxytypes studied by MacDonald.4 Another test series in theBarnes and Mays5 study included the bonding of a thin plateto an I-section steel beam. Load cycles at 2 Hz and varyinglevels of strain were applied to four-point bending specimensfor 48 hours. The specimens were then tested to failure. Theresults indicated that although the strength of the platedbeams decreased with increasing amplitude of cyclic strainduring curing, there was an unexpected strength increase ascompared with the noncycled specimens. The researcherstheorized that this strength increase was due to the heatgenerated by the cyclic loading apparatus, which acceleratedthe curing of the adhesive.

A third investigation by Barnes and Mays5 involvedlarger-scale concrete beams strengthened with bonded CFRPor steel plates. Load cycles at a frequency of 1 Hz and astrain range of 150 microstrain were imposed while theepoxy cured. The cycles were imposed for 48 hours, and thebeams were tested to failure 7 days after strengthening. Eachcyclically loaded beam failed at the same load as its respectivecontrol beam. Thus, the ultimate load capacity of thesestrengthened beams was not affected by vibration appliedduring curing of the epoxy.

ACI 440.2R design procedure to prevent debonding failure

According to the “Guide for the Design and Constructionof Externally Bonded FRP Systems for StrengtheningConcrete Structures (ACI 440.2R-02):”3 “...there are manydifferent varieties of debonding failure that can govern thestrength of an FRP-strengthened member...” and that “...moreaccurate methods of predicting debonding are still needed.”

Therefore, the ACI 440.2R design procedures limit the FRPstrain levels “to conservatively account for debondingfailure modes.” The limiting effective FRP strain εfe used tocompute the nominal moment capacity of a strengthenedmember is defined in Section 9.2.1 of ACI 440.2R as beinga function of only the axial stiffness of the FRP laminate perunit of bonded width. The expression prescribed to computethe limiting strain “is only based on a general recognizedtrend and on the experience of engineers practicing thedesign of bonded FRP systems.”

EXPERIMENTAL PROGRAMTest specimens and instrumentation

Each of the eight specimens of the present study representedan approximate one-half scale model of the most design-critical girder of the War Memorial Bridge. The cross-sectional dimensions and reinforcement details of the T-beamspecimens are shown in Fig. 2. To correct for self-weightscaling effects and therefore obtain dead-load stressconditions compatible with those in the actual bridge, thebeam web was widened, and a superimposed dead load of28 lb/ft (0.41 kN/m) was applied. To match the behavior ofthe design-critical bridge girder, steel reinforcement wasproportioned to match the mechanical reinforcement ratio ofthe girder. Likewise, the amount of FRP was selected tomatch the FRP reinforcement ratio of the bridge girder. FRPtensile reinforcement consisted of a precured, unidirectional,carbon-fiber laminate strip with a width of 2 in. (51 mm) anda bonded length of 108 in. (2743 mm). The thickness of theFRP strip was a variable in the test program. The bondedlength of the FRP was chosen to correspond to the locationof the curtailment points on the actual bridge girder. Thesecurtailment points were designed according to the anchoragemodel outlined by El-Mihilmy and Tedesco;6 application ofthe procedure is detailed by Swenson and Barnes.1

Each specimen had a span length of 15.5 ft (4.72 m) andwas tested in four-point bending. The test setup is shown inFig. 3. Material properties are listed in Table 1. Concrete andsteel reinforcement material properties were measured usingstandard ASTM methods. The FRP modulus of elasticity Ef

ACI member Michael W. Reed is a structural engineer at K&S Engineering, PLLC,Clarksville, Tenn. He received his BS from the University of Tennessee at Martin,Martin, Tenn., and his MS from Auburn University, Ala. His research interestsinclude innovative semi-active frame earthquake systems and fiber-reinforcedpolymer strengthening of reinforced concrete bridges.

ACI member Robert W. Barnes is an associate professor, Department of CivilEngineering, Auburn University. He is a member of ACI Committee 408, Bond andDevelopment of Reinforcement; Joint ACI-ASCE Committee 423, Prestressed Concrete,and is Secretary of Joint ACI-ASCE Committee 445, Shear and Torsion. He receivedhis BCE from the Georgia Institute of Technology, and his MSE and PhD from theUniversity of Texas at Austin, Austin, Tex.

ACI member Anton K. Schindler is Gottlieb Assistant Professor, Department ofCivil Engineering, Auburn University. He received his MSE and PhD from theUniversity of Texas at Austin. He is a member of ACI Committee 231, Propertiesof Concrete at Early Ages. His research interests include nondestructive testing,hydration modeling of cementitious materials, concrete properties, early-agebehavior of concrete structures, and concrete pavement performance modeling.

