Sunteți pe pagina 1din 11

Composite Structures 64 (2004) 285295

www.elsevier.com/locate/compstruct

Plate bending behavior of a pultruded GFRP bridge deck system


Thomas Keller *, Martin Schollmayer
Composite Construction Laboratory CCLab, Swiss Federal Institute of Technology, BP Ecublens, CH-1015 Lausanne, Switzerland

Abstract
The bi-directional plate bending behavior of a pultruded GFRP bridge deck system with orthotropic material and system
properties was investigated by means of two full-scale experiments and numerical modeling. Furthermore, the maximum span of the
deck between two main girders was determined using Eurocode loading and presupposing a maximum deection ratio of span/300.
Compared to an isotropic plate-strip, the bi-directional behavior of the orthotropic deck was not very pronounced. Nevertheless,
selecting experiment specimen with ve proles instead of only three reduced the maximum deections at the serviceability limit
state (SLS) by ca. 50%.
Therefore, since the SLS governs in the design of GFRP bridge decks, the possible contribution of bi-directional plate bending is
not insignicant and should be considered and improved in order to fully exploit the potential of pultruded GFRP bridge decks.
 2003 Elsevier Ltd. All rights reserved.
Keywords: Composite structures; Bridge decks; Pultrusion; GFRP

1. Introduction and objective


Bridge decks are a promising application of ber reinforced polymers (FRP) today. Some of the favorable
characteristics of these decks are a high strength combined with a small dead load; a large tolerance for frost
and de-icing salts; short installation times with minimum trac interference; as well as possible increase in
the live load of existing bridges via replacement of the
heavy concrete decks.
Dierent deck systems were already developed and a
multiplicity of demonstration projects with smaller
spans were realized [1,2]. In principle, two construction
forms can be dierentiated today: deck elements from
pultruded structural shapes, also called proles, that are
glued together and large sized sandwich slabs with different core structures [1].
Most of the bridges constructed to date use pultruded
deck systems. Several systems were already developed,
consisting of dierent basic prole cross-sections: hexagonal plus half-depth trapezoidal sections [1], triangular single or dual-cell sections [35], box sections [6] as
well as trapezoidal dual-cell sections [7].
Today, the uni-directional load-carrying behavior of
these proles in the pultrusion direction is rather well
*

Corresponding author.
E-mail address: thomas.keller@ep.ch (T. Keller).

0263-8223/$ - see front matter  2003 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruct.2003.08.011

investigated and known. In most cases, this knowledge


is based on beam experiments with a maximum of three
bonded proles [2,5,8]. A possible bi-directional plate
bending behavior, however, was neglected until today,
probably because of the awareness of the strong orthotropy of these decks. Yet, despite the probable limited bi-directional action, it seemed to be a worthy
pursuit to investigate this plate bending behavior in view
of the potentially favorable inuences: increase in the
maximum span due to reduction of deections and increase in the resistance to wheel loads.
The plate bending behavior of the most commonly
used deck system up to now, consisting of proles with
trapezoidal dual-cell section (cf. Fig. 1), was investigated
within a research project presented in this paper. Further aims were the determination of the maximum span
of the deck system according to Eurocode loads and the
development of an FE-model for future applications.

2. Pultruded GFRP bridge deck system


The deck system used consists of E-glass ber rovings
embedded in a isophthalic polyester resin. The rovings
are stitched into multi-ply structural fabrics with engineered orientations. The fabrics are combined with additional rovings and mats. Two primary lay-up schemes
were used. The laminate around the cells is biased with

286

T. Keller, M. Schollmayer / Composite Structures 64 (2004) 285295

3. Numerical modeling
3.1. Road trac loads
Road trac loads according to the European Code
EC1 were used [9]. Within the scope of this project, load
model 1 had to be considered. Load model 1 consists of
the following concentrated and uniformly distributed
loads, which approximate most of the actions of the
trac of trucks and cars:

Fig. 1. Schematic of the pultruded GFRP bridge deck.

