Composites: Part B 32 (2001) 309±322

www.elsevier.com/locate/compositesb

Strengthening steel bridge sections using CFRP laminates

Rajan Sen a,*, Larry Liby b, Gray Mullins c
a
Samuel and Julia Flom Professor, Department of Civil and Environmental Engineering University of South Florida, Tampa, FL 33620, USA
b
Formerly graduate student, Department of Civil and Environmental Engineering, University of South Florida, Tampa, FL 33620, USA
c
Assistant Professor, Department of Civil and Environmental Engineering, University of South Florida, Tampa, FL 32620, USA
Received 17 December 2000; accepted 15 February 2001

Abstract
This paper presents results from an experimental investigation to determine the feasibility of using carbon ®ber reinforced polymer
(CFRP) epoxy laminates to repair steel composite bridge members. Six specimens, each consisting of a 6.1 m long W8 £ 24 wide ¯ange A36
steel beam acting compositely with a 0.114 m thick by 0.71 m wide reinforced concrete slab, were ®rst loaded past yield of the tension ¯ange
to simulate severe service distress. The damaged specimens were then repaired using 3.65 m lengths of 2 or 5 mm thick CFRP laminates
bonded to the tension ¯ange and tested to failure. The results indicated signi®cant ultimate strength gains but more modest improvement in
the elastic response. Non-linear ®nite element analyses were in good agreement with the experimental results. The study suggests that it is
feasible to strengthen steel composite members using CFRP laminates. q 2001 Elsevier Science Ltd. All rights reserved.
Keywords: A. Laminates; Carbon ®ber reinforced polymer

1. Introduction has yielded. Finally, for this system, the CFRP/adhesive

bond is the weakest link and will always control the mode
Fiber reinforced polymers have been successfully used to of failure. Thus, consideration of surface preparation alone
strengthen structural elements made of concrete, masonry will not suf®ce. Instead, more attention must be paid to
and wood. Numerous applications have been cited for augment the capacity of the adhesive through the use of
bridge structures (e.g. repair of prestressed concrete and appropriate fasteners.
wood bridges in Switzerland [10], multi-span hollow box Successful strengthening of steel bridge sections with
girder bridge in Germany [11], slab bridges in Japan [12], CFRP would be particularly useful as steel stringer bridges
and reinforced/prestressed concrete bridges in the United constitute a large segment of aging bridges in the United
States [13]). States that are nearing the end of their useful service life. If
Although there have been some studies which involved service life could be extended, highway authorities would
strengthening steel members (e.g. Refs. [3,4]), this exten- have access to a lightweight, speedy and therefore econom-
sion has not been so widespread since it poses a different set ical strengthening method that would minimally disrupt
of problems. First, the likelihood of lateral buckling makes traf®c.
it necessary to fabricate composite steel sections where the The opportunity to use CFRP laminates to strengthen
compression ¯ange is continuously supported by a rein- steel came unexpectedly. In 1992, the University of
forced concrete slab. These sections are much more dif®cult South Florida, Tampa, FL were developing a steel
to fabricate and test in comparison with rectangular concrete composite bridge model to evaluate the restraint effect
or wood sections. Second, steel's high strength and stiffness of elastomeric bridge bearings [7]. For this study, the
make it a more dif®cult material to strengthen especially girder spacing, slab thickness or the number of shear
since carbon ®ber reinforced polymers (CFRP) that are connectors was not critically important. Consequently,
commonly used have a smaller modulus. This means that advantage was taken to alter these parameters to boost
substantial load transfer can only take place after the steel the ultimate capacity of the slab and to ensure that if
the concrete slab were cut longitudinally, three identical
steel composite beams would result. In addition, an
* Corresponding author. Fax: 11-813-974-2957. identical second bridge was fabricated to provide addi-
E-mail address: sen@eng.usf.edu (R. Sen). tional specimens.
1359-8368/01/$ - see front matter q 2001 Elsevier Science Ltd. All rights reserved.
PII: S 1359-836 8(01)00006-3
310 R. Sen et al. / Composites: Part B 32 (2001) 309±322

Fig. 1. Cutting bridge into three composite beam sections.

