Você está na página 1de 13

Composite Structures 69 (2005) 103115 www.elsevier.

com/locate/compstruct

Computer simulations and parametric studies of GFRP bridge deck systems


Baolin Wan, Dimitris C. Rizos *, Michael F. Petrou, Kent.A. Harries
Department of Civil and Environmental Engineering, University of South Carolina, 300 South Main Street, Columbia, SC 29208, USA Available online 3 August 2004

Abstract This research studies the structural behavior of a GFRP bridge deck system. The study develops FEM models of the GFRP bridge deck system using ANSYS 7.0 software package. Data collected from eld measurements and laboratory testing is used to validate and calibrate the proposed FEM models. Parametric studies through computer simulations are conducted to investigate the characteristics of the system. Four parameters are considered for service load conditions: (a) presence of diaphragms, (b) girder stiness, (c) girder spacing, and (d) composite action. In addition, geometric and material nonlinearities are considered in the investigation of the capacity of the new bridge system. The analysis results are discussed and conclusions on the GFRP deck system are presented. 2004 Elsevier Ltd. All rights reserved.
Keywords: Finite element method; GFRP bridge deck; Load distribution; ANSYS

1. Introduction Bridge decks made of ber reinforced polymers (FRP) have been widely studied and increasingly used in highway bridges, both in new construction and replacement of existing bridge decks. FRP composite materials have a number of advantages including, high specic stiness and specic strength ratios, good fatigue behavior, and corrosion resistance. However, compared to traditional construction materials, such as steel, timber, and concrete, GFRP materials have more complex material properties and structures exhibit distinctive behaviors [1]. Investigations of the behavior of FRP bridge decks have been conducted through laboratory tests on FRP deck components [24], and eld tests on FRP bridges [57]. Field tests are performed under service loads which are signicantly lower than failure loads. Experimental investigations on the other hand are more suitable for strength/capacity assessment studies through destructive testing. However, such tests seldom consider

Corresponding author. Tel.: +1-803-777-6166/8953; fax: +1-803777-0670/8953. E-mail address: rizos@engr.sc.edu (D.C. Rizos). 0263-8223/$ - see front matter 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.compstruct.2004.05.012

the entire structure due to equipment limitations and associated costs. Furthermore, parametric studies in experimental procedures are time consuming and prohibitively expensive. Computer simulations based on advanced methods, such as the nite element method (FEM), are reliable and cost eective alternatives in structural analysis for the study of structural response and performance. FEM procedures have been successfully employed in research studying the performance of FRP bridge decks or their components [1,810]. This study conducts computer simulations within the framework of parametric investigations of the bridge on State Highway 655 (S655) in South Carolina which has a glass ber reinforced polymer (GFRP) composite deck on steel girders. To this end, 3-D models of the bridge are developed suitable for FEM analysis using software ANSYS 7.0. The models and FEM solutions are calibrated and validated through comparisons with experimental tests conducted in the present study and the eld measurements reported in [7]. Subsequently, parametric investigations are conducted by systematically varying the following parameters associated with the behavior of the deck: (a) presence of diaphragms, (b) girder stiness, (c) girder spacing, and (d) composite action. Material nonlinearity and large strains are also considered for investigating the behavior of the bridge under loads

104

B. Wan et al. / Composite Structures 69 (2005) 103115

beyond the design loads and assessing the strength and capacity of the deck system. The following sections introduce the physical model and discuss the development, calibration and validations of the computer models. Subsequently, the details of the parametric investigations are presented and the ndings are discussed.