66% 45 and 34% 0 plies to increase the web shear


stiness; the face sheet laminate, top and bottom, is
biased with 60% 0 and 20% each 90 and 45 to increase the overall stiness of the deck. The resulting
deck properties are listed in Table 1. The total ber
content by weight is 60% [8].
The trapezoidal dual-cell prole is formed through a
pultrusion process. Fiber reinforcements are wetted with
the resin and pulled through heated metal dies which
binds and cures the composite into a precise cross-section. The proles are assembled by bonding with a
structural polyurethane adhesive (cf. Fig. 1).
The deck shows a strongly orthotropic load-carrying
behavior due to the dierent ber architectures (material
orthotropy, cf. Table 1) and the dierent structural behaviors (system orthotropy) in longitudinal and transversal directions. From the structural point of view, the
webs and the face sheets form relatively strong I-beams
in the longitudinal direction, strengthened by the face
sheet reinforcements in the web areas (cf. Fig. 3). In the
transversal direction, however, the deck shows a relatively week Vierendeel behavior due to the cellular
structure. The loads are carried primarily by transversal
bending in the webs and face sheets. The low stiness
of the bonded joints further degrades the transversal
behavior.

(a) Double axle concentrated loads (aQ  Qk load for


each axle), distance between the two axles 1.20 m,
distance between the wheels of one axle 2.00 m, load
patch dimensions 0.4 0.4 m per wheel.
(b) Uniformly distributed loads aq  qk , in function of
the numbers of lanes.
aQ and aq are national adjustment factors. The corresponding and considered values for the serviceability
limit state (SLS) in this project are a concentrated load
of 150 kN per wheel and a distributed load of 9 kN/m2
for lane 1, assuming that the distance between two main
girders corresponds about to one lane. The dynamic
amplication is included. The adjustment factors were
set to aQ aq 1. For the ultimate limit state (ULS),
the SLS loads have to be increased by a load factor of
cF 1:35. The Eurocode does not specify limits for
vertical deections due to road trac loads for the main
girders or for the bridge decks.

3.2. FE-model
For the numerical calculations, a nite element (FE)
bridge deck model consisting of six pultruded proles
was developed with the FEM program ANSYS, cf. Fig.
2. The volume element Solid 186 was used. The adhesively bonded joints between the structural elements
were not modeled. The edges parallel to the proles are
not supported, while the edges perpendicular to the

Table 1
Material properties of the pultruded deck [8]
Parameter

Face sheets

Web walls

EL
ET
EZ
GLT
GLZ
GZT
mLT
mLZ
mZT

21,240 MPa
11,790 MPa
4140 MPa
5580 MPa
600 MPa
600 MPa
0.32
0.3
0.3

17,380 MPa
9650 MPa
4140 MPa
7170 MPa
600 MPa
600 MPa
0.3
0.3
0.3

Fig. 2. ANSYS FE-model with volume element Solid 186.

T. Keller, M. Schollmayer / Composite Structures 64 (2004) 285295

proles are simply supported. The mechanical properties represented in Table 1 were used.
An even number of proles were chosen because this
allows, due to double-symmetry, the modeling of only
one quarter of the deck and, therefore, to minimize the
number of nite elements. The specimen for the experiments, whereas, consisted of ve (specimen S1), respectively three (S2) proles.
The maximum vertical deections and stresses at SLS
were calculated with the FE-model as follows: The FEstructure was loaded, according to load model 1, with a
concentrated wheel load of 150 kN in the center of
gravity of the deck, distributed on a 0.4 0.4 m load
patch, as well as with a uniformly distributed load of 9
kN/m2 . In view of possible thin polymer concrete surfacings, no load dispersal of the concentrated load was
assumed. Only one wheel was used for the loading because it was assumed that the wheels spheres of inuence
do not interact with each other. The resulting maximum
spans for maximum deection ratios of span/300, 400
and 500 are listed in Table 2.
Subsequently, FEM calculations were used to determine the parameters of the physical experiments. The
specimen were subjected to only one concentrated load
at the center of gravity of the deck, distributed on a
0.4 0.4 m load patch. An SLS value for this load was
determined in such a way that the calculated maximum
stresses in the FEM model, subjected to EC1 loads on
the one hand and to the experimental concentrated load
on the other hand, matched.
A reference span of 2.70 m was considered for the
experiments which corresponds to a deection limit of
span/300. This ratio seems reasonable to the authors for
GFRP bridge deck applications in view of a sucient
serviceability. A precondition is that the adherence of
the surfacing to the deck is ensured at this ratio.
According to this procedure, an SLS concentrated
load of 165 kN and a ULS concentrated load of 220 kN
at the reference span of 2.70 m was determined for the
experiments.
The results of the numerical modeling of the behavior
of specimen S1 with ve proles at the SLS are illustrated in Figs. 7 and 8 (deections) as well as Fig. 12
(stresses) and are compared with the measured results.
The dierent number of proles (six proles FEM, ve
proles experiment specimen) was not estimated to be of
signicant inuence.

