Structural performance of composite sandwich bridge decks with hybrid GFRPsteel core

Composite Structures xxx (2010) xxx–xxx

Contents lists available at ScienceDirect

Composite Structures

journal homepage: www.elsevier .com/locate /compstruct

Structural performance of composite sandwich bridge decks with hybridGFRP–steel core

Hyo-Seon Ji a,⇑, Jae-Kwan Byun b, Chang-Soo Lee b, Byung-Jik Son c, Z. John Ma d

a Department of Civil Engineering, Daewon University College, 599 Sinwol-Dong, Jecheon-Si, Chungbuk-Do, Republic of Koreab Department of Civil Engineering, University of Seoul, 90 Jeonnong-Dong, Dongdaemun-Gu, Seoul, Republic of Koreac Department of Civil Engineering, Konyang University, 26 Nae-Dong, Nonsan-Si, Chungnam-Do, Republic of Koread Department of Civil and Environment Engineering, University of Tennessee, Knoxville, TN 37996, USA

a r t i c l e i n f o

Article history:Available online xxxx

Keywords:Composite sandwich bridge deckHybrid GFRP–steel coreStructural performanceANSYS

0263-8223/$ - see front matter � 2010 Elsevier Ltd. Adoi:10.1016/j.compstruct.2010.08.037

⇑ Corresponding author. Tel.: +82 43 649 3266; faxE-mail address: [email protected] (H.-S. Ji).

Please cite this article in press as: Ji H-S et al. S(2010), doi:10.1016/j.compstruct.2010.08.037

a b s t r a c t

This paper presents the static and fatigue performance of composite sandwich bridge decks with hybridGFRP–steel core. The composite sandwich bridge deck system is comprised of wrapped hybrid core ofGFRP grid and multiple steel box cells with upper and lower GFRP facings. Its structural performanceunder static loading and fatigue loading with a nominal frequency of 5 Hz was evaluated. The responsesfrom laboratory testing were compared with the ANSYS finite element predictions. The failure mode ofthe proposed composite sandwich bridge deck was more favourable because of the yielding of the steeltube when compared with that of all-GFRP decks. The ultimate failure of the composite sandwich deckpanels occurs by shear of the bonded joints between GFRP facings and steel box cells. Results from fatigueload test indicated no loss in stiffness, no signs of de-bonding and no visible signs of deterioration up to 2million load cycles. The thickness of the composite sandwich deck retaining the similar stiffness may bedecreased to some extent when compared with the all-GFRP deck. This paper also presents design of aconnection between composite sandwich deck and steel girder.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

It was recognized that most of structural deficiencies of bridgeswere shown in the bridge deck geometry and deck condition. Thebridge deck is a structural component that distributes and trans-fers loads transversely to the girders that bear on piers or abut-ments. The most commonly used type for slab-on-girder systemsis a cast-in-place concrete deck. Structural deficiencies of a bridgedeck are related to physical deterioration due to environmentalconditions, unanticipated factors like heavy vehicle traffic, fatigueand dynamic responses. They usually cause substantial problemsfor structural repair and replacement of a concrete deck. For hightraffic volume bridges, the speed of construction, especially forthe cases of bridge deck repair and replacement, has become a crit-ical issue more than ever. Strong momentum exists for the spreadof precast construction for bridges with a push to expand the lim-its, especially for the use in long-span bridges. One of the promis-ing systems for precast bridge construction has been concrete-decked, precast and prestressed concrete (DPPC) girders for super-structure [16]. The DPPC girder bridge does not need a cast-in-place deck, or any wearing course. By having the concrete deck

ll rights reserved.

: +82 43 649 3508.

tructural performance of comp

as an integral member, several benefits are obtained, for instance,an accelerated time frame for construction, a high quality plant-produced concrete deck, and an entire cross section prestressedfor enhanced section properties and durability. Despite several ma-jor benefits, one of the perceived disadvantages with the DPPCgirders is its increased weight in long-span bridges. The writershave been working on finding potential solutions including theuse of fiber-reinforced polymer (FRP) decks replacing the concretedeck in DPPC girder bridges [15].

In recent years the interest in using FRP in construction field hassignificantly increased worldwide. However, the use of FRP panelsfor bridge decks is still limited, due to the relatively higher cost ofthe FRP materials than the conventional ones [10,20]. Currentlymost FRP deck panel systems have the cellular-box geometries fab-ricated by the pultrusion process [2,9,13,17]. It is believed thatinnovative designs based on FRP materials could play a great rolein extending the application of FRP materials in bridge decks andcutting down the cost. One such design method is to use the hybridconcept. To make the most of FRP materials, combinations of FRPand conventional materials have recently been investigated.Deskovic et al. [7] investigated a hybrid FRP–concrete beam thatconsists of filament-wound glass fiber-reinforced polymer (GFRP)box section combined with a layer of concrete in the compressionzone and a thin carbon fiber-reinforced polymer (CFRP) laminate in

osite sandwich bridge decks with hybrid GFRP–steel core. Compos Struct

http://dx.doi.org/10.1016/j.compstruct.2010.08.037

mailto:[email protected]