2.1. Bridge deck and GFRP deck panels The bridge deck consists of Duraspan 766 GFRP composite deck panels supplied by Martin Marietta Composites (MMC). These panels are 11.3 m (37.2 ft) long, 195 mm (7.6 in.) deep and are formed by interlocking elements 305 mm (12 in.) wide. The deck panels are delivered to the site in pre-assembled 3.05 m (10 ft) widths. The longitudinal axis of the panels is oriented perpendicular to the trac and the nal interlocking connections are made on site. Fig. 1(b) shows the major dimensions and average thickness of the cross section geometry of the GFRP panel. 2.2. Bridge girder system The GFRP deck is supported on ve W36 l50 girders spaced at 2.438 m (8 ft) on center. The girders are designed assuming a conventional reinforced concrete deck and noncomposite action, although composite action with the GFRP deck is provided in the real bridge. Consequently, the girders are overdesigned, resulting in, a relatively sti supporting system. At the ends of the bridge, the girders are embedded into the concrete abutment wall and the deck sits at the top of the wall as shown in Fig. 2(a). MC18 42.7 steel channels are used as diaphragms in the bridge and are placed at the third points. 2.3. Connection between GFRP deck and girders The GFRP deck is intended to act compositely with the supporting girders through shear stud connections. Stud pocket openings spaced at 610 mm (2 ft) are provided along the entire length of the span. Three studs are arranged transversely across the girder ange at each pocket (Fig. 2(b)). The studs are installed onto the girders through the stud pocket openings and then the stud pockets are grouted. A 25 mm (1 in.) high strength grout haunch, placed as the pockets are grouted, is provided between the deck and girders. 2.4. Instrumentation and load testing Measurements of total girder deections, relative deck-to-girder deections and strains were taken at the quarter and eighth points of the bridge through surveying procedures, mechanical Draw Wire Transducers (DWT), and strain gauges, as reported in [7]. The instrumentation arrangement is shown in Fig. 2(c). Measurements of the relative deck deections are obtained through DWT attached on sti steel beams spanning the top anges of adjacent girders. Two dump trucks loaded with aggregate are used to apply loads on the bridge in the eld. The dimensions and weights of

2. Physical model The bridge studied in this research is the bridge on State Highway 655 (S655) spanning a Norfolk/Southern railway line near Landrum South Carolina and is shown in Fig. 1(a). The bridge is simply supported, its length is 18.3 m (60 ft), and is designed to carry a two lane trac and has an overall width of 11.3 m (37.2 ft). This bridge is a demonstration structure and is the rst use of GFRP deck panels in the state of South Carolina. Load tests have been performed and eld measurements are reported in the framework of a research program that assesses the performance of the GFRP deck [7]. The details of the bridge components are outlined in the following sections.

Fig. 1. The GFRP bridge system: (a) the bridge, (b) cross section geometry of GFRP panel showing average dimensions.

B. Wan et al. / Composite Structures 69 (2005) 103115

105

Fig. 2. Physical model: (a) integral abutment, (b) stud pocket arrangement to deckgirder connection [7], (c) location of instrumentation placed on underside of bridge [7], and (d) load conguration.

the trucks, as well as the dierent loading congurations are described in detail in [7]. The contact area of each tire was measured to be approximately 230 mm2 (9 in2 .) The rst load test, Load Test 1, was conducted on July 9, 2002 and the second, Load Test 2, was conducted on February 3, 2003. All loading cases reported in [7] were considered for calibration and validation of the computer models. This paper presents one validation study with the load case of two trucks placed symmetrically with respect to the middle girder, as shown in Fig. 2(d). The two axles of the rear tandems are located symmetrically across the mid-span of the bridge, as shown in Fig. 2(d).

models use element type Shell 93 to model the GFRP deck anges and webs and steel girders. This element is a general shell element dened by eight nodes, and accommodates varying thickness and orthotropic material properties. The following sections present a number of modeling aspects related to the FEM analysis of bridge components and loading cases. 3.1. Model of GFRP deck The geometry of the webs connecting the top and bottom face sheets varies along X direction, as shown in Fig. 1(b). In order to t the actual pattern of the deck, a variable element size of the bottom sheets is chosen as 127, 178, 178, and 127 mm (5, 7, 7, and 5 in.) along the X -direction. Similarly, the element size of the top sheets varies as 178, 127, 127, and 178 mm (7, 5, 5, and 7 in.) along X -direction, and a single element is dened on the webs. In the Z direction (transverse), the size of the elements is 152 mm (6 in.) except the elements at the ends of the deck panels which are 178 mm (7 in.). The discretized deck model is shown in Fig. 3(a). The material properties of the GFRP deck are assumed as recommended by Martin Marietta Composites and

