Accepted Manuscript

Structural behaviour of an all-composite road bridge

Tomasz Siwowski, Damian Kaleta, Mateusz Rajchel

PII: S0263-8223(18)30064-3
DOI: https://doi.org/10.1016/j.compstruct.2018.03.042
Reference: COST 9489

To appear in: Composite Structures

Received Date: 5 January 2018

Revised Date: 22 February 2018
Accepted Date: 12 March 2018

Please cite this article as: Siwowski, T., Kaleta, D., Rajchel, M., Structural behaviour of an all-composite road bridge,
Composite Structures (2018), doi: https://doi.org/10.1016/j.compstruct.2018.03.042

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Structural behaviour of an all-composite road bridge

Tomasz Siwowski (corresponding author)

Rzeszow University of Technology
35-959 Rzeszow
Al. Powstancow Warszawy 12
Poland
siwowski@prz.edu.pl

Damian Kaleta
Promost Consulting
35-308 Rzeszow
Ul. Niemierskiego 4
Poland
kaleta@promost.pl

Mateusz Rajchel
Rzeszow University of Technology
35-959 Rzeszow
Al. Powstancow Warszawy 12
Poland
mrajchel@prz.edu.pl

Abstract

Fiber reinforced polymer composites have become an integral part of the bridge industry
because of their versatility, high strength-to-weight ratio and enhanced durability. The novel
idea of an all-composite structural system for road bridges has been proposed for the first time
in Poland. The FRP bridge is a simply supported structure with 10.0 m long span and 7.66 m
wide deck. The superstructure consists of four U-shaped girders bonded with sandwich deck
slab, fabricated by means of a vacuum infusion. The bridge configuration, a finite element
model developed for design and the proof test results are described in this paper. The test has
shown that an all-composite bridge can meet the relevant strength and deflection design
criteria. To develop an understanding of the long-term performance of the FRP bridge, a
monitoring scheme utilizing distributed fibre-optic sensors was implemented to assess any
changes in the bridge structural behaviour.

Keywords

FRP bridge, all-composite superstructure, FEM model, proof test, monitoring, DFOS

Declarations of interest: none

Info: All changes according to review we marked in red.

1
Introduction

Ageing of the highway infrastructure have led to an increasing demand for new technologies
in bridge engineering. Due to their corrosion resistance, high strength and low self-weight,
fibre-reinforced polymers (FRPs) offer promising options. The FRP composites have become
an integral part of the construction industry because of their versatility, high strength-to-
weight ratio, enhanced durability, resistance to fatigue and corrosion, accelerated construction
as well as lower maintenance and life-cycle costs. Advanced FRP composite materials are
emerging for a wide range of civil infrastructure applications. Recently, smaller pedestrian
and road bridges, bridge decks for new and rehabilitated structures, bridge enclosures, repair,
strengthening, seismic retrofitting and other structural applications have been built or
conducted by means of FRP composite components. A very good application of FRP
composites has been proved in case of hundreds of pedestrian bridges. These projects have
benefited from the lightweight nature of composites and the ability to carry components into
remote locations. However, the scope of this paper is limited to road (highway) bridges,
because there exists only very few of all FRP composite bridges that are suited for the heavy
traffic load classes.

The prototype bridges constructed entirely from FRP composites were first conceived in
Europe and China in the late 1970s. The first FRP bridge built in Europe is believed to be the
10 m span bridge constructed at Ginzi, Bulgaria, in 1981/82 [1]. It was a GFRP slab structure
fabricated as one element by the hand lay-up method. The second all-composite bridge to be
constructed was the Miyun bridge in Beijing, China, completed in 1982. The GFRP span with
the total length of 20.7 m consisted of six rectangular sandwich beams with the honeycomb
core, bonded together to create a slab span [2]. Then by the 1990s, interest in FRP as a
construction material accelerated with the increased acceptance and demand for new
technologies in the world. This led to a drop in costs for manufacturing FRP composites and
more cases of FRP bridge construction began to come to light. Although FRP composites
were first introduced into the market in Europe and Asia, the nation with the most funding and
research put into their development has been by far the USA, starting in the mid 1990s [3].

FRP composites have been in service as complete load bearing superstructures on public
roads worldwide for over 20 years. The vast majority of these all-composite bridges is
performing well, although there have been a few instances where they have created problems
for the bridge owner and were taken out of service. The review of selected all-composite road
bridges is presented in Table 1 and 2. These bridges have been fabricated by means of one of
two methods: manufacturing the primary structural elements as customized, project-specific
components specially designed for a particular structure (see Table 1) and combining and
joining the proprietary FRP components, mostly pultruded shapes (see Table 2). Most of all-
composite bridges constructed to date have a slab or a beam & slab scheme of superstructure.
In the latter case the composite action between slab and beams is generally induced by the
shear connectors, both mechanical or bonded. However, there is a one example of the fully
integrated superstructure, having some beams and relevant part of a slab manufactured jointly
in one piece. The total length of most all-composite road bridges is about 10 m, with two
exceptions to date (see Table 1).

The all-composite bridge superstructures deployed nowadays have been made primarily by
means of one of three manufacturing technologies: hand lay-up, vacuum assisted resin
transfer molding (VARTM) or its US variation – Seemann composites resin infusion molding
process (SCRIMP) and pultrusion. Each of these technologies have had particular benefits in

2
individual cases. The all-composite bridge structures have been made of glass fibre (such as
E-glass). Carbon, aramid and other fibres are available but in most instances, glass provides
the best value for the money (i.e. ability to meet established design criteria for the least cost).
In some cases the GFRP elements of superstructure have been strengthened with CFRP strips
or mats bonded to bottom surface to increase the stiffness of the whole span. Vinyl ester resin
is the predominant choice for the matrix that binds the fibres together. It is cost effective, yet
durable over the long run. Lower cost polyesters and more expensive epoxies have also been
used but to a lesser extent. Initially the honeycomb composites and, later, the structural foams
as PVC or PUR as well as balsa wood have been used for cores in sandwich structures.