http://www.sciencedirect.com/science/journal/02638223

http://www.elsevier.com/locate/compstruct


2 H.-S. Ji et al. / Composite Structures xxx (2010) xxx–xxx

the tension zone. Burgueno et al. [6] presented a hybrid bridge sys-tem with light weight concrete filled circular carbon/epoxy shellgirders and an E-glass/polyester deck. They showed that the designis stiffness-driven for the modular composite slab bridge systemwith the beams of CFRP materials shell type. Lopez-Anido andHan [14] presented the hybrid fiber-reinforced polymer-glulampanels for bridge decks. Mufti et al. [19] developed hybrid FRP–concrete bridge decks using pultrusion and filament-winding fab-rication methods. Van Erp et al. [23] developed hybrid FRP–con-crete beams for a bridge application in Australia. The hybridFRP–concrete beam consists of two parts: concrete in the compres-sion zone and GFRP box section in the tensile zone. Additional CFRPlaminate was added to the tensile flange of the box section to in-crease stiffness of the hybrid beam. A layer of concrete is bondedon the GFRP box section by an epoxy adhesive. The weight of thehybrid FRP–concrete beam was claimed to be about 1/3 of that ofa reinforced concrete beam. Aref et al. [4] investigated a hybridFRP–concrete multi-cell bridge superstructure that consists ofGFRP trapezoidal box beams combined with a layer of concretein the compression zone. As stated above, the research on the hy-brid structures has been mainly applied to the combinations of FRPand concrete. Ji et al. [11] proposed the composite sandwich decksystem with a hybrid FRP–steel core that has the potential to beused with DPPC girders. The main advantages of composite sand-wich panels for bridge decks are intended to not only improve stiff-ness and buckling response by the continuous support of coreelements with the face laminates, but also be cost efficient com-pared to all-GFRP decks.

This paper is focused on the static and fatigue performances ofthe composite sandwich panels with hybrid FRP–steel core forbridge decks. They were investigated by a combination of experi-mental and analytical programs. Research has shown that the de-sign of the composite sandwich panels is governed by stiffnessand fatigue. This paper also presents the connection design of acomposite sandwich deck for a steel I-girder bridge.

GFRP Facing

25.0 6.5UFL Grid

CF

Steel Tube

4@125 5=

+Bonding layer LFL

4.8

GFRP Facing

Fig. 1. The proposed composite sandwich de

Please cite this article in press as: Ji H-S et al. Structural performance of comp(2010), doi:10.1016/j.compstruct.2010.08.037

2. Experimental program

The experimental program was to investigate the static and fa-tigue performance of the proposed composite sandwich deck sys-tem [11] and to verify that the proposed deck system meets thestiffness requirement and has significant reserve strength. Threedifferent tests in this study were conducted on the compositesandwich deck panels as following: (1) design load test for obtain-ing flexural strength and structural stiffness of the deck; (2) ulti-mate load test up to failure; and (3) fatigue load test underservice load up to 2 million to inspect the reduction in stiffnessor strength due to fatigue.

2.1. Specimen description

The composite sandwich bridge deck system is comprised ofwrapped hybrid core of GFRP grid and multiple steel box cells withupper and lower GFRP facings [11]. Fig. 1 shows the cross section ofthe composite sandwich deck. The design procedure of the com-posite sandwich deck profile was summarized in [11].The dimen-sion details of the GFRP grid are shown in Fig. 2. The compositesandwich panel specimen used in this laboratory testing is shownin Fig. 3. The detailed manufacturing process of the compositesandwich panel specimens was described in [11].

2.2. Properties of the composite sandwich deck system

The main constituents of the composite sandwich bridge decksystem are fibers, resins and steel. Unidirectional E-glass roving,biaxial E-glass roving, continuous strand mat, polyester and vinylester resins were selected as the constituents for the GFRP lami-nates in the composite sandwich panels. The core of steel HSSs inthe composite sandwich panel was made of steel, grade KSD3568 steel [18]. The constituents’ material properties used forthe composite sandwich panels are shown in Table 1. Considering

GFRPGrid

+L Bonding layer

00

5.6

25.04.04.5

116

4.54.84.8

Steel Tube

ck panel (all dimensions in millimeter).



100

3007

100

100

6.5 25

100725

2525

Longitudinal

Transverse

Fig. 2. Plan view of the GFRP grid (all dimensions in millimeter).

Fig. 3. Composite sandwich deck panel.

Table 1Material properties.

Material E (GPa) G (GPa) m q (g/cm3)

E-glass fiber 72.50 27.60 0.22 2.54Polyester resin 3.38 1.38 0.38 1.24Vinyl ester resin 3.91 1.43 0.37 1.15Steel 205.8 79.15 0.3 7.85

H.-S. Ji et al. / Composite Structures xxx (2010) xxx–xxx 3


the fact that the vinyl ester resin offers a superior failure elonga-tion which is about twice that of polyester, the vinyl ester resinwas used at the bonding layers between the steel box cells andGFRP laminates as shown in Fig. 1. This selection took into accountthe coefficient of thermal expansion (CTE) mismatch between theGFRP and steel. The properties of the layers in the proposed decksystem are shown in Tables 2 and 3. The detailed informationabout the fiber layups and material design was summarized in[11].

2.3. Experimental setup, instrumentation, and procedure

2.3.1. Static load testThe experimental setup for the static load test is shown in Fig. 4.

A servo-hydraulic universal testing system was used for the flex-ural tests under the three-point loading. Its capacity was1000 kN. The effective span length of the composite sandwichpanel was 2.7 m, which is a little different from 2.5 m providedby the design manual [18]. The Korean Highway Specifications

Fig. 4. Experimental setup (static load test). (a) Locations of strain gauges at top. (b)Locations of LVDTs and strain gauges at bottom. (c) Locations of strain gaugesthrough the cross section.