3. FEM model ANSYS 7.0 nite element software is used to generate the computer models and perform a FEM analysis in this research. By denition, the global X , Y , and Z axes of the bridge models coincide with the longitudinal, vertical, and transverse direction of the bridge, respectively. The FEM model of the real bridge is considered as the control model in the parametric investigations. All

106

B. Wan et al. / Composite Structures 69 (2005) 103115

Fig. 3. FEM model: (a) deck and girder, (b) stud pockets, (c) cross bracing, (d) end supports, and (e) loading conguration.

North Carolina State University [11] for FE models of the Duraspan GFRP material. The GFRP material is assumed linear elastic to failure. The material is orthotropic and the properties are shown in Table 1. 3.2. Model of steel girders The element size of the girder in the X -direction varies as 127, 178, 178, and 127 mm (5, 7, 7, and 5 in.) in order to conform to the mesh of the GFRP deck. In the Z direction, the size of the elements of the girder anges is 152 mm (6 in.). The web is divided into ve equal size elements in Y direction. The view of the girder model is shown in Fig. 3(a). The material properties of the girders are assumed isotropic with Youngs modulus E 200 GPa (29,000 ksi) and Poissons ratio m 0:32. For nonlinear analysis, an elastic perfectly plastic behavior is

assumed with a yield strength dened, as fY 344:5 MPa (50,000 psi).

3.3. Model of the connection between GFRP deck and girders A 25 mm (1 in.) gap is provided between the bottom of the deck and the top ange of the girder to represent the geometry of the real bridge. The grout haunch is not modeled in this work. It is assumed that the shear studs fully transfer shear forces between deck and girders. Consequently, at each shear stud location, deck and girder nodes with the same X and Z coordinates are coupled in X , Y and Z directions. The model of the connection between GFRP deck and girders is shown in Fig. 3(b).

B. Wan et al. / Composite Structures 69 (2005) 103115 Table 1 Material properties used in the FE model for the FRP components [11] Property Elastic modulus x, GPa (ksi) Elastic modulus z, GPa (ksi) Elastic modulus y , GPa (ksi) Poisson ratio, xy Poisson ratio, yz Poisson ratio, xz Shear modulus xy , GPa (ksi) Shear modulus yz, GPa (ksi) Shear modulus xz, GPa (ksi) Thickness, mm (in.) FCL , MPa (ksi) FTL , MPa (ksi) FCT , MPa (ksi) FTT , MPa (ksi) FSLT , MPa (ksi) Element Top sheet 13.2 (1910) 6.6 (955) 27.8 (4030) 0.28 0.24 0.24 12.8 (1850) 1.3 (185) 1.3 (185) 16.8 (0.66) )261 ()37.88) 261 (37.88) )150 ()21.72) 147 (21.38) 71 (10.25) Webs 17.0 (2460) 6.6 (955) 19.4 (2820) 0.28 0.24 0.24 12.8 (1850) 1.3 (185) 1.3 (185) 11.2 (0.44) )2 15 ()31.25) 221 (32.01) )190 ()27.58) 130 (18.9) 91 (13.16) Bottom sheet 14.9 (2160) 6.6 (955) 27.2 (3940) 0.28 0.24 0.24 12.8 (1850) 1.3 (185) 1.3 (185) 16.8 (0.66) )261 ()37.89) 261 (37.89) )150 ()21.72) 147 (21.38) 71 (10.25)

107

Notations: FCL laminate ultimate strength in longitudinal direction under compression; FTL laminate ultimate strength in longitudinal direction under tension; FCT laminate ultimate strength in transverse direction under compression; FTT laminate ultimate strength in transverse direction under tension; FSLT laminate ultimate strength under shear.

