Quasi-static and fatigue performance of a cellular FRP bridge deck adhesively bonded to steel girders

Quasi-static and fatigue performance of a cellular FRP bridge deck adhesively bonded to steel girders

Composite Structures 70 (2005) 484–496 www.elsevier.com/locate/compstruct Quasi-static and fatigue performance of a cellular FRP bridge deck adhesive...

965KB Sizes 0 Downloads 1 Views

Composite Structures 70 (2005) 484–496 www.elsevier.com/locate/compstruct

Quasi-static and fatigue performance of a cellular FRP bridge deck adhesively bonded to steel girders Thomas Keller *, Herbert Gu¨rtler Composite Construction Laboratory CCLab, Swiss Federal Institute of Technology, BP Ecublens, Lausanne 1015, Switzerland Available online 19 November 2004

Abstract This paper describes the quasi-static and fatigue performance of hybrid bridge girders composed of cellular FRP bridge decks and steel girders. The FRP bridge deck is connected adhesively to the steel girders and acts as the top chord of the hybrid section. Compared to a reference steel girder, the stiffness and quasi-static load-carrying capacity of the hybrid girders were considerably increased due to composite action between the FRP decks and the steel girders. Failure due to quasi-static loading occurred in the FRP decks during yielding of the bottom steel flanges. The adhesive bond between the FRP decks and the steel girders showed no signs of damage due to fatigue loading. The results of the investigation showed that the well-established design method for steel– concrete composite girders with shear stud connections can essentially be used for the design of such FRP-steel girders. The principal modifications necessary for design are proposed.  2004 Elsevier Ltd. All rights reserved. Keywords: Adhesives; Bridges; Bridge decks; Composite action; Hybrid girders; Composite structures; Fatigue; FRP; Pultrusion

1. Introduction Fiber-reinforced polymers (FRP) are becoming a promising alternative construction material for bridge decks [1,2]. Some of the favorable characteristics of these decks are high strength combined with a small dead load; a large tolerance for frost and de-icing salts; short installation times with minimum traffic interference; a possible increase in the live load or deck width of existing bridges via replacement of the heavy concrete decks. In most cases, however, FRP bridge decks must compete with concrete decks. In addition to their transverse load-carrying function, concrete decks usually contribute as part of the top chords of the main girders in the longitudinal axis of the bridge. In this way, the stiffness and load-carrying capacity can be increased several

*

Corresponding author. E-mail address: [email protected]fl.ch (T. Keller).

0263-8223/$ - see front matter  2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.compstruct.2004.09.028

times in comparison with a simple steel or concrete girder. The shear stud or stirrup connections provide full composite action over the cross-section. In order to be a competitive option, FRP decks must also offer a transverse load-carrying and longitudinal top chord function. The transverse load-carrying performance of pultruded FRP bridge decks was investigated in [3]. The investigation showed limited two-dimensional plate bending behavior due to the orthotropic material properties and deck geometry. Particularly in the case of concrete deck replacement, FRP decks must be capable of maintaining the longitudinal top chord function otherwise the main girders must be strengthened. Furthermore, the FRP deck-to-girder connection must provide full composite action between the different parts of the cross-section in a predictable and reliable way. This paper reports on results of a three-year research project funded by the Swiss Federal Roads Authority and industry partners to develop temporary and permanent bridge systems using FRP decks and steel girders for spans up to 50 m. More specifically, adhesively

T. Keller, H. Gu¨rtler / Composite Structures 70 (2005) 484–496

bonded deck-to-girder connections should provide full composite action, which implies a linear strain distribution through the depth of the cross-section. The bridge systems are to have specific design guidelines and recommendations on construction detailing. The design should conform as much as possible to the established design method for steel–concrete composite bridges, in order to facilitate the work of practicing bridge engineers [4]. According to the design method for steel–concrete composite bridges, global girder failure must occur before the failure of the deck-to-girder connection, in addition to the full composite action [5,6]. For compact FRP-steel girders, this means that the full plastic moment must be developed in the girder. In the case of single span bridges, therefore, compression failure of the FRP chord should occur during yielding of the steel girders bottom flange. The paper concentrates on the performance of singlespan hybrid FRP-steel bridges in the longitudinal direction. FRP-steel hybrid girders were investigated analytically and experimentally. The transverse behavior, the behavior of continuous bridges deck chords subjected to tension, as well as the durability of the adhesive bond under environmental loads are also part of the research project, but are not discussed in this paper.

2. Material properties The FRP deck system used in the experimental part of the project was the ASSET bridge deck system used for the 2002 constructed West Mill Bridge, UK [7]. This deck system is composed of pultruded shapes bonded together with two triangular cells, as shown in Fig. 1. The shapes consist of approximately 62% E-glass fibers (volume fraction) and an isophthalic polyester matrix.