Table 3Layer stiffness properties predicted using micro-mechanics model.

Ply name Ply orientation E1 (GPa) E2 (GPa) G12 (GPa) m12

UFL Random 7.99 7.99 1.69 0.349UFL 0�/90� 19.55 4.35 1.77 0.343LFL Random 7.99 7.99 1.69 0.349CFL 0�/90� 19.55 4.35 1.77 0.349Grid core 0� 20.38 4.42 1.80 0.341Steel core Isotropic 204.00 204.00 76.90 0.300Bonding layer Random 8.17 8.17 1.73 0.343

Table 2Layer properties of the proposed deck.

Ply name Ply type Weight(g/m2)

Ply thickness(mm)

Plies perlayer

Vf

UFL CSM 450 3.2 4.0 0.194UFL Biaxial roving 570 2.4 2.0 0.234LFL CSM 450 3.2 4.0 0.194CFL Biaxial roving 570 2.4 2.0 0.234Grid core Uniaxial roving 570 25.0 25 0.246Steel core – – – – –Bonding layer CSM 450 1.6 2.0 0.183




define a tire contact area. The tire contact area shall be assumed asa rectangle with a length-to-width ratio equal to 1/2.5. Its areashall be determined as 1.416P (mm2), where P is the design wheelload in N. As a result, the length L (in the traffic direction) andwidth W in mm can be calculated as follows:

L ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

0:566Pp

Þ; W ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

3:541Pp

ð1Þ

Therefore, the size of loading steel patch in the tests was580 mm� 230 mm as shown in Fig. 5a. The load was applied tothe deck specimen through a 580 mm� 230 mm loading patch torepresent a design truck wheel loading DB-24 [18]. A rubber padwith the same dimensions as the steel plate was placed betweenthe deck and the steel plate. A spreader beam was used so as to ap-ply force from the actuator. The load was applied at the center ofthe span. Two composite sandwich deck specimens were tested.Each test sample was instrumented with both strain gages and dis-placement transducers. The vertical deflections of the specimenswere measured using linear variable deflection transducers(LVDTs) at three points [D1, D2, D3 in Fig. 5b] at the mid-spanand another point [D4 in Fig. 5b] at the quarter span from the rightsupport. Strains were measured using strain gauges at the top faceof the specimen [T1, T2, TR1, TR2, TR3, and TR4 in Fig. 5a], at the

(a) Locations of strain gauges at top

(b) Locations of LVDTs and strain gauges at bottom

(c) Locations of strain gauges through the cross section

2.7m

1D 1BR

2D4D

4BR

1B

3BR3B

2B2BR3D

0.75m

3TR

1TR1T

Loading 0.23m

0.58m4TR

2T2TR

2.7m

1M

2M

3M

4M

5M

Fig. 5. Locations of LVDTs and strain gauges.


bottom face [B1, B2, B3, BR1, BR2, BR3, and BR4 in Fig. 5b], andthrough a cross section [M1, M2, M3, M4, and M5 in Fig. 5c]. Itshould be noted that ‘‘TR” and ‘‘BR” gauges mean the 45� strain ro-sette. The test samples were preloaded up to 100 kN for three cy-cles at regular intervals. Then they were subjected to service loadtests and ultimate load tests. In service load tests, they were testedup to the design wheel load 122.3 kN first and then to 491 kN dueto the 500 kN capacity of the load cell. In ultimate load tests, theywere loaded to failure by the setup of 1000 kN capacity to examinethe failure progress and strength.

2.3.2. Fatigue load testThe experimental setup for the fatigue load test of the compos-

ite sandwich deck panel is shown in Fig. 6. A servo-hydraulic uni-versal testing system was used for the fatigue test. A MTS 436controller was used to control the load range, the loading fre-quency and the number of cycles of the hydraulic actuator. Thecomposite deck panel in the test had two continuous spans andwas supported on one steel beam and two rollers as shown inFig. 7. The span between the steel beam on the right end and themiddle roller is 0.6 m and the other span is 2.1 m. This configura-tion was intended to get the fatigue responses of the compositesandwich deck panel at the critical sections for both the positivebending moment and negative bending moment. It can also studythe fatigue responses of the deck-to-girder connection which issimilar to those in actual bridge conditions. Fig. 8 shows a sideview of the deck panel before testing with the steel through-boltsattaching the deck to the steel support beam. A square steel tube150 mm� 150 mm� 6 mm was additionally used for the deck-to-beam connection due to upward displacement. A 600 mm�

Fig. 7. Schematic of the fatigue load test geometry.

Fig. 6. Experimental setup (fatigue load test).



Fig. 8. Side view of composite sandwich deck in fatigue test configuration. (a)Locations of strain gauges at top (units in millimeter). (b) Locations of strain gaugesat bottom (units in millimeter). (c) Locations of LVDTs at bottom (units inmillimeter).