Hwan-Woo Lee is an associate professor, Department of Civil Engineering, PukyongNational University, Namgu, Busan, Republic of Korea. He received his BS from SeoulNational University, Seoul, South Korea, and his MS and PhD from the KoreaAdvanced Institute of Science and Technology, Daejeon, Republic of Korea. Hisresearch interests include behavior and design of prestressed concrete bridges,evaluation and rehabilitation of existing structures, and reliability analysis andoptimal design of structural systems.

Fig. 2—Cross section of laboratory test specimen (1 in. =25.4 mm).

Table 1—Material properties

Concrete

No. 3reinforcing

bar

No. 5reinforcing

barFiber-reinforced

polymer strip Epoxy

f ′c , psi

Ec, ksi

ft , psi

fr, psi

fy,ksi

Es,ksi

fy,ksi

Es, ksi

fu, ksi

Ef ,ksi

εfu , %

Ea,ksi

7100 6470 560 920 66 29,000 49 29,700 405 20,000 1.8 245

Note: 1 psi = 6.895 kPa, 1 ksi = 6.895 MPa.

ACI Structural Journal/November-December 2005 825

was measured by the researchers; the ultimate strength of theFRP and the epoxy modulus of elasticity were taken asreported by the manufacturer. All eight specimens were castfrom the same batch of ready mixed concrete.

Beam instrumentation consisted of electrical resistancegauges for measuring steel reinforcement tension strains,FRP strains, and concrete surface compressive strains.Figure 4 indicates the locations of the strain gauges anddisplacement potentiometers. The two inner bars of thetension steel reinforcement were gauged at each instrumentedcross section. These gauges were located at 18-in. (460-mm)intervals at cross sections where flexural cracking wasexpected (stirrup locations).

Surface-mounted strain gauges were used to measureconcrete compressive strains. These gauges were placedalong the longitudinal axis on the top surface of the flange.Each gauge was placed 12.75 in. (324 mm) on either side ofmidspan (within the maximum moment region). These werethe closest locations to the load points that would not resultin damage to the gauges. As indicated in Fig. 4, surface-mounted strain gauges were used at 14 locations to measureFRP strains.

Specimen deformation was applied by means of ahydraulic actuator as a part of a closed-loop system. A built-in load cell was used to measure the applied force. Linearpotentiometers were employed for measuring deflections.The deflections measured at the supports were averaged andsubtracted from the midspan deflection to compensate forpossible support deformations.

A high-speed data acquisition system and computer wereused to process and store the test data collected from thestrain gauges and potentiometers.

Test procedureThe testing procedure was the same for all the beams with

a few exceptions. One beam was not strengthened, and twoFRP-strengthened beams were not exposed to traffic loadcycles during epoxy curing. Test variables included intensityof cyclic loads during strengthening, thickness of the epoxylayer tb, and FRP thickness tf. The values of each variable foreach specimen are given in Table 2.

So that the concrete strength would remain approximatelyconstant during testing of all eight specimens, testing wasdelayed until 110 days after specimens were cast. To simulateactual in-service bridge conditions, the following steps wereperformed:

1. Loading the specimen to induce flexural cracking;2. Applying pre-strengthening service load cycles;3. Preparing the concrete substrate for strengthening;

4. Initiating traffic load cycles for FRP installation;5. Applying primer epoxy to the concrete surface;6. Applying tack-coat epoxy and bonding the FRP strip to

the member;7. Allowing epoxy to cure for 48 hours while undergoing

traffic load cycles;8. Stopping load cycles and bonding strain gauges to FRP;9. Applying post-strengthening service load cycles; and10. Loading member to failure.A detailed description of the procedure follows.Cracking of specimens (Step 1)—To distribute flexural

cracking throughout each specimen to an extent comparableto that of the War Memorial Bridge, a 14 kip (62 kN) loadwas applied monotonically and then removed. This loadcorresponds to approximately 70% of the capacity of the

Fig. 3—Four-point bending test setup.

Fig. 4—Beam geometry, reinforcement, and instrumentation(1 in. = 25.4 mm).

Table 2—Test program and summary of results

Specimen tf , in. Cycle intensity tb, in. Mn,exp, kip-in. ∆fail, in. My,exp, kip-in. ∆y,exp, in. εfe,exp, × 10–6 Memberductility index

Unstrengthened None None NA 829 >8.82 845 0.45 NA >19.7

B1 0.055 None 0.070 989 0.96 914 0.47 4392 2.0

B2 0.055 Low 0.052 1022 1.03 932 0.46 5110 2.2

B3 0.055 Medium 0.052 922 1.03 931 0.46 4923 2.2

B4 0.055 High 0.050 1025 1.06 937 0.48 5337 2.2

B5 0.055 High 0.090 992 0.98 943 0.49 5147 2.0

B6 0.055 None 0.100 1010 0.99 943 0.48 5632 2.1

B7 0.075 High 0.062 1060 0.96 957 0.50 4691 1.9

Note: 1 in. = 25.4 mm; 1 kip-in. = 0.113 kN-m.