The main goal of the R&D project carried out by a research consortium, was to develop and
demonstrate the first Polish all FRP composite road bridge, suited for the heavy traffic load
classes. The innovative idea of FRP composite girder-deck structural system for small bridges
has been proposed. This paper describes the demonstrative bridge itself and presents the
research results on its structural behaviour. The finite element model (FEM) analysis and
some results of the research carried out on the bridge system have been briefly presented.
According to experience gained from worldwide applications, the most important issues in
structural behaviour of all-composite bridges are: stiffness, especially when GFRP is used,
dynamic characteristics due to lightweight superstructure and structural health monitoring to
assess the bridge durability. These three performance characteristics of the new all-composite
bridge have been evaluated on the basis of static and dynamic load tests and are presented in
the following chapters. The first Polish all FRP composite bridge was built in late autumn
2016 and is in service since the beginning of 2017. Along with the former demonstrative FRP
bridge project described in [20], the output of this R&D project gives a very promising future
for the FRP composite bridge application in Poland.

Bridge description

The first Polish road bridge fully made of FRP composites is situated in Rzeszow along the
urban road over a small local stream. This is a 10.7 m long single-span simply supported
bridge with 7.7 m wide deck, carrying 2 × 2.5 m wide roadway and two 0.75 m and 1.1 m
wide sidewalks. Its nominal carrying capacity amounts 300 kN according to the Polish bridge
standard. The all-composite bridge superstructure is formed by four FRP composite girders
with an overlying 0.13 m thick FRP sandwich deck slab (Fig. 1a). The deck equipment
consists of two lightweight concrete sidewalk slabs reinforced with GFRP bars and
encompassed by stone curbs and polymer cornice plates, thin insulation and pavement layer,
two expansion joints and steel balustrades. Eight elastomer bearings are used to support the
FRP span on the RC abutments. The solid abutments are placed on 10 micropiles with
diameter of 110 mm and length of 4.0 m (Fig. 1b).

The FRP girders have an U-shaped cross-section with slightly inclined webs, two top 220 mm
wide and 15 mm thick flanges and one bottom 340 mm wide and 15 mm thick flange (Fig.2a).
The maximum width of the girder amounts 1380 mm and the depth is 715 mm. The top and
bottom flanges are made of solid GFRP composites whereas the webs are made in form of the
sandwich panels with PVC foam layer in-between two GFRP laminates. To increase the
torsional stiffness of the FRP girder and to prevent buckling of webs, nine internal
diaphragms are placed and bonded along the length of the girder. The diaphragms are made in
form of 46 mm thick sandwich plates with a structure similar to the webs. The sandwich
bridge deck slab consists of two 12 mm thick GFRP laminates and 105 mm thick PUR foam

3
core stiffened with the internal vertical GFRP ribs (Fig.2b). The deck slab is bonded to the top
flanges of the girders with epoxy adhesive.

The unidirectional and biaxial stitched glass fabrics were used as a reinforcement of the FRP
composite superstructure. The unit weight of glass fabrics ranged from 800 to 1200 g/m2. As a
core material for sandwich parts of girders and deck panels, the PVC foam with density of 80
kg/m3 and the PUR foam with density of 105 kg/m3 were applied respectively. All composite
superstructure was fabricated by the VARTM infusion technology and epoxy resin was used
as a matrix of all composite parts.

To facilitate manufacturing, transport and assembling of the all-composite superstructure it

was divided and fabricated in several parts. Each girder was fabricated entirely in one infusion
process. Due to manufacturer constrains, the sandwich deck slab had to be fabricated in total
12 panels with four various A to D types and dimensions (Fig.3). The individual transverse
and longitudinal adhesive joints were elaborated and tested to enable assembling all panels in
one deck slab, at first in the workshop and finally on site. Two assembling parts of the bridge
superstructure were manufactured in the workshop. The A, B and C panels of one deck side
with two relevant girders were assembled together in the tandem by joining all parts with
epoxy adhesive. After curing adhesive joints two superstructure tandems were transported on
site and rested on bearings, mounted earlier on RC abutments. Finally, the closing D panels
were attached to the deck with the adhesive and bridge equipment was placed on the
superstructure.

Bridge superstructure design

The first step of the all-composite superstructure design was to shape the detailed cross-
sections of FRP girder and deck slab assuming the composite action of both elements under
service load. Basing on own designing experience gained in the former R&D projects devoted
to FRP bridges [20], [21], the overall cross sections as well as the initial structure of each
particular laminate in both U-shaped girder body and the sandwich panel were assumed. The
values of properties of materials were determined by the laboratory tests and elaborated from
a statistical point of view to obtain characteristic values. During verifying the resistance, the
deformability as well as the stability, the characteristic values corresponding to the 5% fractile
of the statistical distribution were used according to guidelines [22]. The characteristic
mechanical properties of all GFRP laminas and two structural foams are presented in Table 3.

The detailed FEM model of the FRP superstructure was prepared in order to use it in design
process and to analyse bridge structural behaviour in different stages of construction and
during service. This model was also used to check code requirements. The four-node shell
finite elements were used for superstructure discretization (Fig.4). Material characteristics
obtained from testing were used in FEM modelling (Table 2). Using the classical laminate
theory and starting from the experimental characterizations of laminas and the selected
laminate stacking sequences, the material properties of the laminates were determined and
taken into account in FEM model of the FRP superstructure.

In the first stage of FEM analysis the stiffness of the entire superstructure was checked
assuming standard design loads according to Eurocode 1 [23] with relevant adjustment
coefficients. The main design criterion used for selection of laminas and optimization of
laminates was the allowable deflection in mid-span, which was applied according to the
British code as span/300 [24]. In such way, the final laminate structure for all girders and

4
deck elements was established taking into account the stacking sequence of constituent
laminas. This FEM model of the superstructure was also used for prediction of theoretical
values of its self-frequencies, damping parameters and modes of vibrations.
In the second stage of the FEM analysis, much more sophisticated model of the superstructure
was used with refined mesh. Thanks to software code possibilities, the exact layered structure
of the laminates could be discretized. However, due to FEM code limitations up to ten various
layers (laminas) were modelled in finite shell elements using substitute material properties
established with the help of the classical laminate theory (Fig.5). Such kind of substitution
enabled the exact material properties to be assumed in the FEM model and the verification of
laminate resistance at ply level.