Following completion of the experimental phase of the 3. Specimen details

bearing restraint study, the concrete slabs of the two bridges
were cut (Fig. 1) to obtain the six test specimens used in this As mentioned already, the six specimens used in the study
study. This paper presents a brief summary of the technical were obtained by cutting each of two concrete/steel compo-
problems that were overcome and highlights some of the site bridge models into three parts (Fig. 1). The resulting
principal ®ndings. A comprehensive report describing all cross-section is shown in Fig. 2. It consisted of a 114 mm
aspects of the research is available [6]. thick, 710 mm wide reinforced concrete slab that was
attached to a W8 £ 24 steel section by 36 shear connectors.
The spacing of the 50 mm long, 19 mm diameter stud-type
2. Objectives shear connectors was varied Ð they were more closely
spaced near the supports than at mid-span. The concrete
The overall goal of the study was to assess the feasibility slab deck was reinforced top and bottom in the longitudinal
of using CFRP laminates for strengthening steel composite and transverse directions by 20M (#6 bars). These were
bridge members. Since only distressed sections would be spaced at 230 mm in the transverse direction and 280 mm
strengthened it was necessary to simulate such damage in the longitudinal direction.
prior to any strengthening. The main objectives of the The specimens were identi®ed as North, Center or South
study may be summarized as follows: depending on their location in the original bridge deck.
Although the six specimens were geometrically identical,
1. To develop a procedure that could be used for strength- test coupons provided by the fabricator indicated that the
ening steel composite members using CFRP laminates. yield strength of the A36 steel in the two bridge models
2. To evaluate the potential bene®t of strengthening were 310 and 370 MPa. Thus, three of the specimens had
damaged steel composite specimens using CFRP lami- wide ¯ange beams with a yield stress of 310 MPa and three
nates. others had a yield stress of 370 MPa. In the following, speci-
3. To assess the extent to which experimental results may mens are identi®ed by their position in the original bridge
be predicted using available non-linear ®nite element deck and the yield stress that is used as a suf®x, e.g. North-
computer programs. 310 MPa, Center-370 MPa, etc.

Fig. 2. Slab reinforcement details.

 R. Sen et al. / Composites: Part B 32 (2001) 309±322 311

Fig. 3. Loading of simply supported steel composite bridge section.

4. Materials specimens. This was achieved by pre-loading each speci-

men (see Fig. 3 for schematic arrangement) past yield of the
4.1. Concrete tension ¯ange, a much higher stress level than the maximum
stress of 55% of yield permitted by the AASHTO Bridge
Normal weight concrete used in casting the bridge deck Speci®cations [1]. As a result of this loading, there was
was supplied locally and was selected to satisfy Florida permanent deformation in each member prior to strengthen-
Department of Transportation's speci®cations for Class II ing. A summary of the applied loads, de¯ections and maxi-
concrete. Specialist help was used for pouring, screeding mum recorded steel strains for all the specimens appears in
and ®nishing. The 28 day compressive strength of Table 1. The applied loads averaged 142 and 187 kN for the
43.6 MPa was 40% higher than the 31 MPa design strength. 310 and 370 MPa girders, respectively.
An additional 18% compressive strength gain was realized
at the time of testing, which resulted in a compressive
strength range of 48.5±52.6 MPa. 6. Repair of damaged specimens