826 ACI Structural Journal/November-December 2005

unstrengthened specimen. Flexural cracks were spaced atapproximately 9 in. (230 mm) intervals. As anticipated, thecracks formed at sections coincident with stirrup locations.Thus, the steel strain gauges were successfully located at orvery near flexural cracks.

Application of prestrengthening service load cycles(Step 2)—To simulate years of traffic exposure, each specimenwas exposed to 100,000 sine-wave cycles applied at afrequency of 1 Hz. The magnitude of the cycles was selectedto induce a live-load stress range of 10.6 ksi (73 MPa) in thetension steel. According to live load test results from the WarMemorial Bridge, this stress range approximately correspondsto the range caused by a legal-load-limit truck. To cause areinforcement stress range of this magnitude an applied loadof 5.6 kip (24.9 kN) was required. The applied load cycleconsisted of a sine wave centered about 3.3 kip (14.7 kN)with an amplitude of 2.3 kip (10.2 kN). Because the WarMemorial Bridge has a relatively low traffic volume, it isestimated that 100,000 cycles of this magnitude are morethan the bridge would experience over 10 years of service.

Concrete substrate surface preparation (Step 3)—According to ACI 440.2R, localized out-of-plane variations,including form lines, should not exceed 1/32 in. (1 mm). Anyvoids should be filled with epoxy and all laitance, dust, dirt,oil, curing compound, existing coatings, and any othermatter that could interfere with the bond between the FRPand the concrete should be removed.3

Because the test specimens were cast and cured in alaboratory environment, the amount of contaminants on thesurface was minimal. Form lines on the surface of the beamswere removed using a grinding stone. On Beam B5, a smallpatch of honeycombing was filled using the same epoxyused to bond the FRP. Laitance and form oil were removedby applying isopropyl alcohol and scrubbing the surface witha stiff-bristled brush. This procedure was then repeated witha neutralizing agent. The surface was then allowed to dry.

Application of cycles during epoxy cure (Steps 4 to 7)—Toinvestigate the effects of leaving a bridge open to trafficduring FRP installation, five of the test beams were exposedto load cycles during strengthening and the epoxy-curingperiod. The cycles were continued for 48 hours after instal-lation of the FRP. Three different levels of traffic cycleswere simulated. The relative intensities and frequencies ofthe three cycle types—labeled “low,” “medium,” and“high”—are depicted in Fig. 5. The same amount of timeelapsed between FRP installation and ultimate strength

testing for all of the strengthened specimens regardless ofwhether they were subjected to load cycles during the epoxy-curing period.

The “low” intensity cycles were selected to represent thetruck loads experienced during actual strengthening operationsfor the War Memorial Bridge. Because traffic was divertedaway from the lane above the exterior girder beingstrengthened, the peak intensity of each cycle was approx-imately half of that used in the pre-strengthening cycles. Thiscycle was targeted to induce a live-load stress range in thesteel tension reinforcement of 4.8 ksi (33 MPa). Althoughthis level of loading is denoted “low” for the purposes of thisstudy, it was characterized by strains at the epoxy-concreteinterface that exceeded those of the Barnes and Mays5

concrete beam specimens discussed previously. The returnperiod of each load event was 5 minutes (300 seconds).

The “medium” intensity cycle consisted of the same waveused in the “low” cycles, but the frequency of occurrencewas increased to one per minute. Also, every fifth loadingwas approximately twice as intense and was selected toinduce a live-load stress range in the steel reinforcement of10.6 ksi (73 MPa). This loading regimen was developed torepresent a series of lighter vehicles interspersed withoccasional heavy vehicles traveling over the War MemorialBridge with no traffic mitigation measures in place.

The “high” intensity cycle resulted from increasing thefrequency of the “medium” cycle by a factor of 20. Thus, thelower magnitude event (4.8 ksi) occurred every 3 seconds, andthe large magnitude event (10.6 ksi) occurred every 15 seconds.This “high” intensity load pattern, which featured more than5700 heavy truck events and 23,000 lighter events per day,was selected to represent a hypothetical bridge with a highvolume of traffic.

FRP installation (Steps 5 and 6)—Except for the twospecimens with a thicker bond layer, the FRP reinforcementwas installed according to manufacturer recommendations. Thefirst step of the FRP installation process was the coating ofthe beam soffit with a saturant epoxy that served as a primer.This was done to fill small voids in the concrete and give asmooth surface for the application of the FRP. The epoxywas applied evenly to the surface of the concrete using apaint roller.