4. Experimental set-up

Table 2
Maximum spans according to FEM calculations

S2

287

4.1. Specimen
Two specimen S1 and S2, consisting of ve, respectively three pultruded proles were prepared for two
static experiments. The dimensions of the specimen are
reported in Table 3 and the cross-sections are illustrated
in Fig. 3. The specimen were bonded by the manufacturer.
It should be noted that the underside of the specimen
S1 was not exactly plane. Therefore, the specimen was
only supported in the middle region in the unloaded
state. At the edges, a gap between the underside of the
specimen and the elastic strip of up to 2 mm remained.
The width of the gap was carefully measured and the
measured values of the outer displacement transducers
were corrected accordingly.
4.2. Set-up and load equipment
The experiment set-up is illustrated in Fig. 4. The
specimen were simply supported by three concrete
blocks on each side. 200 mm wide steel proles with 50
mm wide and 10 mm thick elastic strips were installed in
between the concrete blocks and the specimen. The
support axis was assumed to be in the middle of the
elastic strip. The possible horizontal displacement of
the steel shapes and the elastic strips on the concrete
blocks allowed for the variation of the span of the
specimen from 1.90 to 2.70 m. The concentrated load
was applied with a hydraulic jack of a capacity of 1000
kN via a steel plate 0.4 0.4 m. Between the steel plate
and the deck a 20 mm thick timber plate of the same
dimensions was placed.
Table 3
Dimensions of the specimen
Dimensions

S1

S2

Number of proles
Width
Length
Height

5
1626 mm

3
1016 mm
2900 mm
195 mm

195

S1

Max. deection ratio

Corresponding span l

l/500
l/400
l/300

1500 mm
1950 mm
2700 mm

195

1626

1016

Fig. 3. Cross-sections of specimen S1 and S2.

288

T. Keller, M. Schollmayer / Composite Structures 64 (2004) 285295

Fig. 6. Arrangement of the load transducers in experiment S1dist2700.

Fig. 4. Experiment set-up.

4.3. Instrumentation and measurements


22 (16) displacement transducers were used in order
to measure the vertical deections of the specimen S1
(S2) on the top and bottom side. Taking advantage of
symmetry, the transducers were mainly concentrated in
one quarter of the specimen. The arrangement of the
transducers is illustrated in Fig. 5 for specimen S1.

42 52
14
41 51 57 66 58 67 60 68
23

13

69

22

24

40 50 55 62 56 63 59 64 61 65
12

17

45

46
19

18

16

1626

53 43 54 44
15
load patch
400 x 400mm

4.4. Experimental program

displacement transducer

upper face

strain gage

2900

31 15

16 35 17

18

19

30 10

11 34 12

13

14

21
24

29
3

25
33
23

28

36 7

11

32 2

9
0

37 5

1626

22

5
27 20
6
26 21
7

In one experiment, on one support side, the steel


prole was replaced by load transducers and load
transducer dummies in order to measure the distribution
of the support reactions along the support line (cf. Figs.
6 and 15). Load transducers were placed under ve webs
on the right side of the support (no. 2529), on the left
side dummies were used (no. d1d4) with the same geometry as the load transducers. For the measurement
and the storage of the data delivered by the transducers,
a HBM UPM 60 amplier was used in connection with a
computer.
In order to measure the strains of the specimen at
dierent locations, 65 strain gages for S1 as well as 19
strain gages for S2 of the types 6/120 LY13 and 6/120
LY1, produced by Hottinger Baldwin Messtechnik
GmbH, were used. The carrier dimensions of these gages
were 6 13 mm, the grids sizes were 2.8 6 mm with an
electric resistance of 120 X. The arrangement of the
gages on the top and bottom side is illustrated in Fig. 5
for specimen S1. The data were registered by a HBM
UPM 100.

bottom face

Fig. 5. Displacement transducers and strain gages of specimen S1.