4.2. CFRP laminates 6.1. Beam/CFRP combination

The CFRP laminates used were fabricated by Composite Since there were two different sets of yield strengths (310
Horizons, Covina, CA using a simple hand layup process. and 370 MPa) and two different CFRP laminate thicknesses
Widely available T300 carbon ®bers and an epoxy matrix (2 and 5 mm), several beam/laminate combinations were
were used. The laminate had a uni-directional elastic considered. However, to obtain the most data, each steel
modulus of 114 GPa and an average unidirectional tensile type was strengthened using each of the two different
strength of 1840 MPa. A linear stress-strain relationship CFRP laminate thicknesses.
gave an ultimate failure strain of 1.6%. Three of the six Two of the three lower capacity 310 MPa steel composite
laminates were 2 mm thick, the remaining three were beams were strengthened using thicker 5 mm CFRP lami-
5 mm thick. These thicknesses are considerably greater nate since this would provide a measure of the maximum
than those used for strengthening concrete or wood and strength gain that was attainable. The remaining specimen
were established from non-linear ®nite element analyses was strengthened using the 2 mm laminate. This meant that
to ensure comparable strengthening [9]. The width of the two of the three 370 MPa specimens were strengthened
laminate was 0.165 m to match the width of the ¯ange of the using 2 mm CFRP laminates (providing a lower bound on
W8 £ 24 beam but its length was only 3.65 m, being limited strength gain) and one by the 5 mm laminate.
by fabrication considerations of the CFRP laminate.
6.2. Adequacy of adhesive
4.3. Adhesive Because the epoxy adhesive is so much weaker than steel,
The two-part epoxy adhesive FR1272 was selected based non-linear ®nite element analyses were carried out to eval-
on the recommendation of the CFRP laminate fabricator. It uate its adequacy under ultimate conditions. The results of
is relatively inexpensive and is widely available. However, the analyses indicated that the epoxy would fail at the ends
it has a pot life of 10±15 min at 248C making its ®eld of the CFRP laminate due to high peeling stresses. To
applicability limited. prevent this type of failure, a three-piece steel clamp was
designed so that it could be positioned precisely where
needed to withstand the predicted peeling stresses. The
5. Simulation of damage clamps were attached to the bottom ¯ange of the W8 £ 24
section using two, 19 mm A490 bolts and two smaller 8 mm
In order to meet the objectives of this study, it was hardened steel bolts. Only two sets of clamps were needed
®rst necessary to simulate severe service distress in the since they could be re-used. Brief details on the analysis and
 312
Table 1

R. Sen et al. / Composites: Part B 32 (2001) 309±322

Summary of results

Specimen ID Mid-span values Plain section Strength gain Comments

(CFRP thickness) ultimate capacity (%)
(kN)
Simulating service damage Repaired specimen

Load De¯ection Tensile strain Ult. load De¯ection Concrete strain CFRP strain
(kN) (mm) (me ) (kN) (mm) (me ) (me )

North 310 (2-mm) 142 40.90 1524 237 102.9 2258 6646 196 21.0 No failure bolts
Center 370 (2-mm) 186 44.96 1744 271 . 100.0 2267 8164 249 a 8.7 No failure no bolts
South 370 (2-mm) 187 44.96 1717 272 . 100.0 2284 7811 249 a 9.1 No failure no bolts
Center 310 (5-mm) 142 39.69 1239 b 298 96.0 1918 5099 196 52.2 CFRP rupture along bolts
South 310 (5-mm) 142 30.23 1275 b 169 32.3 786 1538 196 213.7 Adhesive failure no bolts
North 370 (5-mm) 187 45.72 1695 329 87.9 1778 4859 249 a 32.0 CFRP rupture along bolts
a
Predicted value.
b
indicates gage failure at speci®ed strain (after value).
 R. Sen et al. / Composites: Part B 32 (2001) 309±322 313

Fig. 4. Clamp assembly used to withstand peeling stresses.

design of the clamps may be found elsewhere [8]. Complete matically recorded by a SYSTEM 4000 data acquisition
details of the study are presented in Ref. [6]. system.

