Engineering Structures 40 (2012) 5163

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Performance of a scaled FRP deck-on-steel girder bridge model with partial degree of composite action
Julio F. Davalos a, An Chen b,, Bin Zou a
a b

Department of Civil and Environmental Engineering, West Virginia University, Morgantown, WV 26506-6103, United States Department of Civil Engineering, University of Idaho, Moscow, ID 83844-1022, United States

a r t i c l e

i n f o

a b s t r a c t
The applications of FRP decks have been increasingly implemented in the United States since the mid-1990s. Although extensive research has been conducted on stiffness and strength evaluations of various types of FRP decks, studies on FRP deck-on-steel girder systems are limited. The performance of the FRP deck-on-steel girder system depends substantially on the connectors used. Recently, a prototype shear connector was developed and its advantages were illustrated through both laboratory study and eld applications, where partial degree of composite action could be achieved between the deck and stringers. This paper aims to study the effectiveness of the shear connector at bridge-system level, including the static and fatigue performance of the shear connector and the bridge system, the degree of composite action of the system; and more importantly, the inuence of the partial degree of composite action on load distribution factor and effective ange width since no provision is provided in AASHTO Specications. For this purpose, a 1:3 scaled FRP deck bridge model was tested, with an FRP sandwich honeycomb deck connected to steel girders using the prototype shear connector. The test program consisted of three phases: Phase I and Phase II included static and fatigue load tests on the scaled bridge model. In Phase III, a T-section was cut-out from the bridge model and then loaded under three-point bending until failure. A Finite Element model was constructed to simulate the tests. It was shown that the shear connection was able to provide partial composite action of about 25%, and sustain a cyclic fatigue loading equivalent to 75 year bridge service life span. It is shown that AASHTO Specications can still be applicable for load distribution factor, while the effective ange width needs to be re-dened for bridge-system with partial degree of composite action. The ndings from this paper can be used for design purposes. 2012 Elsevier Ltd. All rights reserved.

Article history: Received 23 November 2010 Revised 2 February 2012 Accepted 19 February 2012 Available online 23 March 2012 Keywords: FRP bridge deck Scaled bridge model Testing Finite Element analysis Load distribution factor Effective ange width Partial degree of composite action

1. Introduction Fiber Reinforced Polymer (FRP) bridge decks offer signicant advantages for rapid replacement and new construction due to their favorable characteristics for durability, lightweight, high strength, rapid installation, lower or competitive life-cycle cost, and high quality manufacturing processes under controlled environments. The applications of FRP decks have been increasingly implemented in the United States since the mid-1990s. Although extensive research has been conducted on stiffness and strength evaluations of various types of FRP decks [14], only limited studies are available on FRP deck-on-steel girder bridge systems, which were mostly evaluated based on eld- or lab-scale testing. Keelor et al. [5] conducted a eld study on a short-span bridge located in Pennsylvania, USA. This bridge had a pultruded FRP deck over ve steel girders equally spaced at 1.75 m; the span-length

Corresponding author.
E-mail address: achen@uidaho.edu (A. Chen). 0141-0296/$ - see front matter 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.engstruct.2012.02.020

was 12.6 m, and the deck thickness was 195 mm. The FRP deck was assumed to achieve full composite-action through grouted stud connections welded to the stringers. Their results showed that under service load condition, this full composite action resulted in effective-widths corresponding to about 90% for interior stringerspacing and 75% for exterior half stringer-spacing. Keller and Gurtler [6] conducted lab tests on two large-scale T-sections to study composite action and effective ange width. Each test model was 7.5 m long with a pultruded FRP deck section of 1.5 m wide adhesively bonded to the top ange of a steel supporting beam. The in-plane normal strain distribution, i.e., strain parallel to the bonded surface, across the width of the FRP section was recorded at both upper and lower FRP facesheet components. The results showed that under service limit state, the normal stress was almost uniform across the panel section. While under failure limit state, the normal stress decreased towards the panel edges, indicating a more pronounced effect of shear lag under ultimate load. Later, two reduced scale T-sections were tested to service limit state and ultimate limit state [7]. One of the T-section was fatigue loaded to 10 million cycles. The FRP pultruded anges, which