Table 4 gives an overview of all experiments carried


out within this project. Listed are the appellation of the
experiments, the type of experiment (SLS, USL, failure),
the span, the maximum load, the type of loading (loador displacement-controlled) as well as the loading rate.
In every experiment, the applied loads, the vertical deections and the strains were measured.
With specimen S1, a series of nine bending experiments with dierent spans from 1.90 m up to the reference span of 2.70 m were performed. The specimen were
loaded up to 210 kN, which slightly exceeds the SLS
load level of 165 kN for the reference span. It was assumed that the experiments did not inuence one another due to progressive damages.
Subsequently, at the reference span of 2.70 m, a creep
experiment during 14 h under the SLS load (165 kN)
was performed. Furthermore, the previously mentioned
support reaction distribution experiment was carried
out.

T. Keller, M. Schollmayer / Composite Structures 64 (2004) 285295

289

Table 4
Overview of the experiments performed
Appellation

Type of exp.

Span (mm)

Max. load (kN)

Loading

Rate

S1sls1900
S1sls2000
S1sls2100
S1sls2200
S1sls2300
S1sls2400
S1sls2500
S1sls2600
S1sls2700
S1creep2700
S1dist2700
S1uls2700
S1fail2700
S2fail2700

Bending SLS
Bending SLS
Bending SLS
Bending SLS
Bending SLS
Bending SLS
Bending SLS
Bending SLS
Bending SLS
Creep SLS
Support reaction
Bending ULS
Bending failure
Bending failure

1900
2000
2100
2200
2300
2400
2500
2600
2700
2700
2700
2700
2700
2700

210
210
210
210
210
210
210
210
210
165
165
210
774
796

Load-controlled
Load-controlled
Load-controlled
Load-controlled
Load-controlled
Load-controlled
Load-controlled
Load-controlled
Load-controlled
Load-controlled
Load-controlled
Load-controlled
Displ.-controlled
Displ.-controlled

7.5 kN/min
7.5 kN/min
7.5 kN/min
7.5 kN/min
7.5 kN/min
7.5 kN/min
7.5 kN/min
7.5 kN/min
7.5 kN/min
7.5 kN/min
7.5 kN/min
7.5 kN/min
2 mm/min
2 mm/min

Following, the specimen was subjected to the ULS


load, again at the reference span of 2.70 m. After the
ULS experiment, the specimen was loaded up to failure.
With specimen S2, only a failure experiment was
carried out, once again at the reference span of 2.70 m.

5. Experimental results
5.1. Experiments at SLS load level
The results of the experiments S1sls1900S1sls2700
on specimen S1 are illustrated in Fig. 7. The maximum
deection on the bottom side at 165 kN load as a
function of the span is shown. At this load level, the
behavior remained linear-elastic and no cracks could be
heard. Therefore, the assumption that the experiments

did not inuence one another due to progressive damages was conrmed.
The deection of both specimen at the reference span,
in the transversal direction at mid-span, is shown in
Figs. 810. The dierent behavior of the two specimen is
clearly apparent in Fig. 8. In specimen S1 with ve
proles, the proles on the outside showed almost no
deection. In contrast, all three of the proles of specimen S2 deected. The maximum deection with the
three proles was about 50% higher than with the ve
proles. Furthermore, the assumption in Section 3.2 is
conrmed: The inuence sphere of the wheel load is
about 1.20 m and approximately corresponds to the
distance between the two axles of load model 1.
The distribution of the deections in longitudinal
direction are illustrated in Fig. 11 for specimen S1.
There is no notable dierence in the shape of the distributions of specimen S1 and S2.
0

90

85

80

1
7

FEM

Experiment 165kN

[mm]

75

[mm]

70

65

FEM
S1 @165kN

S2 @165kN

60

4
55

10

50

12

45

S1 transducer no.4
40
1.8

1.9

2.1

2.2

2.3

2.4

2.5

2.6

2.7

1000

2.8

800

600

400

200

200

400

600

800

1000

position [mm]

span b [m]

Fig. 7. Maximum deection of specimen S1 in function of the span at


165 kN load, comparison of FEM calculations and interpolated
measurements (bottom face sheet, exp. S1sls1900S1sls2700).

Fig. 8. Deections in transversal direction at mid-span at SLS load,


comparison of FEM calculations and interpolated measurements
(bottom face sheet, exp. S1sls2700). The numbers indicate the transducers, cf. Fig. 5.