6.3. Bonding CFRP to steel 7.2. Loading

Surface preparation is important for ensuring a successful The same four-point loading scheme used for simulating
bond between CFRP laminate and steel. Since the steel service damage (see Fig. 3) was used for testing the repaired
girders used in the bridges were provided with a protective specimens. The applied load was transmitted through two
coat of paint, it was necessary to remove this paint. The 0.71 m long stiffened spreader beams that were welded to a
specimens were turned over to provide access to the bottom 1.22 m long stiffened cross beam. The length of the spreader
¯ange where the CFRP was to be bonded and a portable beams was the same as the width of the concrete slab. This
sandblaster was used to obtain the necessary etched surface. meant that the test specimens were subjected to two
The CFRP laminate was bonded to the steel surface using uniform, transverse line loads spaced 1.22 m apart.
the epoxy adhesive in accordance with the manufacturer's The crossbeam was directly connected to the ram of a
recommendations. After the adhesive cured for 48 h, the 45 ton hydraulic jack using a swivel head that was attached
specially designed clamps resisting peeling stresses were by custom-designed connectors. This ensured that the loads
positioned over the ends of the laminate and secured in were transferred equally to the two spreader beams. Four
place using bolts as shown in Fig. 4. 150 mm £ 200 mm fabric pads were placed under both
spreader beams to ensure uniform distribution.

7. Testing of repaired specimens 7.3. Adhesive failure

7.1. Instrumentation The ®rst two specimens tested were the Center- and
South-370 MPa steel composite beams strengthened using
The test specimens were most heavily instrumented at the 2 mm CFRP laminates. The test was stopped when no addi-
quarter and mid-span locations. Strain gages were tional load could be applied (271 and 272 kN) and the
embedded inside the deck slab during casting and were de¯ection exceeded the 100 mm limit of the LVDT (see
also attached to the top concrete surface and to the underside Table 1). There was no adhesive failure, as was predicted
of the steel ¯anges. Strain gages were also af®xed to the by non-linear analysis.
CFRP laminate at mid-span and at 50 mm from each end of The third specimen tested was South-310 MPa steel
the 3.65 m long CFRP laminate, i.e. 1.27 m from the end of composite beam strengthened using 5 mm CFRP laminate.
the beam. Additionally, gages were also attached to the This test ended suddenly with the separation of the CFRP
bottom ¯ange of the steel beam just beyond the CFRP lami- laminate from the steel ¯ange. The clean separation was
nate (1.14 m from the end of the beam). Comparison of believed to be adhesive failure due in part to a manufactur-
strains in the steel beam and that in the CFRP laminate in ing defect that left a small depression on the bonded face of
close proximity allowed slip to be readily detected. the 5 mm thick CFRP laminate.
LVDTs were used to measure de¯ections at the mid-span To prevent a recurrence of this failure, bolts were used to
and quarter-span locations. Load cells were used to measure augment the load transfer capacity of the epoxy adhesive.
applied loads and support reactions. All data were auto- The con®guration and arrangement of bolts was based on
314 R. Sen et al. / Composites: Part B 32 (2001) 309±322

Fig. 5. Load-displacement North-310 MPa-2 mm (epoxy 1 bolted).

non-linear ®nite element analysis and consisted of two lines laminates in Figs. 8±10. A summary of the maximum loads
of three, 19 mm A490 bolts spaced at 41.9 cm, starting and mid-span de¯ections and strains in both concrete and
8.9 cm from the edge of the laminate [6]. In all, a total of the CFRP laminate is presented in Table 1. This table also
12 steel bolts were used (six at each end) in the two remain- contains information on the load applied to simulate damage
ing tests involving 5 mm thick CFRP laminates. Although referred to earlier.
no bolts were needed for the 2 mm CFRP laminate, they Table 1 shows that with one exception South-310 MPa,
were provided nonetheless to assess the effect of load trans- increases in the ultimate strength were substantial. For
fer in the presence of bolts. A pair of bolts was provided example, Center-310 MPa realized an increase in strength
8.9 cm from each end of the CFRP in the last specimen from 142 to 298 kN, a strength gage of 110%. Inspection of
strengthened using 2 mm, (for details refer to Ref. [8]). Figs. 5±10 shows that the strengthening effect is largely
con®ned to the post-yield region. This is not altogether
surprising since the CFRP laminate has a lower elastic
8. Results for repaired beams modulus 256% that of steel. This meant that prior to yield-
ing, stresses in the composite laminate were lower than
8.1. Strengthening those in the bottom steel ¯ange.
The premature local adhesive failure of South-310 MPa,
Load vs displacement plots comparing the responses of enabled it to be re-tested to determine the ultimate capacity
the six repaired and plain specimens are presented in Figs. of the steel section (see Fig. 9). The results of this test were
5±10. The results for specimens repaired with 2 mm CFRP used to ®ne-tune the non-linear ®nite element model to
laminates appear in Figs. 5±7, and those for the 5 mm CFRP predict the ultimate capacities of all the specimens that

Fig. 6. Load displacement Center-370 MPa-2 mm (epoxy only).