485

Table 1 Material properties of ASSET deck (supplier values) Parameter

Face panels (GPa)

Diagonal walls 9.8/7.8 mm (GPa)

Ey Ex

23.0 18.0

16.5/17.3 25.6/22.7

Table 2 Properties of SikaDur 330 epoxy adhesive Type of loading

Failure stress (MPa)

Failure strain (%)

E-Modulus (GPa)

Tension (5 specimens) Compression (5 specimens)

38 ± 2 (failure) 80 ± 3 (maximum)

1.0 ± 0.1 (failure) 3.7 ± 0.1 (maximum)

4.6 ± 0.14 3.1 ± 0.03

Deck properties provided by the deck manufacturer are listed in Table 1 (moduli of elasticity in longitudinal and transversal directions). In the following, the x-axis designates the longitudinal direction of the bridge while the y-axis designates the transverse bridge direction, that is, the pultrusion direction of the deck shapes. The deck chords of the experimental girders were composed of three elements each that were pre-assembled by the deck manufacturer using an epoxy adhesive (Nils Malmgren BPE Lim 465). The surfaces of the bonded deck-to-girder connections were prepared by abrading with a sander until the first appearance of the fiber mats. The steel girders were welded from S355 steel plates according to Eurocode 3 [8]. The mechanical properties from three specimens, measured according to EN 10002, were yield strength 371 ± 10 MPa, ultimate strength 538 ± 5 MPa, modulus of elasticity 210 GPa. The surfaces of the top flanges of the steel girders were sandblasted in order to ensure the adhesion. An epoxy adhesive (SikaDur 330) was used for the deck-to-girder connections and for the joints between the deck elements. The adhesive properties according to ISO 527 and ASTM 695 are summarized in Table 2.

3. In-plane performance of pultruded bridge decks 3.1. Overview

Fig. 1. Pultruded FRP deck system with triangular cell configuration.

FRP bridge decks must fulfill two main performance requirements in order to provide a top chord function. First, the deck must be capable of full shear transmission from the lower to the upper deck face panel to provide full composite action inside the deck and, second, the deck must possess an adequate in-plane compression performance to function effectively as the compression chord in the hybrid girder. In the design of a hybrid girder, therefore, the in-plane shear and compression stiffness, as well as the in-plane shear and compression

486

T. Keller, H. Gu¨rtler / Composite Structures 70 (2005) 484–496

load-carrying capacity of a deck are of primary interest to structural engineers. The term ‘‘in-plane shear’’ is used here to mean shear in the deck plane, similar to the standard concrete deck terminology. In laminate and failure theory of FRP materials the term ‘‘out-ofplane shear’’ would be used for this type of shear. The in-plane deck properties depend not only on the material properties, but also on the cross-sectional geometry and the adhesive joint arrangement in the deck. Since these combined properties are difficult to determine through calculation, an experimental technique for FRP deck specimens was developed to evaluate the structural inplane performance of FRP decks on the full-size or system level. Knowing the in-plane system properties should allow a structural engineer familiar with the design of steel–concrete composite girders to make a preliminary design of a hybrid FRP-steel girder. The two following sections describe how the in-plane system properties of the previously described ASSET bridge deck system were determined. 3.2. In-plane compression performance The in-plane compression stiffness is required for the calculation of the cross-sectional stress distribution and the vertical deflections at the serviceability limit state (SLS) of a hybrid girder. The in-plane compression capacity is needed for the verification of the ultimate limit state (ULS). The experimental set-up for the in-plane compression experiments is shown in Fig. 2. The specimens were composed of four pultruded shapes and cut to a length of 747 mm and a width of 600 mm. Four specimens (1c–4c) were concentrically loaded transverse to the pul-

Fig. 2. Experimental set-up for in-plane compression experiments.

trusion direction along the entire 600 mm width. The load introduction and support of the deck face panels were made by 40 · 40 mm aluminum sections with cutouts in order to fix the face panels in the horizontal direction and to prevent local failure at the points of load introduction. In order to load both deck face panels evenly, thin lead strips minimized inaccuracies of the loading surfaces. The load was applied under displacement-control at a rate of 1.5 mm/min by a Trebel press with a loading capacity of 10,000 kN. The specimens were equipped with eight strain gages at their mid-height to measure the strains in the load and transverse directions on both deck sides. Omega-gages (PI-2-100 from Tokyo Sokki Kenkyujo, Japan) were placed on both face panels over each of the adhesively bonded joints (cf. Fig. 2). The displacements could thus be measured over a length of 50 mm at the joints. The measured load–displacement and calculated axial stress–strain curves of specimens 1c–4c are shown in Fig. 3. The displacements and strains in the load direction are indicated. The axial stresses, r, were calculated by dividing the measured loads by the cross-sectional area of the face panels. The axial strains, e, were calculated by dividing the measured displacements by the specimen height. Table 3 gives average values and standard deviations of the failure loads. A full load transmission from the steel plates to the whole element width occurred after approximately 1 mm of displacement. Subsequently, all specimens showed a linear-elastic behavior up to a brittle failure. The applied load decreased after the failure due to the displacement-control of the experiments (cf. Fig. 3). Only specimen 1c behaved slightly differently, that is, the load could be maintained for approximately 1.5 mm of further deformation, after which it dropped

Fig. 3. In-plane compression: load–displacement and axial stress– strain behavior of specimens 1c–4c.