The design value, Xd, of the generic property of resistance or deformation of a material was
expressed, in a general form, as the following equation [22]:

where:
ηc is a conversion factor which takes into account, in a multiplicative manner, the peculiarity
of the actual case,
Xk is the characteristic value of the property,
γm is a partial factor covering uncertainty in the resistance model and geometric deviations if
these are not modelled explicitly.

The mean strength of each particular lamina Xk was taken into account in calculations. The
conversion factor ηc is obtained by multiplying the specific conversion factors relevant for all
the environmental actions and long term effects affecting the behaviour of the material (creep,
temperature, humidity, fatigue). The values attributed to these factors are indicated in [22].
The values of conversion factors and material partial factors applied in design are presented in
Table 4.

The verification of laminate resistance at ply level (first ply failure method) was carried out
by using the well-known criteria: maximum stress, Azzi – Tsai - Hill and Tsai – Wu [25]. For
each laminate the failure index If was calculated (Table 5). The theory postulates that ply
failure initiates when If > 1.0. Since no such case took place, the strength and stability of each
girder and deck section were numerically revealed and thus confirmed the compliance of the
bridge superstructure with the service and ultimate limit states as defined in relevant code. No
failure criterion was omitted thus proving the proper design and optimization of laminates,
which form the all-composite superstructure.

The bonded joints taken into account in design are made from FRP elements (adherents)
mostly subjected to axial force. The mechanical behaviour of more complex joints (panel –
panel) were reduced to single-lap or double-lap joints. The normal interfacial stress (σ -
orthogonal to plane of the joint) and the shear stress (τ - parallel to the plane of the joint,
along the direction of the axes of the joint) were defined by simplified constitutive interface
laws [26]. The verification of the ULS of a bonded joint required that the following condition
was satisfied in the adhesive:

where:

5
NSd is related to the design normal strain which the joint should transfer,
NRd is related to the design normal strain resistance.

NRd includes the effects of ageing and environmental effects, and is eventually penalized in
order to take into account the presence of shear and flexure stresses. The degradation effects
(ageing, temperature, moisture, chemicals) for the adhesive were taken into account in design
by using the values of the conversion factor ηc given in [22]. Moreover, the principal stresses
within the adherent associated to the stresses σ and τ transferred to the interface were checked
by comparing to the design strength to tension and compression of the FRP matrix.

The whole designing process was strongly supported by the laboratory testing. The
comprehensive material testing was carried out by the UCB-MF of Warsaw University of
Technology, while the structural tests comprising deck panels, all adhesively bonded joints
and the full-scale FRP girder, were carried out at the Roads & Bridges Department of the
Rzeszow University of Technology. In particular the strength of a bonded joint was verified
through appropriate tests (design by testing). According to [22] these could represent a valid
tool in the case of joints with particularly complex geometries (panel – panel joints). The
design resistance was determined in accordance to the procedure indicated in EN 1990 [27].
At least 3 samples of each type of joint were tested. Moreover, to characterize the
homogeneity of the quality of the bonding, with the aim of highlighting any defects including
delamination, debonding as well as the presence of empty voids, non-destructive acoustic
emission tests were used after assembling the superstructure. The structural laboratory tests
and their detailed results are to be published elsewhere.

Static test

The proof test of the bridge comprising static and dynamic loading was carried out by the
Rzeszow University of Technology. Two four-axis trucks with the nominal weight of 300 kN
each and the total weight of 642.2 kN were applied for static tests. Two loading schemes were
realized: the asymmetric scheme with only one truck (SS-1) and the symmetric scheme with
two trucks standing on the bridge (SS-2) (Fig.6). Girder and deck vertical displacements as
well as FRP composite strains were recorded in order to compare them with subsequent
theoretical values, determined in FEM analysis and with allowable values according to codes
used in design. The vertical displacements were measured by means of the LVDT sensors in
12 discreet points of each girder’s bottom flange, in three span cross-sections, i.e. in ¼, ½ and
¾ L. The foil strain gauges were also located in these same discreet points of girders bottom
flange and additionally in the upper flanges as well as in deck slab between girders, in all
three span sections respectively. The FRP longitudinal strains were measured in girder flanges
while transverse and longitudinal strains were monitored in the deck bottom surface. The total
number of 56 strain gauges were used in the static tests.

As far as vertical displacements are concerned, the maximum measured elastic displacement
in the mid-span of internal girder was 13.04 mm, what is only 53% of theoretical value and is
also less than the allowable value equalling 33.4 mm (L/300). The transverse distribution of
girder mid-span displacements under the half of and full static load is shown in Figure 7,
while the displacement versus time at mid-span in Figure 8. About 50% difference between
measured and theoretical values under full loading is due to the influence of the bridge deck
equipment not considered in the design model (expected to be a dead load). Two lightweight
concrete sidewalk slabs, stone curbs and polymer cornices considerably increased the stiffness
of the lightweight all-composite span. Particularly, concrete sidewalks with the total surface

6
of 1/3 of deck surface contribute to span stiffness due to good adhesion induced between fresh
concrete and FRP surface during concreting. The stiffening effect of the concrete sidewalk is
clearly shown for the external not-loaded girder during asymmetric load when the difference
between theoretical and measured values is about 150% (Fig.7a).