 R. Sen et al. / Composites: Part B 32 (2001) 309±322 315

Fig. 7. Load displacement South-370 MPa-2 mm (epoxy only).

were tested. The predicted value was found to be 249 kN for ing effect in the elastic region. This is an important consid-
the 370 MPa girders. Using the measured and the predicted eration particularly for bridge application where live load
failure loads, the absolute increases in ultimate strength of de¯ection limits are very stringent.
the strengthened steel beams were found to range between 9 Figs. 5±10 show that there is very little stiffening in the
and 52% (Table 1). The smaller increases correspond to the elastic region except for specimens strengthened with the
2 mm laminate and the larger increases to the 5 mm ones. 5 mm CFRP laminates (Figs. 8±10). Even for this case,
In addition to increases in ultimate load, the elastic region there was practically no stiffening for Center-310 MPa
of the plain section was considerably extended in the specimen (Fig. 8) due to cracking in the slab from the initial
repaired specimen. The increases ranged from about 20% loading. For specimens strengthened using 2 mm laminates
for South-370 MPa/2 mm (Fig. 7), where the elastic range there is even a reduction in stiffness in the elastic region
was extended from 156 to 187 kN, to over 67% in the case (Fig. 7). This is probably because the concrete damage in
of North-370 MPa/5 mm (Fig. 10). This increase is particu- this beam could not be fully offset by repairs and the stiff-
larly encouraging since it suggests that service loads of ness lost as a result was not regained by bonding the CFRP
repaired specimens could be increased using this technique. laminate.
The lack of increase of stiffness in the elastic range is due
8.2. Stiffening to CFRP's smaller elastic modulus compared with steel. The
increase in stiffness resulting by bonding a CFRP laminate
The load vs de¯ection response of the strengthened and to the steel surface farthest from the neutral axis is partially
unstrengthened beams also provide a measure of the stiffen- offset by this lower modulus. However, with thicker

Fig. 8. Load displacement Center-310 MPa-5 mm (epoxy 1 bolted).

316 R. Sen et al. / Composites: Part B 32 (2001) 309±322

Fig. 9. Load displacement South-310 MPa-5 mm (epoxy only).

laminates, increases are larger provided there is no adhesive Figs. 13±16. The results for the 2 mm CFRP laminates are
failure (Fig. 10). If local adhesive failures occur, strength- shown in Figs. 13 and 14 while those for the 5 mm laminates
ening cannot be expected (e.g. Fig. 9). in Figs. 15 and 16.
Inspection of Figs. 13 and 14 showing the strain variation
8.3. Failure mode for the 2 mm CFRP laminate indicate their similarity. This
is not surprising since there was no separation of the 2 mm
The load de¯ection plots in Figs. 5±10 indicate that, in
laminates, whether bonded or bolted. In each of these cases,
general, there was a considerable amount of de¯ection when
the load was transferred primarily by the adhesive.
the maximum load was reached (Fig. 11). For the beams
The comparison of strain variation along the depth of the
strengthened using 2 mm CFRP the test was stopped when
specimen strengthened with 5 mm CFRP laminate shown in
no additional load could be applied. As such, there was no
Figs. 15 and 16 indicate obvious dissimilarities. For the
failure of the CFRP laminate or its bond with the steel
bonded specimen (Fig. 15) there is load transfer to the
surface. With the beams strengthened using 5 mm CFRP,
laminate by the epoxy until adhesive failure. The bolted
however, failure occurred due to shearing off the CFRP
specimen (Fig. 16) transfers load in the same manner until
laminate by the bolts (see Fig. 12).
the CFRP delaminates from the steel at about the same load.
8.4. Bolted vs bonded The loads hitherto being carried by the epoxy were now
transferred to the bolts with commensurate strain relaxation
In order to assess the relative performance of bonded and in the CFRP laminate. Failure ®nally occurred when
bolted specimens, the strain variations with depth at the excessive strain (and therefore deformation) around the
critical mid-span locations for different cases are shown in bolts led to rupture of the CFRP laminate (Fig. 12). This

Fig. 10. Load displacement North-310 Mpa-5 mm (epoxy 1 bolted).