290

T. Keller, M. Schollmayer / Composite Structures 64 (2004) 285295


0

10

10

20

20

30

30

[mm]

[mm]

40

40

165kN (SLS)

165kN (SLS)

220kN (ULS)

50

50

774kN (failure)

774kN (failure)

60

60

14
70
800

220kN (ULS)

13
600

12
400

15

200

15

position [mm]

200

400

600

70

800

1000

500

16

17

18

500

1000

position [mm]

Fig. 11. Measured and interpolated deections in the main longitudinal section of specimen S1, upper face sheet (exp. S1sls2700,
S1uls2700, S1fail2700).
Fig. 9. Measured and interpolated deections in transversal direction
at mid-span of specimen S1, upper face sheet (exp. S1sls2700,
S1uls2700, S1fail2700).

strain [%]

10

20

31.9

SLS 165 kN
FEM

0.1

stress [MPa]

0.15

21.2

0.05

10.6

m]

1000

0.2

50

strain [%]

31.9

0.1

21.2

0.05

200

300

400

1000

500

500

position [m

m]

500

1000

strain [%]

0.8

Fig. 10. Measured and interpolated deections in transversal direction


at mid-span of specimen S2, upper face sheet (all taken from exp.
S2fail2700).

170.0

failure 774 kN

0.6

127.4

0.4

42.5

85.0

0.2

[m

500

500

position [m

m]

500

1000

sit

1000

po

The measured axial strain, as well as the resulting


axial stress distributions in the longitudinal direction are
illustrated in Fig. 12 for specimen S1 and Fig. 13 for S2.
Stresses were calculated from measured strains using the
E-Modules in Table 1. The shape of the stress distribution of specimen S1 corresponds to the distribution of
the vertical deections, the outer proles received far

stress [MPa]

10.6

500

100

position [mm]

100

[m

200

300

io

400

sit

500

po

70

42.5

ULS 220 kN

0.15

60

io

500

[m
position [m

40

sit

500

po

500

0
1000

796kN (failure)

stress [MPa]

220kN (ULS)

30

io

[mm]

165kN (SLS)

Fig. 12. Interpolated distributions of measured axial strains and resulting axial stresses in dierent longitudinal sections of specimen S1,
bottom face sheet (exp. S1sls2700, S1uls2700, S1fail2700). For SLS,
comparison with FEM results (mid & edge sections).

T. Keller, M. Schollmayer / Composite Structures 64 (2004) 285295

de-loading, a reversible behavior was observed: the


creep deformation went back to zero after ca. 30 min.

212.4

165kN (SLS)

0.9

291

191.2

220kN (ULS)
169.9

796kN (failure)

0.7

148.7

0.6

127.4

0.5

106.2

0.4

85.0

0.3

63.7

0.2

42.5

0.1

21.2

5.3. Support reaction experiment

[MPa]

[%]

0.8

0
1500

0
1000

500

500

1000

1500

position [mm]

Fig. 13. Interpolated distributions of measured axial strains and resulting axial stresses in the main longitudinal section of specimen S2,
bottom face sheet (all taken from exp. S2fail2700).

In experiment S1dist2700, the distribution of the


support reactions along the support line was measured
under the webs of specimen S1 on one support side at
SLS load level (cf. Figs. 6 and 15). The measured results
are shown in Fig. 15. The distribution of the reactions
corresponds well to the transversal deection behavior
(cf. Figs. 8 and 15). The webs on the outside of the
specimen carry much less load than the webs in the
center. The eect is accentuated with increasing load.
The measured values of transducer 25 were corrupted by
the inelastic support of the adjacent dummy no. d4,
which attracted too much load.

less load than the center proles. The maximum resulting longitudinal stress in specimen S1 at SLS was 24.6
MPa in the bottom face sheet and )21 MPa in the upper
face sheet. The corresponding value in the bottom face
sheet of S2 was 34.6 MPa and thus about 41% higher
than in S1.
5.2. Creep experiment
A creep experiment (S1creep2700) was performed on
specimen S1 over almost 14 h in order to examine the
long-term behavior of the deck. The duration of the
experiment was certainly too short to make any conclusive observations, though full SLS loads were considered. During the experiment, the initial elastic
deformation increased about 1 mm (ca. 12%, cf. Fig.
14), the increase attened notedly with the time. After

Fig. 15. Support reaction distribution experiment (S1dist2700), arrangement of load transducers and dummies, interpolated experiment
results.