T. Keller, H. Gu¨rtler / Composite Structures 70 (2005) 484–496

487

Table 3 In-plane compression and shear performance—experimental results Specimen Compression Shear

Failure load (kN) 774 ± 203 273 ± 25

Modulus system (GPa)

E-Modulus strain gages (GPa)

E-Modulus X-gages (GPa)

E: 13.1 ± 2.4 G: 0.047 ± 0.008

18.7 ± 1.5 –

11.3 ± 2.4 –

down. Failures always occurred in one of the deck joints at the locations of the stepped adhesive connection, as shown in Fig. 4. Only in specimen 1c did failure occur in two joints at the same time. At the onset of failure, small de-bonding cracks parallel to the face panels could be observed in the exterior flanges of the stepped joints. Subsequently, the flanges split and buckled at these locations. Failures always occurred in the adherends and never in the adhesives or in the interfaces. Three different transverse elastic compression moduli were determined in the linear-elastic ranges and are given in Table 3 (average values and standard deviations): (a) the elastic system compression modulus calculated from the Trebel press measurements and including the influence of the joints, (b) the elastic compression modulus of the pultruded shapes calculated from the strain gage measurements, and (c) the elastic compression modulus calculated from the omega-gage measurements over the adhesively bonded joints. A large influence of the adhesively bonded joints on the global transverse deck stiffness was observed in that the stiffness over the joints dropped to about 60% of the stiffness between the joints (cf. Table 3). The transverse elastic modulus provided by the deck producer (18 GPa, cf. Table 1) is thereby very similar to the calculated value from the strain gage measurements and neglects the influence of the joints.

Fig. 4. In-plane compression: failure pattern of specimen 4c.

3.3. In-plane shear performance Full composite action in the deck requires a sufficient in-plane shear stiffness of the deck. Furthermore and as previously mentioned, no brittle shear failure should occur in the deck plane (as well as in the adhesive bond) before global ductile girder failure. The in-plane shear experiments used the same experimental set-up as the compression experiments with the exception that the load introduction took place in only one of the two deck face panels of the specimens, and only the opposite face panel was supported. The small horizontal reactions resulting from the eccentricity of the loading and supporting axes were taken by the support angle sections seen in Fig. 2. The same specimen geometry as in the compression experiments was used and four specimens, 1s–4s, were examined. The specimens were equipped with 40 strain gages each on the face panels and diagonals in order to map the transmission of forces from the left into the right deck face panel through the diagonals and to determine the degree of Vierendeel (transverse bending) and truss action. The load was applied under displacement-control at a rate of 1.5 mm/min. The load–displacement curves of the specimens measured by the Trebel press are shown in Fig. 5. The differential displacements (shifts) between the deck face panels in the load direction are shown. The average values and standard deviations of the failure loads are listed in Table 3. The load–displacement curves were

Fig. 5. In-plane shear: load–displacement (differential shift) and shear stress–strain behavior of specimens 1s–4s.

488

T. Keller, H. Gu¨rtler / Composite Structures 70 (2005) 484–496

transformed into shear stress–strain curves, s–c. The shear force was divided by the face panels surface with resulting shear stress whilst the shear strain was calculated from the differential displacement in the load direction and the distance of the deck face panel axes. The corresponding shear stress–strain axes are also shown in Fig. 5. In the linear-elastic range of the shear deformation curve (c from 0.005 to 0.010), a system shear modulus, Gxz, of the FRP deck system was determined and is also given in Table 3. All specimens showed a linear-elastic response until brittle failure. The applied load decreased after failure due to the displacement-control method of load application. Failures always occurred abruptly in one of the truss joints at locations where the tension diagonals were attached, as shown in Fig. 6. At the onset of failure, small cracks perpendicular to the tension diagonals were observed in the joints. Subsequently, one of the tension diagonals pulled away from the truss joint and the unsupported face panel buckled. Deformation of the specimens was then continued at a very low level of load. Further local de-bonding failures occurred in the truss joints and at the upper supports at locations with maximum out-of-plane tension stresses due to transverse bending. The failures always occurred in the adherends and never in the adhesive joints. Using the results of the strain measurements, the axial forces in the deck face panels and webs were determined using the material properties listed in Table 1. The internal forces (axial and shear forces, bending moments) were also calculated using simple structural analysis software. Fig. 7 shows the resulting axial forces, shear forces and bending moment distributions of

Fig. 6. In-plane shear: failure pattern of specimen 3s.

specimen 3s at the onset of failure. From the shear and axial force diagrams, it could be seen that approximately 85% of the load was transferred by truss action (compressive and tensile forces in the diagonals). Less than 15% was transferred by Vierendeel action (transverse bending). However, local moments and high shear forces were calculated in the joints between the intersections of the diagonals with the face panels, which are offset. The results from the strain measurements (gage location cf. Fig. 7) and calculations matched quite well. All specimens showed similar results.

Fig. 7. In-plane shear: axial force, shear force and bending moment diagrams at onset of failure of specimen 3s.

T. Keller, H. Gu¨rtler / Composite Structures 70 (2005) 484–496

489

Table 4 In-plane system properties of the ASSET deck Type of loading

Modulus (GPa)

Compression (4 specimens) Shear (4 specimens)

E = 16.2 G = 0.047

Failure stress (MPa) 41 0.61

3.4. Discussion of in-plane performance For the development of a design method for hybrid FRP-steel bridge girders, characteristic deck system properties for the in-plane shear and compression stiffness and capacity must be known. In the Eurocodes, the characteristic property values are normally reported as 5% fractile values. At this stage in the project, however, reliable fractile property values could not yet be provided since the number of tests performed was too small. Average property values were thus calculated from the experimental results listed in Table 3 as follows: (1) The axial failure stress, rx,u, was calculated from the average specimen failure stress, converted to a 1 m slab width. The elastic system compression modulus, Ex, of 16.2 GPa was estimated from the strain gage and omega-gage values (approximation: 2/3 from strain gages, 1/3 from omega-gages according to the corresponding geometric lengths). Furthermore, an idealized linear-elastic axial stress–strain curve was derived from the measured and calculated results and is shown in Fig. 3. (2) The deck shear failure stress, sxz,u, was calculated from the average specimen failure stress. The system shear modulus, Gxz, from Table 3 was chosen. Again, an idealized linear-elastic shear stress–strain curve was derived and is shown in Fig. 5. The resulting average in-plane compression and shear system properties are listed in Table 4. Based on these values, the experimental FRP-steel girders described in the following paragraph were designed.