In Figure 8, the experimentally and numerically determined load transverse distribution

factors were determined. The load distribution factors of the bridge span were calculated in
accordance with well-known formulas on the basis of the girders deflection under relevant
load position (see f.e. [28]). The maximum values of the transverse distribution factors were
as follows: 0.203 and 0.212 for the external girders as well as 0.280 and 0.305 for the internal
girders. The quite good compliance was obtained between the influence lines designated by
the FEM model and the measured behaviour of the span under the load. For both external
girders the transverse distribution factors, determined using the FEM, were slightly lower than
those obtained experimentally. On the other hand, the FEM values for both internal girders
were greater than experimental ones. However, theses discrepancies are small, between 7 to
15 %, and reflect the abovementioned influence of the bridge deck equipment on the all-
composite span behaviour.

The comparison between measured and theoretical FRP strains in girder bottom flanges at
mid-span under the half and full static load is shown in Figure 10, while the measured strains
versus time at mid-span in Figure 11. The maximum elastic strain measured in bottom flange
composite was 0.760 ‰ and constitutes 62% of theoretical value. It proved the influence of
concrete sidewalks on structural behaviour of the girders. It can be seen more reasonably in
Figure 12, where the vertical strain distribution in most loaded girder is presented. The
contribution of concrete sidewalk plate as a part of girder’s upper flange shifted the neutral
axis up to the bond line level and thus mitigated girder’s bottom flange. However, this
maximum real strain is far less than design limit value for glass fibre laminas used for
manufacturing of girder’s bottom flange (ɛtg = 10.54 ‰).

The considerable discrepancies were shown between measured and theoretical FRP strains in
deck bottom surface at their mid-span (Fig.13). The measured longitudinal strains were much
less due to shifting the neutral axis towards the bond line and thus decreasing bottom surface
strains induced by the girder global bending. The differences in transverse strains induced by
the local deck slab bending, showed the uncertainties of the FEM model and the necessity to
validate it against the experimental results. However, it should be also reported that maximum
FRP elastic strain measured in deck panel (in transverse direction) was 0.106 ‰ and
constitutes about 1% of design limit value for glass fibre lamina.

Dynamic test

The dynamic characteristics of a bridge can by evaluated by means of direct results of

acceleration, strain or displacement measurements of bridge elements under proof or service
load (mostly trucks) passing a bridge with various velocities. Most of published dynamic
characteristics of FRP road bridges were set in this way in frequency domain [29], [30], [31].
Using this method and basing on the measurement of girder and deck displacements such
dynamic parameters as: dynamic coefficients, natural frequencies, logarithmic decrement and
damping ratios were determined for the all-composite bridge. The changes of vertical
displacements in few points of the superstructure were recorded in order to evaluate the
dynamic behaviour of the bridge. The dynamic tests were carried out with one truck passing
the bridge with three velocities: 10 km/h, 20 km/h and 25 km/h (SD-1 to SD-3 loading

7
schemes). The local road constrains precluded from applying higher velocities in the dynamic
test. To enhance dynamic effect, the 5 cm deep transverse threshold was used and also sudden
braking on pavement was applied in two additional schemes (SD-4 and SD-5) with the truck’s
velocities of 10 km/h and 15 km/h, respectively.

In each measurement point and for every load scheme the "time-displacement" plots were
obtained as shown in example in Figure 14. The dynamic coefficients for each scheme were
calculated as the ratio of maximum dynamic displacement (read on the plot) and maximum
static displacement, counted as the average of minimum and maximum displacements on the
plot. Basing on measured displacement values, the dynamic coefficients for all velocities (10 /
20 / 25 km/h) were assessed as 1.079 / 1.100 / 1.102, respectively. These values are much
lower than the dynamic coefficient (1.25) used in bridge design.

The first natural frequency of the all-composite bridge span was determined by means of the
"time-displacement" plot for scheme SD-4 with the transverse threshold (Fig.15) as the ratio
of number of displacement peaks in time ts and this time ts (in seconds). The estimated first
natural frequency of the all-composite span equalled 10.1 Hz and was much higher (means
better) than recommended value according to the Polish bridge requirements (3 Hz).
However, once again the influence of the heavy concrete sidewalks could be seen, because the
first natural frequency determined in FEM analysis was much lower and equalled only 5,36
Hz.

Finally, the experimentally determined logarithmic decrement equalling 0.283 was

determined by means of the "time-displacement" plot for scheme SD-5 (Fig.16) and applied
to estimate the all-composite span’s damping ratio. The value of the relevant damping ratio
amounted 4,5 % and is similar to damping ratios of several US all-composite bridges, where
the minimum damping ratio of 5% was observed (see f.e. [30]). However, this value is much
lower than corresponding damping characteristics for typical concrete bridges with similar
beam-slab superstructure and span length.

Bridge monitoring

Considering the worldwide recognized advantages of fibre optic sensors as measuring devices
in the SHM of the FRP bridges and the unique ability to measure the long range distributed
strain and temperature along the entire bridge superstructure, the distributed fibre optic sensor
(DFOS) technology was chosen as the basic SHM system of the first Polish all-composite
bridge. For distributed strain and temperature measurements, the optical reflectometer based
on linear Rayleigh scattering called Luna OBR 4600 was used. The single-mode telecom-
grade optical fibres were applied and 10 mm virtual measurement sections spaced every 10
mm along optical fibres were set to measure strains and temperature [32].

The fibre optic sensors installation took place in the workshop just after the FRP girders and
deck panels were manufactured. The sensors were placed inside two bridge girders and
bonded to their composite body as showed in Figure 17. The total ten sensors with the length
of about 9.60 ± 0.10 m were installed enabling the strain measurement in about 10 000 virtual
discrete points of the girder. The section of the sensor no. 10 was placed inside a
polypropylene tube in order to measure the ambient temperature to compensate its influence
on strain readings. Only one middle deck panel with dimensions of 5.35 x 1.92 m was
equipped with the fibre optic sensors. Two sensors were installed on the top and bottom

8
surfaces of the panel, i.e. sensors no. 01/02 and 03/04, respectively. Their alignment enabled
to measure strains in both transverse and longitudinal direction of the panel (Fig.18).