800

S1 upper sheet
S1 bottom sheet
S2 upper sheet
S2 bottom sheet

10
700

600

load [kN]

500

[mm]

400

300

ULS

200

SLS
2

100

1
0

S1 transducer no.4
0

10

12

14

16

10

20

30

[mm]

40

50

60

70

time [hours]

Fig. 14. Creep experiment (S1creep2700): measured time-dependant


deection behavior.

Fig. 16. Failure experiments S1fail2700 and S2fail2700: measured


load-deection behavior of the upper and lower face sheets of specimen S1 and S2.

292

T. Keller, M. Schollmayer / Composite Structures 64 (2004) 285295

5.4. Experiments at ULS load level


For specimen S1, the experiment S1uls2700 was performed. For S2, the results at ULS level from experiment S2fail2700 were considered. The results of the
measured deections are illustrated in Figs. 911, the
results of measured strains and resulting stresses in Fig.
12 for specimen S1 and Fig. 13 for S2. In comparison to
the SLS results shown, the ULS results are only increased by the factor 220 kN/165 kN due to the still
linear-elastic behavior of both specimen.
5.5. Failure experiment specimen S1
The measured load-deection behavior of specimen
S1 up to failure, measured on the upper and lower face
sheet, is shown in Fig. 16. The de-loading phase could
only be measured on the bottom face sheet because the
displacement transducer on the upper side fell o at
specimen failure. Furthermore, Fig. 17 illustrates the
vertical deformation of the web under the load patch
between the displacement transducers on the bottom
(no. 4) and upper (no. 15) face sheet.
The global deection behavior was linear-elastic up to
about 300 kN. At this point the stiness changed
slightly. This change could be traced back to the fact
that the webs began to deform at this point (cf. Fig. 17).
Subsequently, the load patch began to punch into the
upper face sheet (cf. Fig. 18). The stiness remained
almost constant then up to the failure load of 774 kN.
Failure occurred in a brittle manner at a maximum
vertical deection of 51 mm (upper face sheet) through
buckling of the upper face sheet and the webs of the
middle three proles under the load patch (Fig. 19). At
the same time, one of the bonded joints near the load
patch opened and delamination failures between webs
800

700

Fig. 18. Failure experiment S1fail2700: vertical deection of the area


around the loading patch just before failure (punching action).

Fig. 19. Failure experiment S1fail2700: buckling of the upper face


sheet and opening of a bonded joint.

and face sheets occurred (cf. Figs. 20 and 21). Thereby,


the failures formed in the outer prole layer and not in
the bonded joint (Fig. 22). Afterwards, the load dropped
down to 470 kN and could then be slightly increased
once again to 490 kN (Fig. 16), where a second internal,
invisible failure occurred. After de-loading, the vertical
deection went back to ca. 3 mm (cf. Fig. 16). Neither of
the two outside proles were damaged during the experiment.

600

load [kN]

500

400

300

ULS

200

SLS
S1 transd. 4-15
100

0
0. 5

S2 transd. 22-3
0

0.5

1.5

2.5

3.5

vertical web deformation [mm]

Fig. 17. Failure experiments S1fail2700 and S2fail2700: measured


vertical web deformation of specimen S1 and S2.

Fig. 20. Failure experiment S1fail2700: delamination failure in the


bottom web to face sheet connection.

T. Keller, M. Schollmayer / Composite Structures 64 (2004) 285295

293

Fig. 23. Failure experiment S2fail2700: buckling of the free ange at


the right edge just before failure.

Fig. 21. Failure experiment S1fail2700: delamination failure in the


upper web to face sheet connection parallel to the bonded joint (detail
cf. Fig. 22).

the transversal direction up to the failure. A brittle


failure occurred at 796 kN, at a slightly higher load than
specimen S1 (+3%).
The specimen failed through buckling of the upper
face sheet and the webs. All three proles failed in the
same manner and at the same time (cf. Figs. 24 and 25).
The maximum deection at failure was 60 mm (upper
face sheet), thus ca. 18% higher than with specimen S1.
After de-loading, the vertical deection went back to ca.
14 mm (cf. Fig. 16).
The distribution of the vertical deections at the onset of failure is illustrated in Fig. 10. The measured axial
strains and resulting axial stresses in the longitudinal
direction are illustrated in Fig. 13. The gures show
once again the already stated behavior: deections were

Fig. 22. Failure experiment S1fail2700: detail of Fig. 21, failure in the
outer prole layer, not in adhesive or in the interface.