500 mm box beam was utilized to apply the load at the mid-span.This spreader box beam was supported by the deck via a stack of50 mm thick steel plates used as loading patches. These rectangu-lar loading patches of a 580 mm� 230 mm with the larger dimen-sion transverse to the direction of traffic, were used to simulate thecontact area of one DB-24’s wheel. The test sample was subjectedto fatigue loading for 2000,000 cycles. The load was cycled be-tween the maximum load of 76.5 kN and the minimum load of7.65 kN at a frequency of 5 Hz under the load control. The loading

(a) Locations of strain g

(b) Locations of strain gau

(c) Locations of LVDT

1050

200500

1T

1050600

1TN

2TN

3TN

2T

3T

4T

1TL 2TL

Loading

1050

1B

1050600

1BN

2BN

3BN

2B

4B

1BL 2BL3BL

3B

1050

1D

10506003D

4D 2D5D

Fig. 9. Locations of LVDT


range was determined from the rear axle load of the DB-24 truckincluding the impact factor 1.3, which is 122.4 kN. Since the distri-bution width of wheel load is 1.6 m, the maximum load for the1.0 m wide test sample should be 76.5 kN (122.4 kN/1.6). The fati-gue load test was stopped periodically for static service load testsup to 122.4 kN, and the composite sandwich deck panel was in-spected for signs of damage periodically as well. The static serviceload tests were conducted on the test sample after 500,000,1000,000, 1500,000 and 2000,000 cycles during the fatigue loadtest. The test sample was instrumented with both strain gagesand displacement transducers as shown in Fig. 9. The verticaldeflections of the deck were measured using linear variable deflec-tion transducers (LVDTs) at three points D1, D2, D3 along the mid-section of the right span, D4 at the quarter-section of the right spanand D5 at the mid-section of the left span as shown in Fig. 9c.Strains were measured using strain gauges at the top face of thespecimen (T1, T2, T3, T4, TL1, TL2, TN1, TN2, and TN3 in Fig. 9a)and at the bottom face (B1, B2, B3, B4, BL1, BL2, BL3, BN1, BN2,and BN3 in Fig. 9b). The BL3 point in Fig. 9b shows the locationof the strain gauge mounted on the bottom surface of the steeltube. The deflections and strains for the test sample were recordedduring the test using a high-speed data acquisition system.

3. Results and discussion

3.1. Stiffness

The displacement at the mid-span of the deck panel was5.65 mm at the service load 122.3 kN in the first test, which was

auges at top (units in mm)

ges at bottom (units in mm)

s at bottom (units in mm)

3 intBL Po

s and strain gauges.




about L/478 for the span of 2.7 m. This was slightly less than thedeflection limit of L/425 of 6.35 mm. The load–displacementcurves of the measuring points are shown in Fig. 10. The panelexhibited a fairly linear behavior up to 122.3 kN. The stiffness ofthe deck panel was calculated to be 16.7 kN/mm. The predicteddeflection value of the steel deck alone at the service load of122.3 kN was compared with the experimental deflection valueof the composite sandwich deck. The maximum deflection atmid-span was calculated to be 6.77 mm, which is 19.8% higherthan the tested deflection of 5.65 mm. The stiffness of compositesandwich deck has slightly increased when compared to the steeldeck with tubes alone. Some audible noise was detected at150 kN. The mid-span bending strains from the top and bottomfaces were compared for the panel at a load of 491 kN in Fig. 11.

Fig. 10. Load–displacement response(s).

Fig. 11. The load–strain responses from design load test. (Note: ex = strain in x-directionfacing and (b) at bottom facing.


For example, the top surface longitudinal strain (X direction) atPoint ‘‘T1”, located at 135 mm away from the loading patch inFig. 5a, was �1006le in compression when the load was122.3 kN, and the strain at Point B1 was 911le at the same load le-vel. These strains are less than 20% of the maximum strain from theFRP coupon test. Please note that, in Fig. 11, the notations TR1 ex,TR1 e1 and TR1 e2 are used to denote the strains measured at thestrain rosette point TR1 in the span direction and two principaldirections. The strains BR1 ex, BR1 e1 and BR1 e2 are similar exceptthat they were measured at the bottom facing of the specimen.From Fig. 11, TR1 e1 and TR1 e2 show 3000le and 1250le, respec-tively, at the load of 400 kN, whereas BR1 e1 and BR1 e2 are 2960leand 1250le at the same load level. The calculated principal stresswas �23.33 MPa in compression at the top facing and 75.08 MPa intension at the bottom facing. The safety factor of the stressesagainst the ultimate strength varied from 2.8 to 8.8.

3.2. Failure mode

The failure mode of the sandwich panel was studied by the ulti-mate load test. The load–displacement curves up to 632.5 kN areshown in Fig. 12. As load was increased beyond 400 kN, theload–displacement curves became nonlinear. A maximum load of632.5 kN was obtained. The maximum measured displacementwas 92.3 mm. When the load of 632.5 kN was reached, a verystrong and loud noise was heard due to the de-bonding of GFRPlaminates from the wrapped GFRP grid and steel tube core. A per-manent displacement was observed beyond the load of 632.5 kN asshown in Fig. 12. The deck panel did not return to its original shapeafter releasing of the load unlike the cellular all-GFRP deck panels.The permanent displacement of the composite sandwich panel wasdue to the yielding of the steel tube core, and the failure of the fi-bers. Fig. 13 shows the failure of the specimen. The specimen faileddue to the delamination of the GFRP grid from the steel tube coreinstead of any local damage around the loading patch on the topfacing. This failure mode was different from the local punching fail-ure mode or other local damages of the cellular all-GFRP panels.The steel tube core yielding created the desired ductility in the

; e1 = the maximum principal strain; e2 = the minimum principal strain.) (a) At top



Fig. 12. The load–displacement responses from ultimate load test. Fig. 14. Strain gage data at a cross-section of the panel.