Fig. 8. Experimental girder under four-point loading.

• serviceability limit state: maximum deflection span/ 600, • failure mode: ductile global failure mode, deck compression failure during steel yielding, • full composite action over entire FRP-steel cross-section up to failure. The resulting cross-section of the experimental girders is shown in Fig. 9. The decks were 1.50 m wide and 225 mm deep. At this width, it was assumed that the decks would participate fully as part of the top chord in the transverse direction (full width effective). To compensate for tolerances of the upper steel flange, a hard and soft shim arrangement was used for the adhesive bond with a resulting adhesive layer width of 196 mm and a thickness between 6 and 10 mm. From calculations according to the previously mentioned design criteria, the following experimental loads per jack and maximum deflections at mid-span were predicted: 2 · 90 kN and 8 mm at SLS (serviceability limit state), 2 · 120 kN and 11 mm at ULS (ultimate limit state) and 2 · 395 kN at FLS (failure state). From the ratio FLS/ULS, a partial safety factor of cM = 3.3 and from the ratio FLS/SLS, a total safety factor of cF Æ cM = 4.4 was expected. 4.2. Set-up, instrumentation and experimental program

4. Performance of adhesively bonded FRP-steel girders 4.1. Girder design and prediction of performance Two experimental steel girders with adhesively bonded FRP deck top chords were designed and fabricated (cf. Fig. 8). The girders were designed in such a way that the shear stresses in the adhesive bond under four-point loading matched those of a 15 m span reference bridge, subjected to Eurocode 1 loading [9] at the serviceability and ultimate limit state. The design criteria used for the reference FRP-steel bridge corresponded to the design criteria used for steel–concrete composite bridges:

The two experimental girders, designated as Fix 3 and Fix 4, had spans of 7.50 m and were loaded by

Fig. 9. Cross-section of girders with instrumentation.

490

T. Keller, H. Gu¨rtler / Composite Structures 70 (2005) 484–496

Fig. 10. Set-up and instrumentation of girders.

two jacks at a 3.00 m distance from the supports. The experimental set-up is shown in Figs. 8 and 10. Load patches of 0.4 · 0.4 m were used according to Eurocode 1 [9]. Deformable 5 mm thick hard rubber plates were placed between the steel plates of the jacks and the FRP decks. The girders were instrumented with 94 strain gages and nine displacement transducers each. Axial strains were measured at 17 locations in the longitudinal direction with strain gages distributed over the depth of the hybrid girder (on the upper and lower deck face panels as well as on the top and bottom steel flanges). Additionally, at five cross-sections, the axial strain distribution in the upper and lower deck face panels were measured in the transverse direction to determine the effective width of the deck (cf. Figs. 9 and 10). The displacement transducers measured the vertical deflections at 5 points as well as the differential in-plane displacements (shifts) between the upper and lower deck face panels at the ends of the girders. A thermocouple measured the temperature increase in the adhesive layer during the curing of the adhesive in girder Fix 4. The maximum temperature reached 34 C, that was far below the glass transition temperature of the deck resin (approximately 85 C).

All loadings of girders Fix 3 and 4 were carried out under displacement-control and at a rate of 1 mm/min. Girder Fix 3 was first loaded to the SLS load (2 · 90 kN, first cycle). After unloading, the girder was reloaded to the ULS load (2 · 120 kN, second cycle). After unloading, the girder was loaded up to failure (FLS experiment, third cycle). Girder Fix 4 was first loaded up to the SLS load (first cycle). After unloading, the girder was subjected to 10 million sinusoidal load cycles between 2 · 10 kN and 2 · 40 kN at a frequency of 2.0 Hz. The fatigue load range was chosen in order to achieve the same variation of shear stresses in the adhesive bond as the stress variation in the reference bridge subjected to Eurocode 1 fatigue loads [9]. The associated maximum vertical deflection amplitude was 4 mm. An SLS load experiment was carried out at intervals of one million cycles. Following the fatigue experiment, the girder was loaded up to failure (FLS) using displacement-control and at a rate of 1 mm/min. In addition, a reference steel girder without a FRP deck was loaded up to failure in order to compare the load-carrying performance of the hybrid and steel girders.