Once the primer became tacky, a thicker tack-coat epoxywas used to bond the FRP to the concrete. First, the FRP stripwas cleaned using a lint-free rag saturated with methyl ethylketone (MEK). The tack-coat epoxy was then mixed andspread over the FRP strip and the surface of the beam.

As mentioned previously, one of the variables investigatedin this study was thickness of the bond layer. Five of thestrengthened beams were installed as directed by the FRPmanufacturer. To apply the epoxy to the FRP, a taping knifewith a v-notch was used. The v-notch was sized so that theknife could be dragged over the FRP strip leaving a volumeof epoxy with a triangular cross section. The fresh epoxy hada thickness of 0.125 in. (3.2 mm) along the centerline of theFRP and zero along the edges. On the surface of the concrete,a trowel with multiple v-notches was used. The notches inthe trowel were 0.125 in. (3.2 mm) deep. This allowed auniform epoxy depth to be applied to the concrete surface.For the specimens with a thicker bond layer, the sameprocedure was used except that glass beads with a diameterwithin the range 0.079 to 0.091 in. (2.0 to 2.3 mm) weresprinkled into the epoxy prior to bonding the FRP strip to the

Fig. 5—Three levels of repair cycles (1 kip = 4.448 kN; 1 ksi= 6.895 MPa.

ACI Structural Journal/November-December 2005 827

beam. These glass beads were introduced to ensure that thedesired epoxy thickness was obtained.

Once the tack-coat epoxy became tacky, the FRP strip waspositioned on alignment marks on the underside of the beam.The strip was pressed onto the beam by means of a J-roller.This allowed excess epoxy to be squeezed out, produced auniform epoxy thickness, and removed all voids. Excessepoxy was then removed, and the outer FRP surface waswiped clean with a rag saturated with MEK. The finalmeasured epoxy thickness (bond layer thickness tb) for eachspecimen is given in Table 2.

Prior to the beginning of the experimental program, smallflexural test specimens were used to determine the necessarycuring time of the epoxy under laboratory conditions. Fromthese tests, it was determined that the epoxy was fullyeffective after curing for 48 hours at room temperature.Therefore, the epoxy on the beam specimens was allowed tocure for 2 days. Five of the strengthened beams were exposedto load cycles during this 48-hour period—two were not.

FRP instrumentation and post-strengthening cycles (Steps 8and 9)—At the conclusion of the 48-hour curing period, theload cycles were stopped so that the FRP strain gauges couldbe bonded reliably. A constant load of 0.5 kip (2.2 kN) wasmaintained on the beams at this time. The beams were heldat this load for 24 hours to allow the strain gauge adhesive tocure. Except where closely spaced at each end of the FRPstrip, strain gauges were bonded to the FRP directly beneathflexural cracks. Data were recorded and compared for loadcycles applied immediately before and after this 24-hourholding period to verify that there was no apparent stiffnessgain in the system.

To simulate years of service after strengthening, an additional20,000 load cycles were then applied to each specimen. Thecycles were of the same frequency, magnitude, and shape asthose used to represent prestrengthening load cycles (Step 2).Once these cycles were completed, the specimens wereloaded to determine their behavior and ultimate strength.

Loading to failure (Step 10)—The beam specimens wereloaded monotonically in selected load increments until thereinforcing steel yielded; further loading was applied indisplacement increments until failure. Except for theunstrengthened specimen, load increments were used up toan applied load of 22 kip (97.9 kN), where reinforcementyielding was evident from the load-deflection behavior. Atyielding of the tension steel reinforcement, the midspandisplacement ranged between 0.4 and 0.5 in. (10.2 and12.7 mm). After yielding of the tension steel reinforcement,the strengthened beams were loaded until midspan displace-ments of 0.65 and 0.9 in. (16.5 and 22.9 mm), respectively,were reached. The subsequent displacement incrementresulted in failure of all the strengthened beams due tocomplete loss of FRP anchorage on one end of each strip.

RESULTSA summary of the test results for the eight flexural ultimate

strength tests are listed in Table 2. The FRP-strengthenedspecimens are subdivided into three series:

1. Four specimens (B1, B2, B3, and B4) with “thin” epoxyand an FRP thickness of 0.055 in. (1.4 mm);

2. Two specimens (B5 and B6) with “thick” epoxy and anFRP thickness of 0.055 in. (1.4 mm); and

3. One specimen (B7) with “thin” epoxy and an FRPthickness of 0.075 in. (1.9 mm).

The relative term “thin” describes the epoxy thicknessresulting from FRP installation according to manufacturer-recommended procedures. Average measured epoxy thick-nesses for these five specimens ranged from 0.05 to 0.07 in.(1.3 to 1.8 mm). “Thick” epoxy was obtained by usingglass beads as explained previously. Measured epoxythicknesses for these two specimens ranged from 0.09 to0.10 in. (2.3 to 2.5 mm).