The monitoring of the bridge state-of-repair has been planned as a sequence of several proof
load tests under controlled loading to carry out twice a year for first five years of the bridge
service. The first proof test was carried out just after bridge completion and before it was
opened to traffic. Therefore except DFOS, the conventional discrete foil strain gauges were
also used in most of critical locations of the FRP superstructure. Thus two different strain
measurement techniques, applied for testing, enabled the validation of the DFOS
measurement by comparing the strains resulting from both techniques. The second proof load
test was carried out under the same loading conditions after 8 months of service and the traffic
was stopped for a few hours to facilitate testing.

In both proof load tests two four-axle heavy vehicles were used as a bridge loading with the
total mass of 64220 kg. The following strain readings were carried out in the subsequent
loading stages: before trucks entering the bridge (zero strains), under the full loading of the
bridge (maximum strains) and few minutes after unloading (residual strains). The temperature
induced strains were compensated due to temperature strain measurement with the fibre optic
sensor no. 10. The ambient temperature differences in various stages of both proof tests
amounted about 1-3 °C, which meant the strains in the range of 30-100 με (microstrains)
ought to be deducted in final strain evaluation.

In the following plots, the strain measurements provided by the selected fibre optic sensors
are presented. The positive (+) values mean the tensile strains, while the negative ones (–) the
compression strains detected in the relevant discrete locations along the fibres. The plots
present strains measured in both proof tests for better comparison. This comparison allows to
evaluate the FRP superstructure performance during first 8 months of service to check both
traffic and environment influence on its structural behaviour. The comparison qualitatively
and quantitatively reveals the strain differences induced during service and, on their basis, the
current state- of-repair of the bridge can be monitored.

In Figure 18, the strains detected in the bottom flange of the external girder (sensor no. 09) are
presented. In turn, in Figure 19, the strain plot in the upper flange of the same girder measured
by the sensor no. 08 are shown. In Figure 20 the plot of the strains measured on the bottom
surface of the deck panel is presented (sensor no. 03/04) with the reference to Figure 17
showing sensor location. Both tension/compression and transverse/longitudinal strains
induced in the deck panel under loading are presented.

Comparing the strain plots for FRP girder and deck panel obtained in both proof tests, it can
be concluded that their character and strain values are very similar and close, which reveals
no negative influence of traffic load and environment on the all-composite superstructure
behaviour after 8 months of bridge service. It should be clarified that the strain value of 100 x
10-6 (denoted 100 με) corresponds to the stress range of 1,21 - 4,21 MPa, depending on the
lamina type in particular girder or deck laminate. The only considerable difference was
detected by the sensor no. 08 mounted on the upper flange of the girder. The characteristic
sharp changes of the strain plots indicate the girder sections where the inner diaphragms are
located. The diaphragms decrease strains in adjacent girder laminates.

The maximum strain in bottom flange of the external girder was about 715 με in both tests
(sensor 09). After the bridge unloading, the remaining strains amounted only 30 and 50 με in

9
proof test 1 and 2, respectively. The mean differences between strains in tests 2 and 1 ranged
only 12 με, which showed the lack of any negative effect during 8 months of service. The
strains in the upper flange of the external girder (sensor no. 08) oscillated around 0 and were
both positive and negative in the proof test 1. The maxima on both sides amounted about 120
με and -45 με respectively. It confirms the fact that the location of girder’s neutral axis seems
to be in the bond line between girder’s U-body and the deck panel. In the proof test 2, the
strains measured in sensor 08 increased by 50-60 με in average and were only positive
(tension). The maximum strain amounted about 170 με. It was due to additional strains
induced by the temperature. During the proof test 2, the ambient temperature was about 25 °C
and as a result, the temperature of the FRP panel was much higher due to black thin pavement
layer. The mean reading of thermal sensor no. 10 was about 60 με during the entire test. This
strain increase, induced by temperature, was visible only in the higher part of the girder (deck,
upper flanges) and disappeared towards bottom flange, where no thermal effect was detected.

The behaviour of the deck panel was identified on the basis of the sensor no. 03 strain
measurement. It was located on the bottom surface of the panel and enabled the strain
detection in both panel directions. The strains along the A section, parallel to the panel
longitudinal axis, indicate that the location of the girder’s neutral axis is a little below the
bond line (full compression of the deck in longitudinal direction). The maximum compression
strain was about -100 με and increased to 0 after bridge unloading. The mean difference
between maximum strains in both proof tests amounting only 5 με is negligible, which shows
no difference in panel behaviour during service. The higher strain difference was detected
after unloading (about 30 με) because the higher remaining strains were left in the panel after
the proof test 2. The strains along the C and G sections, also parallel to the panel longitudinal
axis, were smaller than corresponding A section strains. In this part of the panel, the sensor
no. 03 was located inside the U-girder and the maximum strains of the panel bottom surface
ranged -30 to -45 με. The mean difference between maximum strains in both proof tests was
only 3-13 με and about 30 με after unloading.

The strains induced by panel transverse bending were measured by the sensor no. 03 in the
sections B, D, F and H. These strains were twice as high as in case of the longitudinal sections
and had positive values, which indicated tension of the panel’s bottom surface. The maximum
strains were measured along the D and F sections (close to the middle of the panel), and their
maximum values ranged 180-200 με and 340-360 με in the proof test 1 and 2 respectively.
The mean strain increase in these sections amounted about 60-80 με under full loading and
50-60 με after unloading. It confirms the existence of additional thermal strains in the deck
panel. A little lower strain increase was detected in the sections B and H, where the mean
strain increase between two proof tests ranged 45-65 με, both under and after loading.
However, the maximum strains in the proof test 2 were twice smaller than in the test 1 and
equalled 140-180 με.

To summarize, the DFOS strain measurement of the FRP bridge span during two proof load
tests, carried out 8 months apart, reveals the strain differences having no practical meaning
referring to time dependent structural behaviour of the bridge span. It confirmed the very safe
stress level in the FRP elements. The maximum strain detected during both proof tests was
about 845 με, which corresponds to the stresses in particular laminas ranging 10.2 – 35.6 MPa
and which, in turn, is about 4.2 – 8.0 % of their tensile strength. After 8 months of service,
excluding momentary thermal strains, no serious changes in the FRP composite superstructure
behaviour were observed under full loading and after unloading. The maximum difference

10
between strains, measured during proof test 1 and 2, was about 80 με and is negligible
referring to safety and serviceability of the all-composite bridge.