52

J.F. Davalos et al. / Engineering Structures 40 (2012) 5163

were adhesively bonded to the steel stringers, were able to achieve full composite action. At ultimate limit state, the failure mode was deck compression failure with the yielding of steel stringer. The structural behaviors of this full composite model were established at service and ultimate limit states. The deection and ultimate strength could respectively increase by 30% and 56% by considering composite action. There were strain differentials between top and bottom facesheet due to low in-plane shear stiffness of the core. The strain distribution of top and bottom facesheet showed that the top facesheet fully participated as a top cord, while the bottom facesheet showed more shear lag phenomenon. The effective ange width was smaller than the effective ange width of a comparable concrete deck. Also the T-section could sustain 10 million cycles of fatigue loading, which was comparable to the Eurocode 1 fatigue load on a reference bridge. Fatigue tests on two T-beams with 7.5 m span length and adhesively bonded FRP deck to steel stringers were conducted by Keller and Tirelli [8]. The pultruded FRP ange was full compositely connected to the steel stringers, and the FRP ange participated as the top cord of this T-section. The fatigue limit was considered to be 25% of static failure load at 10 million load cycles, which was far above the actual fatigue load in FRP bridges. The adhesive bond connection was proved to be able to sustain the fatigue loading. The aforementioned studies were focused on FRP deck-on-steel girder system with full composite action. However, in actual bridges, because of the low equivalent modulus of the FRP deck, the contribution of the deck to the bridge system is not signicant even for full composite action. Also, because of the difference in coefcient of expansion between the FRP and steel, where FRP has a typical Coefcient of Thermal Expansion (CTE) in the range of 1.62.7 105/C and steel has a CTE of 1.2 105/C, full composite action may induce an adverse effect on the bridge system if the shear connection is fully constrained. To address the above concerns, a prototype shear-stud sleeve-type connection was developed by Davalos et al. [9] and further studied by Righman et al. [10] and Davalos et al. [11]. This connection has been implemented in the eld, and more recently applied to the Wildcat Creek Bridge, Indiana [12]. It was proved that proper response of the bridge under static and fatigue loads can be achieved and the FRP deck can be secured for uplifting and in-plane forces by using this connection. Since the practical spacing for this connection was about 0.61.2 m, comparing to conventional concrete deck connection spacing of 0.150.25 m, there is partial degree of composite action between the deck and stringers. In the AASHTO Standard [13] and LRFD [14] Specications, slabon-girder bridges can be designed with either non-composite or full-composite action, but there is no provision for partial degree of composite action. Bridges with partial composite are typically designed as non-composite action, which is not economical since the contribution the deck based on partial composite action is neglected. Therefore, it is the aim of this paper to study the behavior of FRP-deck on steel girder bridge system with partial degree of composite action and provide design parameters, using a combined experimental investigation and Finite Element (FE) analysis. In particular, the objectives are: (1) to evaluate the performance of the prototype shear connection at the bridge-system level; (2) to evaluate the performance of the bridge system under static and fatigue loads; and (3) to study the degree of composite action of the bridge system and its inuence on the design parameters including transverse load distribution factor and effective ange width. 2. Test plan The test consisted of three phases. In Phase I, static load tests were conducted on a 1:3 scaled bridge model to investigate trans-

verse load distribution factor and local deection of the FRP deck. In Phase II, the same bridge model was subjected to a cyclic loading to evaluate the fatigue resistance of both shear connection and the bridge system. In Phase III, a T-section was cut-out from the bridge model and then loaded under three-point bending for linear and ultimate response to investigate its effective ange width, degree of composite action, strength, and failure mode.

3. Test models 3.1. Bridge model description A 1:3 scaled bridge model with a span of 5.5 m was constructed consisting of three steel stringers (W16 36, Gr50) spaced at 1.22 m apart (Figs. 1 and 2), based on a reference bridge designed according to AASHTO Specications [15]. A 5.5 m 2.74 m 0.13 m FRP deck, consisting of three 1.8 m wide and 2.74 m long individual FRP honeycomb panels as shown in Fig. 3 and assembled by tongue-and-groove connections along the two 2.74 m transverse joints, was attached to the stringers using a prototype studsleeve connector. The longitudinal direction of the honeycomb core (Fig. 3) is perpendicular to the trafc direction. As shown in Fig. 4, the FRP honeycomb panels consisted of top and bottom facesheet and a sinusoidal core. The facesheet has three layers of CDM 3208 laminate and two layers of Chopped Strand Mat (ChSM) composed of E-glass ber and polyester resin, with the material properties shown in Tables 1 and 2. Another layer of ChSM (0.256 kg/m2) is placed in between the facesheet and core as a bonding layer. The total thickness of the FRP panel is 125 mm, where the facesheet and the core are 12.5 mm and 100 mm thick, respectively. During construction, the steel girders with pre-welded shear studs at 0.6 m apart were rst erected and braced at the supports and mid-spans. FRP panels, with 80 mm circular holes cut at corresponding locations, were installed on the top of the steel girders, and then snug-tted by transverse tongue-and-groove connections, which were joined together using polymer resin and bonded at the top and along the joints using ber glass sheets, as shown in Fig. 5. The tongue and groove honeycombcore connection was about 0.15 m wide and was lled with polymer concrete to strengthen the joint, as shown in Fig. 6. Finally, the shear connections were installed, as shown in Fig. 7. The connection details, including installation; design formulas based on static and fatigue tests conducted at component

Fig. 1. Photo of scaled bridge model.

J.F. Davalos et al. / Engineering Structures 40 (2012) 5163

53

Fig. 2. Details of scaled bridge model.

level on push-out specimens to obtain P D(loaddisplacement) and SN (stress rangefatigue life) relationships; and the perfor-

mance of connections at bridge system level, were reported separately in Davalos et al. [11].