3.4. Model of diaphragms Although the actual diaphragms in the bridge are steel channels, X bracing is chosen to represent the function of the diaphragms. ANSYS Link 10 element is selected to model the X bracing and is assigned the same steel properties as the girders and the area of the element is equal to half that of the actual steel channel diaphragm. The model of the X bracing is shown in Fig. 3(c). 3.5. Model of supports In the actual structure, the steel girders are embedded in the concrete abutment wall over 483 mm (19 in.). The deck also sits on the abutment wall over the same length. In order to model such support conditions, the vertical (Y ) displacements of the bottom anges of the girders and the bottom sheet of the deck are constrained over 483 mm (19 in.) from the end of the bridge. Over the same length, the transverse horizontal (Z ) displacements of the top anges of the girders are constrained. Additionally, the bottom anges of the girders are constrained in longitudinal (X ) and transverse (Z ) directions at one end of the bridge. The model of the end supports is shown in Fig. 3(d). 3.6. Load congurations The front tire load selected in the model is the average of the four front tire loads of the two trucks which is 21,858 N (4912 lb). The rear tire load selected is the average of the eight rear tire loads of the two trucks which is 40,949 N (9202 lb). These loads are distributed

as a uniform pressure over an area of 305 356 mm (12 14 in.) for the front tires and an area of 254 610 mm (10 24 in.) for the rear tires in order to t the mesh of the model. The contact areas approximate the tire contact areas and spreading of the load through the 50 mm (2 in.) asphalt riding surface. The relative distances between tires are also adjusted slightly to t the mesh of the model. The loaded model is shown in Fig. 3(e). The total load applied to the model is 415 kN (93.26 kips).

4. Validation of the model In order to validate and calibrate the FEM model, deection values obtained from the computer simulations are compared to eld measurements and laboratory experiments. 4.1. Comparison with eld measurements The load case and measurements described in Section 2.4 are considered in this comparison study. The deformed conguration of the control model of the bridge is shown in Fig. 4(a). In order to compare the analytical results with eld measurements, deections at the eighth points of the girders need to be monitored. In the case where a FEM mesh node does not exist at the point of interest, the deection at the target point is calculated through interpolation of the deection values at neighboring nodes. Fig. 4(b)(d) show the measured deections from the two load tests and FEM analytical results of girder displacements. The accuracy of the FEM solution is very good for all practical purposes. The relative deck deection is calculated as the dier-

108

B. Wan et al. / Composite Structures 69 (2005) 103115

Fig. 4. Validation of FEM analysis with eld measurements: (a) deformed shape, (b) center girder deection, (c) interior girder deection, and (d) exterior girder deection.

Table 2 Relative deck deections: FEM solutions and eld measurements Position Centerinterior D Interiorexterior D Centerinterior B Model (mm) 0.48 )0.04 0.21 Test 1 (mm) 0.33 )0.20 0.23 Test 2 (mm) 0.38 0.03 0.20

ence between the deection of the deck at the point of interest and a corresponding reference deection. The reference deection is calculated through linear interpolation at the point of interest of the average deections of the center and tip of the top anges of the two adjacent girders. Values of the measured and predicted relative deections are shown in Table 2, where the agreement is good for all practical purposes. 4.2. Comparison with experimental testing The laboratory test specimen uses the same deck system and girder size as the eld bridge and the setup is shown in Fig. 5. The test specimen consists of eight prefabricated deck panels of total specimen length of 2.438 m (8 ft), resting on four girders spaced the same as the eld bridge. The girders are continuously supported on the laboratory oor. A total load of 44.48 kN (10 kips) is applied at the center of the top sheet of the deck over a contact area of 610 356 mm (24 14 in.). The bottom of the deck is heavily instrumented in order to

monitor the deection and strain of the deck. The location of the instrumentation is shown in Fig. 5. The deformed conguration of the computer model of the test specimen is shown in Fig. 6(a). The measured and calculated deck deections are shown in Fig. 6(b). The accuracy of the FEM solution is very good for all practical purposes. It is evident that better accuracy of the deection in the transverse direction is obtained compared to the longitudinal direction. Values of the measured and calculated strains are shown in Fig. 6(c) and (d). Solid and dashed lines represent the longitudinal and transverse direction of the bridge. Squares and triangles represent the longitudinal, ex , and transverse, ez , strain components. Open symbols indicate experimental results and lled symbols indicate computed results. It is observed that the measured transverse strain component in both directions is in very good agreement with the computed values. However, discrepancies are observed in the amplitude of the longitudinal strain components in both directions. These discrepancies are justied as follows. At each monitoring point on the test specimen, one LVR and two strain gages are installed. Therefore, it is not possible to place all these instruments at the same point as assumed in the computer model. In the transverse direction, such small deviations do not aect signicantly the test measurements since the relatively regular geometry does not cause abrupt variations in the stress and strain elds at the bottom of the deck. However, the closely spaced webs of the GFRP deck in