 R. Sen et al. / Composites: Part B 32 (2001) 309±322 317

Fig. 11. De¯ected shape of repaired 310 MPa beams at 218 kN.

problem may be avoided by using a greater number of bolts, dictate transferring shear using bolts, the use of adhesives
thereby limiting their maximum strain and by using CFRP would still be recommended to help distribute the load and
laminates manufactured with bolt holes and multi-direc- provide transfer of shear until the bolts become fully loaded
tional carbon ®ber reinforcement around such holes. in bearing. In this case, the use of adhesives served to secure
the CFRP laminate while the bolt holes were drilled. It
would be preferable to use a larger quantity of smaller
9. Discussion bolts to minimize the magnitude of localized stress concen-
trations at the bolts and to reduce the tendency for transverse
The results from the study indicate clearly that it is possi- CFRP laminate fractures. The magnitude of the localized
ble to strengthen steel composite bridge sections using stress concentrations may also be reduced by implementing
CFRP laminates. Because of the high strength and stiffness high quality drilling techniques, and by utilizing CFRP
of steel the operation is somewhat more complicated than laminates which have more plies with varying ®ber orienta-
that used for concrete or wood. In this study, a combination tion in the bolt hole regions, or by using CFRP laminates
of adhesive, fasteners, and clamps were used. specially manufactured to include reinforced holes for bolt-
Because T300 carbon ®bers were used for cost reasons, ing. Installation would be facilitated by such manufactured
strengthening was more pronounced in the post-yield region holes or by pre-drilling the abrasive, drill-resistant CFRP
than in the elastic region. However, if CFRP laminates with laminates prior to transport to a ®eld site.
higher stiffness are used, an increase in the elastic region Although bolts and clamps were successfully designed,
may be expected. With the prices of carbon ®ber dropping, sized, and positioned to aid in shear transfer and to resist
this may be realizable earlier than previously thought peeling stresses at the laminate ends, the use of bolts some-
possible. what negates the time ef®ciency of CFRP repairs. To ensure
Since the epoxy adhesive is very weak in comparison effective repairs the bond interface of the CFRP must be free
with steel, bolting was required especially to assist in the from defects at the onset to maintain uniform adhesive
load transfer for the thicker laminates. In con®gurations that thickness and reduce incidence of local adhesive failures.

Fig. 12. Rupture of 5 mm CFRP plate (Center-310 Mpa).

318 R. Sen et al. / Composites: Part B 32 (2001) 309±322

Fig. 13. Strain variation at mid-span Center-370 MPa-2 mm (epoxy).

Additionally, tapered laminates in both width and thickness are shown in these ®gures. Additional results, including a
accompanied by a scheme of adhered tapering with an adhe- complete set of comparisons for load vs de¯ection and steel
sive ®llet can reduce the stress concentrations leading to strains for the unreinforced and strengthened beams tested,
high peeling stresses. may be found elsewhere [2].
Inspection of Figs. 18±20 show excellent correlation
between the experimentally measured strains and predictions
10. Non-linear ®nite element analysis from the ®nite element analysis. This is especially true with the
2 mm laminate. The correlation is somewhat poorer for the
The non-linear ®nite element analysis program 5 mm laminate at higher loads. This is not altogether surpris-
PCFRAME [5] was used to predict the response of the six ing since some of the load taken by the bolts could not be
steel composite bridge sections tested. The concrete slab modeled. Use of bolts in the 2 mm specimens proved to
was discretized into 20 layers and the steel reinforcement have no effect as the adhesive suf®ciently transferred the
by two. The wide ¯ange section was modeled as eight steel shear forces without causing bearing in the bolts.
layers with the CFRP laminate and the adhesive represented Agreement for load vs displacement was not as close with
by single layers (see Fig. 17). Along the length, each beam the ®nite element analysis predicting a stiffer post-yield
was discretized using 41 members. response. This was because the effect of pre-damage was
Figs. 18±20 show comparisons of predictions from the only approximated in the analysis. The overall agreement
non-linear ®nite element analysis with test results. The between experiment and ®nite element analysis was gener-
strain variation in the CFRP laminate and concrete strain ally quite satisfactory.