54

J.F. Davalos et al. / Engineering Structures 40 (2012) 5163

4. Test procedures 4.1. Phase I test static behavior of bridge model Three load cases were used for the bridge model (Fig. 1) with shear connection spacings either to be 0.6 m or 1.2 m, as shown in Fig. 9 and Table 3. The model was rst tested with shear connections at 0.6 m intervals. The 1.2 m spacing was achieved by removing the top sleeves of every other shear connections. Load case 1 (LC1, Fig. 9) was designed to study the transverse load distribution factor. A concentrate load was applied over an area of 0.6 m by 0.25 m, using a 245 kN actuator to simulate truck wheel load, at mid-span and over the middle girder, as shown in Fig. 9. The bridge was loaded to 50% service limit load by displacement control rate of 1 mm/min. The 50% service load was based on the level calculated for the reference full-scale bridge. Based on AASHTO Specication [14], the service stress limit for Service II load combination can be calculated as:

Fig. 3. KSCI honeycomb FRP panel.

ff 0:8Rh F yf

Fig. 4. Facesheet lay-up.

Table 1 Properties of constituent materials. Material E-glass liber Polyester resin E, GPa (106 psi) 72.4(10.5) 5.06(0.734) G, GPa (106 psi) 28.8 (4.IS) 1.63(0.237)

where Rh is the hybrid factor, which is 1.0 for the homogenous section used in this model; and Fyf is the yield strength of the ange. Therefore, the service stress limit ff was 275 MPa (0.8Fyf) for Grade 50 (Fyf = 344 MPa) steel girders used in this model, and the model was loaded until the ange stress reached 138 MPa (0.5ff), corresponding to a 178 kN patch load. Load cases 2 and 3 were used to study the local deection of the FRP deck (Fig. 9). A 36 kN patch load, corresponding to a rear wheel-load of an HS20 truck, was respectively applied at the mid-point between girders 1 and 2 and girders 2 and 3 (Fig. 9). Since load cases 2 and 3 were symmetrical, the local deection of the deck was calculated as the average values of these two load cases. Transverse deection prole was obtained by the measurements from ve LVDTs across the mid-span section, as shown in Fig. 10, where LVDTs 13 were placed directly under girders 13, respectively; and LVDTs 4 and 5 were placed under the FRP deck at mid-span. The local deection was dened as the relative displacement of the panel between the two supporting girders by linear interpolation.
m
0.255 0.30

q, g/cm3
(Ib/in.3) 2.55(0.092) 1.14(0.041)

4.2. Phase II test fatigue behavior of bridge model Similar to the static test, a patch load was applied at the midspan of the middle girder. The load range was 0112 kN. The model was subjected to 10.5 million cycles loading, which was equivalent to 75 years bridge service life span [24]. The fatigue load was calculated as comparable to the corresponding design load for a reference full-scale bridge, where the stress range was 3.3 MPa for induced shear stress at each shear connection [15], which corresponded to a cyclic load of 025.1 kips at the mid-span of the center girder. The loading was stopped at every 2 million cycles and the deection of the model was measured to calculate the stiffness. 4.3. Phase III test T-beam behavior

Table 2 Material properties of facesheet. Nominal weight (kg/m2) CDM 3208 0 90 ChSM 0.531 0.601 0.256 Thickness (mm) 0.49 0.55 0.25 1.91 Vf 0.424 0.425 0.396 0.188

Bonding layer ChSM 0.915

3.2. T-section model description After completing the tests on the scaled bridge model, a Tsection was cut-out with a ange width of 1.22 m, as illustrated in Figs. 2d and 8, which was supported by a steel girder. Three brackets were placed on each side of the ange to provide lateral support to the ange section.

A patch load was applied at the mid-span of the T-beam over an area of 0.6 0.25 m2 using a 490 kN actuator to study its effective ange width, as shown in Fig. 8. The system was subjected to three-point bending loading with displacement control at a rate of 1 mm/min within the service load limit, and loaddisplacement relation was recorded. A total of 20 strain gauges were attached at top and bottom surfaces of the deck, with 10 at quarter-span and

J.F. Davalos et al. / Engineering Structures 40 (2012) 5163

55

Fig. 5. Tongue and groove connection with FRP sheet covered.

Fig. 6. Tongue-and-groove connections.

Fig. 7. Installations of shear connection to FRP decks.

10 at mid-span, as shown in Fig. 11, to measure the longitudinal normal strain of the FRP ange and evaluate effective ange width. In addition, strain gauges were bonded along the depth of the girder to measure the longitudinal normal strains, as shown in Fig. 12. Based on the strain distributions, the neutral axis of the system can be determined, and subsequently, the degree of composite action can be calculated, as will be shown in the next section.

e Mi ESi ei ei wi S i LDF PN PN PNl S PN j M ES e e j j j j j 1 j 1 j1 ej wj j1 S

l

Sj

5. Test results 5.1. Phase I test results of bridge model 5.1.1. Load distribution factor As suggested by Eom and Nowak [16], the expression

is adopted to obtain load distribution factor (LDF), where Mi the bending moment at the ith girder; E the modulus of elasticity; Si the section modulus of the ith girder; Sl the typical interior section modulus; ei the maximum bottom-ange static strain of the ith girder, which can be obtained from the test results; wi the ratio of the section modulus of the ith girder to that of a typical interior girder; and N is the number of girders. If the girders have the same section modulus, which is the case in this study, Eq. (2) can be simplied as

LDF PN

ei e

j 1 j

where k is the number of lanes loaded.