Conclusions

The work has shown that a short-span all-composite bridge superstructure can meet the
strength and deflection design criteria according to European bridge regulations [22], [23],
[24], [26] and [27] as well as domestic requirements. The following conclusions can be made
based on analysis of the all-composite bridge structural behaviour and monitoring data:

 the overall stiffness of the all-composite span was about 50% greater than estimated in
design and the maximum deck slab and girder deflections equalled only 1/3 of the
allowable value;
 very good transverse load distribution between all girders, revealed in testing, confirmed
the appropriate structural behaviour of the all-composite span;
 measured maximum FRP strains were about 60% of theoretical values and less than 10%
of material limit values, so the further material optimization could be considered in all-
composite bridge design;
 bridge equipment (concrete sidewalk slabs, stone curbs and polymer cornices)
considerably contributes to structural behaviour of the lightweight all-composite span,
what on the one hand deviates measured and theoretical performance indicators, but on
the other hand provides the additional safety margin;
 the comparison of the proof testing and numerical simulation results shows a considerable
discrepancy due to influence of non-structural bridge equipment on FRP bridge stiffness;
a finite element model developed to simulate the bridge structural behavior should be
improved and validated by field testing;
 despite the lightweight superstructure, dynamic characteristics of the all-composite span
are very good, fulfilling the relevant requirements for road bridges and are similar to those
observed in another FRP road bridges;
 strain monitoring with DFOSs revealed no serious changes in the all-composite
superstructure behaviour after 8 months of bridge service, excluding momentary thermal
strain increases.

All testing results of the all-composite road bridge confirm the necessity of the structural
health monitoring of such kind of prototypes, where not fully proven structural and material
solutions were applied. The structural monitoring system installed on the bridge has provided
valuable field data on the performance of FRP bridge. It is expected that such system will not
only give engineers and road agencies a valuable tool in monitoring the structural
performance of a novel FRP bridge systems, but will also provide important information
related to durability, design criteria, and long-term response of FRP all-composite structures.
Furthermore, as design guidance improves and becomes more widespread, and the civil
engineering profession gains experience and confidence with FRP composites, it is likely that
the use of FRP composite elements will increase in bridge engineering.

Acknowledgment

This work was created within the framework of the project: “COMBRIDGE – An innovative
FRP composite road bridge”. The project was implemented as part of a pilot programme

11
entitled: Support for Research and Development Works in the Demonstrative Scale
DEMONSTRATOR+ (Contract No. UOD-DEM-1-041/001) and was co-funded by the
National Centre for Research and Development (NCBiR).

References

[1] Holloway LC, Head PR. Advanced polymer composites and polymers in the civil
infrastructure. London: Elsevier Science Ltd, 2001.
[2] Ye LP, Feng P, Lu XZ, Qian P, Lin L, Huang YL, Hu WH, Yue QR, Yang YX, Tan Z,
Yang T, Zhang N, Li R. FRP in China: the state of FRP research , design guidelines, and
application in construction. In: Proceedings of the 2nd International Conference on FRP
Composites in Civil Engineering (CICE 2004), Adelaide, 8-10 December 2004. p. 109 –
119.
[3] Bank LC. Application of FRP composites to bridges in the USA. In: Proceedings of the
International Colloquium on Application of FRP to Bridges, Tokyo, 20 January 2006. p.
9-16.
[4] Rodriguez JG, Carlson NM, Torres LL, Blackwood LG, Yoder TS. INEEL test on
composite bridge. Idaho Falls, ID: Idaho National Engineering and Environmental
Laboratory, 1999.
[5] Davalos JF, Qiao PZ, Xu XF, Robinson J, Barth KE. Modelling and characterization of
fiber-reinforced plastic honeycomb sandwich panels for highway bridge applications.
Composite Structures 2001;52(3-4):441-452.
[6] Foster DC, Richards D, Bogner BR. Design and installation of fiber reinforced polymer
composite bridge. Journal of Composites for Construction 2000;4(1):33-37.
[7] Alampalli S, O’Connor J, Yannotti AP. Fiber reinforced polymer composites for the
superstructure of a short-span rural bridge. Composite Structures 2002;58(1):21-27.
[8] Gillespie JW, Eckel DA, Edberg WM, Sabol SA, Mertz DR, Chajes MJ, Shenton HW, Hu
C, Chaudhri M, Faqiri A, Soneji J. Bridge I-351 over Muddy Run: design, testing and
erection of an all-composite bridge. Transportation Research Record: Journal of the
Transportation Research Board 2000;1696(2):118-123.
[9] Nystrom H, Watkins S, Stone D, Nanni A. Laboratory and field testing of FRP composite
bridge decks and FRP reinforced concrete bridge for city of St. James, Phelps County,
MO. Rolla, MO: Center for Infrastructure Engineering Studies, University of Missouri -
Rolla, 2002.
[10] Behrends MA. Testing and health monitoring of an integrally molded fiber reinforced
polymer bridge. Cincinnati, OH: University of Cincinnati, 2012.
[11] Veltkamp M, Haffmans A. New concepts in movable lightweight bridges. In:
Proceedings of 19th IABSE Congress 2016: Challenges in design and construction of an
innovative and sustainable built environment, Stockholm, 21-23 September 2016. p.1288-
1295.
[12] Burgoyne C. Advanced composites in civil engineering in Europe. Structural
Engineering International 1999;9(4):267-273.
[13] Hayes MD, Lesko JJ, Weyers RE, Cousins TE, Haramis J, Gomez J, Masarelli P.
Laboratory and field characterization of the Tom’s Creek Bridge superstructure. Journal
for Composites in Construction 2000;4(3):120-128.
[14] Howdyshell PA, Trovillion JC, GangaRao HVS, Lopez-Anido R. Development and
demonstration of advanced design composite structural components. Washington, D.C.:
Construction Engineering Research Laboratories, US Army Corps of Engineers, 1998.