Behavior of unstrengthened specimenFigure 6 depicts the measured load versus midspan

displacement response for the unstrengthened specimen andthe first series of strengthened specimens (those with 0.055 in.[1.4 mm] FRP installed according to manufacturerrecommendations). The unstrengthened specimen wasunder-reinforced and very ductile. The full stroke of thehydraulic actuator was exhausted without a loss of load-carrying capacity. Testing ceased when the midspan deflectionreached 8.82 in. (224 mm). The experimental momentcapacity Mn,exp of 829 kip-in. (93.7 kN-m) was determinedas the total moment corresponding to the yield plateau of thebeam prior to the onset of strain hardening of the reinforcingsteel. This method corresponds to the conventional designpractice of neglecting any benefits of strain hardening whencomputing the nominal moment capacity.

Behavior of FRP-strengthened specimensAll of the strengthened specimens failed upon separation

between the FRP strip and the concrete beam. Except for afew small areas, a thin layer of surface concrete remained

Fig. 6—Effect of cycle intensity on response (1 in. = 25.4 mm;1 kip = 4.448 kN).

Fig. 7—Anchorage failure of FRP reinforcement.

828 ACI Structural Journal/November-December 2005

bonded to the FRP strip after debonding as shown in Fig. 7.Thus, the failure actually occurred within the concrete andnot at the adhesive interface. Because the entire cover layerdid not separate from the beam, however, this type of failureis referred to in this paper as “FRP debonding” instead of“cover delamination.” In all cases, debonding of the FRPinitiated beneath one of the load points and eventually prop-agated to the end of the strip.

Figure 8 is a plot of the measured FRP strains along thelength of a typical strengthened specimen (B4). Each seriesrepresents the FRP strains at a particular value of appliedload and is labeled according to the fraction of the maximumload Pmax applied to the strengthened beam prior to failure.All of the FRP strain gauges within 45 in. (1140 mm) ofmidspan were located at cross sections with flexural cracks.

For Specimen B4, steel strain measurements indicate thatthe steel reinforcement first yielded at the +18 in. (+460 mm)cross section (beneath one load point) at an applied load of0.89Pmax. The steel at the midspan section yielded shortlythereafter. Once the steel yields at a cross section, the FRPmust resist most of the tension force that develops due to theadditional moment applied to that section. Accordingly, itcan be seen that the FRP strains at these two sectionsincreased rapidly for the next relatively small load incrementshown (0.95Pmax). The steel yielded at the –18 in. (–460 mm)cross section (beneath the other load point) at an applied loadof 0.96Pmax. For a small increment beyond this load, the FRPstrain at this section began to increase rapidly as well.

Beyond the 0.96Pmax load level, the FRP strain at midspanbegan to decrease slightly as the load increased to 0.97Pmax.The test was paused briefly at this load level, which corre-sponded to a midspan displacement of 0.9 in. (22.9 mm).Local FRP debonding like that shown in Fig. 9 could be seenextending from the cracked sections at midspan and beneaththe load points.

As the load increased from 0.97Pmax to Pmax, the FRPstrains in the maximum moment region continued toincrease. Once this peak load was reached, however, furtherdisplacement of the beam (approximately 0.10 in. [2.5 mm]for this specimen) resulted in constant or decreasing FRPstrains in this region accompanied by sharply increasing FRPstrains extending out toward one end of the strip (negativedirection on Fig. 8). This indicates that once the locallydebonded lengths coalesced between the cracked sectionsnear the load points, debonding quickly propagated to oneend of the FRP. Less than 0.5 s after the final strains shownin Fig. 8 were recorded, the FRP strip debonded completelyalong a length that included the entire maximum momentregion and extended to one end of the strip. The total peakmoment corresponding to this anchorage failure of the FRPwas taken as the experimental moment capacity Mn,exp of thestrengthened specimen. After the FRP debonding occurred,the behavior of the strengthened specimens closely matchedthe behavior of the unstrengthened specimen.

The ductility of the specimens was greatly affected by theintroduction of the FRP. This is apparent from a comparisonof the member ductility indexes of the eight specimens listedin Table 2. The member ductility index was calculated as theratio of the midspan deflection at failure to the midspandeflection at yielding of the reinforcing steel. The ductilityindex of the unstrengthened beam actually exceeded thevalue tabulated because this test was terminated prior to theloss of load-carrying capacity. The ductility index of theunstrengthened beam was approximately 10 times that of thestrengthened specimens.

The yield moment My,exp and deflection ∆y,exp listed inTable 2 were obtained by inspection of the load versusdeflection curves. The limiting FRP strain εfe,exp correspondsto the largest FRP strain measured just prior to failure.