Table 5 Experimental results at serviceability limit state (SLS, 90 kN per jack) Girder

Fix 3 Fix 4 Reference steel

Maximum mid-span deflection (mm)

Bending stiffness EI (kN m2)

First cycle

After 107 cycles

First cycle

After 107 cycles

8.1 (span/930) 7.9 (span/950) 13.8 (span/540)

– 7.4 (span/1010) –

186,400 188,800 90,500

– 192,800 –

T. Keller, H. Gu¨rtler / Composite Structures 70 (2005) 484–496

4.3. Performance of girders at SLS The first cycle results for girder Fix 3 and Fix 4 and the results for Fix 4 after 10 million cycles up to the SLS loads are given in Table 5. The maximum deflections at mid-span are indicated as well as the girder bending stiffness. The bending stiffness, EI, was calculated at 7 mm of deflection for both girders using simple beam theory and neglecting shear deformation. The behavior of the girders remained linear-elastic during all SLS experiments and was almost identical for both girders (cf. Fig. 11). During the fatigue experiment, the stiffness, EI, of girder Fix 4, calculated from the SLS experiments at intervals of one million cycles, remained constant up to 10 million cycles as shown in Fig. 12. Neither an in450

491

crease in the hysteresis loop area nor a slope decrease or a lateral shift of the loops was observed. Fig. 13 shows measured axial strain distributions through the depth of the mid-span cross-sections of both girders. The measured axial strain distributions of both girders were almost identical, but did not remain totally linear through the depth of the cross-section. The strains in the upper deck face panels dropped slightly from the theoretically expected linear distribution. The fatigue experiment did not noticeably influence the strain distribution in girder Fix 4. From the maximum strains, the corresponding maximum stresses were estimated using the deck elastic modulus in Table 1 and 210 GPa for the steel girders. The maximum stresses in the upper face panels of the FRP decks attained 7 MPa and +90 MPa in the bottom steel flanges (cf. Fig. 13). The increase in the strains at mid-span up to SLS loading during the FLS experiment are illustrated in Fig. 14 for girder Fix 3. Full load transmission was observed in the

400

Load per jack [kN]

350 300 250 200 150

ULS

100

SLS

Fix 3, 3rd cycle Fix 4, 107 cycles Steel girder

50 0

0

20

40

60

80

100

120

140

160

Deflection at midspan [mm]

Fig. 11. Load–deflection behavior of girders at mid-span (FLS experiments).

Fig. 13. Axial strain–stress distributions in mid-span cross-section at SLS load.

450

5

2.5

x 10

400 350

Load per jack [kN]

Stiffness EI [kNm 2]

2

1.5

1

300 250 yielding

yielding 200 150

Fix 4, fatigue experiment: 100

measured at SLS load level 0.5

1

2

3

4

5

6

Load cycles

7

8

9

10 6

SLS

upper face panel lower face panel top flange bottom flange

50 0

0

ULS

Fix 3 at midspan:

1

0.8

0.6

0.4

0.2

0

0.2

0.4

0.6

Axial strain [%]

x 10

Fig. 12. Girder stiffness at one million fatigue cycle intervals.

Fig. 14. Load–strain behavior of girder Fix 3 at mid-span (FLS experiment).

T. Keller, H. Gu¨rtler / Composite Structures 70 (2005) 484–496

adhesive bonds between the top steel flanges and the lower deck face panels: measured top steel flange and lower face panel axial strains are identical (cf. Fig. 14). Small differential in-plane shifts between the upper and lower deck face panels at the ends of the girders were measured: approximately 0.05 mm at SLS loads (cf. Fig. 15). Regarding the effective deck width, Fig. 16 shows the measured axial strains in the upper and lower deck face panel of girder Fix 3 in the transverse direction at midspan. A fitted parabolic approximation is also seen in Fig. 16. When subjected to SLS loads, the deck was almost evenly loaded along its entire width and participated fully as part of the top chord of the hybrid girder. Similar results were seen for girder Fix 4 and it was observed that the fatigue loading did not influence the transverse strain distributions. Fig. 17 illustrates the measured axial strains on the top steel flanges at SLS loads as well as the correspond450 400

Load per jack [kN]

350 300 250 200 150 ULS 100

SLS

50

Fix 3, 3rd cycle 7 Fix 4, 10 cycles

0 0

0.05

0.1

0.15

0.2

0.25

0.3

0.01

0.005 Load

Load

0

Axial strain [%]

492

-0.005 -0.01

-0.015 SLS, top steel flange: Fix 3, 1st cycle st Fix 4, 1 cycle 7 Fix 4, 10 cycles

-0.02

-0.025 0

1

2

3

4

5

6

7

Girder length [m]

Fig. 17. Axial strains on top steel flanges at SLS load and fitted straight lines.

ing fitted straight lines for both girders (correlation factor 0.98). The axial strain values match well the measured values on the lower side of the FRP deck although the dispersion of the latter was slightly more pronounced. Through derivation of the equations of the fitted curves, the longitudinal shear stress distributions in the adhesive bonds were calculated and are represented in Fig. 18. The predicted stress distribution at SLS loading is also shown (0.5 MPa constant value between supports and jacks). The longitudinal shear stresses in the two girders due to first cycle loading and those seen after fatigue loading are very similar. That is, the fatigue experiment did not influence the longitudinal shear stress distribution in the adhesive bond. In general, the shear stresses at SLS loads were close in value to 0.5 MPa and therefore similar to the predicted value.

Differential shift left end [mm]

Fig. 15. Differential shifts between deck face panels at left end of girders.

3

FLS

Shear stress [MPa]

2

Load

Load

1

SLS 0 SLS -1

-2

Adhesive layer: st Fix 3, 1 cycle st Fix 4, 1 cycle Fix 4, 107 cycles Predicted

FLS

-3 0

Fig. 16. Effective width: axial strains and fitted parabolic curves of girder Fix 3 at mid-span.