Fig. 15. Strain gage data at a section of neutral axis.


hybrid sandwich panel. The ultimate failure of the panel occurredbecause of the shear failure of the bonding layer as was investi-gated by Teixeira de Freitas et al. [22]. They showed that the ulti-mate failure of the composite bonded systems occurred by shear ofthe adhesive layer. A further study should be conducted to investi-gate the shear behavior of the bonding layer due to the influence ofdifferent thickness in the proposed deck. Fig. 14 shows the load–strain curves for the strain measuring points ‘‘M1”–‘‘M5” indicatedin Fig. 5c. From the test results shown in Fig. 15, the location of theneutral axis was found at 94.99 mm from the bottom face of thepanel at the service load level.

3.3. Stiffness degradation

The flexural fatigue load test was conducted to evaluate the fa-tigue performances at the critical sections within the two spansand the deck-to-girder connection. The load–deflection curves atseveral deck locations and load–strain curves during the statictests after 0, 0.5, 1.0, 1.5 and 2.0 million cycles were shown inFig. 16 and 17. The deflection and strain responses remained al-most the same for all of the static load tests, and the compositesandwich deck showed no apparent changes in stiffness of thestructure. The residual deflection in the composite sandwich deckpanel under the fatigue loading was found to be insignificant asshown in Fig. 18. Damage due to the fatigue loading was also

Fig. 13. Composite sandwich


inspected at the area of the deck panel around the bolted connec-tion. No visible signs of deck panel deterioration or de-bonding

panel at a failure load.



Fig. 16. Static deflections after different fatigue cyclings (D2 point).

Fig. 17. Deck strains after different fatigue cyclings at bottom facing (B3 point).

Fig. 18. Mid-span deflection history during fatigue test.

Table 4Deflections and strains from interrupted fatigue cycling (load = 122.4 kN).

Loadcycling(cycles)

Mid-span vertical deflection (mm) Mid-span transversestrain (microstrain)

LeftcenterlineD1

CenterlineD2

RightcenterlineD3

TopfacingT3

Bottomfacing B3

0 3.198 4.441 2.955 �571 201440,000 2.651 3.967 2.541 �548 207800,000 2.689 4.015 2.591 �556 2071200,000 2.689 4.034 2.620 �555 2151600,000 2.670 3.967 2.561 �554 2062000,000 2.708 4.015 2.581 �554 206

Table 5Comparison of strains and strengths from interrupted fatigue cycling(load = 122.4 kN).

Load cycling(cycles)

Bottom facingcenterline(FRP composites)

Bottom facing centerline (steel tube)

Mid-span transversestrain (microstrain)

Mid-spantransverse strain(microstrain)

Mid-spantransversestress (MPa)

0 490 383 78.8440,000 462 363 74.7800,000 475 371 76.41200,000 471 371 76.41600,000 470 383 78.82000,000 465 379 77.9Design limits 3930a 130b

a The allowable strain = 20% of failure strain of the FRP coupon test.b The allowable stress is provided by the Korean Highway Specifications [18].


between the GFRP grid core and steel multi-cell tube core were ob-served. A comparison of deflection and strain measurements be-tween the static load test without any previous fatigue cyclesand the static load test conducted after 2000,000 fatigue cyclesare shown in Table 4. The deflection at D2 for the test after fatiguecycling was 15.88 mm and the maximum strain in the transversedirection measured at the bottom facing was 477le. And the strainof the steel tube measured at the same location was 379le asshown in Table 5. The allowable strain of the FRP composites is20% of the maximum strain from the FRP coupon test, 3930leand the allowable stress of the steel is 130 MPa as summarizedin [11]. The limit normal strain not to cause the fiber de-bondingin the laminates made with chopped strand mat (CSM), woven rov-ing (WR) and uni-directional (UD) glass reinforcement may be esti-mated as 1000le after 1000,000 fatigue cycles as provided in the


handbook [8]. The tensile strain in the FRP composites and steelmembers of the composite sandwich deck panel were found tobe well below the design allowable strain. The value of EI calcu-lated is 8:63� 106N �m2 after the complete fatigue test. The flex-ural rigidity after the fatigue test was reduced about 3.5% whencompared to that calculated from displacements measured in thedesign load test.



Table 6Stresses for the composite sandwich deck (load = 400 kN).

Item GFRP (Experiment) Steel (ANSYS)

Top facing(fx

a)Bottom facing(fx

a)Bottom facing

Stress (MPa) �65.10 23.65 244Ultimate strength (MPa) �255.35b 210.03c 248.4d

Safety factor 3.92 8.88 1.02

Safety factor = stress/ultimate strength.a Longitudinal direction of deck (perpendicular to bridge axis).b Yield point of GFRP laminate (0�/90�).c Yield point of GFRP laminate (chopped strand mat).d Yield point of steel.

Table 7Comparison of deflections and strains (load = 122.4 kN).