Effect of load cycles during epoxy curing (specimens with thin epoxy and FRPthickness of 0.055 in. [1.4 mm])

Four specimens, B1 to B4, were tested that had an FRPthickness of 0.055 in. (1.4 mm) and an epoxy thickness thatwas achieved by following the manufacturer’s installationinstructions. These specimens were designed to represent theactual amount of FRP strengthening provided on the WarMemorial Bridge. The four specimens differed only withrespect to the intensity of load cycles applied duringstrengthening. The load-displacement behavior for thesespecimens is compared in Fig. 6.

By comparing the behavior of the specimens (B2–B4) thatwere exposed to load cycles during strengthening to thebehavior of the strengthened specimen (B1) that was not, itis apparent that load cycles did not cause a reduction in theeffectiveness of the FRP strengthening. On the contrary,each of the low- and high-intensity cycle specimens wasactually slightly stronger (within 4%) as compared with theuncycled specimen. No trend linked the ultimate strength ofthe beams to the intensity of the load cycles. Likewise, eachspecimen that was exposed to cycles during strengthening hada larger strain in the FRP at failure εfe,exp than the uncycledspecimen. The increase in failure strain was 16, 12, and 22%for the beams subjected to low-, medium-, and high-intensitycycles, respectively.

Although the ultimate strengths of the cycled specimenswere not significantly larger than those of the uncycledspecimen, the cycled specimens all reached a larger deflection

Fig. 8—FRP strains along length of beam for Specimen B4(1 in. = 25.4 mm).

Fig. 9—Localized debonding of FRP reinforcement inmaximum moment region.

ACI Structural Journal/November-December 2005 829

prior to failure. The increase of deflection at failure rangedfrom 8 to 10%. This is most likely attributable to the fact thatthe FRP for these three specimens was able to sustain largerstrains in the maximum moment region prior to debonding.

As opposed to the study of Barnes and Mays5 discussedpreviously, there was no significant difference in the ambienttemperature during epoxy curing for the beam specimens inthis study. The laboratory was air-conditioned, and thehydraulic pump was located in a separate enclosure. There-fore, the slight increase in strength of the cycled specimenscannot be attributed to temperature variations. The results ofthis test series indicate that keeping the War MemorialBridge open to traffic during and after FRP installation didnot reduce the effectiveness of the strengthening.

Effect of epoxy thickness (specimens with thick epoxy and FRP thickness of 0.055 in. [1.4 mm])

It has been suggested by Harmon et al.7 that the decreasedbond layer shear stiffness resulting from the use of a thickerepoxy layer can increase the force resisted by the FRP priorto debonding. Specimens B5 and B6, with an intentionallythickened layer of epoxy, were tested. The load-displace-ment response of these two specimens is compared with thatof the corresponding thin-epoxy specimens in Fig. 10. Tobracket the two extremes of load cycles, Speciman B6 wasnot subjected to load cycles during strengthening, whereasSpecimen B5 was exposed to the “high” intensity cycles.

Contrary to the results of the previous series, the specimen(B5) exposed to cycles during epoxy cure achieved a slightlylower ultimate strength than the specimen (B6) not subjectedto cycles. The strain in the FRP at failure was also lower forthe specimen subjected to cycles. These small variations,however, are within the amount of variation that might beexpected for a comparison set of only two specimens.

For the measured bond layer thickness values reported inTable 1, the model introduced by Harmon et al.7 gives 6 to9% larger FRP tensile forces at failure for the thicker epoxyspecimens (B5 and B6) than for the companion specimens(B4 and B1). The measured moment capacity increase due tostrengthening (∆Mn,exp in Table 3) for B6 was 13% largerthan for Specimen B1, and the maximum measured FRPstrain increased by 28%. When the two specimens (B5 andB4) subjected to the high level of load cycle intensity arecompared, however, ∆Mn,exp for the thicker epoxy specimen(B5) is 17% less than for Specimen B4. The maximummeasured FRP strain for Specimen B5 is 4% less than forSpecimen B6. Because of the limited range of epoxy thicknessestested and the small number of specimens, no conclusion can bereached regarding the effect of epoxy layer thickness.

Effect of FRP thickness (specimen with thin epoxy and FRP thickness of 0.075 in. [1.9 mm])

The final specimen, B7, was strengthened with an FRPstrip with a thickness of 0.075 in. (1.9 mm). Its load-displacement behavior is compared with that of thecompanion 0.055-in. (1.4-mm) FRP specimen (B4) in Fig. 11.The FRP was installed according to the manufacturerrecommendations, so the epoxy was classified as “thin.” Thebeam was exposed to the high intensity load cycles duringthe epoxy-curing period. The specimen failed at a momentapproximately 3% greater than that of the companionspecimen (B4). The maximum measured FRP strain in thespecimen with thicker FRP was 88% of that measured in B4.However, when considering the increased cross-sectional

area of the FRP strip, the total FRP force developed in thethicker FRP prior to debonding was 120% of that developedin the thinner FRP. This increase is in agreement with the20% increase in FRP force predicted using the model for thistype of failure from Harmon et al.7, in which the capacity isproportional to the square root of the FRP thickness.