56

J.F. Davalos et al. / Engineering Structures 40 (2012) 5163

Fig. 10. Instrumentation of bridge model.

Fig. 8. T-beam test model.

respectively, for the 0.6 m and 1.2 m connection spacings, as shown in Table 4. The values calculated from AASHTO Standard Specication [13] and AASHTO LRFD Specications [4] are also listed in Table 4. It can be seen from Table 4 that the load distribution factor from the test result with connection spacing of 0.6 m is 1.2% and 12.3% higher than values from AASHTO LRFD and AASHTO Standard Specications, respectively. Although the two specications do not consider bridges with partial degree of composite action, it seems that they can still provide reasonable accuracy in predicting the load distribution factor for a bridge system that has a low degree of composite action. However, the difference will be larger for higher degree of composite actions. More accurate results can be obtained using a series solution, which is presented in a separate study [25]. 5.1.2. Local deck deection As shown in Fig. 13, the induced deection prole clearly displays the localized effect, where the largest deection occurs at the loading position. The local deections are calculated to be 1.65 mm (L/727) and 1.75 mm (L/686) for the connection spacings of 0.6 m and 1.2 m, respectively, as shown in Table 5. In AASHTO LRFD Specications [14], there is no deection limit for FRP bridge decks, although there is a provision for orthotropic bridge deck, which is intended for steel orthotropic deck with ribs. However, some researchers suggested L/400 as the deection limit [18,19]. Therefore, the local deection for this FRP panel is considered to be acceptable. 5.2. Phase II test results of bridge model 5.2.1. Fatigue resistance No apparent damage on the shear connection and the bridge system was observed for the static test. For the fatigue test, the stiffness remained nearly constant throughout the loading history, as shown in Fig. 14, which was not surprising as the applied stress range was much lower than the threshold value obtained from the single connector push-out test of 53 MPa [11]. The test results indicated that the shear connection and the bridge system can meet both strength and fatigue requirements established by the AASHTO LRFD Specications [14].
Label LC1 LC1 LC2 LC2 LC3 LC3

Fig. 9. Load cases on scaled bridge model.

Table 3 Load case designation. Load case 1 1 2 2 3 3 Transverse load position Aligned Aligned Aligned Aligned Aligned Aligned with with with with with with girder 2 girder 2 mid-point mid-point mid-point mid-point Connection spacing (m) 0.6 1.2 0.6 1.2 0.6 1.2

5.3. Phase III test results of T-beam 5.3.1. Degree of composite action Based on the strain distributions from the test, with one example shown in Fig. 15 at 100% service load, the neutral axis of the system, which determines the degree of composite action, can be plotted as in Fig. 16. Subsequently, the degree of composite action (DCA) can be calculated as [20]:

of of of of

girders girders girders girders

1,2 1,2 2,3 2,3

Based on Eq. (3), the load distribution factors under the service load for the middle girder were calculated to be 0.647 and 0.644

DCA%

Np N0 100 N 100 N0

J.F. Davalos et al. / Engineering Structures 40 (2012) 5163

57

Fig. 11. Instrumentation of T-beam FRP deck section.

Fig. 12. Instrumentation of T-beam girder (see Fig. 11 for locations of cross sections).

Table 4 Load distribution factor (LDF) of test model. LDF from AASHTO Specications Test results 0.6 m connection spacing LDF 0.647 AASHTO Standard AASHTO LRFD 0.727 0.655 Diff. (%) 12.3 1.2 1.2 m connection spacing LDF 0.644 Diff. (%) 12.9 1.7

where N0 and N100 are predicted neutral axes corresponding respectively to non- and full-composite actions, which can be calculated respectively as 202 mm and 256 mm [15]; and Np is the neutral axis for partial composite action, which is about 213218 mm from Fig. 16. Thus, the degree of composite action of the system can be calculated to be about 25%. It is interesting to observe that the effect of connection spacing is only marginal for both 0.6 m and 1.2 m spacings, as the neutral axes are very close as shown in Fig. 16. Further study is necessary to evaluate the effect of spacing on the degree of composite action.

58

J.F. Davalos et al. / Engineering Structures 40 (2012) 5163

Fig. 16. Neutral axis position. Fig. 13. Deection prole corresponding to load case 2 (see Fig. 9 for load case denition).