12
[15] Cousins TE, Lesko JJ. Construction of a Virginia short-span bridge with the
Strongwell 36-inch double-web I-beam. Blacksburg, VA: Department of Civil and
Environmental Engineering, VirginiaTech, 2005.
[16] Luke S, Canning L, Collins S, Brown P, Knudsen E, Olofsson I. The development of
an advanced composite bridge decking system – project ASSET. Structural Engineering
International 2002;12(2):76-79.
[17] Canning L. Examples and case studies. In: Zoghi M, editor. The international
handbook of FRP composites in civil engineering. Abington: CRC Press, Taylor &
Francis Group, 2013. p.191-217.
[18] Hollaway LC. A review of the present and future utilisation of FRP composites in the
civil infrastructure with reference to their important in-service properties. Construction
and Building Materials 2010;24(24):2419–2445.
[19] Kendall D. Developments in FRP bridge design. Reinforced Plastics
2010;May/June:38-42.
[20] Siwowski T, Kaleta D, Rajchel M, Wlasak L. The first polish road bridge made of
FRP composites. Structural Engineering International 2017;27(2):308-314.
[21] Siwowski T, Poneta P, Kulpa M. Development and research on the innovative FRP
girder for bridge application. In: Proceedings of the 37th IABSE Symposium: Engineering
for Progress, Nature and People, Madrid, 3-5 September 2014. p.2239-2246.
[22] Ascione L, Gutierez E, Dimova S, Pinto A, Denton S. Prospect for new guidance in
the design of FRP. Ispra: EC Joint Research Centre, 2016.
[23] EN 1991-2 Eurocode 1: Actions on structures - Part 2: Traffic loads on bridges.
Brussels: European Committee for Standardisation, 2003.
[24] BD 90/05: Design manual for roads and bridges, vol. 1, section 3, part 17: design of
FRP bridges and highway structures. The Highways Agency, Scottish Executive, Welsh
Assembly Government, The Department for Regional Development Northern Ireland,
2005.
[25] Kaw AK. Mechanics of composite materials. Abington: CRC Press, Taylor & Francis
Group, 2005.
[26] CNR-DT 205/2007. Guide for the Design and Construction of Structures Made of
Pultruded FRP Elements. Italian National Research Council, Rome, 2008.
[27] EN 1990 Eurocode 0: Basis of structural design. Brussels: European Committee for
Standardisation, 2002.
[28] Rajchel M, Siwowski T. Live load transverse distribution in a road slab-girder bridge
made of FRP composite girders. Roads and Bridges 2017;16(2):131 – 145.
[29] Alampalli S. Field performance of an FRP slab bridge. Composite Structures
2006;72(4):494-502.
[30] Aluri S, Jinka C, GangaRao HVS. Dynamic response of three fibre reinforced polymer
composite bridges. Journal of Bridge Engineering 2005;10(6):722–730.
[31] Aref AJ, Alampalli S. Vibration characteristics of a fibre-reinforced polymer bridge
superstructure. Composite Structures 2001;52(3-4):467-474.
[32] Siwowski T, Rajchel M, Sienko R, Bednarski L. Smart monitoring of the FRP
composite bridge with distributed fibre optic sensors. In: Proceedings of the 9th
International Conference on Fibre-Reinforced Polymer Composites in Civil Engineering
(CICE 2018), Paris, 17-19 July 2018 (in press).

List of figure captions

Figure 1: Cross-section of the bridge superstructure (a) and side view / longitudinal section of
the bridge (b)

13
Figure 2: U-shaped composite girder structure (a) and sandwich deck slab structure (b)
Figure 3: FRP superstructure assembling division and location of bonded joints
Figure 4: FEM model of the bridge superstructure
Figure 5: Actual stacking sequence and its FEM substitution of upper and bottom flange of
FRP girder
Figure 6: Two bridge loading schemes in static test
Figure 7. Transverse distribution of girder mid-span displacements under the half (a) and full
static load (b)
Figure 8: Displacement versus time at mid-span of all girders (colour should be used)
Figure 9: Experimentally and numerically determined load transverse distribution factors
Figure 10: Measured and theoretical FRP strains in girder bottom flanges at mid-span under
the half (a) and full static load (b)
Figure 11: Measured strains versus time at mid-span of all girders (colour should be used)
Figure 12: Vertical strain distribution in most loaded girder
Figure 13: Measured and theoretical FRP strains in deck bottom surface at panel’s mid-span:
(a) longitudinal strains; (b) transverse strains;
Figure 14. Maximum deflection in mid-span of the inner girder in SD-1 scheme
Figure 15: Maximum deflection in mid-span of the inner girder in SD-4 scheme (transverse
threshold)
Figure 16: Maximum deflection in mid-span of the inner girder in SD-5 scheme (braking)
Figure 17: DFOSs layout for the girder and deck panel strains measurement (sensor no. 10 for
thermal strain only)
Figure 18: Plot of strains measured with the DFOS no. 09 - the bottom flange of external
girder (colour should be used)
Figure 19: Plot of strains measured with the DFOS no. 08 – the upper flange of external girder
(colour should be used)
Figure 20: Plot of strains measured with the DFOS no. 03 - the bottom surface of the deck
panel (colour should be used)