1

2

3

4

5

6

7

Girder length [m]

Fig. 18. Shear stress distributions in adhesive bond at SLS and FLS.

T. Keller, H. Gu¨rtler / Composite Structures 70 (2005) 484–496

493

4.4. Behavior of girders at ULS The behavior of all experiments remained linear-elastic up to the ULS loading. All measurements (strains and deflections) at ULS loading increased by a factor of 1.35 from the SLS measurements. 4.5. Behavior of girders at FLS The failure experiments were carried out for girder Fix 3 at the third cycle and for girder Fix 4 after 10 million fatigue cycles. Table 6 shows the main results including the ultimate failure loads and the maximum deflections at failure. The behavior of the girders during FLS loading was very similar to the behavior seen during SLS (first cycle, Fix 3 and 4) and ULS (second cycle, Fix 3) loading, in these load ranges. When the ULS load was exceeded, both girders showed almost identical linear-elastic behavior up to approximately 90% of the ultimate failure loads (cf. Fig. 11). Subsequently, the stiffness of the girders began to decrease slightly due to yielding of the bottom steel flanges (cf. Fig. 14 for Fix 3). After the first audible cracks in the decks at approximately 95% of the ultimate failure loads, the girders were unloaded in order to study the unloading and reloading paths and the residual plastic deformations. The different paths were almost identical and parallel to the initial elastic loading paths (cf. Fig. 11). The residual plastic deformations were very small at approximately 3 mm for both girders. After reloading, a first failure occurred abruptly in the deck of both girders. A crack occurred in the short flange of a stepped truss joint of the upper face panel close to one of the load patches, as can be seen in Fig. 19. The epoxy bond at this location remained undamaged. This failure mode matched the failure mode of the in-plane compression experiments on the small-size specimens (cf. Fig. 4). The ultimate failure loads were 373 kN for girder Fix 3 and 400 kN for Fix 4. Almost at the same time, a second failure in the same deck section occurred in the lower face panel (cf. Fig. 20). Again, the short flange of the stepped joint failed and not the adhesive bond. Subsequently, the lower panel detached at this location from the top steel flange and the flange started to deform plastically, as can be seen in Fig. 21.

Fig. 19. First failure of girder Fix 4 in a stepped joint of upper face panel.

Fig. 20. First and second failure in upper and lower deck panel of same section, girder Fix 4.

Fig. 21. Local detachment of deck at location of second failure. Large plastic deformation of top steel flange, girder Fix 4.

Table 6 Experimental results at failure state (FLS) Girder

Fix 3 Fix 4 Reference steel

Ultimate failure load (kN)

Maximum deflection at failure (mm) 7

Third cycle

After 10 cycles

Third cycle

After 107 cycles

373/jack – 263/jack

– 400/jack –

39 (span/192) – 75 (span/100)

– 41 (span/183) –

494

T. Keller, H. Gu¨rtler / Composite Structures 70 (2005) 484–496

During these first and second failures, the load dropped down to a lower load level due to the displacement-control method of loading. The girder could then be deformed further at an almost constant load up to approximately 115 mm deflection for Fix 3 and 100 mm for Fix 4, as can be seen in Fig. 11. At this deflection, a second deck joint failure developed near the second load patch in girder Fix 3. The load dropped down again and continued to decrease. The same failure developed in girder Fix 4, though less abruptly and the load decreased more evenly. The experiments were stopped at approximately 140 mm of deflection, which corresponded to a ratio of span/54. The unloading paths were parallel to that of the reference steel girder (refer to Fig. 11). A large plastic deformation remained after unloading at 93 mm for girder Fix 3 and 105 mm for Fix 4. In Fig. 11 it can be seen that the top flange of the reference steel girder began to yield at approximately 185 kN per jack. At approximately 250 kN per jack, the yielding top flange began to buckle at one location between the jacks, very similar to the buckling shown in Fig. 21. The load then dropped down after having reached a maximum of 263 kN per jack and 75 mm of vertical deflection. The experiment was stopped at 117 mm of deflection at mid-span. The measured axial strain distributions in the crosssections at mid-span for girder Fix 3 at the third cycle and for Fix 4 after 10 million cycles at the onset of failure are shown in Fig. 22. The measured strains in the deck and adhesive bond were the same for both girders. The strains in the bottom steel flange of girder Fix 4 were slightly higher than in Fix 3. Again, the measured strain distributions of both girders did not remain completely linear through the depth of the cross-section. The strains in the upper deck face panels dropped slightly from the theoretically expected linear distribution. The differential in-plane shifts between the upper and lower deck face panels, however, remained small at approximately 0.3 mm at FLS loads (cf. Fig. 15). Again, full load transmission could be observed in the adhesive