It should be noted that the use of thicker FRP currentlyresults in one or two penalties to the designer. First, analysisusing the limiting effective FRP strain of ACI 440.2R-023

yields no increase in strength for a thicker FRP strip relativeto a thinner FRP strip of the same width. Second, the reductionin limiting FRP strain attainable with the thicker FRP stripalso reduces the strain in the reinforcing steel at failure. Inthe most recent ACI 3188 and ACI 440.2R design recommen-dations, the design strength reduction factor is a function ofthe tensile strain in the reinforcing steel. Thus, depending onthe magnitude of the net tensile strain in the reinforcing steel,a further reduction in the design strength may result when athicker FRP strip is used.

In contrast with the prediction of the ACI 440.2R method,the total amount of force resisted by the thicker FRP in thisstudy was larger relative to the force resisted by the thinner FRP.Increasing the FRP thickness, however, proved inefficient—a36% increase in the FRP area resulted in only a 20% increasein the FRP force at failure.

Fig. 10—Effect of epoxy thickness on response (1 in. =25.4 mm; 1 kip = 4.448 kN).

Fig. 11—Effect of FRP thickness on response (1 in. =25.4 mm; 1 kip = 4.448 kN).

830 ACI Structural Journal/November-December 2005

Without a comparison specimen not subjected to loadcycles, it is impossible to draw a definite conclusionregarding the effect of load cycles on beams strengthenedwith the thicker FRP. This specimen, however, appears tohave performed at least as well as the comparable specimenwith thinner FRP bonded over the same contact area.

Limiting effective FRP strainAs discussed previously, the ACI 440.2R3 design procedure

limits the FRP strain used to compute the nominal capacityof the strengthened member. Computation of the limitingeffective strain εfe in accordance with ACI 440.2R yieldslimiting strain values of 0.0076 and 0.0056 for the specimenswith FRP thickness of 0.055 in. (1.4 mm) and 0.075 in.(1.9 mm), respectively. Figure 12 is a graphical comparisonbetween the largest FRP strain measured at any crackedsection in each strengthened specimen and the correspondingACI 440.2R design value.

The FRP strains never reached the failure level predictedby the current ACI 440.2R method. The experimental valuesranged from 58 to 74% of the predicted value for an FRPthickness of 0.055 in. (1.4 mm). The experimental value forSpecimen B7 was 84% of the predicted value for an FRPthickness of 0.075 in (1.9 mm).

The ramifications of the ACI 440.2R overestimation of thelimiting effective FRP strain may be seen in Table 3. Themeasured moment capacity Mn,exp of the unstrengthenedspecimen was 4% larger than the predicted capacity Mn,ACI;

therefore, the measured increase in moment capacity due toFRP strengthening ∆Mn,exp is compared with the increase inmoment capacity predicted by applying the ACI 440.2Rrecommendations ∆Mn,ACI. The value ∆Mn,ACI was 5 to 29%larger than ∆Mn,exp for the six specimens strengthened witha 0.055 in. (1.4 mm) strip. Only the 0.075 in. (1.6 mm) FRPspecimen exhibited a strength increase larger than the valuepredicted according to the ACI 440.2R recommendations.Thus, the ACI 440.2R procedure for predicting thedebonding strain can be unconservative for the type of FRPstrengthening system used in this study.

The progression of FRP strains depicted in Fig. 8 anddescribed above indicates that the prescription of a limitingFRP strain may still be a valid approach to preclude the typeof progressive debonding failure experienced in this study.Because the actual failure occurred within the concrete,however, it is reasonable to expect that the achievable FRPstrain be limited by the mechanical properties (stiffness,tensile strength, and/or fracture energy) of the concrete aswell as the stiffness of the FRP system.

CONCLUSIONSThe results of the experimental study support the

following primary conclusions:1. Concrete bridges can be strengthened effectively with FRP

even if they remain open to traffic during strengthening; and2. The current ACI 440.2R3 design recommendation for

computing the limiting effective FRP strain resulted inunconservative predictions of capacity increase for six of theseven beams strengthened with externally bonded, precured,unidirectional laminate strips.

Although these results support the first conclusion for thespecific materials used in this study, it is possible that trafficloads may adversely affect other types of FRP or adhesives.Nonetheless, the fact that at least some of these systems canbe installed without closing bridges to traffic significantlyincreases the attractiveness of FRP as a strengtheningalternative. Further research is required to establish aperformance specification for FRP systems that would ensureadequate behavior under these conditions. Ideally, such aperformance specification would require only inexpensive,small-scale materials testing to verify compliance.