The distributions of the vertical deections at the


onset of failure are illustrated in Figs. 9 and 11. The
measured axial strains and resulting axial stresses in
the longitudinal direction are illustrated in Fig. 12. Both
distributions illustrate very well the previously described
behavior: deections, strains and stresses in the outer
proles were considerably smaller than in the central
area. Maximum axial stresses at failure were 163 MPa in
the bottom face sheet and )99.8 MPa in the upper face
sheet (beside the load patch).
The partial safety factor, calculated as ratio between
failure load (774 kN) and USL load (220 kN), was cM
3:52. The total safety factor amounted to cF  cM
1:35  3:52 4:75.

Fig. 24. Failure experiment S2fail2700: delamination and buckling of


the upper face sheet and of the outer web on the right side.

5.6. Failure experiment specimen S2


The measured load-deection behavior of specimen
S2 up to failure, measured on the upper and lower face
sheet, is also shown in Fig. 16. There is a slighter change
in the stiness at about 400 kN in comparison to specimen S1 at 300 kN. The web of S2 deformed more evenly
from the beginning (cf. Fig. 17). At about 400 kN, the
free ange at the right edge began to buckle visibly (cf.
Fig. 23). The specimen deformed much more evenly in

Fig. 25. Failure experiment S2fail2700: delamination and buckling of


the upper face sheet and of the outer web on the left side.

294

T. Keller, M. Schollmayer / Composite Structures 64 (2004) 285295

more evenly distributed in the transversal direction


compared to S1. Maximum stresses at failure were 194
MPa in the bottom face sheet, thus ca. 19% higher than
in specimen S1.
The partial safety factor, calculated as ratio between
failure load (796 kN) and USL load (220 kN), was
cM 3:62. The total safety factor amounted to cF  cM
1:35  3:62 4:89.

6. Discussion and conclusions


The conducted experiments and the numerical modeling can be interpreted as follows and the following
conclusions can be drawn.
6.1. FE-modeling
Numerically calculated and experimental results
match well. Figs. 7 and 8 (deections) as well as Fig. 12
(stresses) illustrate the comparison between the experiment results of specimen S1 with ve proles and the FEmodel with six proles at SLS load level. The model is
wide enough to consider the plate bending action.
Therefore for this deck type in the future, for any kind of
loadings or load combinations, the developed FE-model
can be used for calculations of global vertical deections
and stresses at the SLS. The failure behavior and the
failure loads cannot be predicted since both depend
mainly on the in-thickness strengths of the laminates and
local stability eects which are not taken into account.
6.2. Plate bending behavior
Looking at the results of the support reaction distribution experiment (Fig. 15) as well as the behavior in
Fig. 8, bi-directional plate bending action for concentrated loads does not appear to be very pronounced due
to the orthotropic material and system behavior and the
adhesively bonded joints. At a span of 2.70 m, only four
proles carry about 90% of the concentrated center load
which corresponds to an eective support length of ca.
1.20 m. A corresponding isotropic plate-strip would
have an eective support length that is about three times
longer. Furthermore, compared to the maximum stress
at failure of 163 MPa (S1), the corresponding stress in a
comparable isotropic plate-strip would only amount to
about 40% of this value.
Nevertheless, selecting experiment specimen with ve
instead of only three proles reduced the maximum
deections at the SLS by ca. 50%. Therefore, since the
SLS is determinant for the design of GFRP bridge decks
(cf. Section 6.4), the possible contribution of bi-directional plate bending is indeed signicant and should be
considered in order to fully exploit the potential of
pultruded GFRP bridge decks. For future experiments

on comparable pultruded bridge decks, it is recommended to use at least ve proles.