Table 5 Deection prole of test model. LVDT # (see Fig. 9) 0.6 m connection spacing 1 4 2 5 3 1.2 m connection spacing 1 4 2 5 3 Deection (mm) 1.91 3.31 1.45 0.20 0.076 1.93 3.40 1.40 0.20 0.10 Local deection (mm)

1.65

1.75

5.3.2. Effective ange width The strain proles at mid-span are plotted for both top and bottom surfaces in Figs. 17 and 18, respectively. These strain values are curve-tted by approximate functions. In bridge engineering, the three-dimensional behavior of a bridge system is usually reduced to the analysis of a T-beam section with a reduced width of deck in relation to center-to-center spacing of stringers, over which the longitudinal normal stresses are assumed to be uniformly distributed, which is termed as effective ange width. Based on denition of effective ange width, the longitudinal normal stress is assumed to be uniformly distributed along the panel section. The effective ange width can be expressed as the integral of normal stress distribution divided by the maximum stress rmax at the panel-stiffener intersection as

R b =2 be

b=2

rx dx

rmax

R b =2
0

rx dx rmax

The average normal stress, rx, which can be assumed uniform along its segmentally discrete width h, can be obtained from the results of the strain gages. The integral of the in-plane normal stresses, i.e., normal stress parallel to the deck, can be approximated as the summation of the discrete values over the panel section, which can be expressed as

b =2

rx dx

n P i1

rx;i hi

b=2

Fig. 14. Stiffness ratio variations during fatigue test.

where rx,i is the normal stress calculated from the normal strain at strain gage i, rmax is the stress value at the panel-girder center line, hi is the width for each block, and n is the total number of the blocks. Based on Eqs. (5) and (6), the effective ange width was calculated to be about 0.63 m, which is about 50% of the actual ange width. Using a shear lag model formulation [21], be can be calculated as:

-2787

500
-1207

R b =2 be
1677

T-section Depth (mm)

400 300 -413 200 100 0

540

cosh

coshn1 ydy  
n1 b 2

Beam Section Slab Section

1333

q A11 where n1 p , A11 and A66 are stiffness of FRP panel, a is the a A66 length of the panel along the girder direction, and b is the width of the panel. The effective ange width is predicted to be 1.01 m for full composite action. Therefore, a Reduction Factor R is suggested in order to account for the effect of partial degree of composite action. Accordingly, we can dene be, for partial degree of composite action, as

-3000

-2000

-1000

0 Micro Strain

1000

2000

3000

be Rb100

Fig. 15. Axial strain distribution at mid-span cross section B (see Fig. 11 for locations of cross sections).

where b100 is the effective ange width for full composite action. Details of the formulation of R and how to incorporate R in the design are reported in Chen et al. [17]

J.F. Davalos et al. / Engineering Structures 40 (2012) 5163

59

Fig. 17. Normal strain distribution on top facesheet of FRP panel.

Fig. 18. Normal strain distribution on bottom facesheet of FRP panel.

5.3.3. Failure mode at strength limit The T-section displayed nearly linear elastic behavior until the load reached about 220 kN, and then the loaddeection curve displayed a shallow nonlinear behavior, as shown in Fig. 19. The maximum load was about 356 kN. The T-section failed by local buckling of the ange of the steel girder, as shown in Fig. 20, which

indicated that the FRP deck cannot brace the steel girder as effectively as a concrete deck. This also corroborates the nding of 25% partial composite action as described before.

Fig. 19. Load deection curve.

Fig. 20. T-section prior to failure.

60

J.F. Davalos et al. / Engineering Structures 40 (2012) 5163 Table 6 Equivalent properties of FRP panel. Ex (MPa) In-plane Bending 2560 5640 Ey (MPa) 2300 5640

6. Finite Element model An FE model was created to simulate the tests described above using ABAQUS [22], as shown in Fig. 21. The model consisted of three parts, steel stringers, FRP deck, and shear connections. The steel was modeled as a linear elastic isotropic material with the modulus of elasticity equal to 200,000 MPa and the Poissons ratio equal to 0.3. The elements used consisted of 4-node general-purpose shell elements with reduced integration and hourglass control, and nite membrane strains. These elements in ABAQUS are commonly referred to as S4R elements [22]. The cross bracings were modeled by two node beam elements. As shown in Fig. 4, the FRP deck consisted of top and bottom facesheets, and a core with sinusoidal shape. It is cumbersome to model the actual geometry of the FRP deck. Therefore, the FRP deck was simplied into an equivalent FRP plate. The properties of the FRP plate, as shown in Table 6, were obtained using a stiffness analysis based on homogenization theory [23] and veried against bending tests on discrete FRP panels, as shown in Appendix A. The same S4R elements were used to model the equivalent FRP plate. The most important characteristic of this model is to simulate the interaction between the FRP deck and the steel girder. This was achieved using multiple-point constraint connector elements (CONN3D2 in ABAQUS) among the common nodes over the center contact section between the top ange of the beam and the bottom of the deck, as shown in Fig. 22. Each connector element extended between the center-line of the beam ange thickness and the center-line of the deck thickness. The following boundary conditions for the connector element were specied: (1) All three rotational degrees of freedom are coupled between the beam and the deck (no relative rotation); (2) the vertical displacement in the z-direction is coupled (no relative displacements); (3) elastic displacements are prescribed in the x and y directions, by varying the elastic displacement constants, to simulate partial composite behavior between the deck and the girder, where the stiffness was 1.46 kN/mm based on the testing results from Davalos et al. [11]. The patch load on the physical test specimen consisted of a 600 mm 250 mm area, which was applied as a pressure load over the elements. Pin-roller constraint was used to represent the simply-supported boundary conditions from the test.

mx
0.303 0.303

Gxy (MPa) 560 1400

Fig. 22. Illustration of connector element.