B. Wan et al. / Composite Structures 69 (2005) 103115

109

earity. Subsequently, a series of parametric studies is conducted by systematically varying the main parameters that are expected to aect the behavior of the bridge system, i.e., (a) presence of diaphragms, (b) girder stiness, (c) girder spacing and (d) composite action. These parametric studies are conducted under service load conditions where linear structural behavior is observed. The following sections introduce each study and discuss the ndings. 5.1. Nonlinear analysis Nonlinear analysis, in which both geometric and material nonlinearities are considered, is performed on the control model to nd the capacity of the bridge and identify the load levels below which true linear behavior is exhibited. The steel is assumed to be an elastic-perfectly plastic material with yield stress of 344.5 MPa (50 ksi). The FRP material is assumed to remain elastic until it reaches its ultimate strength which is shown in Table 1. The two-truck loading conguration of the control model is applied. The load amplitude increases gradually until rst failure is observed. At each load step the load increases by two additional truck loads. Each load step is divided into a minimum of 3 and maximum of 20 sub-steps as selected automatically by ANSYS based on the solution convergence. In ANSYS, failure criteria are used to assess the possibility of failure of a material. The maximum stress failure criterion is dened as the maximum of stress ratios of applied stresses over failure stresses at each of the in-plane integration points. All six independent stresses at a point, including tensile or compressive stress in X , Y , Z directions and three shear stresses, are checked. If the value of a stress ratio at a point exceeds 1, the material will fail at that point. The control model uses shell elements in the FRP bridge deck and steel girders, therefore, the out of plane stresses are not considered in the failure criteria. In order to accomplish this in ANSYS, the out of plane failure stress is set to 6.89 107 MPa (1.0 1010 psi) which is extremely high preventing failure in this direction. Failure criteria of the deck and the stress distribution of the center section of center girder are monitored at each sub-step of the analysis. When the applied force reaches the load of 15.1 trucks, representing a total force of 3133 kN (704 kips) on the bridge, the maximum stress ratio is 1.022 indicating failure. This failure takes place at the web of the GFRP deck and at a location of shear studs connecting the center girder and deck. The position of rst failure is shown in Fig. 7. At failure, the maximum in plane shear stress in the web is 92.7 MPa (13.45 ksi) which exceeds the failure shear stress of 91 MPa (13.16 ksi). The maximum stress ratios at the top and bottom sheets of the deck are 0.47 and 0.41, respectively, indicating that

Fig. 5. Test setup of laboratory experiment and placement of instrumentation.

the longitudinal direction may introduce steep gradients in the stress and strain elds. Consequently, measurements are very sensitive to the location of the instrumentation with respect to the webs of the GFRP deck. It is observed in Fig. 6(a) that for the particular support conditions the deck itself is deformed locally indicating a limited ability of the deck to distribute the load.

5. Computer simulations The ANSYS FEM model that was presented in Section 3 and validated in the preceding section is used as the control model in the following computer simulation studies in order to assess the behavior of the bridge system. First, the behavior of the bridge under loads beyond the design loads is considered for assessing the strength and capacity of the deck system and for identifying the load levels that demarcate linear from nonlinear behavior. In this case the development of the control model assumes material and geometric nonlin-

110

B. Wan et al. / Composite Structures 69 (2005) 103115

Fig. 6. Validation of FEM analysis with laboratory data: (a) deformed shape, (b) deck deection, (c) transverse strain, and (d) longitudinal strain.

Fig. 7. Location of rst failure of the web.