Fig. 14. Strain variation at mid-span North-310 MPa-2 mm (epoxy 1 bolted).

 R. Sen et al. / Composites: Part B 32 (2001) 309±322 319

Fig. 15. Strain variation at mid-span South-310 MPa-5 mm (epoxy only).

11. Conclusions tively. Corresponding increases for the bridge sections

containing 370 MPa wide ¯ange shapes were 9 and
On the basis of the experimental and numerical studies 32%, respectively.
presented, the following conclusions may be drawn: 3. Increases in the elastic stiffness were relatively modest
(see Figs. 5±10) particularly when thinner 2 mm CFRP
1. Steel composite members may be strengthened by bond- laminates were used. However, increases in the elastic
ing CFRP laminates. However, much thicker laminates region of strengthened members were substantial indicat-
are needed to achieve comparable strength and stiffness ing the potential for increases in service loading.
gains typically obtained in strengthened concrete or 4. The failure mode of the strengthened sections was gener-
wood members. Additionally, the capacity of the epoxy ally ductile and accompanied by considerable deforma-
adhesive to transfer loads to the CFRP laminates may tion (see Fig. 11). However, where only epoxy adhesive
need to be augmented, e.g. by clamps or fasteners, to was used for bonding a 5 mm CFRP laminate, adhesive
ensure load transfer past yielding of steel. failure led to separation of the CFRP laminate and
2. Test results showed signi®cant increases in ultimate sudden failure (South-310 MPa). No similar separation
capacity of steel composite bridge members strengthened was observed with 2 mm CFRP laminates bonded using
by CFRP laminates. Strength gains for the bridge epoxy alone in that smaller tensile forces developed in
sections containing 310 MPa wide ¯ange shapes were the thinner CFRP and the associated shear stresses never
21 and 52% for the 2 and 5 mm CFRP laminates, respec- exceeded the epoxy's shear strength.

Fig. 16. Strain variation at mid-span Center-310 MPa-5 mm (epoxy 1 bolted).

320 R. Sen et al. / Composites: Part B 32 (2001) 309±322

Fig. 17. Discretization of cross-section for ®nite element analysis.

5. The CFRP laminate must be adequately anchored to the predicted with reasonable accuracy by non-linear ®nite
steel ¯ange to prevent peeling failure at the ends. In this element analyses.
study, a re-usable clamp system (see Fig. 4) was devel-
oped from non-linear ®nite element analysis that proved The study demonstrated the feasibility of using CFRP lami-
to be simple, ef®cient and inexpensive. These peeling nates for strengthening steel composite bridge members. If
forces could be minimized by sculpting the ends of the the intent of the installation is to increase the service load
laminates and adhesive to reduce stress concentrations in capacity and not to simply restore lost capacity due to
this region. However, this would probably require order- damage, both the slab and shear connectors attaching it to
ing the laminates to case-speci®c lengths and would the upper steel ¯ange must have the reserve capacity needed
reduce ®eld installation ¯exibility. for supporting the additional compressive and shear forces.
6. The response of the strengthened specimens was While clamps and fasteners were both used in the study,

Fig. 18. Comparisons: load vs concrete strain (North-310 Mpa-2 mm).