7. FE analysis results Fig. 23 displays the displacement contour of the FRP panel subjected to load case 2 as shown in Fig. 9. The deections are shown in Table 7, where good correlations can be observed between the FE and testing results as shown in Table 5. Using the data reduction techniques as shown in Eqs. (3) and (5), the LDF and effective ange with for the connection spacing at 0.6 m is 0.621 and 0.75 m, respectively, which are close to the testing results of 0.647 and 0.63 m, illustrating the accuracy of the FE model developed herein. The FE model can be used to carry out parametric study.

Fig. 21. FE bridge model.

J.F. Davalos et al. / Engineering Structures 40 (2012) 5163

61

Fig. 23. Deformed shape of panel.

Table 7 Deection prole of bridge model. Deection point 0.6 m connection spacing 1 2 3 4 5 Deection (mm) 1.77 2.81 1.54 0.45 0.12 Local deection (mm)

3.
1.12

4. 8. Conclusions In this study, a one-third scaled bridge model consisting of an FRP sandwich deck attached to three steel girders by a mechanical connector was constructed and tested. A T-beam section was cutout from the scaled bridge to be tested under three-point bending until failure. An FE model was constructed to simulate the tests. Since the practical spacing for the connectors was about 0.6 1.2 m, comparing to conventional concrete deck connection spacing of 0.150.25 m, there is partial degree of composite action between the deck and stringers. In the AASHTO Standard [13] and LRFD [14] Specications, slab-on-girder bridges can be designed with either non-composite or full-composite action, but there is no provision for partial degree of composite action. Therefore, this paper aims to study the effectiveness of the partial composite action on the performance of bridge system. Based on this study, the following conclusions can be drawn: 1. The bridge system and shear connections were able to sustain a cyclic fatigue loading equivalent to 75 year bridge service life span, and showed nearly no degradation for both static and fatigue tests. Therefore, the prototype shear connection used can adequately perform at a bridge-system level and be applicable in practice. About 25% degree of composite action can be achieved with the prototype shear connection. 2. It seems that AASHTO LRFD Specications [14] and AASHTO Standard Specications [13] can be used to predict load

5.

6.

distribution factor for FRP-deck on steel girder bridges with partial degree of composite action, where the former provides better prediction for bridge system considered in this study. Further study needs to be conducted to verify this nding. The effective ange width for bridges with partial degree of composite action (DCA) can be calculated from the effective ange width for full composite action, with the introduction of a reduction factor R, for a specied percent of composite action. The two different connection spacing used in the present model, 0.6 m and 1.2 m, do not have signicant impact on structural behavior or performance, such as load distribution factor or degree of composite action, for the FRP-deck on steel girder system considered in this study. Further study is required to evaluate the effect of connections spacings. The panel local deection ratio is about L/700, which is considered to be acceptable. The failure mode of the bridge system was local buckling of the top ange of the steel girder, while the deck remained relatively intact. Good correlations can be observed between the testing and FE results. The FE model can be used to carry out parametric study.

Acknowledgements We gratefully acknowledge nancial support from the NSF Partnerships for Innovation program and the WVU Research Corporation. We appreciate technical advice and testing samples provided by Dr. Jerry Plunkett of Kansas Structural Composites, Inc., Russell, Kansas. Appendix A. Evaluation of FRP panel properties A.1. Equivalent material properties based on homogenization theory Equivalent properties of the FRP panel were obtained following the method presented in Davalos et al. [23]. First, the layer properties of the facesheet were calculated based on constitute materials

62 Table A1 Stiffness properties of facesheet lamina. E1 (MPa) CDM 3208 CDM 3208 CSM Bonding layer 35,900 17,400 9820 E2 (MPa) 11,100 17,400 9820 G12 (MPa) 2810 6200 3510

J.F. Davalos et al. / Engineering Structures 40 (2012) 5163

G23 (MPa) 3030 6200 3510

m12
0.305 0.406 0.397

m23
0.509 0.406 0.397

concept and mechanics of materials approach, with the results shown in Table A2. Finally, the equivalent 2-D orthotropic properties of the sandwich panel were calculated by modeling it as a three-layer laminated system of top and bottom faces and core using lamination theory, with the results shown in Table A3. A.2. Test setup To evaluate the accuracy of the equivalent properties obtained in Table A3, the three FRP panels, which were used as deck for the scaled bridge, were tested to obtain their stiffness prior to the installation of the scaled bridge. As described earlier, the panels were 1.8 m wide (transverse direction as shown in Fig. 3) and 2.74 m long (longitudinal direction as shown in Fig. 3). As shown in Fig. A1, the panels were supported along the longitudinal direction. A patch load was applied using a 245 kN actuator over an area of 0.6 0.25 m2, as shown in Fig. A1. The panels were loaded up to 45 kN with force control at a loading rate of 9 kN/min at ve different locations, numbered as #1#5, as shown in Fig. A2. For panel #1, 10 uni-directional strain gauges were bonded at the ve locations to measure longitudinal and transverse strains. Five LVDTs were installed below the deck to measure vertical displacements at these locations. For panels #2 and #3, only the displacements at the center of the panel, i.e., at location #3, were measured. A.3. Test results The measured deections and strains are presented in Tables A4 and A5. The units for the displacement and strain are mm/kN and microstrain/kN, respectively. The L represents longitudinal strain and the T represents transverse strain. The transverse strain gage at location 4 failed and no strain data were available. A.4. FE analysis The testing results as shown in Tables A5 and A6 cannot be directly used to evaluate the accuracy of the equivalent properties as shown in Table A3. Therefore, an FE model was constructed to simulate the bending test. Similar to the FE model for the scaled bridge, the FRP panel was modeled as an equivalent FRP plate. S4R shell element, with 2D orthotropic properties as shown in Table A3 was used to model the FRP plate. The pin-roller boundary condition and loading conditions as those used in the test were simulated in the FE model. Since load cases #1 and #2 are symmetric to and load cases #4 and #5, only results for load case #1#3 are reported in Tables A6 and A7 for deections and strains, respectively.