Load cycling(cycles)

Mid-span vertical deflection (mm) Mid-span transversestrain (microstrain)

LeftcenterlineD1

CenterlineD2

RightcenterlineD3

TopfacingT3

Bottomfacing B3

0 3.198 4.441 2.955 �571 2012000,000 2.708 4.015 2.581 �554 206ANSYS 0.757 1.955 0.757 �248 177


4. Finite element analysis

4.1. FE model

FEM procedures have been successfully employed in researchstudying the performance of FRP bridge decks or their components[24]. A FE analysis using software ANSYS [3] was carried out for thetest specimen. The overall FE model for the static load tested spec-imen is shown in Fig. 19. The boundary conditions were modeledas a pin and a roller for the end supports. Displacement beyond aload of 400 kN became nonlinear. This load level was predictedwhen the steel tube core yielded, as shown in Table 6. The compar-isons of displacement and strain results between experiment andanalysis are shown in Table 8. The overall FE model for the fatigueload tested specimen is shown in Fig. 20. The boundary conditionsof the deck in fatigue load test were modeled as one span with afixed support and a roller, while the adjacent other span was sim-ply supported. The steel support girder in fatigue test was modeledby three dimensional beam elements. The interface between thedeck and steel support girder was modeled by rigid elements in or-der to transmit the load and displacement from the deck to themain girder. The boundary conditions at the nodes of the hybridFRP–steel deck’s lower facing and steel support girder’s top flangewere modeled as constrained displacement in the verticaldirection.

4.2. Material and stiffness properties and boundary conditions

The layers’ properties as described in Section 2.2 were used forthe FE analysis of the test specimens in this study. The boundaryconditions in static load test were modeled as constrained

CFL

Steel TubeLFL

Fig. 19. FE model of the specimen f


displacement in the vertical direction at nodes of the compositesandwich deck panel lower facing and steel support girder topflange of testing. The boundary conditions in fatigue load test weremodeled as one span with a fixed support and roller, while the

Grid

UFL

or the static load test (ANSYS).



Table 8Data of all-GFRP deck for comparison.

Decksystem

Depth(mm)

Weight(kg/m2)

Deflectionreported

Flexural rigiditya

ðNm2ð�106ÞÞ

Superdeck[2]

203 107 L/530 9.83

DuraSpan[17]

203 90 L/450 4.28

Proposeddeck

164 177 L/478 8.95

a Normalized flexural rigidity to DB-24 loading [18] for a 2.7 m center-to-centerdistance between supporting girders.


adjacent other span simply supported. The steel support girder infatigue test was modeled by three dimensional beam elements.

4.3. Results of analysis

The predicted deflection as shown in Table 7 at the design loadlevel of 122.3 kN was 1.955 mm, while the tested deflection was4.44 mm. These deflection values are less than the design limit of6.35 mm. The predicted deflection at D2 is 55.9% lower than thetested average deflection. But predicted strain at mid-span (B3) is11.9% lower than the tested strain. The comparisons of deflectionand strain results from FEA and test data at the load 122.4 kN(27.4 kip) are shown in Table 8. The predicted deflection at the de-sign load level of 122.3 kN (27.4 kip) was 1.955 mm, while thetested deflection is 4.44 mm (0.17 in.). These deflection valuesare less than the design limit of 6.35 mm (0.25 in.). The predicteddeflection at center is 55.9% lower than the tested average deflec-tion. But predicted strain at mid-span (B3) is 11.9% lower than thetested strain. As is pointed out by Wan et al. [24], these discrepan-cies between the responses from laboratory testing and the FEApredictions are justified as follows. Each monitoring point on thetest specimen was instrumented with both one displacementtransducer and strain gages. It is not possible to place these instru-ments at the monitoring point of the specimen as assumed in theFE model. Displacement beyond a load of 400 kN (89.69 kip) be-came nonlinear. The predicted yield load was considered as400 kN (89.6 kip). The laminated facing and the steel tube’s topflange was the parts that were particularly investigated since thedeck failed in these components. The normal stress contours for

Fig. 20. FE model of the specimen fo


the laminated facing and the steel tube’s top flange are shown inFig. 21 and 22. As shown in Fig. 21, the maximum stress in the hy-brid FRP–steel deck panel is 371.309 MPa. At the load of 400 kN,The induced stresses within the steel cells from FE analysis weresmaller than those of the strength limits. The proposed deck sys-tem has a safety factor of 3.92–8.8 against the material failure.The location of the neutral bending axis by the FE analysis was90.36 mm (3.56 in.) from the bottom face of the panel. The locationof the neutral bending axis by testing result was found at94.99 mm from the bottom face of the panel at the service loadlevel.

Fig. 23 shows the contours of deflection, strain and stress for themodel of the deck in the fatigue test. As shown in Fig. 23, the cur-vature slope of the deflection contour is very inclined in the trans-verse direction. As is pointed out by Zi et al. [25,26], it could beconfirmed that the contour’s nonlinearity in the transverse direc-tion is due to the shear deformation of the rectangular core type.An additional research of the composite sandwich deck with thefoam-filled GFRP grid could be conducted in the further researchto improve such a contour’s nonlinearity in the transverse direc-tion. The comparisons of deflection and strain results from FEAand test data at the load 122.4 kN are shown in Table 7. The pre-dicted deflection at the design load level of 122.3 kN for the spanof fatigue load test was 1.955 mm, while the tested deflection is4.44 mm. These deflection values are less than the design limit of6.35 mm. The predicted deflection at center is 55.9% lower thanthe tested average deflection. But predicted strain at mid-span(B3) is 11.9% lower than the tested strain.

5. Performance comparisons with all-GFRP decks

The results of three different tests conducted on the proposedcomposite sandwich panel were investigated. The failure load of632.5 kN was five times larger than the design wheel load of122.3 kN as specified in the specification [18]. The mid-spandeflections of the deck sample at the design wheel load were5.65 mm and 4.02 mm for the static test and the post-fatigue loadtest respectively. A deflection-to-span ratio of the deck sample wasabout L/478, which was less than the span design limitation of L/425 of 6.35 mm as specified in the specification [1,18]. This indi-cates good structural performance of the deck as far as strengthis concerned. In other words, the deflections will control the design

r the fatigue load test (ANSYS).