bonds between the top steel flanges and the lower deck face panels. The linear-elastic development of the strains at mid-span during the FLS experiments is also illustrated in Fig. 14 for girder Fix 3. The bottom flange began to yield shortly before failure while the top flange had not yet attained the yielding strain (0.18%). After deck failure, however, the top flange also began to yield. The calculated maximum stresses at the onset of failure in the upper face panel of the FRP deck were approximately 33 MPa in both girders. As previously mentioned, the bottom flanges and the bottom part of the webs of the steel girders yielded (371 MPa) while the top flanges attained approximately 230 MPa in compression (cf. Fig. 22). With regard to the effective deck width, the axial strain distributions in the transverse deck direction were less uniform when compared with the SLS results (cf. Fig. 16, results for Fix 3). A decrease in axial strain towards the edge areas of the face panels was observed. The decrease was more pronounced in the lower face panel than in the upper panel. Girder Fix 4 showed similar results after the fatigue experiment. Since the axial strains increased linear-elastically almost up to failure and the top steel flanges did not yield before failure of the deck, the axial strains along the top steel flanges increased proportionally to the corresponding SLS strains in Fig. 17. The longitudinal shear stresses at FLS calculated from measured strains were 1.9 MPa for girder Fix 3 and 2.4 MPa for Fix 4, as shown in Fig. 18. The predicted shear stress distribution is also shown in Fig. 18 (a constant value of 2.4 MPa between supports and jacks, assuming full composite action). Again, the longitudinal shear stresses were low and the fatigue experiment did not influence them noticeably. 4.6. Discussion of girder performance 4.6.1. Composite action The prediction of the performance of the hybrid girders assumed full composite action between all components of the cross-section. As the experiments showed, the adhesive bonds between FRP decks and steel girders were sufficiently stiff and resistant to provide full composite action between the top steel flanges and the lower Table 7 Comparison predicted/measured values

Fig. 22. Axial strain–stress distributions in mid-span cross-section at FLS.

Girder

SLS deflection (mm) pred./ measured (variation)

FLS ultimate failure load (kN) pred./ measured (variation)

FLS deck failure stress (MPa) pred./ measured (variation)

Fix 3

8/8.1 (+1%) 8/7.9 ( 1%)

395/373 per jack ( 6%) 395/400 per jack (+1%)

41/ 33 ( 20%) 41/ 33 ( 20%)

Fix 4

T. Keller, H. Gu¨rtler / Composite Structures 70 (2005) 484–496

deck face panels at all load levels up to failure. The participation of the upper deck face panels, however, was slightly reduced due to the in-plane shear flexibility. Compared to the values of the reference steel girder, deflections at SLS of the hybrid girders were decreased by approximately 50% while failure loads were increased by approximately 46% due to composite action (both percentages are based on average values). The measured maximum deflections at SLS were close to the predicted value (cf. Table 7). The measured ultimate failure loads deviated by 6% and +1% from the predicted failure load for girder Fix 3 and Fix 4, respectively. Due to the only small reduction in composite action of the upper face panel, the variations were within the accuracy of measurements and calculations. It should be noted, nonetheless, that the potential contribution of the FRP deck to the stiffness and capacity of hybrid girders decreases markedly with increasing span due to the relatively low compression stiffness of the FRP deck. Simple calculations show that doubling the span from 15 m to 30 m, for example, reduces the stiffness contribution of the deck from 25% to 9%. This result is generally valid for the FRP bridge deck systems available today. In this respect, future FRP deck generations must be detailed as to be much stiffer in the longitudinal direction of the bridge in order to be a competitive option, particularly in the case of deck replacement. 4.6.2. Effective deck width The decks of both girders fully participated as top chords over the whole 1.50 m widths at the SLS and ULS. At the FLS, however, the face panels showed a decrease in participation towards the deck edges. The decrease was more pronounced in the lower face panel than in the upper one. 4.6.3. Fatigue behavior The intermediate SLS experiments at intervals of one million cycles and the final failure experiments performed on girder Fix 4 showed no indications of degradation or damage to the girder nor to the adhesive bond due to fatigue loading. The comparison to girder Fix 3, which was not subjected to fatigue loading, confirmed this result. 4.6.4. Failure behavior The upper deck face panels failed at lower stresses than predicted. For the hybrid girder experiments a failure stress of 41 MPa was expected, as given in Table 4. The measured stresses of 33 MPa were 20% below this value. The stepped joints, which failed first, were only 60 mm from one of the two load patches. The vertical displacement measurements on the top and bottom of the girder in the jack axis showed that the deck was compressed approximately 1.5 mm under the load patch re-

495

gion at failure. Therefore, a direct influence of the load patch on the stress states in the adjacent joints was probable, which led to the ‘‘premature’’ failure in the joints. Without the premature failure in the upper deck panel, a better comparison of the predicted and measured failure stresses in the deck as well as a slightly higher failure load and a more ductile girder behavior could have been expected. The failure mode in the stepped joint flanges was very similar to the failure mode of the system experiments. The first failures in the upper panels led to a shift of the axial forces into the lower panels, which were subsequently overloaded and failed again in the stepped joints at the same cross-sections as where the first failures had occurred. Compared to the reference steel girder, however, the decks were still able to maintain some compression capacity (cf. Fig. 11). Only with the progression of deck failures did the loads decrease and approach the failure load of the reference steel girder, but at girder deflections approximately twice as big as for the reference steel girder at failure. Fig. 11 also implies possible redundant behavior in the case of unexpected failure of the deck-to-girder bond. The reference steel girder could still carry the SLS and ULS loads. 4.6.5. Adhesive girder-to-deck bond The longitudinal shear stresses in the adhesive steel girder-to-deck bond between supports and jacks were evenly distributed and very small at the investigated load levels, which included the failure load. The shear stresses at failure ranged between 1.9 and 2.4 MPa and were therefore far from the ultimate stresses that can be expected in these adhesive bonds [10]. The absence of peeling stresses made the stress state even more favorable. Although the adhesive layers were relatively thick 6–10 mm, full composite action was achieved and no creep deformation due to fatigue loading occurred. The 10 million fatigue cycles were seen to cause no visible or measurable degradations or damage. 4.6.6. Conceptual design method The hybrid girders were designed using the established design method for steel–concrete composite girders (see Section 4.1). The SLS design criterion was determinant for the girder design. The expected ductile failure behavior (second criterion) was observed, but it was limited by the premature failure in the deck joints caused by the geometric arrangement of the experimental set-up. The partial safety factors (FLS/ULS ratio) was found to be cM = 3.1 for girder Fix 3 and cM = 3.3 for Fix 4 while the total safety factors (FLS/SLS ratio) was found to be cF = 4.1 (Fix 3) and cF = 4.4 (Fix 4) and were close to the predicted values of cM = 3.3 and cF = 4.4. As previously discussed, a slight reduction in composite action was observed only within the deck and not in the adhesive layer (third criterion). It can