Table A2 Stiffness properties of facesheet and core. Ex (MPa) Facesheet Core 13,600 530 Ey (MPa) 14,100 1.0

mx
0.304 0.431

Gxy (MPa) 3500 0.7

Table A3 Equivalent properties of FRP panel. Ex (MPa) In-plane Bending 2940 6488 Ey (MPa) 2641 6488

mx
0.303 0.303

Gxy (MPa) 648 1600

Fig. A1. Panel test setup.

as shown in Table 1 and the volume fractions from Table 2, with the results shown in Table A1. The stiffness of the facesheet was then predicted using lamination theory, as shown in Table A2. Stiffness evaluation of the core was based on a homogenization

Fig. A2. Loading positions and instrumentations.

J.F. Davalos et al. / Engineering Structures 40 (2012) 5163 Table A4 Test results deections. Test Load case 1 2 3 4 5 Panel 2 Panel 3 Deection (102 mm/kN) 1 (LVDT1) 15.3 14.2 16.7 14.1 12.2 15.0 17.1 2 (LVDT2) 15.4 24.4 18.2 12.4 15.7 18.8 19.6 3 (LVDT3) 15.9 17.5 19.8 17.5 16.1 20.6 21.5 4 (LVDT4) 15.7 12.8 18.3 25.0 16.1 19.4 20.0 5 (LVDT5) 12.6 15.4 16.9 15.3 16.3 17.4 18.4

63

Table A5 Test results longitudinal and transverse strains. Test Load case 1 2 3 4 5 Longitudinal strain (microstrain/kN) 1L 19.5 15.7 13.2 12.4 8.9 2L 13.8 27.1 19.4 13.1 14.6 3L 9.6 16.6 24.1 14.8 9.1 4L 13.2 12.5 19.9 25.4 12.8 5L 8.4 15.2 12.9 11.3 18.9