Fig. 8. Truck load vs. center girder deection.

the stress in the top and bottom sheets are 47% and 41% of failure stress when web fails in shear. Furthermore, the maximum stress in steel girders is 290 MPa (42 ksi) which is 84% of the yield stress and develops at the bottom ange of the center girder 18 mm (7 in.) from the center of the bridge towards the front tires. Therefore, the initial failure of the FRP bridge will occur in the web of the deck due to insucient shear capacity. Fig. 8 shows the scaled truck loads vs. deection of center girder. The stresses of the steel girders are still in the linear portion of the material when the web of the FRP deck fails because the maximum stress in the girders is lower than the yield stress,. However, when the load reaches 7.15 trucks an apparent nonlinear load deection behavior is observed which is attributed to the ef-

fects of the girder rotation. The nonlinear considerations indicate that the design load is well below critical values that will lead to nonlinear behavior, or failure, of the bridge system, justifying, thus, the linear analysis assumption adopted in the following parametric studies. 5.2. Eect of diaphragms In order to study the eect of diaphragms on the stiness and behavior of the bridge, the diaphragms are removed from the control model while the remaining geometry and loading are the same as in the control model. The deformed conguration is shown in Fig. 9(a). It is observed that the girder rotation in the

B. Wan et al. / Composite Structures 69 (2005) 103115

111

Fig. 9. Eects of diaphragms: (a) deformed congurations, (b) deck relative deection, and (c) center girder deection.

Table 3 Girder bottom ange maximum deection and deck maximum relative deection for dierent parameters Parameters Control model Without diaphragms Girder stiness Girder spacing Noncomposite +10% Size )10% Size 4 Girders 3 Girders Girder bottom ange deection (mm) Center girder 6.56 8.07 5.13 8.53 n.a. 10.68 10.87 Interior girder 5.33 5.97 4.14 6.99 7.35 n.a. 7.79 Exterior girder 2.78 1.50 2.08 3.78 3.56 4.95 1.08 Deck relative deection (mm) Centerinterior 0.48 0.67 0.44 0.54 2.59 5.90 0.87 Interiorexterior )0.04 0.02 )0.05 )0.03 0.48 n.a. 0.09

structure without diaphragms is evident, especially for the external girders. More ecient distribution of the load in the transverse direction of the bridge is observed when diaphragms are considered. The maximum deections of girder bottom anges and maximum relative deections of deck are shown in Table 3. The deck relative deection in the transverse direction is shown on Fig. 9(b) and the center girder deection along the bridge length is shown in Fig. 9(c). The maximum girder deection increases 23% and the maximum deck relative deection increases 40% compared to the control model. This represents a signicant dierence compared to the composite reinforced concrete deck studied in [12] where it is observed that the presence of diaphragms has little eect on the net deection while the load remains small. This may be attributed to the dierent relative rigidity between the girders and the deck of the two systems, and

further investigations are in progress. Nevertheless, it is evident that the diaphragms prevent the rotation of girders, increase the stiness of the entire structure and assist slightly in the load distribution in the transverse direction of the deck. 5.3. Eect of girder stiness The physical dimensions of the girders, including web height, web width, ange height and ange width, are altered by 10%, compared to the control model, to study the eect of girder stiness on the bridge behavior. The corresponding change of the moment of inertia of the girder with respect to its neutral axis is +46.41% and )43.39%, respectively. A third case considers innitely sti girders and corresponds to the case where the deck is continuously supported at the girderdeck interface.

112

B. Wan et al. / Composite Structures 69 (2005) 103115

Fig. 10. Eects of girder stiness: (a) deformed congurations, (b) deck relative deection, and (c) center girder deection.