Fig. 21. Stress contour in bottom flange of steel tube of composite sandwich deck panel (load = 400 kN).

Fig. 22. Stress contour in top flange of steel tube of composite sandwich deck panel (load = 400 kN).


of this composite sandwich panel bridge deck system. From theexperimental observations and failure mode analysis, it can beconcluded that the failure mode of the proposed hybrid deck wasmore favourable because of the yielding of the steel tube whencompared with that of all-GFRP decks. The structural performanceof the test sample with regards to the design criteria showed thatthe deck had sufficient stiffness against the design wheel load. Noreduction in stiffness or strength was found after 2000,000 cyclesof fatigue load in the service load condition. The proposed compos-ite sandwich deck met the service load criteria after fatigue cycling.For the comparison purpose, the flexural rigidity, depth, anddeflection of the pultruded all-GFRP deck made by creative pultru-sions (CP), and MMC were considered [2]. Table 8 shows the com-parisons of the commercial decks with the proposed compositesandwich deck. Normalized flexural rigidity to DB-24 loading fora 2.7 m center-to-center distance between supporting girderswas calculated without considering shear deformation. Comparing


with those all-GFRP decks, the thickness of the proposed deck canbe decreased by 19%. The minimum depth [tmin (ft)] of a reinforcedconcrete bridge deck is stated in Eq. (2), where the span is in feet[5].

tmin ¼1:2ðspanþ 10Þ

30ð2Þ

The depth of the tested deck panel was 164.4 mm, which is lar-ger than the minimum thickness of a reinforced concrete deck usedat the span length of 30 m. The required minimum depth for a RCdeck from Eq. (2) in this case is 110.2 mm. The estimated cost ofthe available all-GFRP decks in the United States is about $700–807/m2 [5], whereas the estimated cost of the proposed deck isabout $375/m2. The cost of constructing concrete decks is about$200/m2 in Korea. This deck system may be used for a deckreplacement and repair projects of the bridges having a high trafficvolume to save the cost.



Fig. 23. Displacement contour of composite sandwich deck panel in fatigue load test (Z-direction).

Polystrene

Haunch

Grout

Grid

ShearStud

SteelGirder

CompositeSandwichPanel

Fig. 24. Schematic of proposed composite sandwich deck-to-girder connection.


6. Connection between deck and girder

The connection of FRP decks to steel or concrete main girdersremains a challenge. The connection of FRP decks to main girderswill be focused on reducing construction time and ensuring com-posite action between decks and girders. Currently, Keller et al.[12,21] investigated adhesively bonded joints between pultrudedFRP decks and steel girders. The proposed connection system be-tween the composite sandwich deck and a steel girder was givenin Fig. 24. The components of the proposed deck connection consistof the module of all-GFRP grids and steel shear stud. After the steelgirders are placed, polystyrene haunches are attached to the edgeof the top flanges of the steel girders with an epoxy adhesive. Ina composite bridge, the haunch height will vary due to thesuper-elevation of the roadway geometry. The modular hybriddeck panels with longer connection grid are then placed on thefoam. The surface of the longer connection grid of the modular hy-brid deck panels will be coated by a resin sand to adhere the grout


or polymer mortar to the horizontal surfaces of the top flanges ofthe steel girders. This method will be a design method for hybriddecked steel girders with flexible hybrid shear connections includ-ing conventional steel studs and FRP grids. After the installation ofall of the composite sandwich deck panels, the non-shrinkagegrout or polymer mortar will be placed up to the bottom of thelonger connection grid. These deck-to-girder connections shouldbe capable of sustaining the required number of loading cycleswith no loss in structural capacity. The modular composite sand-wich deck panels are light enough for two construction workersto carry. The assessments of connection behaviours between thecomposite sandwich deck panels and girders will be studied inthe future based on the investigations in [12,21].

7. Conclusions

The performance of the proposed hybrid deck panel for highwaybridges was evaluated using an experimental approach. Based on




the results of the investigations presented in this paper, followingconclusions are summarized:

(1) The proposed hybrid FRP–steel core for bridge decks wasable to not only improve stiffness and buckling response,but also be cost efficient compared to all-GFRP decks.Besides, it was also able to reduce the weight of all-steeldecks. This paper shows a combination of FRP and conven-tional materials like steels that can fully take advantage ofeach material’s strengths and characteristics.

(2) The thickness of the hybrid deck may be decreased by 19%when compared with the all-GFRP decks with the same flex-ural rigidity. A deflection-to-span ratio of the hybrid deck, L/478 was less than the span design limitation of L/425 asspecified in the specification.

(3) The failure mode of the proposed hybrid deck was morefavourable because of the yielding of the steel tubes whencompared with that of all-GFRP decks.

(4) The estimated cost of the available all-GFRP decks in theUnited States is about $700–807/m2, whereas the estimatedcost of the proposed deck is about $375/m2. In Korea, thecost of constructing concrete decks is about $200/m2. Theproposed deck system may be used for a deck replacementand repair projects having a high traffic volume to save thecost.

(5) No reduction in stiffness or strength was found up to2000,000 cycles of fatigue load in the service load condition.The proposed composite sandwich deck met the service loadcriteria after fatigue cycling.

(6) The assessments of connection behaviours between thecomposite sandwich deck panels and girders will be studiedin the future.