Tables

14
Table 1. All FRP composite road bridges made of solid laminates and sandwich plates

Length &
Bridge name, Bridge Fabrication
No. Year width FRP structure FRP constituents Remarks Ref.
location type method
(meter)
Ginzi, L= 10.0 glass fibres, world’s first FRP
1 1981 slab no data hand lay-up [1]
Bułgaria B=? polyester resin bridge
Miyun, glass fibres,
L= 20.7 bonded
2 Beijing, 1982 slab polyester resin, hand lay-up [2]
B = 9.9 rectangular beams
China honeycomb core
hand lay-up
U-shaped beams, glass fibres,
INEEL, Idaho beam & L = 9.1 (beams), demonstrative
3 1995 deck from polyester resin, [4]
Falls, USA deck slab B = 5.5 pultrusion bridge
pultruded profiles vinylester resin
(deck)
No-Name glass fibres,
L = 7.1 sandwich slab first US public
4 Creek, Russell 1996 slab polyester resin, hand lay-up [5]
B = 8.5 (KSCI system) FRP bridge
County, USA honeycomb core
U-shaped beams, hand lay-up
Smith Road glass fibres,
beam & L = 10.1 deck from (beams),
5 Bridge, Butler 1997 polyester resin, Tech-21 bridge [6]
deck slab B = 7.3 pultruded profiles pultrusion
County, USA
(DuraSpan) (deck)
Bennett’s sandwich slab
glass fibres,
Creek, L = 7.8 (Hardcore
6 1998 slab polyester resin, VARTM [7]
Steuben B = 10.1 Composites
PET foam core
County, USA system)
Muddy Run sandwich slab
glass fibres,
(no. I-351) L = 9.8 (Hardcore
7 1998 slab polyester resin, SCRIMP [8]
Glasgow, B = 8.5 Composites
PET foam core
USA system)
St. Francis
glass fibres,
Street, L = 8.0 sandwich slab
8 2001 slab polyester resin, hand lay-up [9]
St. James, B = 8.3 (KSCI system)
honeycomb core
USA
Eight Mile integrated beams
glass fibres,
Road, beam & L = 6.7 and slab
9 2007 vinylester resin, VARTM [10]
Hamilton deck slab B = 18.7 (SuperFiberSPAN
PUR foam core
County, USA system)
10 Hoofdbrug, 2010 slab L = 12.0 sandwich slab glass fibres, VARTM lift bridge [11]

15
 Oosterwolde, B = 11.2 (InfraCore polyester resin,
The system) PUR foam core
Netherlands
Nelson
Mandela, sandwich slab glass fibres,
L = 22.5
11 Alkmaar, 2016 slab (InfraCore polyester resin, VARTM lift bridge [11]
B = 14.5
The system) PUR foam core
Netherlands

Table 2. All FRP composite road bridges made of pultruded shapes

Bridge name, Bridge Length & width Pultrusion profiles

No. Year FRP constituents Connections Remarks Ref.
location type (meter) system (company)
tongue &
Bonds Mill, L = 8.2 glass fibres,
1 1994 beam (box) ACCS (Mouchel) groove, bascule bridge [12]
Gloucester, UK B = 4.6 polyester resin,
adhesive
Tom’s Creek,
L = 5.3 EXTREN glass fibres,
2 Blacksburg, 1997 beam bolts timber deck [13]
B = 7.3 (Strongwell) vinylester resin,
USA
Laurel Lick, Superdeck
beam & L = 6.1 glass fibres, hook bolts,
3 Lewis County, 1997 (Creative timber deck [14]
deck slab B = 4.9 vinylester resin, adhesive
USA Pultrusions)
Route 601,
L = 5.3 EXTREN glass fibres,
4 Sugar Grove, 2001 beam bolts timber deck [15]
B = 7.3 (Strongwell) vinylester resin,
USA
West Mill, glass fibres,
beam & L = 10.0 ASSET
5 Oxford County, 2002 carbon fibres, adhesive [16]
deck slab B = 6.8 (Fiberline)
UK polyester resin,
Klipphausen,
L = 6.3 ASSET glass fibres, first German FRP
6 Dresden, 2005 slab adhesive [17]
B = 6.0 (Fiberline) polyester resin, bridge
Germany
Chamberlain Composolite, tongue &
L = 11.7 glass fibres,
7 Bridgetown, 2006 beam (box) former ACCS groove, bascule bridge [18]
B = 8.9 polyester resin,
Barbados (Strongwell) adhesive
glass fibres,
Standen Hey, L = 10.0 ASSET
8 2007 slab polyester resin, adhesive [19]
Lancashire, UK B = 4.9 (Fiberline)
structural foam

16
Table 3. The material parameters of FRP components

Material Notation Unit Type of material

parameter Glass fabrics PVC foam PUR foam
Fibre θ [deg] 0/90 0 0/90 ---- ----
orientation
Density ω [g/m2] 1200 1200 800 1200 1050
Nominal t [mm] 0.80 0.80 0.80 15.0 8.0
thickness
Modulus of Ex [GPa] 20.50 42.13 20.00 0.09 0.029
elasticity Ey [GPa] 20.50 10.87 20.00 0.09 0.029
Poisson’s νxy ---- 0.019 0.29 0.029 0.4 0.41
ratio νyx ---- 0.019 0.075 0.029 0.4 0.41
Shear Gxy [GPa] 3.90 4.40 3.90 0.027 0.0015
modulus Gyz = Gxz [GPa] 3.04 2.71 2.83 0.027 0.0015
Normal Xt [MPa] 520.0 855.0 522.0 ---- ----
strength Xc [MPa] 320.0 537.0 321.0 1.40 1.03
Yt [MPa] 520.0 44.0 522.0 ---- ----
Yc [MPa] 320.0 84.0 321.0 1.40 1.03
Shear strength Sxy [MPa] 60.0 51.0 60.0 1.15 0.28
Syz = Sxz [MPa] 30.0 25.0 30.0 1.15 0.28

Table 4. Values of conversion factors and material partial factors

Conversion and partial factors

Limit states in code checking
ηc γm γm / ηc
Ultimate limit states (ULS) 0.55 1.55 2.82
Serviceability limit states (SLS) 0.73 1.00 1.37

Table 5. Maximum failure index If for all superstructure laminates

Girder Deck slab

Criterion
Bottom flange Upper flange Web Bottom face Top face
Maximum stress 0.704 0.481 0.312 0.874 0.408
Azzi-Tsai-Hill 0.714 0.541 0.391 0.886 0.410
Tsai-Wu 0.649 0.565 0.323 0.935 0.441

17
18