6.3. EC1 loading
The maximum inuence sphere of a concentrated
wheel load transverse to the pultrusion direction is ca.
1.20 m for the deck used. It corresponds to the distance
between the two axels of EC1 load model 1 in the
bridges longitudinal direction. Therefore, EC1 loading
can be accurately modeled in experiments through one
concentrated load.
6.4. Design criteria
The total safety factor (ratio failure load/SLS load)
for the deck used is of the magnitude of 4.8. Therefore, it
can be concluded that the SLS governs the global deck
design.
6.5. Maximum deck span
The maximum single span for the deck used subjected
to EC1 loading and at a maximum deection ratio of
span/300 is 2.70 m. This span can be increased by ca.
20% by considering the possible continuous beam eect
over several main girders. The exact amount of increase
can be calculated case by case with the FE-model developed. A maximum deection ratio of span/300 seems
to be acceptable to the authors from the serviceability
point of view, though the ratio is not prescribed in the
Eurocode. A precondition is that the adherence of the
surfacing to the deck is ensured.
6.6. Creep behavior
The creep experiment performed indicates that creep
due to trac loads seems not to be a determinant issue
in the global deck design. However, the inuence of
possible heavy surfacings (thick asphalt layers) must be
investigated.
6.7. Failure behavior
Failure modes were always characterized by local
stability failures (buckling) after initial interlaminar
failures such that the material strength could not be
exploited. The maximum resulting stress at failure in the
experiments was 194 MPa.
Due to the plate bending action and corresponding
smaller axial stresses, it was expected that the failure
load of specimen S1 would exceed that of S2. The failure
load of specimen S2, however, exceeded the failure load
of specimen S1 by about 3%. The corresponding axial
stresses at failure in specimen S2 were about 20% higher
than in S1. This inexpected behavior can be explained as

T. Keller, M. Schollmayer / Composite Structures 64 (2004) 285295

295

follows: due to local bi-directional punching deformation in specimen S1 in the compression zone, corresponding high local curvatures caused high in-thickness
stresses and premature interlaminar failures that led to
earlier buckling of the top face sheet. In specimen S2, the
deformation around the load patch was more even and
uni-directional with less local curvatures, thereby allowing for higher stresses.

Acknowledgements

6.8. Possible improvement of the plate bending behavior

[1] Keller T. Recent all-composite and hybrid ber-reinforced polymer bridges and buildings. Prog Struct Eng Mater 2001;3(2):132
40.
[2] Solomon G, Godwin G. Expanded use of composite deck projects
in USA. Struct Eng Int 2002;12(2):1024.
[3] Brown RT, Zureick AH. Lightweight composite truss section
decking. Marine Struct 2001;14:11532.
[4] Crocker H, Shehata E, Haldane-Wilsone R, Mufti A. Innovative
ber reinforced bridge deck modules. In: Proceedings of 3rd
International Conference on Composites in Infrastructure ICCI02,
San Francisco, 2002. Paper 076 on CD-ROM.
[5] Luke S, Knudsen E, Taljsten B. Advanced composite bridge
decking systemproject ASSET. Struct Eng Int 2002;12(2):769.
[6] Hayes MD, Ohanehi D, Lesko JJ, Cousin TE, Witcher D.
Performance of tube and plate berglass composite bridge deck.
J Comp Construct 2000;4(2):4855.
[7] Cassity P, Richards D, Gillespie J. Composite acting FRP deck and
girder system. Struct Eng Int 2002;12(2):715.
[8] DARPA Final Technical Report. Advanced composites for bridge
infrastructure renewalphase II, Task 16Modular Composite
Bridge, March 2000.
[9] European Committee for Standardization (CEN): Eurocode 1
Actions on Structures, part 2: trac loads on bridges. prEN 1991-2,
2002.

An improvement in the bi-directional plate bending


behavior of pultruded bridge decks is, as already stated,
advantageous and desirable. The maximum span and
the resistance to wheel loads can be increased considerably, with the latter due to the possible reduction of
local punching deformation through increased transversal stiness.
In the case of the deck system used, the most eective
method to improve the stiness transverse to the principal span is to increase the system stiness, as increasing the material stiness through changes in the ber
architecture are limited. The relatively weak Vierendeel
load-carrying mechanism can be improved with less
transversal bending action and more axial force action
by means of a change in the geometry of the webs.
Furthermore, the change from the soft polyurethane
joints to stier epoxy joints may be advantageous and is
worth investigating.

The authors wish to acknowledge the support of


Martin Marietta Composites, Raleigh, USA, the manufacturer and supplier of the DuraSpan 766 bridge deck
system used.
References

S-ar putea să vă placă și