Acknowledgments

The research described in this paper was conducted with thesupport from Seong-won Construction Ltd., the University of Seoul,and the National Science Foundation under Grant No. 0550899awarded to the fifth author. The authors deeply thank for the assis-tance. Any opinions, findings, and conclusions or recommendationsexpressed in this paper are those of the authors, and do not neces-sarily reflect the views of the sponsors.

References

[1] AASHTO. AASHTO LRFD bridge design specifications, 4th ed. Washington (DC);2007.


[2] Alagusundaramoorthy P, Harik IE, Choo CC. Structural behavior of FRPcomposite bridge deck panels. J Bridge Eng, ASCE 2006;11(4):384–93.

[3] ANSYS 10.0: documentation ANSYS Inc. distributed with ANSYS 10.0; 2005.[4] Aref AJ, Kitan Y, Lee GC. Analysis of hybrid FRP–concrete multi-cell bridge

superstructures. Compos Struct 2005;69:346–59.[5] Bakis CE, Bank LC, Brown VL, Cosenza E, Davalos JF, Lesko JJ, et al. Fiber-

reinforced polymer composites for construction – state-of-the art review. JBridge Eng, ASCE 2002;6(2):73–87.

[6] Burgueno R, Karbhari VM, Seible F, Kolozs RT. Experimental dynamiccharacterization of an FRP composite bridge superstructure assembly.Compos Struct 2001;54:427–44.

[7] Deskovic N, Triantafillout, Meier U. Innovative design of FRP combined withconcrete: short-term behavior. J Struct Eng, ASCE 1995;121(7):1069–78.

[8] European structural polymeric composites group (EUROCOM). In: Clarke JL,editor. Structural design of polymer composites – EUROCOMP design code andhandbook. UK: E&FN Spon, Ltd.;1996.

[9] Federal Highway Administration (FHWA). FRP decks and superstructures:current practice; 2002. <http://www.fhwa.dot.gov/bridge/frp/deckprac.htm>.

[10] Ji HS, Son BJ, Chang SY. Field testing and capacity-ratings of advancedcomposite materials short-span bridge superstructure. Compos Struct2007;78(2):299–307.

[11] Ji HS, Son BJ, Ma ZJ. Evaluation of composite sandwich bridge decks withhybrid FRP–steel core. J Bridge Eng, ASCE 2009;14(1):36–44.

[12] Keller T, Gurtler H. In-plane compression and shear performance of FRP bridgedecks acting as top chord of bridge girders. Compos Struct 2006;72:151–62.

[13] Kim HY, Hwang YK, Park KT, Lee YH, Kim SM. Fiber reinforced plastic deckprofile for I-girder bridges. Compos Struct 2005;67:411–6.

[14] Lopez-Anido R, Han X. Structural characterization of hybrid fiber-reinforcedpolymer glulan panels for bridge decks. Journal of Composites forConstruction, ASCE 2002;6:194–203.

[15] Ma ZJ, Choppali U, Li L. Cycling tests of a fiber-reinforced polymer honeycombsandwich deck panel at very cold temperatures. International Journal ofMaterials and Product Technology 2007;28(1/2):178–97.

[16] Ma ZJ, Chaudhury S, Millam J, Hulsey JL. Field test and 3D FE modeling ofdecked bulb-tee bridges. Journal of Bridge Engineering, ASCE2007;12(3):306–14.

[17] Martin Marietta Composites (MMC), Ltd. DuraSpan fiber-reinforced polymerbridge deck system; 2005. <http://www.martinmarietta.com/products>.

[18] Ministry of Construction and Transportation (MOCT). Standards specificationsfor highway bridge, 2nd ed., Korea; 2000.

[19] Mufti AA, Labossiere P, Neale KW. Recent bridge applications of FRPs inCanada. Struct Eng Int 2002;12(2):96–8.

[20] Nystrom HE, Watkins SE, Nanni A. Financial viability of fiber-reinforcedpolymer bridges. J Manage Eng, ASCE 2003;19(1):2–8.

[21] Schollmayer M, Keller T. Modeling of through-thickness stress state inadhesive joints connecting pulturded FRP bridge decks and steel girders.Compos Struct 2009;90:67–75.

[22] Teixeira de Freitas S, Kolstein H, Bijlaad F. Composite bonded systems forrenovations of orthotropic steel bridge decks. Composite Structures 2010;92:853–62.

[23] Van Erp G, Heldt T, McCormick L, Carter D, Tranberg C. An Australian approachto fibre composite bridges, fibre composites design and development.Toowoomba, Australia: University of Southern Queensland. <http://www.fcdd.com.au:2002>.

[24] Wan B, Rizos DC, Petrou MF, Harries KA. Computer simulations and parametricstudies of GFRP bridge deck systems. Compos Struct 2005;69:103–15.

[25] Zi G, Kim BM, Hwang YK, Lee YH. An experimental study on static behavior of aGFRP bridge deck filled with a polyurethane foam. Compos Struct 2008;82:257–68.

[26] Zi G, Kim BM, Hwang YK, Lee YH. The static behavior of a modular foam filledGFRP bridge deck with a strong web-flange joint. Compos Struct 2008;85:155–63.


http://www.fhwa.dot.gov/bridge/frp/deckprac.htm

http://www.martinmarietta.com/products:2005

http://www.fcdd.com.au:2002

http://www.fcdd.com.au:2002


Documents

Structural performance of composite sandwich bridge decks with hybrid GFRPsteel core