The second conclusion should not be taken as acondemnation of the entire ACI 440.2R design procedure.The portion of the procedure devoted to determining the FRPdebonding strain, however, resulted in an overestimation ofthe actual amount of strengthening achieved in all but one ofthe specimens in this study, regardless of whether or not loadcycles were applied during strengthening. As stated inSection 9.2.1 of ACI 440.2R3: “Further research into themechanics of bond of FRP flexural reinforcement should resultin more accurate methods for predicting delamination.”

ACKNOWLEDGMENTSSupport for this study was provided by the Federal Highway Administration

(FHWA) and ALDOT. The funding, cooperation, and assistance of theseorganizations are gratefully acknowledged. The contents of this paperreflect the views of the authors who are responsible for the facts and accuracyof the data presented. The contents do not necessarily reflect the officialviews or policies of the FHWA or ALDOT. FRP materials for the laboratorystudy were donated by R. J. Watson, Inc., and Fyfe Co., LLC.

NOTATIONEa = modulus of elasticity of adhesiveEc = modulus of elasticity of concreteEf = modulus of elasticity of FRP

Fig. 12—Measured versus calculated limiting FRP strains(1 in. = 25.4 mm).

Table 3—Measured and predicted moment capacities

SpecimenMn,ACI, kip-in.

Mn,exp, kip-in.

∆Mn,ACI, kip-in.

∆Mn,exp, kip-in.

∆Mn percent difference

Unstrengthened 798 829 NA NA NA

B1 1004 989 206 160 +29

B2 1004 1022 206 193 +7

B3 1004 992 206 160 +26

B4 1004 1025 206 196 +5

B5 1004 992 206 163 +26

B6 1004 1010 206 181 +14

B7 1001 1060 203 231 –12

Note: 1 kip-in. = 0.113 kN-m.

ACI Structural Journal/November-December 2005 831

Es = modulus of elasticity of steelfc′ = compressive strength of concretefr = modulus of rupture of concreteft = splitting tensile strength of concretefu = ultimate strength of FRPfy = yield strength of steelMn,ACI = moment capacity according to ACI 440.2R-023

Mn, exp = experimental moment strengthMy, exp = experimental yield momentPmax = maximum load applied to strengthened beam prior to failuretb = thickness of epoxytf = thickness of FRP∆fail = midspan deflection at failure∆Mn,ACI = increase in moment capacity according to ACI 440.2R-023

∆Mn,exp = measured increase in moment capacity∆y = midspan deflection at yield of steel reinforcementεfe = effective strain level in FRPεfu = ultimate tensile strain of FRP

REFERENCES1. Swenson, K. S., and Barnes, R. W., “Design Procedure for FRP

Strengthening of War Memorial Bridge,” Interim Report 930-466, AuburnUniversity Highway Research Center, Auburn, Ala., May 2002.

2. Reed, M. W., and Barnes, R. W., “Effects of Traffic Loads during FRP

Strengthening of the War Memorial Bridge,” Interim Report RP-930-466-1,Auburn University Highway Research Center, Ala., Oct. 2004.

3. ACI Committee 440, “Guide for the Design and Construction ofExternally Bonded FRP Systems for Strengthening Concrete Structures(ACI 440.2R-02),” American Concrete Institute, Farmington Hills,Mich., 2002, 45 pp.

4. Macdonald, M. D., “Strength of Bonded Shear Joints Subjected toMovement during Cure,” International Journal of Cement Composites andLightweight Concrete, V. 3, No. 4, Nov. 1981, pp. 267-272.

5. Barnes, R. A., and Mays, G. C., “The Effect of Traffic Vibration onAdhesive Curing during Installation of Bonded External Reinforcement,”Proceedings of the Institution of Civil Engineers—Structures & Buildings,V. 146, No. 4, Nov. 2001, pp. 403-410.

6. El-Mihilmy, M. T., and Tedesco, J. W., “Prediction of AnchorageFailure for Reinforced Concrete Beams Strengthened with Fiber-ReinforcedPolymer Plates,” ACI Structural Journal, V. 98, No. 3, May-June 2001,pp. 301-314.

7. Harmon, T.; Kim, Y. J.; Kardos, J.; Johnson, T.; and Stark, A., “Bondof Surface-Mounted Fiber-Reinforced Polymer Reinforcement for ConcreteStructures,” ACI Structural Journal, V. 100, No. 5, Sept.-Oct. 2003,pp. 557-564.

8. ACI Committee 318, “Building Code Requirements for StructuralConcrete (ACI 318-05) and Commentary (318R-05),” American ConcreteInstitute, Farmington Hills, Mich., 2005, 430 pp.