All other parameters of the control model remain unchanged. Fig. 10(a) shows the deformed congurations of the structures with variable girder stiness. Fig. 10(b) and (c) show the distribution of the relative deck deection in the transverse direction and girder deection in the longitudinal direction, respectively. The maximum deection at the bottom ange of the girder and the deck maximum relative deection values can be found in Table 3. As expected, the maximum deection of the girder and deck are proportional to the girder exibility. However, it is observed that increase of the girder stiness reduces the ability of the deck to distribute the load, leading to undesirable local deck deformations. 5.4. Eect of girder spacing Girder spacing also alters the stiness of the deck supporting system. In order to study these eects, one

and two girders are removed from the control model and the remaining girders are repositioned within the deck width of the control model. All other parameters of the control model remain the same. In the case of a structure with four girders (one girder removed) the two exterior girders are kept at their original locations. The other two girders are rearranged so that the mesh of the top ange of the girder conforms to the mesh of the deck while keeping approximately equal girder spacing. By this arrangement, the space between the exterior girder and the interior girder is 3200 mm (126 in.) and the space between the two interior girders is 3353 mm (132 in.). The girder spacing for 4-girder structure is approximately 33% larger than that in the original model. For the structure with three girders, the two interior girders are removed while the center and the two exterior girders are kept at the original locations. Therefore, the girder spacing for 3-girder structure is 4877 mm (192 in.) which is twice of that of the control model.

B. Wan et al. / Composite Structures 69 (2005) 103115

113

Fig. 11. Eects of girder spacing: (a) deformed congurations, (b) deck relative deection, and (c) center girder deection.

Fig. 11(a) shows the deformed congurations of 4girder and 3-girder structures. For the 4-girder structure, most of the tire loads are applied on the deck between interior girders. It is shown that the deck deformation is mainly in the center area between interior girders for the 4-girder structure. It can be also observed that the loads are distributed over a smaller area of the deck compared to the control model. For the 3-girder structure, two trucks are separated by a center girder. It is evident that each half of the deck supports the truck separately and the deformations cannot be distributed over the center girder. The maximum girder deections and deck relative deections for 4-girder and 3-girder structure can be found in Table 3. The maximum girder deection increases by 12.0% and 62.8% when girder spacing increases 33% and 100%, respectively. As expected, fewer girders, and, therefore, larger girder spacing, decrease the overall bridge stiness. However, the maximum deck relative deections for 4-girder and 3-girder structures are 5.4 and 12.3 times the maximum deections of the control model, respectively. These numbers are signicantly higher than the control model and are comparable to the girder deections. It indicates that the girder conguration plays a key role on the distribution of deformations and that the FRP deck itself cannot distribute the deformations eectively.

5.5. Eect of composite and noncomposite action The last study pertains to the eects of composite action on the distribution of deformations. In this case the coupling at the girderdeck interface at the shear stud pockets and shear stud locations is ignored with the exception of the vertical direction. By this arrangement, the vertical displacement of the deck can be transferred to girders, but no composite action develops for exure. All other parameters for the noncomposite structure remain the same as those of the control model. Fig. 12(a) shows the deformed congurations of the control model and the noncomposite bridge and the maximum value of deections is shown in Table 3. Fig. 12(b) and (c)show the distribution of relative deck deformations in the transverse direction and the girder deformation in the longitudinal direction, respectively. The maximum girder deection and maximum deck relative deection in the noncomposite structure are 65.7% and 81.3% higher than the composite structure, respectively. The structure with composite action shows, as expected, a stier overall behavior. Signicant deck deformation is observed only in the immediate area where the rear tire loads are applied. The noncomposite structure, on the other hand, exhibits distribution of the deformations over a much larger area, yielding a more eective conguration.

114

B. Wan et al. / Composite Structures 69 (2005) 103115

Fig. 12. Eects of composite and noncomposite action: (a) deformed congurations, (b) deck relative deection, and (c) center girder deection.