 R. Sen et al. / Composites: Part B 32 (2001) 309±322 321

Fig. 19. Comparisons: load vs CFRP strain (North-310 Mpa-2 mm).

ideally clamps alone will be more desirable for ®eld instal- Acknowledgements
lation as it eliminates the need for heavy magnetic drills that
will otherwise be needed to secure the CFRP laminates. This The investigations reported were carried out with the ®nan-
will greatly facilitate the speed with which repairs can be cial support of the Florida Department of Transportation and
carried out particularly in congested urban areas Ð a very the US Department of Transportation. The authors gratefully
important consideration for using the light-weight CFRP acknowledge the contribution of Dr Mohsen Shahawy,
laminates in the ®rst place. Director, Structural Research Center, Florida Department of
The effects of moisture and humidity on the long-term Transportation and Mr Ken Spillett, formerly a graduate
performance of the adhesive must be evaluated before ®eld student in the Department of Civil and Environmental
applications. However, this issue was not included within Engineering, University of South Florida, currently a Profes-
the scope of this study. In addition, appropriate measures sional Associate, Parsons Brinckerhoff Construction Services,
developed by the aerospace industry should be used to Tampa, FL. They also wish to express their sincerest
prevent possible galvanic action due to the direct contact appreciation to Composite Horizons, Covina, CA for their
of the CFRP laminate with steel in the presence of moisture. invaluable advice and assistance. However, the opinions,

Fig. 20. Comparisons: load vs CFRP strain (North-370 Mpa-2 mm).

322 R. Sen et al. / Composites: Part B 32 (2001) 309±322

®ndings and conclusions expressed in this publication are [7] Sen R, Spillett K. Restraint effect of bridge bearings. Final Report
those of the writers and not necessarily those of the Florida submitted to Florida and US Department of Transportation, January,
National Technical Information Service, VA, 1994.
or US Department of Transportation.
[8] Sen R, Liby L, Spillett K, Mullins G. Strengthening steel composite
bridge members using CFRP laminates. In: Taerwe L, editor. Non-
metallic (FRP) reinforcement for concrete structures. London,
References
England: E and FN Spon, 1995. p. 551±8.
[9] Wang Y. Bridge strengthening using advanced composites,
[1] American Association of State Highway and Transportation Of®cials. MSCE thesis submitted to the Department of Civil Engineering
Standard speci®cations for highway bridges. Washington, DC., 15th and Mechanics, University of South Florida, Tampa, FL, December,
ed., 1994.
1992.
[2] Liby L. Steel composite bridge sections strengthened by carbon ®ber
[10] Meier U, Deuring M, Meier H, Schwegler G. Strengthening of struc-
reinforced plastic laminates, MSCE Thesis submitted to Dept. of Civil
tures with CFRP laminates: research and applications in Switzerland.
Engng and Mechanics, University of South Florida, Tampa, FL,
In: Neale KW, Labossiere P, editors. Advanced Composite Materials
November, 1993. 353 p.
in Bridges and Structures. CSCE, Montreal, P.Q., 1992. p. 243±51.
[3] Mertz DR, Gillespie JW. Rehabilitation of steel bridge members
[11] Rostasy FS, Hankers C, Ranisch E-H. Strengthening of R/C and P/C
through the application of advanced composites. Final Report,
structures with bonded FRP plates. In: Neale KW, Labossiere P,
NCHRP-93-ID01, 1996.
editors. Advanced Composite Materials in Bridges and Structures.
[4] Mosallam AS, Chakrabarti PR. Making the connection, Civil Engi-
CSCE, Montreal, P.Q., 1992. p. 253±63.
neering, ASCE, April, 1999. p. 56±9.
[12] Uemura M, Aoyama K, Kliger H. A new reinforcing ®ber sheet
[5] PCFRAME. Computer program for non-linear analysis of reinforced
material and its application in the repair of concrete structures. In:
and prestressed concrete frames. Developed by Y. Kang, Ph.D disser-
tation, University of California, Berkeley, 1977. 39th International SAMPE Symposium, April, Vol. 39, 1994. p. 347±
[6] Sen R, Liby L. Repair of steel composite sections using CFRP lami- 54.
nates. Final report submitted to Florida and US Department of Trans- [13] Shahawy M. Strengthening highway bridges with carbon ®ber mate-
portation, August, National Technical Information Service, VA, rials. In: 74th Annual Meeting of the Transportation Research Board,
1994. Washington, D.C., January, 1995.