and strains obtained from the tests are not symmetric with minor differences, due to geometry imperfections and manufacture errors. Therefore, average values of load cases #1 and #5; and load case #2 and #4 are shown in Tables A6 and A7. As can be seen from Tables A6 and A7, the difference between the testing and FE results are consistent for both deections and strains. For deection, the differences between FE and test range from 6.7% to 16.7%, with an average value of 13%. For strain, the difference between test and FE analysis varies from 7% to 20%, with an average value of 13%, except for load case #2, since the strains are small in the transverse direction. To compensate for the differences between the testing and FE model, the equivalent properties as shown in Table A5 are reduced to 87%, as shown in Table 6, which are used in the FE model of the scaled bridge. References
[1] Bakis CE, Bank LC, Brown VL, Cosenza E, Davalos JF, Lesko JJ, et al. Fiber-reinforced polymer composites for construction-state-of-art review. J Compos Construct, ASCE 2002;6(2):7387. [2] Davalos JF, Chen A. Buckling behavior of honeycomb FRP core with partially restrained loaded edges under out-of-plane compression. J Compos Mater 2005;39(16):146585. [3] Chen A, Davalos JF. Transverse shear with skin effect for composite honeycomb sinusoidal core sandwich. J Eng Mech, ASCE 2007;133(3):24756. [4] Chen A, Davalos JF. Strength evaluations of sinusoidal core for FRP sandwich bridge deck panels. Compos Struct 2010;92(7):156173. [5] Keelor DC, Luo Y, Earls CJ, Yulismana W. Service load effective compression ange width in ber reinforced polymer deck systems acting compositely with steel stringers. J Compos Construct 2004;8(4):28997. [6] Keller T, Gurtler H. Composite action and adhesive bond between berreinforced polymer bridge decks and main girders. J Compos Construct 2005;9(4):3608. [7] Keller T, Gurtler H. Quasi-static and fatigue performance of a cellular FRP bridge deck adhesively bonded to steel girders. Compos Struct 2005;70:48496. [8] Keller T, Tirelli T. Fatigue behavior of adhesively connected pultruded GFRP proles. Compos Struct 2004;65:5564. [9] Davalos JF, Barth KE, Qiao P. Development of an innovative connector system for ber reinforced polymer bridge decks to steel girders. NCHRP stage I rep. Washington (DC): National cooperative Highway Research Program; 2000. [10] Righman J, Barth KE, Davalos JF. Development of an efcient connector system for ber reinforced polymer bridge decks to steel girders. J Compos Construct 2004;8(4):27988. [11] Davalos JF, Chen A, Zou B. Stiffness and strength evaluations of a new shear connection system for FRP bridge decks to steel girders. J Compos Construct, ASCE 2010;15(3):44150. [12] Machado1 MAS, Sotelino ED, Liu J. Modeling technique for honeycomb FRP deck bridges via nite elements. J Struct Eng, ASCE 2008;134(4):57280. [13] AASHTO. AASHTO standard specications for highway bridges, 16th ed. Washington (DC): American Association of State Highway and Transportation Ofcials; 1996. [14] AASHTO. AASHTO LRFD bridges design specications, 3rd ed. Washington (DC): American Association of State Highway and Transportation Ofcials; 2004. [15] Zou B. Design guidelines for FRP honeycomb sandwich bridge decks. Ph.D. dissertation, West Virginia University, Morgantown (WV, USA); 2008. [16] Eom J, Nowak AS. Live load distribution for steel girder bridges. J Bridge Eng 2001;6(6):48997. [17] Chen A, Davalos JF, Justice AL, Michaelson GK, Perisetty N. Effective ange width for orthotropic FRP bridge decks with degrees of composite action. J Bridge Eng, ASCE, in press. [18] Demitz JR, Mertz DR, Gillespie JW. Deection requirements for bridges constructed with advanced composite materials. J Bridge Eng 2003;8(2):7383. [19] Zhang Y, Cai CS. Load distribution and dynamic response of multigirder bridges with FRP decks. Eng Struct 2007;29(8):167689. [20] Park KT, Kim SH, Lee YH, Hwang YK. Degree of composite action verication of bolted GFRP bridge deck-to-girder connection system. Compos Struct 2005;72:393400. [21] Zou B, Chen A, Davalos JF, Salim HA. Evaluation of effective ange by shear lag model for orthotropic FRP bridge decks. Compos Struct 2011;93:47482. [22] Dassault Systmes. ABAQUS/CAE Users Manual, Version 6.7. Pawtucket (RI); 2007. [23] Davalos JF, Qiao PZ, Xu XF, Robinson J, Barth KE. Modeling and characterization of ber-reinforced plastic honeycomb sandwich panels for highway bridge applications. Compos Struct 2001;52:44152. [24] Moon FL, Gillespie JW. Experimental validation of a shear stud connection between steel girders and a ber-reinforced polymer deck in the transverse direction. J Compos Construct 2005;9(3):2847. [25] Zou B, Davalos JF, Chen A, Ray I. Evaluation of load distribution factor by series solution for orthotropic bridge decks. J Aerosp Eng, ASCE 2011;24(2):22739.

Transverse strain (microstrain/kN) 1T 1 2 3 4 5 13.9 2.4 6.2 1.7 1.4 2T 2.0 10.4 2.8 3.8 1.2 3T 7.0 3.1 14.8 2.5 6.8 4T N/A N/A N/A N/A N/A 5T 2.5 2.1 6.2 1.7 13.5

Table A6 Deection correlations between test and FE results. Location LC #1 Test 1 2 3 4 5 Diff (%) 15.9 15.5 16.0 15.9 12.4 16.7 FE 14.0 13.6 13.8 13.6 10.2 LC #2 Test 14.2 24.7 17.5 12.6 15.4 6.7 FE 13.6 23.4 16.8 11.9 13.6 LC #3 Test 16.7 18.2 19.8 18.3 16.9 13.9 FE 13.8 16.8 18.2 16.8 13.8 Panel 2 Test 15.0 18.8 20.6 19.4 17.4 15.0 FE 13.8 16.8 18.2 16.8 13.8 Panel 3 Test 11.4 19.6 21.5 20.0 18.4 13.9 FE 13.8 16.8 18.2 16.8 13.8

Table A7 Strain correlations between test and FE results. Strain loc. LC #1 Test 1L 2L 3L 4L 5L Diff (%) 1T 2T 3T 4T 5T Diff (%) 19.2 14.2 9.4 13.0 8.6 7.7 13.7 1.6 6.9 N/A 1.9 7.2 FE 18.3 11.7 12.1 11.7 7.1 12.7 2.2 5.0 2.2 1.7 LC #2 Test 14.0 26.3 15.7 12.8 13.2 12.6 2.1 10.4 5.3 -3.8 1.9 207.3 FE 11.4 24.8 16.1 10.6 11.4 1.2 10.1 0.9 -0.7 1.2 LC #3 Test 13.2 19.4 24.1 19.9 12.9 14.7 6.2 2.8 14.8 N/A 6.2 20.7 FE 12.1 16.3 20.8 16.3 12.1 5.0 2.3 13.0 2.3 5.0

A.5. Correlations between testing and FE results As can be seen from Tables A4 and A5, although load cases #1 and #2 are symmetric to load cases #4 and #5, the deections