496

T. Keller, H. Gu¨rtler / Composite Structures 70 (2005) 484–496

therefore be concluded that the design method for hybrid girders with FRP decks can, in principal, be based on the well-established design method for steel–concrete composite girders while respecting the following main modifications: (1) The in-plane deck properties of compression stiffness and load-carrying capacity of the FRP deck must replace the concrete deck properties. (2) Depending on the in-plane shear stiffness of the FRP deck, the possible reduction of participation of the upper deck face panel must be considered. The in-plane shear capacity must be sufficiently high in order to prevent premature deck failure. (3) The effective width of the FRP deck must be considered. (4) A verification of the shear stud behavior must be replaced by a verification of the adhesive bond behavior.

(3) Established design methods for steel–concrete composite bridges with a few modifications can be used. The serviceability limit state governs the design of adhesively bonded hybrid FRP-steel girders. (4) The proposed in-plane system properties of FRP decks are useful since they allow for the application of established design methods for girders with concrete decks. The system properties comprise the material properties and the effects of the cross-sectional geometry and the joint arrangement. The proposed experimental technique to obtain the system properties could be standardized. (5) The level of composite action in the FRP deck and the effective deck width must be verified. The level of composite action in the deck depends on the inplane shear stiffness of the deck.

Acknowledgments Based on these conceptual aspects and the experimental results presented herein, the development of a detailed design method for hybrid girders consisting of steel girders with adhesively bonded FRP decks, independent of the deck system type, is in progress at the CCLab. In particular, the in-plane shear stiffness-dependant loss of participation of the upper deck panel will be modelled analytically.

The authors wish to acknowledge the support of the Swiss Federal Roads Authority (FEDRO); Fiberline A/S, Kolding, Denmark (supplier of the ASSET bridge deck elements); and Sika AG, Zurich, Switzerland (supplier of the SikaDur 330 epoxy adhesive).

References 5. Conclusions The main conclusions at this stage in the project are (1) Composite action between FRP bridge decks and steel girders can increase the stiffness and reduce the deflections of hybrid girders considerably. However, the contribution of the deck as the top chord of the hybrid system depends on the deck stiffness in the longitudinal direction and decreases with increasing girder span. The contribution of the currently available FRP bridge decks is limited for longer spans and should be improved in future deck generations. (2) The adhesive bond between the FRP bridge deck and steel girders behaved well under quasi-static and fatigue loading. The bond did not fail and showed no damage after 10 million fatigue cycles. An overall ductile failure mode of the hybrid girder system was achieved. The FRP deck failed in compression during yielding of the steel girders. In the case of unexpected bond failure, the hybrid girders behave in a redundant manner in that the steel girders alone could still carry the full dead and live loads.

[1] Keller T. Use of fibre reinforced polymers in bridge construction. Structural Engineering Documents, No. 7, International Association for Bridge and Structural Engineering, ISBN 3-85748-108-0, 2003, p. 131. [2] Reising RMW, Shahrooz BM, Hunt VJ, Neumann AR, Helmicki AJ. Performance comparison on four fiber-reinforced polymer deck panels. J Compos Construct 2004;8(3):265–74. [3] Keller T, Schollmayer M. Plate bending behavior of a pultruded GFRP bridge deck system. Compos Struct 2004;64:285–95. [4] Mertz DR, Chajes MJ, Gillespie JW, Kukich DS, Sabol SA, Hawkins NM, et al. Application of fiber reinforced polymer composites to the highway infrastructure. NCHRP Report 503, Transportation Research Board, Washington, 2003. [5] Eurocode 4. Design of composite steel and concrete structures. Part 2: Rules for bridges. European Committee for Standardization, Brussels, prEN 1994-2, 2002. [6] Iles CD. Design guide for composite highway bridges. New York: Spon Press; 2001. p. 241. [7] Luke S, Canning L, Collins S, Knudsen E, Brown P, Ta¨ljsten B, et al. Advanced composite bridge decking system—project ASSET. Struct Eng Int 2002;12(2):76–9. [8] Eurocode 3. Design of steel structures. Part 1-1: General rules. European Committee for Standardization, Brussels, prEN1993-11, 2002. [9] Eurocode 1. Actions on structures. Part 2: Traffic loads on bridges. European Committee for Standardization, Brussels, prEN 1991-2, 2002. [10] Keller T, de Castro J, Schollmayer M. Adhesively bonded and translucent GFRP sandwich girders. J Compos Constr 2004;8(5): 461–70.