6. Conclusions This study investigates through computer simulations the main parameters that aect the analysis and design of the bridge on State Highway 655 (S655) in South Carolina which has a glass ber reinforced polymer (GFRP) composite deck on steel girders. To this end, 3D computer models that account for nonlinear behavior are generated suitable for FEM analysis using software ANSYS 7.0. The computer models are validated with in situ measurements and experimental data. Subsequently, a series of parametric studies considered the following parameters for service load conditions: (a) presence of diaphragms, (b) girder stiness, (c) girder spacing, and (d) composite action. The following conclusions can be drawn from this research: 1. The FEM solution is in very good agreement with experimental data of the girder deections and the relative deection of the deck. The measured transverse strain component in both directions of laboratory test specimen is in very good agreement with the computed values. However, the strain measurements are very sensitive to the location of the instrumentation with respect to the webs of the GFRP deck. 2. It is observed in laboratory tests, eld measurements and computer simulations that the deck itself deforms locally when the supporting girder structure is excessively sti. 3. Nonlinear analysis shows that the initial failure of this GFRP bridge will occur in the web of the deck due to insucient shear capacity. The computer models indicate that the bridge fails locally because this kind of FRP bridge deck has a limited ability to dis4.

5. 6.

7.

tribute load when the girders are much stier than the deck. Use of diaphragms increases the stiness of the supporting girder system by limiting girder rotation. However, distribution of deformations on the deck is more eective when diaphragms are omitted. It is observed that the GFRP deck can distribute load better if girder stiness decreases. Fewer girders, or larger girder spacing, also decrease the overall bridge stiness resulting to more eective distribution of the deck deformations. However, a model with three girders shows signicantly increased relative displacements of the deck. It is concluded that girder spacing plays a key role on the performance of the deck. The composite structure can carry higher loads than noncomposite structure with identical girders, however, the noncomposite structure shows a more ecient distribution of deformations on the deck.

The preceding discussion indicates that in order to meet design strength and serviceability demands a good balance between the rigidity of GFRP deck and supporting girder system is required. Such balanced designs will achieve good load distribution on the deck and alleviate local deck deformations that may lead to local deck failure. Specic recommendations for balanced designs are beyond the scope of this work and require extensive research.

Acknowledgements This research was supported by the South Carolina Department of Transportation. The opinions expressed

B. Wan et al. / Composite Structures 69 (2005) 103115

115

are those of the authors and do not reect the opinions of the sponsors. References
[1] Qiao P, Davalos JF, Brown B. A systematic analysis and design approach for single-span FRP deck/stringer bridges. Composites Part B 2000;31:593609. [2] Chandrashekhara K, Nanni A. Experimental testing and modeling of a FRP bridge. Report of the Missouri Department of Transportation, Research, Development and Technology. Report No. RI98-032. 2000. [3] Godwin G, Toddie J, Steggall P. Lewis County, New York, bridge using a low prole 5 deep ber-reinforced polymer composite deck. In: Proceedings of Third International Conference on Composites in Infrastructure, San Francisco. 2002 [CD-ROM]. [4] Kalny O, Peterman RJ, Ramirez G, Cai CS, Meggers D. Evaluation of size eect and wrap strengthening on structural performance of FRP honeycomb panels. In: Proceedings of 82nd Annual Meeting of the Transportation Research Board, Washington DC. 2003 [CD-ROM]. [5] Shenton HW, Chajes MJ. Long-term health monitoring of an advanced polymer composite bridge. In: Proceedings of the

[6]

[7]

[8]

[9]

[10]

[11] [12]

SPIEThe International Society for Optical Engineering, vol. 3671. 1999. p. 14351. Alampalli S, Kunin J. Load testing of an FRP bridge deck on a truss bridge. Transportation Research and Development Bureau, New York State Department of Transportation. Report No. FHWA/NY/SR-01/137, 2001. Turner MK, Harries KA, Petrou MF, Rizos D. In situ structural evaluation of a GFRP bridge deck system. Compos Struct 2004;65(2):15765. Davalos JF, Qiao P, 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. He Y, Aref AJ. An optimization design procedure for ber reinforced polymer web-core sandwich bridge deck systems. Compos Struct 2003;60:18395. Wu H, Mu B, Warnemuende K. Failure analysis of FRP sandwich bus panels by nite element method. Composite Part B 2003;34:518. Martin Marietta Composites. DuraSpan ber reinforced polymer (FRP) bridge deck. Technical Literature. 2002. Zhou S, Rizos DC, Petrou MF. Eects of superstructure exibility on strength of concrete bridge decks. Comput Struct 2004;82(1): 1323.

Você também pode gostar